Introduction

Before launching into this very lengthy case study, I feel I should say some­thing about why I think a study this long and detailed is necessary. The short answer is this: I am proposing an alternative to standard theories of grammar. That is not something to be taken lightly. One of the principal merits claimed for the standard approaches is that they are rigorous and precise. For an alternative to be taken seriously, it must show that it is every bit as rigorous and precise. In addition, a new approach must show that it can succeed where previous approaches fail. That is why I have chosen there-constructions. They are classical cases that every theory of grammar must deal with, and where all previous approaches have been found wanting. Previous approaches have failed in two respects: First, they have not been able to account for the incredibly complex detail in these constructions. Second, they have not been able to account adequately for the relationship between the there of There's Harry on the porch and the there of There's a man on the porch. Standard theories simply throw up their hands and claim that there is no relationship. The principal purpose of this case study is to demonstrate how cognitively based grammar can be done with as much rigor as generative grammar, and why it succeeds in a complicated case where generative grammar fails.

There is, of course, not just one theory of generative grammar; there are dozens, if not hundreds, of actual theories, and the variations on them go into the millions, as McCawley's title Thirty Million Theories of Grammar suggests. We will be arguing that all generative approaches to grammar are inadequate. The argument will be an empirical one—based on a very large body of data. But before we get to the details, it should be pointed out that generative approaches to the study of language are inadequate not just for empirical reasons, but for theoretical reasons as well.

A generative grammar is technically a kind of formal syntax—a collection

462

Cognitive Grammar 463

of principles for manipulating symbols without regard to their meaning. As such, generative grammars typically presuppose an objectivist approach to cognition. A semantics for a generative grammar is either model-theoretical or it consists of a translation into another system of symbols—a "mentalese" usually called a logical form or semantic representation—where the symbols are taken as being internal representations of external reality. In either case, generative grammar is subject to Putnam's critique (see chap. 15), which means that it cannot be provided with a consistent theory of meaning. We must simply look elsewhere for a theory of grammar. In particular, we must find a theory of grammar in which the syntax is not independent of the semantics.

We will be proposing a theory of cognitively based grammar that makes use of cognitive model theory as described in chapter 17. A grammar from this point of view will be a radial category of grammatical constructions, where each construction pairs a cognitive model (which characterizes mean­ing) with corresponding aspects of linguistic form. We will refer to this as a cognitive grammar.

Cognitive grammars avoid Putnam's critique for the following reason:

Cognitive models are internal in Putnam's sense and are given meaning via their connection with experience, especially bodily experience (see chap. 17). The parameters of linguistic form in grammatical constructions are not independent of meaning; rather they are motivated, and in many cases even predicted, on the basis of meaning. On this view, many aspects of syntactic structure are motivated by, or are consequences of, the struc­ture of cognitive models.

Cognitive Grammar

The present case study is intended to show that radial categories occur in grammar and that they have the same function as radial categories in the lexicon (see over), namely, motivating correspondences between form and meaning. We will be claiming that the category of clause structures in a language is radially structured, with a central subcategory and many noncentral subcategories. The central clause structures exhibit a direct and regular relationship between form and meaning, specified by general principles which we will refer to as central principles, Noncentral clause structures are systematically related to central clause structures, and their form-meaning correspondences derive in large part from those which are more central. Thus, the form-meaning correspondences of noncentral clause structures are highly motivated—by their relation to more central

464 Case Study 3

structures and by the form-meaning correspondences given by the central principles.

This general idea is not new. The idea of central and noncentral clause structures goes back to Zeilig Harris's "kernel" sentences (Harris 1957) and has its roots in the tradition of descriptive grammars in which the "most basic" sentence types are described first, and the "less basic" types described later. The idea that form-meaning correspondences are direct and regular for central clause types goes back to Katz and Postal's inte­grated. theory (Katz and Postal 1964) and to generative semantics (Lakoff 1963). The most popular version of this theory was Chomsky's Aspects theory (Chomsky 1965), in which "deep structures" played a role similar to that of central clause types. A recent incarnation of this view of central clause types is Gazdar, Klein, Pullum, and Sag's Generalized Phrase-Structure Grammar (1985), which makes a grammar-metagrammar dis­tinction, which is roughly equivalent to the^ cehtral-noncentral dis­tinction. Other contemporary theories of grammar make a similar distinction.

None of these theories, or other contemporary theories, have been able to account adequately for the phenomena we will be discussing. There are a number of reasons fob this.

First, the theory of meaning used in these theories has, on the whole, been a version of objectivist semantics. As we will see, a cognitive seman­tics is required instead—one that makes use of prepositional, meta-phoric, and metonymic models, as well as the theory of mental spaces.

Second, such theories do not have an adequate account—if they have any account at all—of grammatical constructions, direct pairings of parameters of form with parameters of meaning.

Third, such theories do not have an adequate concept of a category. As we will see, radial categories with prototypical centers are needed.

Fourth, other theories of grammar operate with the usual dichotomy between predictability and arbitrariness. Instead, a concept of motivation is "needed.

Fifth, other theories relate, central to noncentral cases by syntactic transformations, or roughly equivalent devices—metarules, redundancy rules, and the like. These usually provide only for syntactic relationships. We will argue that what is needed instead is a concept of an "ecological location" within a grammatical system as a whole. Ecological locations are defined by central subcategories plus links of noncentral to central categories. Such links are characterized by what we shall call "based on" relations. These differ from transformations and their equivalents in that they may specify semantic and pragmatic relationships, including some

Cognitive Grammar 465

defined by metaphoric and metonymic models, and others defined by the addition (and sometimes the subtraction) of whole ICMs.

Sixth, most other theories (with the notable exception of generative semantics) assume that syntactic categories and grammatical relations are "autonomous," that is, entirely independent of meaning and use. We will see that, instead, syntactic categories and grammatical relations have radial .structure, with a prototypical center that is predictable on semantic grounds; the noncentral members constitute extensions which are not pre-dictable on a semantic basis, but which are typically semantically or pragmatically ^motivated.

Seventh, most other theories (again with the notable exception of gen­erative semantics) assume that syntactic constraints on the occurrence of a construction cannot be predicted from the meaning of that construc­tion. We will show that the opposite is the cas&, and that a great many syn­tactic constraints can be predicted on semantic grounds.

Eighth, other theories of grammar assume some form of atomism, namely, that the meaning of a grammatical construction is a computable function of the meanings of its parts. We wiU argue instead that gram-1. matical constructions in general are holistic, that is, that the,,me.aningJaLu f^e'whole constryctwn^is.moriyated by theJnea^gs.ofihe.-parts^huUs'f notconnputable from^ them.

Ninth, just about all other theories assume that there is a clear division between the grammar and the lexicon, with the grammar providing struc­tures and the lexicon providing meaningful words to plug into gram­matical structures. We will see that such a clear division is problematic, and that there is more likely a continuum between the grammar and the lexicon (see Fillmore, Kay, and O'Connor, to appear; Sadock 1984).

Tenth, most other contemporary theories assume that grammar is independent of the rest of cognition. As we will show, grammar is depen­dent on many other aspects of cognition, which we will have represented via prototype theory, cognitive models, mental spaces, etc. In fact, we will argue that grammars are defined relative to nonuniversal conceptual systems.

The purpose of this .case study is to show that, because of these differ­ences, a cognitive grammar can provide both an adequate description and explanation of the complexities of fftere-constructions, while other theories cannot. Existential f/iere-constructions, as in There's a fly in my soup, were originally viewed as a phenomenon-that-could be handled naturally and easily by generative theories. Over the years, however, they have proven to be extremely recalcitrant, It was also thought at first that the relationship between existential and deictic t/iere-constructions,

466 Case Study 3

like There goes a fly, into your soup, could be easily accounted for by transformational grammar (cf. Fillmore 1968, Lyons 1968, Thome 1973, Kuno 1971). Ultimately, this proved impossible for reasons that we will discuss below. The problem was not that there was no systematic relation­ship between the constructions; the problem, instead, lay with the theory of transformational grammar.

This case study thus brings to bear empirical evidence concerning the following ten issues:

1. Is objectivist semantics adequate, or is a cognitive semantics necessary?

2. Are grammatical constructions epiphenomena, or do they have a real cognitive status?

3. Are classical categories sufficient for grammar, or is prototype theory required?

4. Will a predictable-arbitrary dichotomy suffice for grammar, or is the concept of motivation necessary?

5. Can transformations (and their rough equivalents, metarules, re­dundancy rules, and the like) account for relationships among construc­tions, or do we need instead the concept of "ecologial location" defined in terms of radial structure and the "based on" relation?

6. Are syntactic categories and grammatical relations autonomous, or are they radially structured with a semantically defined center?

7. Must all syntactic constraints be accounted for only by rules of syn­tax that are oblivious to meaning, or can a great many syntactic con­straints be accounted for on semantic grounds?

8. Is the meaning of every grammatical construction a computable function of the meaning of its parts, or are there constructions whose meaning is motivated by the meanings of its parts but is not computable from them by general rules?

9. Is there a strict dichotomy between grammar and lexicon, or is there a continuum between the two?

10. Is grammar a separate "module," independent of other aspects of cognition, or does it make use of other aspects of cognition such as proto­type categorization, cognitive models, and mental spaces?

On the whole, generative approaches to grammar assume the first al­ternative in each case, while cognitive grammar maintains the second alternative. These are not the only differences between generative and cognitive grammar, but they are the ones that matter for an understand­ing of fAere-constructions. As we go along, we will try to show that the generative assumptions are inadequate, and that a cognitive approach is needed.

Grammatical Constructions 467 Grammatical Constructions

I am using the term grammatical construction in a somewhat enriched ver­sion of its traditional sense. Traditional grammarians took it for granted that the grammar of a language could be described in terms of a collection of constructions, where each construction was a configuration of syntactic elements (like clause, noun, preposition, gerund, etc.) paired with a meaning and/or use associated with that syntactic configuration. I will be speaking of grammatical constructions in just this sense. Each construc­tion will be a form-meaning pair (F,M), where F is a set of conditions on syntactic and phonological form and M is a set of conditions on meaning and use. I have been working with my colleagues Charles Fillmore and Paul Kay on a general theory of such constructions (see Fillmore, Kay, and O'Connor, to appear).

I will try to be somewhat more precise than traditional grammarians were, and I will also be concerned with such contemporary issues as the relationships among constructions and the psychological reality of indi­vidual constructions and categories of constructions. It is my ultimate goal to place the theory of grammatical constructions within a general theory of symbolic models. Symbolic models are pairings of models of form with other cognitive models. A general theory of language would include an account of lexical items, motivated idioms, grammatical con­structions, morphemes, etc. An even more general theory of symbolic models would account for the understanding of all sorts of form-meaning correspondences that have a cognitive reality. Such a general theory is far beyond the present study; for the present we will be concerned only with grammatical constructions.

It should be pointed out the concept grammatical construction is extremely controversial in contemporary linguistics. In most contempo­rary formal theories, grammatical constructions in our sense have no status whatever. They are considered epiphenomena—consequences of more general rules of a very different character. The types of general rules differ from theory to theory. What should be borne in mind throughout the following discussion is that grammatical constructions are left out of other contemporary formal theories not for any empirical rea­sons, but for reasons internal to those theories. Those theories are very much the poorer for it. Theories of grammar without grammatical con­structions simply do not account for anything approaching the full range of grammatical facts of any language. In fact, they are limited to a rela­tively small range of phenomena. It is part of the objective of this case study to demonstrate the utility of grammatical constructions, as well as the utility of prototype theory. The point is to show that when prototype

468 Case Study 3

theory is taken together with grammatical constructions, it is possible to state regularities that cannot otherwise be stated. To my knowledge, there is no contemporary generative theory that accounts for anything approaching the range of phenomena discussed in this case study.

Some Basic Facts

Let us begin with certain basic and relatively well-known facts. There are two types of (Aere-constructions, the deictics and the existentials:

Deictic: There's Harry with his red hat on. Existential: There was a man shot last night.

The term deictic is used to refer to words like this and that that are used in pointing or are interpretable only relative to the context in which the sen­tence is uttered. In There's Harry with his red hat on, the word there is used to pick out a location relative to the speaker. It is a locative adverb in that it picks out a location, and it is deictic in that the location it picks out is relative to the speaker. In the existential case, it is not location but exis­tence which is at issue, in this case the existence of an event. The there in There was a man shot last night does not pick out a location. Therefore, while the deictic there may take an accompanying pointing gesture the existential there cannot. In addition, the deictic there bears stress (not necessarily primary sentence stress), while existential there does not bear stress at all and its vowel may reduce.

Moreover, there are major syntactic differences between the construc­tions. The differences involve subjecthood, negatibility, embeddability, and alternation with here.

Subjecthood: Existential there is grammatical subject; deictic there is not.

There are at least two clear tests that show this. Tag-questions require that a pronoun corresponding to the subject appear in the tag. This works fine for existentials, but not for deictics (asterisks indicate unaccept-ability):

- There was a man shot, wasn't there?

- "There's Harry with his red hat on, isn't there?

The existential works like the simple sentence,

- Debbie saw Harry last night, didn't she? where she refers back to the subject, Debbie. Since deictic there is not a

Some Basic Facts 469

subject, it cannot occur in the tag. Similarly, only subjects can occur in what are called /•awn^-constructions. Thus, in a central clause like/o/w is sick, John is subject. It can therefore be subject of cases of raising-constructions like ___ is believed to ___ and ——— is likely to

- John is sick.

- John is believed to be sick.

- John is likely to be sick.

This also works for existential f/iere-constructions.

- There was a man shot.

- There was believed to have been a man shot.

- There is likely to be a man shot.

However, it does not work for deictic fAere-constructions.

- There is Harry with his red hat on.

- *There is believed to be Harry with his red hat on.

- "'There is likely to be Harry with his red hat on.

In the above example. Harry is the subject, not there. This is clear from the fact that it goes into the nominative case when it is pronominalized.

- There he is with his red hat on.

- ""There him is with his red hat on.

Note that inverted word order is forbidden with a pronoun subject:

- ""There is he with his red hat on. This restriction will be discussed below.

Negatability: Existential (/lere-constructions can be negated; deictic ^lere-constructions cannot be negated.

- There wasn't anyone in the room.

- ""There isn't Harry with his red hat on.

Embeddability: Existential constructions can be freely embedded in sub­ordinate clauses; deictic constructions can almost never be embedded (with exceptions to be discussed below).

- If there's anyone in the room with a red hat on, I'll be surprised.

- *If there's Harry in the room with his red hat on, I'll be surprised.

- I doubt that there's anyone in the kitchen.

- *I doubt that there's Harry in the kitchen.

470 Case Study 3

Alternation with here: Here can occur in deictic constructions, but not in existential constructions.

- There's Harry with our pizza!

- Here's Harry with our pizza!

- There will be a man shot tomorrow.

- '"Неге will be a man shot tomorrow.

Yet, despite these very considerable differences, it is possible to find cases where the deictic and existential differ superficially only in stress. (Capitals indicate stress.) The following sentences minimally differ in form, and one has to think twice to differentiate them.

Deictic: there's an ape flirting with Harriet. Existential: There's an APE flirting with Harriet.

Such similarities are by no means accidental. As Fillmore, Thorne, Kuno, and Lyons observed, it is no accident that the same word—there—occurs in both constructions. It is generally agreed that the deictic there is in some sense the more basic of the two and that the existential use is extended from the deictic use. The commonsense explanation seems to be that things that exist must exist in a location, or in slogan form: to be is to be located. I think this is essentially correct as far as it goes, but it is not an explanation, and needs to be made into one. However, the authors mentioned above attempted to relate the deictics and the existentials via a transformational rule. This attempt failed, and one of the principal rea­sons was that no transformational derivation could account for all the dif­ferences between the constructions. Thus, one of the central problems that we will have to deal with is:

How can we show the relationship between the existential and the deictic constructions while simultaneously accounting for the differences?

We will approach this question, among others, via a cognitively based grammar.

Some Pragmatic Distinctions

One of the things that grammatical constructions will enable us to repre­sent is the pairing of pragmatic conditions (namely, conditions on the use of a construction) with syntactic conditions (e.g., conditions on word order, etc.). In fact, one of the major ways we have of identifying gram­matical constructions is to find cases where pragmatic conditions are asso­ciated with syntactic conditions. In the case of deictic t/tere-constructions,

Some Pragmatic Distinctions 471

we can best reveal the pragmatic conditions on the construction by com­paring instances of (Aere-constructions to the corresponding simple sentences. This is especially important, since it is commonly believed that ^re-constructions have no special status as constructions at all, but rather are merely derived by regular transformational rules from the cor­responding simple sentences. According to such a proposal, a trans­formation moving adverbs to the front of a sentence would apply, say, to He comes here to yield Here he comes, which would presumably be a mere stylistic variation of the simple sentence. Since transformations are, by definition, not paired with either pragmatic conditions or semantic condi­tions, the transformational analysis predicts that there should be no important semantic or pragmatic differences between /Aere-constructions and simple sentences. To show that such differences do exist is to provide prima facie evidence for the existence of a construction, rather than a mere transformational variant. In other words, if we can show that Here he comes has different semantic and pragmatic properties from He comes here, then we can show that they are not merely related by a syntactic transformation and that Here he comes is really an instance of grammati­cal construction—a symbolic model that pairs syntactic conditions with semantic and pragmatic conditions.

Let us begin by considering the difference between sentences like:

- Here comes Harry. and

- Harry comes here.

In Here comes Harry, there is a true present tense, that is, comes makes reference to the time the sentence is uttered and it is instantaneous. In Harry comes here, the present tense form has a generic meaning, that is, it picks out a general area of time around the present, and comes desig­nates many instances of coming, not just one. This difference shows up in the way that these two constructions treat modifiers like/row time to time, which is restricted to generic, not instantaneous, time reference.

- Harry comes here from time to time.

- *Here comes Harry from time to time.

The here also works differently in the two constructions. In Harry comes here, here designates the location of the speaker. In Here comes Harry, come places Harry on a trajectory aimed toward the speaker. Here has two functions. It both designates the end point of that trajectory (the location of the speaker) and designates Harry's location as being closer to

472 Case Study 3

the speaker than there (the contrasting term) would indicate. This differ­ence shows up with locative adverbs that cannot refer to the location of the speaker, for example, around the corner:

- Here comes Harry around the corner.

- *Harry comes here around the corner.

In the first sentence,, here designates Harry's location as he rounds the comer, and places him on trajectory aimed toward the speaker. But in the second sentence, the here is at the location of the speaker, which is

inconsistent with around the corner.

A similar point can be made with a sentence where there is some varia­tion among speakers. Roughly half the speakers of .English I have asked find sentences like the following grammatical:

- %There comes Harry. (%indicates variation)

In this sentence, there indicates Harry's, location, which is on the trajec­tory defined by comes, but is not the end point of that trajectory. But in the next sentence, there indicates the end point of the trajectory defined

by comes:

- Harry comes there (regularly).

The two constructions also differ in their pragmatics, that is, in the con­ditions on their use. Thus, Harry comes here is a simple assertion like any other declarative sentence. But Here comes Harry is not only an asser­tion. In addition, the speaker is directing the attention of the hearer to the

location specified by here.

Correlating with these differences in semantics and pragmatics are major syntactic differences. The simple sentences do not have any of the syntactic constraints that гйеге-constructions have. Thus, they can take tag-questions.

- Harry comes here, doesn't he? They can occur in raising constructions.

- Harry is likely to come here. They can be negated.

- Harry doesn't come here anymore. They can be embedded freely.

- If Harry comes here, let me know.

- I doubt that Harry comes here.

Pragmatics and Syntax 473

In short, deictic rA^re-constructions have syntactic, semantic, and prag­matic properties that are markedly different from the corresponding sim­ple sentences. This suggests that fftere-constructions really are construc­tions—pairings of form and meaning—rather than mere Stylistic variants of simple sentences.

Pragmatics and Syntax

We have defined a grammatical construction in such a way as to include semantic and pragmatic constraints. As we have seen, ordinary simple sentences like Harry comes here differ considerably in such constraints from instances of deictic r/iere-constructions like Here comes Harry. We are now in a position to see that defining constructions to include both semantic and pragmatic constraints allows us to predict much of their syn­tactic behavior.

We will begin by asking a very traditional question in syntax: what kinds of syntactic configurations can occur in what kinds of subordinate clauses? It is a syntactic question of the most basic kind, and it is usually assumed that a general answer to this question can be given in syntactic terms alone. We will try to show that this is not the case and that a general answer can be provided only within a theory that contains grammatical constructions that include semantic and pragmatic constraints.

We will approach the question by looking at what I will call per forma­tive subordinate clauses. The basic properties we will be concerned with can be seen in the sentences:

- I'm leaving, because here comes my bus.

- *I'm leaving, if here comes my bus.

Z/-clauses and because-clauses are both adverbial clauses. It is usually assumed that whatever can occur in one kind of adverbial clause can occur in another. These examples indicate that that simply isn't true. Because-dauses permit constructions that (/-clauses forbid.

What is of particular interest in these examples is that the construction that differentiates them, as exemplified by Here comes my bus, is usually assumed to be a main-clause construction, that is, a construction that occurs only in main clauses. (See Emonds 1971, Hooper and Thompson 1973, and Green 1976.) What is a main-clause construction doing in a sub­ordinate clause at all? And why does it occur in because-c\auses but not in (/-clauses? Do other main-clause constructions act in the same way? And if so, which ones do, and why?

A quick look at other constructions that are supposed to occur only in main clauses shows that Here comes my bus is not alone. Take, for exam-

474 Case Study 3

pie, a negative question like Isn't it a beautiful day?! It shows exactly the

same behavior:

- We should go on a picnic, because isn't it a beautiful day!

- *We should go on a picnic, if isn't it a beautiful day!

This shows that whatever is going on, it is not restricted to individual con­structions. It must involve some general property of a class of construc­tions that usually occur only in main clauses. Exactly what property is it?

Before going on to try to answer these questions, one more peculiarity of this phenomenon should be pointed out. Main-clause constructions occur in frecauye-clauses only when they are in final position. Proposed because-c\wses do not permit them:

- *Because isn't it a beautiful day, we should go on a picnic.

- *If isn't it a beautiful day, we should go on a picnic.

- ""Because here comes the bus, I'm leaving.

- *If here comes the bus, I'm leaving.

Any adequate account of this phenomenon must explain why this is so.

Some Speech Act Constructions

I would like to suggest, as a first approximation, that what unites the con­structions in question is that they are all speech act constructions, that is, constructions that are restricted in their use to expressing certain illocu-tionary forces that are specified as part of the grammar of English. Let us consider a number of such constructions, together with their illocutionary

force constraints.

Deictic there-constructions direct the hearer's attention to something

present.

- There goes Harry! Negative questions convey positive hedged assertions.

- Didn't Harry leave? Inverted exclamations express surprise (see N. McCawley 1973).

- Boy! Is he ever tall! WH-exclamations express something that has just come to mind.

- What a fool he is! Rhetorical questions convey corresponding negative statements.

Some Speech Act Constructions 475

- Who on earth can stop Bernard? Reversal tags convey hedged assertions.

- He's coming, isn't he? All of the above occur in final because-c\auses:

- I'm gonna have breakfast now, because am I ever hungry!

- Tm gonna have breakfast now, if am I ever hungry!

- We should have another party, because what a good time everyone had at the last one!

- *We should have another party, if what a good time everyone had at the last one!

- The Knicks are going to win, because who on earth can stop Bernard?

- ""The Knicks are going to win, if who on earth can stop Bernard?

- I guess we should call off the picnic because it's raining, isn't it?

- "I guess we should call off the picnic if it's raining, isn't it?

However, not all speech act constructions occur in final because-c\aases. Compare, for example, the differences between reversal tags and redupli­cative tags.

- It's raining, isn't it? (reversal tag)

- It's raining, is it? (reduplicative tag)

Although reversal tags can occur in because-clauses, reduplicative tags cannot.

- *I guess we should call off the picnic because it's raining, is it?

The relevant difference between reversal and reduplicative tags seems to be that, whereas reversal tags convey assertions, reduplicative tags con­vey challenges to the assertions of others. We can see that this is the rele­vant distinction by looking at other speech act constructions that cannot occur in frecawe-clauses, for example, true questions and imperatives.

- *l'm staying because go home!

- *I'm leaving because which girl pinched me?

Why should rhetorical questions like Who can stop Bernard? occur in because-c\auses, while true questions like Which girl pinched me? cannot? The difference is that rhetorical questions convey statements (e.g., No one can stop Bernard) while true questions are requests for information. Such cases suggest the following hypothesis:

476 Case Study 3

Only speech act constructions that (directly or indirectly) convey statements can occur in performative subordinate clauses.

This hypothesis predicts that certain overt performatives but not others should be able to occur in because-clauses. The prediction is borne out, since only statements occur:

- I'm going to vote for Snurdley because I maintain that he's the only honest candidate.

- *I'm staying because I order you to leave.

- *I'm leaving because I ask you which girl pinched me.

The hypothesis also accounts for the range of speech act constructions that do occur. All of the cases cited above happen to convey correspond­ing statements:

- Here comes the bus! conveys

- The bus is coming.

- Isn't it a beautiful day? conveys

- It's a beautiful day.

- Am I ever hungry! conveys

- I'm hungry.

- What a good time everyone had!

conveys

- Everyone had a good time.

- Who on earth can stop Bernard? conveys

- No one can stop Bernard.

- It's raining, isn't it? conveys

- It's raining.

Thus we can account in a straightforward way for the relationship between the speech act constructions and the semantics of the because-clauses:

Some Speech Act Constructions 477

When speech act constructions occur in because-c\auses, the content of the statement conveyed by the speech act construction equals the con­tent of the because-c\ause.

Thus, in a sentence like

- I'm gonna have breakfast now, because am I ever hungry!

the exclamation Am I ever hungry! conveys the statement I'm hungry. Thus, the reason given in this sentence is the same as the reason given in

- I'm gonna have breakfast, because I'm hungry.

Thus performative subordinate clauses perform two functions at once. They perform a speech act that conventionally conveys a statement, and they give the content of that statement as a reason for the first statement.

This general principle covers all the cases given above and excludes nonoccurring cases like pure questions and orders. It allows rhetorical questions that convey statements, and it predicts that if an imperative construction were to conventionally convey a statement, it could occur in these clauses. As it happens, there is such an imperative. Compare the pure imperative.

- Find out which girl pinched me. with a sentence of the same syntactic form

- Consider which girl pinched me.

The first is simply an order. The second, however, assumes that the hearer already knows the answer. It directs the hearer to think about the answer and assumes that if the hearer does so, he will reach a specific con­clusion that the speaker already has in mind. It is a roundabout, but none­theless conventionalized, way of conveying a statement which is never overtly mentioned. As predicted, the difference between these two im­peratives is reflected in their ability to occur in because-clauses. The second can occur, while the first cannot.

- '"I'm staying because find out which girl pinched me.

- I'm staying because consider which girl pinched me.

Thus, we see that it is not the imperative construction per se that is ruled out. Most imperatives cannot occur in such clauses simply because they do not conventionally convey statements. Imperatives that do conven­tionally convey statements do occur in these because-c\smses.

478 Case Study 3

Initial Clauses

We are now in a position to explain why speech act constructions can occur in because-c\aases in final position, but not in initial position. Because-dwses in sentence-initial position are presupposed, while those in sentence-final position are not. Consider the sentence:

- I'm going to vote for McSwain instead of Polanski because McSwain can win in November.

In this sentence the speaker is asserting that McSwain can win in Novem­ber. However, in the following sentence, the speaker is taking it as a fore­gone conclusion that McSwain can win in November.

- Because McSwain can win in November, I'm going to vote for him

instead of Polanski. This can be seen even more clearly in the following examples.

- Do you think that John left early because he was tired?

- Do you think that, because he was tired, John left early?

- I doubt that John left early because he was tired.

- I doubt that, because he was tired, John left early.

In the first sentence of each pair, it is not necessarily taken for granted that John was tired, whereas in the second sentence of each pair it is.

Given the presuppositional character of initial fcecaM5e-clauses, we can explain why speech act constructions cannot occur in such clauses. In order for speech act constructions to occur in because-c\auses at all, they must convey statements. However, it is impossible to both state and pre­suppose something simultaneously. It is for this reason that speech act constructions cannot occur inside initial because-dauses.

Further evidence for this explanation has been brought to my attention by James D. McCawley, who observes that in certain situations proposed because-dmses do not have to be presupposed. Thus, in sentences like

- My boss wants me to vote for Polanski, but because McSwain can win in November, I'm going to vote for him.

In this case, the speaker can be asserting, not presupposing, that McSwain can win in November. McCawley observes that in such cases speech act constructions can occur in preposed because-dauses.

- I want to stay, but because here comes my bus, I'd better leave.

This shows that the constraint on the occurrence of speech act construc­tions is pragmatic rather than syntactic. That is, it is not the preposed

The Range of Clauses 479

position of the because-clauses that rules out speech act constructions;

rather, it is the presuppositional character of the position that rules out the speech act constructions. When that presuppositional character is re­moved, then speech act constructions can occur in preposed because-clauses.

The Range of Clauses

Let us now turn to the question of what kind of clauses speech act con­structions can occur in. We have seen that they occur in because-dauses. They also occur in clauses that begin with although, except, since and but.

Although

- I'm not going to vote for Snurdley, although I maintain that he's the best candidate.

- I've decided to stay, although here comes Harry—and you know what I think of him!

- I'm going to stay on my diet, although could I ever go for a dim sum brunch!

Except

- I'd stay a little longer, except here comes my bus!

- We really shouldn't go on a picnic, except it is a nice day, isn't it?

- I'd go swimming with you, except am I ever tired!

Since

- I'd better leave, since here comes my bus!

- I'm going to cheat on my taxes, since who will ever find out?

- No one's going to be there, since it's going to be boring, isn't it?

(It should be noted, incidentally, that since is formal and many of these constructions are mainly used in informal speech. This can lead to register incompatibility, as in ?He must be a great player, since what a shot he hit! This incompatibility is irrevelant to the present discussion.)

But

- I really should stay, but here comes my bus.

- I'm on a diet, but am I ever hungry!

In addition to (/-clauses, all other adverbial subordinate clauses exclude these speech act constructions, e.g., where, when, while, as, etc.

- ""There's Bill where isn't John sitting.

480 Case Study 3

- *Harry left when did he ever get hungry!

- -John left as didn't Bill come in?

- *John was sitting in his favorite chair reading while was Harry ever

sneaking up on him!

Speech act constructions occur in two classes of adverbials: reason adver-bials (because, since) and concessives (although, except, but). Actually, these two classes form a single more general class for the following rea­son: In sentences of the form "A although B" and "A but В," В is a reason for not A. For example, in a sentence like John stayed up although he was tired, being tired would be a reason for not staying up. In short, conces­sive clauses give reasons for the opposite of the main clause. The general­ization seems to be that speech act constructions occur in clauses express­ing reasons of either sort.

What we have arrived at is a single general principle:

Clauses expressing a reason allow speech act constructions that convey statements, and the content of the statement equals the reason

expressed.

The observant reader will have noticed that not all of the examples given have been subordinate clauses. But-clauses are coordinate. Yet with respect to this phenomenon they work the same way as other clauses ex­pressing reasons. In fact the general rule does not mention subordinate clauses at all. It only mentions "clauses expressing a reason." Such clauses may be subordinate and marked with subordinators like because, although, etc. They may be coordinate and marked with but. In addition, coordinate clauses marked with and and or may also take such construc­tions when one coordinate clause is expressing a reason for another.

- Here comes my bus and so I'd better leave.

Conclusions

The problem we began with was a conventional syntactic problem: What kinds of constructions can occur in what kinds of subordinate clauses? The solution to the problem requires constructions to be paired in the grammar with the illocutionary forces they express (both directly and indirectly). Once this is done, a general rule can be stated in purely semantic terms. In other words, a complex syntactic problem can be solved by a simple semantic principle—provided we have the means in the grammar to pair constructions with the meanings they convey. This phenomenon therefore provides support for a theory of grammatical con­structions of the sort we will be using in the remainder of this case

Pragmatics in the Grammar 481

study—one in which semantic and pragmatic information is paired with syntactic information, which permits generalizations about syntax to be stated in semantic and pragmatic terms.

Of course, grammatical construction theory is not the only contempo­rary theory that permits the direct pairing of syntactic and semantic infor­mation. However, it is the only theory I am aware of that permits the pairing of complex syntactic configurations with the appropriate prag­matics—in this case, conveyed illocutionary force. Grammatical con­struction theory permits such pragmatic factors to enter directly into the composition of sentences. Generative theories with an autonomous syn­tax cannot do this. For example, they have no way of generating speech act constructions in exactly the right subordinate clauses while not gen­erating them in the wrong subordinate clauses.

Pragmatics in the Grammar

Before we continue, an important but subtle distinction needs to be noted. The pragmatic constraints, that is, constraints on use, that we have been discussing are in the grammar of English. Constraints on use which are part of the grammar of a language are very different from knowledge about normal use which is not part of the grammar. Compare, for exam­ple, sentences like:

- There's Sadie.

- That's Sadie.

These two sentences are normally used in pretty much the same way: to direct the hearer's attention to something in the perceptual field of both speaker and hearer and to identify it with the expression given (in this case, Sadie). The semantic difference is minor: there directs attention to a location which has an entity located there, while that directs attention to the entity in that location.

But there is a major difference. In the case of There's Sadie, the speech act condition on the use of the sentence is absolutely required by the grammar of English. Deictic (/lere-clauses can be used only when the speaker is directing the attention of the hearer. However, in the case of That's Sadie, the given condition is part of our knowledge of how the con­struction is normally used, but not part of the grammar of English. Just because a construction is normally used in a given way, it doesn't follow that it necessarily has to be used in that way. That's Sadie is, therefore, free to be used in ways that There's Sadie cannot be used. For example, it can be negated, questioned, and embedded freely.

482 Case Study 3

-That isn't Sadie.

- Is that Sadie?

- I wonder if that's Sadie.

- I doubt that that's Sadie.

- I'm leaving if that's Sadie.

When we speak of pragmatic constraints on constructions, we will be speaking about those that arc part of the grammar of English, not those that are part of our knowledge about how constructions happen to be normally used.

Grammatical constructions, as we will speak of them below, are com­plex cognitive models with two dimensions: one characterizing para­meters of form and one characterizing parameters of meaning. We have seen that deictic f/iere-constructions fit this characterization. At this point we are in a position to begin to discuss prototype theory. We will show that there is not just one deictic f/iere-constructions, but a whole category of them, and that the category is not a classical category, but a category centered around a prototype.

Deictic TTie/'e-Constructions

We have seen nothing so far to indicate that there is more than one deictic йеге-construction. But as we will see shortly, there are many "subcon-structions," and they form a prototype-based category—a category con­sisting of constructions. It is also a category that cannot be defined by necessary and sufficient conditions. That is, it is impossible to give condi­tions that all and only the constructions in this category share. What we can do is characterize the category by giving the properties of the central construction and then give the properties that minimally distinguish each of the subconstructions from the central construction. The result is a cog­nitive structure that is very much like the category structures that occur elsewhere in the conceptual system. The point is that structures like this are to be expected, given the way people normally construct categories. Prototype theory thus explains why such a grouping of constructions should exist in a language. According to traditional generative theories, such a clustering of constructions is simply an anomaly.

Here is a list of examples of the various subconstructions in the cate­gory of deictic f/iere-constructions:

THE DEICTICS

Central: There's Harry with the red jacket on. Perceptual: There goes the bell now!

Deictic ТЪете-Constructions 483 Discourse: There's a nice point to bring up in class. Existence: There goes our last hope. Activity Start: There goes Harry, meditating again. Delivery; Here's your pizza, piping hot! Paragon: Now there was a real ballplayer! Exasperation: There goes Harry again, making a fool of himself. Narrative Focus: There I was in the middle of the jungle .... New Enterprise: Here I go, off to Africa.

Presentational: There on that hill will be built by the alumni of this uni­versity a ping-pong facility second to none.

Some are distinct subconstructions, some are not. Their status, as we will see, is anything but obvious. As we go through them one by one, we will see what the differences are. But before we begin, it is important to see that such sentences present problems. Consider the following phenomena:

Only presentational constructions permit locative phrases after there and full auxiliaries; the others do not permit locatives and allow only simple tenses:

- There on that hill will be built a new library, (presentational)

- There goes Harry, (central)

- "There on that hill goes Harry, (central plus impermissible locative phrase)

- There goes Harry with the red sneakers on. (central)

- "There is going Harry with the red sneakers on. (central plus imper­missible full auxiliary)

Only the narrative focus construction permits the past tense, and it does not permit come.

- Here I was, with a silly hat on. (narrative focus)

- Here he comes, with a silly hat on. (central)

- *Here I came, with a silly hat on. (narrative focus with come)

Only the delivery construction takes an elongated vowel, and it requires present tense:

- There was my pizza, ready to go. (narrative focus)

- The e с ore's your pizza, ready to go. (delivery)

- *The e e ere was your pizza, ready to go. (delivery with past tense)

484 Case Study 3

A remarkable phenomenon occurs in the perceptual deictic. A percept (say, the pain in my knee) can be replaced by the thing perceived (in this

case, my knee).

- There goes the pain in my knee.

- There goes my knee.

But this is not possible when the verb is come. Suppose you feel a twinge in your knee slightly before the pain appears. You can report this with the first sentence, but not the second.

- Here comes the pain in my knee.

- "'Неге comes my knee.

Similarly, suppose you have an alarm clock that clicks a few seconds before it is set to go off, and when it goes off it makes a beep. Upon hear­ing the warning click, one might say

- Here comes the beep. but not

- '"Here comes the alarm clock. However, when the alarm clock goes off one can say either

- There goes the beep. or

- There goes the alarm clock.

One last example: The discourse deictic construction, unlike the central construction, does not permit go. To see the contrast, suppose that you see a beautiful car go by. You can say either

- There's a beautiful car. or

- There goes a beautiful car.

However, if you are listening to a lecture and you hear the lecturer make a beautiful point, you can say

- There's a beautiful point. but not

- "There goes a beautiful point.

These are subtle distinctions, and previous accounts of these construc-

Notation 485

tions have not been able to explain them. Within the theory of gram­matical constructions, an explanation is possible. In attempting to pro­vide both descriptions and explanations, I will adopt the following strategy: Describe the central construction in very great detail, enough to permit a description of all the variations on it. Enough detail must also be given to distinguish the deictic constructions from the existential con­structions. After this is done, the noncentral constructions can be de­scribed very simply as variations of the central construction, and their peculiarities can be explained on this basis.

Notation

Because the phenomena we are discussing are extremely complex, it is necessary to use some sort of notation to keep track of everything in­volved in the analysis. The notation will be essentially the same as the notation used in case study 1 to characterize cognitive models. This nota­tion has a number of virtues. First, it is the only notation I know of that is sufficiently precise to describe these phenomena. Second, it shows the unity between the theory of cognitive models and the theory of gram­matical constructions, since cognitive models are used in grammatical constructions to represent meaning. Third, it has empirical conse­quences, since it is the only notational system now in use that can charac­terize syntactic amalgams (for details, see Lakoff 1974). An amalgam is a case where more than one proposition is represented in a single clause. This situation arises in deictic ritere-constructions. Consider the example

- %There comes Harry with his red hat on.

This is syntactically a single clause. But it corresponds to three semantic propositions:

- Harry is there.

- Harry is coming, that is, moving toward the speaker.

- Harry has his red hat on.

We will be using the following notation: a syntactic element, for instance the clause There's Harry with his red hat on, will be represented by an integer, say zero. The corresponding semantic elements, for instance the three propositions mentioned above, will each be represented by zero fol­lowed by one, two or three prime signs. When a syntactic element is rep­resented by an integer, the corresponding semantic element will be repre­sented by the same integer followed by one or more primes. The number of primes distinguishes among the different semantic elements corre­sponding to individual syntactic elements. Cases of this sort are common

486 Case Study 3

in natural language and the notation adopted is one way (by no means the only possible way) of describing such cases.

Cognitive Systems versus Formal Systems

The term formal system has been used in linguistics in two distinct senses:

Practical formal system; A system of principles of some sort expressed precisely, often in a notation which permits one to give precise names to concepts, to state hypotheses in appropriate detail, and to make detailed predictions.

Technical formal system: A special kind of mathematical system of pro­duction rules, in which arbitrary symbols are manipulated in an algorithmic fashion without regard to their meaning.

Generative linguistics makes use of technical formal systems, not merely practical formal systems. The difference is crucial, since technical formal systems have many properties that practical formal systems do not have. For example, the technical concept of "generative capacity" is defined only for technical formal systems, not for practical formal systems in gen­eral. Moreover, technical formal systems are defined in such a way that the meaning of the symbols does not enter into the rules of the system. Practical formal systems have no such limitations.

The basic metaphor of generative linguistics is that a grammar of a lan­guage is a technical formal system restricted to certain kinds of formal rules (e.g., phrase structure rules, transformations, etc.). However, many linguists have used the notational system of transformational gram­mar as a practical formal system, and a great deal of confusion has arisen as a result.

As a working linguist and cognitive scientist, I will not be using the technical formal system metaphor for grammars, since it is irrelevant to the concerns of cognitive science, and since it cannot be the basis for a theory of meaning. My concern, like that of other cognitive linguists, is to ask what we need to hypothesize about the nature of cognition in general in order to account for the phenomena of natural language. The notation I will be using is a practical formal system that has the merit of permitting appropriate precision. I view it as a practical notation for characterizing certain relevant aspects of cognitive models. It has no cosmic significance. It is simply one way of providing a precise description of certain things that need to be described precisely.

Our notation, incidentally, does not capture every relevant aspect of grammatical constructions, For example, grammatical constructions may constitute gestalts, where the whole is conceptually simpler than the sum

Cognitive Systems versus Formal Systems 487

of the parts. Our notation does not capture the gestalt characteristics of constructions. Another very important thing it cannot characterize ade­quately is the spatialization of form. As we argued above, syntactic form is characterized in image-schematic terms—using part-whole schemas, linear order, closeness, centrality, size, containers, etc. The notation we will be using describes such schemas but doesn't utilize them directly in the notation.

Throughout what follows it should be borne in mind that we are de­scribing different kinds of entities than those described by any technical formal system of the sort used in generative linguistics. Here are some differences:

- Cognitive models and grammatical constructions use prototype theory to account for categorization. Technical formal systems do not use prototype theory.

- Cognitive models and grammatical constructions have gestalt proper­ties. That is, the wholes may by psychologically simpler than the parts. Technical formal systems, on the other hand, are atomistic— that is, the wholes are just collections of the parts.

- In technical formal systems, syntactic elements are arbitrary symbols. In cognitive grammar, syntactic elements are not arbitrary but are motivated by corresponding semantic elements. The concept of motivation does not exist in technical formal systems. It will be dis­cussed in some detail below.

- In technical formal systems, there is no notion of the ecology of the system and no concept of an ecological niche, a place where some­thing fits into the system. Cognitive systems have such an ecology, and we will be discussing it below. The concept of a ecological niche will turn out to be an important part of the theory of grammatical constructions.

- Within classical formal linguistics, there is a concern with such mathe­matical notions as the generative power of formal systems and with ways of limiting generative power by constraining the forms of rules. These technical mathematical concerns are of no interest whatever in a cognitively based grammar. We are interested in the way general cog­nition places constraints on human languages and not in constraints having to do with cognitively irrelevant concepts like generative power. Any choice among alternative descriptions must be based on such cognitive criteria as the ability to account for language acquisi­tion, the ability of a grammar to cohere with a conceptual system, the ease of cognitive processing, and the overall ecology of the grammar.

488 Case Study 3

- In technical formal systems, syntax is independent of semantics and pragmatics. In cognitively based linguistics, syntax is to a very signifi­cant extent (though by no means entirely) dependent on semantics, pragmatics, and communicative function.

For these reasons, the system that we will be discussing is not just another technical formal system of the sort that a great many linguists have justifiably become disillusioned with. It is, instead, an attempt to provide a precise characterization of some central concepts in cognitive

linguistics.

Before we go on, there is one further possible confusion that ought to

be cleared up. It is sometimes maintained that the mind uses technical formal systems (e.g., phrase structure or transformational grammars). If this were true, some type of generative linguistics would be cognitively correct. One of the goals of this case study is to show that that is not the case, by actually working out cognitively based descriptions and explana­tions of subtle linguistic phenomena that technical formal systems seem

not to be able to deal with.

Linguists trained within the generative tradition have sometimes claimed that only technical formal systems allow for the kind of precise formulation of principles that is necessary to adequately describe and explain linguistic phenomena. What is implicit in this claim is that cogni­tively based accounts of linguistic phenomena are necessarily imprecise and therefore incapable of providing adequate descriptions and explana­tions. Another goal of this case study is to take a problem area in English grammar and provide for it an account that is much more adequate, both descriptively and explanatorily, than any account in terms of mathemati­cal formal systems can be. This is one reason why the present case study is

as detailed as it is. Finally, the present study is by no means antimathematical. It is quite

conceivable that a mathematics appropriate to the kind of cognitive sys­tems discussed here could be developed. In fact, 1 am rather hopeful about this. It is one of the reasons that I have taken the trouble to de­scribe the phenomena as precisely as I have. I object only to the use of an inappropriate mathematics such as technical formal systems.

Some Notational Prerequisites

Each construction will be described in terms of the following parameters of form and meaning:

Parameters of Meaning

The Central ICM: An Experiential Gestalt 489

-Semantic elements (e.g., entities, locations, predicates, etc.)

-Overall semantics (e.g., any idealized cognitive model associated with the construction as a whole). The ICMs will often include back­ground conditions, speech act conditions, and a specification of the function of the construction.

Parameters of Form

- Syntactic elements (e.g., clause, noun phrase, verb, etc.)

- Lexical elements (e.g., here, there, come, go, be, etc.)

- Syntactic conditions (e.g., linear order of elements, grammatical rela­tions such as subject and object, optionality of elements, etc.)

- Phonological conditions (e.g., presence or absence of stress, vowel length, etc.)

Form-Meaning Pairings

Whenever there is a pairing between a syntactic element and a seman­tic element, we will express the pairing in the following way:

- The syntactic element will be represented by an integer, ('.

- The semantic element with be represented by Г, the same integer followed by a prime.

No primacy of either syntax or semantics is indicated by this notation, though in general we will be trying to predict as much as possible of the syntax from the semantics. All that is indicated is a pairing of the form ((',('}. In some cases, one syntactic element will be paired with more than one semantic element. In such cases we will use more than one prime, i'", etc.

We are now in a position to begin the first of our major projects, moti­vating the syntax of deictic (/lere-constructions on the basis of their parameters of meaning. We will begin with the idealized cognitive model (ICM) that represents the parameters of meaning of the central deictic construction. Once we do that, we can go on to account for various parameters of form of the central construction.

The Central ICM: An Experiential Gestalt

It is one of the principal findings of prototype theory that certain clusters of conditions are more basic to human experience than other clusters and also more basic than individual conditions in the cluster. Lakoff and Johnson (1980) refer to such a cluster as an experiential gestalt. Such a gestalt is often representable by an ICM. It should be borne in mind that in such cases the entire ICM is understood as being psychologically sim-

490 Case Study 3

pier than its parts—hence the term gestalt. Although a great many condi­tions may enter into the description of such a gestalt, it is important to realize that the complexity of the description is an artifact of our nota-tional system. If we had a notational system that reflected psychological reality, the entire ICM would be representable by a simple description and its parts by a more complex one.

The reason it is important to bear this in mind is that it is normal to think of descriptions in the following way; the more complex the descrip­tion, the more complex the concept. In the case of experiential gestalts, however, the reverse may hold: a complex description may correspond to a cognitively simple concept, while a relatively simple description of one of the parts of the concept may be cognitively more complex. This is so in the case of experiential gestalt we are about to describe.

One of the most basic things people do is point out things to other peo­ple. English-speaking children start to do this at the two-word stage with utterances like "Da shoe" (There's a/the shoe). When we pick out all the relevant parts of this very simple experiential gestalt, we find a fairly com­plex cluster of conditions:

It is assumed as a background that some entity exists and is present at some location in the speaker's visual field, that the speaker is directing his attention at it, and that the hearer is interested in its whereabouts but does not have his attention focused on it and may not even know that it is present. The speaker then directs the hearer's attention to the location of the entity (perhaps accompanied by a pointing gesture) and brings it to the hearer's attention that the entity is at the specified loca­tion. Additionally (for older children and adults), if the entity is mov­ing, the motion may be indicated. And the speaker may choose to describe the entity or its location.

Adopting the format given above for representing parameters of mean­ing, we can pick apart this experiential gestalt in even more detail and represent it as follows:

The Pointing-Out ICM

Semantic Elements 5: Speaker Я: Hearer

1': a location

2': a locational predicate

2": a predicate of motion

3': an entity

4': a predicate

Restricted Prediction and Cognitive Explanation 491

0': a proposition of the form 2'(3',1') (the entity is at the location) O": a proposition of the form 2"(3') (the entity is moving) 0'": a proposition of the form 4'(3') (the predicate holds of the entity) (0" andO'" are optional.) Overall Semantics Speech Act Background

B-l: 3' exists

B-2: 0' (3' is located at Г)

B-3: 3' is in 5's visual field

B-4: S is focusing his attention on 3'

B-5: S assumes H is not focusing his attention on 3' Speech Act

5 directs fTs attention to Г Functional Condition

To focus /fs awareness on 3' Gesture

S gestures in the direction of Г, typically by pointing or turning the head. (optional)

In what follows, I will refer to all of the above simply as the pointing-out ICM.

Restricted Prediction and Cognitive Explanation

Part of the program of cognitive grammar is to show how aspects of form can follow from aspects of meaning. We will try to show, for example, that almost every syntactic property of the central deictic construction is a consequence of the fact that it is used to express the pointing-out ICM. There is a long tradition in linguistics of attempting to predict form from meaning. To date, none of these attempts has succeeded. Take some well-known and relatively simple-minded attempts:

- Actions are expressed by verbs.

- States are expressed by adjectives.

- Physical objects are expressed by nouns (or noun phrases).

If we take such principles as absolutes that hold for all cases, then there are abundant counterexamples. Thus, states like knowing can be expressed by verbs, as in I know the answer. Actions can be expressed by adjectives, as in Be aggressive! or Be careful! An expression denoting a physical object can be a verb, as in He hammered in the nail.

492 Case Study 3

Yet, despite such counterexamples, the principles proposed above are by no means completely wrong. They are wrong only in having been pro­posed as absolutes. There are a number of reasons why I think attempts

to do so have failed.

First, it was assumed that they should work for all cases in a language. Second, it was assumed that they were all to be universal. Third, it was assumed that they were to be absolute, rather than rela­tive to other factors in a language.

Fourth, it was assumed that everything was either fully predictable or arbitrary. No other alternative was possible.

I would like to suggest an approach that avoids these pitfalls.

- Such principles may be limited to prototypical cases, in particular, to central subcategories of radial categories. We will refer to these as

central principles.

- Such principles may be language-particular, rather than universal.

- Such principles may make use of the "ecological location" of a con­struction within a linguistic system.

- Specific cases may preempt general principles.

Central principles play a dual role. First, they characterize form-meaning regularities for central subcategories, e.g., prototypical clauses, nouns, verbs, adjectives, subjects, etc. Second, they characterize the way in which noncentral cases are like central cases. That is, they help charac­terize what it means for a noncentral case to be motivated by central

cases.

A grammatical system is characterized as follows:

First, given a conceptual system, the central categories of the language are characterized by the central principles linking form and content.

Second, noncentral categories are "located" in the grammatical system via relations to central (or more central) categories. The relation which "locates," say, one constructional category relative to another is called the based-on relation. : ;

Third, less-central categories are characterized by those minimal differ­ences that distinguish them from the categories they are based on.

Fourth, central principles are used to characterize redundancies in the following way: Any property of a noncentral category that can be pre­dicted by central principles from a specification of minimal differences is redundant. Such prediction will be referred to as "restricted predic­tion," since it is prediction restricted to a given ecological location.

Restricted Prediction and Cognitive Explanation 493

Fifth, relative motivation is characterized as follows: the more the properties of a given category are redundant, the more it is motivated by its ecological location, and the better it fits into the system as a whole.

Sixth, an optimal ecological structure for a language is a structure that maximizes motivation.

There is a tradition in linguistic analysis of identifying an optimal linguis­tic analysis (according to some criterion of optimality) with the cogni-tively correct analysis for some "ideal" speaker of the language (cf. Chomsky, 1965). I will follow that tradition, even though I am acutely aware of its pitfalls, which are legion. None of us, after all, is ideal. Our individual grammars may not be structured optimally. Moreover, there may be a point in development at which our grammars "freeze" and no further optimization of what we have learned up till then takes place. Per­haps the role of such an optimal grammar is that it represents the maximal motivations used in the speech community in general, though no individ­ual may make full use of them. That is, a speaker may learn certain con­structions by rote rather than fitting them into his grammar as a whole. But it is unlikely that any individual learns all constructions by rote, with­out understanding them as being motivated in any way by the rest of the grammar.

One thing we can be sure of is that a great deal of optimization of the sort we will be describing does occur. We know this from the fact that when new constructions are added to language, they are not random. Rather, they are motivated by existing structures, just as new senses of words are motivated by existing senses. New constructions and words that "fit well" into a language are easier to learn than those that don't. The theory we are proposing offers a characterization of just what it means to "fit well."

The kind of optimal analysis we will be giving—one which maximizes motivation—also allows us to characterize actual systems that may be nonoptimal. It will allow us to characterize individual variation and to ask whether speakers really use maximally motivated systems, and if not, just how their systems are motivated. It will also allow us to investigate the role that motivation plays in 4inguistic change and in language-'acquisi-tion.

Before we embark on a spate of restricted predictions, it is important to point out what cognitive relevance such predictions are supposed, to have. We are not claiming that speakers of natural languages go about making such predictions every time they produce an utterance. Quite the opposite. What speakers do is use a linguistic system that fits together

494 Case Study 3

well and is cognitively efficient for their purposes. What we are trying to do is characterize exactly what it means for a linguistic system to "fit to­gether well" and be "cognitively efficient." We are attempting to charac­terize these concepts using the analytic notion "restricted prediction" as we have defined it. Another way of thinking of restricted prediction is that it characterizes the concept of "systemic redundancy," that is, the redun­dancy of the system as a whole given its overall ecological structure. We will discuss this concept in somewhat more detail after we have seen some examples.

We are trying to give cognitive explanations for linguistic phenomena. Part of our goal is to understand why languages are structured the way they are and what that can tell us about cognitive organization. Toward this end, we are trying to develop techniques of linguistic analysis that will allow us to get insight into cognitive organization. The hypothesis that we are offering is that restricted prediction (in the special sense in which we are defining ft) is a measure of cognitive efficiency, given the linguistic and conceptual system as a whole. This in turn can allow us to explain why languages are structured the way they are.

If we can show, for example, that most of the syntax of the central deictic construction can be predicted, in our sense of the term, from the pointing-out ICM, then we will have shown that in English the central deictic construction is a cognitively efficient way to express the pointing-out ICM. Moreover, we will have shown in detail exactly why it is a cogni­tively efficient construction, given the rest of English. That is, we will have shown why it is easy to learn, remember, and use. And we will have explained, in cognitive terms, why it has the peculiar properties it has.

In the predictions to be made below, I am taking for granted certain central principles governing the correspondence of form and content in prototypical cases. Some appear to be universal, while others are English-specific. It is beyond the scope of this study to justify all of them or even to state all of them as precisely as necessary. That will have to be done in a fully worked out and fully explicit cognitive grammar. What we can do is give a few simple examples of the sort that we will be making use of below. Most of the central principles are relatively obvious, so obvious as to hardly be worth stating. We are bothering to state them only to pro­vide some simple clear examples. We will state each principle both in sim­ple English, and in the notation we will be using. Each of the examples given will be used in the discussion below.

Some Central Principles (CP) CP1: Clauses correspond to propositions.

CLAUSE (X) iff PROPOSITION (X')

Restricted Prediction and Cognitive Explanation 495 CP2: Noun phrases correspond to entities.

NP (X) iff entity (X') CP3: Motion predicates are expressed as verbs.

If motion predicate (X'), then V(X) CP4: Location predicates are expressed as verbs.

If LOCATION PREDICATE (X'), then V(X)

CP5: Locative adverbs correspond to locations.

If LOCATIVE ADVERB (X) iff LOCATION (X")

CP6: Parts of a semantic structure correspond to parts of the correspond­ing syntactic structure.

ракт(л", У) iff part(a', У)

CP7; Semantically optional parts of ICMs correspond to syntactically optional elements.

If part (А",У) and optional (X'), then optional (X)

Some Detailed Predictions The central deictic construction is based on the simple central clause and expresses the pointing-out ICM. Given this, we can make the following predictions:

Prediction 1: 0 is a clause.

A consequence of CP1. Since 0', О", and 0'" are propositions which are expressed by 0, 0 is a clause.

Prediction 2: 0 has parts 1, 2, 3, and 4.

A consequence of CP6. Since Г, 2', 3', and 4' are parts of the proposi­tions 0', О", and 0'", it follows that 1, 2, 3, and 4 are parts of 0.

Prediction 3: 1 is a deictic locative adverb.

It is a consequence of CP5 that 1 is a locative ad verb. Moreover, an ele­ment located relative to the context of utterance is deictic. Since Г is a location relative to the speaker, it is therefore relative to the context of utterance; thus, 1 is a deictic locative adverb.

Prediction 4: 3 is a noun phrase.

A consequence of CP2. Since 3' is an entity, 3 is a noun phrase. Prediction 5: 3 is not a sentential complement.

Since 3' is a physically present, visible entity, it cannot be a proposition or an event or a state. Since only propositions, events, or states are ex­pressed by sentential complements, 3 cannot be a sentential complement.

496 Case Study 3

Prediction 6: 2 is a verb.

A consequence of CP3 and CP4. Since 2' is a locational predicate and 2" is a predicate of motion, it follows that 2 is a verb.

Prediction 7: 2 is in the simple present tense.

In the pointing-out ICM, the predication of location or motion is simul­taneous with the speech act. Therefore, it must be present tense, which expresses simultaneity with the speech act. Since the pointing-out event (the speech act) is instantaneous, and since the present tense marks simultaneity with that speech act, there is no progressive or perfect aspect, and hence no auxiliary verbs. Since the pointing-out ICM has no indication of modality, there are no modals. Consequently, 2 is in the sim­ple present tense. Prediction 8: 4 is optional.

A consequence of CP7. In the pointing-out ICM, the proposition describ­ing 3' is optional. The predicate 4' occurs only in that proposition. Since 4' is optional, the syntactic element expressing it, namely 4, is optional.

Prediction 9: 1 is first.

The speaker is directing immediate attention to location 1'. To direct im­mediate attention to something, you have to mention it immediately. Therefore, the syntactic element expressing Г must come first.

Prediction 10: 3 is subject of 0.

In central clauses, the subject of the clause must be a noun phrase ex­pressing an argument of one of the propositions expressed by the clause. It must also be included among the semantic elements in the parameters of meaning of the construction. There are only two possible candidates, 3 and 1, since 3' and Г are the only arguments listed among the parame­ters of meaning. But 1 is ruled out because, here, as in central clauses, an adverb cannot be the subject. That leaves 3.

Lexical Choices

In this construction there are three cases where lexical items are chosen:

the deictic locative adverb (here or there), a locational predicate (be, sit, stand, or lie), and a predicate of motion (go or come). Most of these specific lexical items, but not all, are predictable from general principles. And the ones that are not predictable are nonetheless not arbitrary.

This construction contains lexical elements because it has semantic ele­ments that are tightly restricted: a location relative to the speaker, a predicate of location, and a predicate of motion. These can only be

Lexical Choices 497

expressed lexically; they cannot be expressed (at least not in English) by bound morphemes or word order or constituent structure. The general principle governing the choice of lexical elements can be called the lexical choice principle. Suppose:

(a) there is a semantic element x' in a construction,

(b) there is a semantic constraint on x',

(c) x, the syntactic element corresonding to x', is a lexical category (e.g., noun, verb, adverb, etc.), and

(d) L is one of the most general individual lexical items available in the lexicon that is in the syntactic category of x and meets the semantic con­straints on x'.

Then:

(e) L expresses x in the construction.

The lexical choice principle makes the following predictions:

Prediction 11: 1 is expressed by here or there. Prediction 12: 2 can be expressed by be, go, or come.

•\

Here and there are the most general individual lexical items that are deictic locative adverbs. Be is the most general single lexical item that expresses a predicate of location. And go and come are the most general verbs that express a predicate of motion.

What is not predicted by the lexical choice principle is that sit, stand, and lie can occur in the construction as element 2. But this particular collection of verbs is not an arbitrary collection. They happen to be the basic-level verbs of location. This construction allows verbs of location to be one level of generality lower than verbs of motion. The basic-level verbs of motion are run, walk, etc. These, of course, cannot occur. Their absence is to be expected from the lexical choice principle, but the fact that sit, stand, and lie do occur is not predicted. Instead it must be listed among the unpredictable parts of the construction that 2 may be ex­pressed by basic-level locational verbs.

It may seem strange to have a condition mentioning basic-level verbs as part of a construction. However, the ability to refer to cognitive concepts like basic-level categorization may make sense of some otherwise inexpli­cable facts. Moreover, basic-level categorization must be appealed to in the description of grammatical facts in languages other than English. For example, in Swedish, one cannot say the equivalent of The lamp is in the comer or The rug is on the floor. One must say instead the Swedish equiva­lents of The lamp stands in the corner and The rug lies on the floor. Swedish appears to require basic-level verbs in simple locative clauses. The impor-

498 Case Study 3

tant thing in both the Swedish and English constructions is that one doesn't, have to give a list of apparently unrelated items. What unites them is their level of categorization and that can be specified in a gram­matical construction.

Before we leave the realm of the lexicon, we should make one more prediction that is related, at least tangentially, to lexical items.

Prediction 13: 1 bears stress.

Lexical items-in major'grammatical categories, including adverbs, normally bear some stress. Lack of stress occurs in a relatively small range of such cases—anaphoric pronouns and backgrounded elements that do not take part in any minimal contrast. As we will see below, the existential there is a backgrounded element that does not minimally contrast with anything else. It is unstressed for that reason. But the deictic here and there are not backgrounded and they do enter into a mini­mal contrast—with each other. Thus, there is no reason whatever for them to lack stress.

The Final Phrase

The final phrase—syntactic element 4—expresses semantic element 4', which is a predicate. Predicates arc typically expressed syntactically as verb phrases. For example, in a sentence like Harry is carrying a huge herring, the verb phrase is carrying a huge herring is predicated of Harry. The final phrase'in the central deictic construction is, however, different from normal verb phrases in an important respect. Take the sentence

- There's Harry, carrying a huge herring.

The final phrase is only carrying a huge herring; there is no is, that is, no overt manifestation of the verb be, which is the head verb of the verb phrase is carrying a huge herring. This is typical of final phrases in this construction. They are phrases which, in simple sentences, normally occur in verb phrases where be is the main verb. We can see this in the following examples:

- There's Sally in front of her house.

- Here comes Max with his new girlfriend.

- There's Sammy at work.

- There goes the cop, afraid of his shadow.

- Here's the victim, shot in the back.

- There's Mary, about to go into the bookstore.

The final phrases in such examples seem very diverse. What all the ex-

With Meaning "Have" 499

pressions. in italics have in common is that they are all remnahts of verb phrases whose head verb would be be in a corresponding simple sentence. This suggests the followng tentative hypothesis about final phrases:

There is a construction in English that is arrived at in the following way:

take a verb phrase whose head verb is be, and subtract the be. Let us call this: YP/be. The final phrase of the deictic f/iere-constructions is VPIbe. • . • •

Thus far we have hypothesized that final phrases are verb phrases with a missing be. Let us now consider a further bit of evidence supporting this position.

With Meaning "Have"

All of the cases we have mentioned happen to have corresponding sim­ple sentences where the be is actually expressed:

- Sally is in front of her house.

- Max is with his new girlfriend.

- Sammy is at work.

- The cop is afraid of his shadow.

- The victim was shot in the back.

- Mary is about to go into the bookstore.

These are some of the many situations where be is required to be the head of the verb phrase in a simple sentence: with prepositional phrases in the first three examples, with adjective phrases (e.g., afraid of his shadow), with passive phrases'(e.g., shot in the back), as part of the phrase be about to, and with progressive participial phrases (e.g., carrying a huge herring). Thus, where be occurs as the head of a verb phrase, the remainder of the verb phrase can occur as a final phrase in a central deictic there-constractwn.

We can find further support for this theory of final phrases in the behav­ior of a special class of with-phrases that have the meaning of have and appear to occur with an understood but unexpressed be. An example of such a phrase is with her shoes off in:

- There's Tammy with her shoes off.

At first, such expressions might appear anomalous because they cannot occur with an overtly expressed be.

- '"Tammy is with her shoes off. The corresponding full sentence has have instead of be with.

500 Case Study 3

- Tammy has her shoes off.

If we look at a variety of contexts where there is an understood but unex-pressed be, we find that this with can occur there. We will consider two such cases: reduced relative clauses and the complement of the \erbfind. Full relative clauses that have a be following the relative pronoun, such as

- Anyone who is carrying a herring .. . usually have variants without the relative pronoun and be:

- Anyone carrying a herring . ,.

Here carrying a herring is a reduced relative clause, and the be that usually accompanies the progressive participle carrying is understood but not overtly expressed. Other constructions that usually occur with be, such as passives and about to, can occur here.

- Anyone shot in the back .. .

- Anyone about to leave ...

Special with can occur in reduced relative clauses as well.

- Anyone with his shoes off...

Similarly, constructions that usually take be can occur in the complement oifind without the be, as in:

- I found Harry carrying a herring.

- I found the victim shot in the back.

- I found Sally about to leave.

Special with can also occur here:

- I found Tammy with her shoes off.

The generalization governing the occurrence of special with seems to be that it expresses certain senses of have and that it occurs where one would expect to find a VP/be. Thus, the fact that special with can occur in the final phrases of central deictic (/lere-constructions lends further support to our analysis of a final phrase as a YP/be.

There is, however, a slight wrinkle here. Not quite all verb phrases with be can occur as final phrases. There are two types that cannot: predi­cate nominals and adjectives that indicate what has been called an "inher­ent" property.

- Harry is a lawyer, (predicate nominal)

- Harry is tall, (inherent adjective)

The Last Word on Final Phrases 501

- ""There goes Harry a lawyer.

- ""There goes Harry tall.

Noninherent adjectives, like drunk, specifying properties that vary with the occasion, can occur as final phrases:

- There goes Harry drunk! This contrasts with

- ""There goes Harry a drunk!

(Incidentally, we are discussing only sentences without comma intona­tion; the sentence There goes Harry, a drunk is, of course, well-formed but has a structure that is irrelevant here.) Noun phrases tend to express persisting properties. Thus a drunk indicates persistent drunkenness, while drunk does not. This explains why noun phrases tend to be ruled out as final phrases. However, some can occur, but only when they are clearly nonpersistent and noninherent.

- There goes Citation the winner by a nose! This condition can be described in our format as follows:

Syntactic Elements 4: a VPibe

Semantic Elements

4': a noninherent, nonpersistent predicate

The Last Word on Final Phrases

The description given above won't quite handle all final phrases. It misses cases where there is motion indicated and where the final phrase is direc­tional, for example,

- There goes Harry into the bar.

- Here comes the bus into the terminal.

Since into the bar and into the terminal are directional phrases, they can­not occur with be, but rather require a verb of motion, like go or come. Since each of the sentences has an occurrence of a motion verb in it, it might seem reasonable to assume that the directional phrase modifies that motion verb—for example, that into the bar modifies goes and into the terminal modifies comes in the above sentences.

But this solution cannot be made to work in all cases. The reason is that the final phrases are part of a separate proposition from the rest of the

502 Case Study 3

construction. We can see this with the help of some added adverbs that have the function of separating out the component propositions. Let us begin with a nondirectional final phrase and then show the same for the

directionals.

- There is Derek, running around the track.

Here running around the track is the final phrase. We can show that there are two propositions here by adding the adverbs again and still.

- There is Derek again, still running around the track.

Still presupposes that what it modifies has been going on continuously. Again presupposes that what it modifies has happened before, but that there has been some discontinuity. If they were modifying the same proposition, there would be a contradiction. What we have is two propo­sitions equivalent to the separate sentences

- There is Derek again.

- Derek is still running around the track.

The same effect can be shown with directionals. Compare

- There goes Harry into the bar.

with

- There goes Harry again, into the bar this time.

Again indicates that Harry went by before, and this time indicates that he did not go into the bar before. There are two separate propositions, equivalent to the sentences:

- There goes Harry again.

- Harry is going into the bar this time.

The verb goes designates a number of occurrences of going, most of which did not include going into the bar. Thus, we cannot simply have into the bar modifying the goes of There goes Harry again,

We can describe such cases by making a minimal modification of the description of final phrases given in the preceding section. There we de­scribed the final phrase as a verb phrase missing a be. In this case the miss­ing head verb would be either go or come, depending on which was chosen as syntactic element 2, the main verb of the whole construction. Since 2 = be, go, or come, the modified lexical condition in our descriptive format would now read as follows:

Syntactic Elements

4: a VP/2

The Order of Verb and Noun Phrase 503

There is one further complication concerning final phrases: More than one of them can occur in a sentence. In fact, there can be an indefinitely large number.

- There's Sally at work on her new book, writing as quickly as she can, afraid that she might miss her deadline.

We can describe this in a straightforward way by indicating that syntactic element 4 can be iterated indefinitely. A standard way of notating itera­tion is to place an asterisk after the numeral marking the syntactic ele­ment, that is, changing "4" to "4*."

The last detail that needs to be mentioned is that the noun phrase of the construction—syntactic element 3—functions as the grammatical subject of the final phrase. This is predicted from the semantics. The final phrase is semantically a predicate, and it is predicated of 3', the semantic entity corresponding to the syntactic noun phrase. Since 3' is the only required argument of this predicate, it is the only possible candidate to be the grammatical subject of the final phrase. Thus, in a sentence like

- There's Derek again, still running around the track.

Derek is the grammatical subject of (is) still running around the track. In the notation given, 3 is grammatical subject of 4. It will thus govern such phenomena as reflexivization in final phrases, for example. There's Derek again, still admiring himself.

Technically, the grammatical subject relation is usually defined so that it holds between a noun phrase and a clause, not a verb phrase. To be technically correct, we need to pick out the clause that 3 is subject of. The entire construction constitutes a clause, represented by syntactic element 0. Let us define a subconstruction (0; 3,4), which consists of the clause, 0, together with the noun phrase, 3, and the verb phrase, 4. We will now say that 3 is the subject of (0; 3,4), or, in other words, that "Derek" in the above example is the subject of "Derek (is) still running around the track."

The Order of Verb and Noun Phrase

Thus far, we have not considered the relative order of syntactic elements 2 (the verb) and 3 (the noun phrase). The order is not fixed, and varies depending on whether or not the noun phrase is a pronoun. In general, the noun phrase precedes the verb if it is a pronoun and follows the verb if it is not.

- There he goes.

- *There goes he.

- There goes Harry.

504 Case Study 3

These facts seem to be constant across all speakers. There is variation when the noun phrase is not a pronoun and precedes the verb. Some speakers find proper names well-formed in this position. (The "%" indi­cates variation.)

- %There Harry goes.

But in general, the longer the noun phrase, the worse this construction seems to be.

- ""There the mayor goes.

- -There the tall man goes.

- *There the man with the purple hair goes.

- *There the man who I met yesterday goes.

This ordering of verb and noun phrase seems not to be confined to the central deictic construction. It also occurs with directional deictic ad­verbial constructions:

- Away ran Harry.

- Out jumped the cat.

- Up popped a mole.

- Into the room came a priest.

The same variation in the order of verb and noun phrase occurs with this construction:

- Away he ran.

- *Away ran he.

- Away ran Harry,

- %Away Harry ran.

Again, for those who find Away Harry ran grammatical, the sentences get worse as the noun phrases get longer:

- *Away the major ran.

- *Away the tall man ran.

- *Away the man with the purple hair ran.

- *Away the man who I met yesterday ran.

Incidentally, the pronoun constraint seems to work only for definite pronouns, like he, she, etc. The indefinite pronoun one works like other noun phrases:

- There goes one!

- %There one goes!

The Central Deictic Construction 505

- I was waiting for a mole and up popped one!

- %I was waiting for a mole and up one popped!

This construction functions deictically for the following reason: it is used in what are sometimes called "vivid narratives," where a story that is supposed to be visualized is being told. Thus the construction functions as though the participants were in the presence of the speaker and hearer.

Consequently, there seems to be a word-order constraint that extends beyond the central deictic (Acre-construction to spatial deictic adverbial constructions in general:

In spatial deictic adverbial constructions, the noun phrase precedes the verb if it is a definite pronoun, and the verb precedes the noun phrase otherwise.

Given this constraint, the relative order of the verb and noun phrase in the central deictic f/iere-construction is predicted.

One aspect of the central deictic construction that does not seem predictable is the choice of locational verbs in case there is no motion. In this construction, the basic-level locational verbs sit, stand, and lie may optionally appear as element 2.

- There sits Harry, chomping on a cigar.

- There he stands, ready to defend our country.

- Here lies a great American.

This extension from the superordinate locational verb be to the basic-level locationals seems to occur only in the central construction.

The Central Deictic Construction

We are finally in a position to represent the central deictic construction in maximally general terms. The following is a minimal representation, leaving out all predictable details.

The Central Deictic Based on: The Central Clause

Parameters of Meaning The pointing-out ICM

Parameters of Form

Semantic Elements 4: not inherent

506 Case Study 3

Syntactic Elements

4*: aVP/2

Lexical Elements . If there is no 2", 2 may be expressed by a basic-level locational verb.

(These are sit, stand, lie.)

This is all that needs to be said about the construction. Everything else is

predictable.

Although most of the syntax of the central construction is a conse­quence of its semantics, it is important to list all of the parameters of form even the predictable ones, since many of them will be different in the n'oncentral deictic constructions. Here is the full representation, with the pointing-out ICM listed as a single unit.

The Central Deictic Based on: The Central Clause

Parameters of Meaning The pointing-out ICM

Parameters of Form

Syntactic Elements

0: clause, a whole with parts:

1: DEICTIC LOCATIVE ADVERB

2: VERB 3: NOUN PHRASE

4*: a verb phrase minus 2 Lexical elements

1; here or there

2: either be, or a superordinate motion verb if there is a 2 , or a

basic-level locational verb if there is no 2" Syntactic Conditions

S-1: 1 is first

S-2: 2 is in the simple present tense

S-3: 2 precedes 3 unless 2 is a definite pronoun

S-4: 4 is optional

S-5: 4 is last

S-6: 3 is subject of 0

S-7: 3 is subject of <0; 3,4)

S-8: 3 is not a sentential complement Phonological Condition

P-l: 1 bears stress

One Construction or Many? 507 One Construction or Many?

The bulk of this study will be concerned with the distinction between cen­tral and noncentral constructions. It is not an uncontroversial distinction, and since the validity of the entire study depends on the distinction, it is important to discuss the question of how one can tell whether one is deal­ing with one construction or many.

First, let us distinguish between a special case and a minor variation. Given a proposed analysis of a central construction, a subconstruction that meets all of the conditions of the analysis is a special case. A subcon­struction that comes close to meeting all the conditions of the analysis, but doesn't quite make it, is a minor variation. The noncentral cases we will be discussing are all minor variations.

Incorrect Predictions

Given a proposed central construction and one class of minor variations, can one always revise the analysis of the central constructions to be more general, so that the minor variations can be considered special cases of the revised central construction? In other words, can one take what is in common between the old central and the old noncentral constructions and come up with a new, more general central construction that will work just as well?

The answer in general is no. Such a strategy can lead to incorrect pre­dictions concerning form-meaning pairings. Such a reanalysis is possible only when the new, generalized central construction predicts no incorrect form-meaning pairings. In each of the cases we will be discussing below, we will argue that any attempt to generalize the central construction to include the noncentral construction will result in such incorrect predictions.

Increased Complexity Elsewhere

It may also be the case that even if such an analysis is possible, it may not be desirable. For example, consider an analysis with one central construc­tion and ten noncentral constructions. Suppose that there is a possible reanalysis of the central construction that would eliminate the need for one noncentral construction, but that this would considerably increase the complexity of the analyses of the other nine noncentral constructions. In such a case, one might argue that such a generalization in the central construction is too costly. But to make such an argument one would have to be able to weigh generalizations in central constructions against gener­alizations in noncentral constructions. As yet, no such cases have been found, and so the question has not arisen for any real cases.

508 Case Study 3

Lack of Full Predictability

Another important distinction that needs to be made is that between a special case and a fully predictable variation. In all of the noncentral con­structions that we will discuss below, most of their properties are predict­able from their relation to the central construction and from other more general principles of both universal grammar and the grammar of English. However, in no case that we will discuss can every aspect of the noncentral construction be predicted. That is, in each case there is at least some minimal part that we claim that a speaker of the language must

leam.

It should be noted that we cannot prove that any given property is unpredictable. To do that, one would have to know all the correct general principles of both universal grammar and the grammar of English, and be able to show that each property listed is unpredictable. All that I can say is that at present I don't see any way of predicting them. But, although the analyses are not verifiable, they are falsifiable. To argue against any given analysis, all one has to do is show how all properties of a given subcon-struction are predictable from other general principles.

But even if one could do this for all the cases given below, one would still not have succeeded in arguing against the applicability of prototype theory in grammar. If each of the constructions is a variation and not a special case, then all one would have succeeded in doing is arguing that the noncentral cases were predictable from the central cases. To argue against the applicability of prototype theory in general, one must provide an analysis according to which all of what we have called variations turn out to be special cases and which makes no incorrect predictions about form-meaning pairings.

In the following section, we will discuss a case that indicates that any such analysis is improbable, if not impossible.

The Noncentral Constructions

The noncentral constructions are variations on the central construction and will be described as such. To indicate the linkage, each noncentral construction will be marked "based on: the central deictic." This will be taken to mean that each noncentral construction will inherit from the cen­tral construction all parameters of form and meaning except for those that are explicitly contradicted by the parameters listed for the noncentral construction.

But this is too simple a concept of inheritance. Some of the subcon-structions we will be discussing involve metaphoric and metonymic models—models that have a mapping function. For example, the next

The Noncentral Constructions 509

construction we will be looking at is the perceptual deictic, in which refer­ence is made to sounds and other percepts in one's nonvisual perceptual field. The central deictic is adaptable for this purpose because we model perceptual space metaphorically on physical space. This allows us to pick out perceptual "locations," as in

- There's the buzzer!

What is "inherited" by such perceptual deictics is not the semantics of physical space used in the central construction, but a metaphorical map­ping of physical space into perceptual space. One thing that we need to know about the perceptual deictic construction is that it makes use of this metaphorical mapping. Thus, one part of the representation of a subcon-struction will be a listing of the names of any mapping models—met­aphoric or metonymic—that it makes use of.

The Perceptual Deictic -It is commonly believed that the senses of lexical items are independent of the constructions that the lexical items appear in. Thus, it is assumed that if a word has a given sense when used in a special grammatical construc­tion, it will also have that sense outside of that construction. This is true in many cases. For example, take the word there. In the central deictic con­struction, there designates a location in physical space.

- There's Harry.

Similarly, when used outside of this construction, say in a simple sen­tence, there also designates a location in physical space.

- Harry is there.

According to the usual belief, there has this meaning in the English lexi­con, independent of any particular construction. When there happens to be used in a special construction, such as the central deictic, it has its normal lexical meaning. This view can be expressed as follows:

The Lexical Independence Hypothesis: The meanings of words are independent of any grammatical constructions that the words occur in.'

This hypothesis is assumed to be true by most contemporary theories of grammar. The fact that it is false is therefore by no means a trivial fact.

We can see that it is false by comparing the central deictic and percep­tual deictic constructions. Corresponding to the central deictic

- There's Harry. we have the simple sentence

510 Case Study 3

- Harry is there. But given the perceptual deictic

-There's the beep. there is no corresponding simple sentence

- "The beep is there.

The perceptual sense of there, which refers to a perceptual rather than spatial location, only occurs in the perceptual deictic construction. If the lexical independence hypothesis were correct, the perceptual sense of there should occur in such simple sentences, since simple sentences do not have any constraints that would prohibit it from occurring there. The fact that the perceptual sense of there cannot occur in simple sentences must therefore be considered prima facie evidence against the independence

hypothesis.

Incidentally, these facts can also be accounted for in an ad hoc way—by

marking the appropriate sense of there in the lexicon to indicate that it can only occur in this construction. This solution misses the generalization that we stated. It also assumes that the lexical independence hypothesis is

false.

One of the effects of the lexical independence hypothesis has been to

confine the application of metaphor and metonymy to the lexicon and to keep the grammar independent of them. This phenomenon suggests, however, that this cannot be done. Instead, the perceptual deictic con­struction must, as a whole, make use of the relevant metaphors. This con­clusion fits very harmoniously with grammatical construction theory as we have seen it at work so far. Grammatical constructions have complex parameters of meaning that are not just associated with individual lexical items. For instance, the central deictic has the entire pointing-out ICM as its semantics—and this goes well beyond the semantic contributions of the lexical items. The metaphors that relate spatial location to perceptual location must apply to the entire pointing-out ICM, not just to the mean­ing of there, in order to provide correct meanings for the perceptual deic­tic construction. To get the meaning right, metaphors must operate on entire constructional meanings, not just on the meanings of words. Let us turn to some examples that show why this is so.

In the perceptual deictic construction, there refers to a location in non-visual perceptual space at a time that either is present or in the recent past, sufficiently recent so that the image (say the auditory image) still persists- Here refers to a location in nonvisual perceptual space at a time that is in the immediate future. Thus, one can say of a beep alarm that has just gone off, or may or may not have finished beeping:

The Noncentral Constructions 511

- There's the beep.

If the beep is preceded by a clicking noise a few seconds before going off, one can say upon hearing the clicking noise:

- Here comes the beep,

Go and come are used in this construction not to indicate motion, but to indicate activation, for example, activation of a signaling device like an alarm or activation of a pain, as in:

- There goes the throbbing in my head again. There appear to be three conceptual metaphors involved in these cases:

NONVISUAL PERCEPTUAL SPACE IS PHYSICAL SPACE; PERCEPTS ARE ENTITIES. REALIZED IS DISTAL; SOON-TO-BE-REALIZED IS PROXIMAL. ACTIVATION IS MOTION.

Once we know that these metaphors apply, they predict most of the uses of the construction. For example, in the description of the semantics these metaphors predict that:

Г = a location in nonvisual perceptual space 2' = a predicate of perceptual location 2" = a predicate indicating activation 3' = a percept

B-3: 3' is in S's nonvisual perceptual field

Incidentally, the gesture often accompanying this construction is pointing upwards with the index finger, as in

- There's the bell now! (speaker points upward and tilts ear upward)

This gesture can be used independently of the syntactic construction.

As I pointed out above, this construction interacts with metonymy in an interesting way. There seems to be a conceptual metonymy of the form

THE THING PERCEIVED STANDS FOR THE PERCEPT.

Among the percepts are sounds, smells, pains, etc., while the things per­ceived are entities that give rise to the percepts like alarm clocks, chemi­cal factories, injured knees, etc. This metonymy shows up in the synon­ymy of sentences like

- There goes the beep.

- There goes the alarm clock.

when the alarm clock makes a beep. In the second sentence, the alarm

512 Case Study 3

clock isn't going anywhere; it is standing metonymically for the beep it makes when it goes off. Similarly, if you have an old knee injury that is acting up, you can say either of the following:

- There goes the pain in my knee.

- There goes my knee.

The second is, in fact, more idiomatic. Again the knee isn't going any­where; it is becoming active in producing a perceived pain. The knee is standing metonymically for the pain.

But things are not so simple. The metonymy, at least as we have stated ' it so far, does not appear to work when the verb come is used. Thus, when one hears the click a few seconds before the beep of the alarm clock, one can say the first sentence, but not the second.

- Here comes the beep.

- *Here comes the alarm clock.

Similarly, when one senses that one's knee is about to act up, one can say the first sentence but not the second:

- Here comes the pain in my knee.

- *Here comes my knee.

There are two ways to account for this phenomenon. One can try a description that says that when the verb come appears in the construction, the metonymy cannot be used. One can state this in the following

manner:

Lexical Condition

2 = come, only if it is not the case that 3' stands for the percept of 3'.

Such a condition would claim that the phenomenon is not predictable, but must be listed as part of the nonpredictable part of English grammar.

However, such a condition can be made predictable by reanalyzing the metonymy. Instead of the formulation given above, one can revise the metonymy to read as follows:

THE THING PERCEIVED STANDS FOR THE PERCEPT WHILE THE PERCEPTION IS IN PROGRESS.

Adding the qualification "while the perception is in progress" permits the metonymy with there goes, but rules it out with here comes, since the per­ception has not yet started at the time it is appropriate to say here comes. This formulation of the metonymy accounts for the linguistic facts. It would predict the following change in the parameters of meaning:

The Noncentral Constructions 513 3': a percept or something perceived

But such a formulation of the metonymy does not merely account for the linguistic facts. It is also an account of what we normally do when we per­ceive—that is, we take our percepts as being actual things that give rise to our perceptions. Although it is possible for us to question our percep­tions, by separating percepts from things perceived, it is not what we nor­mally do—at least not in the case of vision. Distinguishing between per­cepts and the things that give rise to them is more common with the nonvisual modalities, where one is less likely to be certain what one hears, smells, feels, etc. Perhaps that is why the choice is grammaticized in the case of nonvisual modalities—so that one can indicate the distinc­tion between the percept and the thing perceived, since that is a useful distinction to make.

Such a distinction does not show up everywhere in English. For exam­ple, words like sight and sound are not distinguished in cases like:

- I can't stand the sight of the bell.

- I can't stand the sound of the bell.

But such a grammaticized distinction between the modalities does show up in deictic there-constructions:

- There's the sound of the bell.

- ""There's the sight of the bell.

The central deictic makes reference to entities, not to their visual per­cepts such as the sight of the bell. The noncentral deictic only makes refer­ence to percepts and only to nonvisual percepts. This leaves visual per­cepts out in the cold, so far as f/iere-constructions are concerned.

What a speaker of English has to learn about the perceptual deictic is that it is based on the central deictic and that the metaphoric and metony-mic models concerning perception (which exist independently in the con­ceptual system) are used in the construction. Everything else is predict­able. Therefore, all that needs to be mentioned in a minimal specification of the perceptual deictic is the name of the construction it is based on and the fact that the mapping models used are those that concern the domain of nonvisual perception. This will invoke the use of all of the metaphoric and metonymic models mentioned above, namely,

NONVISUAL PERCEPTUAL SPACE IS PHYSICAL SPACE; PERCEPTS ARE ENTITIES. REALIZED IS DISTAL; SOON-TO-BE-REALIZED IS PROXIMAL. ACTIVATION IS MOTION.

514 Case Study 3

THINGS PERCEIVED STAND FOR THEIR PERCEPTS WHILE THEY ARE BEING

PERCEIVED

What the following minimal representation says is that there is a variant of the central deictic that is concerned with nonvisual perception.

The Perceptual Deictic Based on: The Central Deictic

About: NONVISUAL PERCEPTION

This location in the grammatical system defines both a target domain (nonvisual perception) and a source domain (space), and thus the con­struction will automatically make use of the appropriate metaphorical mapping models in the conceptual system. Thus spatial here and there in the central deictic will be mapped onto the appropriate realized and soon-to-be-realized senses. Spatial come and go will also be mapped onto their activation senses. And this will occur only relative to their uses in this construction and will not apply to their use in the general lexicon. This will allow the appropriate sense of there in There's the beep, while block­ing the occurrence of that sense in *The beep is there.

Such a minimal representation, which mentions only the ecological niche of the construction, will generate a full grammatical construction with the same parameters of form as the central deictic has and the revised parameters of meaning cited above.

Learning

This description accounts automatically for how such a construction can be learned. Once a child has the appropriate conceptual system for comprehending perception—including the metaphoric and metonymic models mentioned above—all that the child has to learn is that there is such a construction based on the central deictic and about nonvisual per­ception. To learn that such a construction exists, all he has to do is hear instances of the construction used in appropriate contexts.

One Construction or Two?

This is a convenient point at which to discuss some of the problems that arise from trying to generalize the central deictic to account for the cases covered by the perceptual deictic. This is something I actually tried to do in a previous version of this analysis, and a description of my failure in this matter may be instructive.

One Construction or Two? 515

My first instinct was to keep the number of senses of lexical items and the number of constructions from proliferating. Ideally, I preferred to have a single deictic sense of the word there in both central clauses and (Aere-constructions, and I preferred to have a single construction to cover both the central and perceptual cases. Initially, I had placed two condi­tions on what I thought would be a reasonable analysis.

Condition 1: The Same Lexical Sense Condition

The there in the simple sentence Harry's there is the same sense of the same lexical item as the there in There's Harry.

Condition 2: The Same Construction Condition

There's Harry is an instance of the same construction as There's the beep.

The analysis of the central construction given above meets condition 1, but conditions 1 and 2 cannot be met simultaneously, while maintaining a maximally general account of the phenomena.

The first problem concerns the nonoccurrence of *The beep is there. Suppose that the construction mentions the lexical items here and there. To meet condition 2, the same lexical item there with the same sense must occur in both There's Harry and There's the beep. Thus, there must be a single sense of there neutral between the spatial and perceptual senses. But now there is a problem with condition 1. This there that is neutral be­tween the spatial and perceptual senses must occur in the simple sentence Harry's there, if the same lexical sense condition is to be met. But if the there in Harry's there is neutral between the spatial and perceptual senses, then there is no natural way to keep it from occurring in *The beep is there.

There are, of course, some unnatural ways that could be tried. One might, by brute force, add a condition that says that this sense of there cannot occur in simple sentences whose subject refers to a nonvisual per­cept. This is a purely ad hoc constraint, and it would not only have to be placed on there, but also on here. Such ad hoc constraints are not in the spirit of conditions 1 and 2, which seek maximum generality.

Incidentally, the same problem arises even if we do not assume that the construction mentions the lexical items here and there. The construction might simply mention a deictic locative adverb, and the lexicon might in­clude there and here with both spatial and perceptual senses. But this would give rise to the same problem. To block *The beep is there, one would have to have the same ad hoc constraints keeping the perceptual

516 Case Study 3

sense from occurring in simple sentences with a subject referring to a non-visual percept.

But it is not even clear that such a solution could be made to work at all. Having a construction neutral between the spatial and perceptual cases would require that the parameters of meaning be neutral between space and perception. But it does not appear that this is possible, when one looks in detail at what it would involve. The meaning of there would have to be neutral between distal and realized; the meaning of here would have to be neutral between proximal and soon-to-be-realized. And the meanings of come and go would have to be neutral between activation and motion. While the members of each pair are obviously related in some way, there is no present account of semantics that permits such neu­tral meanings—and until such an account is forthcoming one cannot even consider the possibility of such a neutral semantics. Thus, there can be no general construction covering both the spatial and perceptual cases if it

has any realistic semantics at all.

There is still one more possibility to be considered, namely, that the general construction covering both the spatial and perceptual cases does not mention any parameters of meaning at all. There are three things wrong with such a solution.

First, the construction would not account for the pairing of form and meaning. Since that job has to be done by maximally general principles somewhere in a grammar, the same problems raised above would arise elsewhere.

Second, the parameters of form would all have to be listed as arbitrary, instead of being mostly predictable from the parameters of meaning. Moreover, the syntactic constraints that are predictable from the pragma­tics under the analysis given above would need to be accounted for in some other way.

Third, such a solution would be unable to account in any nonarbitrary way for the contrasts:

- There goes my knee.

- *Here comes my knee.

- There's the sound of the bell.

- "There's the sight of the bell.

- Harry is there.

- *The beep is there.

Arguments like these can be constructed for the noncentral constructions given below. I will not go into each case in as much detail but will present the data on which such arguments would be based.

The Discourse Deictic 517

We should take this opportunity to give a clear example of the distinc­tion mentioned above between a special case of a construction and a pre­dictable variant of a construction. As we saw above, the only thing about the perceptual deictic that we could not predict is that such a construc­tion—a variant of the central deictic concerned with perception—exists. Suppose we could even predict that. What we would have would be a completely predictable variant of the central deictic. We would not have a special case of the central deictic. The reason is that no neutral semantics exists for the domains of space and perception. In order for the percep­tual deictic cases to be special cases of the central deictic, the central deic­tic would have to have a semantics neutral between space and perception. Since that is impossible, the perceptual deictic construction cannot be a special case of the central deictic. It must be a variant—in this case, an almost but not quite predictable variant.

The Discourse Deictic

Here and there can be used to refer to something in a discourse. Let us first take up the case where the speaker is commenting on something that someone else has said. In that case, there is used to refer to something said in the immediate past. If the speaker thinks he can predict what is about to be said, he can refer to it with here.

- Now there's a good point, (past)

- Here comes the best part. (future)

The verb go, however, cannot be used in reference to a discourse.

- ""There goes a nice point. The most basic metaphor involved here is:

DISCOURSE SPACE IS PHYSICAL SPACE; DISCOURSE ELEMENTS ARE ENTITIES.

This metaphor is the basis of expressions like

- I'm lost.

- Where are we?

- Can we go back to your last point?

and many others. There also seems to be a metaphor such as:

IMMEDIATELY PAST DISCOURSE IS IN OUR PRESENCE AT A DISTANCE FROM US. DISCOURSE IN THE IMMEDIATE FUTURE IS MOVING TOWARD US.

This would allow us to account for the fact that there in There's a nice

point refers to discourse in the immediate past, and for the fact that here

f

518 Case Study 3

comes indicates something that has not yet been said but is anticipated. It would also account for the absence of the verb go, since past discourse is not conceptualized as moving. Here is what such a metaphorical account would predict:

Г = a location in discourse space

Т = a predicate of discourse location

2" = a predicate of discourse motion toward the speaker

3' = a discourse entity

B-3: 3' is in or about to enter 5's perceptual field

2 ^ go

If 1 = there, then discourse entity occurred in the immediate past.

If \ = here, the discourse entity is anticipated in immediate future.

The minimal representation for the discourse deictic need only men­tion that it is based on the central deictic and that the mapping models used are the metaphors given above. The conditions given above will be consequences, and the full representation will be that of the central deic­tic with the above changes made.

The Discourse Deictic

Based on: The Central Deictic

About: discourse

The Existence Deictic

Things that exist exist in locations. To be is to be located. Moreover, we know that something exists if it is in our presence; otherwise, we cannot be sure. These common facts form the basis of a widespread metaphor;

EXISTENCE IS LOCATION HERE; NONEXISTENCE IS LOCATION AWAY.

This metaphor is the basis for many common expressions reporting birth and death:

- There's a baby on the way.

- The baby is here.

- The baby has arrived.

- The doctor delivered the baby.

- He's left us.

- He's gone.

- We've lost him.

- Let us pray for the dear departed.

The Activity Start Deictic 519 The same general metaphor seems to be used for existence in general. Thus, something comes into existence and goes out of existence. The existential senses of come, go, here, and there are used in the deictic there-construction.

- There goes our last hope.

- There goes the possibility of a peaceful settlement.

- Here comes another outburst.

- Here comes the chance of a lifetime!

The metaphor would predict that elements of the central construction would take on the following values:

Г = a location in a conceptual space, divided into two parts so that entities in locations near the speaker exist and those in locations far from the speaker do not exist.

B-l: 3' is either coming into existence or going out of existence B-3: 3' is in 5's conceptual field

Again, the minimal representation only needs to mention that the construction is based on the central deictic and that the mapping model used is the metaphor given above.

The Existence Deictic Based on: The Central Deictic

AbOUt: EXISTENCE

The Activity Start Deictic Consider the sentence

- There he goes, meditating again.

There is no motion indicated—only an activity, which in this case involves a lack of action and motion. In general, one can point out the beginning of an activity by using a variant of the central deictic. The central deictic is made applicable by a metaphor in which activities are conceptualized in terms of motion along a path. The construction does, however, add one thing to the conceptual metaphor: the location designated by the deictic locative adverb is the beginning of the metaphorical activity path. We can represent this as follows:

The Activity Start Deictic

Based on: The Central Deictic About: activity

520 Case Study 3

Semantic Elements

Г: designates start of activity (starting point on activity path)

Outside of this construction, there does not have the meaning of designat­ing the starting point of an activity, and this description of the construc­tion accounts for that. In addition, it also predicts the absence of come in this sense. Thus, compare

- There goes Harry, thinking about linguistics again.

- Here comes Harry, thinking about linguistics again.

Both can be instances of the central deictic, in which go and come desig­nate motion. But only the first sentence can indicate the beginning of an activity with no indication of motion.

Cases like Here we go again! will be discussed below when we get to exasperation constructions.

The Presentational Deictic

The presentational deictic differs from the central deictic in a number of ways. The most striking difference is in the syntax. Syntactic element 2, rather than being just a verb in the simple present tense, may be either a verb or a multiword verbal idiom (e.g., laid to rest), and it may be accom­panied by a full auxiliary phrase. And syntactic elements 4*, rather than just occurring at the end of the construction, may occur in second position immediately following the deictic locative adverb, or in final position or both. Here are some examples with syntactic element 4 in second position and an auxiliary phrase.

- There in the alley had gathered a large crowd of roughnecks.

- There on the stage wearing an outrageous costume was standing one of our most distinguished public figures.

- Here honored by thousands of her fellow citizens will be laid to rest our beloved former mayor. Sally Stanford.

- There without a stitch on, staring at the ceiling, lay the beautiful KGB agent shot through the heart.

- Here isolated from all the noise of the city will be built large, comfort­able estates to house our former presidents.

The auxiliary + verb combinations include had gathered, was standing, will be laid to rest, and will be built. The "final phrases" in second position in­clude in the alley, on the stage wearing an outrageous costume, etc.

The permissible verbs are those that, directly or indirectly, specify a location. Indirect specifications of location include will be laid to rest, will be built, and even can be seen:

The Presentational Deictic 521

- There in the southern sky can be seen the largest comet ever to enter our solar system.

The presentational deictic has two uses. In a narrative, it can be used to indicate a discovery, thus introducing (or reintroducing) the referent of the noun phrase into the discourse:

- There in my favorite chair sat a fat man with a monocle.

It can also be used to point out something the speaker considers ex­tremely significant. This makes it useful in pompous speeches, public relations pamphlets, public announcements, etc.

- If I am elected, I guarantee that here in this town will be built a multi­purpose athletic facility second to none!

Moreover, the there can be omitted, provided the first non-null ele­ment of the construction still designates a location relative to the speaker.

- Behind the desk was sitting a bald-headed man.

- Strewn about the room were lying the victim's personal effects.

- Standing atop that mountain is the largest radiotelescope ever built.

However, if the initial phrase does not designate a location, syntactic ele­ment 1 cannot be null.

- There without a stitch of clothes on lay the victim.

- "Without a stitch of clothes on lay the victim.

- There, enjoying his solitude, can be found the country's most famous hermit.

- *Enjoying his solitude can be found the country's most famous hermit.

- There happy at last could be seen the smiling refugees.

- *Happy at last could be seen the smiling refugees.

The presentational deictic can be represented minimally as follows:

The Presentational Deictic

Based on: The Central Deictic Discourse Conditions

Used in narratives or announcements when S considers 3' significant Lexical Conditions

2 = a verb designating a location, directly or indirectly Syntactic Conditions

2 may have a full auxiliary. Instances of 4* may optionally occur before 2 and 3.

524 Case Study 3

aspect of the delivery scenario and express that aspect. Attention can be directed either to the delivered object and its location or to the hearer and

his location on the activity path.

In the following examples, attention is directed to the location of the

delivered object.

-'Here it comes! This focuses on the movement of the object toward the hearer.

- Here it is!

This focuses on the arrival of the object in the hearer's presence.

-In the following examples, attention is directed to the location of the hearer on the activity path.

-Here you are!

This focuses on the presence of the hearer at the beginning of the activity path.

- Here you go!

This focuses on the start of the activity, that is, on the beginning of motion by the hearer along the activity path. It follows that only come, and not go, can refer to the motion of the delivered entity to the speaker and that only go, and not come, can refer to the "movement" of the hearer along

the activity path.

Our knowledge about deliveries can be represented as a cognitive

model. Here is a representation of such a model, restricted to the issues at hand.

The Delivery ICM Metaphorical Mapping: activity is motion along an activity path.

Semantic Elements D: deliverer R: receiver E: entity

LI: location of £ LI: location of R Ml: predicate indicating motion through space

A: activity path Ml: predicate indicating motion along A

Scenario Background

The Delivery Deictic 525 R is waiting for D to bring E to L2 R requires E in order to move along A (perform the activity)

Activity D brings E to L2

Resultant State

E is at L2

R is at beginning of A

As in the case of the pointing-out ICM, the delivery ICM constitutes an experiential gestalt, which is psychologically simple, but whose analysis looks complex.

The delivery deictic can be used in situations that are structured by the delivery ICM. In such cases, the speaker is the deliverer (S = D). The construction is marked phonologically—either the deictic locative adverb is unstressed or its vowel is elongated.

We can represent this as follows;

The Delivery Deictic

Instance of: The Deictic Category Overall Semantics

The Delivery ICM, with S = D. Phonological Conditions

1 may be elongated with rising intonation and need not bear stress. What is particularly interesting about the delivery deictic is that it can be

an instance of more than one of the deictic constructions discussed above. Thus

- H e e e e re comes your pizza!

is an instance of the central deictic applied to the delivery situation, while

- H e e e re you go!

is an instance of the activity start deictic applied to the delivery situation. The reason this is possible is that the delivery ICM includes both motion of an object to a destination and the start of an activity.

We have represented this situation simply by indicating that the deliv­ery ICM is an instance of the deictic category. This means that it can be an instance of any construction in the category that it can fit. It happens to be able to fit both the central deictic and the activity start constructions. It fits them in two different ways. That is, each construction takes a different perspective on the delivery situation. In the central deictic perspective, attention is focused on the entity being delivered, which is the thing being

522 Case Study 3

1 may be null if an initial instance of4* designates a location, directly

or indirectly.

Since the verb must pick out a location, rather than indicate motion, we can account for why come and go cannot occur with a full auxiliary:

- There goes the president.

- *There is going the president.

- *Here will come the president.

The Delivery Deictic There is a deictic (Acre-construction that is used to indicate delivery, as in:

- Here's your pizza, piping hot.

- Here's your hot and sour soup, with no MSG.

- There's your car, all washed and waxed.

One might think that this is just a special case of the central deictic con­struction, where the speaker happens to be delivering something to the hearer. For the most part this is correct. However, there is a phonological peculiarity that is linked to the delivery situation: the vowel in here and there can be elongated and the elongation only occurs with a "delivery" situation. The most famous case is Ed McMahon's introduction of Johnny Carson:

- H e e e e re's Johnny! This occurs in delivery situations in general, for example:

- H e e e e re comes your soup!

- Th e e e e re's your car, all fixed and ready to go!

In delivery situations, the thing being delivered is either being brought to the hearer (as in the first sentence) or has been left in a temporary nearby location to be picked up (as in the second sentence). Consequently, the verb go cannot be used in the delivery deictic construction, as we can see from the incompatibility of go with the elongation of the vowel in there:

- *Th e e e e re goes your car, all washed and waxed!

Such a sentence might be used in a bizarre situation, say, on a TV comedy where the carwash attendant started to inform his customer that his car was ready and in the process noticed a thief driving off with it. But even then, it would be a joke rather than a normal thing to say.

As we observed above, deictic f/iere-constructions can occur in the past tense in certain narrative situations:

The Delivery Deictic 523

- There I was in the middle of the jungle .. .

- There was my car with the key in the ignition .. .

Since the delivery deictic only occurs in the present tense, such sentences cannot have the elongated vowel that signifies delivery:

- *Th e e e e re I was in the middle of the jungle . . .

- *Th e e e e re was my car with the key in the ignition...

In addition, the here and there of the delivery deictic is unique among the deictics in that it need not bear stress and may take a reduced vowel.

Finally, this construction is unique in that it allows the second person pronoun, you, to replace syntactic element 3.

- H e e e e re you are.

- There you are—one pizza with sausage and mushrooms.

However, when this occurs the verb gn is pn^.hip hut rn^a ”• ^

- Here you go!

- There you go!

- *Here you come!

- "There you come!

The choice of go versus come with 3 = you is exactly the opposite of the choice when 3 + you. We can see the reason for this difference of lexical choice if we compare the sentences:

- Here comes your pizza!

- Here you go!

Either could be used by a waiter delivering a pizza to a customer's table. But there are subtle differences. iq the cowe-sentence, come refers to the movement of the pizza toward the hearer. It is typically uttered when the pizza hasn't quite yet reached the hearer. The go-sentence differs in both respects. It is usually uttered as the pizza (or whatever) reaches the hearer. And the go does not necessarily refer to any movement of any sort. Instead the go seems to refer to the hearer's embarking on some activity that could only begin with the delivery. Thus, the come-sentence focuses on the delivery, while the go-sentence focuses on what happens after the delivery.

Part of our understanding of what delivery is typically about is that the recipient is going to do something with what is delivered—if only to store it for later use. Given the activity is motion metaphor, it is natural forgo to be used to refer to what the hearer is about to do with the delivered entity upon delivery. Thus, all of the following sentences pick out some

526 Case Study 3

pointed out. Thus we have the following fit between the delivery ICM and the semantic elements of the central deictic:

The central deictic perspective on the delivery ICM:

S = D

H = R

3' = E

Г = LI

2' = Ml

In the activity start perspective, the receiver (R) is both the hearer (Я) and the entity being pointed out (3'). The activity start perspective:

H = R = 3'

S = D

Г = L2

2' = Ml Jb&Jielivery f^ictic is a case where there is a minor variation in phono-

logical form which is used to signal the presence of a delivery situation. Minor variations in phonological form are often used to signal something about the nature of the situation. Such cases involve form-meaning pair­ings that can be represented by minimal constructions. We will now turn to such a case—one where an expression of awe is indicated phonologi-cally and where this can intersect with deictic constructions.

Paragons

When one thinks that something is very good—among the best of its kind—it is common to direct attention to it and express awe at how good it is. (The capitals followed by three dots indicate extra-heavy stress, op­tionally accompanied by breathiness.)

- Now that's ... a real cup of coffee!

- Now THIS ... is chicken soup the way mama made it!

- THIS ... soup is g о о о d!

This can also be done using a deictic (Aere-construction.

- Now THERE ... is a great centerfielder!

- Now HERE ... is a great cup of coffee!

The semantics that is special to these cases can be represented by what I will refer to as the expression of awe ICM.

The Expression of Awe ICM Semantic Elements

Paragons 527

S: Speaker

H: Hearer

Г: an entity

2': a predicate indicating the best of a type

О': 2'(Г) Speech Acts

S is expressing 5's awe at 0'

The paragon-intonation construction is quite simple and involves no syn­tactic or semantic parameters: it pairs the expression of awe ICM with the appropriate phonology—extra-heavy stress and breathiness on the ap­propriate element. It is independent of, and can be superimposed on, any construction with a consistent pragmatics.

The Paragon-Intonation Construction

Parameters of Meaning

The expression of awe ICM Parameters of Form

1 is pronounced with extra-heavy stress and breathiness.

Such a pairing of form and meaning is so simple and obvious that it seems barely worth mentioning. Yet it produces some complex and not entirely obvious results. These complexities come from the fact that the paragon-intonation construction can form intersections with other constructions, among them simple sentences and deictic (ftere-constructions.

Let us begin with simple cases of the sort discussed above. Suppose we take ordinary simple sentences like

- That's a cup of coffee.

- This is chicken soup.

Superimposing the paragon-intonation construction yields

- that's ... a real cup of coffee!

- THIS ... is chicken soup the way mama made it!

Here "real" and "the way mama made it" indicate that it is a paragon.

It is part of the semantics of the central deictic that there is an entity at the location referred to by here and there. There is a general metonymic mapping, according to which a place may stand for something located at that place. For example, Toledo is late may mean that our sales represen­tative located in Toledo is late. When this metonymic mapping is em­ployed, the location referred to by here or there stands for the entity at that location. This permits the paragon-intonation construction to be

528 Case Study 3

superimposed on the central deictic, and for here or there to refer to the entity to be described as a paragon.

- Now THERE ... is a real cup of coffee!

- Now HERE ... is chicken soup the way mama made it!

One source of complexity is the fact that the paragon-intonation con­struction brings along its own pragmatics. The speech act it conveys is expressive—in the same class of speech acts as exclamations. It can thus potentially intersect with other exclamation constructions. Consider the exclamation construction in

- Boy! Is HE ever tall!

This has the syntactic form of a question, but the phonological form and the illocutionary force of an exclamation. It requires that the heavily stressed element (HE, in this case) be a subject. Thus this construction can intersect with those instances of the paragon-intonation construction where the heavily stressed element is a subject.

- Boy! Was THAT ever a cup of coffee!

This construction cannot intersect with paragon (Acre-sentences, since the stressed element in those cases is not a subject.

- *Boy! Was THERE ever a cup of coffee!

Since the paragon-intonation construction expresses awe at something positively being a paragon, it cannot take true negations.

- *Now THAT .. . wasn't a real cup of coffee!

Here the semantics of the negated construction, that the referent is not a paragon, contradicts the pragmatics, which expresses awe at its being a paragon.

However, it can occur with a negation that expresses a positive propo­sition.

- Now THAT.. . can't be beat!

Here can't be beat expresses a positive value, despite being syntactically negative. Another example is the construction that occurs in

- If JIM won't eat these ducks' feet, I don't know who will!

The effect of this construction is to suggest that Jim probably will eat the ducks' feet. Since this negation conveys a positive not a negative, it can occur with paragon-intonation constructions.

- If THAT .. . isn't a real cup of coffee, I don't know what is!

Exasperation 529

Not only can the paragon-intonation construction intersect with the cen­tral deictic, it can also intersect with other deictics, for example, the dis­course deictic.

- Now THERE . . . was a nice point!

If one happens in the discourse to have been talking about old-time base­ball players, one can say

Now THERE ... was a great centerfielder!

If we are at a baseball game, we can point to a ballplayer and say, using the central deictic plus the paragon-intonation construction,

- THERE . . . goes a great centerfielder!

But if you are talking about old-time ballplayers, you cannot say

- *Now THERE .. . went a great centerfielder!

This would be an intersection of the paragon-intonation and the discourse deictic constructions. But the discourse deictic does not permit go, as we observed above.

- *There went a nice point!

The question raised by this analysis is whether the paragon deictic con­struction needs to be listed in the grammar of English at all. The tradi­tional answer is, of course, no. If all of its properties can be predicted from the general paragon-intonation construction and other deictics, then there is no need to say anything more. However, one might argue, in response, that such complex constructions have some cognitive status as speech formulas—special things one says on certain occasions. We will discuss this issue in general below.

Exasperation

v

A deictic there-construction can be used to express exasperation over someone's behavior. Examples include:

- There goes Harry, sounding off again!

- There she is, making a fool of herself!

Go here does not refer to motion, but to actions on Harry's part. It appears that the activity is motion metaphor is at work here and that the there designates the beginning of an activity. Exasperation may be indi­cated by a sigh, a throat constriction, slight nasalization, or an intonation contour. But exasperation can be indicated by these means, not only with

530 Case Study 3

deictic f/iere-constructions, but also with just about any sentence at all with the appropriate meaning.

- (sigh) He's insulting her again!

Let us, for the sake of discussion, represent such a pairing of form and meaning with a very simple construction, the exasperation construction. Speech act

S expresses exasperation with 0', a proposition. Phonological Condition

The sentence is pronounced with a sigh, a throat constriction, slight

nasalization, or the exasperation intonation. Sentences like

- (sigh) There goes Harry, telling the same old jokes!

can now be seen as intersections of the activity start deictic and the exas­peration construction. This explains why we don't have deictic exaspera­tion expressions with come, except in the central, spatial sense. Thus,

- Here comes Harry, sounding off again!

suggests that Harry is actually moving toward the speaker, while the cor­responding sentence with there goes does not require any motion at all. In the activity is motion metaphor, which is used in the activity start deic­tic, the metaphorical motion is never toward the speaker. This would explain why the here comes in the above example can only have a physical motion reading. The use of this metaphor in the activity start construction would also explain why such cases can have here without spatial motion, but only with a first-person subject.

- Here we go, making fools of ourselves again!

- *Hcre they go, making fools of themselves again!

According to the metaphor, the deictic locative adverb refers to the be­ginning of the activity path. If the speaker is at the beginning of the activ­ity path, here may be used. This is not the case if someone other than the speaker is at the beginning of the activity path. The above distinction is, thus, a consequence of the activity is motion metaphor as used in the activity start deictic and the ordinary constraints on the use of here.

Narrative Focus

In the course of a vivid narrative, a speaker may want to focus on a par­ticipant in the narrative, perhaps himself. This can be accomplished with a deictic (/lere-construction:

Narrative Focus 531

- There I was, alone on a desert island...

- There I am, alone on a desert island ...

- Here I am, alone on a desert island ...

In a vivid narrative, the hearer is expected to imagine the action—that is, to create a mental image of what is happening. Here and there function to focus on a location in the imagined scene. Both the past tense and the present may be used. The difference has to do with cognitive "distanc­ing." The choice of here and the present tense bring the hearer "closer" to the imagined scene of the narrative. They do so in somewhat different ways. The present tense identifies the action of the narrative with the time of the telling. "Here" invites the hearer to imagine the action close to him and the speaker, rather than at a distance.

When the present tense is used and the imagined scene of the narrative is taken to occur simultaneously with the time of the narration, the narra­tive focus deictic merges with central deictic via the pretense that the ele­ments of the narrative exist and are present. The full range of central deictic expressions can be used in such a situation, including verbs of motion:

- Here come the killer bees, blackening the sky . . .

- There go the jets, screaming overhead . ..

This seems to be a matter of general conventions of narration rather than anything peculiar to (Aere-constructions.

Let us call these the vivid narrative conventions, and specify them as follows:

The speaker assumes the hearer is forming a vivid mental image of the narrative. The things in the mental image are taken to be present. If the present tense is used, there is a pretense that the action is occurring at the time of the telling. If the past tense is used, the action is taken as happening prior to the telling.

' When we apply the vivid narrative conventions to the use of the central deictic, we get the trivial consequence:

Semantic Elements

1' = a location in the imagined scene of the narrative 3' = an entity in the world of the narrative

The vivid narrative conventions may apply to many of the deictic con­structions, to yield past tense forms that can be used in vivid narratives.

- There were the helicopters at last, coming to rescue us ...

- There went Harry, out to buy some bread . . .

532 Case Study 3

- There came the bus at last—and just as you might expect, it was full.

- There went the buzzer, and the children ran out the door!

- There went our last hope.

The fact that these conventions require one to imagine a scene vividly (presumably to form a mental image of it) explains the absence of certain forms. For example, things that have not yet happened are not part of a vivid mental image. Since here comes in certain constructions refers to the future, which has not yet happened, those cases are impossible as inter­sections with the narrative focus construction.

- '"Неге came the beep!

- *Here came another outburst!

Jeanne van Oosten has pointed out (personal communication) that some of the deictic constructions simply cannot occur at all as intersections with the vivid narrative conventions. Those ruled out are the delivery deictic and the exasperation deictic.

- *Here you went! (delivery plus past tense vivid narrative)

- ""There came our pizza! (delivery plus past tense vivid narrative)

- "There he went again! (Exasperation plus past tense vivid narrative)

I do not know exactly why these intersections are impossible. My best guess is that there is some pragmatic constraint on the vivid narrative con­ventions that I have missed and that it is inconsistent with the pragmatics of these constructions. However, at present, these phenomena remain a mystery.

Enthusiastic Beginning

A continuous intonational rise with a fall at the very end can often be used to mark the enthusiastic beginning of some enterprise.

- Let's get going!

- Let's hit the road!

- Off we go!

- They're off!

Let us call this pairing of intonation with an expression of enthusiasm at the beginning of an enterprise the enthusiastic beginning construction. This can form an intersection with the activity start deictic:

- Here we go, off to Africa!

This seems to be nothing more than a predictable intersection of the two constructions.

The Issue of Cognitive Status 533 A description of the idiom Here goes nothin' can be given as follows:

Intersection of: Activity Start Deictic and Enthusiastic Beginning S = person making attempt 3 = nothin' 3' = attempt with nothing to lose

The rest of the syntax, semantics, and phonology of the idiom is inherited from the enthusiastic beginning deictic. This idiom is therefore not com­pletely arbitrary. The arbitrary parts are listed above in the minimal rep­resentation. The rest of the idiom is motivated by the constructions it is an instance of.

The Issue of Cognitive Status

The grammar of a language is a cognitive subsystem. Anything in the grammar that has to be learned by a native speaker must exist as part of that cognitive subsystem, and thus has some cognitive status. For exam­ple, anyone learning English must learn how to form passives (like Harry was hit by someone) and how to form questions (like Did someone hit Harry?). Presumably they must learn passive and question constructions. But having learned those, they can superimpose them to form passive questions (like Was Harry hit by someone?). They presumably do not have to leam a passive-question construction. They just intersect the pas­sive construction and the question construction. The result, the passive-question construction, is made up anew each time and has no permanent cognitive status. That is, it does not exist as a unit in the conceptual sub­system we were calling a grammar.

According to the assumptions of traditional grammar, nothing that is predictable has any permanent cognitive status. The assumption was that if it can be predicted, then it is computed anew each time it is used. Though this seems plausible in the case of passive questions, it is not true in all cases. Perhaps the best-known example of this sort is the negative question. In terms of syntactic conditions, negative questions are just intersections of the question construction and thesentential negative con­struction. But negative questions have a special pragmatics—they are not questions of negatives, but rather hedged positive suggestions. For exam­ple, Didn't Harry leave? is a hedged suggestion that Harry did leave. Thus, negative questions must have a cognitive status of their own. The construction would be represented as an intersection of the negative and question constructions, with a special pragmatics indicated.

Some of the complex cases we have been discussing may be like this, though I do not at present have evidence to demonstrate it. Paragon deic-

534 Case Study 3

tics (like THERE.. .is a cup of coffee!) and exasperation deictics (like There he goes again! are intersections of other constructions, yet they seem to have the status of speech formulas—things that one learns to say on certain occasions. When one learns a language, one does not only learn those constructions from which you can form other, more complex. constructions. One also learns which constructions to use when. Even if such constructions are intersections of other constructions, they still have cognitive status because one must learn something about when to use them. For example, suppose one could specify conditions under which the exasperation deictic is used as a formula. All you would have to learn is that there is such a formula with such conditions, and the details would

follow as a consequence.

What Do the Deictics Have in Common?

As we have just seen, the deictic f/iere-constructions can be analyzed as forming a natural category with the central deictic as the prototype. It will be instructive to go back over the data we have considered to see exactly why the deictics cannot be analyzed as forming a classical category. If the deictics formed a classical category, then the properties they have in com­mon would characterize all and only the members of the category. One way of seeing just how far away this is from being true is to survey what

the various deictic there-constructions all share. The Generalized Deictic Parameters of Meaning Semantic Elements

S: speaker

H hearer

Г an argument

2' a predicate

3' an argument

4' a noninherent predicate

0' a proposition, 2'(3',Г) 0'": a proposition, 4'(3')

Speech Act

S is directing ffs attention to Г

Parameters of Form Syntactic Elements

What Do the Deictics Have in Common? 535

1: a deictic locative adverb

2: a verb

3: a noun phrase

4*: VP/be or VP/2 Lexical Conditions

1 = here, there Syntactic Conditions

1 is first

2 precedes 3 unless 3 is a pronoun

4 is last

4* is optional

These shared properties can be viewed as necessary conditions for a construction to be a deictic (Acre-construction. But according to the class­ical theory, the shared properties should be both necessary and sufficient to characterize all the constructions. But these conditions are so vague that they are not sufficient to characterize any of the constructions. To get the individual constructions one would have to add to these necessary conditions various combinations of syntactic, semantics, pragmatic, lexi­cal, and phonological properties. There is no way to preserve the classical theory and still get the correct pairings of form and meaning. Moreover, even if one left meaning out entirely, there would be no way to get suf­ficient conditions on the parameters of form. The presentationals take full auxiliaries, element 4 optionally in second position, and an optional deic­tic adverb. The others do not. Of the remaining constructions, two take the past tense. Lexically, one doesn't allow verbs of motion, another doesn't allow come,'and one doesn't take be. The phonological condi­tions introduce even more complications. Any attempt to state shared sufficient conditions, even on parameters of form alone, simply will not

work.

No matter what one does, one is going to wind up with a radially struc­tured category of constructions. One possible analysis would be to have the abstract, generalized construction in the center and the actual con-' structions structured around the center and specifying it further. As I see it, there are two things wrong with such an analysis. First, it would fail to make correct predictions about prototype effects. Since none of the ac­tual constructions are central on this analysis, the analysis fails to predict what appears to be a crucial fact, that the central deictic seems to be a bet­ter example of the category than the other cases. Intuitively,

- There's Harry.

536 Case Study 3

seems to be a better example of a deictic construction than any of the fol­lowing:

- there's ... a real cup of coffee!

- There he was in the jungle . . .

- He ее re's Johnny!

- There will stand the new stadium for the 1996 Olympics.

This is predicted on our analysis, but not on the analysis where the gener­alized deictic is central.

Second, such an analysis would preclude stating the generalizations

that we pointed out above. The predictions that can be made on the basis of the central deictic cannot be made on the basis of the generalized deic­tic. If all of the actual constructions are based on the generalized deictic, then all one can do for each construction is give a big list of whatever needs to be added. This is an especially serious flaw in the area of seman­tics. Under our proposed analysis, it is possible to make use of meta­phoric and metonymic models that exist independent of these construc­tions. By naming which metaphoric and metonymic models are used, we can predict most of the semantics of the noncentral constructions from the semantics of the central deictic. This is possible because the meta­phoric models all take physical space as their source domains. However, if the generalized deictic is taken as the center of the natural category and if all the actual constructions are just extensions of it, then the metaphoric models that exist independently in the conceptual system cannot be used to predict semantic properties. One would be reduced instead to giving a list of the semantic properties in each case and would thereby miss the regularities that are possible with the analysis that makes use of the meta­phoric models. .

To summarize: Despite the fact that there are necessary conditions for

membership in the category of deictic (/Hire-constructions, there are no sufficient conditions. This is why the classical theory will not work for such constructions. In addition, any attempt to analyze the generalized construction as central will fail for two reasons:. First, it cannot predict prototype effects; in particular, it cannot predict that examples of the cen­tral deictic are better examples of the category than examples of, say, the presentational, paragon, or delivery subconstructions. Second, it does not permit the statement of generalizations, especially semantic general­izations.

Given that there are necessary conditions which can be stated in a gen­eralized construction, one ought to ask whether such a generalized con­struction has any psychological reality or plays any role in the grammar of a language. At this point, there is no reason to believe that it plays any

Cognitive Organization and Cognitive Efficiency 537

role whatever. It simply seems to be too general to be of any use, and in the absence of any reason to hypothesize it, I would tentatively suggest that it has no psychological reality whatever.

This is not a trivial matter. The generalized deictic represents the result of abstracting out the common properties of the various kinds of deictic f/iere-constructions. To claim that it plays no cognitive role is to claim that seeking common properties—as linguists are trained to do—does not al­ways lead to the best analysis, even when one can find such common prop­erties.

Cognitive Organization and Cognitive Efficiency

The deictic (Acre-constructions present a bewildering range of phenom­ena. We have found that we could make sense of these phenomena using the concept grammatical construction, together with prototype theory. Accordingly, if the deictic t/iere-constructions are viewed as constituting a radially structured category of grammatical constructions, a wide range of phenomena can be seen to follow from general principles. The central deictic construction's parameters of form can be seen almost entirely to be consequences of the pragmatic function of the construction, namely, to express the pointing-out ICM. Given this, the noncentral constructions can be seen as minimal variants of the central construction. This account seems to explain a wider range of data than any other accountavailable.

The key to this mode. of explanation is the concept of motivation. We have characterized motivation in terms of (1) central principles of form-meaning correspondence; (2) ecological location within a grammatical system, defined by the based-on relation; (3) minimal differences; and (4) restricted prediction, using central principles, metaphors and metony­mies already in the conceptual system, and minimal differences. If we as­sume that languages are structured to maximize motivation, we would ex­pect radially structured categories to be prevalent, since such structures have the effect of maximizing motivation. Thus, radial structures in gram­mar have the same function that they do in the lexicon—that of reducing 'the arbitrariness of form-meaning correspondences.

Motivation is, therefore, concerned both with parameters of form and with parameters of meaning. In each language, parameters of meaning make use of concepts from a conceptual system which, in certain respects, is not universal. Since parameters of form depend on parameters of meaning, it follows that some grammars may be better-suited to certain cultures and conceptual systems than to others. For example, the English perceptual deictic construction, with expressions like There goes my knee, is very well suited to a conceptual system that has metaphoric and meto-

538 Case Study 3 nymic models like activation is motion, realized is distal, and the

THING PERCEIVED STANDS FOR THE PERCEPT. Such a construction WOuld not

fit at all well into a language based on a conceptual system without a realized is distal metaphor. Motivation is therefore a global property of both a conceptual system and a grammar that makes use of it.

Motivation may best be thought of as "systemic redundancy"— redundancy as denned by the overall structure of the grammar and con­ceptual system taken together. We have made a specific proposal for how to characterize such redundancy in terms of the analytic techniques at our disposal. In the theory we have outlined, it is possible for us to make cer­tain "restricted predictions." It is our claim that there is a correspondence between each such "prediction" that can be made using this theory and some cognitively real redundancy within the grammatical system of a

speaker.

We have not been claiming that speakers of a language go about mak­ing such predictions as part of everyday language use. Speakers do not al-gorithmically compute most of the properties of the constructions they use every time they speak. Quite the opposite. Speakers use construc­tions that are present in the grammatical system—present with a high de­gree of systemic redundancy.

We assume that constructions are used as wholes, as entire gestalts.

But in order to make sense of viewing grammatical constructions as ge­stalts, one must have some account of gestalt formation for grammatical constructions, that is, an account of what it is that allows a construction to function as a gestalt rather than just as a random collection of properties. The account we are proposing is the following:

-The more a construction's properties are motivated, the better it functions as a gestalt.

There is a great deal that follows from this principle. Good gestalts are cognitively simple, easy to learn, easy to remember, and easy to use. They also have the following very important property:

- If A is a good gestalt, and В is a minimal variation of A, then В is al­most as good a gestalt as A.

In other words, there is a very good reason why cognitive structure is or­ganized in terms of good gestalts and minimal variations on them: it maxi­mizes cognitive efficiency. Thus, we can see why there should be so many radially structured categories. If a cognitive model is a good gestalt, then minimal variations on it will be easy to learn, remember, and use. When there are many minimal variations on a cognitive model, the result is a ra­dially structured category.

Cognitive Organization and Cognitive Efficiency 539

This provides a critical link between cognitive organization and cogni­tive efficiency. Radial category structure can make it easier for people to leam, remember, and use cognitive models, provided that the prototypi­cal centers of the radial categories are good gestalts. But our proposed concept of a good gestalt is itself dependent on overall cognitive organiza­tion. The reason is that motivated properties are motivated not merely by the properties of an adjacent construction. They are also motivated by general principles governing the entire linguistic and conceptual system. The good-gestalt principles for geometric figures are local: they have to do with things like symmetry, continuity, etc. But the good-gestalt princi­ples for grammatical constructions are global. They have to do with the overall ecology of the conceptual and linguistic systems. All sorts of fac­tors outside the construction itself may enter in: cognitive models, includ­ing those metaphoric and metonymic models in the conceptual system, language-particular word-order constraints, lexical contrasts, rules gov­erning stress and intonation, etc. To be a good gestalt, a construction must fit well into the linguistic and conceptual system as a whole. It is our claim that the more a construction is motivated, the better it will fit into

the ecology of the system.

The concept of motivation therefore provides a cognitively based eval­uation metric for grammars. Suppose we think of a language as a collec­tion of form-meaning pairs, where the meanings are concepts in a given conceptual system. Incidentally, this is not an uncontroversial character­ization of a language, since it takes a language as being defined relative to a conceptual system. We can then define the "expressive power" of a lan­guage as the collection of concepts in that conceptual system that the lan­guage can distinctively express. Incidentally, that collection is infinite, since it includes complex concepts. This is an intuitive way to define "ex­pressive power" since it is based on what is expressed. For a language of a given expressive power, an optimal grammar is one that maximizes moti­vation. And there may be many equally optimal grammars since motiva­tion may be maximized in many ways. An evaluation metric for grammars would rate grammars as to how well motivated they are (that is, how well they fit together overall) relative to the concepts expressed. The worst grammars would be those with no motivation at all, grammars where the form-meaning correlations are maximally arbitrary. Grammars of real natural languages (as opposed to hypothesized "possible" natural lan­guages) always show a very high degree of motivation. People seem to learn and remember highly motivated expressions better than unmoti-vated expressions. We thus hypothesize that the degree of motivation of a grammatical system is a measure of the cognitive efficiency of that system relative to the concepts the system expresses.

540 Case Study 3

Contemporary generative theories do not incorporate a concept of mo­tivation at all, and they do not permit prototype-based categorization. The concept of expressive power used in generative theories has nothing whatever to do with what is expressed. Conceptual systems play no role. Expressive power there is concerned only with the set of strings of unin-terpreted symbols that a formal grammar can generate. Generative theo­ries are thus incapable of even approaching the kinds of issues raised in this case study, just as they are incapable of accounting for the kinds of regularities we found in the study of deictic there-constructions,

We will now turn to the relationship between the central deictic and central existential constructions. This has traditionally been a stumbling block for generative theories. Various attempts have been made to ac­count for the relationship within transformational theories, and they have all been failures. One reason is that the relationship between these con­structional categories is not a purely syntactic relationship. Another rea­son is that generative grammars cannot adequately characterize the over­all ecology of grammatical systems. The relationship between the deictic and existential Леге-constructions in English is very much a matter of that

overall ecology.

Comparison of Deictics and Existentials

The deictics and the existentials can look similar, as the following exam­ples show (capitals indicate stress).

- there's a new Mercedes across the street, (deictic)

- There's a new MERCEDES across the street, (existential)

Despite the superficial similarities, the deictic and existential there's differ in at least the following ways:

- Deictic there refers to a specific location; existential there does not.

- Deictic there contrasts with here; existential there does not.

- Deictic there occurs independently of the deictic (/lere-constructions;

existential there does not occur outside of existential there-

constructions.

- Deictic there is a locative adverb; existential there is not.

- Deictic there is not a grammatical subject; existential there is a gram­matical subject.

- Deictic there almost always bears stress; existential there almost never

does.

The lone case I know of where deictic there does not bear stress is the de­livery deictic (Th eee re's your pizza!). The only case I know of where

Comparison of Deictics and Existentials 541

existential there does bear stress is when it occurs in WH-exclamations, which require a stressed subject:

- Were THERE ever a lot of problems!

Existential there is basically unstressed but appears to be able to acquire stress from a specialized construction. The only apparent counter­example I know to this generalization is the What, ME worry? construc­tion, which requires a stressed subject, but disallows existential there as

subject (Akmajian, 1984):

- *What, THERE a fire?

However, this construction excludes third person inanimate pronouns in general:

- "'What, IT on fire?

This independently needed restriction also rules out sentences containing existential there and accounts automatically for the ill-formedness of

-What, THERE a fire? which eliminates such cases as counterexamples. The generalization, therefore, seems to stand:

- Existential there is unstressed in the existential (ftere-constructions, but stress can be imposed by a construction imposing stress on a sub­ject.

From the list of properties given above, it is clear that the spatial deictic there of the central deictic construction thus differs considerably from the existential there. Yet there are intermediate types of deictic there that are closer to the existential there. For example, the there's of the perceptual and discourse subconstructions do not refer to a concrete location in space, but rather to abstract locations in perception and discourse. The there of the existence deictic construction, as in There goes our last hope, refers to a location in conceptual space that characterizes existence. Semantically, this comes very close to the existential there. And the there of the delivery deictic, when it does not bear stress, is phonologically in­distinguishable from the existential there. Thus there is a continuum be­tween the spatial deictic there and the existential there.

I would like to suggest that this continuum has a significance in the or­ganizational structure of English grammar. 1 will argue that there is a category of existential constructions parallel to the deictic constructions, and that the central existential construction is based on the central deictic construction. I contend that there is only one fundamental difference be­tween the constructions—the reference of existential there—and that all the other major differences are a consequence of that one distinction.

544 Case Study 3

struction, except for those that are incompatible with the assumption that

there designates a mental space. Here are some predictions;

Prediction 1: Existential there is not a locative adverb.

The reason is that it does not designate a location. This makes there eligi­ble to serve as grammatical subject, though we have not yet predicted why it does in fact serve that function.

Prediction 2: Existential there does not contrast with here.

Again the reason is that it does not designate a location and therefore cannot contrast with another location.

Prediction 3: Existential there does not exist as a "free" lexical item, independent of existential constructions.

Existential there, on our hypothesis, occurs in English only as a conse­quence of its appearance in this construction.

Prediction 4: Existential there does not bear stress.

Elements that are backgrounded, do not convey any new information, and are not involved in any contrast are unstressed. Nothing can be more backgrounded than the space itself. There is not involved in any contrast and conveys no new information. Because there is unstressed, the vowel in it is subject to vowel reduction.

Prediction 5: The existential there does not take a pointing gesture.

Since it doesn't designate a specific location, the pointing gesture that goes with the deictic there is inappropriate.

Prediction 6: The basic-level locational verbs sit, stand, and lie cannot occur in the central existential construction.

These verbs concern orientation in physical space and are not applicable to mental spaces.

Prediction 7: TTiere-constructions cannot be used to point something out and therefore do not have the speech act condition of the central deictic.

Since existential there does not designate a location, it cannot serve the locating function that deictic there serves in the deictic constructions. The fact that existential there designates a mental space is incompatible with the speech act condition of the central deictic, which is to direct the hearer's attention to the specific location designated by there (or here).

Comparison of Deictics and Existentials 545

Prediction 8; TTiere-constructions can take negatives, questions, etc., and can be freely embedded.

This is a consequence of their not having the speech act condition associ­ated with the deictic constructions.

Prediction 9: Existential fftere-constructions are not limited to simple present tense, but can take full auxiliaries.

This is also a consequence of their not having the central deictic's speech act condition.

Prediction 10: The central existential construction has the same func­tional condition as the central deictic, namely, to focus the hearer's awareness on the referent of the noun phrase.

Elimination of the speech act condition does not affect the functional con­dition. The reason is that the speech act condition made reference to ele­ment Г, the location, whereas the functional condition made reference to element 3', the entity located. Thus the existential construction, like the central deictic, functions to focus the hearer's awareness on the entity designated by the noun phrase.

Prediction 11: The noun phrase is not both definite and specific.

The function of the construction is to focus the hearer's awareness on the referent of the construction. If the construction is to serve this function, then either the hearer must not have been aware of the referent, or he must have forgotten about it. Let's put the case of forgetting aside for now, since we will discuss it in detail shortly. Reminding aside, it must be assumed that the hearer has not been aware of the entity referred to. If the noun phrase refers to a specific entity, then the definite article indi­cates that the hearer is already aware of it. This violates the functional condition on the construction. Hence, the noun phrase cannot be both specific and definite, unless it is serving a reminding function. We can fac­tor out cases of reminding by introducing a negative into the sentence. This allows us to get clear cases where nonreminding existentials cannot have a specific definite noun phrase. Compare the following:

- There wasn't a man in the room.

- "There wasn't the man in the room.

- There isn't one in the house, (one is indefinite)

- *Therc isn't it in the house, (u is definite and specific)

546 Case Study 3

Note, incidentally, that the definite article can be used freely with nonspecific noun phrases, for example:

- There was the usual argument in class today.

- There's the strangest smell coming from the refrigerator.

- There's this woman next door who plays the stereo too loud.

These uses of definite determiners the and this in these constructions do not function to pick out specific referents that the hearer is assumed to be already aware of. That is why they are permitted in the existential con­struction.

Let us return to our predictions.

Prediction 12: There is first; syntactic element 4, the "final phrase," is last.

These are inherited from the central deictic construction. But there is an additional reason why there is first. Backgrounded elements tend to occur earlier in a clause than foregrounded elements. Nothing can be more backgrounded than the space in which entities occur.

Prediction 13: The verb precedes the noun phrase.

The syntactic condition inherited from the central deictic is that the verb precedes the noun phrase, unless the noun phrase is a definite pronoun. Since definite pronouns are ruled out of the central existential construc­tion altogether, the inherited condition has the verb always preceding the noun phrase.

Is There the Subject?

In the central deictic, syntactic element 3 (the noun phrase) is the gram­matical subject. One might expect this condition to be carried over to the central existential. But it cannot be. The reason is that there occurs first. Like element 3, there is a noun phrase, not an adverb, and there is an ar­gument of the main verb be. English is a subject-initial language, which means that when there are two nonadverbial noun phrase arguments of the main verb in a central clause, the subject must be first. Since there is first, only it can be subject. Otherwise one of the most basic constraints of English grammar would be violated. This constraint does not hold in spe­cial topicalization constructions, but there doesn't have the right kind of semantics to be a topicalized element—one doesn't topicalize the element referring to the background space.

Is There the Subject? 547

Prediction 14: There is subject.

This prediction is borne out. There are three relevant tests for sub-jecthood. The first concerns tag formation. Tags like "will you?" and "didn't he?" can be appended to the end of declarative sentences. One of the conditions is that the pronoun in the tag agree with the subject, as in:

- John kicked the dog, didn't he?

- *John kicked the dog, didn't it?

In existential constructions, tags agree with there, indicating that there is subject:

- There was a dog in the yard, wasn't there?

- *There was a dog in the yard, wasn't it?

The second test has to do with a construction known as "raising." Con­sider a sentence like

- It is believed that Harry is sick.

In such situations, the subject of the {hat-clause, in this case, Harry, can be "raised" and can appear as the subject of the higher clause is believed:

- Harry is believed to be sick. Now suppose there is an existential construction in the (йж-clause:

- It is believed that there is a dog in the yard.

Only subjects can be raised. To test for subjecthood of the existential con­struction, we ask whether there or a dog can be raised to become subject of is believed. The answer is there:

- There is believed to be a dog in the yard.

- *A dog is believed there to be in the yard.

The third test is subject-verb agreement. Here the results are not at all clear-cut. Sometimes there acts as subject and governs agreement, and sometimes the noun phrase does. There is also considerable variation from speaker to speaker. In my own speech, here is what happens: If the verb is contracted, it agrees with either there or the noun phrase. If the verb is not contracted, it agrees with the noun phrase.

- *There is problems with your proposal.

- There are problems with your proposal.

- There's problems with your proposal.

- There're problems with your proposal.

548 Case Study 3

There are other speakers for whom only the noun phrase can govern agreement. For them, there does not act like a subject at all with respect to agreement, and the third sentence above {There's problems with your proposal) is ill-formed for such speakers. (For a general discussion of this phenomenon see Dixon 1977, Nathan 1981, and Sparks 1984.)

But even though the verb may agree with the noun phrase, that does not necessarily mean that the noun phrase is acting as subject. Cases where the verb agrees with the noun phrase can be handled in two ways:

either straightforwardly, with the verb agreeing directly with the noun phrase, or indirectly, with there agreeing with the noun phrase and the verb agreeing with there. The second alternative—indirect agreement— may sound strange, but there seems to be strong evidence in its favor. The evidence comes from raising cases. Suppose the noun phrase is plural, like problems in the above example. Suppose further that the noun phrase is embedded in a that-clause:

- It is believed that there are problems with your proposal.

In the raising version, there is subject of be believed and be agrees with there. Problems remains in the that-c\saise and is not close enough to be believed to be available for direct agreement. Yet the be of be believed can

be plural:

- There are believed to be problems with your proposal.

In fact, the same agreement paradigm occurs in my speech:

- *There is believed to be problems with your proposal.

- There are believed to be problems with your proposal.

- There's believed to be problems with your proposal.

- There're believed to be problems with your proposal.

These facts suggest that the indirect agreement proposal is correct, and that there must be represented as agreeing with the noun phrase, and all verb agreement occurs via there. This makes there undisputed grammati­cal subject of the construction. Yet even if such an analysis is adopted, there remains the question of why there should agree with the noun phrase, or to put it another way, why the subject of the construction should have the agreement properties of the noun phrase. The intuitive reason seems to be that the noun phrase, even if it is not the grammatical subject, has some right to the title, since it is subject in the base construc­tion, that is, in the central deictic construction. The agreement properties might be construed as indicating that the noun phrase would have been subject if other considerations had not intervened. I believe that this

Final Phrases 549

should be included in any explanatorily adequate account of the agree­ment properties of this construction.

Why There and Not Here?

According to our hypothesis, semantic element 1', which designated a lo­cation in physical space in the central deictic construction, will designate a mental space in the central existential construction. In the central deictic, the corresponding syntactic element, 1, was expressed by either of two contrasting lexical items—here and there. In the existential construction, the possibility for such a contrast has been eliminated and at most one of these modes of expressing semantic element Г can be inherited. The fact is that there is inherited and here is not. Is there a reason for this?

There seem to be at least two reasons why there is chosen, and no rea­sons at all for choosing here. The simplest and most straightforward ra­tionale for the choice of there has to do with the fact that the referent is a mental space—the medium in which entities occur. A space is understood as the ultimate background relative to which the entities in the space are foregrounded and is therefore comprehended as being distant. This way of comprehending a space makes the choice of the distal there natural and the proximal here unnatural.

The second rationale concerns markedness. When a binary contrast is eliminated, it is most natural for the unmarked member of the pair to be chosen. There seems to be the unmarked member of the hero-there, pair. Deictic there appears earlier than here in children's speech. And it is much more frequent.

Prediction 15: 1 = there.

Final Phrases

Existential constructions, like deictic constructions, have final phrases. The exact nature of the final phrase in the central existential construction is a direct consequence of the nature of the final phrase in the existential construction. Here are some examples (with final phrases in italics).

- There is someone in the yard. (locative phrase)

- There's a boy running away. (progressive participial phrase)

- There wasn't any money stolen, (passive phrase)

- There isn't anyone taller than Harry, (adjective phrase)

- There was no one with his shirt on. (special with-phrase)

- There is a concert at noon. (temporal phrase)

- There is a man about to leave, (about to-phrase)

550 Case Study 3

The range of final phrases permitted in the central existential construction is the same as that in the central deictic construction. Moreover, the same phrases are ruled out:

- "There is a man a doctor. (No predicate nominals)

- "There is a man tall. (No "inherent" adjectives)

As with the central deictic, the final phrases are optional and iterable. Here are some cases with no final phrase:

- There won't be any trouble.

- There's someone who wants to see you.

- There may be problems.

and a case of iteration:

- There's a man standing at the front door mad as hell about to call the

cops.

Prediction 16: 4 is an optional VP/be and can be iterated, Prediction 17: 3 is subject of 4.

From the central deictic construction, we also predict that syntactic ele­ment 3, the noun phrase, functions as grammatical subject of the final phrase, if it is present. This will allow us to account automatically for a great many common syntactic properties of final phrases, such as the oc­currence of reflexive pronouns, equi constructions, floating quantifiers, etc. (For an idea of how such phenomena can be accounted for in a theory of constructions, see Lakoff 1977.)

No Verbs of Motion

Another prediction that can be made is that the central existential does not contain any verbs of motion—neither go nor come. As we shall see, some noncentral existentials do contain verbs of motion, but they func­tion in special ways which we shall discuss below. The following are not

possible existentials:

- "There will go a boy to the ballgame.

- "There can't come any muggers in here.

In the central deictic construction, the occurrence of go and come were consequences of the fact that one can point out things that are moving through physical space. But entities do not move through mental spaces (though they may come to be in them). Thus, the semantics of mental spaces rules out motion through a mental space, and hence predicts the absence of go and come in the central constructions.

Constituent Structure 551

Prediction 18: 2 = be.

To allay objections at this point, I should point out that perfectly fine existential sentences like

- There will come a time when you'll be sorry. and

- There ran into the room three strange men dressed as walruses.

will be discussed below when we get to noncentral existential construc­tions.

Incidentally, we can now restate our description of 4 to read:

4 = VP/2 since 2 is be.

Constituent Structure Prediction 19: 0 is a clause with parts 1,2, 3, and 4.

In the central deictic construction, syntactic elements 3 and 4—the noun phrase and the final phrase—are sister constituents within the clause and not parts of a single larger noun phrase. For example, in

- There's Harry in the room.

Harry in the room is not a single unit—and in particular, not a single noun phrase. Our hypothesis predicts that this constituent structure is inherited by the existential construction. Thus, in the existential sentence

- There's a man in the room.

a man in the room has at least one analysis as a sequence of two disjoint constituents—a noun phrase followed by a final phrase. It is not just a sin­gle noun phrase.

We can prove this by making use oFRoss constraints (see Ross 1967). Ross showed that WH-questions cannot leave gaps in either relative clauses or reduced relative clauses. The italicized portions of the follow­ing sentences are noun phrase units. The first contains a relative clause, while the second contains a reduced relative clause (with who is missing):

- John knows a man who is in the room extremely well.

- John knows a man in the room extremely well.

If we attempt to question the room, gaps will be left behind in the relative clause and the reduced relative clause.

552 Case Study 3

- *Which room does John know a man who is in __ extremely well?

- *Which room does John know a man in __ extremely well?

Similarly, the italicized portion of the following sentence is a noun phrase unit containing a relative clause.

- There is a man who is in the room.

Thus we cannot get:

- * Which room is there a man who is in __?

However, in the sentence

- There is a man in the room.

the italicized portion does not function like a single noun phrase unit with a reduced relative clause. Thus we have questions like:

- Which room is there a man in ——? , • • ,

Thus, a man in the room functions not like a single noun phrase but like a sequence of two constituents, a man and in the room, just as predicted.

Minimal and Full Representations

Given all of tjie above predictions, we can represent the central existen­tial (йеге-construction in a remarkably minimal fashion:

The Central Existential Based on: The Central Deictic Semantic Element V: a mental space

This is all that needs to be said. The rest follows from the principle of in­heritance, principles of language in general, and independently needed principles particular to English. One way of understanding this is that the central existential is just the central deictic applied to mental spaces in­stead of locations in physical space.

To see just how much is predicted, let us take a look at a full represen­tation:

The Central Existential Based on: The Central Deictic

Parameters of Meaning

An Alternative 553

Semantic Elements 1': a mental space

a predicate locating an entity in a mental space an entity

2' 3' 4': a noninherent predicate

0': a proposition, 2'(3',1') 0': a proposition, 4'(3') S: Speaker H: Hearer Function To focus H's awareness on 3'

a proposition, 2'(3',1') a proposition, 4'(3')

Parameters of Form

Syntactic Elements

0: a clause, with parts 1,2,3,4

1: noun phrase

1: verb .

3: noun phrase

4*: VP/2 '. Lexical' Conditions

1 = there

2 = be Syntactic Conditions

S-1: 1 is first . . S-3: 2 precedes 3 '

S-4: 4 is optional

S-5: 4 is last

S-6: 1 is subject of 0

S-7: 3 is subject of (0; 3,4)

S-8: number (1) agrees with number (3); optional if 1 governs con­traction Phonological Conditions

1 is unstressed

This should give some idea of the explanatory power of grammatical construction theory. All of the full representation is predictable from the ecological location and one semantic condition.

An Alternative

Suppose that, instead of analyzing the central existential as based on the central deictic, we had analyzed the central existential as being based on the central clause type. Something akin to this, though by no means ex-

554 Case Study 3

actly the same, is done in transformational grammar, where there is an at­tempt to derive existentials from central clauses.

Perhaps the best way to compare the analyses is to ask how much of the full representation of the central existential is predictable if we try to base the central existential on the central clause type. To find this out, what we have to do is construct a minimal representation in which the central exis­tential is based on the central clause type, and then see how complex it

is.

The syntactic elements of the central existential that correspond to sim­ple sentences are elements 0,2,3, and 4—the clause, made up of the verb be, the noun phrase, and the final phrase. For example, corresponding to There is a thief running away, would be the central clause 0 = A thief is run­ning away,v/hete2 = (5,3 = a(/ue/,and4 = runningaway. The semantic ele-mcnts corresponding to such central clause types are proposition O", en­tity 3', and predicate 4', where 0" = 4'(3'). In this case, 0" is the meaning of A thief is running away. The noun phrase, 3, is the subject of the final phrase, 4—or as we have been writing it, {0; 3,4).

The parts of the central existential that are predictable are:

Semantic Elements

O": a proposition, 4'(3')

3': an entity

4': a predicate Syntactic Elements

0: a clause, with parts 3, 4

3: a noun phrase Syntactic Conditions

S-7: 3 is subject of <0; 3,4)

This is what is predictable simply from saying that the central existential is based on a central clause type. Most of the central existential construction is not predictable. Here is what a minimal representation would have to

be like. The Central Existential (Alternative Version)

Parameters of Meaning

Semantic Elements Г: a mental space 2': a predicate locating an entity in a mental space

4'*: noninherent 0': a proposition, 2'(3',1')

Based on Versus Derived from 555

S: Speaker H: Hearer Function To focus Ifs awareness on 3'

Parameters of Form

Syntactic Elements

0: a clause with parts 1, 2

1: a noun phrase

2: a verb Lexical Conditions

1 = there

2 = be Syntactic Conditions

S-1: 1 is first

S-3: 2 precedes 3

S-4: 4 is optional

S-5: 4 is last

S-6: 1 is subject of 0

S-8: number (1) agrees with number (3); optional if 1 governs con­traction. Phonological Conditions

1 is unstressed

Just about all of the central existential would be unpredictable if it were based on the simple sentence. In such an analysis, these properties would be arbitrary and unmotivated. The analysis given above can thus explain almost everything that this analysis would have to state as arbitrary.

Based on Versus Derived from

Those linguists who tried to relate existentials to deictics—Fillmore, Kuno, Lyons, and Thorne—had a correct intuition. Their problem was that they tried to use transformational grammar to show the relationship by transformationally deriving the existential from the deictic. Of the twenty predictions we were able to make using such an analysis in gram­matical construction theory, a transformational analysis could make only two of the minor predictions—prediction 17 (that 3 is subject) and predic­tion 19 (the constituent structure).

Thus, the concept based on in grammatical construction theory is con­siderably different from the concept derived from in transformational grammar.

556 Case Study 3

One Category or Two?

We have argued that the deictic constructions form a natural category with the central deictic as prototype. For the most part, the nonprototypi-cal deictic constructions are based on the central deictic. But we have just argued that the central existential construction is also based on the central deictic. Does that make the central existential part of the deictic cate­gory?

The answer is no. The central existential, in its full representation, is so

different from any of the deictics that it is in a category of its own. There are well over a dozen differences between the constructions, and many of them are major differences. The categories may be "adjacent," in that there are also many similarities as well as examples whose superficial form is differentiated only by stress. Yet the categories are distinct.

What is remarkable about this is that the difference between the cen­tral deictic and the minimal representation of the central existential is so small. Given just one difference, the difference between a location in physical space and a mental space, a great many major differences follow. This is an illustration of how a small difference, placed within a system of complex principles, can yield a very large difference. It is a dramatic illus­tration of the effects that the overall ecology of a grammatical system can

have.

Some Consequences

We have claimed that the central existential construction is based on the central deictic and not, as many linguists have suggested, on the central clause. This allows us to explain a great deal more about this construction than we could explain by analyzing the construction as being based on a central clause. At the same time, it permits us to show how the central ex­istential is related to the central clause. This is done via one of the propo­sitions expressed, 0'", which predicates 4' of 3', and which corre­sponds to the substructure (0; 3,4). Thus we can account for the fact that

in a sentence like - There's a man shaving himself.

the noun phrase a man (element 3) is understood as the subject of. shaving himself (element 4). This allows us to relate existentials to central clauses corresponding to (0; 3,4) without basing the construction on those central

clauses.

Explaining the relationship in this way avoids the problems that trans­formational grammarians have run into by trying to transformationally

Some Consequences 557

derive existentials from central clauses. Transformational treatments face problems like trying to derive the existential

- There's a man who wants to see you. from the impossible central clause

- *A man who wants to see you is.

On our account there is no problem, since the existential construction is based not on the central clause type, but on the deictic construction.

Our analysis also allows us to account automatically for phenomena like the following.

- There's a Japanese executive in the waiting room.

- A Japanese executive is in the waiting room.

- There's a Japanese executive in our company.

- *A Japanese executive is in our company.

The difference is a consequence of a general principle of English gram­mar plus the difference between rooms and companies. Companies are institutions that are in part constituted by the people who are employed in them. Rooms are not like this. People in a room do not in part constitute the room, unless they are specifically placed there to be part of the makeup of the room—say, a receptionist.

- There's a receptionist in the waiting room.

- *A receptionist is in the waiting room.

The last sentence is not appropriate if the receptionist is there in her func­tion as the receptionist in that room. It is however fine if someone who is a receptionist somewhere else happens to be in the waiting room.

A transformational analysis would have difficulties with these cases if it were to try to derive all existentials from corresponding simple sentences. The problem is that, in these cases, the corresponding simple sentences don't exist. Such a problem doesn't arise in the analysis we have pro­posed, since the existential sentences are not derived from simple sen­tences. All the existential sentences are straightforward cases of the cen­tral existential construction. The problem lies not with the occurrence of the existentials, but with the nonoccurrence of the simple sentences. The general principle involved, the part-whole predication constraint, seems to be something like the following:

In a central clause with an indefinite subject and be as the main verb, the verb phrase cannot predicate that the subject is part of some whole with respect to some conceptual schema.

558 Case Study 3

In the sentences given above, there is a company schema in which execu­tives are understood as part of the company. Thus the principle rules out *A Japanese executive is in our company. Similarly, there is a reception room schema in which the receptionist is part of the reception room. The principle will therefore rule out *A receptionist is in the reception room if the receptionist is the one working in the reception room. Other exam­ples abound:

- There is a flaw in the diamond.

- *A flaw is in the diamond.

- There is no lid to this jar,

- *No lid is to this jar.

- There are Szechuan peppercorns in this dish.

- *Szechuan peppercorns are in this dish.

Incidentally, English has a special construction to mark cases where there is a schema of the sort described above, where one entity is under­stood relative to that schema to be part of another entity. Here are some examples of that construction:

- The diamond has a flaw in it.

- The jar has no lid to it.

- This dish has Szechuan peppercorns in it.

- The reception room has a receptionist in it.

- *The reception room has a Japanese executive in it.

This accounts for the following phenomenon:

- There is a vase on the table.

This can be understood in two ways. Either a vase happened, inciden­tally, to be placed there, or the vase is part of the table setting. The exis­tential is neutral between these two situations. However,

- A vase is on the table.

is restricted to the incidental-occurrence sense, while - The table has a vase on it.

suggests that the vase is part of an overall image including both table and vase, as in the case of a table setting.

The part-whole predication constraint will also account for cases where the noun phrase designates a literal or metaphorical substance that in­heres in some entity.

Some Consequences 559

- There's not much to him.

- There's a great deal of merit in his theory.

- There's not much substance to his claims.

Here the final phrase is a prepositional phrase beginning with either to or in. The noun phrase which is the object of the preposition (e.g., his the­ory) designates the entity that the substance (e.g., merit) inheres in. Be­cause substances are understood as part of the entities that they help to constitute, this construction fits a part-whole schema. As a consequence there are no corresponding simple sentences, but there are corresponding /love-sentences:

- ""Not much is to him.

- He doesn't have much to him.

- *A good deal of merit is in his theory.

- His theory has a good deal of merit in it.

- *Not much substance is to his claims.

- His claims don't have much substance to them.

Incidentally, the to in this construction has a bizarre and as yet unex­plained quirk. Whereas in can occur in postnominal modifiers like the merit in his theory, to cannot occur in such cases:

- The merit in his theory is considerable.

- *The substance to his theory is considerable.

Though I cannot explain this quirk, the fact that it exists is of interest. It provides a test to distinguish final phrases from postnominal modifiers, and it shows that in

- There is not much substance to his claims.

to his claims is a final phrase, as suggested above, and not a postnominal modifier attached to substance.

Although the distribution of the prepositions in these examples is ex­tremely peculiar, there is no reason to think that they are tied to the exis­tential г/геге-construction, and hence no reason to set up a special subcon-struction to account for them.

The part-whole predication constraint does not cover a range of other cases where indefinite subjects cannot appear in simple sentences with be as the main verb.

- There's an hour before lunch.

- *An hour is before lunch.

560 Case Study 3

- There's a concert at ten o'clock.

- *A concert is at ten o'clock.

These involve temporal, not part-whole, relationships, as the following examples show:

- *Lunch has an hour before it.

- *Ten o'clock has a concert at it.

Another principle will be required to rule out temporal predicates of in­definite subjects in central clauses.

To summarize: Cases like those above, where there is no central clause corresponding to an existential sentence, pose no problem for the analy­sis we have given of existentials. Principles to account for the ill-formedness of those central clauses are needed independently. Since the principles are restricted to central clauses, they have no effect on existen­tials.

Incidentally, one automatic consequence of our analysis of the central

existential is that examples like the following are automatically predicted to be ill-formed;

- *There was being a man shot.

This is a clause of the form: NP-be-NP -.... Such a clause can take the progressive auxiliary on be (is being) only when the "be + noun phrase" can be understood as designating some action over which the subject is exercising control. For example,

- John is being a fool.

- Sally is being a nuisance.

are possible because one can actively be a fool or a nuisance, but

- *Harry is being the man she married.

- *Sam is being the worst hitter in the American League.

are impossible because the subject cannot exercise control over such things. Similarly, a mental space cannot exercise control over the entities in it. Thus, the be of the existential can never take the progressive auxil­iary, and the ill-formedness of

- ""There was being a man shot. follows as a consequence. However,

- There was a man being shot.

Lists 561

is permitted in a straightforward fashion. Being shot is a final phrase, which is based on simple clauses. It is permitted because A wan was being shot is permitted.

Reminders

The function of existential constructions is to focus the hearer's aware­ness on semantic element 3', the referent of the noun phrase. This can happen in two situations: either the hearer has never been aware of 3', or he has been aware of 3' and forgotten about it. It is only in the latter case, where the speaker is reminding the hearer of 3', that the noun phrase can be both definite and specific. Examples include:

- There's always Harry.

- But there's the dog!

- There's still the remains of Christmas dinner in the freezer.

Pronouns used deictically can also occur here.

- If you need a backup catcher, there's always me.

- If you need a good pinch-hitter, there's always him. (said pointing to someone)

It is predicted, however, that anaphoric pronouns cannot occur in this construction at all. Since their antecedents, if they are to function as ante­cedents, cannot have been forgotten, anaphoric pronouns cannot func­tion as reminders.

On our analysis, all of these reminding uses are simply cases of the cen­tral existential. They just happen to be used in a reminding function. When the noun phrase is definite and specific, reminding is the only func­tion compatible with the general constraints on the construction.

Lists Examples of reminders are most commonly given in the form of lists:

- There's the cat to feed, the dog to walk, the horse to brush,. ..

- There'll be Max at the head of the table. Sally next to me,. ..

Lists of this sort are a general feature of English and are not peculiar to reminding uses of existentials. They apply to all sorts of other construc­tions:

- Joan is prettier than Sue, richer than Melanie, smarter than Eliza,. . .

- Bring the camera, the backpack, the canteen,. . .

562 Case Study 3

- I want to give Tom a sweater, Jeff an espresso-maker,. . .

- Tom likes cats, Sally horses, Mike dogs,. . .

Existential lists are simply cases where the list construction has applied to an existential sentence. They are not part of the analysis of existentials at all.

A sentence with a definite NP like

- There's Harry.

can be either a central deictic (if there is stressed) or, under special cir­cumstances, a central existential (if there is not stressed). The "special cir­cumstances" are those that characterize reminders, namely, the hearer has been aware of the referent of the NP and has forgotten about it. The has-been-aware-of-it condition sanctions the definiteness, while the has-forgotten-about-it condition sanctions the existential fAere-construction, whose function is to introduce it to the hearer's awareness. If these "spe­cial circumstances" do not obtain, then it is pragmatically odd to take There's Harry as an existential construction.

This completes our study of the central existentials. We can now turn to the noncentral cases.

The Noncentral Existentials

Let us begin by considering examples of each of the constructions we will be discussing.

Central: There's a masked man outside.

Strange: There's a man been shot.

Ontological: There IS a Santa Claus.

Infinitival: There's making dinner to start thinking about.

Presentational: There walked into the room a tall blond man with one

black shoe.

The Strange Existential

Guy Carden (1978) has observed that there is an existential construction that is a minimal variation of the central existential, but has some unusual properties. Compare the sentences:

- There's been a man shot. (central existential)

- There's a man been shot. (strange existential)

The Strange Existential 563

Although many speakers, on first inspection, don't notice that the strange existential is in any way different from the central existential, a little in­vestigation reveals that it is unusual. It appears at first that the contracted verb is a contraction of is, but that is false. Is cannot appear in place of the contraction 's in these sentence types.

- "There is a man been shot.

Moreover, an inspection of possible tag questions shows that the verb is not is, but has\

- "There's a man been shot, isn't there?

- There's a man been shot, hasn't there?

Carden also observes that there is another construction that includes a pronominal copy of the subject and the main auxiliary.

- John's left, he has. This also reveals that the contracted 's is has, not is.

- There's a man been shot, there has.

- "There's a man been shot, there is.

Yet has cannot appear uncontracted.

- "There has a man been shot.

Apparently, contracted has can appear in this construction because it looks like contracted is. Notice that the strange existential does not occur in the plural, since 've (the contracted form of plural have) is not the same form as the contracted plural of be, namely 're.

- *There've many people been killed this week. The corresponding central existential

- There've been many people killed this week.

is fine, as expected.

The strange existential not only occurs with passive final phrases, but also with just about any verb phrase minus have of the perfect.

- There's a strange dog been hanging around outside.

- There's an actor been president for almost six years now.

- There's a man robbed the drugstore down the street.

- There's someone fallen overboard.

We can represent this construction minimally as follows:

564 Case Study 3 The Strange Existential Based on: The Central Existential Lexical Condition

2 = have of the perfect Phonological Condition

2 = 's

Given that the central construction contains the condition that 4 = a verb phrase minus 2, the fact that 2 = the have the perfect will predict that 4, the final phrase, will be a verb phrase minus the have of the perfect.

Rational Properties

Although such an account of this construction is perhaps most perspicu­ous, it does miss something. It is no accident that the phonological form chosen is 's. This is the contracted form of is, which is the verb in the cen­tral construction. And the contracted form is, in fact, the most typical form used. That is, one usually finds there's rather than there is in casual

conversation.

In short, there is a reason why the phonological form in the construc­tion is's and not some other random element. It is a rational thing to oc­cur, not an arbitrary form. Yet it is not a motivated property of the con­struction, as we are using the term. To be "motivated" it would have to be predictable from the construction's ecological location (what it is based on) plus general principles of grammar. But it is predictable from one other piece of information—the fact that it is the same as the typical form of the corresponding element of the central construction.

We can represent this information in the following way. Let us refer to element 2 of the central existential construction as the "ancestor" of 2 in the strange existential construction. Let us abbreviate "the typical form of the ancestor of 2" as TFA(2). Were we to list "TFA(2)" as the phono­logical condition of the strange existential, it would follow that 2 = 's.

Given that 2 = 's, there are only two things that 2 could be—either is or has of the perfect auxiliary. Note that it is only the perfect has that con­tracts. Thus, John has a book does not contract to John's a book, while John has been here contracts to John's been here. Since there are only two possibilities, all we need to know about the lexical item chosen as 2 is that it is not is, that is, it is not the same as the lexical item chosen in the central construction, or equivalently that it is not the same as its ancestor. If we abbreviate "ancestor of 2" as A(2), then we can represent as:

The Ontological Existential 565

Lexical Condition 2 = A(2).

With these two pieces of information about the relationship between element 2 in the strange construction and the corresponding elements in its ancestor construction, we can predict both of the above conditions that define the strange construction. The fact that this can be done shows that these are not arbitrary conditions. But they are not predictable in the way that motivated conditions are. That is, they lean somewhat more toward arbitrary properties than motivated properties do, but they are not by any means completely arbitrary. We will refer to such properties of construc­tions as rational properties.

I will use the term rational property for all nonarbitrary properties that are not motivated, in the technical sense in which we are using the term in this case study. Language” contain a great many rational properties, properties that are not arbitrary but make sense in terms of the rest of the system. I do not at present have anything approaching a general theory of rational properties. However, I think they are one of the most interesting aspects of natural languages, and they deserve serious study.

Like motivated properties, rational properties seem to be easier to learn, remember, and use than arbitrary properties. This suggests that they too have an important cognitive status.

The Ontological Existential

One of the existential constructions is actually concerned with ontology, that is, with whether an entity exists or not. In this construction, the verb exist can appear instead of be, and the verb bears stress. It is also common for the noun phrase to be an indefinite version of a proper name, since dis­cussions of existence often center upon whether some entity for which there is a proper name exists.

- There is a Santa Claus.

- There is no God.

- If there exists a Valhalla, I want to go there.

Other examples include:

- There are five prime numbers below twelve.

- There weren't cars in 1876.

- How could there exist a God without a mother?

- There is work to be found.

566 Case Study 3 We can represent this construction minimally in the following way:

The Ontological Existential

Semantic Elements

2" = a predicate of existence

0" - 2" (3') Lexical Condition

2 may be exist Phonological Condition

2 bears stress

Incidentally, this is another construction that cannot have correspond­ing simple sentences:

- *A Santa Claus is.

- *No God is.

There are, however, corresponding simple sentences with the verb exist:

- Santa Claus exists.

- No God exists.

The above analysis accounts for the relationship between the existentials and such simple sentences via the proposition, O", in the semantics, which expresses the proposition that the referent of the noun phrase exists. This is the proposition expressed by the corresponding simple sentences. In a sentence like There exists a God, the subpart corresponding to this propo­sition consists of syntactic elements 3 (a God) and 2 (exists) and bears a direct relationship to the corresponding simple sentence Л God exists.

Moreover, a sentence like

- There exists a Santa Claus living quietly at the North Pole.

is related to two simple sentences via the propositions 0" and 0'" in their semantics. 0" is of the form 2'(3'), which corresponds to the syntactic subpart consisting of syntactic elements 3 (A Santa Claus) and 2 (exists). 0'" is of the form 4'(3'), which corresponds to the syntactic subpart con­sisting of elements 3 (Л Santa Claus) and 4 ((is) living quietly at the North Pole). The above existential sentence is thereby related systematically to the corresponding simple sentences;

- A Santa Claus exists.

- A Santa Claus is living quietly at the North Pole.

The difference between the way that be and exist function in these exam­ples is the following: * A Santa Claus is ruled out as a simple clause because the copula be must take a following NP. A Santa Claus exists is allowed as

The Infinitival Existential 567

a simple clause because exist is intransitive. Both There is a Santa Claus and There exists a Santa Claus are instances of the ontological existential construction. The be in this construction is permitted because the con­struction is based on the central existential, which in turn is based on the central deictic (which has a be). Exist can occur in this construction by the lexical choice principle, since the semantic conditions list a predicate of existence.

To my knowledge there is no transformational analysis of such existen­tials that adequately accounts for the way they are systematically related to two such simple sentences.

The analysis given above also accounts for the fact that definite specific noun phrases cannot occur with this construction.

- "There exists the negative square root of nine.

The reason is that one cannot both be predicating the existence of an entity and reminding someone about it. One can remind someone of something specific, but only if its existence is taken for granted.

The Infinitival Existential

In morphology, there is a distinction made between bound and free morphemes. Bound morphemes are those that cannot occur indepen­dently, but only occur as part of larger units. For example, the -ing of working and the -s of cats are bound morphemes. There is a similar dis­tinction to be made in grammar between bound and free constructions. A bound construction is one that occurs only as part of other constructions.

There is an existential construction that contains a bound construction within it. I have in mind cases where the bound construction is an in­finitival clause which contains a gap that is not in subject position. In the following examples, the gapped infinitival clause is in italics, and the noun phrase which is understood as fitting into that gap is in bold.

- There is a bed for you to sleep in ___.

- There's food to eat ___,

- There's still getting himself into college for John to start thinking about ___.

- There won't be any producers there for Stephanie to be seen by———.

In these existential sentences, the gapped infinitival clause is the final phrase, and the noun phrase that is understood as fitting into the gap is syntactic element 3.

568 Case Study 3

This construction can be given the following minimal representation:

The Infinitival Existential

Based on: The Central Existential

Syntactic Elements

4: a GAPPED INFINITIVAL CLAUSE

Incidentally, here are two other constructions that make use of the gapped infinitival clause—the bound construction used in the infinitival

existential:

Object-to-Subjcct Raising

- This bed is easy to fall asleep in ___.

- Getting myself to exercise regularly is tough for me to even consider

Nominal with an Infinitival Relative Clause

- Harry is the man for you to see ___.

- This is a good bed to sleep in ___.

- I just found a good paper to assign ___ in class.

The gapped infinitival clause construction can be characterized as follows:

The Gapped Infinitival Clause Based on: The Central Clause Type

Parameters of Meaning Semantic Elements

0': a proposition, with Г as a part.

Г; an entity. 2': an entity, not a part of 0',

Semantic Conditions

Г is coreferential to 2'

Parameters of Form

Syntactic Conditions S-1: 0 is infinitival. S-2: 1 is not subject of 0.

S-3: 1 is null. S-4: 2 immediately commands 0.

Condition S-3 specifies that there is a gap. Since Г is an entity, 1 is an NP gap. The semantic element 2' indicates that this is a bound construction.

The Infinitival Existential 569

The conditions on the pair (2', 2) specify that this construction must fit an "external" entity, which is coreferential to 1'. Condition S-4 tells where this external entity must be relative this construction. Essentially, 2' plays the same role that lambda-abstraction plays in theories such as gen­eralized phrase-structure grammar and Montague grammar. That is, it links the equivalent of a variable inside the construction (Г) to a corefer­ential noun phrase entity (2') in the next highest clause. The "immediate command" condition of S-4 specifies that there must be a noun phrase coreferential with the gap in the next highest clause. This condition must be consistent with whatever constraints there are on that next-highest clause. When the infinitival existential and the gapped infinitive clause constructions are superimposed, element 3' of the infinitival existential will fit onto (that is, be identical to) element 2' of the gapped infinitival clause. It is by such .a mechanism that the theory of grammatical construc­tions in general achieves the same purpose as lambda-abstraction.

From this perspective, we can characterize the gapped infinitive clause in a minimal fashion, by first characterizing the bound clause construc­tion.

The Bound Clause Construction Based on: The Central Clause Type

Parameters of Meaning

Semantic Elements

0' a proposition with Г as a part.

Г an entity.

2' an entity that is not a part of 0'. Semantic Condition

Г is coreferential to 2'. Syntactic Conditions

S-1: 1 is null.

S-2: 2 immediately commands 0.

The bound clause construction is a general construction that has many instances—all clausal constructions containing noun phrase gaps. These include the cqui-construction, the raising construction, the WH-question construction, the relative clause construction, etc. Among its instances is the gapped infinitival clause:

The Gapped Infinitival Construction Based on: The bound clause Construction

Parameters of Form

570 Case Study 3

Syntactic Conditions S-1: 0 is infinitival. S-2: 1 is not the subject of 0.

The Presentational Existential

Narratives create mental spaces of their own, and when an existential construction is used within a narrative, it is the mental space of the narra­tive that is designated by there. Moreover, it is the function of the there-construction to bring the entity designated by the noun phrase into the narrative. Thus when a sentence like There was a loud noise! occurs in a narrative, it functions to bring the noise into the mental space of the

hearer, which is also the space of the narrative. A good narrative is one that allows the hearer to vividly picture the

^•vppts pf the narrative as they are told. English has an existential con-^- - ~ -'—•^ “^ notTotivpg it allows speakeis tu iutioduce-a-

AVP.niS П1 lilt: началу ^ >..^ -.- . -

struction that is specialized to narratives. It allows speakeis tu intiuduca new narrative element, while simultaneously sketching a scene.

- From an asylum near Providence, R.I., there recently disappeared an

exceedingly singular person.

- In the cubicle there was sitting alone a pretty young woman writing a

term paper.

- Suddenly there burst into the room an SS officer holding a machine

gun.

- There once lived in Translyvania an old woman with three sons.

This construction differs in certain important ways from other existen-tials. It permits an entire verb phrase, not just a verb alone, to appear as syntactic element 2. Thus, in the above examples, there are the verb phrases (including auxiliaries): recently disappeared, was sitting alone, and burst into the room. The verb phrase in such a sentence must be in­transitive. Thus, we do not have sentences like:

- *Suddenly there hit Harry over the head a mugger brandishing a base­ball bat.

- *Last year there sent me a Valentine's card an attractive older woman

in my office. These are ill-formed because the verb phrases hit Harry over the head and

sent me a Valentine's card are transitive.

In my speech there is also a constraint on this construction having to do with the length (or possibly the information content) of the noun phrase

and the verb phrase:

The Presentational Existential 571

The verb phrase cannot be much longer (or have much more informa­tional content) than the noun phrase.

Thus, I find the following sentences ill-formed:

- *In the cubicle there was sitting alone a girl.

- * Suddenly there burst into the room a man.

- *Thcre once lived in Transylvania a woman.

The only difference between these sentences and the ones cited above is the length of the noun phrase. If we hold the verb phrase constant and vary the length of the noun phrase, we can see the sentences getting pro­gressively better.

- *Suddenly there burst into the room a cop.

- * Suddenly there burst into the room a tall cop.

- ?*Suddenly there burst into the room a brawny cop.

?'''Suddenly there burst into th“ room a mcan'-tookiog-сор-

- Suddenly there burst into the room a tall, brawny, mean-looking cop.

Here ?* indicates pretty bad and ? indicates pretty good.

We can represent this constraint as a syntactic condition of the form:

Size (3) > Size (2)

This seems at first like a strange constraint, until we consider the function it serves. The refcrcnt of the noun phrase is being set up as a new narra­tive element via the description give in the verb phrase and the noun phrase. The verb phrase must therefore set up an appropriate back­ground against which the noun phrase will count as new information. The noun phrase must therefore convey more new content than the verb phrase. Given the fact that, for the most part, the longer a phrase is the more content it conveys, it follows that the size of the noun phrase must be greater that the size of the verb phrase. Thus, the syntactic condition is a consequence of the function of the construction—bringing a new ele­ment into a narrative via the descriptions given the verb phrase and the noun phrase.

We will refer to the predicate expressed by the verb phrase as semantic element 2". This is predicated of the referent of the noun phrase, 3'. We will refer to the resulting proposition as O". We can now provide a mini­mal representation of this construction.

572 Case Study 3

The Presentational Existential

Based on: The Central Existential

Semantic Elements

Г: a narrative space 2": a predicate O": 2" (3')

Syntactic Elements

2: a verb phrase, with an intransitive head verb.

Given the narrative function of the construction, the syntactic constraint

Size (3) г Size (2)

will follow.

The account we have given of the presentational existential has a num­ber of important consequences. Note that the only constraint that we have placed on the head verb of the verb phrase is that it must be intransi­tive. This is an implicit claim that any intransitive verb should be able to work here providing that the verb phrase functions to set up an appropri­ate background for the noun phrase. This, of course, will depend on many factors—the nature of the verb, the nature of the auxiliary verbs, the time and locational adverbs present in the sentence, etc. Take, as an example, the intransitive verb bleed. By itself, this verb describes a process, but does not set up an appropriate background for the introduction of a narra­tive element. Thus we do not get sentences like:

- "There bled a hemophiliac.

But this does not mean that bleed cannot occur as the head verb of the verb phrase. With the right auxiliary verbs and adverbs modifying bleed, it too can function as the head verb of a presentational existential:

- For two hours there had been bleeding on the emergency room floor a poor hemophiliac who had fainted before he could sign his Blue Cross

form.

So far, I have not found any intransitive verbs that cannot be made to fit this construction. It is just a matter of finding the right combination of aux­iliaries and adverbs, a noun phase of the appropriate size, and a context

that makes sense.

Of course, some intransitive verbs fit more easily than others. Verbs

that have such meanings as "coming into existence," or "changing to a new state," or "moving to a new location," can occur in this construction most easily, usually without any help from auxiliaries or adverbs. Hence the oft-cited examples:

The Presentational Existential 573

- There arose a commotion.

- There ensued a riot.

- There entered a policeman.

In a reminding use, the verb remain can be used, as one would expect.

- There remains one major problem.

These are just special cases of the presentational construction. The verb phrase happens to contain just a single verb. And since the verb phrase is short, the noun phrase can be relatively short. However, it cannot be too small; for example, the indefinite pronoun one is somewhat smaller than the above verbs, and so the following sentences are ill-formed:

- "There arose one.

- "There ensued one.

- "There entered one.

However, with the right context and a short enough verb, the indefinite pronoun is possible:

- Bill had said that he didn't expect any Christmas cards, so when there finally came one in the mail, he was surprised.

With the somewhat longer verbs arrived and was delivered, there is a cor­responding decrease in acceptability.

- ?Bill had said that he didn't expect any Christmas cards, so when there finally arrived one in the mail, he was surprised.

- *Bill had said that he didn't expect any Christmas cards, so when there was finally delivered one in the mail, he was surprised.

Thus, the cases where the verb phrase is just a simple verb are just special cases of the presentational existential construction. No special construc­tion is necessary to handle them.

The analysis given above makes a further prediction about the syntax of this construction, namely, that it cannot be negated. This is a conse­quence of our claim that the verb phrase functions either to predicate an entrance of the new element into the scene, as in

- There emerged from the cocoon a beautiful black and red butterfly.

or to set up a background against which a new element may be intro­duced, as in

- There were singing in the alley below a hearty group of carolers undaunted by the snow and the cold.

574 Case Study 3

In either case, negation is inconsistent with the function of the construc­tion. In the first case, it negates the entrance predicate, communicating the lack of an entrance:

- "There didn't emerge from the cocoon a beautiful black and red butterfly.

In the second case, it negates the predicate setting up the appropriate background, which results in the lack of such a background:

- *There weren't singing in the alley below a hearty group of carolers undaunted by the snow and the cold.

The analysis given also predicts that not all negatives will be excluded. It is possible for certain kinds of negatives to help set up a background, rather than negate one. Never, for example, sets up an extended counter-factual mental space which can serve as a background.

- While I was sheriff, there never occurred any demonstrations of significant proportions.

Even predicates like end can play a role in setting up a background for this construction.

- There recently ended the longest scoring streak in the history of pro­fessional hockey.

The pragmatic function of this construction also predicts where it can and cannot be embedded and what other constructions it can interact with. Let us begin with the sentence

- There jumped out of the hole a jackrabbit with enormous ears.

As Aissen (1975) points out, this sentence can form a relative clause modifying the hole under the right conditions.

- That's the hole out of which there jumped a jackrabbit with enormous ears.

However, it cannot be used as a relative clause modifying jackrabbit. - "That's the jackrabbit that there jumped out of the hole.

In the construction, jackrabbit is constituent 3, the element being intro­duced into the narrative. If hole is relativized, it is still possible ioi jack-rabbit to be introduced into the narrative via this construction. But if jackrabbit is relativized, then it has been introduced prior to the back­ground given in the construction, which violates the pragmatic condition on the construction.

Whafs in Common? 575 Let us now turn from the details of analysis to some theoretical issues.

What's in Common?

As in the case of the deictic constructions, it is possible to find a set of con­ditions shared by all the existential constructions.

The Generalized Existential Parameters of Meaning

Same as the central existential Parameters of Form

Syntactic Elements

0: clause, with parts 1, 2, 3, and 4

1: noun phrase

3: noun phrase Lexical Conditions

1 = there Syntactic Conditions

S-1: 1 is first

S-3: 2 precedes 3

S-6: 1 is subject of 0

S-8: number (1) agrees with number (3); optional if 1 governs

contraction Phonological Conditions

1 is unstressed

This is a very strange set of conditions. There is no specification of the syntactic categories of elements 2 and 4. Since 2 is a verb phrase in the presentational construction, but a verb in all the other constructions, there is no syntactic category that can be uniformly assigned to 2 in all constructions. For the same reason, be cannot be assigned as the lexical representation of 2 in all constructions. Since 4 is a gapped infinitival clause in the infinitival construction, there can be no common value of 4 that is shared in all constructions. Similarly, the syntactic conditions that accompany 4 are not shared and so are not part of the generalized existen­tial. What the existentials have in common is so skewed a set of conditions that it is difficult to imagine that it could have any linguistic significance whatever.

Since these are necessary but not sufficient conditions, no classical account of the category of existentials is possible. It would be technically possible to give a prototype-based account with the generalized existen­tial as the prototypical construction, and all the other constructions as

576 Case Study 3

noncentral. As in the case of the similar proposal for the deictics, there are two objections. First, this would not account for the fact that central existentials are considered better examples of existentials than noncen­tral cases. Thus

- There is a boy in the room. is a better example of the category than:

- There's a boy been hurt.

- There exists a God.

- Into the room there burst a policeman.

In addition, many of the generalizations that we stated above cannot be stated if the generalized existential is taken as the prototypical construc­tion, with the other constructions based on it. Most notably, we could not use all the predictions that come from basing the central existential on the central deictic. That predicts more than is in the generalized existential, and there is no way to predict exactly what is in the generalized existen­tial. The point here is that the concept "based on" allows us to predict exactly what is in the central existential, but there is no corresponding concept that would allow us to predict exactly what is in the generalized existential. This suggests that the shared properties of the existentials— those in the generalized existential—have no cognitive status at all.

Whereas we can predict all but one property of the central existential, we can predict nothing of the generalized existential. Were we to choose it as the central member, we would miss a great deal. We would also miss a great deal in the specification of each of the subconstructions. Instead of simply listing where the individual subconstructions diverge from the cen­tral existential, we would have to list considerably more for each subcon-struction. The minimal representations would be much more complex. As in the case of the deictics, there seems to be no reason to accord any cognitive status whatever to the conditions that happen to be shared by the members of the category of f/iere-constructions.

Distinguishing Versus Defining Properties

In the classical theory, shared properties, defining properties, and distin­guishing properties are coextensive. The idea is that a category is defined by shared properties, and these distinguish the members of the given cate­gory from everything else. As we have seen, the collection of properties that incidentally happen to be shared by the constructions in the category seem to have no special cognitive status. If another subconstruction were

Distinguishing versus Defining Properties 577

added to the category, the shared properties might well change, but the category would otherwise remain intact.

We have also seen that there is no set of properties that exactly defines all and only the subconstructions in the category. The two constructional categories that we have studied are each defined by a central member and conventionalized variations on it.

Minimal distinguishing properties may, however, be found and may even be psychologically real. As their name suggests, distinguishing prop­erties are those that minimally distinguish one category from another. They are a subset of the shared properties, but they do not define the cate­gory. In the case of deictic and existential f/iere-constructions, it does seem possible for speakers to distinguish one category from the other directly, without using the prototype structure we have uncovered. My guess is that they use properties such as these:

Deictic: First element is either here or there and refers deictically to a

real or abstract location.

Existential: First element is there and does not refer to any entity or

location.

Such conditions will suffice to distinguish the deictics from the existen­tials. I have no evidence whatever that distinguishing properties play any role in the grammar of a language. It is possible that they play no role at all and that speakers, if asked, can come up with some minimal distin­guishing characteristics that will enable them to make the distinction. But in ordinary language use, people don't have to engage in the task of dis­tinguishing one category from another. Instead, they just have to pair form and meaning, and constructions of the sort we have discussed are sufficient for the task. This does not rule out the possibility that linguists will someday find that minimal distinguishing characteristics have some cognitive function in grammar. If so, something like the above may have to be added to what we have already described.

Let us suppose for the sake of discussion that there is some reason for the grammar of English to include minimal distinguishing features, like those cited above, to distinguish the deictics as a group from the existen­tials as a group. Such minimal distinguishing features would still not do very much of what has to be done to define the constructions—that is, to tell exactly what form-meaning pairings are, and are not, in the language. Such conditions would not exclude what needs to be excluded and would not get the constructional meanings and pragmatic conditions right. Nor would they account for the fine details of the syntax that differ from sub-construction to subconstruction. It might be the case that we need both

578 Case Study 3

the prototype-based analysis and minimal distinguishing features, but there is no way to do without the prototype-based characterization.

There Is Where

This case study would not be complete without a discussion of a class of sentences that look like they are deictic ^eye-constructions but aren't. I have in mind sentences like

- There is where he put the money. Such sentences may look superficially like

- There's Harry.

But they have different properties. They can be negated, questioned, embedded, and raised.

- There isn't where he put the money.

- *There isn't Harry.

- Is there where he put the money?

- *Is there Harry?

- There is believed to be where he put the money.

- *There is believed to be Harry.

- If there is where he put the money, our problems are over.

- *If there's Harry, our problems are over.

It is possible to get conjunctions of noun phrases of the form Harry and where he put the money, as in

- I keep thinking about Harry and where he put the money.

Such conjunctions, however, are impossible in deictic there-constructions. For example, if one discovered Harry getting the hidden money out of a secret cache, one could say either

- There's Harry. or

- There's where he put the money. but not

- "There's Harry and where he put the money. Nothing needs to be said about гАгге-constructions to account for these

A Short Recapitulation 579

facts. They are automatic consequences of another phenomenon in English grammar.

English has a construction called the pseudo-cleft construction. Here are some examples:

- What Harry ate was a bagel.

- What bit me was a mosquito.

- Where he put the money was under the bed.

Under certain conditions, pseudo-clefts can be inverted:

- A bagel is what Harry ate.

- A mosquito is what bit me.

- Under the bed is where he put the money.

Inverted pseudo-clefts like these are remarkable in that they allow adverbs like under the bed to function as grammatical subjects. We can see that under the bed is a subject in the last sentence above from the fact that it can raise:

- Under the bed is believed to be where he put the money,

Inverted pseudo-clefts can also be negated, questioned, and embedded with relative freedom.

- Under the bed isn't where he put the money.

- Is under the bed where he put the money?

- I doubt that under the bed is where he put the money.

- I would be surprised if under the bed were where he put the money.

Since locative adverbs like under the bed can occur in pseudo-clefts and inverted pseudo-clefts, so can deictic locatives like there.

- Where he put the money was there (pointing).

- There is where he put the money.

Although There is where he put the money looks superficially like a deictic (/tere-construction, it isn't one. It is simply an inverted pseudo-cleft, and functions exactly as one would expect.

A Short Recapitulation

We are now in a position to give a simple and elegant characterization of the f/iere-constructions. We will make use of the following notation.

For: What the construction is used for.

580 Case Study 3

About: What the construction is used to talk about. On: What other construction it is based on.

The integers denote the following syntactic elements, which are conse­quences of their semantics:

1: there (or here)

2: V or VP

3:NP

4; Final Phrase

A slash indicates "minus." And lexical items are capitalized, as are rule names. In the "based-on" statements, the central deictic and existential constructions will be abbreviated by CD and CE.

The central deictic and existential constructions can be described sim­ply as follows:

CENTRAL DEICTIC: For: Pointing OUt

2: If locational, then basic level.

4': Not inherent.

4: VP/2. [That is, a VP minus a 2.]

CENTRAL EXISTENTIAL: On: CD

V: Mental space.

This says that the central deictic construction is used for pointing out things (using the pointing-out ICM), and it gives the three properties of the construction that are not consequences of its use. The on-condition says that the central existential is based on the central deictic, and gives the single condition that differentiates the central existential from the central deictic. All the other differences are consequences.

There are nine minimal variants of the central deictic, and they can be characterized very simply. All of them are linked to the central deictic by the on-condition. First, there are five that arc about various subject matters.

perceptual deictic: On: CD. About: Perception discourse deictic: On: CD. About: Discourse existence deictic: On: CD. About: Existence activity start deictic: On: CD. About: Activity

Г: Starting point. delivery deictic: On: CD: About: Delivery

5: Deliverer

Option: 1: elongated, rising, unstressed

There are also three minimal variants that are simply intersections of the

Summary 581

central deictic and another minimal construction that pairs a pragmatic condition with an intonation. All that needs to be said is that these con­structions exist and are based on both the central deictic and an additional construction.

PARAGON DEICTIC: On: CD and PARAGON

EXASPERATION DEICTIC: On: CD and EXASPERATION

ENTHUSIASTIC BEGINNING DEICTIC: On: CD and ENTHUSIAT1C BEGINNING

There are also two special purpose constructions, one of which is a mini­mal variant and one of which is more complex.

narrative focus deictic: On: CD. For: Vivid narrative presentational deictic: On: CD. For: Announcements 2: Verb of location Options: 2 has full auxiliary 4 precedes 2 and 3 If 4' = location, then 1 is null.

The presentational deictic is the only complex variant, having one special condition and three options.

There are also three variants based on the central existential.

STRANOE EXISTENTIAL: On: CE

2: Perfect have & 's

ontological existential: On: CE. About: Existence 2: stressed Option: 2 = exist

INFINITIVAL EXISTENTIAL: On: CE

4: Gapped infinitival

PRESENTATIONAL EXISTENTIAL: On: CE

1': Narrative space 2: Intransitive VP 2': No restriction

Given the enormous complexity of the data, it is hard to see how a more elegant account could be given that covers the same range of syntactic, semantic, and pragmatic phenomena, while also showing the relation­ships among the constructions.

Summary

This case study has had both a narrow and a broad purpose. The narrow purpose was to provide an adequate description of both deictic and exis-

582 Case Study 3

tential f/ierc-constructions and to show how they are related. The broad purpose was to answer a number of deep and important questions about the nature of language. Here are the answers that have emerged from our study:

- Cognitive semantics is necessary for the description of the meanings of grammatical constructions. This includes mental spaces and cogni­tive models of many sorts, including metaphoric and metonymic models.

- Grammatical constructions have a real cognitive status. They are not mere epiphenomena arising from the operation of generative rules.

- Prototype-based categorization occurs in grammar. Radially struc­tured categories exist there, and their function is to greatly reduce the arbitrariness of form-meaning correlations.

- The concept of motivation is needed in order to account for a great many of the regularities that occur in grammar.

- Syntactic categories are not autonomous, nor are they completely predictable from semantic considerations. Instead, their central sub-categories are predictable from semantic considerations, and their noncentral subcatcgories are motivated extensions of central subcate-gories.

- A great many syntactic properties of grammatical constructions are consequences of their meanings.

- The meanings of whole grammatical constructions are not com­putable by general rules from the meanings of their parts. They are, however, motivated by the meanings of their parts.

- There is a continuum between the grammar and the lexicon.

- Grammars are not separate "modules" independent of the rest of cog­nition. The reason is that they make use of prototype categorization, which arises in other aspects of cognition, and they also make use of various aspects of conceptual systems, such as cognitive models (including metaphoric and metonymic models) and mental spaces.

Generative Semantics Updated

Cognitive grammar, as I have presented it, has developed gradually over a number of years, evolving from generative semantics and case grammar through the theory of linguistic gestalts to its present form. In recent years, I have been working on the theory of grammatical constructions with Berkeley colleagues Charles Fillmore and Paul Kay. I view cognitive grammar as an updated version of generative semantics.

Generative Semantics Updated 583

Certain basic principles of cognitive grammar were also basic princi­ples of generative semantics:

- Language is part of general cognition and makes use of general cogni­tive mechanisms.

- The primary function of language is to convey meaning. A grammar should therefore show as directly as possible how parameters of form are linked to parameters of meaning.

- Since meaning and communicative function are primary, grammars should attempt to explain as much as possible about parameters of form on the basis of parameters of meaning and communicative function.

- Pragmatics is taken to be the semantics of communication, and the same theoretical apparatus is used in the description of both domains.

The differences between generative semantics and cognitive grammar are a matter of implementation rather than a matter of basic assumptions. They have come about because our understanding of the basic mecha­nisms involved in cognition has increased considerably in recent years. Here are the principal areas of difference for semantics.

- Generative semantics assumed that model-theoretical approaches to logic could be adapted to account for semantics and pragmatics.

- Cognitive grammar recognizes that model theory will not account for most semantic and pragmatic phenomena. It hypothesizes that a theory of experientially grounded cognitive models can adequately account for semantic and pragmatic phenomena.

And here are the principal areas of difference for syntax:

- Generative semantics assumed that the classical arguments given for transformational grammar were essentially correct. It attempted to link form and meaning by taking logical forms as underlying syntactic structures. Global rules served the function of directly linking logical form and surface syntax and were viewed as preferable to transforma­tions (which largely got in the way of direct form-meaning correspon­dences).

- Cognitive grammar recognizes recent results showing that trans­formational rules are not necessary. It also recognizes the empirical failure of generative theories of grammar to account for most gram­matical phenomena. Grammatical construction theory proposes in­stead that symbolic models—direct form-meaning pairings—are basic

584 Case Study 3

elements of grammar. Grammatical constructions are organized via prototype theory, using radially structured constructional categories.

Generative semantics succeeded in a great many ways:

- Generative semantics got linguists to pay attention to the role of logi­cal form and model theory.

- Generative semantics showed how predicate-argument structure, logical operators, coreference, binding, propositonal functions, etc., played a role in such syntactic phenomena as negative polarity, anaphora, constraints on the occurrence of quantifiers, etc.

- Generative semantics showed how the internal structure of lexical items had syntactic consequences.

- Generative semantics got the linguistic community to pay attention to the relationship between syntax and pragmatics, especially phenom­ena involving performatives, presuppositions, implicatures, eviden-tials, etc.

- Generative semantics brought to the attention of the linguistic com­munity an enormous number of syntactic and semantic phenomena.

Primarily as a result of the work of generative semanticists, such matters

are now part of mainstream syntactic and semantic theories.

But generative semantics also failed as an empirical theory in many ways:

- Because it used transformational syntax, it could not deal adequately with syntactic amalgams (Lakoff 1974).

- Because it depended on a theory of classical syntactic categories, it could not deal adequately with prototype effects in syntax and with the semantic basis of syntactic categories.

- Because it used a generative grammar, it could not account for moti­vated grammatical constructions, which are not fully predictable by generative rules. (See Fillmore, Kay, and O'Connor, to appear.)

- Because the lexicon only had redundancy rules available to it, but not radial categories, it could not deal adequately with polysemy (Brug-man 1981, Dixon 1968, Lindner 1981).

- Because it had no account of the spatialization of form, it could not deal with syntactic iconicity (Lakoff and Johnson 1980, chap. 20, Haiman 1980).

- Because it used an objectivist semantics, that is, a semantics contain­ing logical forms and model theory, it could not deal adequately with:

(a) basic-level categories and prototype effects in semantics (fc) Fillmore's frame-semantic phenomena

Generative Semantics Updated 585

(c) the Lakoff-Johnson metaphor phenomena

(d) the Borkin-Nunberg-Fauconnier metonymy phenomena (see Fauconnier 1985)

(e) the Lindner-Brugman-Janda image-schema phenomena (/) alternative conceptual systems (Casad 1982, Brugman 1983) (g) inconsistent cognitive models (Kay 1979, Gentner and Gentner

1982)

(и) such mental space phenomena as referential opacity and pre­supposition (Fauconnier 1985)

(f) the semantics of classifier systems (Dixon 1968, Downing 1984) (/) metalinguistic negation (Fillmore 1984, Horn 1985) (k) the inconsistency in the theory of meaning pointed out by Putnam (1981)

It should be pointed out that it was not only generative semantics that failed in these ways. Every theory that makes use of a generative syntax and either model-theoretic semantics or a semantics that involves only logical form has failed in the same ways. The appearance of failure has been avoided only by ignoring such phenomena. Luckily, a significant number of linguists have paid attention to them, and the result is cogni­tive linguistics.

Afterword

I began work on this book with the knowledge that objectivist views of the mind have a very wide currency in the academic world. Among my principal aims has been to characterize that view, name it, point out that it is an opinion, not a fundamental truth, and raise the question of its va­lidity, so that it can be discussed in the open and no longer be presup­posed automatically as part of an unquestioned background.

Many scholars do indeed take it for granted without question that con­ceptual categories are denned solely by the shared essential properties of their members, that thought is the disembodied manipulation of abstract symbols, and that those symbols get their meaning solely by virtue of cor­respondences to things in the world. The view of reason as abstract, dis­embodied, and literal is well-established.

It is well-established partly because there are certain aspects of it that are correct. Some conceptual categories do have the structure of a classi­cal category. Certain aspects of classical logic do seem to be used in reason. But it does not follow from that that the whole objectivist para­digm must be accepted. Another of my goals has been to provide an alter­native to objectivist views that preserves what is right about objectivism, while permitting research into the nature of reason that goes beyond the limits that objectivism had placed on that research.

What I could not have guessed when I began this book is just how many researchers have found evidence of one sort or another that contradicts objectivist views. One of the most gratifying aspects of this research has been the discovery that there is a large body of evidence from many fields that supports a view of mind that is centered in the bodily and imaginative capacities of human beings.

During the course of the writing, I had an opportunity to lecture on these topics not only at universities throughout the United States, but in many other countries, particularly in Europe, Asia, and South America. Everywhere I have gone, I have encountered researchers whose findings

586

Afterword 587

are, in one way or another, inconsistent with the objectivist view of mind, but who had no idea that so many others had reached similar conclusions. Such meetings have been important to me. They are like meetings be­tween long-lost relatives, eager to share news of the family. As long as this book is, I have barely begun to follow up all the leads I have been given. My initial aim was to bring together whatever results I could find that supported an experientialist view of mind. I have come to the conclu­sion that it would take many lifetimes to do that. I am stopping here with the knowledge that there are many others involved in the enterprise.

The ideas I have put forth are, thus, by no means uniquely my own. Most of them are shared by a diverse group of scholars whose research has led them to question traditional views of reason, categorization, and language, but who have not yet formed themselves into a community. It is important to develop such a sense of community, and I hope that this book can contribute to it.

References

Adams, Karen, and Nancy Faires Conklin. 1973. Toward a Theory of Natural Classifica­tion. In Proceedings of the Ninth Regional Meeting, Chicago Linguistic Society, pp. 1-10. Chicago: Chicago Linguistic Society.

Aissen, Judith. 1975. Presentational There-Insertion: A Cyclic Root Transformation. In Papers from the Eleventh Regional Meeting, Chicago Linguistic Society, pp. 1-14. Chicago: Chicago Linguistic Society.

Akmajian, Adrian. 1984. Sentence Types and the Form-Function Fit. Natural Language and Linguistic Theory 2, no. 1, 1-24.

Armstrong, Sharon Lee, Lila Gleitman, and Henry Gleitman. 1983. What Some Con­cepts Might Not Be. Cognition 13:263-308.

Amheim, Rudolf. 1974. Visual Thinking. Berkeley: University of California Press.

Austin, John L. 1961. Philosophical Papers. Oxford: Oxford University Press.

Barsalou, Lawrence W. 1983. Ad-hoc categories. Memory and Cognition 11:211-27.

———. 1984. Determinants of Graded Structure in Categories. Psychology Department,

Emory University, Atlanta.

Barwise, Jon. 1980. Scenes and Other Situations. In Barwise and Sag, 1980. Barwise, Jon, and John Perry. 1980. The Situation Underground. In Barwise and Sag,

1980.

———. 1984. Situations and Attitudes. Cambridge, Mass.: MIT Press.

Barwise, Jon, and Ivan Sag, eds. 1980. Stanford Working Papers in Semantics, vol. 1. Stanford University.

Bates, Elizabeth, and Brian MacWhinney. 1982. Functionalist Approaches to Grammar. In L. Gleitman and E. Wanner, eds.. Language Acquisition: The State of the Art, pp. 173-218, Cambridge: Cambridge University Press.

Backer, Alton L. 1975. A Linguistic Image of Nature: The Burmese Numerative Classi­fier System. Linguistics, Whole 165, pp. 109-21.

Beneke, Timothy. 1982. Men on Rape. New York: St. Martin's Press.

Bennett, David. 1975. Spatial and Temporal Uses of English Prepositions. London:

Longmans.

Berlin, Brent. 1968. Tzeltal Numeral Classifiers. The Hague: Mouton.

———. 1976. The Concept of Rank in Ethnobiological Classification: Some Evidence

from Aguaruna Folk Botany. In Casson, 1981, pp. 92-113. Berlin, Brent, Dennis E. Breedlove, and Peter H. Raven. 1974. Principles ofTzeltal Plant

Classification. New York: Academic.

589

590 References

Berlin, Brent and Paul Kay. 1969. Basic Color Terms: Their Universality and Evolution. Berkeley: University of California Press.

Berlin, Brent, and A. Kimball Romney. 1964. Descriptive Semantics of Tzeltal Numeral Classifiers. In A. K. Romney and R. D'Andrade, eds., Transcultural Studies in Cogni­tion, pp. 79-98. Special publication of The American Anthropologist 66, no. 3, part 2.

Bernstein. Richard. 1983. Beyond Objectivism and Relativism. Philadelphia: University of Pennsylvania Press.

Bolinger, Dwight. 1972. Degree Words. The Hague: Mouton.

———. 1976. Language and Memory. Language Sciences.

———. 1977. Meaning and Form. New York: Longmans.

———. 1982. Intonation and Gesture. In Papers from the Parasession on Nondeclara-

tives, pp. 1-22. Chicago: Chicago Linguistic Society. Borges, J. L. 1966. Other Inquisitions. New York: Washington Square Press. Breivik, Leif Egil. 1975. The Use and Non-use of Existential There in Present-Day

English. Forum Linguisticum 7:57-103. Brown, Roger. 1958. How Shall a Thing Be Called? Psychological Review 65:14-21.

———. 1965. Soci'a/ Psychology. New York: Free Press. Brugman, Claudia. 1981. Story of Over. M. A. thesis. University of California, Berkeley. Available from the Indiana University Linguistics Club.

———. 1983. The Use of Body-Part Terms as Locatives in Chalcatongo Mixtec. In Report No. 4 of the Survey of California and Other Indian Languages, pp. 235-90. University of California, Berkeley.

———. 1984. Metaphor in the Elaboration of Grammatical Categories in Mixtec. Lin­guistics Department, University of California, Berkeley.

Bargess, Don, Willett Kempton, and Robert MacLaury. 1983. Tarahumara Color Modifiers: Category Structure Presaging Evolutionary Change. American Ethnolo­gist, no. 1, 133-49.

Bybee, Joan, and Carol Moder. 1983. Morphological Classes as Natural Categories. Language 59, no. 2, 251-70.

Cain, A. J. 1958. Logic and Memory in Linnaeus's System of Taxonomy. Proceedings of the Linnaean Society of London 169:144-63.

Carbonell, Jaime. In press. Towards a Computational Model of Metaphor in Common-sense Reasoning. In Holland and Quinn, in press.

Garden, Guy. 1978. TTiere-insertion Makes a Mistake: A Global Rule and the Validation of Introspective Judgments. Linguistics Department, University of British Columbia, Vancouver.

Carpenter, P. A., and P. Eisenberg. 1978. Mental Rotation and the Frame of Reference in Blind and Sighted Individuals. Perception and Psychophysics 23:117-24.

Casad, Eugene. 1982. Cora Locationals and Structured Imagery. Ph.D. diss., University of California, San Diego.

Casad, Eugene, and Ronald Langacker. 1985. "Inside" and "Outside" in Cora Grammar. International Journal of American Linguistics 51.3:247-81.

Casson, Ronald, ed. 1981. Language, Culture, and Cognition: Anthropological Perspec­tives. New York: Macmillan.

Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton.

———. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press.

————. 1981. Lectures cm Government and Binding. Dordrecht: Foris. dark, Eve V., and Herbert Clark. 1978. When Nouns Surface as Verbs. Language 55, no. 4, 767-“11.

References 591

Coleman, Linda. 1975. The Case of the Vanishing Presupposition. In Proceedings of the

First Annual Meeting of the Berkeley Linguistics Society, pp. 78-89. Berkeley: Berkeley

Linguistics Society. Coleman, Linda, and Paul Kay. 1981. Prototype Semantics: The English Verb Lie.

Language 57, no. 1, 26-44. Craig, Colette, ed. 1986. Categorization and Noun Classification. Philadelphia: Benjamins

North America. Cruse, D. A. 1977. The Pragmatics of Lexical Specificity. Journal of linguistics

13:153-64. Davidson, Donald, and Gilbert Harman, eds. 1972. Semantics of Natural Language.

Dordrecht: Reidel. De Sousa, Ronald. 1980. The Rationality of Emotions. In A. 0. Rorty, cd.. Explaining

Emotions. Berkeley and Los Angeles: University of California Press. DeValois, R. L., I. Abramov, and G. H. Jacobs. 1966. Analysis of Response Patterns of

LGN Cells. Journal of the Optical Society of America 56:966-77. DeValois, R. L., and G. H. Jacobs. 1968. Primate Color Vision. Science 162:533-^0. Denny, J. Peter. 1976. What are Noun Classifiers Good For? In Papers from the Twelfth

Regional Meeting, Chicago Linguistic Society, pp. 122-32. Chicago: Chicago Linguistic

Society.

Dixon, R. M. W. 1968. Noun Classes. Lingua 21:104-25. ———. 1977. Semantic neutralization for phonological reasons. Linguistic Inquiry

8:599-602.

-. 1982. Where Have All the Adjectives Gone? Berlin: Walter de Gruyter.

Dobzhansky, T. 1955. Evolution, Genetics, and Man. New York; John Wiley & Sons. Dougherty, J. W. D. 1978. Salience and Relativity in Classification. American Ethnologist

15:66-80. Reprinted in Casson, 1981.

Downing, Pamela. 1977. On the Creation and Use of English Compound Nouns. Lan­guage 53, no. 4, 810-42.

———. 1984. Japanese Numeral Classifiers: Syntax, Semantics, and Pragmatics. Ph.D.

diss.. University of California, Berkeley. Dubois, D., and H. Prade. 1980. Fuzzy Sets and Systems: Theory and Applications. New

York: Academic Press. Ekiof, P. C. 1976. Whitehead's Problem Is Undeddable. American Mathematical

Monthly 83, no. 10, 775-87. Ekman, Paul. 1971. Universals and Cultural Differences in Facial Expressions of

Emotions. Nebraska Symposium on Motivation series, James K. Cole, ed. Lincoln:

University of Nebraska Press. Ekman, Paul, Wallace V. Friesen, and P. Ellsworth. 1972. Emotion in the Human Face.

Elmsford, N.Y: Pergamon Press. Ekman, Paul, Robert W. Levenson, and Wallace V. Friesen. 1983. Autonomic Nervous

System Activity Distinguishes among Emotions. Science 221:1208-10. Emonds, Joseph. 1971. Root and Structure-Preserving Transformations. Ph.D. diss.,

MIT, Cambridge.

Fauconnier, Gilles. 1985. Mental Spaces. Cambridge, Mass.: MIT Press. Feyerabend, Paul. 1975. Against Method: Outline of an Anarchistic Theory of Know­ledge. London: NLB. Fillmore, Charles. 1968. The Case for Case. In E. Bach and R. Harms, eds., Universals in

Linguistic Theory, pp. 1-90. New York; Holt, Rinehart & Winston.

———. 1975. An Alternative to Checklist Theories of Meaning. In Proceedings oftht

592 References

First Annual Meeting of the Berkeley Linguistics Society, pp. 123-31. Berkeley: Berke­ley Linguistics Society.

-. 1976. Topics in Lexical Semantics. In Peter Cole, ed., Current Issues in Linguistic

Theory, pp. 76-138. Bloomington: Indiana University Press

——. 1978. The Organization of Semantic Information in the Lexicon. In Papers from the Parasession on the Lexicon, pp. 1-11. Chicago: Chicago Linguistic Society.

. 1982a. Towards a Descriptive Framework for Spatial Deixis. In R. J. Jarvella and

W. Klein, eds.. Speech, Place, and Action, pp. 31-59- London: John Wiley.

1982&. Frame Semantics. In Linguistic Society of Korea, ed.. Linguistics in the

Morning Calm, pp. 111-38. Seoul: Hanshin.

1985. Frames and the Semantics of Understanding. Quaderni di Semantica 6,

no. 2, 222-53.

Fillmore, Charles, Paul Kay, and Mary Catherine O'Connor. To appear. Regularity and Idiomaticity in Grammar: The Case of Let Alone. Department of Linguistics, Univer­sity of California, Berkeley.

Fodor, J. A. 1975. The Language of Thought. New York: Crowcll. Frege, Gottlob. 1966. On Sense and Reference. In P. Geach and M. Black, eds., Transla­tion from the Philosophical Writings of Gottlob Frege. Oxford: Blackwell.

Gazdar, Gerald, Ewan Klein, Geoffrey Pullum, and Ivan Sag. 1985. Generalized Phrase-Structure Grammar. Cambridge, Mass.: Harvard University Press.

Geertz, Clifford. 1973. The Interpretation of Cultures. New York: Basic Books.

Gensler, Orin. 1977. Non-syntactic Antecedents and Frame Semantics. In Proceedings of the Third. Annual Meeting of the Berkeley Linguistics Society, pp. 321-34. Berkeley:

Berkeley Linguistics Society.

Centner, Dedre. 1981. Generative Analogies as Mental Models. In Proceedings of the Third Annual Conference of the Cognitive Science Society, pp. 97-100. Berkeley: Cogni­tive Science Society.

Gentner, Dedre, and Donald R. Centner. 1982. Flowing Waters or Teeming Crowds:

Mental Models of Electricity. In D. Gentner and A. L. Stevens, eds.. Mental Models. Hillsdale, N.J.: Eribaum.

Gibson, James J. 1979. The Ecological Approach to Visual Perception. Boston: Houghton Mifflin.

Gilchrist, Alan, and Irvin Rock. 1981. Rational Processes in Perception. In Proceedings of the Third Annual Conference of the Cognitive Science Society, pp. 50-56. Berkeley:

Cognitive Science Society.

Gleitman, Lila. 1981. What Some Concepts Might Not Be. Address to Annual Meeting of the Jean Piaget Society. Psychology Department, University of Pennsylvania, Phila­delphia.

Goguen, J. A. 1969. The Logic of Inexact Concepts. Synthese 19:325-73.

Goodman, Nelson. 1978. Ways of Worldmaking. Indianapolis: Hackett.

Gould, Stephen Jay. 1983. Hen's Teeth and Horse's Toes. New York: Norton.

Green, Georgia. 1976. Main Clause Phenomena in Subordinate Clauses. Language 52, no. 2, 382-97.

Hacking, lan. 1983. Representing and Intervening: Introductory Topics in the Philosophy of Natural Science. Cambridge: Cambridge University Press.

Haiman, John. 1980. The Iconicity of Grammar: Isomorphism and Motivation. Language 56, no. 3, 515^10.

Hanson, Norwood R. 1961. Patterns of Discovery: An Inquiry into the Conceptual Foundations of Science. Cambridge: Cambridge University Press.

References 593

Harris, Zeilig. 1957. Co-occurrence and Transformation in Linguistic Structure. Lan­guage 33, no. 2, 293-340.

Hawkins, Bruce. 1984. The Semantics of English Spatial Prepositions. Ph.D. diss.. Uni­versity of California, San Diego.

Heider, Eleanor (Eleanor Rosch). 1971. "Focal" Color Areas and the Development of Color Names. Developmental Psychology 4:447-55.

———. 1972. Universals in Color Naming and Memory. Journal of Experimental Psy­chology 93:10-20. Heider, Eleanor, and D. Olivier. 1972. The Structure of the Color Space in Naming and

Memory for Two Languages. Cognitive Psychology 3:337-54. Hesse, Mary. 1963. Models and Analogies in Science. London: Sheed & Ward. Hinton, Leanne. 1982. How to Cause in Mixtec. In Proceedings of the Eighth Annual

Meeting of the Berkeley Linguistics Society, pp. 354-63. Berkeley: Berkeley Linguistics

Society. Holland, Dorothy, and Naomi Quinn, eds. In press. Cultural Models in Language and

Thought. Cambridge: Cambridge University Press. Hooper, Joan, and Sandra Thompson. 1973. On the Applicability of Root Transforma­tions. Linguistic Inquiry 4:465-98. Hopper, Paul, and Sandra Thompson. 1984. The Discourse Basis for Lexical Categories

in Universal Grammar. Language 60, no. 4, 703-52. Horn, Laurence. 1985. Metalinguistic Negation and Pragmatic Ambiguity. Language 61,

no. 1, 121-74. Hull, David L. 1984. Contemporary Systematic Philosophies. In Sober, 1984, pp.

567-602. Hunn, Eugene S. 1975. A Measure of the Degree of Correspondence of Folk to Scientific

Biological Classification. American Ethnologist 2, no. 2, 309-27.

———. 1977. Tzeltal Folk Zoology: The Classification of Discontinuities in Nature. New

York: Academic Press. Jaeger, Jeri. 1980. Categorization in Phonology: An Experimental Approach. Ph.D.

diss., University of California, Berkeley. Janda, Laura. 1984. A Semantic Analysis of the Russian Verbal Prefixes ZA-, PERE-,

DO-, and ОТ-. Ph.D. diss., University of California, Los Angeles.

———. 1986. Band 192. In Slavistische Beiirage. Milnchen: Veriag Otto. Jespersen, Otto. 1909. A Modem English Grammar on Historical Principles. 1 vols.

London: Alien & Unwin. Reprinted, 1954, by Barnes & Noble, New York. Johansson, Gunnar. 1950. Configurations in Event Perception. Uppsala: Almkvist &

Viksell. Johnson, Mark. 1981. The Preconceptual Basis of Experiential Metaphor. Department of

Philosophy, Southern Illinois University, Carbondale.

———. 1983. What Rule Theories of Morality Can't, but Ought to, Do. Department of Philosophy, Southern Illinois University, Carbondale.

-. In press. The Body in the Mind: The Bodily Basis of Reason and Imagination.

Chicago: University of Chicago Press. Katz, Jerrold, and Paul Postal. 1964. An Integrated Theory of Linguistic Descriptions.

Cambridge: МГТ Press. Kay, Paul. 1979. The Role of Cognitive Schemata in Word Meaning: Hedges Revisited.

Department of Linguistics, University of California, Berkeley.

-. 1983я. Three Properties of the Ideal Reader. Cognitive Science Report, no. 7.

Institute for Cognitive Studies, University of California, Berkeley.

594 References

———. 19836. Linguistic Competence and Folk Theories of Language: Two English

Hedges. In Proceedings of She Ninth Annual Meeting of the Berkeley Linguistics

Society, pp. 128-37. Berkeley: Berkeley Linguistics Society. Kay, Paul, and Willett Kempton. 1984. What Is the Sapir-Whorf Hypothesis? American

Anthropologist 86, no. 1, 65-79. Kay, Paul, and Chad McDaniel. 1978. The Linguistic Significance of the Meanings of

Basic Color Terms. Language 54, no. 3, 610-46. Keenan, E. L., and A. Faltz. To appear. Logical Types for Natural Language. Synthese

Language Library Series. Dordrecht: D. Reidel. Kempton, Willett. 1981. The Folk Classification of Ceramics: A Study of Cognitive Proto­types. New York: Academic Press. Ken-, Nancy H. 1983. The Role of Vision in "Visual Imagery" Experiments: Evidence

from the Congenitally Blind. Journal of Experimental Psychology: General 112, no. 2,

265-77. Kimball, John. 1973. The Grammar of Existence. In Papers from the Ninth Regional

Meeting, Chicago Linguistic Society, pp. 262-70. Chicago: Chicago Linguistic Society. Kleene, Stephen. 1967. Mathematical Logic. New York; John Wiley & Sons. Kosslyn, Stephen M. 1980. Image and Mind. Cambridge, Mass.: Harvard University

Press.

———. 1983. Ghosts in the Mind's Machine. New York: W. W. Norton. Kripkc, Saul. 1972. Naming and Necessity. In Davidson and Harman, 1972, pp. 253-355. Kuhn, Thomas. 1970. The Structure of Scientific Revolutions. 2d ed. Chicago: University of Chicago Press.

———. 1977. The Essential Tension. Chicago: University of Chicago Press.

Kuno, Susumu. 1971. The Position of Locatives in Existential Sentences. Linguistic

Inquiry 2, no. 3, 333-78.

Labov, William. 1973. The Boundaries of Words and Their Meanings. In Joshua Fish-man, ed., New Ways of Analyzing Variation in English, pp. 340-73. Washington, D.C.:

Georgetown University Press.

Lakatos, Imre. 1976. Proofs and Refutations. Cambridge: Cambridge University Press. Lakoff, George. 1963. Toward Generative Semantics. Syntax and Semantics, vol. 7, ed. James D. McCawley. New York: Academic Press, 1976.

———. 1968. Counterparts, or the Problems of Reference in Transformational Gram­mar. Department of Linguistics, Harvard University. ___. 1970. Linguistics and Natural Logic. Studies in Generative Semantics, no. 2. Ann

Arbor: University of Michigan Linguistics Department. Reprinted in Davidson and

Hannan, 1972, pp. 545-665.

, 1972. Hedges: A Study in Meaning Criteria and the Logic of Fuzzy Concepts. In

Papers from the Eighth Regional Meeting, Chicago Linguistic Society, pp. 183-228. Chicago: Chicago Linguistic Society. Reprinted in Journal of Philosophical Logic 2

(1973); 458-508.

1974. Syntactic Amalgams. In Papers from the Tenth Regional Meeting, Chicago

Linguistic Society, pp. 321-34. Chicago: Chicago Linguistic Society. ——. 1977. Linguistic Gestalts. In Papers from the Thirteenth Regional Meeting, Chicago Linguistic Society, pp. 236-87. Chicago: Chicago Linguistic Society.

. 1980. Getting the Whole Picture: The Role of Mental Images in Semantics and

Pragmatics. In Proceedings of the Sixth Annual Meeting of the Berkeley Linguistics Society, pp. 191-95. Berkeley: Berkeley Linguistics Society.

References 595

-. 1982a. Experiential Factors in Linguistics. In T. Simon and R. Scholcs, eds., Lan­guage, Mind, and Brain, pp. 145-57. Hillsdale, N.J.: Lawrence Eribaum.

-. 1982&. Categories and Cognitive Models. Cognitive Science Report, no. 2, Institute

for Cognitive Studies, University of California, Berkeley.

1984. Classifiers as a Reflection of Mind. Cognitive Science Report, no. 19. Insti­

tute for Cognitive Studies, University of California, Berkeley. Lakoff, George, and Mark Johnson, 1980. Metaphors We Live By. Chicago: University of

Chicago Press. Langacker. Ronald. 1981. The Nature of Grammatical Valence. Linguistic Notes from La

Jolla 10:33-59. ———. 1982. Space Grammar, Analysability, and the English Passive. Language 58, no.

1, 22-80.

. 1983, Remarks on English Aspect. In P. Hopper, ed.. Tense and Aspect: Between

Semantics and Pragmatics, pp. 265-304. Amsterdam: John Benjamins.

-. 1986. Foundations of Cognitive Grammar, vol. 1. Stanford: Stanford University

Press.

Levy, Robert 1.1973. Tahitians: Mind and Experience in the Society Islands. Chicago: Uni­versity of Chicago Press.

———. 1984. Emotion, Knowing, and Culture. In Richard Shweder and Robert A. LeVinc, eds.. Culture Theory: Essays on Mind, Self, and Emotion, pp. 214-37. Cam­bridge: Cambridge University Press.

Lewis, David. 1972. General Semantics. In Davidson and Hannan 1972, pp. 169-218.

———. 1984. Putnam's Paradox. Australasian Journal of Philosophy 62, no. 3, 221-36. Lewis, Marshall. 1980. Joy in Mudville: A Syncretic Account of There-sentences. Depart­ment of Linguistics, Indiana University, Bloomington. Lindner, Susan. 1981. A Lexico-Semantic Analysis of Verb-Particle Constructions with

Up and Out. Ph.D. diss.. University of California, San Diego. Available from the

Indiana University Linguistics Club.

———. 1982. What Goes Up Doesn't Necessarily Come Down: The Ins and Outs of Opposites. In Papers from the Eighteenth Regional Meeting, Chicago Linguistic Society, pp. 305-23. Chicago: Chicago Linguistic Society.

Lounsbury, Floyd. 1964. A Formal Account of the Crow- and Omaha-Type Kinship Terminologies. In W. H. Goodenough, ed., Explorations in Cultural Anthropology, pp. 351-94. New York: McGraw-Hill. Reprinted in Stephen A. Tyier, ed.. Cognitive Anthropology, pp. 212-54. New York: Holt, Rinehart & Winston, 1969.

Lyons, John. 1968. A Note on Possessive, Existential and Locative Sentences. Founda­tions of Language 3:390-96.

McCawley, James D. 1981. Everything That Linguists Have Always Wanted to Know about Logic—But Were Ashamed to Ask. Chicago: University of Chicago Press. -. N.d. A Selection of There-insertion Verbs. Department of Linguistics, University

of Chicago.

McCawley, Noriko A. 1973. Boy, Is Syntax Easy! In Papers from the Ninth Annual Meet­ing, Chicago Linguistic Society, pp. 369-77. Chicago: Chicago Linguistic Society.

McCune, Keith Michael. 1983. The Internal Structure of Indonesian Roots. Ph.D. diss., University of Michigan.

Mac Lane, Saundcrs. 1981. Mathematical Models: A Sketch for the Philosophy of Mathe­matics. American Mathematical Monthly, Aug.-Sept., pp. 462-72.

. To appear. Mathematics: Form and Function. Mathematics Department, Univer­

sity of Chicago.

596 References

MaeLaury, Robert. In preparation. Color Categories in Mesoamerica: A Cross-Linguistic Study. Ph.D. diss., University of California, Berkeley.

Malotki, Ekkehart. 1983. Hopi Time. Berlin: Mouton.

Marmor, Gloria Strauss, and Larry A. Zaback. 1976. Mental Rotation by the Blind: Does Mental Rotation Depend on Visual Imagery? Journal of Experimental Psychology;

Human Perception and Performance 2, no. 4, 515-21.

Martin, D. A. and R. M. Solovay. 1970. Internal Cohen Extensions, Annals of Mathe­matical Logic 2:143-78.

Mayr, Ernst. 1963. Animal Species and Evolution. Cambridge, Mass.: Belknap Press.

———. 1984e. Species Concepts and Their Applications. In Sober, 1984, pp. 531-40.

———. 1984&. Biological Classification: Toward a Synthesis of Opposing Methodologies.

In Sober, 1984, pp. 646-62. Merrffl, G. H. 1980. The Model-Theoretic Argument against Realism. Philosophy of

Science. 47:69-81. Mervis, Carolyn. 1984. Early Lexical Development: The Contributions of Mother and

Child. In C. Sophian, ed.. Origins of Cognitive Skills. Hillsdale, N.J.: Lawrence

Eribaum Associates.

———. 1986. Child-basic Object Categories and Early Lexical Development. In U.

Neisser, ed.. Concepts Reconsidered: The Ecological and Intellectual Bases of Catego­rization. New York: Cambridge University Press. Mervis, Carolyn, and Eleanor Rosch. 1981. Categorization of Natural Objects. Annual

Review of Psychology 32:89-115. Miller, George, and Philip Johnson-Laird. 1976. Language and Perception. Cambridge,

Mass.: Harvard University Press. Milsark, Gary. 1974. Existential Sentences in English. Ph.D. diss., MIT. Available from

the Indiana University Linguistics Club. Minsky, Marvin. 1975. A Framework for Representing Knowledge. In P. H. Winston,

ed.. The Psychology of Computer Vision. New York: McGraw-Hill. Murphy, Gregory L. 1982. Cue Validity and Levels of Categorization. Psychological

Bulletin 91, no. 1, 174-77. Murphy, Gregory L., and Douglas L. Medin. 1984. The Role of Theories in Conceptual

Coherence. Psychology Department, Brown University, Providence, Rhode Island. Nathan, Geoffrey S. 1981. What's These Farts About? Linguistic Inquiry 12:151-53. Newport, Elissa, and Ursula Bellugi. 1978. A Flower Is a Flower Is a Flower. In Rosch

and Lloyd, 1978, pp. 49-72. Osherson, Daniel, and Edward Smith. 1981. On the Adequacy of Prototype Theory as a

Theory of Concepts. Cognition 9, no. 1, 35-58.

Putnam, Hilary. 1975a. Mind, Language, and Reality. Philosophical Papers, vol. 2. Cam­bridge: Cambridge University Press.

———. 1975b. The Meaning of "Meaning." In K.. Gunderson, ed.. Language, Mind, and Knowledge. Minnesota Studies in the Philosophy of Science, 7. Minneapolis: Uni­versity of Minnesota Press. Reprinted in Putnam, 1975a, pp. 215-71.

————. 1978. Meaning and the Moral Sciences. London: Routledge & Kegan Paul.

———. 1980. Models and Reality. Journal of Symbolic Logic 45:464-82.

———. 1981. Reason, Truth, and History. Cambridge: Cambridge University Press. Pylyshyn, Zenon. 1981. The Imagery Debate: Analogue Media versus Tacit Knowledge.

Psychological Review 87:16-45. Quine, W. V. 0. 1939. A Logistical Approach to the Ontological Problem. In Quine,

1966, pp. 64-69.

References 597

. 1951. On Carnap's Views on Ontology. In Quine, 1966, pp. 126-34. •. 1966. The Ways of Paradox and Other Essays. New York: Random House.

-- • • • • " • " - - Ч-... 17„..l. „„,1 Г „„^„r,. Г-п1„г

-. 1969. Ontological Relativity and Other Essays. New York and London: Columbia

University Press.

————. 1983. Ontology and Ideology Revisited. Journal of Philosophy 130:499-502. Quinn. Naomi. In press. Understanding the Experience of Marriage in Our Culture. In

Holland and Quinn, in press.

Quirk, Randolph, Geoffrey Leech, Sidney Greenbaum, and Jan Svartvik-1974. A Gram­mar of Contemporary English. London: Longmans.

Rando, Emily, and Donna Jo Napoli. 1978. Definites in There-Sentences. Language 54,

no. 2, 300-313.

Reddy, Michael. 1979. The Conduit Metaphor. In A. Ortony, ed.. Metaphor and Thought, pp. 284-324. Cambridge; Cambridge University Press.

Rey, Georges. 1983. Concepts and Stereotypes. Cognition 15:237-62.

Rhodes, Richard. 1977. Semantics in Relational Grammar. Proceedings of the Thirteenth Annual Meeting of the Chicago Linguistic Society. Chicago: Chicago Linguistic Society.

Rhodes, Richard, and John Lawler. 1981. Athematic Metaphors. Papers from the Seven­teenth Regional Meeting, Chicago Linguistic Society. Chicago: Chicago Linguistic

Society. Rips, Lance J. 1975. Inductive Judgments about Natural Categories. Journal of Verbal

Learning and Verbal Behavior 14:665-81. Rosaldo, Michelle. 1980. Knowledge and Passion. Cambridge: Cambridge University

Press. Rosch, Eleanor (Eleanor Heider). 1973. Natural Categories. Cognitive Psychology

4:328-50.

———, 1975я. Cognitive Reference Points. Cognitive Psychology 7:532-47.

——. 1975ft. Cognitive Representations of Semantic Categories. Journal of Experi­mental Psychology: General 104:192-233.

——, 1977. Human Categorization. In N. Warren, ed., Studies in Cross-Cultural Psy­chology. London: Academic.