12.3.2 The construction of Roget’s Thesaurus

Roget’s Thesaurus (1852) of the English lexicon was a quantum leap in providing an onomasiological account of a European language. Though Roget can be seen as redeveloping the resources of synonym dictionaries, the conceptual structure he devised was original, an analytical system he had begun decades earlier (in 1805), and developed in the background to his professional work as a doctor, scientist and as secretary of the British Royal Society from 1827 to 1848. His interests were also philosophical – like Wilkins, he engaged with Leibniz’s philosophy and questions of how human thought was represented in language. That apart, he had a practical concern with finding the right word to express an idea, and for many users of his Thesaurus, it was effectively a ‘word-finder’. The macrostructure of Roget’s Thesaurus is a set of hierarchically ordered concepts, articulated vertically down from just six at the highest level:

• abstract relations

• space

• matter

• intellect

• volition

• sentient and moral powers.

Beneath these ‘superconcepts’, he developed a structure of 1,000 concepts and subconcepts, so that the words labelling them are in hyponymic relations with those above and below. Some of the nodes at the second and third levels down had to be labelled by means of phrasal descriptors, e.g. ‘power in operation’, ‘indirect power’ (both under CAUSATION) because there was no single word for the concept. These phrasal descriptors show gaps in the hierarchies of nomenclature for ordinary words, unlike scientific terminology where there is hyponymic nomenclature at every level within its taxonomic systems. At the fourth level of the Thesaurus hierarchy, Roget provided two sets of words that express a given concept: one containing analogous words, the second correlative words, both including nouns, verbs, adjectives and sometimes adverbs. For example, under the concept degrees of power there is first a set of analogous words whose meanings affirm the subconcept of strength (‘vigour’, ‘to force’, ‘strong’, ‘strongly’), and a second set (the correlative words) whose members negate the subconcept, e.g. weakness, fail, impotent. In the original layout of the Thesaurus, the two sets were presented in parallel columns, so that readers could fi nd the word they were seeking under a positive or negatively related concept. Roget thus anticipated one of the findings of cognitive linguistics, that linguistic opposites are closely associated in the human brain (Aitchison 1992: 74–7). Roget’s parallel layout has been lost in modern versions of the Thesaurus, which simply interleave the two sets in successive paragraphs down a single-column page. Likewise Roget’s original designations ‘analogous words’, ‘correlative words’ are now overlooked,10 so that words grouped together in his Thesaurus sets are sometimes thought of as synonyms, though they may be synonymous in only one of their denotations and divergent in their connotations. There are of course no definitions to detail their meanings, as in a dictionary. Within the Thesaurus, word meanings are simply indicated by their place in the hierarchy of concepts, and their positive/negative polarity. This was in keeping with Roget’s aim of providing partial synonyms from which literate users might choose the most effective alternative for their purposes. The original Roget’s Thesaurus included about 15,000 words, a relatively small inventory by comparison with modern dictionaries of the same size. It was however ‘greatly enlarged and improved’ after Peter Roget’s death (1869) in two-yearly editions by his son John, and annual editions by grandson Samuel from 1925 to 1953. From c.1880 there were separate British and American editions, in which the British reviewed and enlarged the inventory within Roget’s original subcategories, and the American reworked the material and the overall hierarchy of concepts (Burger 1991). In Britain ‘Roget’s Thesaurus’ was trademarked, whereas in the US the Roget name became generic, and used by imitators and publishers of more and less thesaurus-like publications.