| For a translation into rumanian, see this link. |
The Pronominal Approach (PA) has been presented as a research method for the study of the basic structures of natural languages. It was conceived by Karel van den Eynde at the Catholic University of Leuven in 1969 and was developed at the universities of Leuven and Aix-en-Provence through the following two decades (Eynde 1978 & 1988, Blanche-Benveniste et al. 1987, Eynde et al. 1988-89). Since 1992 further development took place in cooperation between the universities of Leuven and Odense (Schøsler & Kirchmeier-Andersen 1997 a).
In the PA we find a combination of ideas from two different areas: linguistics and constructive mathematics. From linguistics the PA has received inspiration from the distributional analysis principles of Togeby and Harris, from mathematics the PA has been influenced by the notion of proportionality or isomorphism and by the constructivism known from constructive mathematics, especially Goodman (1951).
As described by Eynde (1995) the PA is on a par with the linguistic principles of Knud Togeby (1965), the distributional analysis of Zellig Harris (1951, 1954 and 1965), and the componential analysis of Eugene Nida (1975). They all argue that the proof of the existence of a semantic property for any linguistic unit should lie in the study of their syntactic behaviour. Harris himself has suggested statistical observations in a corpus as a possible means for generalization, however, having in mind the obvious shortcomings of a statistical approach (Gebruers 1991:238), it seems worthwhile instead to try to derive a classification from a representative subset of linguistic material.
The development of a method for determining such a representative subset is the main contribution of the PA. The PA relies on two basic assumptions:
According to the PA, pronouns may be used as a set of basic primitives which enable us to investigate a large fragment of a language on the basis of a smaller and more manageable one.
The relation of proportionality concerns the pronoun and the syntagms which may stand in its place. This relation is constant in the sense that we cannot freely associate pronouns and syntagms. The pronoun 'he' is always related to syntagms like 'the boy', 'the man' 'the father' etc. and never to 'the girl', 'the woman',' the mother' etc. It follows also that the relation is a one to many relation as one pronoun may be associated with an almost infinite number of syntagms.
The relation is proportional in the sense defined in the context of constructive mathematics especially Goodman's work on constructional systems (Goodman 1951). Goodman develops the notion of extensional isomorphism as follows:
A relation R is isomorphic to a relation S in the sense here intended if and only if R can be obtained by consistently replacing the ultimate factors in S (Goodman [1951](1977); p. 10-11).
Thus, for two relations a:b and a':b' we can state: iff a is related to a' in the same way as b is related to b', they are isomorphic.
| a | a' | |
| -- | :: | -- |
| b | b' |
Along the same lines, the relation between pronouns and syntagms can be described as isomorphic or, in terms of the PA, as proportional, as illustrated in:
| she | he | it | ... | |||
| --------- | :: | --------- | :: | --------- | :: | --------- |
| the girl | the boy | the book | ... |
The series of equations should be read: 'she' stands to 'the girl' as 'he' stands to 'the boy' and as 'it' stands to 'the book'.
| he | him | |
| --------- | :: | --------- |
| the boy1 | the boy2 |
A relation of proportionality may not only exist between pronouns and syntagms but also between different categories of pronouns and even between different categories of syntagms.
With these background assumptions in mind, the methodology of the PA can be outlined as follows: The PA is based on the hypothesis that the valency kernel, i.e. the verb, noun or adjective, is the basic unit of analysis because it associates a number of valency-bound elements and determines the relation between them. The first aim of the PA is to ascertain a complete coverage of the possible valency relations of the valency kernel. Following the principles of constructive mathematics it does not attempt to do so in one go by taking into account all possible realizations of entities which may be valency-bound, but instead it restricts the investigation to a representative subset of the language consisting of elements from a closed word class, the pronouns. Since each of these elements can be shown to be proportional to a much larger set of syntagms, the results of investigations based on the restricted inventory can be extrapolated to the rest of the language.
A distinction is made between simple predicators ("to pull") and complex predicators ("to pull somebody's leg"). In the latter case, several lexical elements form a unit which has its own valency.
The first phase wants to cover not only the main sentence structure (containing a verbal or adjectival predicator with their pro-syntagm paradigms), but also all types of embedded constructions that show some valency kernel of their own (e.g. nominal, attributive adjectival, adverbial) and that appear with identical or similar pro-forms (pro-kernels, pro-referents, pro-determiners) (Eynde 1998:145-167).
The secondary aim is to identify in a negative way (i.e. through absence of proportionality relation with pronominal paradigms) all elements that are not part of the former "endotactic" skeleton of the language, and may thus be analyzed as "epitactic" elements. These elements will then be classified in distinct categories that receive their syntactico-semantic definition from their status of (left or right) adjunct to the former endoctactic structure or to one of its different (sub)constituents (either predicator or pronomial paradigms.
For the identification of a given valency slot the proportionality relation between one or several pronouns and a set of syntagms must be established. For the syntactico-semantic description of a valency slot the complete paradigm of the acceptable pronouns in this slot must be determined.
An introduction to the pronominal approach and its application to French within the proton dictionary is available here.