The topic of this section is at the heart of descriptive linguistics and, therefore, of all theories of grammar. It is the area of dispute among models such as Generative Grammar and Construction Grammar, to name but these. In what follows here, emphasis is not so much on the form that a linguistic description must take that properly accounts for the relation between lexicon and grammar, but rather on the function of these two components in linguistic activity.
The uppermost task in the teleonomic hierarchy is “the unlimited creation of interindividually available meanings”, consequently to produce and understand an unlimited amount of messages. This requires an interplay of two faculties:
| faculty | calculating the relation between an expression and its meaning by |
|---|---|
| semantic | rules which presuppose and spell out compositionality |
| pragmatic | world-knowledge and orientation in the linguistic and extralinguistic context |
In a simplistic model, the description of the language system consists of two disjunct parts, the grammar and the lexicon. With this hand in hand goes a supposition of the grammar being the core of the system, the lexicon just an appendix. This was codified already by Bloomfield:
The lexicon is really an appendix of the grammar, a list of basic irregularities. (Bloomfield 1933:274)
The basic fallacy underlying these approaches to the relationship between grammar and lexicon is the rule/list fallacy (thus named in Langacker 1987, ch. I.A.2). It consists in the supposition that each given linguistic unit must either be generated by rules or must be stored in the inventory. As a matter of fact, it is quite normal, both at the level of the individual mind and at the level of a linguistic description, that a given unit may be arrived at in either way, or that it is partially accessed holistically, partially analytically.
A proper account of the relation between lexicon and grammar starts from the presupposition that human beings may take either a holistic or an analytic approach to just anything. This is also the basic alternative underlying the opposition between lexicon and grammar, as shown in the following figure:
The lexicon is, thus, the set of significative units that are accessed holistically, while the grammar is the set of significative units that are accessed analytically. Needless to repeat, the two sets are not mutually exclusive.
The availability and use of the alternative approaches to cognition is universal in human beings, and so is the structure of individual cognitive activities, including importantly language, in terms of this alternative. This, however, does not imply that for any particular linguistic item, it is predetermined whether the individual speaker accesses it holistically or analytically. Here like in all other respects, linguistic theory is not a theory of the individual mind and its activity. In other words, a given linguistic unit such as lecture hall may be accessed holistically by one individual, but analytically by another individual; or even one way at one time and another way at another time by the same individual (cf. again the section on the holistic vs. analytic approach). Therefore no individual correspondence is to be expected between the degree of lexicalization or grammaticalization of an individual linguistic unit and the type of access an individual speaker takes in using it.
Trade-off between storage and computation, between strain of long-term memory and working-memory.
Lexicalization is a gradual process. It may affect some parts of a complex expression more than others.
Complex lexical items – phrases, idioms etc. – are not “complex words”. Instead, the lexicon contains units of all the grammatical levels. They differ from units of the corresponding levels in the grammar by a higher degree of lexicalization.
Methodologically, it is the task of the grammarian to maximize the role of computation, i.e. to minimize the amount that needs to be stored as unanalyzable.
There is a wide-spread supposition (esp. in the variety of generative grammar defended by Chomsky) that only the first of these faculties is involved and that, consequently, the infinite character of language is guaranteed only by compositionality, regularity and, thus, by grammar. This is probably a sublimation of the afore-mentioned methodological principle to the level of a theoretical principle.
The following conception appears more adequate: The long-term memory has stored units which differ in abstractness. Some serve as schemata for others, which are more specific. The schematic is the structural aspect; the specifics are increasingly narrow classes of linguistic signs down to just one item that contributes to the identity of a construction. The schemata are more or less productive, i.e. they may be used to form new messages. In this function, they correspond to rules of grammar.
In this conception, there is no borderline between lexicon and grammar. The two principles of associating sense with an expression are in equilibrium and are both at work in every single case:
Lexicalization of a construction may be described as follows: The process starts with a construction C1 and derives from it a construction C2 as a more lexicalized variant of C1. This subtype C2 of C1 is constituted by constraining either selection or combination of items, as compared with C1. At the same time, certain semantic features are specified at the level of C2 as a whole which were derivable by unification in C1.
Grammaticalization presupposes and strengthens the former of the above two faculties, lexicalization presupposes and strengthens the latter one.