In this article I present some examples of the way in which cognition, pragmatics and grammar can be integrated. This view of language, which underlies the theories of language use like cognitive linguistics, contrasts with models like structuralism or generative grammar whose object of study is clearly separated from use. The general principles of cognitivism and the discussion and exemplification of the concepts of construction, grammaticalisation or subjectivisation point to the fact that grammar is a continually evolving entity under the effect of discursive use (pragmatics) and it reflects to some extent aspects of cognition.

The scientific study of language, or more precisely of grammar, has gone forward through different models such as structuralism, generativism and models based on language use among which cognitive linguistics is specially relevant. Each of these models has focused in different ways on what, according to Geeraerts (2003), can be viewed as three complementary perspectives on language: language as a social system, as an individual system and as an individual activity.

a) Language is a social system to the extent that languages are means of socialisation and develop in collective environments.

b) Language is an individual system because we all have our own knowledge of the social system identifed as a specific language.

c) Language is an individual activity because any language is an abstraction resulting from the integration of the individual activities of all its speakers, and the concept of speaker has an inherently social dimension, since human communication presupposes more than one person.

This triple perspective explains the interdisciplinary nature of the study of language and helps understand the different approaches in Linguistics. Ferdinand de Saussure’s structuralist model, which is considered strictly speaking the first scientific approach to linguistics, is based on the dichotomy between language and speech.

Language is a set of collectively codified conventions (a social system), while speech is the ensemble of combinations effectively produced through the code (and thus, an individual psychological activity). For structuralists, linguistics can only concern itself with the system; as a consequence, what we have is a divided model of grammar, in which the system as an activity is ignored, and the individual system (the individual knowledge of the social system) is neglected.

From a generativist point of view, the faculty of language can be identified with linguistic competence, that is, the innate ability of the speaker-hearer to understand and produce a theoretically infinite number of linguistic strings. A finite number of elements (grammatical categories: noun, verb, adjective, etc.) combine according with generative and interpretative rules resulting in grammatical sentences.

Figure 2. Generative grammar

Competence, which corresponds to the individual system, is opposed to performance, the actual realisation of competence in specific linguistic outputs, which can contain errors and features not obviously derivable from the system. The mind is seen as a computer which, among other modules, has one devoted to language. It combines categories according to generative rules for well-formedness and gives rise to sentences which take on physical substance by means of phonetic and semantic interpretative rules. This leads to the isolation of grammar, which is considered to be a separate module from the rest of the cognitive capacities and is studied at the individual level as a system, leaving aside the linguistic activity (performance) as unsystematic. Similarly, the social side of language - its inherent social nature - is also ignored.

After the appearance on the scene of generativism, by the end of the fifties, different disciplines filled the spaces left empty by this model: the social context (sociolinguistics), the situational context (language use) and the cognitive context (the relation between meaning and experience). The bases of the theories of language use emerged, seeking to "recontextualise" grammar and link it to the social setting in which it is developed and realised, to the world and to experience of the world. In this category we find, variously, the functionalism, formal semantics, pragmatics and discourse analysis, sociolinguistics and cognitive linguistics, amongst others.

Cognitive Linguistics can be considered the most comprehensive theory of use because it aims at studying and interrelating the three aspects of language. Some of the basic principles of Cognitive Linguistics will serve to show the importance given to context:

a) Language is not an autonomous faculty, but rather is related to other human cognitive abilities. Thus, the philosophy that inspires this theory is what George Lakoff calls experientialism, in opposition to objectivism, which regards language as a faculty which is separate from the other cognitive faculties or capacities and so to speak disconnected from the world.

b) The main object of study is not system, but language use. Therefore, the system is conceived in a dynamic way, interrelating linguistic structure (syntax), meaning (semantics), language use (pragmatics) and conceptual structure (cognition).

c) Grammar is the result of structuring and symbolising a semantic content through a phonological form.

d) Meaning is not seen from a purely denotative point of view, but includes connotative aspects, so that semantics and pragmatics cannot be separated.

e) Semantico-pragmatic structure in many cases motivates grammatical structure and, therefore, syntax cannot be treated as autonomous.

f) This dynamic view of language breaks down the divisions between the different linguistic levels of language (semantics and pragmatics, semantics and grammar, grammar and lexicon) and shows the weakness of dichotomies such as dichrony and synchrony, competence and performance, denotation and connotation.

Some examples can illustrate this conception of language and linguistics. From the contrastive point of view, it can be noted that a common concept may be expressed by different linguistic forms in three relatively closely related languages. For instance, what in Catalan is called oficina d’objectes trobats (literally, ‘the office of found objects’), in Spanish is oficina de objetos perdidos (literally, ‘the office of lost objects’) while in English this is called lost and found. Obviously, the meaning is the same in all cases, i.e., a place where lost objects are stored, and where people who have lost objects go to get them back. However, each of the languages focuses on an aspect of this reality: in Catalan the term expresses the place (office) and the end of this process (finding), in Spanish the place and the origin are focused upon (losing), and in English the two parts of the process (losing and finding) are highlighted. The reality behind the three expressions is the same, but the conceptualisation is not. The other ways of expressing it could be valid in the respective languages (oficina d’objectes perduts i trobats in Catalan, oficina de objetos encontrados in Spanish, found objects office in English), but, simply, these are not the terms used. In each language the former expressions and not the latter have been conventionalised. This difference cannot be explained by means of a linguistic theory that limits itself to the linguistic system or to competence and does not take into account pragmatics and cognition.

Turning now from lexicon to syntax, we can now review some examples of constructions that, from a logicist point of view, have the same basic meaning:

(1)
a) Joan va trobar l’error / John found the mistake.
b) L’error ha estat trobat per Joan / The mistake was found by John.
c) L’error ha estat trobat / The mistake was found.

Traditionally, passives (1b) have been considered a transformation of the corresponding active sentences (1a): the subject of the active has become an agent complement and may be left out as in (1c). This analysis, which constitutes one of the basic examples of transformations in generative grammar, raises a number of questions. If actives and passives are purely structural variants: Why do both structural options exist (one form would have been enough)? And why, if we analyse any Catalan corpus, the passive with an explicit agent turns out to be unfrequent? And under what conditions is one structure used rather than the other?

By looking at the conditions on use, at the context, answers to these questions can be found. Passive constructions are means of relegating the subject of an agentive construction to a second place or eliminating it altogether, and, as a consequence, highlighting the object. The latter is shifted to the prominent position in the sentence, that of grammatical subject.

In line with this perspective, it is clearer why the agent tends not to appear: in this way the object is fully highlighted. So, though actives and passives are related, they do not comply with the mathematical principle that "the order of the factors does not alter the product". There might be a correspondence between the two structures, but they are not equivalent, since they exhibit differences in meaning and in discourse conditions on use. It can easily be shown that periphrastic passives are little used in informal conversation, and in the case of Catalan, these passives only occur in formal registers, such as academic or legal language.

Defocusing the agent is not a exclusive property of the passive. The same or similar function is carried out by reflexive constructions (middle voice forms) (2a), or by impersonals with a generic interpretation (2b-c).

(2)
a) S’ha trobat l’error / Se (REFLEX) has found the mistake.
b) Hom ha trobat l’error / One (INDET) has found the mistake.
c) Han trobat l’error / they (INDET, elliptic) have found the mistake.

The sentences in (2) are almost equivalent to the passive L’error ha estat trobat (‘the mistake has been found’); however, they differ in not highlighting the object of the action, because the agent is still implicit (‘someone has found the mistake’). It is an indeterminate agent, whose specification is, for whatever reason, of no interest in the context of the discourse. Thus, the action is focused upon. The concept of construction gives an adequate account of examples as the previous ones since it shows that form and meaning are not related in a totally arbitrary, nor totally predictable way. It can be concluded that certain aspects of the form or the meaning of a construction cannot be derived from its components but are effects of the construction itself, which conveys pragmatic aspects that speakers know.

On the other hand, it is evident that Catalan uses the periphrastic passive much less than English, and this may be due to the existence of reflexive constructions, which occupy the space taken up by the passive in English - a language that lacks reflexive structures to express passive conceptualisations. Just as in the case of the "lost and found" languages conceptualise situations in a different way and codify them with structures which in turn imply differentiated discourse effects. Thus, grammar is not independent from cognition and conditions of use, but rather a constant interrelation is set up between the three levels: cognition, pragmatics and grammar.

The connection between pragmatics and grammar is pointed out by the research carried out within grammaticalisation theory by cognitive linguists like Sweetser, as well as functionalists like Traugott, Hopper, Thompson or Heine. Grammaticalisation theory has its roots in the work of Meillet or the structuralist Kurylowicz, who define grammaticalisation as the process by which "a lexical unit or structure takes on a grammatical function, or […] a grammatical unit takes on a more grammatical function" (Heine et al. 1991: 2). The modern version of this line of research argues that grammaticalisation is more than a process of turning lexical elements into grammatical ones; it is a more complex phenomenon that entails modifications in the discourse function and syntactic structure of languages.