Grammaticalization is a process profoundly affecting and changing the linguistic system. Most linguistic investigations of grammaticalization have analyzed its manifestations in the language system, rather than in linguistic activity. Nevertheless, the linguistic system is just the systematic aspect of the linguistic activity of a speech community (Coseriu 1958:271f). As W. von Humboldt (1836: 418) put it:
Language if conceived in its actual nature is something that constantly and at every moment passes by. ... It is itself not a fact (ergon), but an activity (energeia). Consequently its actual definition can only be a genetic one [i.e. one referring to its origin]. Namely, it is the ever repeated work of the mind to enable the articulated sound for the expression of thought. If one takes this directly and strictly, it is the definition of actualized speech; however, in a true and essential sense one can consider as language, as it were, only the totality of this speech.
Linguistic activity, i.e. speaking and understanding, is composed of many acts, operations and processes. From a linguistic point of view, all of them share two fundamental aspects, viz. the selection and combination of units: Every linguistic unit is selected from a set of units that could take its position, and is combined with other units of the same level into a larger unit. This happens at all linguistic levels regardless of whether the speaker is taking an analytic-compositional or a holistic approach. In other words, a unit of a certain level may be composed of units of the next lower level. At the same time, however, the higher level unit is chosen from among a set of units of its own level.
The operations of selection and combination are freer at higher levels of linguistic structure and more constrained at lower levels (s. the section on the Penthouse Principle). The constraints relevant here are ones of the specific language system. They are frozen conventions of the speech community regarding use of its language. The diminished freedom of the speaker in selecting and combining lower-level units is mirrored in the diminished autonomy of these units. Disregarding the subsystem of distinctive units, selection and combination of low-level significative units are conditioned by rules of grammar. For example, most allomorphy is entirely conditioned by its immediate context.
All of this means that what is commonly called ‘linguistic activity’ is not only composed of controlled actions and acts, but also of automatic processes. In short, linguistic activity develops in a teleonomic hierarchy as explained in a preceding section: At the highest level, the speaker freely determines his cognitive and communicative goals; at the lowest level, the constraints of his internalized grammar determine linguistic structure. Recall that the fact that we are conscious of what we are speaking about was taken as a definitional feature of consciousness.
The aspects and components of the activity of speaking and understanding can therefore be arranged on a continuum between the poles of maximum consciousness and total subconsciousness. Each of these components has the two fundamental aspects of any linguistic activity, selection and combination of units. In this sense, it is the selection and combination of linguistic units which is conscious to different degrees. Some of these are shown in the following diagram (cf. Levelt 1989:21f and Knobloch 1994:215f).
consciousness | aspect |
---|---|
high | the current communication problem: illocutionary force and content of the speech act |
↕ | information structure, high-level constructions, lexemes |
mid-level constructions, free grammatical formatives | |
low-level constructions, bound grammatical formatives | |
low | articulation and audition; neural co-activation of syntagmatically and paradigmatically related units |
This conception must be protected against some possible misunderstandings. First of all, we are here focusing on the speaker’s consciousness while executing the operations of selection and combination. Once these are executed, their products leave a trace in his short-term memory and hit his ear; and he is now free to apply any degree of control and consciousness to their analysis (cf. Levelt 1989, ch. 12 and Knobloch 1994:214). This latter feedback process, though important to linguistic activity, is not what is analyzed in the above diagram.
Second, the consciousness levels of the diagram are related to a standard situation of speaking. Many speech situations are not standard in this sense: a phonetician demonstrating some aspect of speech sounds, a cabaret artist imitating the way of speaking of a politician, speaking a foreign language, speaking under the influence of alcohol etc. Even in standard situations, the level of consciousness of component processes varies, e.g. when we say something that we habitually say in this kind of situation or when we have problems of phrasing and wording. Within certain limits, the level of consciousness of certain low-level processes can be raised willfully; s. the explanation of the teleonomic hierarchy. However, the lower the level on which processing is by default automatic, the more difficult its raising to consciousness gets. The lowest-level aspects of speaking are in principle inaccessible to consciousness.
Actually, the speaker’s freedom concerning use of linguistic units is realized at two logical levels: On the one hand, at the lower levels of linguistic structure, he forfeits his freedom to manipulate linguistic units, and instead the language system dictates him what he can, must and cannot do. It is necessary that linguistic activity have these totally automatic, uncontrolled aspects because any complex problem-solving activity must be partially automatized if it is to be executed with at least some efficiency (s. the section on automation). On the other hand, for any given semantic act and unit, the speaker is free to choose the level of control with which he wants to execute and manipulate it. It is not the case that there was a certain inferior layer of cognitive and communicative functions that was in principle inaccessible to free control. Instead, for every meaningful operation and unit, the speaker first chooses the degree of freedom with which he wants to execute and manipulate it. What is currently not at stake may be relegated to inferior levels and may safely be left to the rules of the linguistic system and, thus, to automatisms. Leaving it there unburdens linguistic activity, so the speaker’s capacity to achieve maximum effect for those aspects of his activity which matter to him is increased.
This brings us back to the capacity of the human being to subject to conscious control almost any aspect of the actions and processes that he is involved in. If there is a human skill of some importance, then there is an intellectual activity which reflects on it. Full mastery of some skill therefore comprises two levels of competence, procedural and reflexive competence, where the latter is declarative knowledge of the skill. In the case of language, this leads to the distinction between proficiency in a language and metalinguistic knowledge of a language (Lehmann 2007). Linguistics is, of course, at best metalinguistic knowledge of language at the highest level.
The character of linguistic activity as a teleonomic hierarchy and the position of linguistics at the level of reflexive competence also has consequences for linguistic analysis. Every element of linguistic structure has a function. However, for lower-level elements, this function typically abides within the linguistic system. In other words, a functional analysis is not a mapping of every bit of linguistic structure onto some cognitive or communicative function. Quite on the contrary, a functional analysis has to move the entire teleonomic hierarchy upwards without skipping a step. A clear example of how sound linguistic method proceeds is Kaznelson’s (1974:35; cf. p. 93) analysis of the grammatical process of agreement:
The interrelations between form and content in agreement thus have several layers and levels. What appears as content on one level turns out as the form of a new content at a different, higher level.
For instance, the feminine desinence of an adjectival attribute signifies feminine gender. However, feminine gender has no direct interpretation as a feature of an adjective. Instead, its function lies in the agreement of the adjective with its head noun, which, in turn, serves attribution, which, in turn, serves the specification of a notion.