Introduction

Languages (and hence the grammars that characterize them) are used for two very different purposes. On the one hand, language plays a central role in cognition. Our thoughts are often (but not always) formulated in language and most linguists would agree that a central part of grammar is a level at which meanings are represented. On the other hand, language serves as the primary medium of communication among members of our species. While one might reasonably take the position that vocal communication involves, to a certain extent, the transfer of meaning from one person to another, that is not one of its necessary features. Indeed, it has been shown that a sizeable percentage of our utterances are primarily social in character, rather than involving an exchange of (propositional) information.

Now, clearly, a system ‘ideally’ designed for cognition is not necessarily a system that is ‘ideally’ designed for verbal interaction. On the one hand, the former admits to a complexity not found in the latter. That is, introspection tells us that we are capable of forming in our heads complex concepts that we could never have the ability (or, often, the desire) to communicate to another individual. On the other hand, verbal interaction is subject to constraints of real time that are not found in the simple representation of meaning.

The purpose of paper, then, is to puzzle out different contributions of cognition-aiding factors and vocal interaction-aiding factors in the origins and evolution of grammar. In keeping with standard practice among linguists — albeit a practice that has often created confusion among non-linguists — I will refer to the latter factors as ‘functional’ ones. This usage has arisen as a result of the belief of a large number of linguists that the primary ‘function’ of language is communication (rather than knowledge representation). Hence a ‘functional explanation’, for many linguists, is one that is based on enabling more efficient communication among speakers and hearers. In the remainder of this paper, then, a factor will be said to be ‘cognitive’ if it is based on representations of meaning or thoughts, and a factor will be said to be ‘functional’ if it is based on pressure for (more) efficient use of language among members of the species. I will argue that both played a role in evolution, but an imbalanced one. Natural language grammars at their onset reflected primarily cognitive factors. Functional pressure exerted itself only after the basic structure of grammar was in place. Furthermore, those factors based in cognition are reflected in the genetically-transmitted aspect of grammars (or ‘Universal Grammar’, as it is often called), while communicative factors are primarily historical and have become more and more manifest with the passage of time.

On the functional grounding of grammar

At first blush, the structure of language seems good for communication and at same time not particularly well designed for the representation of meaning or thought. There are several ways that functional factors seem to have shaped language. Perhaps the most important is derived from pressure to process sentences rapidly. For example, cross-linguistic generalizations about language structure reflect the fact that it is in the language user’s interest to recognize the major grammatical elements of the sentence as rapidly as possible. One of the most longstanding typological generalizations in syntax (see Greenberg 1963) is that languages in which the verb precedes the object tend to have prepositions and that languages in which the verb follows the object tend to have postpositions (i.e., preposition-like elements that follow the noun that they are associated with). As Hawkins 1994 has shown, in languages in which this generalization holds, the length of time that it takes the hearer to identify all of the parts of the verb phrase is rather short, in that it is simply a function of the distance between the verb and the preposition. All that lies between these two elements is the object noun phrase. But in that minority of languages where the generalization linking verb-object and preposition-object order does not hold, the identification time is longer, since the object of the preposition as well as the object noun phrase lie between the verb and the preposition. The reasonable conclusion, then, is that the typological generalization about the correlation between verb-object order and adposition type reflects the preference of language users to process input rapidly.

Iconic motivation for grammatical structure is a theme in much functionalist writing. For our purposes this means that the form, length, complexity, or interrelationship of elements in a linguistic representation reflects the form, length, complexity or interrelationship of elements in the concept that that representation encodes. For example, it is well-known that syntactic units tend also to be conceptual units. In his classic study of the effects of iconicity in syntax, Haiman 1985 points to a multitude of cases where grammatical distance and conceptual distance are correlated. Consider, for example, the following well-known pair of examples:

1. John caused Bill to die by inadvertently buying him a ticket on a plane that ended up crashing.
2. ?John killed Bill by inadvertently buying him a ticket on a plane that ended up crashing.

The oddness of (1b) results from the fact that causatives that are single words (e.g. ‘kill’) tend to convey a more direct causation than periphrastic causatives (e.g. ‘cause to die’). So, where cause and result are formally separated, conceptual distance is greater than when they are not.

The third type of usage-based explanation appeals to the flow of information in discourse. Such explanations start from fact that language used to communicate and communication involves the conveying of information. Therefore, it is argued, the nature of information flow should leave and has left its mark on grammatical structure. Information flow has been appealed to, for example, in order to explain the ordering of the major elements within a clause. There are three ways to say: ‘Daddy has brought a Christmas tree’ in Russian, a typical so-called ‘free word-order’ language:

1. Pápa prinyós yólku. (Daddy — bought — a Christmas tree)
2. Yólku prinyós pápa. (A Christmas tree — bought — Daddy)
3. Yólku pápa prinyós. (A Christmas tree — Daddy — bought)

Each sentence is interpreted with the initial element representing old information and the final element representing new information. A common functionalist claim is that the discourse principle of Communicative Dynamism governs the ordering. The passage of time from past to present to future is mirrored iconically in discourse by the ordering of old information before new information (see, for example, Firbas 1987).

Not only do grammars in certain respects seem well designed functionally, but there are also a number of ways that they seem extremely poorly designed for the representation of meaning or thought. Hurford 2002 has catalogued some examples of such (seeming) poor design. Most importantly, since we do use language to communicate, grammatical structure necessarily has phonological properties as well as syntactic ones. Phonology is necessary for the expression of ideas, but not their formulation. Likewise, language is replete with morphological complexity, which serves no obvious cognitive function. For example, in English a concept can be represented by a stand-alone word (‘book’), by a prefix (‘un-able’), or by a suffix (‘king-dom’); indeed a single word can contain all three (‘under-talent-ed’). If language evolved solely in the service of cognition, it is hard to imagine why it would manifest morphological complexity.

Hurford points to several other ways that grammars make distinctions that serve no evident cognitive function. For example, virtually all languages distinguish in their grammars between the grammatical relation ‘Subject’ and the grammatical relation ‘Direct Object’. These grammatical notions conflate a relatively large number of semantic ones. Hurford also notes that all languages contain what he calls ‘quirky mismatches’, that is, they express parallel semantic notions in quite different grammatical ways. Hence past and present tense in English are encoded by suffixes, but future tense by a modal verb. Along the same lines, markers of clause boundaries such as complementizers serve no obvious role in cognition, nor does the ‘displacement’ of an element from another element with which it is associated semantically (e. g., in ‘Who did you see?’, ‘who’ is displaced from the verb ‘see’, of which it is the direct object). Finally, Hurford points to the phenomenon of grammatical agreement as one which serves no cognitive function. So in Spanish, the article and the adjective must agree in gender and number with the noun that they modify — a puzzling fact from the perspective of a cognitive account of the origins of grammar.

Grammar as a reflection of cognitive processes

The core of my argument for the importance of cognitive factors is based on the many ways that grammars seem overdesigned for efficient communication and at the same time well designed from the standpoint of cognition. First, consider the importance of full argument structure. An interesting fact about actual utterances produced by language users is that they rarely contain a subject, a verb, and an object, where the subject and the object (the ‘arguments’ of the sentence) are full non-pronominal words. Most utterances consist of a verb with one full argument, which is either the subject of an intransitive verb or the object of a transitive verb. Other arguments are either reduced to pronomimal or affix status or omitted entirely (the latter in languages like Spanish and Chinese which allow the wholesale omission of arguments). One’s first thought might be that this fact presents another piece of evidence against the idea of the origin of grammar in cognition. However, when one looks at things more closely, we see that grammars are in fact ‘propositional’, that is they consist of structures specified by formal rules that take the sentence to be the basic unit of grammar, where sentences are in a rough mapping with propositions, verbs with predicates, and noun phrases with logical arguments. Evidence for this claim is provided both by the fact that the process of speech production involves calling upon the full argument structure of the sentence (Levelt 1989) and by the fact that sentence fragments can be interpreted only by reference to full grammatical structure (Newmeyer 2003). In other words, whatever one might do in actual speech, one’s cognitive representation embodies all the arguments of the sentence. These are ingredients of cognition, not communication.

Second, every language on earth allows for the possibility of recursion, that is, sentences embedded inside of sentences inside of sentences, ad infinitum. For example, in principle, there is no limit to the number of times that another subordinate clause can be added in sentences like the following:

(3) Mary thought that John said that Sue insisted that Paul believed that …

Is recursion necessary for communication? Apparently, it is not. We virtually never have any reason to utter complex sentences like (3). And the desired message conveyed by sentences with recursion like (4a) can easily be communicated by a sentence like (4b), employing juxtaposition of two clauses:

1. Mary thought that John would leave.
2. Here is what Mary thought. John was going to leave.

Why does human language have recursive properties? The obvious answer is that human thought has recursive properties.

Third, human languages are horribly designed for communication from the point of view of the amount of ambiguity that they allow. Virtually any sentence imaginable is loaded with potential ambiguity. Of course, we deal with this problem in actual language use by means of complex systems of inference and implicature, conveyed meanings, and so on. Hence, in actual conversation, real ambiguity is normally a minor problem. But our concern here is whether languages are well shaped for communicative purposes. Based on the ambiguity that they permit, the conclusion has to be that they are not well shaped. If we focus on cognitive representations instead of on communication, however, structural ambiguity is a much less serious problem. The reason is that many (communicatively) ambiguous sentences are disambiguated by their structures. From the point of view of language use, the possibility of dual (i.e. ambiguous) representations for the same sequence of words is not communicatively desirable. But since the different meanings are represented differently from the cognitive standpoint, we must conclude that in this respect grammars seem well adapted to cognition.

Fourth, a central fact about language that it allows us say anything that we can conceptualize, regardless of whether we would actually have any need, desire, or likelihood to convey the information conceptualized. No more effort is required to say an obviously false sentence or an obviously true one than one that might be genuinely communicatively relevant. Along the same lines, we have no more trouble uttering pure nonsense sentences than grammatically parallel ones that contain an easily accessible semantic content. In other words, communicatively useless sentences provide another example of how language is ‘overdesigned’ for communication.

Fifth, grammatical categories tend to have a closer relation to cognitive categories than to communicative ones. As we have seen, the mental representation of transitive sentences contains a subject, a verb, and an object, even though actual utterances are pared down considerably. The classical definition of parts of speech in terms of meaning ('a noun is a person, place, thing, or idea') is close to correct. Units of word formation (morphemes) are almost alway definable semantically. On the other hand, communicative categories such as 'topic' and 'focus' are less likely to be marked in languages by a special category than semantically-defined categories are. In other words, the properties of the basic grammatical building blocks of a sentence suggest that in an important respect, grammar is better designed for cognition than for communication.

Finally (and more controversially), covert levels of grammatical structure represent aspects of meaning. In the classic Government-Binding model of grammar, the covert level of D-structure is the pure representation of predicate-argument structure and the covert level of Logical Form represents quantification structure, that is, the relations between quantifiers and the variables that they bind. If this picture is right, it suggests that cognitive relations are at the heart of grammar in a way that communicative aspects are not.

Cognition, functional pressure, and the origins of language

Where are we now? On the one hand, we have seen six important design features of language that pertain little — if at all — to communication. Importantly, none of them are ‘learnable’ in the ordinary sense of the word, suggesting that they were there from the dawn of human language itself. On the other hand, those aspects of language that seem designed to better aid communication are historical in nature. That is, unlike predicate-argument structure and so on, we can see how they developed over time. For example, take a communicative aspect of language par excellence, namely discourse markers. These are expressions like the following:

(5) then, I mean, y’know, like, indeed, actually, in fact, well, …

Even though they are essential to the makings of a coherent discourse, they invariably arise historically from something else, most typically out of conceptual meanings and uses constrained to the argument structure of the clause. This fact is not surprsing if if vocal communication itself is derivative. Nouns and verbs trace back to nouns and verbs, because they were there from the start. The derivative nature of discourse markers points to a time when we had structured conceptual representations, but they had not yet been coopted for communication.

There is another reason to posit that cognition left its mark on language before communication. We have learned that the conceptual abilities of the higher apes are surprisingly sophisticated. However, their communicative abilities are remarkably primitive. There is very little calling on their conceptual structures in communicative settings. These facts suggest a three-stage process in language evolution. First, there was the inherited level of conceptual structure. Secondly, the level became linked to the vocal output channel, creating for the first time a grammar that was independent of the combinatorial possibilities of conceptual structure per se and making possible the conveying of thought — in other words, communication. And once grammars started to be drawn upon for real-time purposes, the constraints of real-time use begin to affect their properties.

Conclusion

This paper began by raising the question of the relative roles of cognitive and functional factors in the evolution of grammar. The conclusion is that they have both played important roles, though unbalanced ones. Human language was jump-started by the linking of conceptual structures and the vocal output system. In other words, cognitive factors were the first to shape grammars. But with the passage of time, the exigencies of communication came to play an ever-more important role in grammar. Human language today therefore reflects the influence of both types of factors.

References

Firbas, Jan 1987. "On the operation of communicative dynamism in functional sentence perspective." Leuvense Bijdragen 76: 289-304.

Greenberg, Joseph H. 1963. "Some universals of language with special reference to the order of meaningful elements." In Greenberg, Joseph, Ed. Universals of language, 73-113. Cambridge, MA: MIT Press.

Haiman, John, Ed. 1985. Iconicity in syntax. Typological Studies in Language. Amsterdam: John Benjamins. Hawkins, John A. 1994. A performance theory of order and constituency. Cambridge: Cambridge University Press.

Hurford, James R. 2002. "The roles of expression and representation in language evolution." In Wray, Alison, (ed.) The transition to language, 311-334. Oxford: Oxford University Press.

Levelt, Willem J. M. 1989. Speaking: From intention to articulation. Cambridge, MA: MIT Press.

Newmeyer, Frederick J. 2003. "Grammar is grammar and usage is usage." Language 79: 682-707.

Traugott, Elizabeth C. and Richard B. Dasher 2002. Regularity in semantic change. Cambridge: Cambridge University Press.