Two strikingly different opinions exist about the origins and ontogenesis of theory of Mind. One position posits it as a culturally transmitted meme, a theory of naïve psychology that has proved sufficiently useful so as to occur to most if not all the human groups. The second posits a module genetically specialized for the understanding of mind-reading, sprung forth anew in every infant by virtue of being human (or maybe just a social primate). The meme view explains the four-year delay for false belief reasoning, though perhaps not the surprising invariance in the timetable. The module view explains the uniformity, but only at the cost of positing a moderating influence of “processing capacity” or its kin, executive control, which must mature.

Not surprisingly, two parallel opinions also exist about the origins and ontogenesis of human language. One position posits a cultural transmission, the other a genetic module.  On the first view language learning will take time and experience, on the second it won’t, except for the vexing issues of the time needed to learn a arbitrary lexicon, and to sort out from primary data which set of parameters is appropriate.

I venture that a popular opinion is a mixed model: a module for language, and a meme for theory of mind. That is, genetic transmission for the first, cultural transmission for the second, using language as the medium. On the basis of poverty of the stimulus for language learning, the abstractness of what is learned, and the lack of direct instruction, linguists have concluded that language learning is too complex for it to be a meme. Gold’s mathematical proof of the unlearnability of a human grammar within a finite time (without UG to guide it) has held firm after 35 years of debate (8). But the idea of cultural transmission of theories about the mind not only seem plausible, but is apparently supported by data showing the impact of different rich or impoverished linguistic circumstances for learning the theory.

The literature on Theory of Mind in animals and humans leads me to the conclusion that whereas animals such as chimpanzees may share many of the mind-reading skills of humans, humans uniquely can do false belief reasoning. The literature on natural communication systems and on artificial language training with animals and humans convinces me that whereas several properties once held unique, such as referentiality and combinatorial ability, might be shared, humans uniquely have recursive syntax (10).

Might there be a plausible evolutionary story connecting the two findings? A cultural meme account of the transmission of theory of mind entails a complex enough language in which to couch talk about false beliefs, and that might in turn entail recursive syntax. This is a sensible account, but let us not stop at sensible accounts. It so happens that I have a theory about this that makes a connection between syntax and false belief reasoning in ontogenesis, that is not due to either

a) both being products of cultural transmission or

b) mature Theory of Mind arising by cultural transmission via language.

I use this essay to ask: Can that theory be extended to the case of language evolution?

Language for thinking?

Several philosophers of mind have argued that we take seriously the notion of a language of thought, a language rich enough to express the contents of our propositional attitudes and therefore as rich as natural languages. We can make inferences across the propositions of this language of thought and deduce consequences e.g.:

If George thinks his keys are in the hall, George will search for them in the hall.

Sometimes we engage in this reasoning overtly, often it does not reach ‘consciousness’, and may be in some form prior to full linguistic formulation. The normal condition under which we test 3 and 4 year olds’ understanding of another’s false beliefs are simple: we set up a condition under which someone would act one way if she knew a fact, and another way if she believed something contrary to that fact. We ask: can the child represent the contents of that other person’s mind? So on the classic task can the child retain:

 Maxi believes the chocolate is in the basket

even when it has now been moved from the basket to the cupboard, unseen by Maxi. If the child can retain that, he can deduce:

            Maxi will look in the basket

Or explain:

            Maxi looked in the basket because he left his chocolate there

The question is, how is that belief about Maxi’s belief represented? I have argued that the representation must be as rich as the statement in our natural language, and have the crucial property of a recursive sentence, namely one sentence embedded in another, in which the embedded sentence is false while the overall sentence remains true. This type of structure is uniquely associated with verbs of mental state and communication and this fact provides an important key to its learnability on my account. I have argued that the child can recognize the appropriate recursive structure using communication verbs:

            Maxi said the chocolate was in the basket

for which the child can experience overt events, speech acts, and recognize how the truth conditions hold against the speech. Only when a child has this linguistic structure established, will he have the representational capacity to reason about false beliefs, by analogy with attributed false statements. That is, the verbs of communication and mental state share a crucial recursive syntactic/semantic structure, and identifying it in the case of overt speech provides a bootstrap to the understanding and generation of statements about false beliefs. The predictions have been borne out in several ways empirically, though there is still much theoretical work to be pursued. In brief, normally-developing children (both English-speaking and native ASL) show evidence of mastery of complement syntax via understanding wh-question-movement:

            The girl bought apples, but she said she bought pears.

            What did she say she bought? (pears, not apples)

before they pass the false belief tasks, even non-verbal false belief tasks (3,4,5). So do deaf children (both non-native born ASL and orally-trained) who are language delayed (6), and children with SLI (7), and high-functioning children with autism (20), though this often means on a timetable several years beyond normal.  Training studies to “teach” children complements with communication verbs show transfer to false belief reasoning, just as if the language mastery had enabled their representational skills(9,12).

Is it trivial?

There are several ways that the relationship between language skills and theory of mind could be relatively trivial, and critics have been quick to point them out.  First and foremost, it is not just that passing the verbal FB task itself requires certain linguistic skills, e.g. the ability to follow a narrative or answer a question. Instead, several different nonverbal (or at least low verbal) FB tests show the same dependency on the syntax of complementation. Second, it is not that that the language task and the FB task share in common executive function or meta-representational skills that provide the resource underpinning: language-delayed deaf children succeed on those at a younger age. Thirdly, it is not that the language task entails false belief understanding, because it precedes success on the FB task, and because the task does not require the child to make an attribution of belief. Furthermore, the older deaf children should pass the nonverbal FB task before the language in the case that FB is conceptually prerequisite, and they do not (6). (For a full review of these and answers to claimed counter-evidence, see (2)).

Why children take four years

It is important to recognize the peculiarity in this story, that children take almost four years to ‘recognize’ the right structure as a recursive complement. Why, if they have recursion as part of the evolutionary bio-program for language? I argue that it is critical that they avoid false analogues such as adjunct clauses (which don’t have the right truth properties):

            Maxi spoke and the chocolate was in the basket

 and irrealis clauses (which don’t have truth conditions of the same sort)

            Maxi wanted the chocolate to be in the basket

To get this far, the child has to make a multitude of subtle lexical and morphological distinctions that will take time and experience to learn. Recursion may be there from the start but recognizing it cannot be instantaneous (see also (17)).

Evolution and language

Hauser, Chomsky and Fitch (2002) identify recursion as the fundamental property of human languages that distinguishes them from other animal communication systems. The authors conclude that recursion may be present in other systems not related to language, such as navigation through space, but not in the communication systems of animals.

If true, the natural question arises, why not? Is all such speculation tainted with the shadow of Dr. Pangloss? Bickerton (1) argues for instantaneous change or punctuated equilibrium deriving syntax in toto from a previous stage of “proto-language”- essentially just a lexicon and non-syntactic strings.  Although some thinkers admit the possibility of  gradual change, Bickerton and others  scorn the notion of e.g. generative syntax minus Binding rules, or without Subjacency, because they are adamant that these pecularities of human syntax are not functional and therefore could never have been favorable adaptations. Furthermore, Bickerton claims these “partly evolved” languages should show up still in remote cultures, and that they don’t. However, new data on early forms of Nicaraguan Creole Sign hints at the possibility of “intermediate stage” languages (16). While highly sympathetic to Bickerton’s objections, I am tempted into speculation, not just what use is syntax, but more particularly, what use is recursion? Note, I am NOT suggesting that recursion was “invented” for a particular semantic or pragmatic need. I can still believe that a sudden mutation allowed for a formal property that did not exist in language before, which then conferred advantages on its speakers when it was used for a certain purpose. Bickerton’s own argument takes the same form.

The mathematician Nowak and colleagues (13,14,15) have begun asking questions about the adaptive value and learnability of simple protolanguages such as a primitive lexicon linked to referents. Not surprisingly, they find that the demands of memory, discriminability and time to learn constrain such a system to a fairly small finite set of signals, just what is seen in animal communication repertoires such as monkey calls. Their analysis of the evolution of grammar considers the conditions under which a combinatorial system would be more advantageous than a finite lexicon. In their modeling, if the number of distinct pairings of objects with actions is above a certain threshold, then there is an advantage to a combinatorial syntax that allows creativity for novel events over a large lexicon. They specifically compare two kinds of transmission: cultural, versus genetic, so this work is getting closer to the kinds of questions linguists have asked. That is, are we talking about the evolution of languages here, or the evolution of speakers of those languages? Can we talk about the adaptive value of the first without the second? One must assume that for something to be coded in the genes, it has to confer reproductive advantage on the carrier of those genes, yet languages change (on the surface) at a much more rapid rate than populations.  It is important to consider both possibilities of transmission.

Adaptive value of recursion

A simple combinatorial syntax is a long way from the kind of recursive syntax that linguists find uniquely human. Nowak and colleagues have so far steered clear of claims about particular linguistic properties, so Dr. Pangloss hoves into view.  The only way to proceed it to consider carefully what recursion might gain us, and then ask if this is enough to give languages with it an advantage over languages that do not have it, and their speakers too! Hauser et al do not specify which variety of recursion they regard as critical, and there may be several forms (17).

First, recursion often provides economy, that is, it avoids redundancy across multiple independent sentences. I do not find his kind of recursion, such as in relative clauses, particularly convincing in terms of evolutionary adaptation:

The boy was stuck in a tree. John helped the boy.

 

àJohn helped the boy who was stuck in a tree.

 The boy who was stuck in a tree? John helped him.

Not only are there alternative non-embedded structures that reduce redundancy, but human languages also show variation in which types of relative clauses they permit (3). For example, someone might argue that it is economical to say,

            “I met the man that my brother is taller than”

but you can’t even say that in French! Since forms of recursion vary across languages, this variation is probably a function of cultural transmission. Consider cases also like:

            “Fred’s mother’s dog’s bowl”

or:

“The oak tree by the pond down the path beside the cliff…”

These forms specify a referent adroitly, though in real oral communication the information is likely to be less densely packed:

“You know that cliff? Remember that path by the cliff? If you follow it you come to  this pond, …..”

Although there is an economy of sorts in dense recursion, it is often at the expense of parseability for the listener, so one has to wonder what the evolution modelers would enter as constraints and whether languages with this feature would be selected for. These seem at most slight economies, and not likely to confer survival advantage on the speakers.

Is there another kind of recursion that is more significant from the point of view of speaker advantage? Bickerton argues that syntax permitted off-line thinking, and retention of representations of events not in the here-and-now, but in the past, or the future, or hypothetical. In this he makes the same point as Newmeyer (this conference) that language has essential cognitive advantages. However, once language is not longer about the immediate situation, certain new problems arise in communication. 

It is not too far-fetched to argue that social creatures other than humans monitor the reliability of signalers. Vervet monkeys may be encoding referents (eagle, leopard, snake) in their cries, but monkey mothers can ignore the “mistakes” of their babies who make the cry for “eagle” (or whatever it means) at the sight of a falling leaf (18), and others claim that social primates might monitor the excitability and status of their group-mates in deciding whether to react to e,g. alarm calls(19). Animals may make these qualitative judgments, though it is unlikely they refer to notions such as truth or falsity of the call: reliability is all one needs. Young children are also attentive to the reliability of language-users and prefer to learn from the individual who previously showed themselves to be using the code in the same way as the child and her family (11). Nowak et al. take it as a fundamental premise that reliable signalers gain reproductive advantage.

It seems to me that syntactic recursion is most useful to mark that an event was experienced indirectly:

Mary says that a deer is by the tree.

The form makes use of recursion  to provide the content of Mary’s speech and mark it as her Point of View, allowing the speaker to evade responsibility for a false report. The form signals that the event was only attested through the speech of another. The recursive form above is not the only device, but it is the most common linguistically. Can it be done other ways? Here are several:

Type 1:

“The deer is by the tree.”

“Mary saw the deer.”

This relies on questioning about the source of evidence, and would add complexity to conversations if it had to be done continually.

Type 2:

              Enact as Mary (e.g. taking on her voice, her stance):  “The deer is by the tree”

This successfully marks the speech as uttering from another, but relies on mimicry and acting ability. Signed languages such as ASL adopt roles for the speaker, by adopting a different stance and speaking as the speaker, e.g. Mary. This is grammaticalized in ASL, but to my knowledge no spoken human language has been shown to rely solely on gesture or paralinguistic features like vocal quality to convey speaker identity.

Type 3:

            “The deer is (allegedly) by the tree”

By “allegedly” I am paraphrasing an evidential morpheme, which English does not have. Some languages have evidentials, indicating the source of knowledge, such as hearsay, or directly witnessed, or inferred. Again, it is unclear that any language relies solely on such a morpheme, rather, this is in addition to the usual mechanism of recursive complementation under a verb of communication. One can see the advantage of the latter: if speaker reliability is key, how does it help to indicate simply ‘hearsay’?

If my speculation is correct, a good case for the real utility of linguistic recursion may lie in describing testimony. If so, then languages with this property would confer an advantage to their speakers over languages that did not. The other advantages of recursion (e.g. economy and efficiency) come for free. Once the machinery is established for distinguishing point-of-view on testimony, it is but a short step to using that same machinery for representing beliefs.  Without it, no such representation is possible, however much talk one might hear about minds.

Why invent ‘think that”?

If Mary utters the sentence and there is no deer by the tree, what can be concluded? One could just write her off as unreliable, but there are in fact three possibilities that refine that conclusion:

A. If Mary is a child, perhaps one could doubt she can use the words deer, tree, is properly, rather like the mother Vervet ignores her young.

B If Mary is an adult, then there is a quandary: why is a speaker mis-using an already-learned code?

1. The first alternative is that Mary had something to gain by saying it. We use language not just to describe, but to persuade, deceive, cajole, seduce. Perhaps Mary has learned that when she says this the group moves away from the fire and vacates a nice warm spot for her. If Mary is using the code to deceive, this represents a violation of the rules of communication to which the group should be highly attuned and resist. (And teach their children about in cultural memes like ‘The Boy Who Cried Wolf!’)

2. In the second alternative, Mary “made a mistake”.  This is a sophisticated notion that relies on the notion of intention, but is probably within reach of our closest primate relatives and very young children. We don’t, however, always invoke the notion of false beliefs to account for mistakes. If I trip over a crack in the sidewalk, it would be odd to say “Jill thought the sidewalk was smooth”. We recognize that our perceptual and motor systems have certain error limits, and when we see someone else fall or fail to achieve a goal we recognize the goal and the persistence towards it, and the behavior as an error.  But what about when Mary says she saw a deer where there was none? An alternative that we might appreciate only if we have had like experiences is that she was

a) not failing to use the code properly

b) not deceiving us     but

c) perhaps she saw a shadow as a deer.

Seeing “as” is an expression that reflects the fact that there is interpretation in the visual system. This is where Simulation theory might have a point: I catch myself in the act of shouting “deer” when I re-see it as a shadow, and find that experience distinct from failing to know the right referent for ‘deer’. Having recognized that fine distinction, I project it onto Mary. But simply saying “Mary made a mistake” or Mary was wrong” glosses over these fine conceptual distinctions. We need to say, “Mary thought she saw a deer”.  At that point the representation is sufficient to allow computation, prediction and explanation.

It seems quite possible that this last step in the chain requires cultural transmission, i.e. the forms, once invented, were sufficiently useful to be retained across generations.  For example, it is possible that not every language has lexicalized the difference between ‘say” and “think”. But the tale I have spun here does NOT entail cultural transmission of testimony about mind, i.e. of the theory of mind itself. Instead, the theory – that people have different points of view- is “encapsulated” in the linguistic forms themselves, so a child will encounter them in a wide range of environments without requiring special instruction or situational encounters. Any language is the embodiment of inventions of hundreds of generations of ancestors, only some of whom were bright or lucky enough to have the right experiences and persuade others to follow their rhetorical style. Every new child who learns the language benefits anew from that collective wisdom. It is this kind of reasoning that persuaded Ken Hale to write about the death of languages as being as devastating as the death of species: we lose the collective inventiveness of a whole line of human beings when the last speaker dies.

The model I have proposed acknowledges:

a) The genetic basis of language structures.

b) That despite a), there are still language acquisition “problems” to solve.

c) That a mature theory of mind rests on the representational structures that language provides, so language is a fundamental aid to thought.

d) That it is through communication that the value of recursion to speakers might have first arisen.

References

1. Bickerton, D. (1995) Language and Human behavior. Seattle, WA: University of Washington press.
2. de Villiers, J.G. (in press). Can language acquisition give children a point of view? In J. Astington & J. Baird (Eds.) Why Language Matters for Theory of Mind. Oxford Press.
3. de Villiers, J. G. & de Villiers, P. A. (2000). Linguistic determinism and the understanding of false beliefs. In P. Mitchell & K. Riggs (Eds.), Children’s reasoning and the mind. Hove, UK: Psychology Press. pp.189-226.
4. de Villiers,J.G. & de Villiers, P.A. (2003). Language for thought: Coming to understand false beliefs. In D. Gentner & S. Goldin-Meadow, (eds) Language in Mind: Advances in the Study of Language and Cognition.  MIT Press.
5. de Villiers, J. G., & Pyers, J. E. (2002). Complements to cognition: A longitudinal study of the relationship between complex syntax and false-belief-understanding. Cognitive Development, 17, 1037–1060
6. de Villiers, P. A. (in press) The role of language in theory of mind development: What deaf  children tell us. In J. Astington & J. Baird (Eds.) Why Language Matters for Theory of Mind. Oxford Press.
7. de Villiers, P. A., Burns, F. & Pearson, B. (2003) The role of language in the Theory of Mind development of language-impaired children: Complementing theories. In B. Beachley, A. Brown, & F. Conlin (Eds.), Proceedings of the 27th Annual Boston University Conference on Language Development. Pp. 232-242. Cascadilla Press, 2003.
8. Gold, E.M.  (1967) Language identification in the limit.  Informat.Control, 10,  447-474.
9. Hale, C.M. & Tager-Flusberg, H. (2003). The influence of language on theory of mind: A training study. Developmental Science, 6, 346-359.
10. Hauser, M, Chomsky, N. & Fitch, W.T. (2002) The faculty of language: what is it, who  has it  and how did it evolve? Science, 298, 1569-1579.
11.  Koenig, M.A., Clément, F. & Harris,P. (in press). Trust in Testimony: Children's use of  true and  false statements. Psychological Science.
12. Lohmann, H. & Tomasello, M. (2003). The role of language in the development of false belief understanding: A training study. Child Development, 74, 1130-1144.
13. Nowak, M., Komarova, N & Niyogi, P. (2001) Evolution of universal grammar. Science,  291, 114-118.
14. Nowak, M., Komarova, N & Niyogi, P. (2002) Computational and evolutionary aspects of language. Nature, 417, 611-617.
15. Plotkin, J & Novak, M. (2001) Major transitions in language evolution.  Entropy, 3, 227-246.
16. Pyers, J. (2001) Three stages in the understanding of false belief in Nicaraguan signers: The interaction of social experience, language emergence, and conceptual development. R. Senghas (Chair), The Emergence of Nicaraguan Sign Language: Questions of development, acquisition, & evolution. Invited symposium conducted at the 31^st Annual Meeting of the Piaget Society, Berkeley, CA.
17. Roeper, T. & Synder, W. (in press) Language Learnability and the Forms of Recursion. In Di Sciullo, A. M. and R. Delmonte (Eds.) UG and External Systems. Amsterdam: John Benjamins.
18. Seyfarth, R.M. & Cheney, D.L. (1986) Vocal development in vervet monkeys. Animal Behavior, 34:1640-58.
19. Smith, W, (1977) The Behavior of Communicating. Cambrudge, Ma: Harvard University press.
20. Tager-Flusberg, H. & Joseph. R. (in press).  How language facilitates the acquisition of false belief in children with autism. In J. Astington & J. Baird (Eds.) Why Language Matters for Theory of Mind. Oxford