The origins of language: Akhaṇḍakhaṇḍanam

My title refers to a recent post on Language Log by Herb Terrace, which is a response to an article by Hauser et al. that appeared in Frontiers in Psychology last year. To review briefly: Hauser et al. argue that, despite claims to the contrary, we simply have no idea how human language originated. They define language as “abstract units of information that are organized and combined following specific computational procedures,” and thus putting those procedures into evidence is necessary for any account of human language evolution. As it turns out, this is really hard: none of the most recent evidence from comparative animal behavior, paleontology and archaeology, and molecular biology brings us any closer to understanding the origins of language in this strict sense, which appears only with anatomically modern humans (and evidently not Homo neanderthalensis). Terrace characterized Hauser et al.’s position as “syntactocentric,” and that there are many other behaviors besides the computational procedures of syntax that are implicated in, and accompany, language use: for example, mutual eye gaze (whether “dyadic,” i.e. between the mother and child, or “triadic,” i.e. the mother and child’s gaze is jointly directed toward objects of events), and along with it, the use of visual cues to express and interpret emotional states. The ability to “infer referential intent” is presupposed by the use of “declarative vocabulary,” and this in turn is presupposed by the combination of words into complex units; these are logical presuppositions (I think), but Terrace suggests that the evolution of language had to retrace these steps.

What does any of this have to do with Indian Philosophy? The “syntactocentric” view represents a kind of akhaṇḍapakṣa, in a limited sense: there is no linguistic unit that is not composed of smaller parts arranged according to computational rules; a linguistic unit consists, above all, in the way that its constituent parts are put together, and not in the parts themselves; thus it is akhaṇḍa, indivisible. Terrace represents a khaṇḍapakṣa, according to which language requires discrete units, namely words, that can be used apart from the rules that generate units of greater complexity; language is thus khaṇḍa, divisible. These map onto two opposed views regarding human language origins: a punctuated, “all-or-nothing” position that identifies the origins of language with the evolution of a “Faculty for Language in the Strict Sense” in Homo sapiens, and a gradualist position that has the “Faculty for Language in the Strict Sense” follow upon other, more basic and more widely shared, social and cognitive behaviors.

It’s worth thinking about how these positions might have sounded to participants in premodern Indian debates about the meaning (and compositionality) of linguistic units. There was widespread agreement that language acquisition (vyutpatti) has to follow a bootstrapping, gradualist path: the child learns the meaning of the linguistic unit “bring the cow” by comparing and contrasting his observations with observations in other cases (e.g., “bring the horse”), in a process called anvayavyatireka. And I think Hauser et al. would have to agree with this, too. The disagreement relates to how we human beings ever found ourselves in the position of saying “bring the cow” at all: the production of such sentences requires the use of the complex syntactic rules that Chomsky (one of Hauser’s co-authors) has detailed, and the speed of language acquisition suggests that these rules in some form actually guide the acquisition process and aren’t completely induced from examples, somewhat in contrast to the Indian model. While Buddhists were generally content to say that language was conventional, Brahmanical authors were generally skeptical about the possibility of humans collectively establishing language conventions on their own. Thus Mīmāṃsakas argued that language does not have origins, and Naiyāyikas argued that it originated with God. Although these positions aren’t likely to appeal to us, I think they share a lot of motivation with people who have the “syntactocentric” view of language sketched above (mostly linguists as opposed to psychologists): without complex syntactic rules, we have no language, but individual-level language acquisition in some sense presupposes those rules, and perhaps species-level language development does too, something akin to flipping an evolutionary switch. The khaṇḍapakṣa, with its declarative use of words (“horse!”), gets us part of the way there, but Indian thinkers generally did not give much credence to these declarative functions, given the widely-acknowledged constraint that language use needs to be purposeful (most prominently among the Prābhākara Mīmāṃsakas).

One thought on “The origins of language: Akhaṇḍakhaṇḍanam

  1. Thank you for this very insightful post, Andrew.
    I would add that the position of Hauser et al. (as represented by you) also evokes the Mīmāṃsā tenet that language cannot have originated out of someone’s deciding to set up a convention, since in order to arrange for that convention to work, one would have already needed words, i.e., language. In other words, Hauser et al. seem to share with Mīmāṃsā the idea that language is not incrementally there, a claim which also harmonises with the fact that among actual languages we cannot speak of “primitive” languages (unless we believe that there is a golden language, say, Latin and that whatever fails to achieve it is “primitive”, but this idea has, luckly enough, declined long ago) and that language is either extant (in human beings) or not, with little intermediate steps which could explain its gradual acquisition. (Personal note: I am fascinated by the topic of language among non-human animals, especially Cetaceans, particularly insofar as this might imply some syntax, but I/we do not know enough about it to situate within a gradualist perspective).

Leave a Reply

Your email address will not be published. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>