Language as a complex adaptive system

ResearchBlogging.orgA prominent idea in linguistics is that humans have an array of specialised organs geared towards the production, reception and comprehension of language. For some features, particularly the physical capacity to produce and receive multiple vocalizations, there is ample evidence for specialisation: a descended larynx (Lieberman, 2003), thoracic breathing (MacLarnon & Hewitt, 1999), and several distinct hearing organs (Hawks, in press). Given that these features are firmly in the domain of biology, it makes intuitive sense to apply the theory of natural selection to solve the problem: humans are specially adapted to the production and reception of multiple vocalizations.

In his 1980 book Rules and Representations, Noam Chomsky extends this notion further, arguing that humans also contain specialised mental organs, or modules, for acquiring language. Tracing the history of this argument stems initially from the widespread assumption of generative grammar that all languages are essentially the same in structure, but differ in their sound systems and vocabularies (Evans & Levinson, 2009). To quote Pinker (1994), “According to Chomsky, a visiting Martian scientist would surely conclude that aside from their mutually unintelligible vocabularies, Earthlings speak a single language.” (pg. 232).

As such, under the stewardship of Pinker, Chomsky and others, the origin, evolution and acquisition of language is primarily seen as a biological question to be answered. Whilst it is certain that biology plays a role in the evolution of language, its exact purpose is still contentious in light of a whole host of arguments emerging from research into cognitive, developmental, cultural and social traditions. Not only are the foundations of language universals (Evans & Levinson, 2009), natural language as a formal language (Deacon, 1997) and poverty of the stimulus (Pullum & Scholz, 2002), being challenged at both a theoretical and empirical level, but the recent resurgence of social and cultural factors in explaining language structure (Smith & Kirby, 2008; Lupyan & Dale, in press), language change (Kirby, 2000) and language acquisition (Chater & Christiansen, in press) are providing independent and alternative theories in our understanding.

In highlighting the differences in structure across the spectrum of languages, the renewed emphasis of socio-cultural linguists is to explain these differences. Lupyan (in press), and others (Wray & Grace, 2007), have reported a distributional pattern of structures across the world’s languages, with “strong relationships between linguistic factors related to morphological complexity and demographic/socio-historical factors such as the number of language users, geographic spread, and degree of language contact.” (Lupyan & Dale, in press, pg. 2). Specifically, both Wray & Grace (2007) and Lupyan & Dale (in press) make a distinction between two types of instances in which languages are learned and used: exoteric and esoteric niches. An exoteric niche contains languages with a large variety of speakers, thus a pressure is asserted on the communal language to be suited for communication between strangers (Lupyan & Dale, in press). English, Swahili and Hindi are all examples of languages emerging from exoteric niches, in that they are more likely to “(1) be non-native speakers or have learned the language from non-native speakers, (2) use the language to speak to outsiders – individual from different ethnic and/or linguistic backgrounds” (ibid, pg. 7). Conversely, esoteric niches are composed of languages like Tatar, Elfdalian and Algonquin (ibid). All of these languages belong to relatively small populations – individuals are part of a tightly knit community, based on a shared cultural and social identity.

Given these differing linguistic niches, Lupyan and Dale’s main proposal is that the morphological features of a language are the product of adaptation to learning constraints and the communicative requirements of the speaker population. Therefore, a language being subjected to a greater number of outside learners, for instance, via colonization or large-scale migration, is thus under a greater pressure to become learnable – and subsequently simplifies its morphology and increases its productivity of existing grammatical patterns (ibid). Languages in exoteric niches also become more analytical, which according to Wray & Grace (2007) increases their compositionality in that “meanings of expressions can be determined from their composition, because the system approximates a one-to-one relationship between forms and meanings.” (pg. 9). Furthermore, Wray & Grace (2007) claim given the dynamics of these niches, language must have originally evolved in esoteric communities. On the basis of these assertions, Carstairs-McCarthy (2005) notes: “It therefore becomes an open question whether what linguists take for granted as grammatical universals (even such fundamental features as recursion) may be not biologically based but rather cultural add-ons, resulting from millennia of increasingly exoteric language use.” (pg. 508).

Geographic distribution of the 2,236 languages included in Lupan and Dale's study
Geographic distribution of the 2,236 languages included in Lupan and Dale's study

Given the arguments presented thus far, it is becoming increasingly indisputable that socio-cultural factors are at work in accounting for linguistic structure. Suffice to say, this certainly weakens the position of language being purely a cognitive phenomenon, yet it does not completely rule out the possibility of an innate acquisition devise being responsible for the structural properties of language. One such problem in removing the dependency on solely biological explanations is the logical problem of language acquisition: that children become competent users of language, despite having an incomplete and noisy input (Kirby, 1999; Chater & Christiansen, in press).

According to Chater & Christiansen, an alternative to the competing views of adaptationists and non-adaptationists is that the “fit between the neural mechanisms supporting language and the structure of language itself is better explained in terms of how language adapted to the human brain, rather than vice versa.” (2009, pg. 3). Specifically, they adopt the general view of language being a product of multiple constraints; an evolving system, with features of language emerging from repeated processes of acquisition and transmission across continuous generations of language users (see Deacon, 1997; Kirby & Hurford, 2002; Tomasello, 2003; Christiansen & Chater, 2008).

Therefore, the cognitive architecture underlying the processing of language is likely to be the product of natural selection and developmental processes. For instance, pre-linguistic features such as sequential processing (Christiansen & Delvin, 1997) and working-memory (Fiebach et al., 2005) undoubtedly influence our processing of language, yet it is unlikely these features were exclusively selected for language – rather, they are domain-general and probably selected on their functional flexibility (Chater & Christiansen, in press). This somewhat moves away from a common delineation offered by Kirby & Hurford (2002), with language sitting at three complex, dynamical systems: biological evolution, cultural transmission, and individual learning. As Ferdinand & Zuidema (2008) highlight, this tripartite separation is partially misleading “[…] because it implies that evolution acts directly on learning as an adaptive system. This view essentially deletes cognition from the picture, because it is the embodied cognitive agent that ultimately roots its high-level process of language induction within the biologically evolved wet-ware that is the true processor of language.” (pg. 3). Still, Kirby & Hurford’s central tenant remains: language is not solely a biological problem to probe, but rather a conundrum comprised of many factors.

Thus far, the argument laid out is that the constraints of language arise from two systems: “the embodied cognitive agent and the socio-cultural system in which these agents communicate with one another.” (Ferdinand & Zuidema, 2008, pg.2).  Amalgamating these perspectives of language being shaped by human interaction and domain-general cognition is a strong argument for moving away from the static system of grammatical principles embodied by the generativist position, and towards a new paradigm of language being a complex adaptive system (CAS) (Beckner et al., 2009).

First coined by Holland (1998), CAS are like other complex systems in that they exhibit emergent properties as a result of multiple interconnected elements. What differentiates CAS is its ability to adapt: these systems can learn from past experiences, and subsequently change their behaviour as a result (ibid). Such instances of CAS are found throughout nature, from social insects and ecosystems to immune systems and cellular mechanisms (Ahmed, Elgazzar & Hegazi, 2005). More importantly, these adaptive systems are used to explain the properties of human social endeavours, including: stock markets, communities and political parties (ibid).

Language displays all the hallmarks of being a CAS: “(1) The system consists of multiple agents (the speakers in the speech community) interacting with one another. (2) The system is adaptive, that is, speakers’ behavior is based on their past interactions, and current and past interactions together feed forward into future behavior. (3) A speaker’s behavior is the consequence of competing factors ranging from perceptual mechanics to social motivations. (4) The structures of language emerge from interrelated patterns of experience, social interaction, and cognitive processes.” (Beckner et al., 2009, pg. 3).

Critically, this view places the existence of language at two inter-dependent junctures, consisting of an idiolect (the individual language user) and the communal language (the community of users). Both of these aspects are emergent, with an idiolect emerging from each individual’s use of their language through interactions with other individuals enmeshed in the community, and the communal language being the product of the interactions of all the idiolects (ibid).

These two distinct but connected levels provide a myriad of features and explanations in truly understanding language as a whole, dynamic system. For instance, the inherent diversity between individuals dictates that there is no idealised notion of a speaker-hearer, with each individual’s unique linguistic experience resulting in heterogeneous idiolects (Weinreich, Labov, & Herzog, 1968; Bybee, 2006). But this is not all – language is in a constant state of flux at both the level of communal and idiolect (Beckner et al., 2009); it adapts through an upward spiral of competing factors (see speaker-listener conflict: Chater & Christiansen, in press); small phenotypic differences in humans (vocal tract control, shared attention, memory capacity, etc) can build up and result in a phase transition (Elman, 2005); the immersion, sensitivity and dependency of language users in social networks (Lupyan, in press); and the view of language being a form of cultural adaptation to both the human mind (Christiansen & Chater, 2008) and the transmission vector (Cornish, Tamariz & Kirby, in press).

Thus, seeking a solitary exposition for language is not a fruitful, or even possible, task. That so many features are intertwined in an inextricable system is something to see as liberating, instead of disparaging. For instance, by exploring language through socio-cultural and domain-general cognition, a picture emerges where language is adapting to the vast linguistic environment in which it is being learned and used (Lupyan & Dale, 2009). As such, some of the features of language may be explained via multiple constraints stemming from “different learnability and communication pressures.” (ibid, pg. 18). Languages then, not only undergo a process of gradual drifting until they become mutually unintelligible, they also adapt in response to any significant pressures that are present. Our role as linguists is to see what constraints play a significant role in shaping the features of language. Two obvious candidates are culture and the transmission vector, from which we can pose this question: how does cultural transmission influence, and interact with, language? It is this question which I will attempt to answer in a post about my own research.

Main References
Beckner, C; Blythe, R; Bybee, J; Christiansen, M.H.; Croft, W; Ellis, N.C.; Holland, J; Jinyun Ke; Larsen-Freeman, D; Schoenemann, T (2009). Language is a complex adaptive system Language Learning

Lupan, G. & Dale, R (2009). Language structure is partly determined by social structure Social and Linguistic Structure

WRAY, A., & GRACE, G. (2007). The consequences of talking to strangers: Evolutionary corollaries of socio-cultural influences on linguistic form Lingua, 117 (3), 543-578 DOI: 10.1016/j.lingua.2005.05.005

5 thoughts on “Language as a complex adaptive system”

  1. “An exoteric niche contains languages with a large variety of speakers, thus a pressure is asserted on the communal language to be suited for communication between strangers”

    Wouldn’t the fact that children must be both initiated into language and culture (they are in fact strangers), mean that there are exoteric pressures exerted on every language? The facility children have with languages may mean that the pressures are less, but would this be enough to force human grammars to be mostly compositional, mostly context-free, etc., even if not more analytic, as wray and grace show?

  2. Thanks for the comment. I wholeheartedly agree with your assertion: that a child’s initiation into their language and culture does provide pressures to become increasingly learnable, and subsequently compositional. In fact, this is what we are studying here at Edinburgh (see Kirby, Cornish & Smith, 2008) and is the subject of my next post, which is why I neglected to focus on it in this one.

    Still, I think there is an important difference between adult strangers and child strangers (excuse the very dodgy terminology): the former is a product of many accumulated biases, each of which having been moulded and constrained over the lifetime of the individual, whilst the latter is introduced to their linguistic environments with relatively relaxed biases. So, in an esoteric linguistic environment, children born into this culture are very likely to have their biases conform to the current population, and as such the pressures may not be strong enough to shift the entire population towards a new niche. However, when large groups of adults and children from different linguistic and cultural environments are mixed together (exoteric), then there is a significantly greater pressure for the languages to adapt – both the exoteric environment and transmission vector are acting in concert to make the language more learnable. Whether or not the transmission vector alone is enough to achieve a significant shift towards compositionality is yet to be established. I think it’s most likely that both factors are at work, and it will require experimentation and more research to clarify which was more significant.

    I think this answers your question, but feel free to clarify. Thanks again.

  3. Hi Gary, sorry about the late reply — I’ve been away from the blog for a while. Congratulations on the paper btw. It’s generated quite a bit buzz in some circles. In particular, I was impressed with a post over at The Adventures of Auck.

Leave a Reply