Introduction to Part V: Language change, creation, and transmission in modern humans
Abstract and Keywords
This article covers the processes of (modern) language creation and change, and the roles played in language evolution by socio/cultural transmission. Bernd Heine and Tania Kuteva report on the uses of well-known processes at work in observable language change to reconstruct a plausible scenario for the development of the earliest languages. Joan Bybee reports on the concept of grammaticalization that refers to a bundle of processes causing diachronic change that are known to occur in all languages. Grammaticalization is posited as a critical driving force in the evolution of language, and grammaticalization theory gives us a scientific tool for reconstructing earlier linguistic states. Paul Roberge argues against the prevailing view concerning the role of child learners in language change in connection with the formation of creoles. He argues that native acquisition of pidgins is not necessary to form creoles, which are full linguistic systems. Roberge compares the factors leading to the evolution of full language from protolanguage with the factors involved in the formation of pidgins and creoles. Susan Goldin-Meadow reports on the theme of language creation. She explores the role of the manual modality when used alongside speech, and then investigates what changes occur when this modality fulfils all the functions of language, without speech. Sign languages are fully-fledged languages, but more primitive gestural communication occurs in homesign systems.
The chapters in Part V move beyond the core features of the language faculty that are the focus of Part IV, discussing the processes of (modern) language creation and change, and the role played in language evolution by socio/cultural transmission. Another factor affecting language change is population movements, since contact between different populations can result in massive structural changes to languages, and indeed, can trigger language shift. Authors investigate what light these modern ‘windows’—creation, transmission, and contact events which can be directly examined—shed on language evolution itself. A number of chapters discuss a type of evidence that is relatively new to evolutionary linguistics: computational, mathematical, and robotic modelling.
(p. 506) Various themes recur in both Part IV and Part V of the volume. How much of the human language faculty should be ascribed to a putative Universal Grammar (i.e. a genetic endowment for language learning)? How much can be accounted for by other mechanisms, including sophisticated statistical learning? How much of observed language structure should be ascribed to natural selection, in other words what properties of (proto)language were adaptive as hominins started along the trajectory towards linguistic communication? Alternatively, how much structure arises spontaneously as a result of processes of self‐organization, including linguistic changes that adapt to the learning mechanisms best suited to the human brain? Our view is that there are really no dichotomies here, and that in all cases, complex gene–environment interactions give rise to linguistic behaviours and structures. For instance, it is likely that external functional pressures have, in part, shaped linguistic structure; a well‐studied example concerns principles of linearization of phrases and sentences (see Chapter 1 for some discussion). However, such functional pressures are of course not ‘external’ in any way: they arise from the way humans process the world and learn about its properties, and are, in most cases, phylogenetically primitive. We would expect the types of learning mechanisms occurring in our hominin ancestors to have directly shaped the way that languages developed, and in turn, assuming that an evolving language faculty is adaptive, we expect that early humans evolved in ways that would better enable them to learn ambient languages quickly, and making the most of a fragmented input.
Chapter 54, by Bernd Heine and Tania Kuteva, uses well‐known processes at work in observable language change to reconstruct a plausible scenario for the development of the earliest languages. ‘Grammaticalization’, a concept also discussed by Joan Bybee (Chapter 55), refers to a bundle of processes causing diachronic change that are known to occur in all languages. Over time, lexical forms (such as nouns and verbs) turn into grammatical forms (such as auxiliaries), and grammatical forms turn into forms that are even more grammatical (morphosyntactic items, such as tense markers). These developments are directional—for instance, an auxiliary is not expected to become a lexical verb. Since these processes are ubiquitous in observed language change, linguists propose that they also played a vital role in the initial stages of language evolution. The earliest (proto)languages thus were probably simple concatenations of lexical (proto)words, lacking exponents of grammatical categories such as tense and aspect, case and agreement, passive, and so on. Many propose that nouns and verbs are the two word classes from which all other classes develop (or rather, proto‐nouns and proto‐verbs, since these categories would not initially have the fully‐modern set of properties), but Heine and Kuteva suggest that noun‐like words were the first category, and that all other classes, including verbs, derive from nouns. Grammaticalization is posited as a critical driving force in the evolution of language, and grammaticalization theory gives us a scientific tool for reconstructing earlier linguistic states.
(p. 507) Bybee's chapter extends this theory to constructions. Once words start to be grouped together in sequences that are repeated, ‘chunking’ occurs; units are formed in the memory that can be accessed directly. Novel constructions sometimes involve grammaticalization, meaning that a new grammatical morpheme arises (for instance, the future gonna construction, deriving from (literal) going to), but not all constructions create new grammatical elements (an example is the drive X crazy construction). Bybee argues that capacities for acquiring and processing language are mostly not domain‐specific, but rather, are domain‐general processes, applying to other types of learning. Moreover, language change mostly occurs in language use by adults, and not during language acquisition by children, as has often been proposed.
Paul Roberge also argues against the prevailing view concerning the role of child learners in language change, here in connection with the formation of creoles. He argues that native acquisition of pidgins (which are restricted systems, not full languages) is not necessary to form creoles, which are full linguistic systems. Roberge compares the factors leading to the evolution of full language from protolanguage with the factors involved in the formation of pidgins and creoles. Modern instances of language creation should shed light on language evolution itself—in Botha's terms, they form a ‘window’. Certain ‘primitive’ linguistic features typify pidgins, such as free concatenation of words, and pragmatically rather than syntactically regulated word order; these are also posited as properties of protolanguage (see also Tallerman, Chapter 51). Creators of pidgins seem to resort to the pre‐grammatical mode of protolanguage, thus putatively reflecting the earliest hominin use of language. However, as Roberge notes, modern speakers already all possess full language; does this influence (pidgin) language creation?
The theme of language creation also appears in Susan Goldin‐Meadow's chapter (57) on gesture. She first explores the role of the manual modality when used (as it is in all cultures) alongside speech, then investigates what changes occur when this modality fulfils all the functions of language, without speech. This situation arises in existing and newly emerging sign languages. Sign languages are fully‐fledged languages, but more primitive gestural communication occurs in homesign systems. These ad hoc creations are used between deaf children and hearing (non‐signing) parents—thus, they are pidgins of a sort. Like pidgins, they are not full linguistic systems, but they exhibit some intriguing language‐like properties, including segmentation and combination. Also like pidgins, homesign systems may become fully‐fledged languages, if conditions for socio/cultural transmission are met. A well‐documented example is the emergence of Nicaraguan Sign Language, which became a full sign language when deaf children with different homesign systems were brought together in a school, and had to communicate for the first time outside of their families. In evolutionary terms, Goldin‐Meadow argues that the vocal and manual modalities may have evolved simultaneously, complementing each other as they do today.
(p. 508) The next two chapters further examine the role of contact (and lack of contact) between groups of speakers in creating language diversity and language change. Johanna Nichols (Chapter 58) notes that the principles of historical reconstruction only allow linguists to go back to around the last 7000 years, whereas language itself is assumed to have arisen at least 100,000–200,000 years ago, and pre‐language (what is elsewhere in this volume termed ‘protolanguage’) had probably been evolving for at least a million years. What, then, can examination of modern situations tell us about the origins of language? Nichols shows that, given the plausible size and dispersal of early human populations in Africa, extensive linguistic diversity must always have been the norm, just as it is today in New Guinea and in Northern Australia. There is no single ancestral language or, indeed, protolanguage. Nor does any evidence indicate that all non‐African languages descend from a single ancestral language, as might be expected had there been a single, fairly recent Out‐of‐Africa diaspora.
Brigitte Pakendorf (Chapter 59) looks at instances of language change via contact, but rather than contact between small, relatively isolated groups of individuals in pre‐history (Nichols), or of the kind that occurs today in emerging sign languages (Goldin‐Meadow), her chapter investigates what happens when population‐based contact occurs. In the most extreme case, the result is language shift: a population formerly speaking one language will switch to a new language entirely. Intriguingly, if linguistic studies of contact situations are married with molecular anthropology, then as Pakendorf's work shows, prehistoric language shifts can be detected via genetics. Pakendorf outlines details of a case study from Siberia, showing how evidence from language and from genes in a modern population—and the mismatches between the two types of evidence—can reveal prehistoric population contacts resulting from immigration.
The next four chapters (Smith, Kirby, Cangelosi, and de Boer) all investigate language creation and transmission, particularly the role played by social interaction. All four chapters discuss the evidence that can be obtained from mathematical, computational, and robotic modelling in evolutionary linguistics. Kenny Smith (Chapter 60) argues that verbal reasoning can only take us so far in working out the predictions of a theory, and that formal models can test assumptions in a scientific manner. This is particularly useful in language evolution theory, where direct evidence is almost non‐existent. Smith explains how formal models are used, and why it is that the results themselves are built into the models; this is not a drawback, but the entire point of using a model—it allows the researcher to test assumptions by varying the parameters of interest, then seeing what results from the model.
Simon Kirby (Chapter 61) uses formal modelling to examine the way that language is passed on via socio/cultural transmission. Kirby assumes that a genetically‐specified language faculty exists (i.e. there is some innate capacity for language). Of course, individual languages do not emerge spontaneously from this (p. 509) language faculty, but rather are transmitted between generations of a speech community. To understand language evolution, we must consider the interaction between three distinct systems: biological evolution of the language faculty in the species, socio/cultural transmission of languages in populations of speakers, and individual learning in the child, which involves building a grammar on the basis of language data received. An interesting extension to computer modelling involves cultural transmission in the laboratory, using real human subjects and a simplified artificial ‘language’ that participants transmit. Kirby's results show that the very process of transmission (iterated learning) can lead to the appearance of design in language‐like data, without either natural selection (or an intelligent designer) having intervened. Kirby concludes that languages have themselves evolved to be learnable in this way: they are in fact adaptive systems.
Angelo Cangelosi's chapter (62) on the use of robots and embodied agents (i.e. simulated agents) in modelling considers the implications of these experiments for the evolution of signalling behaviour and coordination, the lexicon, reference, and syntactic compositionality. Central questions involve the initial capacities for social coordination and prelinguistic communication before any specific communication channel evolved; and also the emergence of a shared lexicon in a situation where the agents initially share no meanings. The results show that limited verbal communication can arise under these circumstances, via social interaction. Thus, the importance of socio/cultural transmission is once again emphasized.
These themes of interaction between population‐level and individual learning, and the spontaneous emergence of order in a system are also taken up by Bart de Boer (Chapter 63), in his contribution on self‐organization. As de Boer notes, if certain complex linguistic structures can be shown to result from self‐organization, then we need not ascribe to them a genetic basis. Many factors must still be explained by biological evolution, but these generally involve simpler behaviours, more likely to be dependent on widespread cognitive capacities also found in other animals. Reducing the involvement of biological evolution is desirable, and helps to explain how language may have evolved relatively rapidly from simpler cognitive precursors. Once a system emerges from self‐organization, then the usual biological mechanisms select for adaptations which enable the user to handle the resulting system. Impressive results concerning self‐organization come from investigations of sound systems, as de Boer describes; for instance, the structure of small to medium vowel systems can be explained on the basis of acoustic distinctiveness. Self‐organization also occurs at population level, and can help explain what aspects of language structure emerge during transmission from one generation to the next, as both de Boer and Kirby discuss.
A process with similarities to self‐organization is statistical learning, the process of detecting existing structure by tracking patterns in the input. Of course, this process also involves socio/cultural transmission, since infant learners receive language input from competent speakers. Katharine Graf Estes (Chapter 64) (p. 510) discusses how, via statistical learning, humans of all ages can track regularities in sounds, words, and grammars on the basis of distributional information. For instance, how are words segmented by infants from the incoming speech stream, where word boundaries are not explicitly flagged up? How is it that adult speakers of a language generally share grammaticality judgements on syntactic data? At 8 months, infants are capable of tracking the statistical cues that enable them to start to detect words. One‐year‐old infants can distinguish grammatical from ungrammatical word orders. And infants of 18 months can learn non‐adjacent dependencies. Graf Estes concludes that ‘[h]umans readily learn regularities that commonly occur across human languages, but have greater difficulty learning regularities that do not’. If linguistic structure of various kinds can be acquired on the basis of statistical learning, the role played by an innate universal grammar may be reduced or eliminated. However, the learning abilities discussed by Graf Estes must all have evolved in their turn, and many of them may well be species‐specific. Thus, the predisposition to acquire linguistic structure on the basis of statistical regularities can itself be seen as a central component of the language faculty; see also Anderson; Boeckx in Part IV.
The kinds of learning mechanisms discussed by Graf Estes are brought to centre stage by Nick Chater and Morten Christiansen (Chapter 65), who propose that abstract and arbitrary features of language (for instance, agreement, or case‐marking) are shaped entirely by the processes involved in socio/cultural transmission, rather than by a genetically specified universal grammar. They develop Graf Estes’ view that features which are common cross‐linguistically are easily learned. The idea is that properties which humans cannot learn readily or which hinder communication will be discarded by languages, whereas features that are easily learned and processed, and which are communicatively effective, will be retained and expanded. Thus, languages themselves evolve to fit the kind of learning best suited to the human brain. A central theme in this chapter is that language change is too fast for genetic change to keep up with, thus eliminating the possibility that abstract linguistic features are shaped by natural selection; see, however, Számadó and Szathmáry (Chapter 14) for a different view. For Chater and Christiansen, since language is shaped by socio/cultural transmission, the same processes of language change and creation which are explored in other chapters in Part V also underpin language evolution. However, there is in addition a central role for biological and cognitive constraints in this model: ‘the processes of cultural transmission that have shaped the creation of natural languages [are] grounded in prior human neural and cognitive capacities’. This also means that natural selection can be involved in shaping functional aspects of language (for instance, duality of patterning).
In conclusion, specific human learning and processing capacities (which, of course, have themselves evolved) are the ones that languages must adapt to, and in turn, human learning biases guide the rapid acquisition of languages (this last property being the main impetus in generative grammar for proposing a dedicated (p. 511) universal grammar). The construction of the human niche (see Gibson and Tallerman, Chapter 12, for some discussion) has undoubtedly required adaptations to an evolving language faculty, and the extreme phenotypic plasticity which characterizes humans in particular has also undoubtedly played a critical role. The investigation of ongoing processes of language creation and change, both in the field and in the computer lab, is becoming an increasingly important subfield, and should hopefully lead to a deeper understanding of the processes involved in language evolution itself.