14.1 Introduction
In the course of learning language, children make a number of systematic errors. Explaining these errors provides challenges for all theoretical approaches. From the perspective of an innate universal grammar (UG), explanations centre on the problems of mapping between UG and specific features of the language that children are learning. From a usage-based (UB) approach, explanations centre on interactions between learning processes and features of the input, including, importantly, the relative frequencies of different strings (see Chapter 2 by Baayen et al. and Chapter 3 by Hilpert, this volume).
Since diachronic language change involves variability, language-learning children and the systematic errors that they make have been seen by some researchers as the source of that variability. This is particularly the case from a UG perspective, partly because of the idea that the core UG cannot change and partly because of the idea of a critical period for language acquisition after which the parameters of UG are set. However, UG approaches to diachronic change encompass a wide range of different positions. In some, different children may entertain different parameter settings in their grammars, and over time, as the number of children with the changed parameter setting increases, so the language changes (see, e.g., Lightfoot Reference Lightfoot2006). In others, while children are the source of language change, this occurs in situations of second language learning and multilingual contact (see Meisel, Elsig and Rink Reference Meisel, Elsig and Rinke2013; Kroch Reference Kroch, Baltin and Collins2001,) rather than with children who are growing up in monolingual environments (for an evaluation of both views see Thomason Reference Thomason, Schreier and Hundt2013 and for other approaches that come to a similar conclusion see Hickey Reference Hickey, Schreier and Hundt2013; Cheshire, Kerswill, Fox and Torgesen Reference Cheshire, Kerswill, Fox, Torgersen, Schreier and Hundt2013).
Usage-based accounts start from the position that children learn language from re-using what they hear, in smaller or larger chunks. In the UB approach, linguistic categories such as noun, verb, noun phrase, subject and object are not pre-given but emerge as the child constructs language. Early on, children’s constructions will consist not only of single words but also of ‘big words’, i.e. rote-learned, unanalyzed strings of words (e.g. Whassis?, Whatchadoing?, Dąbrowksa Reference Dąbrowska2000; see also Peters Reference Peters1983; Lieven, Pine and Dresner Barnes Reference Lieven, Pine and Dresner Barnes1992; Bannard and Matthews Reference Bannard and Matthews2008; Arnon and Snider Reference Arnon and Snider2010; McCauley and Christiansen Reference McCauley and Christiansen2014) or of stems with a specific array of morphemes. The development of lexical categories is tied to children starting to develop low-scope, slot-and-frames patterns based on the frequencies in the input. Examples from English are What’s X do-ing?, It’s X-ing, I want a Y, That’s a Z. The slots in these patterns are the basis of emergent categories, initially of low-semantic scope such as ‘THING’ or ‘ACTION’, but showing increasing evidence of abstraction. In principle, morphological learning could follow the same pattern, with children first developing slot-and-frame templates from which morphological categories emerge. Thus, grammar is learned through a continuous process of abstraction. Constituency and more complex syntax emerge through this process. Frequency at different levels (for instance the effects of type and token frequency) is critical in driving this process, and large numbers of studies, not only for English, have found that frequency in the input is closely associated with what children learn (Ambridge, Kidd, Rowland and Theakston Reference Ambridge, Kidd, Rowland and Theakston2015).
A second important thrust of the UB approach is that most or all of early language learning can be explained by general cognitive processes (e.g. working memory, processing speed, the development of prototypes) rather than by any syntactically dedicated (innate) factors. Thus the roles of frequency, salience, processing speed and memory are crucial to UB explanations of language learning and development. In what follows I will demonstrate how these factors interact to generate the systematic errors we see in children’s language development.
From a usage-based perspective, errors either arise from the use of a rote-learned string or a low-scope schema in an inappropriate context (e.g. errors such as Why does he doesn’t like peas? which are suggested to arise from combining a Why does X schema with a declarative statement of what the child wishes to question: He doesn’t like peas), or from the use of an item in a schema with which it is less than optimally compatible (e.g. generalising an intransitive into a transitive schema: You cried her for ‘You made her cry’, for details see Ambridge and Lieven Reference Ambridge, Lieven, MacWhinney and O’Grady2015: sections 6.3 and 7.1.2). During development, there is competition in a network of form-meaning mappings, some target-like and others not. The non-target-like errors are eventually out-competed. Below, I give a number of examples of research illustrating these processes. I will deal with the evidence for variability in early syntactic development in terms of the errors or departures from the adult system that children make. The question is whether these errors could be the source of diachronic change. I will argue, along with many researchers of all theoretical persuasions, that young, monolingual, language-learning children recover from these errors and that, therefore, they are highly unlikely to be the source of diachronic change. I will examine a number of systematic errors made by children and consider the explanations provided for them before returning to the issue of any relationship between errors in language development and diachronic change.
14.2 Errors
When deciding whether or not an error is systematic or, alternatively, can just be attributed to online processing glitches or ‘noise’, it is important to realise that this depends crucially on the level of granularity at which the error is counted. To give a telling example, Aguado Orea and Pine (Reference Aguado-Orea and Pine2015) analysed two Spanish children’s early development of verbal morphology. The study found that, although the overall error rate of person-marking in present tense was low at around 4.5 per cent, there were pockets of very high error rates (for instance, the marking of third-person plural was wrong 31 per cent of the time for one child and 67 per cent for the other). When the data were analysed verb by verb, the error rate went up inversely with the relative frequency of each verb form. Thus the overall error rate for the 58 verbs requiring first-person singular was 4.9 per cent. However, quiero (want-1st-SING) and puedo (can-1st-SING) accounted for around 60 per cent of the children’s correct first-person usage. Once these two correctly marked (and likely rote-learned) verb forms were taken out, the error rate climbed to 10.4 per cent and largely involved the use of the third-person singular inflection instead of the correct first- or second-person singular inflection. The third-person singular inflection was the most frequent form in the speech addressed to the children and thus, in these early stages, the children were incorrectly using this inflection when they did not know the correct form. Two important implications follow from these results. First, despite an overall low error rate, agreement morphology is actually being learned gradually, with pockets of high error which are related to input frequencies. Second, the level of abstraction at which one counts the frequency of a form is important: Errors are low if one treats AGREEMENT as a category, but much higher when one looks at the system in terms of either marking for PERSON or at individual forms.
14.3 ‘Optional Infinitive’ Errors
An area that has been a focus for both sides of the generativist versus usage-based debate is that of so-called ‘optional infinitive’ (OI) errors – utterances that lack finiteness marking (e.g. He going, He do that). In probably the most influential generativist approach to this issue, the Agreement-Tense Omission Model (ATOM), Wexler (Reference Wexler1998) suggested that while children have correctly set the tense and agreement parameters of their language from a very early stage, they are subject to a unique checking constraint in early development, which means that one of these features may be optionally underspecified. This theory makes the prediction that, in languages in which subjects can be dropped (‘prodrop’ languages), the rate of OI errors will be very low because children do not have to deal with checking the agreement feature on the subject. In support of this, young Spanish-speaking children make few OI errors while German- and Dutch-speaking children, who are learning non-prodrop languages, show very high rates, with rates in Dutch-learning children even higher than in German-learning children. However, Freudenthal, Pine, Aguado-Orea and Gobet (Reference Freudenthal, Pine, Aguado-Orea and Gobet2007) have shown that differential rates in OI errors can be accounted for by the relative frequency of utterance-final, non-finite verbs (Dutch 87 per cent, German 66 per cent, Spanish 26 per cent), which results from the verb-second rule in Dutch and German (see also Wijnen, Kempen and Gillis Reference Wijnen, Kempen and Gillis2001). The suggestion is that the child’s processing mechanism may be differentially picking up forms at the ends of utterances and it is this that gives rise to the different rates of OI errors.
A more recent study by Freudenthal, Pine and Gobet (Reference Freudenthal, Pine and Gobet2010) compared the UB position on the quantitative variation in rates of OI errors in different languages with the ‘Variable learning model’ (VLM) of Legate and Yang (Reference Legate and Yang2007). In this model, the child’s system is seen as composed of a number of different grammars or parameter settings, each associated with a particular probability. Relevant data from the input is seen as increasing or decreasing the probability of these competing grammars. The time taken to fix on a +-tense-marking or −tense-marking grammar (e.g. Mandarin Chinese) will depend on the quantitative extent of tense marking in the input. Both the Freudenthal et al. (Reference Freudenthal, Pine and Gobet2010) model and the VLM model account reasonably well for the quantitative differences in OI errors between languages. However, Freudenthal et al. (Reference Freudenthal, Pine and Gobet2010) also found that the particular verbs with which children were more likely to make OI errors were those that occurred more frequently in complex verb phrases in the input, suggesting a specific lexical effect on learning. This cannot be explained in the Legate and Yang (Reference Legate and Yang2007) model, which has no role for lexical effects since the model operates at the abstract level of a TENSE parameter.
Theakston, Lieven and Tomasello (Reference Theakston, Lieven and Tomasello2003) showed a similar effect of input in an experimental study in which children who only ever heard novel verbs in complex verb phrases (i.e. without tense marking) were significantly more likely to produce them as OI errors. However, in support of the idea that children are engaged in gradual abstraction, the results indicated that these children’s learning of finiteness marking was developing: if they heard the novel verb in the finite form, they never incorrectly produced it as non-finite. Thus, development is gradual and piecemeal: children are not leaping from rote-learned utterances to a fully abstract adult grammar.
14.4 Pronoun Case Errors
In a study with a somewhat similar logic to the Freudenthal et al. (Reference Freudenthal, Pine, Aguado-Orea and Gobet2007) study on OI errors, Kirjavainen, Theakston and Lieven (Reference Kirjavainen, Theakston and Lieven2009) investigated whether complex utterances in the input (e.g. Let me do it) might explain the origin of English-speaking children’s first person pronoun case errors, where accusative pronouns are used in nominative contexts (me do it). Naturalistic data from 17 two- to four-year-olds were searched for first-person singular, accusative-for-nominative, case errors and for all first-person singular, preverbal, pronominal contexts. Their caregivers’ data was also searched for first-person singular, preverbal, pronominal contexts. The data show that the proportion of children’s me-for-I errors was correlated with their caregivers’ proportional use of me in first person singular, preverbal contexts. Furthermore, the particular verbs that children produced in me-error utterances appeared in complex sentences containing me in the input more often than the verbs that did not appear in the children’s errors.
Of course, there is no direct mapping from me used as a subject in the input (which never happens), but children are learning and using lexical strings from the input in which me appears before the verb. For at least one of the children, this becomes a productive pattern: she uses the me +V pattern very often, using verbs with which me has not appeared in her input, including some which are tense-marked and would never appear after me in adult speech (e.g. me got, me goes, me can’t). This is interesting since it indicates a process of schema formation in which children initially learn a string from the input and then generalise it to produce utterances which they will never have heard (which will, of course, eventually become of very low or non-existent frequency as the frequency of competing correct constructions builds up).
Pronoun case-marking errors in German are reported as being rare, but this may well be the result of thin sampling. Stumper and Lieven (Reference Stumper and Lieven2011) investigated the acquisition of case on first- and second-singular German personal pronouns (nominative: ich ‘I’, du ‘you’; accusative/dative: mich/mir ‘me’, dich/dir ‘you’) by one German-speaking child, Leo, on the basis of extensive longitudinal data from spontaneous speech (Behrens Reference Behrens2006). Contrary to earlier results, pronoun case errors in this corpus range from 2 per cent to 24 per cent, depending on the particular error, and show a systematic pattern. On the one hand, Leo has a clear-cut preference for erroneous use of ACC pronouns in reflexive events. Between the ages of 2;7 and 3;0, he repeatedly chooses ACC instead of DAT when referring to himself reflexively (e.g. Du reibst *dich/dir die Augen ‘You rub yourself*ACC/DAT the eyes.’; 2;7.2). The ‘you_yourself*ACC/DAT’ error, which comprises both a case and a reversal error, was recorded first followed by a two-month period of ‘I_myself*ACC/DAT’ errors. On the other hand, between the ages of 2;4 and 3;1, there was a clear-cut preference for erroneous use of DAT pronouns in two-participant events that involve a preposition (e.g. Ich habe für *dir/dich Getränke. ‘I have for you*DAT/ACC something to drink’). Eisenbeiss, Bartke and Clahsen (Reference Eisenbeiss, Bartke and Clahsen2006) explained German children’s case-marking errors as being mainly due to having to learn verbs that exceptionally require DAT on the direct object or ACC on the indirect object (termed ‘lexical case marking’) rather than errors arising from the mis-assignment of the case of a particular phrase structure position: ACC for direct objects and DAT for indirect objects. This is known as ‘structural case marking’, and children were expected to have hardly any problems with this (e.g. Ich wasche dichACC ‘I’m washing you’ vs. Ich wasche dirDAT die Haare ‘I’m washing your hair’). However, Leo’s errors do not fit this account. In contrast, in his ACC for DAT errors, he is mostly struggling with DAT for recipients in ditransitive events. His errors with DAT for ACC, mostly following the preposition für, which never assigns DAT, also suggest these errors do indeed involve structural case. Again, the errors seem to be due to schemas that the child has constructed, in this case possibly on the basis of semantics, which will eventually be out-competed by the correct constructions.
14.5 Competition between Constructions
The idea that constructions might compete and give rise to errors was directly addressed in a recent experimental study. In two corpus analyses, Kirjavainen, Theakston, Lieven and Tomasello (Reference Kirjavainen, Theakston, Lieven and Tomasello2009) and Kirjavainen and Theakston (Reference Kirjavainen and Theakston2011) showed that the omission of infinitival to might result from competition between two constructions involving the verb WANT: want to + [verb] and want + [X] giving rise to errors of the type want + [verb] (e.g. want go, want eat cookie) in which the infinitival marker is omitted. Following this, Kirjavainen, Lieven and Theakston (Reference Kirjavainen, Lieven and Theakston2016) tested this idea experimentally. In a within-subjects design1, they primed 2- and 3-year-old children with want + X (e.g. I want my big bike now) and want to + verb (e.g. I want to get my bike) constructions. After each priming episode, the children were asked to produce a sentence starting with I want in an obligatory infinitival-to context, and the rate of to-omission as a function of the preceding prime was measured. Both age groups showed priming effects, but these differed: the two-year-olds provided more correct want to + verb constructions after a want-to prime. The older children showed a reduction in the provision of to after priming with want + X constructions and also produced correct infinitival to after a want-to construction, provided they had both heard and repeated the prime.
14.6 Non-Inversion Errors
One approach to children’s systematic errors is to see them as a reflection of an initially abstract linguistic system in which a parameter is not yet, or wrongly, set. The alternative view assumes that many errors arise from the entrenchment of high-frequency strings, which then compete with differential rates of semantic and syntactic generalisation in related parts of the network. An example are the well-attested non-inversion errors that English-speaking children make with the syntax of questions (e.g. Why she can’t do it?, Dąbrowska Reference Dąbrowska2000). These have been explained in terms of relatively abstract structures. So, for instance, Stromswold (Reference Stromswold1990) suggests that the wh-word is mis-analysed as unmoved and generated in place, and Santelmann, Berk, Austin, Somashekar and Lust (Reference Santelmann, Berk, Austin, Somashekar and Lust2002) argue that errors can be explained by specific features of English syntax (the main verb inversion of copula BE and DO-support). However, children are significantly less likely to make errors with question frames that are frequent in the input (Rowland and Pine Reference Rowland and Pine2000; Rowland Reference Rowland2007). Non-inversion errors occur on low-frequency strings and, crucially, children can show patterns of alternation between correct and incorrect inversion while the system is developing. In an experimental study in which children had to produce questions using combinations of wh-words and auxiliaries, Ambridge, Rowland, Theakston and Tomasello (Reference Ambridge, Rowland, Theakston and Tomasello2006) showed that lexically specific strings of particular wh-words and particular auxiliary forms could account better for the pattern of errors than either the wh-word or the auxiliary type alone (i.e. more non-inversion errors for who do … ? and what do … ? than for what does … ? and who does … ?).
On the other hand, entrenched high-frequency strings can also lead to errors. Ambridge and Rowland (Reference Ambridge and Rowland2009) found that if the children had an entrenched positive question frame such as What does X?, they were significantly more likely to make auxiliary doubling errors when the related negative question was elicited (i.e. to say What does (+) she doesn’t like? instead of What doesn’t she like?). In addition, when a declarative chunk (e.g. She can’t X) was highly frequent, relative to a question frame (e.g. what can’t), the children were more likely to use the declarative chunk and put a question word on the front (e.g. what + she can’t X), giving an error. However, when the declarative string was of low frequency (e.g. She does X), this did not happen.
14.7 Recovery from Over-Generalisation Errors
A problem that has received continuous attention over the last 40 years is how children recover from over-generalisation errors (Bowerman, Reference Bowerman and Hawkins1988; see Ambridge and Lieven Reference Ambridge and Lieven2011 for a detailed discussion). This is a particularly hard problem because, on the one hand, children must be able to form generalisations in order to use language productively. On the other, they are unlikely to be able to retreat from incorrect generalisations purely on the basis of corrective evidence in the input, particularly since there is often no clear competition between different forms – unlike the case with the me-for-I errors discussed above. For instance, what would be the competitors with the generalisations produced by Eva Bowerman, How do you unsqueeze it? (aged 3;11, coming to her mother with a clip earring on and wanting it removed). Much of the discussion of this type of pre-emption concerns how to determine the distance between the error and the potential pre-empting form which could act as a competitor and replace it (e.g. is ‘the man who is/was afraid’ a pre-empting form for ‘the afraid man’?, Boyd and Goldberg Reference Boyd and Goldberg2011). Pinker’s (Reference Pinker1989) solution to the problem of verb argument structure overgeneralisations was based on the interaction of verb meaning with syntax. He suggested that children start with broad-range rules that allow a wide range of verbs to appear in a particular syntactic structure (e.g. action verbs in a causative transitive), thus giving rise to these types of errors. Subsequently, children start to form narrow-range semantic verb classes that define which verbs can alternate and which cannot, and this eliminates the errors. An alternative, though, as it turns out, not conflicting line of research, is based on the finding that frequency is very important in determining which verbs children are likely to over-generalise, with consistent results showing that low-frequency verbs are much more vulnerable (Brooks, Tomasello, Lewis and Dodson Reference Brooks, Tomasello, Lewis and Dodson1999; Theakston Reference Theakston2004). Recent research has shown that both frequency and semantics are independently important and that, rather than narrow and discontinuous verb classes, the issue in semantics is the degree of fit with the construction as a function of what the child already has learned, i.e. there is a semantic continuum. Thus, Blything, Ambridge and Lieven (Reference Blything, Ambridge and Lieven2014) primed children as young as 4;0 to produce verbs of different frequencies with un-prefixes. They found the production probability of verbs in un-form was negatively related to the frequency of the target verb in bare form (e.g. squeez/e/ed/es/ing), as well as being independently, negatively predicted by the frequency of synonyms to a verb’s un-form (e.g. release/*unsqueeze). Ambridge, Pine, Rowland, Chang and Bigood (Reference Ambridge, Pine, Rowland, Chang and Bidgood2013) provide a recent review of this debate.
14.8 Conclusions
Children do indeed make errors in the course of language development. In fact, when analysed at the level of form rather than abstract linguistic category, there are pockets of systematic errors which can be quite substantial. Many of these errors can be analysed as the result of the learning of strings from the input and the abstraction of low-scope schemas from these strings (e.g. some OI errors). Others result from relative frequencies in the input, which means that the child learns some forms much faster than others and defaults to the more frequent form (e.g. third person in Spanish present tense in place of other person markers on less-frequent verbs). Still others result from the blending of two schemas (e.g. some non-inversion errors). These errors co-exist with correct forms and gradually reduce as the correct forms out-compete the errors and as the child learns the more subtle distinctions that assign words to constructions.
This chapter has thus illustrated the role of errors in children’s language acquisition. However, addressing the main question of this volume, can these errors be assumed to lead to or be otherwise connected to language change? Many of these errors/innovations/over-generalisations reduce fairly early in development and are all but gone by school age. Kerswill’s (Reference Kerswill1996) survey of three groups (children aged 0–6, preadolescents aged 6–12 and adolescents aged 12–17) suggests that, indeed, younger children are developing toward the norms of the adult population and that it is the older preadolescents and adolescents who are the youngest sources of change, as argued by Labov (Reference Labov, Lehmann and Malkiel1982; see also Cheshire et al. Reference Cheshire, Kerswill, Fox, Torgersen, Schreier and Hundt2013). Bowerman (Reference Bowerman and Hawkins1988) reports many fewer errors after the age of five, although ‘sporadic’ errors do still occur up to the age of 12 (see Croft Reference Croft2000: 49). Also on past tense, Bybee and Slobin (Reference Bybee and Slobin1982) argue that there is a major difference between the ways in which pre-schoolers deal with some aspects of past tense and the ways that school-age children and adults do. They conclude,
These differences between children and adults have important implications for the relation between child language and historical change. It has often been suggested that children’s innovations in the morphological system are the source of diachronic changes in this system; but this could be true only if children’s creations survived with them into adulthood. Our data show that, in some areas of English verb morphology (e.g. the treatment of no-change verbs), children’s and adults’ strategies are somewhat different, and that a real change in approach takes place before age eight.
In the most detailed study of one child’s past tense and plural overgeneralisations between the ages of two to four, Maslen, Theakston, Lieven and Tomasello (Reference Maslen, Theakston, Lieven and Tomasello2004) show that the over-generalisation rate for past tense shows a major peak between 2;11 and 3;0 and then reduces to zero at 3;11. Interestingly, in this study, there were a small number of errors that, despite massive amounts of evidence in the input as to the correct form, seem to be highly resistant to the exclusive production of the correct form. Although goed was only produced for less than a month, this was in the face of very high input frequencies of the correct form, so why did the child persist in this over-generalisation? The explanation given by the authors (as well as by Bybee and Slobin Reference Bybee and Slobin1982 who found a similar result) is that the present and past tenses of this verb are very far apart and, if the child does not yet realise that all forms of go are related, this may account for the error. Over-generalisations of come (comed) lasted much longer – at least 13 months. Come only undergoes vowel change in the past tense and the present tense ends in a consonant that does not carry ‘high past-tense salience’ (Maslen et al. Reference Maslen, Theakston, Lieven and Tomasello2004: 1,327). Note that in both cases, the child has produced a number of examples of the correct form before the errors start to occur. So these are examples where the child has formed a schema which, perhaps partially through self-reinforcement, becomes, for longer or shorter intervals, resistant to highly frequent evidence in the input of the correct form. These errors and the process of schema formation that underlies them may also be characteristic of some adults and second language learners. However, Bybee and Slobin (Reference Bybee and Slobin1982) argue that although some innovations that are found in both child and adult speech can be related to language change, children are unlikely to be the source of these changes. It is more likely that adolescents, adults and second language learners are the source: a conclusion that the majority of authors in this field, whatever their theoretical approach, appear to agree with (Aitchison Reference Aitchison1991; Kroch 2002; Labov Reference Labov, Lehmann and Malkiel1982; Meisel, Elsig and Rink Reference Meisel, Elsig and Rinke2013; Kerswill Reference Kerswill1996; Ravid Reference Ravid1995).
Young children conform remarkably closely to the language of their environment. Thus, their errors are not maintained into adulthood and therefore cannot become part of the changed language of the next generation (Croft Reference Croft2000). Language-learning children will, of course, have to be the ultimate propagators of change if they learn innovations made by others in adolescence or adulthood and then maintain them and pass them on to the next generation, but there seems to be widespread agreement by many researchers that there is little evidence that they are the innovators.
Yet may it be possible that language change and language acquisition are related beyond the influence of errors? If we are looking for commonalities between child language development and historical language change, my guess is that these lie much more in the processes involved (i.e. for instance the roles of frequency, salience, entrenchment, schema formation, semantics and expressive need) in driving the ways in which children, adolescents, adults and second language learners innovate in their communicative exchanges (also see the chapters on frequency and salience in this volume). The study of how these processes operate as children build their language may well help to inform how the same processes operate in language change.
15.1 Introduction
For some time now, research has identified remarkable similarities in developmental patterns between diachronic change and first language acquisition in various domains. The existence of such parallels can be interpreted as lending support to the generativist idea that grammar change (as opposed to shifts in language use) is located in the L1 acquisition process (see, e.g., Kiparsky Reference Kiparsky, Bach and Harms1968; Lightfoot Reference Lightfoot2006). For others, by contrast, the child learner does not play such a causal role in change; rather, the attested cases of sequential isomorphism between historical development and child language acquisition result from the operation of similar psychological mechanisms, such as analogy and entrenchment, in both diachrony and ontogeny (see Diessel Reference Diessel, Bergs and Brinton2012).
The aim of this chapter is to contribute further to this ongoing debate by examining a number of potential parallels between diachrony and ontogeny in English. Section 15.2 sets the scene for the discussion, introducing recapitulationist views on the relation between ontogeny and phylogeny. Section 15.3 adds diachrony in the picture, presenting some attested developmental parallels between L1 acquisition and historical change (Section 15.3.1) and paying special attention to whether similar or different dynamics are involved in such parallels (Section 15.3.2) and to whether child language acquisition can safely be taken as the locus of change (Section 15.3.3). A comparison of the historical and ontogenetic developments of the going to-future is the focus of Section 15.3.4, which highlights the most significant similarities and differences between the two types of processes. The concluding remarks question the validity of a child-centred theory of language change, while admitting that ‘[w]e can understand certain kinds of change by understanding how acquisition happens, and, vice versa, we can learn much about acquisition by understanding how structural shifts take place’ (Lightfoot Reference Lightfoot2006: 6).
15.2 Setting the Scene: Recapitulation and the Relation between Ontogeny and Phylogeny
Under different formulations, recapitulation has been a recurrent notion in a wide variety of fields, from evolutionary biology to developmental psychology and anthropology. Analogical relations between human development and organic history are already found in Aristotle’s De generatione animalium and his epigenetic views. However, recapitulationism became a contentious issue particularly with the nineteenth-century German biologist Ernst Haeckel and his biogenetic law. According to Haeckel’s recapitulation theory, succinctly summarized as ‘ontogeny recapitulates phylogeny’, the development of every individual organism (ontogeny) repeats the same sequence of developmental stages as have occurred in the evolutionary history of its species (phylogeny). Though largely discredited during the first decades of the twentieth century (see Rasmussen Reference Rasmussen1991), the Haeckelian view of recapitulation and adaptations of it were taken up by various scientists in the latter part of the century, particularly by the paleontologist and evolutionary biologist Stephen Jay Gould. Making use of a modified version of Haeckel’s biogenetic doctrine and focusing on a number of parallels between ontogeny and phylogeny, Gould (Reference Gould1977) maintains that the relationship between the two developments cannot be denied in biological theory and that ‘[e]volutionary changes must be expressed in ontogeny and phyletic information must therefore reside in the development of individuals’ (Reference Gould1977: 2). Recent evolutionary research has shown, however, that a strict interpretation of the biogenetic law, with its strong Darwinist character, is no longer tenable. A critical appraisal of recapitulation is found in, for example, Langer’s (Reference Langer2000) constructivist theory of the evolution of cognition in primates. His findings suggest that the ontogeny of human cognition does not simply recapitulate its primate phylogeny, as testified by the development of specific cognitive specializations in certain species of primates, as opposed to others.
From a rather different perspective, parallels between ontogeny and phylogeny have also been recognized in the field of developmental psychology. Thus, for example, one of the tenets of Piaget’s genetic epistemology is that the development of thought in children resembles the evolution of consciousness in humans, so that studies of the former can definitely contribute to a more complete understanding of the latter. In Piaget’s words,
The fundamental hypothesis of genetic epistemology is that there is a parallelism between the progress made in the logical and rational organization of knowledge and the corresponding formative psychological processes. With this hypothesis, the most fruitful, most obvious field of study would be the reconstituting of human history – the history of human thinking in prehistoric man. Unfortunately, we are not very well informed in the psychology of primitive man, but there are children all around us, and it is in studying children that we have the best chance of studying the development of logical knowledge, mathematical knowledge, physical knowledge, and so forth.
Nevertheless, Piaget explicitly rejects recapitulation as the mechanism behind such parallels; rather, in his view ontogeny and phylogeny follow comparable paths of development under the influence of similar external constraints.
Over the last few decades, recapitulationist and analogical arguments have also been adduced by various linguists. Lamendella (Reference Lamendella, Harnad, Steklis and Lancaster1976) and Bickerton (Reference Bickerton1981, Reference Bickerton1990), among others, explicitly defend the existence of parallels between L1 acquisition and the historical development of language. In fact, some of Lamendella’s statements are strongly reminiscent of strict recapitulation as described above: ‘Most recently encoded genetic information tends to unfold later in ontogeny so as to preserve the temporal sequence in which the new components of the genetic information code were laid down’ (Lamendella Reference Lamendella, Harnad, Steklis and Lancaster1976: 398). Bickerton is slightly more cautious when he maintains that ‘[t]he evidence of children’s speech could thus be treated as consistent with the hypothesis that the ontogenetic development of language partially replicates its phylogenetic development’ (Bickerton Reference Bickerton1990: 115; emphasis mine). A similar perspective is taken by Givón (Reference Givón1979, Reference Givón1998, Reference Givón, Givón and Malle2002, Reference Givón2009), who presents a number of arguments for a neo-recapitulationist view (Reference Givón1979: 273 ff.), proposing proto-grammars for both children and phylogeny (Reference Givón1998: 93 ff.)1 and maintaining that ‘an analogical, recapitulationist perspective on language evolution is both useful and legitimate’ (Reference Givón, Givón and Malle2002: 35). Interestingly, Givón also incorporates diachrony into the recapitulationist picture, when he affirms that ‘[i]n each of the three developmental processes pertaining to human language, evolution, acquisition, diachrony – the very same sequence seems to have been involved’ (Givón Reference Givón1998: 102).
15.3 On the Parallels between Diachronic and Acquisitional Developmental Pathways
In Givón’s neo-recapitulationist scenario, developmental shared trends between the domains of evolution, acquisition and diachrony can be identified in, for example, the emergence of syntactic complexity. Summarizing the development as ‘simple clauses before complex clauses’ (Reference Givón, Givón and Malle2002: 38), Givón explains how simple clauses overwhelmingly predominate both in spontaneous primate communication (see Tomasello and Call Reference Tomasello and Call1997) and in early child syntax (see Bowerman Reference Bowerman1973). Moreover, in both language acquisition and language history, complex sentences typically arise from the merging of simple clauses. The development of complex sentences from earlier paratactic structures in both ontogeny and diachrony is precisely one of the central arguments of Givón’s The genesis of syntactic complexity (Reference Givón2009), where the author contends that functional motivations underlie the development of syntax in both L1 acquisition and language change.
Regardless of whether or not some sort of recapitulatory relationship is considered to exist between diachrony and ontogeny, and regardless of whether it is believed that the same or similar cognitive processes lie behind both the diachronic and the ontogenetic development of language (cf. Section 15.3.2, below), the existence of striking parallels between the way in which language develops in L1 acquisition and how languages develop over time can certainly not be denied.
15.3.1 Some Attested Parallels
The literature provides numerous examples of such parallels for a wide range of linguistic domains in various languages. For example, comparable ontogenetic and diachronic tendencies have been attested between the sequence of acquisition of the Brazilian Portuguese verbal endings by children and the historical development of such inflections (see Hooper Reference Hooper and Fisiak1980). Similarly, frequency effects have been shown to operate in a large number of developments both in early child speech and in the history of languages. Thus, for example, high-frequency irregular forms in morphology are more resistant to analogical pressures than low-frequency items, both in child language acquisition (see Hooper Reference Hooper and Fisiak1980: 178) and in processes of language change (see, among others, Hooper Reference Hooper and Christie1976; Bybee and Slobin Reference Bybee and Slobin1982; Bybee Reference Bybee1985; López-Couso Reference López-Couso and Mazzon2007; Hilpert, Chapter 3 in this volume).2
Modality constitutes a good testing ground for the identification of potential parallels between ontogeny and diachrony. In an analysis of the acquisition of deontic and epistemic modality in Greek and English, Stephany (Reference Stephany, Fletcher and Garman1986) demonstrates that children express the deontic modal meanings of obligation and permission earlier than the epistemic meanings of possibility and necessity.3 This ontogenetic pathway agrees nicely with the development of deontic and epistemic uses of the modals in the history of English, as described in, for example, Goossens (Reference Goosens and Ahlqvist1982), Traugott (Reference Traugott1989), van der Auwera and Plungian (Reference van der Auwera and Plungian1998) and Traugott and Dasher (Reference Traugott and Dasher2002: 105 ff.).
One of the most-often cited examples of possible parallels between ontogeny and diachrony is that of the development of the present perfect construction. Slobin (Reference Slobin and Pagliuca1994; see also Reference Slobin, Givón and Malle2002: 383–4) shows that the present perfect is first used by English-speaking children in resultative contexts (e.g. She has caught a cold), much earlier than in other contexts, such as the perfect of experience (e.g. She has been to Hong Kong a couple of times) and the continuative perfect (e.g. She has known him for five years now). Significantly, as is the case in Child English, the resultative perfect represents the core meaning of the emerging perfect construction in the Old English period (see Carey Reference Carey, Hall, Koenig, Meacham, Reinman and Sutton1990), while the other types of perfect constructions develop later in the history of English.4
Examples of further resemblances between the ontogenetic and the historical development of grammatical markers and constructions are not difficult to find, both for English and for other languages. Slobin (Reference Slobin, Givón and Malle2002: 381–2), for instance, notes the parallel between the acquisition of the Mandarin Chinese accusative marker bǎ in constructions of the type Subject + bǎ + Direct Object + Verb and its diachronic grammaticalization from the full verb bǎ meaning ‘take hold of’. Erbaugh (Reference Erbaugh and Craig1986) also observes interesting similarities between the acquisition of the Chinese noun classifier system and its diachronic pattern: in both language history and L1 acquisition, classifiers appear first with single prototypical nominal referents and then extend their scope to class reference.
Enlightening approaches to how first language acquisition can contribute to a better understanding of diachronic processes in the history of English are found in López-Couso (Reference López-Couso2011) and van Kemenade and Westergaard (Reference van Kemenade, Westergaard, Meurman-Solin, López-Couso and Los2012). The former compares the diachronic grammaticalization of English existential there with the developmental relation between deictic and existential there identified for Child English by Johnson (Reference Johnson1999, Reference Johnson, Cienki, Luka and Smith2001, Reference Johnson2005), who explains the process with his theory of ‘constructional grounding’ or ‘developmental reinterpretation’. In Johnson’s approach, children use occurrences of one pattern (the ‘source construction’) as a model for the acquisition of a related structure (the ‘target construction’), through ‘overlap utterances’ which exemplify formal and semantic-pragmatic properties of both patterns (Johnson Reference Johnson, Cienki, Luka and Smith2001: 124).5 Taking Johnson’s longitudinal evidence from the CHILDES database (MacWhinney Reference MacWhinney2000) as a point of departure and drawing on data from the Old and Middle English sections of the Helsinki Corpus, López-Couso (Reference López-Couso2011) concludes that, as is the case in ontogenetic development, in diachronic terms, there is originally a distal deictic adverb, which comes over time to be used in overlap (bridging) contexts, where it performs both a deictic and an existence-informing function; at a later stage, there starts occurring in contexts which are incompatible with the deictic reading (e.g. negative clauses), thus testifying to the split between locative there and existential there. Similarly, van Kemenade and Westergaard (Reference van Kemenade, Westergaard, Meurman-Solin, López-Couso and Los2012) look into the variation between verb-second and non-verb-second word order in declarative clauses in Early English and the decline of the former pattern in the medieval period by using developmental data from varieties of Present-Day Norwegian with variable verb-second orders similar to those attested in the history of English. Their findings show that syntactic factors and information structure factors closely interact in both language acquisition and language history (e.g. in developments affecting rather infrequent word order patterns where inversion is triggered by information structure factors). Their analysis proves that valuable insights can be gained by looking at historical changes from the perspective of attested processes and preferences in L1 acquisition.
15.3.2 Similar Principles behind the Parallels?
Recognizing the existence of these and similar parallels between acquisitional and diachronic developmental routes could be taken as a strong indication that diachrony recapitulates ontogeny. Yet, even though it is certainly tempting to look at some of the developments outlined in Section 15.3.1 as true cases of sequential isomorphism in ontogeny and diachrony, it may well be the case that the contexts in which these developments take place in child language acquisition differ from those found in the history of language and that, in fact, different psychological processes are at work in the two types of superficially similar development.
Slobin (Reference Slobin and Pagliuca1994, Reference Slobin, Givón and Malle2002), for example, argues that the L1 learner acquires the resultative perfect first (cf. Section 15.3.1) because it is more salient,6 more accessible and less complex than other uses of the perfect construction.7 This explains why the English resultative perfect is already part of the inventory of two-year-old children, while the perfect of experience or the continuative perfect, being cognitively more complex, normally develop only after age four. By contrast, processes of pragmatic extension and reanalysis, rather than degrees of cognitive complexity, seem to lie behind the diachronic development of the different types of perfect in English. This leads Slobin to conclude that the parallel between ontogeny and diachrony is simply ‘illusory’ (Slobin Reference Slobin and Pagliuca1994: 128) and that ‘the dynamics of early grammar formation and the dynamics of adult grammaticization are quite different’ (Slobin Reference Slobin, Givón and Malle2002: 382). In other words, Slobin suggests that the same mappings between conceptual domains are involved in ontogenetic and diachronic grammaticalization, while denying that children simply recapitulate the pragmatic inferences that give rise to the development of grammatical markers historically. Slobin’s characterization of the parallels between ontogenesis and diachrony as illusory is certainly appealing, though probably as difficult to sustain empirically as claims in favor of the existence of genuine parallels between the two domains. The lack of direct evidence to actual data from earlier stages of the language, which can only be approached by means of corpus material, represents a serious methodological problem here.
Ziegeler (Reference Ziegeler1997) takes a somewhat different perspective to that of Slobin (Reference Slobin, Givón and Malle2002). For her, the attested correlations between developmental patterns in ontogenetic and diachronic grammaticalization can be attributed to the existence of a common ‘semantic space’: in both child language acquisition and historical change, grammaticalization makes use of a limited number of basic source concepts (with relatively concrete meanings) for the development of more complex (abstract) and semantically underspecified concepts (see also Bowerman Reference Bowerman and Slobin1985). It is probably not coincidental, therefore, that verbs which children acquire very early across different languages and which are frequently used in the early stages of L1 development (see Clark Reference Clark, Farkas, Jacobsen and Todrys1978), such as those meaning ‘have’, ‘go’, ‘finish’, ‘want’, etc., also represent the most common input for the grammaticalization of tense, aspect and modality markers cross-linguistically (see, e.g., Bybee et al. Reference Bybee, Perkins and Pagliuca1994: 10). An explanation based on shared general cognitive principles is also given by Diessel (Reference Diessel, Bergs and Brinton2012), who maintains that ontogenetic and diachronic developments are often parallel because child language acquisition and historical language change are driven by similar psychological mechanisms, such as analogy,8 entrenchment and categorization (Reference Diessel, Bergs and Brinton2012: 1,601, 1,609–10).
15.3.3 The Controversial Role of Child Language Acquisition in Diachronic Change
The existence of parallels between child language acquisition and diachronic change such as those outlined in the preceding sections has led some linguists to attribute to children a causal role in the historical development of languages. Though the child-centred theory of language change is most readily associated with generative linguistics (see below), the idea that the child plays a major role in diachronic change has been around since at least the late nineteenth century. Paul (Reference Paul1960 [1880]: 34), for instance, maintains that the process of L1 acquisition is the most important cause for changes in language use. A more explicit defense of the relevance of language acquisition for linguistic change and of the transmission of a language from one generation to the next is found in Sweet (Reference Sweet1888: 15), for whom children’s imperfect or defective imitation of the speech of their elders is the primary cause of sound change.9 Similarly, Müller (Reference Müller1890: 75) suggests that the gradual regularization of irregular nominal, adjectival and verbal morphology results from the influence of the ‘dialect of children’.
The role that imperfect learning plays in triggering diachronic change has also come to the fore in more recent non-generative linguistics. Andersen (Reference Andersen1973), for example, proposes a model of phonological change which distinguishes between two modes of change, abduction and deduction, which apply cyclically in the process of L1 acquisition. In Andersen’s model, in the learning process children observe the speech of their models and postulate hypotheses about the laws that govern language, inferring that something may be the case (abductive change) (Andersen Reference Andersen1973: 775); then, they evaluate those abductive inferences by producing new data in conformity with the rules of language (deductive change) (Andersen Reference Andersen1973: 777). It seems, however, that the application of abduction and deduction on the part of children in the process of language acquisition does not always yield the expected output.10 It is precisely in the production of such deviations from the norms of their elders that, in Andersen’s view, diachronic change takes place. Croft (Reference Croft2000: 119) also recognizes the relevance of abduction in processes of form-function reanalysis. In contrast to Andersen (Reference Andersen1973), however, he maintains that abductive changes occur in language use, not in language acquisition.
As mentioned above, it is in generative linguistics that child language acquisition is given a more prominent position in explanations of language change. Starting from the assumption that all humans are genetically equipped with a knowledge of universal aspects of core grammar (see Chomsky’s (Reference Chomsky1981) Universal Grammar),11 L1 acquisition is considered the locus of grammar change. Halle (Reference Halle1962), Kiparsky (Reference Kiparsky, Bach and Harms1968) and Lightfoot (Reference Lightfoot1991, Reference Lightfoot1999, Reference Lightfoot2006), among many others, stress the relevance of the transmission of grammatical knowledge from generation to generation and of children as the vehicles for structural change. A crucial distinction here is that between Internal or I-languages, i.e. those that ‘exist in people’s brains’, and external or E-languages, which are ‘part of the outside world’ (Lightfoot Reference Lightfoot2006: 12). Lightfoot argues that I-languages ‘are formed within a critical period, the first few years of a person’s life’ (Reference Lightfoot2006: 162), when elements of sound structure and properties of basic phrase structure are determined and fixed.
It has been shown, however, that many of the assumptions of the child-based theory of language change are untenable. Serious doubts have been cast on the innateness hypothesis by Sampson (Reference Sampson1997), Croft (Reference Croft2001), Dąbrowska (Reference Dąbrowska2004) and Christiansen and Chater (Reference Christiansen and Chater2008). Moreover, Tomasello (Reference Tomasello1999: 41–5), Croft (Reference Croft2000: 44–59) and Bybee (Reference Bybee2010: 114–9), among others, have provided convincing arguments in favor of a usage-based approach to language change, which contrasts with the aforementioned generative view of I-language vs. E-language. In usage-based approaches, language use itself, governed by different cognitive and social principles, is the primary cause for language change.12 This is also the view adhered to here. As Tomasello (Reference Tomasello2008: 300) puts it, ‘[l]anguage creation and change result from the fact that human communication is open and dynamic, with interlocutors constantly adjusting to one another in order to communicate effectively and accomplish other social goals’.
In this context, an interesting prediction of the usage-based approach to early language development is the existence of considerable variation across children. This is in stark contrast with the idea prevalent within the generative approach that children’s acquisition of grammatical competence follows homogeneous developmental pathways, as predicted by the notion of Universal Grammar. In a comprehensive analysis of the acquisition of auxiliaries, Richards (Reference Richards1990) effectively demonstrates that there exist important inter-individual differences in the process of language learning as regards both the age at which auxiliary verbs emerge and the rate of development. Crucially, Richards’s study proves that such variation is closely linked with the input received by children in conversational interaction. Differential developmental patterns across children have also been identified by Schmidtke-Bode (Reference Schmidtke-Bode2009) in his analysis of the acquisition of the related patterns going-to-V and gonna-V, where the relevance of the nature and frequency of the input in the process of L1 acquisition is also stressed.
15.3.4 Probing the Parallel(s) between the Diachronic and Ontogenetic Developments of the Going to-Future
In the grammaticalization literature, the historical development of the going to-future is considered one of the paradigm cases of grammaticalization,13 which makes it a suitable testing ground for probing the apparent parallels between child language acquisition and diachronic change.
In diachronic terms, the going to-future construction is taken to originate in a biclausal structure involving the main verb go in its motion sense, combined with a non-finite purpose clause, as in (1a). At a later stage, this motion-with-purpose pattern is syntactically reanalysed, as shown in (1b), a change which is made possible by the metaphoric and metonymic inference of futurity from purposive directional constructions (i.e. ‘he is moving in order to get a beer’ > ‘he will get a beer in the near future’).14 Reanalysis is made manifest in cases which are incompatible with a purpose reading, such as those where going to is followed by state verbs, such as like in (1c), as well as instances with inanimate subjects, as in (1d). Once reinterpretation takes place and the construction generalizes to a wider range of contexts, going to can be subject to processes of phonological reduction and word boundary loss, resulting in the form gonna, as in (1e). Here frequency effects (cf. Bybee and Thompson Reference Bybee, Thompson, Bailey, Moore and Moxley1997; Krug Reference Krug1998; Bybee and Scheibman Reference Bybee and Scheibman1999; Scheibman Reference Scheibman2000; Hilpert, Chapter 3 in this volume) and automatization (cf. Haiman Reference Haiman and Pagliuca1994) clearly played a decisive role.
-
a. [He is going] [to get a beer]
b. [He is going to] [get a beer]
c. He is going to like that beer
d. The stock market is going to collapse
e. He’s gonna like that beer
An often-cited early example of the going to-future in the history of English is the late fifteenth-century instance in (2) below. Both Hopper and Traugott (Reference Hopper and Traugott2003: 89) and Eckardt (Reference Eckardt2006: 93), however, rightly acknowledge that the motion interpretation cannot be completely discarded here, since the sequence can also be understood metaphorically as referring to the soul’s journey after death.
(2) Thys onhappy sowle … was goyng to be broughte into helle for the synne and onleful lustys of her body.
Similar indeterminate examples of the use of the be going to-construction are rather common throughout the Early Modern English period, though unambiguous uses also occur at this time and become fairly common by the end of the seventeenth century (see Rissanen Reference Rissanen and Lass1999: 223). An example is given in (3).
-
(3)
he is fumbling with his purse-strings, as a Schoole-boy with his points, when hee is going to be whipt.
(1628, Earle; quoted from Eckardt Reference Eckardt2006: 93)
Interestingly, the construction is already acknowledged as a marker of the imminent future in Joshua Poole’s grammar book The English Accidence (1646):
About to, going to, is the signe of the Participle of the future …: as my father when he was about [to] die, gave me this counsell: I am [about; or going to] read.
As to the reduced form gonna, occasional occurrences are found in the course of the nineteenth century,15 as shown in (4a, b), though its actual expansion is mostly a twentieth-century development.
-
a. Now, Willie lad, I’m ganna gie You twa or three directions.
(1806, A. Douglas Poems 70; OED s.v. gonna v.1)
b. ‘The Heart o’ Mid-Lothian’ is gaunna be acted.
(1856, D. Pae Jessie Melville vii. 76; OED s.v. gonna v.1)
Since the different layers of the construction shown in (1) above are still available in Present-Day English, Schmidtke-Bode (Reference Schmidtke-Bode2009) examines in detail the interplay of such layers in ontogenetic development, using longitudinal data of two monolingual American children from the CHILDES archive and taking into account not only the children’s production, but also their respective input. Applying explorative Configurational Frequency Analysis (von Eye Reference von Eye1990) and the notion of constructional networks from construction grammar, Schmidtke-Bode (Reference Schmidtke-Bode2009) is able to look into the formal and structural similarities between the different production types of going-to-V and gonna-V in the corpus16 as well as the developmental routes followed by the various types. For one of the children (Adam), the development can be summarized as follows (Schmidtke-Bode Reference Schmidtke-Bode2009: 520–6):17
(a) The earliest type in the child’s production is Gon(aux) to V (attested at about 2;3), which forms a cluster with two other types, Gon(aux) V and Gon(ambig) V.
(b) The next cluster to emerge is that containing the first going forms (e.g. I going V, We going V) either with literal (motion) meaning or in auxiliary function. Within this cluster, the most highly entrenched type is I going(aux) V.
(c) The data reveal that at the age of 3;0 the going-types of the previous stage increase in complexity in Adam’s production by means of the introduction of an inflected form of be, the insertion of to and the expansion to a non-declarative sentence-type (e.g. You going V?; They going to V).
(d) Also at around 3;0, Adam starts using the construction with inanimate subjects in the types It’s going V and It’s going to V, the latter winning out from 4;2 onwards. These two types, which show the reduced form of the auxiliary be (’s), belong in a rather large cluster in the network which also comprises the first gonna forms in the child’s production (also at about 3;0). Unexpectedly, gonna occurs with third-person subjects (e.g. It’s gonna V; NP’s gonna V, etc.) about 18 months earlier than with the first-person subject (I’m gonna V). The latest type to emerge within the gonna cluster (at around 4;3) is that involving subordinate clauses (e.g. (that/if) it’s gonna V; (that/if) I’m gonna V).
A comparison of the history of the going to-future outlined above with Schmidtke-Bode’s (Reference Schmidtke-Bode2009) results for Child English reveals a number of striking similarities between the diachronic and the ontogenetic developments. For example, the earliest occurrences of the going-to-V pattern in the children’s production correspond to cases in which the construction is used with its literal motion-with-purpose meaning, in contexts similar to that in (1a, b) above. Illustrative examples are given in (5a, b) below (quoted from Schmidtke-Bode Reference Schmidtke-Bode2009: 526), where the conversational context suggests the meaning of literal motion in order to achieve a goal. Ambiguous instances of this kind clearly invite a reinterpretation of going to as a marker of futurity, before the construction is extended to non-ambiguous grammaticalized contexts such as those in (6a, b), which feature third-person inanimate subjects (cf. (1d) above).
-
a. going wash a hands (Adam 2;8.01)
b. I goin(g) wash em (Sarah 3;0.18)
-
a. This going be a dog (Sarah 3;4.09)
b. It’s going to fall (Sarah 3;5.20)
Data from L1 acquisition research therefore shows that ontogeny follows a parallel path to that suggested by diachronic evidence, at least as regards the early stages of development of the going to-future, with literal motion meanings preceding grammaticalized meanings. The obvious question which arises here is whether the same mechanisms lie behind the grammaticalization of the going to-future in child language acquisition and in historical change (see Section 15.3.2). The answer seems to be in the negative. To begin with, the circumstances in which the two ‘parallel’ processes of grammaticalization take place are noticeably different. Historically, the going to-future emerged as a result of the conventionalization of pragmatic inferences in specific contexts, giving rise chronologically to different layers of the construction. By contrast, in the process of L1 acquisition, children are exposed to the various layers of the construction simultaneously and acquire less-complex meanings (here the literal motion-with-purpose meaning) earlier than extended metaphorical meanings (in this case, the meaning of futurity), which show a higher degree of cognitive complexity.
Be that as it may, certain similarities between historical change and child language acquisition can also be recognized in the later stages of development of the going to-construction. Thus, Schmidtke-Bode’s (Reference Schmidtke-Bode2009) data for the emergence of the reduced form gonna in Adam’s speech appear to be consistent with the diachronic facts summarized above (cf. (1e)): gonna occurs only after the earlier going to-type is firmly consolidated in the child’s production. The trends identified in Adam’s speech cannot, however, be directly extrapolated to Child English as a whole. As shown by Schmidtke-Bode (Reference Schmidtke-Bode2009: 529–30), in Sarah’s output gonna is attested very early, side by side with the going-to pattern, certain types of the former cluster even occurring earlier than some types of the latter. It seems therefore that the existence of some kind of recapitulatory relationship between the ontogenetic and diachronic processes of grammaticalization of the going to-future has to be dismissed. A likely explanation for the differential developmental patterns in the linguistic production of the two children is found in the nature of their respective inputs: while Adam’s input shows very low frequencies for the reduced form, Sarah’s abounds with gonna’s, especially in the earliest transcripts from the CHILDES archive (Schmidtke-Bode Reference Schmidtke-Bode2009: 529). This seems to lend support to the predictions of usage-based approaches to language change discussed in Section 15.3.3, above, and especially to Richards’s (Reference Richards1990) findings that the input frequency of a particular pattern plays a key role in the process of L1 acquisition. Lieven (Chapter 14 in this volume) also reports on the influence of input frequencies on the production of pronoun case errors made by English-speaking children, such as the use of accusative pronouns in nominative contexts (e.g. me do it).
15.4 Closing Remarks
The foregoing discussion has shown that the existence of remarkable parallels between ontogenetic and diachronic developments cannot be taken as definite evidence in favor of some sort of recapitulationist view between child language acquisition and historical change. Striking parallels can indeed be established in various domains, such as the resistance of high-frequency irregular formations in morphology to become regular (cf. Section 15.3.1), the emergence of epistemic modal meanings out of deontic ones (cf. Section 15.3.1), the development of the different meanings of the present perfect and that of the existential there from the locative there (cf. Section 15.3.1) and the directional change in the going to-pattern from motion to future meaning (cf. Section 15.3.4). Nevertheless, matching developmental pathways cannot be adduced for other types of change. Thus, while developmental parallels between ontogeny and diachrony are relatively common on the morphological and syntactic levels (as shown by the examples discussed in the preceding sections), it seems that sound change does not produce such good results. Although certain phonetic processes, e.g. the reduction of consonant clusters, bear strong similarities in child speech (Menn and Stoel-Gammon Reference Menn, Stoel-Gammon, Fletcher and MacWhinney1994) and in language history (Hock Reference Hock1991: 80 ff.), other potential parallels are more questionable, and cast serious doubts on the idea that it is the child that causes historical sound change, especially by failure. Starting from Stampe’s (Reference Stampe1969) conjectures along these lines, Drachman (Reference Drachman and Fisiak1978) compares a number of phonetic processes in language acquisition and historical change,18 and concludes that ‘some kinds of sound change cannot be attributed to learning failures during primary acquisition, while other kinds must remain ambiguous in this respect’ (Drachman Reference Drachman and Fisiak1978: 123) and that, therefore, ‘the role of primary acquisition in language change seems to have been exaggerated’ (Drachman Reference Drachman and Fisiak1978: 138).19
Even in cases where ontogeny and diachrony exhibit strong similarities, different processes and factors may lie behind the two types of development and, as a consequence, not all stages identified in the former can be recognized in the latter, and vice versa. Thus, whereas pragmatic inferencing plays a decisive role in diachronic grammaticalization, cognitive complexity seems to be a more relevant factor in ontogenetic grammaticalization. The acquisition of the going to-future discussed in Section 15.3.4 provides a paradigmatic example (Schmidtke-Bode Reference Schmidtke-Bode2009), to which those of the perfect construction (Slobin Reference Slobin and Pagliuca1994, Reference Slobin, Givón and Malle2002) and of deontic and epistemic modal meanings (Stephany Reference Stephany, Fletcher and Garman1986) could be added (cf. Sections 15.3.1 and 15.3.2). In all likelihood, degrees of cognitive complexity also play a role in other L1 developments mentioned in this chapter, such as the emergence of the more abstract existential meaning in there out of its concrete spatial meaning (Johnson Reference Johnson1999; Reference Johnson, Cienki, Luka and Smith2001, as referenced in Section 15.3.1). It seems therefore that some of the attested parallels between ontogeny and diachrony are likely to be just ‘illusory’, to use Slobin’s (Reference Slobin and Pagliuca1994: 128) term.
Further arguments against the idea of recapitulation are provided by the existence of significant inter-individual differences in the process of L1 learning, as shown by Richards (Reference Richards1990) for the acquisition of auxiliaries (cf. Section 15.3.3) and Schmidtke-Bode (Reference Schmidtke-Bode2009) for the development of the going to-/gonna-patterns (cf. Section 15.3.4). Such inter-individual differences are very likely caused by input differences, therefore dismissing the belief prevalent in generative linguistics that child language acquisition is the locus of diachronic change (cf. Section 15.3.3). Rejecting the idea that L1 learning plays a causal role in language change does not imply, however, ruling out the parallels between ontogenetic and diachronic developments. Although the cases discussed in this chapter question the validity of a child-centred theory of language change, it has to be admitted that ontogeny sometimes ‘retraces’ historical change and that certain diachronic developments can be more fully understood by observing how acquisition happens, and vice versa.
