There is a long linguistic tradition in which language change is explained in terms of first language acquisition. In this tradition, children are considered to be the agents of language change, or at least the agents of changes in the underlying grammar. Since the early 1980s, this has been formulated in the (generative) terminology in terms of parameters set by children: whereas an older generation acquires one particular setting of a parameter (during childhood), a next generation of L1 children may set a parameter differently, based on the input of their parents, and this may lead to a different output. For obvious reasons this argumentation had to be built on theoretical rather than empirical work on language acquisition. There are no children acquiring Old English or Middle Dutch, and, in fact, the field of acquisition research was until recently much less developed and very often not focused on the type of facts that happened to play a role in discussions of language change.
Perhaps the most remarkable development in linguistics since the 1980s is the growth of the field of acquisition research. So, what was not possible then, makes much more sense now, namely, to confront ideas on change with ‘real’ evidence from acquisition. Jürgen Meisel's keynote article in this issue rightly makes a strong argument for the type of collaboration between linguists from several (sub)fields to do precisely this. I also agree with his view that at first sight not much seems to remain of the idea that L1 children are the agents of change if we take into account recent findings on child L1 acquisition. Basically, the finding is that in ‘normal’ transmission from generation to generation children are simply too good to be responsible for transmission errors. What was and is one of the starting points of the generative enterprise (why is it that children are so good at learning their L1), makes it difficult to consider them the agents of change. From the work of my own team in Amsterdam (Blom, Polišenská & Weerman, Reference Blom, Polišenská and Weerman2008; Polišenská, Reference Polišenská2010) the example of the acquisition of Dutch adjectival inflection is, I think, very suggestive in this respect. The central observation is that L1 children are capable to pick up even the opaque aspects of the inflectional system of the target language very early. If so, it is very unlikely that they will deviate from the input (under normal conditions).
Our work on adjectival inflection also corroborates another point Meisel makes. It turns out that in particular the opaque aspects of the Dutch adjectival inflection are very vulnerable in child L2 and adult L2 acquisition. In general, these learners overgeneralize default forms, and may acquire a more simple inflectional system for adjectives. Of course, if this output is the input for a new process of child L1 acquisition, this new generation of L1 children will not acquire the original opaque Dutch adjectival rule and the rule may disappear from the language since children will only acquire this rule if they are forced by the input. So, child L2 acquisition and adult L2 acquisition could be crucial ingredients in change, as Meisel suggests (and how this might have worked in the case at hand of Dutch adjectival inflection in some variants related to Dutch is briefly discussed in Weerman, Reference Weerman2002).
Regardless of this support for crucial claims in Meisel's argumentation, I will argue here for a slightly different perspective on the issue, which may solve some problems that Meisel's hypothesis meets. In Meisel's view, bilingualism is the central key in our understanding of language change, as made explicit in the subtitle of his paper: “bilingualism as cause and effect of grammatical change”. It is certainly true that in many cases change goes hand in hand with bilingualism. This seems evident if indeed an L2 factor is crucial in change. And since a new variant in change usually is in competition with an older variant, in a somewhat more technical sense the effect is indeed bilingualism. Let alone that the latter use of the notion of bilingualism for the coexistence of variants in competition (the effect of change) may be far-fetched, my main concern is Meisel's formulation that bilingualism should be seen as the cause of change. Instead I would like to propose that the motor of change is the opposition between early and late acquisition.
Note first of all that the opposition between early and late acquisition covers the central claims made by Meisel that I just summarized. Child L2 and adult L2 acquisition are crucial ingredients of change according to Meisel. What they share is, of course, that they are both instances of ‘late acquisition’ compared to child L1 acquisition, as prototypical ‘early acquisition’. Typically ‘late acquisition’ is much less successful than ‘early acquisition’ and may hence be responsible for transmission errors. As shown by Meisel himself, bilingualism itself does not have to be problematic in this respect as long as two (or even more) languages are acquired early simultaneously. So, the proper generalization does not seem to be bilingualism but rather the distinction between early and late acquisition.
Several factors are held responsible for this qualitative difference between early and late acquisition. On the one hand, the input for late learners may be less rich both qualitatively and quantitatively, but I think Meisel is right in claiming that at least in addition a ‘fundamental difference’ between early and late acquisition plays an important role, whether or not this has to be formulated in terms of the accessibility of (parameters of) Universal Grammar or a distinction in the way learners may exploit procedural and declarative memory along the lines of Ullman (Reference Ullman2001), or along other lines.
It is important to underline here a consequence of the idea of a ‘fundamental difference’ between early and late acquisition for change. If we claim that some abstract feature can only enter the language (or better: the grammar) via early acquisition as a result of the ‘fundamental difference’ between the two types of acquisition, and if this feature becomes visible at some point in the history of a language, we cannot escape the conclusion that early acquisition is responsible for its appearance. In this respect L1 children must be the agents of change, although the first step in this process could very well be a form of late acquisition. A transmission failure resulting from child or adult L2 acquisition may function as the input of a new process of L1 acquisition, where the new abstract feature is introduced. Indeed, in this way the interplay between early and late acquisition could be the driving force behind change.
So far it might look as if the difference between Meisel's perspective and mine is merely terminological. However, there are at least three related issues, where in my opinion this idea of a distinction between early and late acquisition as a driving force of change works better than Meisel's bilingualism.
For the first issue we turn to those linguists who in fact usually do not refer to acquisition at all when they discuss language change and who sometimes even explicitly reject the role of child L1 acquisition in this process. An example is Croft (Reference Croft2000), who claims that language change emerges from language use. This is in general also the position of those working in the broad and heterogeneous field of grammaticalization. Whatever the precise formulation of this process, some of its steps are in fact cases of what I would call ‘late acquisition’. Take for instance Haspelmath (Reference Haspemath1999) and Keller (Reference Keller1994), who sketch how presumably adult speakers try to talk in such a way that they are noticed and how other speakers take over this novelty. It does not make sense to see this as a form of adult L2 acquisition, since no L2 is involved. Rather, this type of change is an extension or some reanalysis of an aspect of the L1 by adults. It would make the concept vacuous to call this bilingualism, but it does make sense to see it as a form of late acquisition, which we could call adult L1 acquisition. The same age effects apply as in other forms of late acquisition. If adult L2 learners have no access to parameters, or if they are more dependent on declarative memory, then this should also be the case for adult L1 acquisition.
In many cases this type of adult L1 acquisition may not lead to extra rules compared to the grammar acquired earlier as a child and may be restricted to lexical changes. However, this is not necessarily the case, and whatever its precise character, in the end the result of adult L1 acquisition will feed a new process of early acquisition, as in the case of child and adult L2 acquisition discussed above. Even if the changes made by the L1 adults are mainly of a lexical nature, this does not necessarily mean that the grammar acquired by L1 children based on this input is not crucially different from the grammar acquired by the earlier generation, given the fundamental difference between early and late acquisition.
How important is this type of late acquisition that does not seem to fit in Meisel's bilingualism perspective? If it only plays a marginal role, Meisel is right in neglecting it. However, my impression is that it is as crucial as adult L2 and child L2 acquisition. This brings me to my second issue, namely that it is impossible to maintain that there is always a direct connection to L2 acquisition in change. The missing link is what I have called adult L1 acquisition.
If we speculate for one moment on the prehistory of language, the problem may become clearer. In order to understand the spread of languages around the globe, it should be possible to split languages into two or more. They could never become systematically different if some kind of an L2 factor is needed to force this. As may be clear from Meisel's paper and the argumentation above, the first steps needed in the process of splitting can also not be due to child L1 acquisition. What remains is an effect of late acquisition by L1 adults along the lines sketched above, which may be entirely different in the new languages involved.
In fact, we can stay much closer to home to make the same point. Even in the cases where it is evident that some L2 factor is crucial, like the Viking influence on English or the linguistic melting pot in 17th-century Amsterdam, it is impossible to argue that all linguistic consequences are directly linked to L2 speakers. Both in time and in space a bridge is needed between these speakers and new generations of L1 learners. This bridge is built by late acquisition of another type, namely what I have called adult L1 acquisition. So, this type of late acquisition is crucial to understand change.
There is one further issue where the opposition between early and late acquisition might offer a more fruitful perspective. Meisel concludes that “reanalysis affecting parameter settings is much less likely to happen than is commonly assumed in historical linguistics” (p. 142). This indeed makes sense if L2 factors play an important role in change and if L1 acquisition is very robust, as Meisel describes. What makes the claim somewhat problematic, however, is that the notion parameter is nowadays in use for all sorts of micro variation and there hardly is a principled distinction with ‘macro parameters’. Nobody will probably deny that change does take place regularly in the micro domains, but the implication is that we do not really understand why this would not be the case for the macro domains. On the other hand, the claim that some aspects of a language withstand historical developments more easily than others is not only valid for what is covered by parameters. The fact that Dutch attributive adjectives are inflected with a schwa, for instance, seems to be very robust, and is a matter of a spell out rule. So, it looks as if parameter theory falls short on both sides. Again, the opposition between early and late acquisition might help to save the essence of Meisel's observation. What is acquired early is historically robust; what is acquired relatively late is vulnerable to change. Of course, for late learners all aspects of language are acquired late by definition, so the distinction makes sense only for L1 learners. Only if the result of late acquisition seriously influences the input of early acquisition, may the robust features of a language change. It would be an interesting collaborative research program for historical linguistics and the field of acquisition to see if this idea is successful.
You have
Access
There is a long linguistic tradition in which language change is explained in terms of first language acquisition. In this tradition, children are considered to be the agents of language change, or at least the agents of changes in the underlying grammar. Since the early 1980s, this has been formulated in the (generative) terminology in terms of parameters set by children: whereas an older generation acquires one particular setting of a parameter (during childhood), a next generation of L1 children may set a parameter differently, based on the input of their parents, and this may lead to a different output. For obvious reasons this argumentation had to be built on theoretical rather than empirical work on language acquisition. There are no children acquiring Old English or Middle Dutch, and, in fact, the field of acquisition research was until recently much less developed and very often not focused on the type of facts that happened to play a role in discussions of language change.
Perhaps the most remarkable development in linguistics since the 1980s is the growth of the field of acquisition research. So, what was not possible then, makes much more sense now, namely, to confront ideas on change with ‘real’ evidence from acquisition. Jürgen Meisel's keynote article in this issue rightly makes a strong argument for the type of collaboration between linguists from several (sub)fields to do precisely this. I also agree with his view that at first sight not much seems to remain of the idea that L1 children are the agents of change if we take into account recent findings on child L1 acquisition. Basically, the finding is that in ‘normal’ transmission from generation to generation children are simply too good to be responsible for transmission errors. What was and is one of the starting points of the generative enterprise (why is it that children are so good at learning their L1), makes it difficult to consider them the agents of change. From the work of my own team in Amsterdam (Blom, Polišenská & Weerman, Reference Blom, Polišenská and Weerman2008; Polišenská, Reference Polišenská2010) the example of the acquisition of Dutch adjectival inflection is, I think, very suggestive in this respect. The central observation is that L1 children are capable to pick up even the opaque aspects of the inflectional system of the target language very early. If so, it is very unlikely that they will deviate from the input (under normal conditions).
Our work on adjectival inflection also corroborates another point Meisel makes. It turns out that in particular the opaque aspects of the Dutch adjectival inflection are very vulnerable in child L2 and adult L2 acquisition. In general, these learners overgeneralize default forms, and may acquire a more simple inflectional system for adjectives. Of course, if this output is the input for a new process of child L1 acquisition, this new generation of L1 children will not acquire the original opaque Dutch adjectival rule and the rule may disappear from the language since children will only acquire this rule if they are forced by the input. So, child L2 acquisition and adult L2 acquisition could be crucial ingredients in change, as Meisel suggests (and how this might have worked in the case at hand of Dutch adjectival inflection in some variants related to Dutch is briefly discussed in Weerman, Reference Weerman2002).
Regardless of this support for crucial claims in Meisel's argumentation, I will argue here for a slightly different perspective on the issue, which may solve some problems that Meisel's hypothesis meets. In Meisel's view, bilingualism is the central key in our understanding of language change, as made explicit in the subtitle of his paper: “bilingualism as cause and effect of grammatical change”. It is certainly true that in many cases change goes hand in hand with bilingualism. This seems evident if indeed an L2 factor is crucial in change. And since a new variant in change usually is in competition with an older variant, in a somewhat more technical sense the effect is indeed bilingualism. Let alone that the latter use of the notion of bilingualism for the coexistence of variants in competition (the effect of change) may be far-fetched, my main concern is Meisel's formulation that bilingualism should be seen as the cause of change. Instead I would like to propose that the motor of change is the opposition between early and late acquisition.
Note first of all that the opposition between early and late acquisition covers the central claims made by Meisel that I just summarized. Child L2 and adult L2 acquisition are crucial ingredients of change according to Meisel. What they share is, of course, that they are both instances of ‘late acquisition’ compared to child L1 acquisition, as prototypical ‘early acquisition’. Typically ‘late acquisition’ is much less successful than ‘early acquisition’ and may hence be responsible for transmission errors. As shown by Meisel himself, bilingualism itself does not have to be problematic in this respect as long as two (or even more) languages are acquired early simultaneously. So, the proper generalization does not seem to be bilingualism but rather the distinction between early and late acquisition.
Several factors are held responsible for this qualitative difference between early and late acquisition. On the one hand, the input for late learners may be less rich both qualitatively and quantitatively, but I think Meisel is right in claiming that at least in addition a ‘fundamental difference’ between early and late acquisition plays an important role, whether or not this has to be formulated in terms of the accessibility of (parameters of) Universal Grammar or a distinction in the way learners may exploit procedural and declarative memory along the lines of Ullman (Reference Ullman2001), or along other lines.
It is important to underline here a consequence of the idea of a ‘fundamental difference’ between early and late acquisition for change. If we claim that some abstract feature can only enter the language (or better: the grammar) via early acquisition as a result of the ‘fundamental difference’ between the two types of acquisition, and if this feature becomes visible at some point in the history of a language, we cannot escape the conclusion that early acquisition is responsible for its appearance. In this respect L1 children must be the agents of change, although the first step in this process could very well be a form of late acquisition. A transmission failure resulting from child or adult L2 acquisition may function as the input of a new process of L1 acquisition, where the new abstract feature is introduced. Indeed, in this way the interplay between early and late acquisition could be the driving force behind change.
So far it might look as if the difference between Meisel's perspective and mine is merely terminological. However, there are at least three related issues, where in my opinion this idea of a distinction between early and late acquisition as a driving force of change works better than Meisel's bilingualism.
For the first issue we turn to those linguists who in fact usually do not refer to acquisition at all when they discuss language change and who sometimes even explicitly reject the role of child L1 acquisition in this process. An example is Croft (Reference Croft2000), who claims that language change emerges from language use. This is in general also the position of those working in the broad and heterogeneous field of grammaticalization. Whatever the precise formulation of this process, some of its steps are in fact cases of what I would call ‘late acquisition’. Take for instance Haspelmath (Reference Haspemath1999) and Keller (Reference Keller1994), who sketch how presumably adult speakers try to talk in such a way that they are noticed and how other speakers take over this novelty. It does not make sense to see this as a form of adult L2 acquisition, since no L2 is involved. Rather, this type of change is an extension or some reanalysis of an aspect of the L1 by adults. It would make the concept vacuous to call this bilingualism, but it does make sense to see it as a form of late acquisition, which we could call adult L1 acquisition. The same age effects apply as in other forms of late acquisition. If adult L2 learners have no access to parameters, or if they are more dependent on declarative memory, then this should also be the case for adult L1 acquisition.
In many cases this type of adult L1 acquisition may not lead to extra rules compared to the grammar acquired earlier as a child and may be restricted to lexical changes. However, this is not necessarily the case, and whatever its precise character, in the end the result of adult L1 acquisition will feed a new process of early acquisition, as in the case of child and adult L2 acquisition discussed above. Even if the changes made by the L1 adults are mainly of a lexical nature, this does not necessarily mean that the grammar acquired by L1 children based on this input is not crucially different from the grammar acquired by the earlier generation, given the fundamental difference between early and late acquisition.
How important is this type of late acquisition that does not seem to fit in Meisel's bilingualism perspective? If it only plays a marginal role, Meisel is right in neglecting it. However, my impression is that it is as crucial as adult L2 and child L2 acquisition. This brings me to my second issue, namely that it is impossible to maintain that there is always a direct connection to L2 acquisition in change. The missing link is what I have called adult L1 acquisition.
If we speculate for one moment on the prehistory of language, the problem may become clearer. In order to understand the spread of languages around the globe, it should be possible to split languages into two or more. They could never become systematically different if some kind of an L2 factor is needed to force this. As may be clear from Meisel's paper and the argumentation above, the first steps needed in the process of splitting can also not be due to child L1 acquisition. What remains is an effect of late acquisition by L1 adults along the lines sketched above, which may be entirely different in the new languages involved.
In fact, we can stay much closer to home to make the same point. Even in the cases where it is evident that some L2 factor is crucial, like the Viking influence on English or the linguistic melting pot in 17th-century Amsterdam, it is impossible to argue that all linguistic consequences are directly linked to L2 speakers. Both in time and in space a bridge is needed between these speakers and new generations of L1 learners. This bridge is built by late acquisition of another type, namely what I have called adult L1 acquisition. So, this type of late acquisition is crucial to understand change.
There is one further issue where the opposition between early and late acquisition might offer a more fruitful perspective. Meisel concludes that “reanalysis affecting parameter settings is much less likely to happen than is commonly assumed in historical linguistics” (p. 142). This indeed makes sense if L2 factors play an important role in change and if L1 acquisition is very robust, as Meisel describes. What makes the claim somewhat problematic, however, is that the notion parameter is nowadays in use for all sorts of micro variation and there hardly is a principled distinction with ‘macro parameters’. Nobody will probably deny that change does take place regularly in the micro domains, but the implication is that we do not really understand why this would not be the case for the macro domains. On the other hand, the claim that some aspects of a language withstand historical developments more easily than others is not only valid for what is covered by parameters. The fact that Dutch attributive adjectives are inflected with a schwa, for instance, seems to be very robust, and is a matter of a spell out rule. So, it looks as if parameter theory falls short on both sides. Again, the opposition between early and late acquisition might help to save the essence of Meisel's observation. What is acquired early is historically robust; what is acquired relatively late is vulnerable to change. Of course, for late learners all aspects of language are acquired late by definition, so the distinction makes sense only for L1 learners. Only if the result of late acquisition seriously influences the input of early acquisition, may the robust features of a language change. It would be an interesting collaborative research program for historical linguistics and the field of acquisition to see if this idea is successful.