Experience is dynamic and ephemeral, yet humans routinely generate abstract representations of their individualized experience that simultaneously achieve enough stability, plasticity, and interindividual parity to radically facilitate social and cognitive functioning. Christiansen & Chater's (C&C's) ambitious Chunk-and-Pass processing (CPP) proposal offers hope of a comprehensive and elegant account of how this can be. CPP has impressive explanatory breadth, neatly tying language acquisition to language change and language evolution, while also offering promise of a unified account of perception and cognition more generally. By C&C's own acknowledgment, however, many facets of the CPP account cry out for elaboration. In our view, three “working edges” will be (a) accounting for the earliest inception of language acquisition, (b) explaining stability and plasticity differences in learning profiles across knowledge systems (within language as well as across domains), and (c) elaborating CPP on the action processing front.
Regarding the first issue, C&C provide a workable framework for describing language acquisition once basic acoustic units have been discovered (e.g., phonemes, syllables), but do not describe how utter novices initially break into the system. Of course, there is a sizable literature investigating how infants initiate analysis of streaming speech (e.g., Vouloumanos & Werker Reference Vouloumanos and Werker2007; Werker et al. Reference Werker, Yeung and Yoshida2012). One litmus test of the viability of CPP will be its ability to account for the phenomena documented in this literature within a unified Chunk-and-Pass framework. Among the complexities to be confronted here include findings indicating that infants' identification/construction of basic acoustic units may still be taking place at the same time that they are beginning to chunk longer strings of sounds together into words or morphemes. For example, infants remain quite sensitive to phonetic distributions until well into the first year; at 6 to 8 months, just 2–3 minutes of focused exposure to new distributions may be enough to temporarily rearrange infants' phonetic categories (Maye et al. Reference Maye, Werker and Gerken2002). And yet, by this same age, infants typically recognize at least a handful of words, including “mommy” and “daddy” (Tincoff & Jusczyk Reference Tincoff and Jusczyk1999), their own name (Bortfeld et al. Reference Bortfeld, Morgan, Golinkoff and Rathbun2005; Mandel et al. Reference Mandel, Jusczyk and Pisoni1995), and several body part terms, such as “feet” and “hand” (Tincoff & Jusczyk Reference Tincoff and Jusczyk2012). Does CPP somehow build linguistic structure even without clear basic units over which to operate (in contradiction to hypotheses C&C articulate on this matter; e.g., sect. 3.2, para. 1)? Alternatively, does CPP operate on units only as they reach some criterion of availability, so that words composed of early-identified phonemes would potentially be available for chunking, whereas words with more difficult-to-identify phonemes are not? Or do processes other than Chunk-and-Pass need to be brought in to account for the earliest phases of language acquisition?
The second working edge we identify relates to stability and plasticity of representations. C&C note that stability and plasticity trade off: Learning depends on representations being updated to incorporate new content, but at the same time, some degree of stability is needed to avoid new information overwhelming previously acquired information. They argue that stability is a natural product of the compression that occurs during Chunk-and-Pass processing. The processing of linguistic content is “lossy” – the only features retained are those that are captured by a learner's current model of the language, making it difficult to dramatically alter that model since the features necessary to do so are likely the very ones lost in compression. This seems persuasive on the face of it, but leaves unclear how CPP can account for a different stability/plasticity issue: namely, the observation that representations of different types display distinct stability/plasticity profiles. In language, acquired representations of some kinds (e.g., phonetic and syntactic representations) display a strong propensity to stabilize and become markedly resistant to change (e.g., Johnson & Newport Reference Johnson and Newport1989; Kuhl Reference Kuhl2004; Lenneberg Reference Lenneberg and Lenneberg1964; Yoshida et al. Reference Yoshida, Pons, Maye and Werker2010), whereas a variety of evidence suggests that other representational types (e.g., open-class lexical items) seem to display considerably more plasticity (e.g., Curtiss Reference Curtiss1977; Newport Reference Newport1990; Talmy Reference Talmy2000; Weber-Fox & Neville Reference Weber-Fox and Neville1996). In question is whether these different plasticity profiles across representational types arise naturally from CPP. Are there differences in the information to be encoded across various types of representations such that the model would predict an emphasis on stability in some cases versus ongoing plasticity in others? Alternatively, will it be necessary to look to mechanisms beyond CPP to account for such differences, such as diverse neural commitment timetables?
Our third “working edge” focuses on action processing as a particularly fruitful target for broadening the scope of CPP-related investigation. Intuitively, language and action processing seem closely linked. Language can be regarded as one form of action, after all, and both language and action are subject to the Now-or-Never bottleneck, making them amenable to a CPP account, as C&C themselves note. Strikingly, however, investigation regarding action processing lags considerably behind language. One glaring example is the lack of a generally accepted inventory of basic actions, comparable to inventories of phonemes or syllables in language (cf. interesting but small-scale efforts along these lines, such as therblig, Gilbreth & Gilbreth Reference Gilbreth and Gilbreth1919). Another example concerns hierarchical structure, which seems to be a fundamental organizing principle of both action and linguistic representations. To illustrate in the action context, observers typically note that an action such as getting a cup of coffee comprises embedded subgoals, such as getting a mug from a cupboard, placing it on a counter, pouring coffee into the mug, and so on. At the same time, relevant levels of that hierarchy seem not to be as crisp or well-defined as they are in language. A “learning to process” account may provide welcome guidance for continuing attempts to gain purchase on the representation of structure in action, and perhaps also will ultimately help to explain cross-domain differences in representational structure. All in all, as an explicitly domain-general approach, CPP holds promise for accelerating understanding in the action domain in a way that promotes interdisciplinary convergence with theorizing about language.
Experience is dynamic and ephemeral, yet humans routinely generate abstract representations of their individualized experience that simultaneously achieve enough stability, plasticity, and interindividual parity to radically facilitate social and cognitive functioning. Christiansen & Chater's (C&C's) ambitious Chunk-and-Pass processing (CPP) proposal offers hope of a comprehensive and elegant account of how this can be. CPP has impressive explanatory breadth, neatly tying language acquisition to language change and language evolution, while also offering promise of a unified account of perception and cognition more generally. By C&C's own acknowledgment, however, many facets of the CPP account cry out for elaboration. In our view, three “working edges” will be (a) accounting for the earliest inception of language acquisition, (b) explaining stability and plasticity differences in learning profiles across knowledge systems (within language as well as across domains), and (c) elaborating CPP on the action processing front.
Regarding the first issue, C&C provide a workable framework for describing language acquisition once basic acoustic units have been discovered (e.g., phonemes, syllables), but do not describe how utter novices initially break into the system. Of course, there is a sizable literature investigating how infants initiate analysis of streaming speech (e.g., Vouloumanos & Werker Reference Vouloumanos and Werker2007; Werker et al. Reference Werker, Yeung and Yoshida2012). One litmus test of the viability of CPP will be its ability to account for the phenomena documented in this literature within a unified Chunk-and-Pass framework. Among the complexities to be confronted here include findings indicating that infants' identification/construction of basic acoustic units may still be taking place at the same time that they are beginning to chunk longer strings of sounds together into words or morphemes. For example, infants remain quite sensitive to phonetic distributions until well into the first year; at 6 to 8 months, just 2–3 minutes of focused exposure to new distributions may be enough to temporarily rearrange infants' phonetic categories (Maye et al. Reference Maye, Werker and Gerken2002). And yet, by this same age, infants typically recognize at least a handful of words, including “mommy” and “daddy” (Tincoff & Jusczyk Reference Tincoff and Jusczyk1999), their own name (Bortfeld et al. Reference Bortfeld, Morgan, Golinkoff and Rathbun2005; Mandel et al. Reference Mandel, Jusczyk and Pisoni1995), and several body part terms, such as “feet” and “hand” (Tincoff & Jusczyk Reference Tincoff and Jusczyk2012). Does CPP somehow build linguistic structure even without clear basic units over which to operate (in contradiction to hypotheses C&C articulate on this matter; e.g., sect. 3.2, para. 1)? Alternatively, does CPP operate on units only as they reach some criterion of availability, so that words composed of early-identified phonemes would potentially be available for chunking, whereas words with more difficult-to-identify phonemes are not? Or do processes other than Chunk-and-Pass need to be brought in to account for the earliest phases of language acquisition?
The second working edge we identify relates to stability and plasticity of representations. C&C note that stability and plasticity trade off: Learning depends on representations being updated to incorporate new content, but at the same time, some degree of stability is needed to avoid new information overwhelming previously acquired information. They argue that stability is a natural product of the compression that occurs during Chunk-and-Pass processing. The processing of linguistic content is “lossy” – the only features retained are those that are captured by a learner's current model of the language, making it difficult to dramatically alter that model since the features necessary to do so are likely the very ones lost in compression. This seems persuasive on the face of it, but leaves unclear how CPP can account for a different stability/plasticity issue: namely, the observation that representations of different types display distinct stability/plasticity profiles. In language, acquired representations of some kinds (e.g., phonetic and syntactic representations) display a strong propensity to stabilize and become markedly resistant to change (e.g., Johnson & Newport Reference Johnson and Newport1989; Kuhl Reference Kuhl2004; Lenneberg Reference Lenneberg and Lenneberg1964; Yoshida et al. Reference Yoshida, Pons, Maye and Werker2010), whereas a variety of evidence suggests that other representational types (e.g., open-class lexical items) seem to display considerably more plasticity (e.g., Curtiss Reference Curtiss1977; Newport Reference Newport1990; Talmy Reference Talmy2000; Weber-Fox & Neville Reference Weber-Fox and Neville1996). In question is whether these different plasticity profiles across representational types arise naturally from CPP. Are there differences in the information to be encoded across various types of representations such that the model would predict an emphasis on stability in some cases versus ongoing plasticity in others? Alternatively, will it be necessary to look to mechanisms beyond CPP to account for such differences, such as diverse neural commitment timetables?
Our third “working edge” focuses on action processing as a particularly fruitful target for broadening the scope of CPP-related investigation. Intuitively, language and action processing seem closely linked. Language can be regarded as one form of action, after all, and both language and action are subject to the Now-or-Never bottleneck, making them amenable to a CPP account, as C&C themselves note. Strikingly, however, investigation regarding action processing lags considerably behind language. One glaring example is the lack of a generally accepted inventory of basic actions, comparable to inventories of phonemes or syllables in language (cf. interesting but small-scale efforts along these lines, such as therblig, Gilbreth & Gilbreth Reference Gilbreth and Gilbreth1919). Another example concerns hierarchical structure, which seems to be a fundamental organizing principle of both action and linguistic representations. To illustrate in the action context, observers typically note that an action such as getting a cup of coffee comprises embedded subgoals, such as getting a mug from a cupboard, placing it on a counter, pouring coffee into the mug, and so on. At the same time, relevant levels of that hierarchy seem not to be as crisp or well-defined as they are in language. A “learning to process” account may provide welcome guidance for continuing attempts to gain purchase on the representation of structure in action, and perhaps also will ultimately help to explain cross-domain differences in representational structure. All in all, as an explicitly domain-general approach, CPP holds promise for accelerating understanding in the action domain in a way that promotes interdisciplinary convergence with theorizing about language.