A Note on Pedagogy, Topics and Ways of Understanding

Child Language Acquisition:


A working paper



Joseph Galasso


California State University,





Abstract          This working paper is the quick result of some general considerations that went into the formulation of a book review for Wadsworths Publications. The comments herein constitute the first part of a two-part review of the text 'Language Development' (3rd edition, by Erika Hoff: 2004). The theme of this paper is confined to remarks made in the review concerning the general tenor of pedagogical issues: while many of the remarks are in direct response to some of the prompted quotations found in the new edition, the quotes and subsequent comments, I feel, broadly speak to current issues of pedagogy I would like to advance (as opposed to taking on any substantive issues of the text). Hence, I have excluded specific comments aimed at the actual text itself.


Theme                        I would like to narrowly restrict this working paper to a brief outline and review of the suggested treatment of the 'Dual Mechanism Model' (DMM) by which a full range of linguistic concepts can be presented to the student in an entirely exhaustive and comprehensive manner. This potential treatment of the DMM views the classic 'Skinner vs. Chomsky' debate as an all encompassing springboard from which a multitude of linguistic material can be brought forward to the student.


In the main, it will be argued herein that a host of differing theoretical issues can be made more deliberate in showing how leading theoretical positions fall either onto General Domain or Specific Domain theories (sometimes the distinction is labeled under General vs. Special Nativism)[1]. Almost all linguists working on these issues, regardless of their position on nativism, believe that some amount of innate structure in the brain is required in order for the child to acquire language. For instance, Jeff Elman (a connectionists working out of the University of San Diego) has come out saying (pc) that some amount of innate structure must exist, at least at the level of the architectural design of the brain (perhaps in the form of 'hidden units'). This certainly has the flavor of Chomskyan rhetoric, at least from a distance. Elizabeth Bates (an otherwise strong proponent of connectionism) expands on this by saying that any such notion of 'innateness' must also then extent to the lower-scope cognitive capacity ('General Domain'), and not necessarily be restricted to functioning at any higher order level (which might be interpreted as 'Specific Domain') Hence, in today's refashioned terminology, Skinner vs. Chomsky remains vibrant after nearly more than a half century'the same argument is now seemingly reworded as the 'Nature of nature' debate (viz., can potential, innate lower-scope computational processing account for formal rule-based syntax? This question gets somewhat revisited in light of Newport's 'less-is-more-hypothesis' which suggests, at least intuitively to me, that a high-memory/cognitive capacity might somehow be responsible for eventual rule-based syntax. See Newport herein).


More concretely, what I am suggesting here is that some attempt could be made within current pedagogy to show just how individual and otherwise apparently autonomous positions (e.g., behaviorism, constructivism, emergentism, connectionism, etc.) can readily be refashioned as purporting to one side of the larger proposed two-pronged model. Again, when the terminology is pooled in offering a pedagogical devise, students quickly see the larger forces at work. Of course, one always runs the risk of over-simplification and/or mal-representation whenever one attempts to frame all arguments within stark dualism. However, I have seen that when the material is carefully chosen and prepared, with proper background information provided, students utilize the available dichotomy in a comprehensive way which enhances not only the lecture experience, but also buttresses outside fieldwork assignments alongside research proposals. In addition to the general comments that follow, I believe pedagogical practice could be even better enhanced by keeping a keen eye on potential topics of research students could pursue since it very well may be the case that many of the students today taking linguistic courses are required to submit research proposals which have some bearing on language development in the elementary through high-school setting[2].

[0]       There will always be a need to continuously reinvent the overall lecture experience. The ability to cast new and oft-times convoluted material in ways that allow students to gain early traction of the topic is an ever-mounting challenge for most lecturers. Let us be plain and start by saying that teaching is essentially story-telling. I believe most would agree with this, and that every field has its own rich tradition of themes and anthems (from Physics through to the Arts). This is quite natural: people tell themselves stories to work on this or that. We sell ourselves on various points of digression in order to bring upon some inherent opposition. In following a variety of topics surrounding Language Development, one very important story comes to my mind. It is a story with several twists and turns and has intrigued the most imaginative amongst us linguists: so much so that is has taken on a role for itself as 'the quintessential pedagogical device'. Perhaps the more exciting aspects of this singular story are the embedded debates and infighting that carry on still today and that have taken hold over the years and centuries, spawning an embarrassing wealth of writing and discourse. The essential 'story-of-learning' I am on about here has been labeled and relabeled accordingly throughout the years as e.g., Nurture vs. Nature, Environment vs. Genetic, Bottom-Up vs. Top-down, and, (more specific to language), as Frequency (type/token) vs. Variable, Associative vs. Rule-based, and Learning vs. Parameter Setting'in short, I suggest, all reduce to the classic Skinner vs. Chomsky debate (1958). I feel the debate, still today, can be used as a valuable devise in portraying various linguistic attitudes and research models currently being employed in an attempt to explain the maturational development underpinning child language acquisition: (a strong 'maturational-based' language development here as understood in the framework of Borer & Wexler (1987), Radford (1990), Radford & Galasso (1998)).



This paper attempts to (i) outline current and highly influential thinking on the aforementioned areas of debate and (ii) proceed to find creative ways in which the material could serve as a cohesive/coherent springboard from which to base classroom discussion. The idea here is not only to reconsider how one should go about thinking and (re-)thinking the nature of language development, but to show how one might better access the material within a lecture-style presentation.



B.F. Skinner vs. Noam Chomsky   The theme at hand is the classic 'Skinner vs. Chomsky' Debate (Chomsky 1959). The following is an attempt to extract the essence of the debate and to transform the material into an overall pedagogical device. The three major areas of language development Biological Basis, Phonology, and Morphology and Syntax are looked at in turn.


                        It might be shocking to some, even disturbing'given what appeared to be the wide-spread and abrupt demise of Behaviorism as an immediate result of the debate'that many of the arguments which actually fueled the classic debate still linger with us to today. How are we to account for the robust nature of these arguments? In brief, the arguments have survived much in the same way as have the classic debates between Plato vs. Aristotle, or Descartes vs. Locke'viz., the mind/body 'Cartesian Cut' continues to haunt us as does the 'ghost in the machine' analogy which was brought to light by Newton as he fought off Cartesian mechanics. (Recall, although Newton killed off the 'machine', he graciously left us the 'ghost' to fend off). These arguments are well grounded in the larger problem of how one chooses to understand the brain/mind and underscores much of what little we currently know about how the brain actually processes linguistic material. The most articulated theory which couples both sides of the debate is labeled herein as the 'Dual Mechanism Model' (DMM) (see Chomsky 1995, Pinker 1999, and Clahsen 1999 for review).



The DMM maintains that the brain processes linguistic material in two fundamentally different ways: (i) one routing system corresponding to the temporal-lobe region of the brain which governs 'Skinner-like' high-frequency, associative-driven modes of learning and (ii) one routing system corresponding to the frontal-lobe region of the brain which governs 'Chomsky-like' rule-driven modes of acquisition. I am fully aware that such a model could be easily portrayed as over-simplistic and all encompassing: the little known facts that have come out of fMRIs and ERPs do show that the picture is not so well defined as one might hope. (The picture becomes even murkier when trying to spot exactly how/where children with language disorders go wrong with processing language). Nevertheless, the general use of the DMM in analyzing (describing)[3] both language acquisition and language disorders has become increasingly accepted as being on the right track (as noted in Pinker's popular book Words and Rules, though see Fodor 2002 for an opposing view). In addition, when used more as a teaching devise, I am suggesting the DMM serves equally well in a lecture environment in accounting/explaining in a student-friendly way the phenomena of language acquisition. In the following remarks, my overall goal is to tease out just how much of language acquisition follows from this dual mechanism and to show how a maturational story can best be devised to account for the chronological onsets of the different stages attested in child language acquisition (viz. Lexical vs. Functional stages of grammar), onsets determined by neurological developments in the brain (a Biological Bases).




1.1 Language Development


Language development as a basic research topic

[1]       'The modern field of language development emerged in the 1950s when it became clear that language acquisition would serve as a test for rival theories...'


           Stimulus & Response (S&R) methods as an adequate way of describing language development can be questioned here by employing what came out of the 'Wugs Test' as based on rule-morphology: e.g., [N+{s}=> plural], [V+ {ed} => past]. It seems to me that Berko is often relegated in pedagogy either as a means of showing phonological assimilation, or as a means of showing mere knowledge of morphological rules. Both are convincing, but such myopic data analyses tend to stop short of revealing the larger dichotomy at hand'viz., greater effort could be made in current teaching to show just how Berko's data countered connectionism on its own right (as central to the debate and the DMM). This might provide students with a bit more background understanding of the issues that could be easily assimilated before one goes on to complicate matters with, say, connectionism (a kind of super-charged version of behaviorism). In addition, in doing so, this might also be the place to introduce the so-called 'U-shape learning' which could later on be expanded in its own right'regarding 'U-shape' phonological, morphological and syntactic development. (See Appendix 4A). Moreover, once the arguments are spelled-out as such, the same S&R methods can then be more favorable applied in order to account for 1-to-1 mappings of vocabulary and irregular verb storage (as opposed to a 'one-to-many' or 'many-to-one' rule-based analogy). For instance, recent interesting work has come out (Algere & Gordon 1999) showing that very high-frequency regular past tense verbs (e.g., walked)'which under the DMM are considered rule-based and decomposed of a stem+affix {-ed}'may eventually become stored as associative lexical chunks. In this sense, otherwise 'Chomskyan-based' high-frequency decomposed words, over the long haul, begin to take on Skinnerian looks.


It is in this fundamental way that the DMM can be introduced to the student early on as a pedagogical devise. Recall, the DMM states that words, as well as irregular morphological stems, are stored as 'lexical chunks' in the same area of the brain responsible for 1-to-1 associative iconic memory'an area typically located in the back temporal-lobe region (Wernicke's area).


[2]       'A central tenet of behaviorism is the belief that is not necessary to discern what goes on in the mind of the rat in order to explain the change in the rat's behavior, behavior can be fully accounted for in terms of things external to the mind.'



Nature or Nurture


'The alternative view known as nativism [vs. empiricism J.G.], asserts that knowledge cannot come from experience alone. The mind has some preexisting structure...'


            Such typical statements could be coupled and expanded upon in a general way as to show that Behaviorism took learning as a direct input related phenomena with very little (if any) intervening processes (i.e., prescribed preexisting structure) at work. I refer to this mode of learning as 'what comes in, goes out' (like a tape recorder if you will). When coupled with the aforementioned 'Wugs Test' along with 'U-shape learning', students immediately see how Skinner's general assertions of language development were much too strong'viz., his belief that general learning, language included, was the simple and lone bi-product of a general cognitive-based memorization apparatus did not capture developmental aspects of language as attested in the Roger Brown studies. For instance, if language <x,y> goes in (input) but language <x,y,z> comes out (output), as in the attested examples of over-regularization (go-ed, went-ed, sleep-ed, never-ed), then the output cannot be a direct result of the input, but that some (innate) intervening process must be at work which fashions some amount of rule analogy based on variables. In addition, recent psycho-linguistic data collected by Marcus et al. show that the German plural default {-s} (as opposed to say {-en} or {-r}) cannot be the simple result of a threshold of stored memories of regularities (frequency driven) previously experienced in the input, but rather the default arises out of a rule. The Marcus data seem to represent a counter anti-frequency effect (pace the work gathered by Algere and Gordon op.cit) in that the German default {-s} occupies a low-frequency status. I refer to this intervening process (in a general way) as a mechanism that incorporates 'abstraction', eventually leading to the representation of 'true-rules'. This could later be described as Universal Grammar (UG) or the Language Faculty (LF) using more linguistic specific terminology. In my own research, I suggest that abstract rules are not productive in the very early MLU stages of language acquisition and that neuro-linguistic maturational factors contribute to the onset of rules such as INFLectional morphology. Newport's 'less-is-more hypothesis' (1991) might have a bearing on how one could interpret such an early 'non-rule-based' stage of language development. (See Appendix C). In her view, high memory ultimately spawns syntax due to the inability of the mental processes to store and compute this newfound high-volume of material. For instance, this stance can be considered when examining animal communication, particularly chimp ASL studies'since chimps have only a low threshold memory capacity, as attested in their low MLUs, chimps remain till the end in the closed realm of associationism). I would also like to suggest that the abstract ability also supports other species-specific (Human) activities such as Math, Music, Ceremonial beliefs/Taboo, in addition to 'L'anguage. Bickerton (1990) refers to this notion as Language with a capital-'L', as opposed to the generally accepted term language with a small-'l' which characterizes language typology (e.g., French vs. German vs. Korean etc.). Likewise, Pinker's coinage 'Mentalese' (1994) could be expanded as referring to a similar concept'that of 'abstract thought'. These distinctions could be mentioned in relation to general issues dealing with animal communication as presented in contrast to human language. The mere fact that chimps, our closest genetic relative, cannot act in an ultimately altruistic sense, but rather is confined to function in accordance to their own instinctive laws of survival speaks to this unbridgeable gap between human syntax (rule-based) vs. animal communication (associative-based) (Chris Knight, pp.9-16 in Hurford et al. 1998). This can easily get expanded. For instance, it could be argued that human altruism is an ultimate bi-product of the ability to abstract (a frontal-lobe activity). Also regarding the dual mechanism model, it would seem to be the case that Chimps only have access to frequency-driven associations which rely on trained behavior rather than on abstract rules. When students see this aspect of the debate as playing out between animal communication vs. true human language, relevant and highly sophisticated questions emerge in the classroom as to where and how the language processing differs between the two beings. This topic can then later on nicely dovetail into discussions regarding early MLU in child language, a lexical-stage of development which mimics the level of sophistication reached (yet never surpassed) by chimps, etc. etc. Topics to this effect are both numerous and applicable.




            The Chomskyan revolution


[3]       General discussion on Chomsky's theory of generative grammar should always encompass notions of an innately endowed mechanism'a mechanism that not only (i) relies on an external, memorization apparatus which assigns forms to meaning as factored from the environment (Skinner), but also one that (ii) relies on some abstract and internal intervening process which leads to the formation of grammar (Chomsky). To my mind, this apparent two-prong pairing of linguistic material led Chomsky's to devise his finely articulated Principles & Parameters Theory (PPT)'with the former 'principles' seemingly abiding by stated Skinnerian notion of lexical stems and substantive meaning, and the latter 'parameters' abiding by stated Chomskyan formal, abstract features). I do like to suggest that Skinner does make his way into PPT by virtue of (cognitive-based) principles. This is quite fascinating for students in light of Chomsky's strong position after the debate with Skinner. This idea though is never explicitly stated (at least never by Chomsky himself), but, as you can see, it makes for a hefty DMM analogy.


Many forms and/or versions of this general discussion can then later branch off into more developed discussion of a generative phonology, morphological, and syntax (respectively). In general, it could be argued that the aforementioned wugs test data, particularly when coupled with later observations suggesting a U-shape leaning, represent the first attempts to see language processing as taking on a dual routing system'i.e, a system that first utilizes rote-learned strategies then later develops true rule formation. In short, the DMM could be incorporated as follows: (i) Firstly, a phonological dual model could help explain what is behind the Pre-representational vs. Representation stages of development. (The observation that Asperger's children who suffer from a lack of rule formation and thus do not produce typical assimilations might fall under such an example). (ii) Secondly, morphology could likewise be cast under this dual model by addressing inherent distinctions placed on Derivational vs. Inflectional morphology, and where such processes occur in the brain (Temporal lobe vs. Frontal lobe (respectively)). (ii) Thirdly, syntax can capture this dualism by addressing distinctions placed between Lexical and Functional categories'the former being a 'one-to-one' word learning process (more akin to Skinner) by which frequency-based methods of rote-learning establish substantive part-of-speech words (N, V, Adj), while the latter category is heavily based on rule formation.


[4]       '...Behaviorists theories have not stood the test of time...'


This may be a premature judgment given what is now coming out of many connectionist camps. True, it may not be '1950s-style' behaviorism, but it is 'behaviorism' nevertheless. This return to associative-based strategies of language learning could be emphasized in order to construct the full response to the DMM as suggested herein.


[5]       'The issue of whether language acquisition is supported by domain specific mental module or by general cognitive processes..' could be further examined in the light of Neil Smith's work done on man called Christopher (a language savant with very limited cognitive capacity'establishing a Double Dissociation between 'cognitive prerequisites' and 'language' (pace Piaget 1970, Karmiloff-Smith 1992) (Smith et al. 1995: 2).


[6]       One thought could be introduced here to buttress the DMM: the idea being that children may very well start out as (i) Statistical Learners'given that the part of the brain that apparently comes 'on-line' from the very earliest MLU is exactly that limbic/temporal lobe region of the brain which could carry out such statistical frequency learning'then subsequently mature into (ii) Rule Learners (see Wakefield and Wilcox 1994: BUCLD 19, 643-654 for a maturational account of frontal brain development and subsequent functional category processing). The notion here is that the two approaches could complement each other in important developmental ways. The notion of brain development could expand on this maturational account and nicely pave the way for discussion regarding a DMM phonology, morphology and syntactic progression (See Fig. 1.1 for maturational milestones).


            'In other words, a learning mechanism that does nothing more than count the frequency with which things appear together can end up telling word from non-words because the previously heard words result in a different level of activation in a network than non-words. Although no one doubts that the mind can do this, some seriously doubt that a mind that could only do this could ever acquire language.' 'The alternative argument is that to do language, the mind needs symbolic processes, like algebraic rules'.



            For children as Rule learners (Chomsky) (vs. Frequency learners), see Bartke, Marcus, Clahsen (1995) for treatment of German plural {-s} as the default notwithstanding the fact that it constitutes perhaps the lowest frequency of all the German plural inflectional morphemes {-en}, {-er}, {-s}, {-e}, and {-ø} (Bartke et al. 1995: Cognitive Psychology).


            For children as Frequency learners (Skinner), students love the popular examples of how common irregular errors manifests based on frequency learning: e.g., the over-application of the sound pattern *bring>brang>brung (as based on ring>rang>rung) or the Spanish irregular example of Roto (past tense of Rompere, 'to break') which over-regularizes to *Rompido (as based on the high frequency of the stem Romp- and affix {-ido} in the Spanish INFL paradigm). (See Tomasello, Pine and Rowland).



            The above dual analyses can then be nicely elaborated on in follow-up discussion pertaining to question such as e.g., What kind of knowledge does the child acquire? and Continuity vs. Discontinuity in development?


            The subsequent extreme positions Formalism (Chomsky/Top-down theology) and Functionalism (Piaget-Skinner/Bottom-up theology) could then be nicely captured within this same overarching DMM.


1.2 Biological Bases of Language Development


[7]       In addition to the notion of Functional Asymmetry, enhanced brain imagining techniques have been used to further support the DMM (suggesting that there is not only a left/right brain schism regarding language/spatio organization (respectively), but that there exists even finer-grained asymmetries regarding the frontal-lobe/temporal-lobe distinction. The fact that 'seeing/hearing/speaking' words activate areas in the brain which heavily rely on frequency driven, motor-skill capacity (Wernicke's area) seems to me to be a way to expand on what we are saying regarding 'word learning' as a Skinner-type process'a general domain mode of learning. The fact that the frontal lobe (Broca's area) lights up when one 'generates' words incorporating syntax suggests that a different routing system is at work regarding the 'generating process' as opposed to the more prosaic 'word identifying' and 'performance' process. (See comments on Wakefield & Wilcox (§6). Clahsen et al.(2001), Clahsen (1999), Gross et al. (1998), Munte et al. (1998) among others have done studies using Human Brain Potentials and/or ERPs to show how regular vs. irregular verb are stored in the brain in support of the DMM).


[8]       As a follow-up to (§7) above, Broca's vs. Wernickes's aphasia could be discussed as exemplifying the DMM in a more explicit way. Furthermore, the same asymmetries of the brain seem to be at work in identifying two forms of Dyslexia: Surface Dyslexia (SD) vs. Deep Dyslexia (DD). 'Surface dyslexics are able to decode words phonologically, but are not able to recognize whole words'. Their mistakes indicate that phonological representation and segmentation are functioning--as with the mistake tac /tæk/ for cat /kæt/ (an initial-final C stop exchange from /k/ to /t/). The area of the brain impacted is the frontal-lobe (the area responsible for rules). Some research suggests that while SDs can read regular spelling words, they cannot read irregular words correctly. By contrast, Deep dyslexics are unable to decode words phonologically. 'For example, DDs perform a kind of 'gestalt' whole-word learning strategy by which errors arise based on association mistakes: e.g., they may see the word orchestra and read it as symphony' as the two words share common semantic associations. The area of the brain impacted points to the temporal-lobe (the area responsible for associations). In sum, this asymmetry leading to the two forms of dyslexia could be re-categories as Syntactic vs. Semantic Dyslexia. (Italics, Obler & Gjerlow 1999: p 115).


[9]       The DMM: 'Pinker (1999) has argued that both an associative memory system and a rule-based process are the necessary ingredients for language'.


            Here we find mentioned the DMM, but it seems to me more could be shown leading up to this point.


            Alzheimer's and Parkinson's         In addition to these two degenerative disorders, this might be a nice place to introduce and spell-out how Asperger's vs. Williams syndrome seem to parallel (as mirror images) Alzheimer's and Parkinson's. Alzheimer's and Asperger's affect (in different ways) the same area of the brain delegated for normal memory function. For example, Asperger's children (who have an abnormal memory capacity) over-rely on memory to the detriment of their rule-based processing. Regarding the 'Alegre-Gordon' Test (Alegre & Gordon (1996), Asperger's children (as well as SLI children) are happy to keep the intervening inflection between to two stems being compounded:


                        e.g.      (1) *rat-s-eater,


            whereas normal functioning children would delete the plural inflection, thus forming

                                    (2) rat-eater. => [delete plural {-s}]


                                    (3) rats-eater (non-segmented chunk /rats/ = [+plural] )


            This morphological process and deletion of the inflection in (2) is a rule-based frontal-lobe activity. We can account for the error in (1) by saying that the Asperger's child has simply memorized the stem and inflectional {s} as a 'gestalt' whole word (or non-segmented chunk) (3). Williams' children on the other hand have intact functional grammar with abnormal memory function. (See Appendix for profile samples of data taken from a child with Williams' vs. Down's syndrome).


            This material could also simply be introduced here and later developed in Chapter 4 under Morphology.



[10]     'In fact, the brain doesn't reach its mature state until the age of 15'.


            'Biological Maturation' is the null hypothesis. A genetic basis of language development could be pursued throughout with more detail under the proposed DMM as a pedagogical devise. (E.g., Williams' children are affected by a deletion of a strand of genes in chromosome 7 that seem to have antecedes to temporal-lobe processing, among other things.)



[11]     The modular non-modular view


            It seems to me that this is just one more example of how the DMM could explain the distinctions among these two language adaptation theories:


            Special-purpose mechanism (= Special Nativism)

            According to this view, language is housed in an autonomous and special area of the brain completely lacking antecedent to any other lower-level brain processing such as cognition. One could expand on this dualism by utilizing Chomsky's term 'Language Faculty' in such a way that it speaks directly to aspects of the Chomsky vs. Skinner/Piaget Debate. In this sense, the DMM becomes the default hypothesis by virtue that language, as it becomes formally defined as syntax, is now distinct from language as an associative learning mechanism (e.g., root-word learning and irregular word lexical storage). One side of the Dual Mechanism then could be roughly defined as Special Modular, while the other side of the model would be said to be Non-Modular, and thus ultimately tethered to lower-scope cognitive mechanics. In short, Chomsky's quarrel with Piaget (like Skinner before him) was based on the notion that a non-modular theory could never sufficiently provide the whole, comprehensive picture of language development.


1.3. Phonological Development: Learning the Sounds of Language


[12]     Voicing           In additional to phonological assimilation (having to do with voice and place & manner of articulation), it could be mentioned that between the two voicing features, voiced vs. voiceless, children tend to treat the [+ voice] feature as the default--this can be particularly well viewed when one examines the default phonological selection of past tense. Although three allomorphic phone options are available in presenting the past tense inflection {-ed}'(i) /d/ as in the word (ple:d/ (played), (ii) /t/ as in the word /kIkt/ (kicked) (showing phonological assimilation), and (iii) /Id/ otherwise as the default--children start with the /Id/ default form and maintain it up until a certain age of development. Examples of this range from */kIkId/ (kicked), */brokId/ (broke), */kepId/ (kept), */si:Id/ (saw), */kukId/ (cooked), etc. In other words, once children start to employ the phonological rules associated with the past tense {ed}, they over-regularize the /Id/ pronunciation for {ed}. This same pattern of development seems to mimic the 'U-shape learning' as discussed above. (See Marcus et al. 1992) (Also see Pinker's book Words & Rules p. 255).


[13]     In addition to (§12) above, some children with language disorders such as Asperger's syndrome do not seem to apply phonological assimilations as do normal children (cf. the 'Wugs' test Berko), suggesting the phonological 'rule' is not being productive and that rote-learning strategies are imported where stems are memorized as lexical, gestalt wholes along with their morphological endings. (Also see (§9) above regarding Gordon's 'rat-eater' test which shows that Asperger's children may in fact simply be incorporating inflectional endings into their stem as chunks).



[14]     'The maturation of still higher levels of the brain'areas of the motor cortex--may be required for the onset of canonical babbling at 6-9 months'.


            The nature of this comment (i.e., maturation) could then be taken to the next level by further suggesting what type of brain function might be required in order for 'still higher' level processes such as rule-based representational phonology to come on-line. For instance, a Pre-representational stage in speech may only incorporate syllabic discrimination and not phonemic discrimination


[15]     Contrast between /b/ and /v/ This phonological contrast personally reminds me of the Spanish examples I use in class which may also suggest a DMM phonology. For instance, how might we account for the fact the Spanish speakers of Mexican heritage can say the phoneme /v/ as learned from their very earliest MLU speech, e.g., high frequency words such as (Yo) voy (I go), but cannot then say the phoneme in isolation, where the /v/ => /b/ as in the name victor /vIktr/ => bictor /bIktr/, etc. One possible account might be to employ the DMM by suggesting that the /v/ in the early high-frequency word Voy as been processed as a holistic chunk early-on and without any phonological (phonemic) representation. Once rule-based phonology comes on-line, the speaker must make do with his/her rule-based phonemic repertoire (which goes without a true representational /v/). (cf. Walley 1993).


[16]     'This argument is consistent with evidence that phonemic representation of words begin around the time that vocabulary size increases dramatically (18m).


            This could be interpreted as saying that 'high-memory' spawns 'rule-based analogy' simply due to the high amount of material that now must get processed (from phonology through to syntax). As mentioned above, (§2) there might be implications for Newport's 'less-is more' hypothesis, and in a sense a plus for Neo-Darwin adaptive theories of language acquisition.


[17]     Models of Phonological Development

There are four theories summarized in this section'starting from Behaviorism through to cognitive-models. Under the central theme of this paper, some attempt to pin each theory to their respective maturational stage of phonology might make for a very nice concluding section of the chapter. For instance, the behaviorist model is discredited 'because it ignores the role of maturational process...or because it is simply based on imitation and reinforcement'. True, but this type of learning surely enters into the early Pre-representational stage of phonology. Behaviorism is a problem only when one tries to say that it is the whole story. For instance, the observation that there exists a first stage of phonology which may not abide by phonological rules (such as assimilation) speaks directly to a potential behaviorist stage of speech production that is not solely based on holistic, non-segmental speech. (We needn't continue 'to throw out the baby with the bath water' on Chomsky's behalf'that good-fight has been long retired).


[18]     Banana => /nænæ/ is a beautiful example of how such speech cannot be based on a memory bottle-neck of sorts (once attributed to such simplified pronunciation). Here, it is the initial unstressed CV structure that has been deleted. Any attempt to suggest that a lack of memory is behind such errors would undoubtedly run into trouble with this example. More to the point, the example exemplifies the notion of a 'rule-based phonology': (in this instance, a rule-based, syllabic development). (See (§4) Appendix-A on U-shape phonology).



1.4. Lexical Development: Learning Words


[19]     First Words    Context-bound words are iconic in the sense that they are driven more by associative leaning than by symbolic manipulation. The 'Pre-representational stage of lexical development' thus maps onto the 'Pre-representational stage of phonological development' as discussed herein'i.e., both processes are initially governed by the same S&R/reinforcement strategies (more robust and primitive processes which then later get ousted by rule-driven grammar).


            'Some researchers argue that context-bound words appear first and that truly referential words must await some cognitive development in the child' .


This idea'that two different linguistic processes may initially be governed by the same underlying strategy'underscores the notion that (higher-level) formal language development may sit on top of (lower-level) cognition as suggested by the DMM.


[20]     What determines the context of early vocabularies? The fact that children acquire Nouns before Verbs directly speaks to 'saliency factors' which pertain to the child's (lower-level) cognitive mode of learning. Hence, 'meaning' does underwrite language development in this sense.


[21]     'Gopnik and Melzoff (1986) found that children begin to show cognitive understanding that objects can be grouped into categories about the same time that they become very good a learning object labels--which, after all, label object categories, not individual objects'.


            A certain amount of abstraction must take place in order to learn words...moving from context-bound (proto-words) to context-flexible (true-words). This lack of abstraction early on for the child can be seen by his/her lack of super-ordinate level categorization'e.g., children are slow at learning word such as furniture since there is 'one-to-one' iconic mapping of the concept's meaning (as opposed to the basic-level vocabulary chair, table, bed, etc.). Moving from 'context-bound words' (which shows little if any abstraction of category) to 'true words' can be viewed inline with maturation factors that have been discussed regarding the DMM.


[22]     The mapping problems between (i) semantic/cognitive organization and (ii) syntactic/formal organization can be presented as a dualism which has antecedents to the DMM (as based on maturational factors). Ultimately, both forms of bootstrapping are employed by the child in word learning. It just may be that semantic organization precedes syntactic organization (cf. individual research being carried out by Tomasello, Pine & Rowland, and Stromswold).


[23]     Derivational vs. Inflectional Morphology Perhaps one of the most interesting findings that have come out of psycho-linguistic research has been the observation that Derivational Morphology seems to be stored as lexical stems. For some time, it was thought that since derivational morphology was a language processing, that area of the brain that would light-up in such tasks would be the frontal lobe. In fact using fMRI methods, it has been found that words such as teacher /teach/-/{er}/ are stored and processed in the same temporal-lobe area of the brain as the word brother /brother/ (see Clashen 1999, Clahsen et al. 2001).


[24]    Gordon's 'rat-eater test' (1985) similarly suggests that Derivational morphology is more akin to Compounding'both processes involve either (i) taking a stem and adding a meaningful affix to it (the former), or (ii) taking a stem and adding it to a second stem (the latter). Both are seen in fMRI studdies as temporal-lobe tasks. This is a stark contrast with INFLectional morphology by which an affix bears no meaning and by which 'part-of-speech' words are not derived. In sum, Derivational morphology mimics associative word learning on every dimension, whereas Inflectional morphology mimics abstract rule formation.








2. Summary and Concluding Remarks


[25]     While I agree with much of the material in the text'clearly, once again, Hoff has done her homework here'my only substantive critique of the new edition is that perhaps more could be done to pool the many differing data-analyses and theories together under a single roof of debate. I feel that much of the impact is lost regarding the two main schools-of-thought. There are a number of issues that present themselves early on in Chapter 1 that could rightly serve as a means to clearly cast both sides of the argument. It seems to me that while the full range of arguments is included and equally represented without bias (and to Hoff's credit), very little effort is made into bringing the issues to the fore in a meaningful way so that students can come way after reading the text clearly able to assess the differing theories. Whenever students converge the outstanding issues on both sides into a singular cohesive/coherent framework, they emerge with a higher level of understanding that goes far in accounting/explaining what they themselves will eventually witness in their own classroom. One example of how the debate has taken on new life has been a San Francisco based pilot study for reading (called 'Open-Court'). The reading approach utilizes rules of phonology, bottom-up, in totally new and productive ways. One issue that has resurfaced out of the approach is the debate over Whole Language vs. Phonics. Again, as you can see, the two sides more or less embody the classic debate of associative-learning vs. rule-based learning, and my students come way feeling very satisfied that they, in fact, understand the working hypotheses driving both approaches.


[26]     Finally, it could be argued that the 3rd edition might wish to find some room for syntactic tree diagramming for Chapter 5. There might be a way to do this without getting too bogged down in syntactic theory. At least, some discussion of functional features such as [+/-Agr] having to do with INFLection could capture some recent work in syntactic 'Feature Theory'. When early MLU speech is discussed, perhaps (semantic based) NPs could be drawn (as opposed to full DPs) in order to show the lack of D-feature specificity, [-Def]initeness, Case, Number.



            On the other hand, I am in no way advising the path seemingly taken by Guasti in her new text 'Language Acquisition'. As it stands, it is much too syntactic in nature and lopsided in leaning toward Chomsky. Clark's new text 'First Language Acquisition' contrastively falls on the other side of the spectrum in an equally lopsided way. (It is in this sense that)...Hoff continues to serve a broad-based, heterogeneous student population... I'd like to think she got it right.



3. Overall Review and Recommendation


[27]     As a result of past reviews (2nd ed.), the Biological Bases chapter was brought forward from what was originally intended as the final chapter to the second chapter in the text. This was seen as a beneficial move that allowed Prof. Erika Hoff to focus on the 'Biological theme' throughout subsequent chapters. My own 'theme' of this review follows in a similar vain: I suggest that the Biological Bases/Maturational factors, which are underwritten by the DMM, be emphasized early-on in the text and then advanced in the relevant sections as a kind of springboard from which a multitude of topics can be presented and discussed. There are moments, in the second edition, where traces of such a theme can be uncovered, but, as it stands, the 'biological theme' component is scant, lacking any explicit commitment on the part of the author. The central idea that language development is biological, and thus discontinuous by default, can be adequately fashioned in such a way as to preserve both the evenhandedness that Hoff pursues in presenting the range of theories while allowing maturation to serve as a guiding principle behind much of the theory. Recall, it is not so much the debate between 'Nature vs. Nurture', but rather the 'Nature of Nature'. In short, even the strong eliminative connectionists could be convinced into coming down squarely on the notion that language is biologically based. So, much of the meandering here can take a back seat.


[28]     The good word here on this third edition is that Prof. Hoff continues to write in a well detailed manner, a clear narrative replete with the major research paradigms currently being employed today. She has done her homework once again in gathering much of the current data-analyses and theoretical approaches on offer. She continues to take an evenhanded approach in representing the gaggle of personalities who very often cling to a variety of positions at any one time on any one topic. For linguistics, such treatment and professionalism is becoming quite a rare commodity. The chapters all follow a nicely thought-out timetable in how certain material/methodological pursuits follow chronologically/ideologically. The chapters, I find, are iconic in a two-dimensional way: (i) they begin with an Introduction, and slowly work-up into more complicated concepts; (ii) the chapters themselves unfold naturally giving a sense of development, moving from phonology (the first area of language that comes on-line) through to syntax (the last area to mature). I wholeheartedly recommend this new edition as an advanced undergraduate-level reader in language development.











4. Appendix


[29]     A. U-Shape Learning (Phonology/Morphology)

            (i) Phonological U-shape learning (cited from Hildegard, Leopold 1939-1949)


-- Double consonant CC reduced to a sole consonant onset C (= CV stage of development)

-- Default voicing assimilating the [-voice] bilabial plosives /p/ to [+voice] /b/ and alveolar /t/ to /d/.



(ii) Morphology U-shape learning (cited from Marcus)


' Representational rule application: Over-regularization of inflectional morphology {-ed} (24m)

B. Narrative language differences between William's syndrome and Down's syndrome (Lenhoff et al. Scientific America. Dec.1997, pp. 68-73).


            Williams subject, age 17, IQ 50                                  Down subject, age 18, IQ 55

            'Once upon a time when it was dark at                      'The frog is in the jar. The jar

            night, the boy had a frog. The boy was                      is on the floor. The jar on the

            looking at the frog, sitting on the chair,                      floor. That's it'.

            on the table and the dog was looking

            through, looking up to the frog in a jar.

            That night he sleeped and slept for a

            long time, the dog did. But the frog was

            not gonna go to sleep. The frog went out

            from the jar. And when the frog went out,

            the boy and the dog were still sleeping.

            Next morning is was beautiful in the morning.

            It was bright and the sun was nice and warm.

            Then suddenly when he opened his eyes, he

            looked at the jar and then suddenly the frog

            was not there. The jar was empty.

            There was no frog to be found'.

            While William's children have very low non-verbal IQs as well as cognitive motor-skill incapacity, they nonetheless (and in contrast to SLI or Down's children) show quite unique, highly expressive language abilities which exhibit otherwise preserved functional grammar. It has been suggested that was has gone amiss is their associative-based lexical retrieval capacity which might explain why they often use exotic words for mundane things.






C. Sketch of the 'Less-is-More Hypothesis' (Newport)

As suggested by Newport, the inherent problems with learning a second language (L2) has to do with the high memory capacity post-critical learners bring to the data, further enhanced by their high functioning cognitive analytical capacity. Children start out with less memory and so bootstrap their initial grammars to those more salient lexical properties of a language, thus securing true first language acquisition. This initial basic bootstrapping which avoids linguistic complexity is paramount in allowing to children to build-up an eventual formal language system. (See also Elman ( ) for 'computer-learning' grammar models, as well as Felix ( ) for an L2 interpretation).








[30]     D. A Pedagogical Dual Model:       

5. References

Alegre, M.A, Gordon, P. (1996) 'Red rats eater exposes recursions in children's word formation'. Cognition 60, 65-82

Clahsen, H. (1999) 'Lexical entries and rules of language: a multi-disciplinary study of German inflection'. Behavioral and Brain Sciences 22, 991-1060.

Clahsen, H. Sonnenstuhl, I. Blevins, J. (2001) Derivational morphology in the German mental lexicon: a dual mechanism account. Ms. University of Essex.

Elman, J. (1993) 'Learning and development in neural networks: The importance of starting small'. Cognition 48, 71-99.

Felix, S. (1987) Cognition and language growth. Dordrecht: Foris.

Gross, M., Say, T., Kleingers, M., Munte, T.F., Clahsen, H. (1998) 'Human Brain Potentials to violations in morphological complex Italian words'. Neuroscience Letters 241: 83-86

Hurford, J., Studdert-Kennedy, M., Knight, C. (1998) Approaches to the evolution of language. CUP.

Marcus, G.F., Pinker, S., Ullman, M., Hollander, M., Rosen, T. & Xu, F. (1992) 'Over-regularizations in language acquisition'. Monographs of the Society of Research in Child Development 57 (4, serial No. 228).

Marcus, G.F., Brinkmann, U., Clahsen, H., Wiese, R. and Pinker, S. (1996) 'The exception the proves the rule' Cognitive Psychology (cited in BUCLD 19, vol. 1 pp. 60-8).

Munte, T.F., Say, T., Clahsen, H., Schiltz, K., & Kutas, M. (1998) 'Decomposition of morphologically complex words in English: Evidence from ERPs'. Cognitive Brain Research 7: 241-253.

Newport, E. (1990) 'Maturational constraints on language learning'. Cognitive Science 14, 11-28.

Obler, L.K., Gjerlow, K. (1999) Language and the Brain. Cambridge Approaches to Linguistics: CUP.




Radford, A. and Galasso, J. (1998) 'Children's Possessive Structures: A Case Study'. Essex Research Reports in Linguistics, 19. 37-46 (Paper presented to the annual convention of the American Speech and Hearing Association in Nov. 1997).


Smith, N. and Tsimpli, I.M. (1995) The Mind of a Savant. Blackwell.


Wakefield, J. and Wilcox, J. (1994) 'Brain maturation and language acquisition'. Proceedings of the 19th annual BUCLD.

[1] The General Domain Theory here equates to Skinner/Piaget (=higher-scope language sits on top of lower-scope cognition), and the Specific Domain Theory equates to Chomsky (=a language faculty seated in a unique and autonomous module of the brain/mind, unconnected to any other cognitive modules).

[2] A good majority of the undergraduate students currently required to take linguistics due to California State credential requirements will find themselves essentially working amongst an extremely heterogeneous population of language disorder students. The one thing that repeatedly gets expressed in personal evaluations is how the DMM approach to material helped sort through the myriad of issues regarding SLI, as well as Asperger's and Williams' syndrome.

[3] I wouldn't like to give the impression that the DMM is sufficiently explanatory in its own right (regarding explaining language acquisition). The best case scenario suggests that the DMM correctly targets areas of the brain in conjunction with specific language tasks, facilitating our understanding of a modular approach to language. The worst case scenario (which couples the DMM to that of a Computational Theory of the Mind) suggests that the DMM is at best a re-worded effort to promote an ill-conceived and ultimately insufficient neo-Darwinian account of language evolution (cf. Fodor vs. Pinker).