How Children Learn Language
Cambridge: Cambridge University Press, 2005.
Reviewed by Heather Hilton
How Children Learn Language is a delightful introduction to the psycholinguistics of first language acquisition—to be consumed without moderation by linguists, their students, parents of small children, or just the intellectually curious. The book is written for "non-specialists in the field of language acquisition research" , and therefore makes an ideal textbook on L1 development for beginning students in applied linguistics or sciences du langage. Extremely readable and clear (O'Grady carefully limits himself to basic syntax and vocabulary), it is suitable for learners of English—even non-linguists, such as psychology or literature majors. Every point is illustrated with concrete examples of (highly entertaining) real child language, and simple summaries of relevant psycholinguistic research.
At the core of the book are five chapters describing lexical, syntactic, and phonological development in the first language—concentrating on the child's receptive abilities first, before considering production skills. Most of the examples given concern language development in Anglophone children—a potential weakness from a psycholinguistic standpoint, but a plus for English-language specialists.
The chapters devoted to lexical acquisition [chapters 2 and 3] cover various important phenomena: the importance of prosodic features in speech stream segmentation [13-16]; different strategies used by children to identify words [10-11 and 16-18]; the acquisition of inflectional [18-26] and derivational [26-38] morphology. Budding cognitive linguists will be interested in chapter 3, which summarizes universals of meaning construal [52-61], as well as fundamental differences in the acquisition of various parts of speech (nouns, verbs, adjectives, adverbs, prepositions).
Chapters 4 and 5 examine syntactic development, also from a productive and a receptive point of view. The reader is introduced to fundamental concepts in the consideration of syntactic mastery, such as mean length of utterance (and how it develops in early childhood) [83-86], pivotal words [86-88], the "processing bottleneck" in early spoken production [91-92], the acquisition of function words (and the child's capacity to recognize the appropriate function word long before s/he manages to produce it) [92-96], the acquisition of pronouns [99-102] and various complex syntactic structures [102-113]. The summaries of research investigating the child's understanding of various syntactic structures, and the acquisition of narrative expertise are particularly interesting [chapter 5].
The chapter devoted to phonological development [chapter 7] summarizes experiments illustrating the superiority of phonological perception to productive abilities throughout the first years of life, and gives a list of cross-linguistic universals in phonological acquisition. Examples of approximations in the phonology of Anglophone children clearly illustrate the phenomena of consonant deletion, substitution and assimilation [153-59]. The forces at work at this ontogenetic level also play a role in the evolution of human language, of course, and students of historical and general linguistics will gain a concrete grasp of many other phenomena that play a fundamental role in language acquisition, use, and change (the importance of frequency and saliency, for example).
The final chapter of the book provides a useful summary of the "emergentist" position on language development currently held by most cognitive psychologists. Native-language fluency is certainly one of the most complex skills we acquire in our lifetime (if not the most complex), but our L1 expertise depends on the same general learning mechanisms that enable us to acquire other complex skills (such as touch-typing, driving a car, or playing a musical instrument). Emergentists reject the notion of a specialized "language acquisition device," or innate, abstract language "blueprints"  (as postulated by Universal Grammar). A linguist himself, O'Grady attempts to maintain a position open to both linguistic universals and the emergentist perspective, suggesting that certain principles of word class category may be genetically coded, but that most of the work of L1 development follows basic cognitive learning principles [189-90].
Two short sections are appended to the book. One presents methodological tips for recording and analyzing child language, including an explanation of how to calculate mean length of utterance—a productive measure which is useful in assessing L2 proficiency, as well as L1 development. The second appendix provides a mini-introduction to phonetics, and caught my attention because it includes a radically-reduced list of only 12 (American English) vowel sounds. A minimalistic list of this sort is no doubt coherent with the international pronunciation of lingua franca English, and therefore potentially useful for non-specialist students.
O'Grady's bibliography is relatively complete, including essential references in both experimental psychology and linguistics ranging from the 1960's to very recent studies. For the purposes of readability and simplicity, O'Grady has decided to indicate his references via a complicated (and out-dated) footnote system; this is the one laborious feature of an otherwise comfortable scientific reading experience.
Since the European "Reform Movement" of the 1870s, foreign language teaching methodology has been influenced by over-simplified notions of first language acquisition: we learn by simple imitation (La Méthode directe), by conditioning (Audiolingualism), or by simply engaging in communicative interaction (the Communicative Approach). O'Grady's book is therefore particularly useful for researchers and language teachers, who need a better grasp of what actually goes on during first language acquisition. The author makes it clear that L1 learning is by no means simple, even though it is impressive: "several hundred exposures"  may be required before a child gains productive mastery of a single language form; the complex knowledge base that we develop for our native language before the age of five is derived from a spoken corpus of several million sentences of aural input . In most cases, these sentences are tailored to the linguistic needs and abilities of the child. With relatively few exceptions (discussed by O'Grady in chapter 7), the kind of language input children receive is exactly what they need to "bootstrap" their way through the lexicon and syntax of the L1: highly contextualized, very repetitive, with restricted vocabulary and syntactic structure, frequent paraphrasing, more pauses, and exaggerated prosodic features [175-80].
Advocates of the Communicative Approach hypothesize that communicative interaction is all that is necessary for successful L2 acquisition—but with only 2-3 hours of L2 class time per week (nine months of the year), and a certain percentage of this potential exposure time spent communicating in the L1, the implicit memorization of thousands of new language forms simply will not occur. In addition, teaching materials produced for beginning-level language students in France tend to expect them to understand and produce complete syntactic units from the very first lessons ("What's your name?" "I'm seven years old," etc.). One can't help wondering if the functional syllabus of the 1980s could be usefully supplemented at the beginning level by a curriculum devoted to the mastery of sub-syntactic units of oral and written language (contextualized, of course, as child-directed speech always is), before focusing on complete utterance production. We certainly have a tendency to underestimate the importance of receptive work in the language classroom: at all levels, toddlers can distinguish between correct and incorrect language forms long before they can produce the correct forms themselves. Wouldn't it be interesting to design exercises harnessing this ability to learn phonological, prosodic, lexical, morphological and syntactic regularities implicitly? A computerized environment might prove ideal for statistical learning of this type.