This is the first study that empirically investigates preposition placement across all clause types. The study compares first-language (British English) and second-language (Kenyan English) data and will therefore appeal to readers interested in world Englishes. Over 100 authentic corpus examples are discussed in the text, which will appeal to those who want to see 'real data'
Date: Thu, 15 Dec 2005 13:45:35 +0100 From: Annette Hohenberger <firstname.lastname@example.org> Subject: Kids' Slips: What young children's slips of the tongue reveal about language development
AUTHOR: Jaeger, Jeri J. TITLE: Kids' Slips SUBTITLE: What young children's slips of the tongue reveal about language development PUBLISHER: Lawrence Erlbaum Associates YEAR: 2005
Annette Hohenberger, Max Planck Institute for Human Cognitive and Brain Sciences, Munich, Germany
''Kids' slips'', by Jeri J. Jaeger (henceforth, JJJ), is the first monograph about young children's slips of the tongue (henceforth, SOTs), based on the biggest collection sampled to date of child slips from monolingual English-speaking children from 1;7 -- 5;11 years of age, mostly stemming from her own three children, Anna (AN), Alice (AL), and Bob(by) (B).
The book is organized into six chapters. In the first chapter, the SOT approach to studying child language is introduced. In the second chapter, children's and adults' slips are compared. The remaining four chapters are devoted to phonetics/phonology, the lexicon, semantics, and morphology/syntax. Each of these four chapters is self-contained and may be read independently, although, of course, they build upon each other and there is ample cross referencing. The body of the book is followed by the complete and annotated corpus of the children's slips (n = 1383).
The purpose of her book, according to the author herself, is (i) to use children's slips as evidence for the development of linguistic representations and language production planning in phonology, morphology, syntax, and the lexicon and (ii) to lay out a common methodology as to guide future researchers in their (cross-linguistic) study of children's slips.
METHODOLOGY: How can we know if a child has made a slip? This crucial methodological problem is only manageable if (i) the researcher knows the current grammatical system of the child on the background of which a systematic deviation due to lack of competence can be ruled out and/or if (ii) the child possibly indicates a SOT herself by means of a self-repair or otherwise shows signs of having erred (such as looking confused). Given that the author has this intimate knowledge of the three main subjects of the study -- her own children - - and given that external error sources (unavoidable in a pencil-and- paper collection) such as perceptional and collectors' biases have been shown to not systematically distort off-line slip collections (for a survey, see Poulisse, 1999), the methodological problem of child SOT research can be considered as settled, in this study.
CLASSIFICATION: The SOTs have been arranged in a coherent classification system which JJJ mainly takes over from adult SOT research. The errors are classified according to: (i) the LINGUISTIC COMPONENT, resp. the PROCESSING STAGE. Is this a phonological, lexical, syntactic, or propositional error? (ii) the DIRECTIONALITY of the error, if it is a contextual/syntagmatic error. Is this an anticipation, perseveration, or exchange? (iii) the FORM of the error. Is this a substitution, addition, omission, exchange, blend, telescoping, or multiple error?
MODEL: JJJ puts forward a REPRESENTATIONS AND PROCESSING COMPONENTS MODEL (RPC, pp 6-11) which is based on Levelt's (Levelt, 1989, Levelt, Roelofs, and Meyer, 1999) serial-modular model of language production but which extends it in specific ways. The RPC model distinguishes representations in the long term memory such as concepts, lemmas in the content lexicon, and forms in the form lexicon on the one hand from planning processes in short term memory such as lexical selection of words, determination of functional structures and selection of syntactic structures on the other hand. All SOTs can be allocated to respective levels of processing involving specific linguistic representations. In any two-stage model of production, access to a word's syntactic features and meaning (its lemma) and morpho-phonological form happens on two different levels of planning. Misselection due to semantic similarity may arise in the first stage during lexical assignment to functional structure resulting in a semantic substitution such as ''I wanna <b>watch</b> ... I wanna <i>listen</i> to Baby Beluga'' (B-71, 2;6.2). Misordering of phonological segments, however, happens in the second stage when the word's form is accessed and its phonological units are filled into the respective positional slots, resulting in e.g. an anticipation such as ''A stecond story ... a second story'' (AL-235, 4;9.18) where the phoneme [t] of ''s<i>t</i>ory'' is anticipated and added to the planned word ''second''.
CHILDREN'S vs. ADULTS' SLIPS: In order to compare children's to adults' slips, JJJ had collected a reference corpus of n=716 adult slips, which can be accessed at http://linguistics.buffalo.edu/people/faculty/jaeger/adultSOT.html. One crucial precondition in this respect is that both children's and adults' slips must be similar enough to fit into the same classification system and model. In a global survey, JJJ establishes such a comparability in terms of error types and error proportions. She concludes that children's slips are very similar to adults' slips, overall. Differences in the processing span of contextual errors, directionality, formal structure, units, and self-corrections can be attributed to the development of linguistic representation, working memory, and the monitor. Given these overall similarities and specific differences, the value of children's slips can be fully appreciated since ''they allow us to tap into the time-course of the development of specific aspects of language, by seeing how the substance of the errors changes over time'' (p 90).
PHONETICS AND PHONOLOGY: Phonological SOTs are the most common slips in adults as well as in children. Their onset indicates that children no longer treat words as unanalyzed wholes but as units with a structural frame and segmental content. The onset coincides with the vocabulary spurt at around 18-21 months when children have acquired a critical mass of about 50 words suggesting a major reanalysis of words in the mental lexicon in terms of phonological representation. From that time on ''... the segment has primary status in the phonological aspects of young children's production planning'' (p 109). For both adults and children, it is feature similarity which accounts strongest for the likelihood of a phonological SOT. Consonantal phonological SOTs in which error and target segment differ in only one of the five relevant distinctive features, namely place of articulation (POA), +/-continuous, +/-fricative, +/-voice, and +/- nasal, are most frequent. Vowels are less frequently affected by slips than consonants, in both children and adults. In vowel errors, besides feature distance, stress figures prominently: it is two stressed vowels which are most likely involved in a vowel slip.
An excellent example of how children's slips can pinpoint the onset of crucial developments in phonological representation and processing is provided by allophonic variation. In adult slips, if a novel phonological sequence is generated by the slip, the correct allophonic variant is automatically produced, as in the (additive) anticipation ''siŋks inches'' <-- six inches (OC-88, 4;1), where the alveolar nasal [n] becomes a velar nasal [ŋ]. Children, however, have to learn which allophones there are and the rules producing the correct variants. Low-level assimilations such as nasalization of the vowel before a nasal C or POA assimilation of C's within a syllable (see above) are present from the very outset whereas POA assimilation across syllables or (de-) voicing of sonorants in C-cluster occurs only after 2 years of age, when the relevant contexts have been acquired. Specific phonological processes in English such as velarization of [l], +/-aspiration of plosives, or tapping are learned even later.
The syllable as a supra-segmental domain constrains phonological SOTs in adults in important ways. In languages with hierarchical syllable structure such as English, the 'syllable position constraint' allows only segments in identical syllable positions to interact, i.e., onset with onset, nucleus with nucleus, coda with coda. Until the second year of life, however, infants entertain a more holistic representation of the whole word which is successively refined into a linear and hierarchical structure with highly specific phonological content. Only after 2;2 years of age is the syllable position constraint obeyed suggesting that the syllable has superseded the whole word as basic structural unit of phonological processing (p 215). Moreover, the prevalence of consonantal onset errors from 2;6 years on suggests the mastery of the onset-rhyme analysis of English.
LEXICAL ERRORS: Lexical errors come in two kinds: paradigmatic errors (substitution, blend) and syntagmatic errors (anticipation, perseveration, exchange, telescoping). The former are due to selection errors; the latter are due to linearization errors. Substitutions also come in two kinds: semantic substitutions and formal substitutions, so-called 'malapropisms'. While in both adults and children, (besides same lexical category) semantic similarity is the most influential factor accounting for a substitution (children 69%; adults 70%), phonological similarity is less influential in children than it is in adults (40% vs. 55%, p 282). This result suggests that of the two big dimensions by which the mental lexicon is organized -- meaning and form -- semantics is present earlier and remains more influential than phonology which develops only later and plays a secondary role. Besides categorical, semantic, and phonological similarity, other factors, too, contribute to lexical errors, among them tonic stress, previous utterance, rhythmic weight and environmental influences. Usually, 3-5 such factors are involved in a single lexical slip.
Above all, identity of lexical category figures prominently in speech error research as a severe constraint on the possible interaction of words, often referred to as the 'grammatical category constraint' (for a survey, see Poulisse, 1999). According to this constraint, nouns may only interact with nouns, verbs with verbs, etc. On a more general level, it also pertains to the two broad classes of lexical vs. functional categories which also rarely interact. Identifying the onset of the lexical category constraint is relevant to the question if young children organize words in terms of lexical category from the very beginning or not, and if not, in terms of which other system they may conceive of them. At first glance, children seem to respect this constraint as much as adults do (88% vs. 86,5%, p 291), from the outset. However, by looking at the earliest lexical slips in the 1- and 2-word stage, JJJ finds that they are strongly motivated by other than strictly lexical categorical factors, mainly by conceptual but also phonetic similarity, environmental and contextual influences. Given this strong determination by non-linguistic factors, JJJ proposes to link the onset of the lexical category constraint with the onset of syntax which both start at around 2;6 years of age. This linking is also corroborated by the finding that at around 3 years there is a (modest) peak of syntagmatic as well as function word errors creating two coupled U- shaped curves of development.
SEMANTIC RELATIONSHIPS IN LEXICAL ERRORS: Having established that semantics has the strongest influence on lexical errors from early on, JJJ asks further which the semantic relationships are that figure importantly in this respect. For this purpose, she develops a classification scheme (pp 320ff) comprising (i) taxonomic sets ('fruit -- apple, pear'), (ii) non-taxonomic sets ('pillow -- blanket') or co-members of a set ('Mommy -- Daddy'), (iii) partonomic sets ('hand -- finger'), (iv) contrastive relations ('slow -- fast'), (v) synonymy ('bunny -- rabbit'), (vi) associative relations (metonymy and connotation), and (vii) function words (esp. pronouns). Conceptual- semantic categories such as pairs of concrete actors belonging together, often proper names such as 'Mommy -- Daddy', 'Bert -- Ernie', are among the first relationships and may be considered precursors of more elaborate lexical-semantic taxonomies involving co- hyponyms ('apple -- pear') under a common hyperonym ('fruit'). Overall, the semantic relations between words are strikingly similar in children and adults: they mainly wallow in co-hyponymical relations, followed by associative errors. However, the importance of the relationships changes over time. In the beginning, binary coordinated relations dominate. In this context, the ''lemma'' is an important concept, mediating between conceptual and lexical semantics. Consider the two interactants in a semantic substitution, ''watch'' and ''listen''. Conceptually, these are quite different actions -- lexical- categorically, however, they make great co-hyponyms in a common taxonomic hierarchy. The latter organization into lemmas amounts to an abstraction over a semantic domain which supersedes a more holistic conceptual organization of words. This reorganization takes place between 1;7 and 2;6 years of age and is intimately related with the onset of syntax.
MORPHOLOGY: Morphology is traditionally linked to syntax as well as to lexical word formation processes. In syntax, inflectional morphological affixes are attached to the syntactic frames, which provide slots for lexical roots. In the lexicon, morphological affixes attach to word frames and generate novel words. Morphological errors come in three kinds: morphological shifts such as ''two bubble<b>s</b> gum_'' (AL-11, 2;2), morphological substitutions such as ''I've been suck-<b>ed</b> on'' <-- suck-<i>ing</i> (B-456, 5;1) and root morpheme exchanges (so-called 'stranding' errors) such as ''my <b>run</b> is <b>nos</b>-ing'' (AN-57, 2;7). Overall, inflectional errors are more abundant than derivational errors, in children even more so than in adults. Children do not yet have such a rich lexicon with many derived words, whereas they do have already productive syntactic processes. Since the bonds between roots and inflectional affixes are weaker than the bonds between roots and derivational affixes, there are more inflectional errors. Only highly productive derivational morphemes are also stranding-prone (p 409).
Inflectional morphemes are strongly related to syntax and thus are more frequent in syntagmatic errors (e.g. morpheme shifts) than in paradigmatic errors (e.g. substitutions). Again, the onset of inflectional morpheme shifts such as ''two bubble<b>s</b> gum_'' coincides with the onset of syntax (p 405). It takes until 2;2 years of age for the child to factor out the abstract morphemes and make them available as affixes to syntactic frames. Since children do not know so many poly- morphemic words, they have no reason to analyze their morphological structure and to represent it in lexical entries. It takes even longer, until the 6th year of life, until children can reliably segment stem and bound morphemes correctly (p 456). During this time they commit more morpheme shift errors than adults, in poly-morphemic words and multi-word phrases, since the localization of the morpheme in the respective word frame or syntactic template is not yet as fix as in adults. Children show the same prevalence as adults of inflected words to become involved in a stranding error as compared to derived words.
Like phonological similarity, morphological similarity, too, enhances the likelihood of a lexical substitution error. The interactants share the same morphological structure in children's as well as in adults' slips (84% vs. 76.5%), most clearly when two mono-syllabic words are involved. Adults, due to their bigger and more refined lexicons, represent more morphological information and have more poly- morphemic words (esp. Latinate words). The strongest argument for the representation of morphemes in the form lexicon comes from malapropisms: they always show higher morphological similarity as opposed to semantic substitutions (86% vs. 81% for children, 81% vs. 74.5% for adults, p 443).
SYNTAX: Parallel to her claim of an early stage of a purely conceptual organization prior to a lexical organization, JJJ also claims an early semantic prior to a syntactic stage. The evidence against a syntactic account comes from SOTs in the 1- and 2-word stage in which there are (i) no morphological errors, there is (ii) no evidence of lexical categories, (iii) no influence of morpho-semantic categories (such as +/-transitive; mass/count noun), nor of (iv) phrasal tonic stress on SOTs. Only upon entering into the 3-word-stage do all these features appear (p 480).
The only SOT category directly related to syntax are phrasal blends such as ''I <b>won't let you to</b>'' (AL-71, 2;10), a blend of ''I <i>won't let you</i>'' and ''I <i>don't want you to</i>'' (p 468). Along the lines of the 'competing plans hypothesis' (Baars, 1992), phrase blends occur when two equally suited syntactic frames for the same propositional content are activated and the processor, instead of selecting one, blends parts of both frames into a new hybrid phrase.
Early phrase blends tend to be local, involving two individual lexical items whereas later, they become more global, involving advanced syntactic knowledge and necessitating quite complex planning of multi- phrasal sentences (p 474).
EVALUTION OF THE BOOK
With ''Kids' slips'', JJJ has put forward the biggest collection of children's slips to date, analyzed with an unprecedented level of systematicity and specificity. This is the standard work on children's slips -- an absolute must for every researcher in the field of language acquisition and language production. ''Kids' slips'' is a highly significant contribution to the field of ''developmental psycholinguistics'' in that it investigates the on-line production of spoken language in young children. Besides the impressive data base, the merits of this study lie specifically in the theoretical explanation of the empirical data and phenomena. JJJ shows in an illuminating way how children's SOTs bear on specific research questions and longstanding controversies in child language research.
In the following, I would like to highlight the main methodological advantages (M1-M6), point out some problems (P1-4) and then discuss some recurrent themes of the book (R1-R5).
JJJ sets very high methodological standards and lives up to them on every of the 727 pages of this monograph.
M1: OBJECTIVITY, RELIABILITY, AND VALIDITY: JJJ has taken great pains to transcribe the children's utterances phonetically -- which requires a decent amount of ''multi-tasking'' at the time they were collected: being attentive and swift so as to notice the slip in the first place, then bringing to bear her excellent competence in phonetics/phonology to the task of noting down the slip as accurate as possible while managing the ongoing complex everyday situation in which the slip occurred. With such a detailed phonetic representation, a basic requirement of objectivity of the data collection is satisfied. The reliability of the data classification can be taken for granted since she analyzes the child data base with the same classification system as for the adults. With respect to the validity of the inferences drawn from the data analysis, it is an indispensable necessity of the collector to know the current competence of the children under observation, in order to tease apart true performance errors (the SOTs) from competence errors due to the ongoing development (pp 11ff). To fulfill this requirement to the highest extent possible is a privilege of a linguist-mother who, in a short and precious period of life, integrates the otherwise contradictory twofold preoccupations of exhaustive daily child care and excellent scientific research. By comparing the children's slips to an adult reference corpus, she can factor out the developmental effects which are at the heart of her study.
M2: THEORY-NEUTRAL FRAMEWORK: It is JJJ's explicit intention to make her data collection available to researchers from various theoretical approaches to language production (e.g., serial-modular vs. interaction) as well as language acquisition. Therefore, she has abstained from too narrow an analysis and rather chose some very broad conceptions of language production and development such as the 'competing plans hypothesis' (Baars, 1992) as an explanation of the mechanisms behind SOTs and the 'frame-content metaphor' (MacNeilage, 1998, MacNeilage and Davis, 1993) as an account of language production development. However, she also takes a firm stand with respect to the model in terms of which she analyzes her data, viz. the Representations and Processing Components (RPC) Model of Speech Production Planning, which is an advancement of the Levelt model. She explicitly invites scholars from different accounts to re-analyze her data in terms of their own models and encourages them to collect children's slips in other languages in order to be able to make cross-linguistic comparisons -- a big desideratum for future research.
M3: COMPARISON WITH OTHER CHILD SLIP CORPORA: At the same time she always compares her own results to those of other child slip researchers, namely Stemberger (1989), Warren (1986), Aitchison & Straf (1982), Poulisse (1999), and Wijnen (1992) as well as to other adult corpora. Most of the times, her results converge with those of the others. At times, however, she finds differences which call for an explanation. In her comparison of the proportion of paradigmatic lexical substitution errors which are related by semantics or only by phonology, for example, she looks at eight different corpora, child and adult. Although the numbers vary greatly from study to study, they yield converging evidence for a prevalence of semantic over formal errors. Differences in the samples stem from different methodologies and classification systems. In this respect, she underlines again the need for a common methodology and fully explicit classification criteria (p 253). But obviously it is also sample size which matters. While there are many large-scale adult slip corpora, hers is the first child slip corpus at a comparable scale. Without belittling the wealth and value of the other child corpora, JJJ's corpus sets the standards for future child SOT research.
M4: HYPOTHESIS-TESTING: Related to her theory-neutral stance in her basic methodology, she applies a classical hypothesis-testing approach to her empirical data. She formulates a hypothesis, tests it against her corpora, and explains the results in terms of her RPC model. With respect to alternative models, she always makes explicit what the data should look like if they were to fit in a specific model. To give an example, she asks if there is feedback from the formal to the semantic level in the child processor -- a hotly debated topic between modular-serial and interactive models. From the lack of any formal similarity in children's semantic substitutions she concludes that this is most likely not the case (p 373). However, she does not conclude that the interactive model is wrong, since (i) the effect is found in the adult data of hers and other researchers and (ii) its lack in children can be explained independently, e.g., in terms of their weaker monitoring capabilities.
M5. NOVEL MEASURES: In her hypothesis testing approach, JJJ often has developed novel measures or refined existing ones in order to answer very specific questions, e.g., a measure of semantic similarity between the two interactants in lexical errors or a score of multiple converging influences on lexical errors. As for semantic similarity, she comes up with seven categories (coordinate, subsumative, contrastive, associative, and opposed relations, synonyms, and pronouns), each with even finer subcategories to which she submits the subset of lexical errors. She can show that it is the opposed category which provides the first strong semantic coupling between items, such as 'Mommy-Daddy', 'Bert-Ernie', and that only later more taxonomic categories (co-hyponyms) are established. Thus, concrete, frequent, and closely linked pairs are the seeds of a later more refined and comprehensive lexical-categorical system. As for the multiple motivations of lexical slips she has ranked 10 influences according to their causal strength and has determined how many of them are found in each slip: (i) same lexical category, (ii) semantic similarity, (iii) tonic stress, (iv) phonological similarity, (v) utterance, (vi) weight, (vii -- x) various environmental and discourse related influences (pp 282ff). She found that the most common situation for both child and adult errors was that 3-4 influences conspired in a lexical slip (3 for paradigmatic, 4 for syntagmatic slips), suggesting that the planning mechanisms and lexical representations are the same for children and adults. Some of these novel measures are used for establishing baselines of chance frequencies against which the frequencies of particular slip categories of features are tested -- a very laborious undertaking but an indispensable control if valid claims about effects are to be made.
M6. HYPOTHESIS-GENERATOR: The greatest asset of her analyses is, however, not for direct reading but may only become realized in future research. What I find most intriguing in this book is that it may be used as an immensely rich generator of hypotheses to be tested in subsequent experiments. The problems JJJ asks herself are so precise and the application of the results in terms of particular models are so informative that a straightforward translation of the problem into a concrete experimental design is often self-evident. There is a whole set of problems that lends itself to further experimental testing, such as the lack of clear demarcation of syllables and distinction of the syllable position in young children < 2 years of age, the preference for binary oppositions in conceptualization, the interference of language production by contextual distractors, or the stronger coupling of production and perception, to name but a few. Any child psycholinguist will greatly appreciate the research impetus by JJJ's findings, providing a basis for further experimentation. The field of developmental psycholinguistics would immensely benefit if such an extended research cycle -- a broad corpus study followed by specific experiments enabling a reappraisal of the original findings -- could be established in child slip research. It is only recently that the synergy arising from the integration of different kinds of linguistic evidence -- through corpus linguistics and experimental approaches -- has become widely acknowledged (Kepser and Reis, 2005).
There are a couple of problematic aspects which make the reading and working through of this exceptional book quite hard sometimes. Having said this, I hasten to add that these problems are partly inevitably, given the very nature of the data.
P1: QUANTITATIVE AND QUALITATIVE ANALYSIS: This is the most comprehensive book about children's slips ever written with the biggest corpus of children's slips ever collected. This wealth of data is a mercy and a curse at the same time. On the one hand, slips are probabilistic in nature and major insights can be gained by looking at statistical distributions and frequencies, as in a nomothetic/quantitative approach. On the other hand, each slip is unique and deserves a careful analysis on its own, as in an idiomatic/qualitative approach. JJJ runs the risk of giving a too in-depth account of single slips and getting lost in too many details, at times, e.g., what the semantic differences between 'holes' and 'patches' are in a semantic substitution (p 362). After having gathered all these slips, Jaeger falls prey of hunting down each single one in her analysis until it has revealed all its marvels to us. However, this weakness is readily forgiven, since it is scientifically highly satisfying to follow JJJ through all the ramifications of e.g. the phonological structure of a particular SOT. After all, each single slip is a universe and deserves being admired for its intricacy and beauty, take for example the meanwhile legendary slip ''Not by the <b>ch</b>air of my <b>h</b>inny <b>h</b>in <b>h</b>in'' (Jaeger, 1992). Moreover, this meticulous analysis is greatly amended for by excellent summaries at the end of each chapter and subchapter.
P2: ANALYSIS TECHNIQUE: The major analytic tool JJJ uses is descriptive frequency statistics, which she displays in tables. Overall, this monograph contains 91 tables which resemble each other very closely. In a typical table, the frequencies of slip types XYZ are tabulated at the ages 1,2,3,4, and 5, a total is drawn an compared to the adults. But honestly, what else can one do about a slip corpus? JJJ has excessively exploited the tool of descriptive statistics and has gained major insights in doing so. She has made the best of this plain methodology, risking to bore the reader at times. Also, the lack of any inferential statistics is not worth complaining, as JJJ notes herself (p 48). Not much would be gained, indeed, by knowing that e.g. the frequencies for malapropisms in children vs. adults differ significantly. The only reasonable thing one can do about differences in the distributions, be they big or small, is to come up with good explanations in terms of a processing and developmental model and to look for further evidence corroborating particular findings. The search for novel measures (see M5) is surely related to the somewhat restricted statistical possibilities of analyzing the data.
P3. CLASSIFICATION SYSTEM: There is a slight oddity in the classification system insofar as JJJ does not primarily classify SOTs in terms of error unit (segment, morpheme, word, phrase) but with respect to processing stage, directionality, and form of the error. This is fine but in the case of lexical errors it leads to an artificial disruption of the unit 'word' which appears twice, on the two processing levels, 'lexical' and 'syntactic'. While it is certainly true that paradigmatic semantic substitutions and syntagmatic word anticipations occur on different processing levels, they involve the same unit, though. This division is actually overcome in chapter (4) on lexical errors where she looks at e.g. semantic and phonological similarity of both paradigmatic and syntagmatic lexical errors. This is not to say that her classification system is wrong -- it is not. It is just a matter of perspective. A unit such as 'word' can be involved in different kinds of errors on different stages of processing (lemma retrieval, syntactic serialization). However, the unit 'word' can also be tracked across processing levels.
P4. ERRATA: As is inevitable in such an opus, there are some very minor errors and typos. They are not the least disruptive and most of them will go unnoticed by the reader anyway. JJJ has registered them faithfully in a list of errata, which she kindly made available to me. Those who have bought the book may address JJJ by e-mail and ask her to send them the list of errata. The errors will certainly be corrected in the next edition of this book.
There are a couple of recurrent themes in this book relating to deep issues of language production and development.
R1. THE FRAME-CONTENT METAPHOR: JJJ finds MacNeilage's (1998, MacNeilage and Davis, 1993) 'frame-content' metaphor very well suited to account for the development in various processing domains, namely in phonology (where segments are filled into the slots of syllabic frames), morphology (where morphemes are fit into word frames), and syntax (where words are inserted into phrasal or sentential templates). A principled explanation of this similarity across structural levels is given by MacNeilage and Davies in terms of a species-specific organizational property of speech. In the evolution of the speech production process, independent control of structural frames and their content has emerged, possibly sustained by different cortical areas. In language production, the 'slots-and-filler model' of Shattuck-Hufnagel (1979) and subsequent models provide an explanation of how slots in a planned word become filled with segments, thus accounting for serialization and also for errors of serialization.
R2. PARALLEL BETWEEN PHONOLOGICAL AND SYNTACTIC DEVELOPMENT: JJJ finds striking evidence for the coincidence of serialization of segments within syllables with the onset of syntax at around 2;2 years of age (for a diagrammatic representation of the acquisition of syllable structure, see p 183). She accounts for this parallel in terms of a general systems strategy of dealing with growing complexity. ''Thus this linearization seems to appear throughout the linguistic system at approximately the same point in time, as the child's utterances become more complex and require a higher level of organization'' (p 183). The issue of linearization is intimately related to the frames-content issue above. The frames, in MacNeilage's terms, allow for the content units to become linearized by inserting them into the frame's pre-specified slots. From a generative perspective, the relation between hierarchy and linear order has most thoroughly been discussed by Kayne (1993). His 'Linear Correspondence Axiom' (LCA) requires the syntactic phrase marker to exclusively specify binary a- symmetric relations between structural positions such as specifier and head or head and complement so that the respective terminal elements (words) will not be in too symmetric a relation to one another. Only through asymmetry can an unambiguous linearization of elements be provided. Linearity is a crucial constraint on the articulatory-perceptual interface of spoken languages since the single temporal dimension of speech requires units to become strictly serialized. Syntagmatic errors and blends are speech errors that relate to this major constraint on language processing. JJJ has shown in the respective chapters how this ability evolves simultaneously across units (segments, morphemes, words) and levels of processing (phonological, morphological, and syntactic). In phonology, this development -- first frame, then content -- leads to the emergence of segments as basic content units and syllables instead of words as basic supra-segmental frames. Until the second year of life, around 2;2, infants entertain a holistic representation of the whole word as a frame which, successively, is refined into a syllabic representation which is now a linear and hierarchical structure with highly specific phonological content. In this respect, Levelt speaks of the 'phonologization of the lexicon' (1998, 175): by way of indexing structural positions such as onset, nucleus, and coda with matching consonants and vowels of the word, the child becomes able to distinguish huge numbers of syllabic gestures in the syllabary and thereby creates the precondition for rapidly acquiring more lexical items, as in the vocabulary spurt. According to Levelt, the phonologization of the lexicon is followed by the 'grammaticalization of the lexicon' at around 2;6 years of age. It is now the lemma information which becomes refined as to include information about the argument structure and theta grid of verbs. With this information, the processor can build up hierarchical sentential frames into which words are inserted. Note that JJJ assumes a coincidence of phonologization and grammaticalization of the lexicon whereas Levelt assumes a slight developmental prominence of the former over the latter. Both accounts, however, capture the important insight of a parallel structural development and also give the same explanation in terms of an emergent strategy of dealing with growing complexity.
R3. PARALLEL BETWEEN NOUNS AND CONSONANTS AND VERBS AND VOWELS: Another aspect of this parallel is manifested in the similarity in the behavior of nouns and consonant on the one hand and verbs and vowels in SOTs on the other hand. Both nouns and consonants are affected by slips above average, verbs and vowels, however, below average (even if frequency is controlled for). Why is this so? JJJ reasons that verbs as well as vowels are the core anchor of their respective structures, the syllable and the phrase. Their selection is a central issue on which the system obviously spends considerable effort. The significance of this parallel behavior is most clearly expressed in Carstairs-McCarthy's (1999) assumption that the noun-verb distinction actually derived from the consonant-vowel distinction as its evolutionary antecedent. Crucially, both nouns and consonants occupy parallel structural positions around the nucleus -- consonants occupy onset and coda positions, nouns occupy specifier and complement positions. Vowels occupy the nucleus in a syllable and verbs the head position in a phrase, respectively.
R4. THE ROLE OF LEXICON SIZE: Young children build up a mental lexicon in an amazingly short time. JJJ discusses how the lexicon develops with respect to the two major dimensions -- meaning and form. Form does not play a role in very young children's lexicons. Semantics, however, is relevant from the very beginning. She traces how the complex semantic network of adults with various taxonomic levels in the vertical dimension emerges from the first simple, concrete, binary oppositions such as 'Mommy -- Daddy', 'Bert -- Ernie', 'see -- hear' which only exist on a horizontal plane. She argues that lexicon size is the crucial driving force behind structure-building in the lexicon leading not only to an intricate semantic network but also to the emergence of morphology as an interface between form and meaning. It is the increasing number of lexical entries (of increasing length) which necessitates a formal morphological analysis. There has to be a rich substrate from which semantic as well as morphological dimensions can be factored out. This factoring out leads to a more coherent organization of lexical elements in terms of lexical semantics and morpho-phonological structure. Again, hierarchical structure is an emergent solution to the problem of handling a critical mass of lexical items in an economic way.
R5. INFLECTION vs. DERIVATION: Inflected and derived words behave differently in processing, as is well known. In the Levelt model, this difference is accounted for by treating inflected words such as ''read -- read-s'' as a single lemma (with diacritics for the functional information added by the inflection) but treating derived words such as ''read -- read-er'' as separate lemmas. JJJ accumulates rich evidence that inflected and derived words also behave differently in child SOTs. That inflectional suffixes are (i) only loosely integrated into their host word, (ii) extra-syllabic, (iii) detach more easily (iv) belong to syntax more closely is evidenced by a higher frequency of morpheme shifts (''bubble-<b>s</b> gum_''), stranding errors (''my <b>run</b> is <b>nos</b>-ing''), suffix substitution (''<b>turn</b>-ing'' <-- <i>talk</i>- ing) (p 409), and suffix adding. That derivational suffixes show exactly the opposite behavior is evidenced by SOTs such as ''<b>cold</b>'' <-- <i>clos-ed</i> where an un-derived word (''cold'') substitutes for a derived word (''closed''), without stranding its derivational suffix (*cold- ed, p 413). For the same reason, stranding errors and suffix substitutions are rarer with derived words. Thus, the different representation of derivational and inflectional morphology is also corroborated by child and adult SOTs.
''Kids' slips'' is JJJ's life work. It has only become possible by the curiosity and versatility of a very attentive mother-linguist observing her three children growing up and acquiring language during the first five years of their lives. Later on it required a decent amount of stamina to sift the more than thousand SOTs and to analyze them which she did with admirable accuracy and specificity. The fruits of this long lasting enterprise are now available to the benefit of all. JJJ explicitly invites us to re-analyze her data or to start collecting children's slips in the same way as she did, preferably cross- linguistically, so as to evaluate and extend her findings in the future.
Aitchison, J. and Straf, M. (1992). Lexical storage and retrieval: A developing skill? In: A. Cutler (Ed.), Slips of the tongue and language production, 197-241. The Hague: Mouton.
Baars, B. (1992). A dozen competing-plan techniques for eliciting predictable slips in speech and action. In B.J. Baars (ed.), Experimental slips and human error: Exploring the architecture of volition, 129-150. Plenum Press, New York.
Carstairs-McCarthy, A. (1999) The Origins of Complex Language: An Inquiry into the Evolutionary Beginnings of Sentences, Syllables and Truth. Oxford : Oxford University Press.
Jaeger, J. J. (1992): Not by the chair of my hinny hin hin: Some general properties of slips of the tongue in young children. Journal of Child Language, 19, 335-366.
Kayne, R. (1993): The antisymmetry of syntax. Linguistic Inquiry Monographs 25. Cambridge, MA: MIT Press.
Kepser, S. and Reis, M. (2005): Linguistic evidence. Empirical, theoretical and computational perspectives. Studies in Generative Grammar, 85. Berlin, New York: Mouton de Gruyter.
Levelt, W. J.M (1989). Speaking. From intention to articulation. Cambridge, MA: MIT Press.
Levelt, W. J.M (1998). The genetic perspective in psycholinguistics or where do spoken words come from? Journal of Psycholinguistic Research, 27, 167-180.
Levelt, W. J.M., Roelofs, A., and Meyer, A. S. (1999): A theory of lexical access in speech production. Behavioral and Brain Sciences, 22, 1-75.
MacNeilage, P. F. (1998). The frame/content metaphor of evolution of speech production. Behavioral and Brain Sciences, 21, 499-511.
MacNeilage, P. and Davis, B. (1993). Motor explanation of babbling and early speech patterns. In B. de Boysson-Bardies, S. de Schonen, P. Jusczyk, P. MacNeilage and J. Morton (Eds.), Developmental neurocognition: Speech and face processing in the first year of life. Dordrecht: Kluwer.
Poulisse, N. (1999). Slips of the tongue: Speech errors in first and second language production. Studies in Bilingualism 20. Amsterdam: Benjamins.
Shattuck-Hufnagel, S. (1979). Speech errors as evidence for a serial- ordering mechanism in sentence production. In W.E. Cooper, and E.T.C. Walker (Eds.), Sentence processing: Psycholinguistic studies presented to M.F. Garrett, 295-342. Hillsdale, NJ.
Stemberger, J. P. (1989). Speech errors in early child language production. Journal of Memory and Language, 28, 164-188.
Warren, H. (1986). Slips of the tongue in very young children. Journal of Psycholinguistic Research, 15, 309-344.
Wijnen, Frank (1992). Incidental word and sound errors in young speakers. Journal of Memory and Language, 31, 734-755.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Annette Hohenberger is a German psycholinguist who has done research on first language acquisition, spoken and sign language production, and the early cognitive development of infants. In particular, she has done research on slips of the hand and tongue and collected a (yet unpublished) extensive corpus of German child slips.