Date: Wed, 11 May 2005 10:48:04 +1200 From: "Kevin Mendousse (ARTS SELL)" Subject: Causes and Consequences of Word Structure
AUTHOR: Hay, Jennifer TITLE: Causes and Consequences of Word Structure SERIES: Outstanding Dissertations in Linguistics PUBLISHER: Routledge YEAR: 2003
Dr Kevin Mendousse The University of Auckland, NZ and Université de Paris-Sorbonne (FDC), FR
As acknowledged in the author's preface, CAUSES AND CONSEQUENCES OF WORD STRUCTURE originated as a PhD dissertation. As such, its purpose is to make a scholarly, scientific contribution to its field of inquiry and it is therefore primarily intended for the initiated linguist rather than for the novice reader.
Within a few paragraphs of Hay's opening statement that her book is "a book about morphology" (p. 3), the problematic is set within the theoretical framework of morphological decomposition and invokes the pressing need to uncover the factors determining the likelihood that a morphologically complex form will be decomposed during access, some forms being inherently more decomposable than others. While linguistic morphology has traditionally tended to focus on affixes and on accounting for (un)expected (dis)similarities in their behaviour, the author argues here for a different level of morphological abstraction following on from her experimental findings, which strongly suggest that predictions about the behaviour of specific affixes are best made when the focus is on the behaviour of individual words rather than on that of individual affixes.
Recognising that the recognition of words in processing a speech signal is a prerequisite to the listener's higher task of reconstructing the originally intended message, CHAPTER 1 "Introduction" (pp. 3-20) goes on to review major models of speech perception and morphological processing, along with various lexical and prelexical effects triggered by such factors as phonological transparency, temporality and relative frequency (lexical effects), metrical structure, possible word constraint and probabilistic phonotactics (prelexical effects). It further analyses the consequences of such effects for both words and affixes, and provides several clarifying disclaimers, including its acknowledgement that all experiments and claims put forward are strictly limited to the derivational morphology of English. The chapter concludes with a brief outline of the remainder of the book.
CHAPTER 2 "Phonotactics and Morphology in Speech Perception" (pp. 21-37) moves on to demonstrate that the use of phonotactics by English speakers for speech segmentation purposes has direct bearing on morphological processing. The discussion is initiated with a brief review of the growing body of evidence for the role of phonotactic patterns in speech perception, including data from neural network models. The author then offers a close examination of the results obtained from two experiments: experiment 1 was designed to implement a simple recurrent network, initially trained with monomorphemic words to use phonotactics for spotting word boundaries and then tested on a corpus of multimorphemic words; experiment 2 seeks to investigate the degree to which subjects exploit phonotactic probabilities in the morphological parsing of nonsense words.
Following up on these results, which indicate that the learning of word segmentation on the basis of phonotactics is likely to affect morphological decomposition and that phonotactic patterns can be used online for the decomposition of nonsense words into morphemes, CHAPTER 3 "Phonotactics and the Lexicon" (pp. 39-70) evaluates the consequences of this for the processing and the representation of real words. The author reports on the findings of a third experiment in which subjects displayed significant preference for words with low probability junctural phonotactics when asked to make intuitive judgements about the morphological complexity of real words. She goes on to present evidence demonstrating the long-term effects of this within the lexicon, namely that prefixed words that do not contain phonotactic information signalling a boundary are prone to a reduced perception of "prefixedness", a loss of semantic transparency, a proliferation of meaning and an overtaking of their base's lexical frequency. No such effects were however found by the author in the case of suffixes, a predictable fact apparently due to the left-to-right nature of lexical access.
In CHAPTER 4 "Relative Frequency and Morphological Decomposition" (pp. 71- 95), Hay moves away from prelexical to lexical processing in order to explore the role of lexical frequency, arguing that the more relevant frequency effect on morphological decomposition is one of relative frequency between the derived form and its base rather than one of absolute frequency of the derived form as traditionally held by models of morphological access and productivity. The chapter outlines the general assumptions found in the literature about the role of surface frequency, along with some results relating to the role of the frequency of the base. The author then turns to topical models of morphological access, showing that, where such models predict a role of lexical frequency with regard to decomposition, it is one of relative lexical frequency. Finally, she describes two experiments that attest to the fact that maximally decomposable forms are those which are much less frequent than their parts, and conversely: experiment 4 shows that subjects are more likely to rate forms with higher frequency bases as more complex than matched counterparts with relatively lower frequency ones; experiment 5 reveals that prefixed forms with high frequency bases are more likely to attract a pitch accent to the prefix.
CHAPTER 5 "Relative Frequency and the Lexicon" (pp. 97-122) turns to a synchronic examination of the English lexicon to establish the claim that if mechanisms of speech perception and production tend to make complex words more robustly decomposed than others, then evidence of this should be readily available in the lexicon. The author begins with a discussion of the overall frequency distributions of both prefixed and suffixed forms before experimentally investigating the role of frequency in such forms. Results based on the numbers and types of definitions appearing in Webster's 1913 Unabridged Dictionary suggest that relative frequency does indeed affect semantic transparency as well as polysemy, since derived forms that are more frequent than their bases have a tendency to drift away from the meaning of their bases as well as to proliferate in meaning. Polysemy is also shown to be related to absolute frequency, with higher frequency derived forms being less semantically transparent than lower frequency ones. A discussion of some methodological consequences of these results for current experimental work on morphological access concludes the chapter.
CHAPTER 6 "Relative Frequency and Phonetic Implementation" (pp. 123-137) draws on chapter 3's calculations over lexica and its speculation that many phonological violations across suffixal boundaries may, in fact, be resolved in the phonetic implementation because phoneme transitions across suffix boundaries are more likely to be more malleable than those toward the beginning of the word, which are more vital to word recognition. It describes an experiment designed to test the effect of the decomposability of suffixed words upon phonetic implementation. More specifically, the author investigates the implementation of /t/ when it occurs in a consonant cluster that straddles a morpheme boundary. The relative frequency of the derived form and base is found to be relevant to morphological decomposition, as is the strength of a morpheme boundary to the phonetics.
CHAPTER 7 "Morphological Productivity" (pp. 139-152) extends the study to include a discussion of morphological productivity. It provides a brief account of the most widely used metric P (which measures the category- conditioned degree of productivity) for quantifying morphological productivity, along with some of its reported advantages and shortcomings. It also illustrates ways in which morphological productivity has been modelled in the past. Distinguishing between the proponents of the scalar view and those of the absolute view, the author then goes on to examine the relationship of phonotactics and relative lexical frequency to morphological productivity by drawing on frequency counts taken from the CELEX lexical database (which includes counts from the COBUILD corpus) for forms that have productive affixes and monomorphemic bases. Arguing that productivity is a continuum that arises as a function of decomposed words in the lexicon, the author finally settles for the scalar view of productivity: the more an affix is represented by highly decomposable forms, the more likely it is to be productive.
CHAPTER 8 "Affix Ordering" (pp. 153-184) engages with the hotly debated problem of stacking restrictions amongst the derivational affixes of English where only a very small proportion of the numerous combinations is actually realised. Early attempts to account for apparent restrictions on affix ordering have often invoked some form of the Affix-Ordering Generalisation, which divides affixes into two levels L1 and L2, with L1 affixation occurring prior to L2 so that no L1 affix can attach outside L2 affixes. The stated problem is that such accounts of affix ordering are overly restrictive and draw the line at the wrong level of abstraction, while more recent work, which has dismissed altogether the idea of ordering restrictions (beyond selectional restrictions), misses a number of important generalisations. The claim put forward here is that both the range of generalisations about English stacking restrictions, along with their large body of systematic word-based exceptions, are best captured in terms of parsability. Taking the reader through a series of hypotheses, a critical discussion of affix classes, a case study of the denominal suffix -al and two experiments designed to test subjects' intuitions about the likelihood of -al suffixation to a range of -ment final forms to ascertain whether their preferences about non-words reflect the decomposability of the base, Hay explores the idea that affix-ordering constraints are related to the perception and storage of morphologically complex forms. She demonstrates that her maxim "an affix which can be easily parsed out should not occur inside an affix which can not [sic]" (p. 184), when combined with an understanding of the role of frequency in morphological decomposition, provides a better account of affix-ordering restrictions in English. The discussion finally extends to prefixes which, although less likely to co-occur, are involved in bracketing paradoxes contended to be cases in which a highly parsable prefix appears to have attached before a marginally parsable suffix.
The book, which set out to explore possible effects of speech perception strategies on morphological structure, comes to an end in CHAPTER 9 "Conclusion" (pp. 185-189), where the main findings are succinctly summarised. These include the role and use of probabilistic phonotactics for word segmentation, the relevance to and phonetic consequences of relative lexical frequency on morphological decomposition, the relationship underlying decomposability and morphological productivity, and finally the intricate link between affix ordering and decomposability.
Overall, Hay's text is equally informed and informative, and provides the reader with a truly insightful account of morphological decomposition and of how fundamentals of speech processing are responsible for determining the likelihood that a morphologically complex form will be decomposed during access. The originality of this contribution to linguistic morphology lies in its problematic which, because it brings together questions that are usually considered well outside the field of morphology (How do listeners process an incoming speech signal? How do infants learn to spot boundaries between words and begin to build a lexicon?), sheds new light on the kind of abstraction needed.
The proposed level of abstraction enables significant progress to be made on linguistic questions that have not generally been studied together. In particular, Hay's investigation of the causes and consequences of word structure demonstrates that a better understanding of morphological decomposability yields tremendous explanatory power, from fine phonetic details through to predicting how affixes can be used when neologizing and how they may co-occur. Importantly, her research shows that a thorough understanding of these phenomena requires a sophisticated knowledge of the morphological behaviour of individual words rather than of individual suffixes.
Hay's findings provide elements of answers to two problems recently regarded by Aitchison (2003) as central to the psychology of language and to the understanding of how humans cope with words. First (p. 127), is the mental lexicon one of words stored as single items ready for retrieval or is it one of disassembled morphemes pieced back together when needed? Second (p. 151), how and why does the meaning of words change?
The strongest point in this excellent piece of scientific research is undoubtedly the author's outstanding ability to draw on the previous literature to put forward innovative hypotheses, and to then test such hypotheses herself through the careful and creative design of experiments. The descriptions and explanations are always to the point and self- contained, the writing clear and effective, despite the presence of numerous typographical mistakes and occasional grammatical errors.
Evidently, the author has an exceptional ability to communicate sometimes difficult concepts in terms that "speak" to the reader, while encouraging reflection through a discussion that is interesting, relevant and often thought-provoking. This, combined with repeated reader-friendly summaries, theoretically well-grounded hypotheses and claims consistently backed up with statistical evidence, provides the key to the text's extraordinary lucidity.
An added advantage lies in the fact that, because the book does not assume the reader has any prior knowledge of statistics, it is relatively free of mathematical jargon. The author's scholarship is thus readily available to the intended readership as well as to the linguistic community at large. The drawback, however, is that this lack of a description of the mathematical apparatus would make it difficult for the reader ill-versed in parametric and non-parametric measures of evaluation to gauge effectively the significance of the findings or to follow up actively on the research.
As per the disclaimer in chapter 1, the book argues that speech segmentation strategies used by English speakers exert influence on English morphology but it does not examine the consequences of this claim for other languages. A crosslinguistic study to include data from other natural languages would therefore be a very interesting investigation to pursue in order to ascertain the language-specific or universal bearings of Hay's findings.
Aitchison, J. (2003), Words in the Mind - An Introduction to the Mental Lexicon. Oxford; Malden, MA: Blackwell Publishing. First published 1987.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Dr Kevin Mendousse is a lecturer in French at the University of Auckland (NZ) and holds a PhD in linguistics from the Université de Paris-Sorbonne (FR), where he taught English phonetics and phonology as well as grammar and translation. He is a member of that university's linguistics research laboratory Formes-Discours-Cognition (FDC), which has accreditation from the French Ministry of Education and Research. His current research interests include phonetics, (morpho)phonology, and, more generally, psycholinguistics.