"In this book, Richard Kern explores how technology matters to language and the ways in which we use it. Kern reveals how material, social and individual resources interact in the design of textual meaning, and how that interaction plays out across contexts of communication, different situations of technological mediation, and different moments in time."
SUMMARY This book sets out to accomplish the daunting task of providing a comprehensive, up-to-date view of sound change research, incorporating insights from theoretical linguistics, psychology, modeling, and other fields. Part 1 contextualizes phonologization in relation to larger questions of linguistics, psycholinguistics, and cognitive science. Part 2 examines the relationship between phonology and phonetics -- acoustic, perceptual, and articulatory. Part 3 considers the relation between processes of phonologization and the architecture of phonology, morphology, and the lexicon. Part 4 explores psychological and social factors governing sound change within language communities.
Part I: What is Phonologization
1. Enlarging the Scope of Phonologization, Larry Hyman
Hyman begins the volume by providing important historical context to and reviewing multiple perspectives on the scope of phonologization, and argues for expanding its definition to include a range of phenomena spanning change from intrinsic phonetic variation to extrinsic phonological rule. Through case studies involving tonogenesis and ATR harmony, he argues that contrast is not an essential component of phonological change, and that allophony and other processes play important roles. In his view, phonologization can be considered the formation of phonology from any source, and thus is a part of the larger concept of ‘grammaticalization’. In fact, he proposes an admittedly impractical but conceptually useful improvement in terminology to refer to various processes of language change, such as ‘dephonetic phonogrammaticalization’, meaning the conversion of phonetic information into phonological grammar.
2. Certainty and Expectation in Phonologization and Language, Elizabeth Hume and Frederic Mailhot
Like the opening chapter, Hume and Mailhot broaden and contextualize the definition of phonologization, this time in terms of information theory, rather than linguistic theory. This allows consideration of the degree to which grammar-external factors play a role in predicting the targets and results of change. The authors define a model based on the entropy, or “average surprisal”, of a system, and demonstrate that elements at both extremes of surprisal are targeted for change. The linguistic effects of change can also be described in terms of entropy: changes from high to low surprisal are structure preserving, while those from low to high need not be.
Part II: Phonetic Considerations
3. Phonetic Bias in Sound Change, Andrew Garrett and Keith Johnson
Garrett and Johnson consider the phonetic basis of sound change, focusing on questions of typology (why some changes are common, and why some are not -- a “typology of causes”), and actuation (what triggers phonologization of a phonetic element at a particular time and place). The model of phonologization presented has three stages: (i) structured (i.e., biased) phonetic variation in the signal, (ii) linguistically constrained selection of said variants, and (iii) innovation by individuals who propagate changes.
They go on to provide evidence for each component, demonstrating that biases in the production and perception of speech are non-random and directional, due to such factors as motor planning, aerodynamic constraints, gestural mechanics, and perceptual parsing. These individual factors interact with systemic biases (second-order biases) arising from language-specific and universal constraints. Finally, the authors describe a simulation model of actuation based on variation in sociolinguistic awareness within a population of speakers.
4. From Long to Short and from Short to Long: Perceptual Motivations for Changes in Vocalic Length, Heike Lehnert-LeHouillier
Lehnert-LeHouillier attempts to account for the fact that, crosslinguistically, some sound changes are bidirectional and some are unidirectional. Examining the interaction of vowel duration with f0 and vowel height, Lehnert-LeHouiller argues that unidirectional changes are driven by intrinsic phonetic factors (those which affect all listeners in all languages), while bidirectional changes are driven by extrinsic factors (those which affect different languages differently). In this model, tightly associated (intrinsic) cues, like spectral cues, are less likely to be separated due to sound change, which is why vowel length distinctions arise from height distinctions, but not vice versa; cues like f0 are extrinsically associated with vowel height, and are more likely to become dissociated, allowing the development of length contrasts from tonal contrasts, and the reverse. This was supported by the results of a perceptual experiment involving speakers from a number of languages with a vowel length contrast, but varying degrees of association between vowel height, f0, and duration.
5. Inhibitory Mechanisms in Speech Planning Maintain and Maximize Contrast, Sam Tilsen
Tilsen begins by noting that there are many cases where phonetic precursors to phonological change are not phonologized, and that loss of contrast is a possible but not inevitable outcome of sound change. Thus, it is important to consider forces hindering phonologization, along with those favoring it. Dispersion theories capture this notion at a language/typological level, but not at the level of psychological mechanisms at play in individual speakers. According to Tilsen, these are competing or coincident motor plans which interact to push apart or maintain a contrast. This is supported by experimental production studies demonstrating that when motor plans for nearby categories are planned together, acoustic dissimilation results compared to the same categories planned separately.
6. Developmental Perspectives on Phonological Typology and Sound Change, Chandan Narayan
Narayan considers the roles of infants as listeners and caregivers as speakers in shaping typological patterns of sound change, specifically examining the way infant perceptual biases influence typology, and the characteristics of infant directed speech which create the acoustic conditions for sound change. Narayan hypothesizes that acoustic salience and biases unique to infants mediate the relationship between perception and the inventory, leading to the prediction that contrasts which require experience or learning are rarer crosslinguistically. A series of case studies shows correlations between contrast acoustics, infant perception, and typology. Not only do infants operate in a biased way on speech input, but Narayan demonstrates that infant directed speech differs from adult directed speech by analyzing the relationship between pitch cues, voice onset time, and voicing contrasts in English infant directed speech at different developmental stages.
Part III: Phonological and Morphological Considerations
7. Lexical Sensitivity to Phonetic and Phonological Pressures, Abby Kaplan
Kaplan attempts to determine whether the lexical frequency of phonological patterns is driven by phonetic pressures, phonological markedness, or their interaction. To do so, she compares intrinsic phonetic tendencies which have been phonologized in some languages to those which have not been phonologized in any language, finding that lexical patterns are affected by crosslinguistically phonologized phonetic patterns (even if not phonologized in a particular language), but not by phonetic patterns whose phonologization is unattested. From this, Kaplan concludes that phonetic patterns do not affect the lexicon directly, and that abstract phonological markedness intervenes. Another formulation of this conclusion seems to be that only those phonetic patterns that may be phonologized can affect the lexicon. Kaplan goes on to consider alternative formulations of this model, and its extension to phonological alternations.
8. Phonologization and the Typology of Feature Behaviour, Jeff Mielke
Mielke begins by juxtaposing two major theories about the nature of distinctive features; either they are an innate part of Universal Grammar, or they are emergent from phonetic patterns. Mielke analyzes a database of phonetic patterns (P-base), counting and categorizing them according to the type of pattern and the sets of phonological features which best describe them. Patterns are classified as spreading, dissimilation, partitioning (selecting targets or environments for alternations), or other processes.
Some distinctive features are more involved with some categories of patterns than others, which Mielke links to differences in their phonetic correlates, noting that those features involved in partitioning are those which have proven to be the hardest to define cross-linguistically (e.g., laterals). Mielke argues that differences between languages in the behavior of these features should be accounted for by variation in the formation of features from their phonetic correlates, rather than changes or additions to a universal feature set.
9. Rapid Learning of Morphologically Conditioned Phonetics: Vowel Nasalization Across a Boundary, Rebecca Morley
Languages vary not only in their phonological patterns, but in the nature and degree of phonetic implementations (e.g., degree of vowel nasalization by a nasal consonant), so these language-specific phonetics must be learned. These can interact with morphology, and possibly other levels of prosodic structure (i.e., derived environment effects). Morley’s goal is to establish a phonetic origin for such domain-restricted processes, incorporating them into an evolutionary phonology framework. An artificial language learning experiment involving vowel nasalization is presented to determine whether listeners can both attend to subphonemic variants and link them to morphological structure. The results of a perceptual task indicate that listeners are sensitive to the pattern. Morley argues that this sensitivity constitutes a first step in phonologization, and speculates that perhaps all sound change begins at a boundary; if true, derived environment effects could be of central, rather than marginal importance.
Part IV: Social and Computational Dynamics
10. Individual Variation in Socio-cognitive Processing and Sound Change, Alan C.L. Yu
Alan Yu considers psychosocial factors involved in the actuation and diffusion of sound change. Sociolinguistic theories of language change include linguistic ‘innovators’, and ‘propagators’ who spread changes through a linguistic community. Yu attempts to determine the psychological and social characteristics of these innovators, arguing that they arise in part from individual differences in cognitive processing style -- so-called ‘autistic traits’, even among typical individuals. These traits are shown to correlate with perceptual compensation for coarticulation in speech, as well as social characteristics like empathy. Yu suggests that a population of ‘minimal compensators,’ defined by an imbalance between cognitive characteristics, tolerate and use a wider range of variation within linguistic categories and have a wide social network, making them good candidates to be leaders in propagating sound change. Yu speculates about the role of individuals at the other end of the spectrum from these minimal compensators, but the important insight of this chapter is that not everyone operates on linguistic input in the same way, and that variation in language may be accounted for as a by-product of our cognitive, biological, and social makeup.
11. The Role of Probabilistic Enhancement in Phonologization, James Kirby
Kirby considers two central questions: first, why are only selected cues to a contrast targeted for phonologization and not other potential targets; and second, what causes transphonologization (dephonologization of one cue following phonologization of another)? Reminiscent of Hume and Mailhot’s approach (Chapter 2), Kirby’s model posits that the selection of cues for (de)phonologization is based on adaptive enhancement of the contrast -- if contrast precision is lost, cues change to compensate, and those cues which change are those which are most informative. Kirby notes that listeners in the real world are not ideal, and that all cues are not of equal importance. Phonological categories are modeled as weighted mixtures of cues in an agent-based system of listeners and talkers with four stages: production, (talker) enhancement, (listener) bias, and categorization. This model is applied to the case of phonologization of f0 from voicing contrasts in Seoul Korean. Duration, harmonics, and burst amplitude are also cues to the voicing contrast, but the model results in category changes matches those observed in Seoul Korean (phonologization of f0, not other cues), without targeting any specific cues for enhancement. This model clarifies the role of enhancement and bias in change, and suggests that the selection of cues for (de)phonologization can be predicted.
12. Modelling the Emergence of Vowel Harmony Through Iterated Learning, Frederic Mailhot
Mailhot analyzes the diachronic emergence of vowel harmony (focusing on lexical harmony, rather than alternations, as he considers this important for understanding diachronic change) as resulting from a combination of synchronic coarticulation and biased transmission. He presents a simple iterated learning model, consisting of one “adult” and one “child” (learner) transmitting words consisting of four vowels, defined by height and backness features. The acoustic (f1, f2) values of these articulatory features are calculated from category prototypes combined with coarticulation and noise. The learner reverses this process, inferring representations of lexical items based on acoustic prototypes. The properties of the learner’s lexicon vary, depending on levels of noise, coarticulation, and hypocorrection, but there appear to be stable states of the lexicon corresponding to no harmony, complete harmony, and partial harmony across the lexicon. Thus, language-specific phonetic differences, such as vowel-to-vowel coarticulation could be related to lexical tendencies, and possibly phonological alternations.
13. Variation and Change in English Noun/Verb Pair Stress: Data, Dynamical Systems Models, and their Interaction, Morgan Sonderegger and Partha Niyogi
Sonderegger and Niyogi take on the actuation problem: not all phonetic variation results in phonological change, so why does some? The authors examine stress shifts in English noun-verb pairs from 1700-2007, drawing data from many dictionaries and incorporating differences in pronunciation entries as evidence of variation. Several observations emerge from the behavior of pairs over time: some stress patterns are “stable states” -- short term variation around these stable states often occurs, and long-term shift from one stable state to another sometimes occurs, but rarely for both members of the noun-verb pair at the same time. The authors model noun-verb stress shifts in the lexicon using factors based on how speakers deal with such variation in the language, namely mistransmission, discarding (ignoring some variants), and their combination. Mistransmission alone does not produce shifts matching the observed data, but discarding and the combined model lead to bifurcations in noun-verb pairs similar to those observed in the historical data. This model illustrates the importance of combining individual and population level models in accounting for historical change.
EVALUATION This volume does an excellent job at living up to its goal to provide a comprehensive and state-of-the-science view on phonologization. Each chapter raises interesting questions and invites further inquiry, and I have very few substantive criticisms of the work as a whole.
The entire book would be an excellent introduction to phonologization for advanced graduate students. There is a balance between chapters providing a wider view of theories of sound change with those examining particular case studies in detail (the latter nonetheless drawing general principles from said examples). Important theories and models are discussed throughout, which unites these chapters around common themes without being overly repetitive.
Each chapter also provides adequate context and background to stand alone, and individual chapters will likely be of interest to students and researchers in disciplines outside of linguistics, and linguists and psychologists with other specializations who are interested in sound change should find them accessible. For instance, the several chapters incorporating computational models do a good job of describing in sufficient detail for the nonspecialist the computational principles involved while maintaining a focus on the relevant human/linguistic behavior; at the same time, they demonstrate important applications of information theory and various modeling techniques in linguistics. Computer scientists and other specialists will likely find most chapters short on technical detail, but they successfully demonstrate the application of these techniques to the unique characteristics of language and provide an introduction and references to the linguistic theory involved.
As noted by Hyman in Chapter 1, phonologization touches upon and perhaps provides useful evidence for other fundamental questions about language, such as the nature and origin of grammar, and a subset of the chapters examine related issues (e.g., 7, 8). Later chapters (e.g., 10, 13) link phonetic and phonological aspects of language, which might be considered fairly “micro-level” variation to those in other fields, to more general cognitive, biological, and social dimensions of communication, thus linking diverse points of the spectrum of human behavior together, and demonstrating the value of multifaceted approach to understanding sound change taken in this volume, which hopefully lead to the fruitful “cross-pollination” alluded to in the introduction by editor Alan Yu.
ABOUT THE REVIEWER:
Evan D. Bradley is Assistant Professor of Psychology at Pennsylvania State University, Brandywine. He earned a PhD in Linguistics & Cognitive Science at the University of Delaware, and a BA in Cognitive Science and Certificate in Music at Northwestern University. His research interests include acoustic phonetics, auditory perception, and phonological learning, as well as music perception and cognition. His current work examines the perception and learning of musical melody and lexical tone languages.