This book presents analyses of phonological weight in four grammatical systems and proposes novel theories to account for them. ‘Weight’ refers to how phonological systems refer to syllable types for various phenomena. For example, stress will often be attracted to CVː and CVC syllables, so – in such a system – these syllable types are ‘heavy’ while CV syllables are ‘light’. The four systems discussed in the book range from sub-syllabic to phrasal levels, including stress, prosodic minimality, meter, and end-weight. The terms ‘categories’ and ‘continua’ in the subtitle refer to categorical and gradient conceptions of weight, respectively. So, important goals in this book are to demonstrate that weight has such a nature, and to provide cross-linguistic evidence for the various types. The evidence is drawn from descriptions, typological surveys, experimental results, and corpus studies. The theoretical framework employed is Optimality Theory (OT; Prince & Smolensky Reference Prince and Smolensky2004). Maximum Entropy Harmonic Grammar (e.g. Hayes & Wilson Reference Hayes and Wilson2008) is adopted when dealing with gradient weight effects.

This book consists of six chapters. Below I summarize the theory and evidence of each chapter. I then discuss some theoretical and methodological issues raised in this book.

Chapter 2 proposes a theory of vowel prominence to account for complex weight scales for stress. The core constraint is MAIN→VV, which is violated whenever stress falls on anything other than a long vowel or vocalic diphthong. This proposal differs from theories which define weight solely by moraic content (e.g. Hayes Reference Hayes1995). Two independent arguments are raised to support the prominence approach. First, secondary stress sometimes reveals VC to be bimoraic even in words in which VV attracts primary stress. Second, VG (where G is the first half of a geminate consonant) almost always patterns with VC, not VV, in ternary weight systems. For example, Chickasaw exhibits VV > VC > V, with primary stress revealing VV > {VC, V} and secondary stress revealing {VV, VC} > V (Gordon Reference Gordon2004). Moreover, VG syllables are treated as lighter than VV for primary stress, but equivalent to CV and VV for secondary stress. Crucially, VG and VC cannot be treated as monomoraic due to the presence of secondary stress. The prominence approach maintains that VC, VG, and VV are uniformly bimoraic in Chickasaw. VV is more stress-attracting because it contains a long vowel. The book further argues against a previous approach to such complex scales, as advanced by Rosenthal & van der Hulst (Reference Rosenthall and van der Hulst1999) and Morén (Reference Morén1999, Reference Morén2000). The idea is that VV > VC > V can be reduced to μμ > μ by allowing the weight of VC to vary in different environments: it is parsed as monomoraic when unstressed but bimoraic when stressed. However, the book argues that this coercion approach fails to explain cases like Chickasaw because VC can receive secondary stress and VG patterns with VC for secondary stress. The prominence approach, on the other hand, can account for 17 languages known to exhibit such complex scales. An alternative approach, coda prominence, is also rejected as one of the main problems is that it generates systems in which VC > VV, arguably a pathology. Gradient weight for stress is briefly discussed at the end of this chapter, with the gist that the prominence approach can be extended to such systems.

Chapter 3 discusses prosodic minimality: minimum size requirements that languages impose on prosodic words (PWd). Three cases are discussed in this chapter: Latin, Māhārāṣṭrī Prakrit, and Tamil. All three languages have a bimoraic minimum for isolated PWds. However, each has its own response to violations of PWd minimality that arise from resyllabification in the configuration #C₀V́C#V(…). In Latin, VC content words remain stressed after resyllabification, resulting in a subminimal degenerate foot. These degenerate PWds are frequent in prose, but Virgil avoids them almost categorically in verse. The book proposes that such words are neither heavy nor light, but intermediate in weight, with Virgil avoiding them in regulated verse. In Māhārāṣṭrī Prakrit, resyllabification of monosyllables is suppressed due to prosodic minimality, whereas polysyllables are more likely to undergo resyllabification. In contrast, Tamil permits resyllabification but repairs VC monosyllables via gemination. Nevertheless, rhotics are nongeminable in Tamil, as they are argued to be non-moraic (e.g. *[kaɾ] and *[kaɻ]). So, Tamil lengthens the vowel if the syllable is closed by a rhotic. The nonmoraicity of rhotics is implemented by the constraint *μ/R (A rhotic must not head a mora). With *μ/R dominating relevant constraints (e.g. WbyP, ONSET, DEP-μ), the pre-rhotic vowel is lengthened to satisfy prosodic minimality. In short, this chapter illustrates the interaction of minimality and different responses to resyllabification.

Chapter 4 is about syllable weight in poetry, focusing on Finnish, Greek and Tamil. Quantitative meters relate syllable weight or mora count to metrical strength, and most quantitative meters around the world have been described as having a binary weight distinction. However, this chapter makes two main observations and proposals. One is about variable weight, where certain syllable types can be either scanned as heavy or light. Such cases can be analyzed with binary weight using optional processes such as resyllabification, cluster compression, correption in hiatus, and variable final vowel length. The other observation is that many quantitative meters exhibit sensitivity to gradient weight on some variant of the scale V ≤ VC ≤ VV ≤ VVC. For example, in many meters, VV occurs more frequently in strong positions than VC, even though both are heavy. Another way in which meters show gradient weight is found in ‘final indifference’, which refers to a putative universal in quantitative meters where weight sensitivity is suspended in final position. As observed in Homer and Virgil, final position tends to be heavy. This tendency is gradient in that syllable types that are gradiently heavier occur more frequently in the ultima. Another observation about quantitative meters is that they mostly obey the Latin criterion (light iff C₀V) instead of the Khalkha criterion (light iff C₀VC₀), whereas quantitative stress systems are more evenly split between the two. However, the case studies in this chapter show that the Khalkha criterion exists as gradient tendencies in systems that have the Latin criterion categorically. Lastly, the chapter also discusses Interval Theory, according to which weight is assessed not for syllables, but for each interval from the left edge of one vowel to another.

Chapter 5 argues that prosodic-end weight reflects phrasal stress. Prosodic-end weight refers to the tendency to localize prosodically heavy constituents later in phrases. The core proposal is that heavier elements align with phrasal stress. In English and other languages, phrasal prominence is usually right-oriented. As a result, prosodic end-weight is a manifestation of the more general weight-stress interface in phonology. Eight syllabic and sub-syllabic factors are argued to associate with greater end-weight: more stressed syllables, more syllables, longer vowels, lower vowels, longer onsets, less sonorant onsets, more sonorant codas, and longer codas. Previous studies on these factors are extensively reviewed. Evidence from corpora and experiments is provided to fill gaps in the literature. The formal analysis utilizes a set of mapping constraints, schematized as X → φ_s. For example, VV → φ_s requires every long vowel or diphthong to be at the strong (head) position of a PWd (or higher prosodic level). This explains why treat occupies the second position of the phrase trick or treat, other things being equal. The theory advocated here is claimed to be superior than previous accounts. One of the theoretical predictions is that prosodic-beginning weight should exist. This book reports a preliminary study on Turkish revealing potential evidence.

Regarding the theory, one prediction for MAIN → VV is that languages do not distinguish between long vowels and vocalic diphthongs in stress. However, Māori is reported to have this distinction: stress falls on the leftmost long vowel even when vocalic diphthongs are available (de Lacy Reference Lacy and Paul1997 and references cited therein). However, the current formalism of the constraint is unable to account for cases like Māori.

The theory that prosodic-end weight reflects phrasal stress in English is appealing. One of the predictions is that such effect should be muted for non-stress languages. Also, prosodic prominence can be signaled in different ways. For example, Mandarin is argued to be a right-prominent language, which means that the identity of the rightmost tone is maintained while tones are permitted to change in other positions (Chen Reference Chen2000). If prosodic-end weight is also found in Mandarin, perhaps we could have a more unified account of weight in stress and non-stress languages.

The book mentions the importance of evidence for stress, noting recent work that has challenged the correctness of impressionistic descriptions (including my own – Shih Reference Shih2016 et seq.), and aims to ‘heed the cautions … concerning descriptions’ (38). However, the book still accepts many stress descriptions uncritically, to its detriment. For example, Jha’s (Reference Jha1940–44, Reference Jha1958) descriptions of Maithili stress assignment are assumed to be correct, particularly in regard to secondary stress. Yet, there is disagreement in the Maithili literature about the existence of secondary stress. Yadav (Reference Yadav1996: 47–49) does not mark secondary stress for words with final primary stress, e.g. [mirˈcai] ‘hot pepper’, only mentioning secondary stress in compound nouns of four syllables, e.g. [mirˈcaiˌbari] ‘pepper garden’. Moreover, Mishra (Reference Mishra2006) explicitly comments on previous descriptions, rejecting secondary stress:

Unfortunately, the book does not discuss such descriptive disagreements. Thus, a central issue is whether the stress descriptions cited as crucial evidence for the theory are in fact accurate.

To be clear, the book provides evidence for stress and related phenomena that are commensurate with the past and current literature on prosodic phonology. However, the book also lays bare the issue of whether common practice is adequate. Several studies on multiple languages have shown that acoustic realizations do not match previous impressionistic descriptions (e.g. Goedemans & van Zanten Reference Goedemans, van Zanten, van Heuven and van Zanten2007, Tabain, Fletcher & Butcher Reference Tabain, Fletcher and Butcher2014). Similarly, while the book proposes constraints that ‘relate sonority directly to … stress’ (20), agreeing with de Lacy (Reference Lacy and Paul2002, Reference Lacy and Paul2004) and others, it has been demonstrated that many apparent sonority-driven stress cases have highly conflicting descriptions, and those descriptions do not match experimental results (Shih Reference Shih2016, Reference Shih2018a, Reference Shihb; Shih & de Lacy Reference Shih and de Lacy2019).

Overall, this book effectively demonstrates the categorical and gradient effects of weight. This book not only concentrates on weight in speech, but also in poetry. A wide range of methodologies are employed to probe the phonological behavior of weight, including typological surveys, wug-tests, corpus studies, and statistical modeling. There are extensive references and literature reviews throughout the book.

Finally, I recommend this book to advanced graduate students and professors. Each chapter mostly stands on its own, and so could be assigned as individual readings. Readers with a solid background in Optimality Theory will find the analyses easy to follow. Even for scholars who do not use OT in their work, the theoretical insights are easily translated into other theories.