Until recently, research in prosodic morphology (PM) has focused on the role that prosody plays in determining the size and shape of morphemes and words. Laura J. Downing's Canonical forms in prosodic morphology is thus a welcome addition to an emerging area of research that explores the role of morphology in determining prosodic morpheme shape. Her work is positioned within Optimality Theory (OT) and builds on recent morphologically based approaches to reduplication, known as Generalized Template Theory (GTT, McCarthy & Prince Reference McCarthy, Prince, René Kager and Wim1999). GTT eschews reduplicative templates in favour of an approach in which morphological category determines the shape and size of reduplicative morphemes. In reviewing some of this literature, Downing observes that OT lacks a complete theory of the correlation between prosody and morphological shape (32). She proposes an original, morphologically based theory (MBT) of prosodic morpheme shape that builds on the central claims of GTT and extends the domain of investigation from reduplication to truncations, nicknames, root-and-pattern systems and word-minimality effects.

A leading idea of Downing's MBT is that ‘minimality constraints defining canonical forms fall out from a correlation between morphological complexity and phonological complexity’ (34). There are three key components to the proposal: the morphological categories in MBT, head-dependent asymmetries in phonology and morphology, and constraints on the prosody-morphology homology. The morphological categories are stem, root, and affix, where stem is defined as a root+affix complex. Crucially, roots are heads while affixes are non-heads; stems are morphologically complex and composed of roots, and hence subject to some of the same constraints as roots. The key innovation of MBT is that morphological and prosodic complexity – not the prosodic hierarchy – are the determinants of morpheme shape.

The distinction between the three morphological categories stem, root and affix is a central component of MBT. In particular, the idea that some forms are headed (stems and roots), while others are not (affixes), is crucial. Downing adopts and adapts Dresher & van der Hulst's (1998) proposals about asymmetric phonological representations that (i) heads must be complex, and (ii) complexity is satisfied by branching structure. These two requirements are formalized as the constraint in (1).

1. (1) Heads Branch (122)

   Lexical heads (Roots) must prosodically branch.

Prosodic branching is satisfied by any structure in which a mother node has two equal daughter nodes. So, prosodic representations that are bisyllabic, bimoraic or have a branching rhyme would all satisfy Heads Branch. Because stems and roots are headed, they need to satisfy Heads Branch, and would need to have the more complex branching prosodic structure. This fundamental difference between prosodic morphemes that are headed (stems and roots) and those that are not (affixes) provides an elegant morphological account for the sizes and shapes of prosodic morphemes attested in typological studies. This differs from standard models of PM in which bisyllabicity and/or bimoraicity is accounted for in terms of being foot-sized. In MBT, there is no necessary connection between being bisyllabic/bimoraic and being foot-sized. The sizes of prosodic morphemes are the result of the interplay of morphological and prosodic complexity: being headed and needing branching structure. Because affixes are not heads, they do not necessarily branch and thus are smaller.

A number of constraints are proposed to account for the range of PM patterns that Downing examines. Some of the constraints are adapted from standard work in metrical phonology (e.g. Binarity: ‘each daughter of a constituent must be adjacent to some edge of the constituent’, 125) and are informed by Inkelas's (Reference Inkelas1989) prosodic and metrical hierarchy (e.g. ProsodicStem: ‘the morpho-prosodic constituent’ that is bisyllabic, 124). As an interesting result, the bisyllabicity of ProsodicStem follows directly from being morphologically complex, requiring branching structure, and the constraint in (2), according to which each morpheme corresponds to a single syllable.

1. (2) Morpheme-SyllableCrrelation (MorphSyll) (120)

   Each morpheme contains exactly one syllable.

By adopting MorphSyll and HeadsBranch, Downing is able to account for differences between monosyllabic roots and bisyllabic stems. As noted above, MorphSyll gives an explanation for the well-attested pattern of stem bisyllabicity: because stems are composed of two morphemes, bisyllabicity would satisfy MorphSyll. The constraints in (1) and (2) also serve as a guide in proposing other PM constraints, like ProsodicStem and Trunc=PStem. By way of illustrating how MBT accounts for differences in shape between stem, root and affix, I will summarize an analysis of Japanese compound loan-word truncations.

While Japanese loan-words are minimally bisyllabic when truncated, in compound truncations, each half of a compound need only be bimoraic; so each part can be monosyllabic or bisyllabic (Itô Reference Itô1990), as illustrated in (3).

1. (3) Japanese truncation compounds

   

Downing proposes two constraints to account for truncations in general: in one, the truncated form is a prosodic stem (4a), and in the other, it is a prosodic root (4b).

1. (4) Competing prosodic parses of truncated stems (174)

   1. (a) Trunc(ation)=PrStem:A truncated Stem is parsed as a Prosodic Stem.

   2. (b) Trunc(ation)=PrRoot:A truncated Stem is parsed as a Prosodic Root.

For Japanese, Trunc=PrRoot, HeadsBranch and Binarity dominate MorphSyll. Trunc=PrRoot ensures that each half of the compound is a prosodic root with branching prosodic structure. Having MorphSyll as a lower ranked and violable constraint allows each part of the prosodic root to exceed a syllable, while having Binarity ranked higher ensures that prosodic roots do not exceed two syllables or moras in size. This constraint ranking ensures that candidates like *[hebii-metaru] which exceed bisyllabic size are ruled out, while the actual output [hebi-meta] is selected as optimal. Notice that the output violates MorphSyll twice, because each part of the compound is bisyllabic.

Downing's coverage of PM phenomena is vast, as she both analyzes well-known patterns and case studies and introduces and discusses new data from over 150 languages. MBT is informed by standard observations about reduplication and root-and-pattern morphology and newer findings on various topics, like word-minimality patterns. The book contains five chapters, which I will summarize briefly below.

Chapter 1, ‘Introduction’, sets the stage and provides the context for the current study, defining PM and the significance that it has had in the development of phonological theory. One domain of inquiry of McCarthy & Prince's (1986) seminal work that does not seem to be included in this investigation is infixing of any type. However, Downing clearly specifies the domain of investigation clear from the outset to be those types of morphological constructions ‘which are required to have a constant minimal and/or maximal shape’ (4). Still, one wonders whether infixes that are located proximally to a foot (as many are) may be affixed to a prosodic root of some form in MBT.

In chapter 2, ‘Prosodic hierarchy-based templates’, there is a thorough overview of the types of phenomona that the prosodic hierarchy-based version of GTT (PBT) has accounted for. Downing's overview is empirically extensive, covering a range of PM phenomena from a diverse set of languages, including standard data like Arabic root-and-pattern morphology, Makassarese and Axininca Campa reduplication, and truncation patterns from a number of languages. One recurring point made throughout the book is that PBT is not able to explain a number of phenomena because of the coupling of the prosodic hierarchy in (5) with a constraint requiring stems to be prosodic words (Stem=PWd).

1. (5) Prosodic hierarchy

   PWd (prosodic word)

   |

   Ft (metrical foot)

   |

   σ (syllable)

   |

   μ (mora)

Downing observes that coupling the prosodic hierarchy above with Stem=PWd accounts for patterns in which reduplicants coincide with stress feet, as in Diyari tjílpa-tjílparku ‘red-bird species’ (25f.). However, she goes on to note that many languages have disyllabic reduplicants that do not have metrical stress feet, as well as other instances where the prosodic morpheme does not correspond with the stress foot (as in Arabic, which has trochaic stress but iambic templates). Another shortcoming of PBT concerns the observation that the most common word-minimality effects impose a C(onsonant)V(owel) shape. This lack of correlation between stress feet and prosodic morpheme shape is claimed to be a central failing of PBT. However, many of these arguments are difficult to make in OT, in which constraints are regularly violated. Specifically, OT is readily able to account for situations in which one language would require prosodic words to correspond to stress feet and another language where that requirement is regularly violated. In the following, I will outline an analysis that would render it possible to account for some of the potential problems raised in Downing's work.

In translating key concepts from the prosodic hierarchy to OT, Selkirk (Reference Selkirk1995) provides evidence that the constraint on Exhaustivity is violable. Exhaustivity requires each layer of the prosodic hierarchy (cf. (5)) to be present. However, the constraint is crucially violated in cases where a syllable is dominated by a PWd, with no intervening foot. So, in situations in which the minimal word is CV or when a function word is attached directly to a PWd, without an intervening foot, Exhaustivity is violated. Because these two conditions would arise if Exhaustivity is low-ranking in a language, PBT would seem to predict that there are CV minimal words in languages which also have clitics, rather than predicting that bisyllabic prosodic morphemes are impossible in languages without stress systems.

The model of MBT is exemplified in chapter 3, ‘Morpheme-based templates’. As noted above, MorphSyll figures prominently in ensuring monosyllabic affixes and roots vs. disyllabic stems. However, while Downing states that there is considerable evidence for MorphSyll (120), she only cites evidence from word-minimality effects. It is not clear how robust the generalization is regarding the monosyllabic size of morphemes cross-linguistically. In terms of the question of how morphological theory informs MBT, Downing notes that many morphologists assume that the affix is the head of a word – an assumption based on morphosyntactic evidence, such as the affix being category-changing. However, the position taken in the book is that in phonology, roots are heads because they are ‘typically the locus for distinctive phonological information’ (121). This reference to phonology in identifying the root as head seems to weaken Downing's claim that her proposal is purely morphological.

In chapter 4, ‘The role of phonology in defining canonical form in MBT’, Downing shows that apparent problems for MBT, such as CV being the most common minimal word size and the existence of disyllabic roots, ‘are straightforwardly accounted for by taking a closer look at the phonotactics of the particular language’ (211). Constraints against coda consonants or long vowels conspire to produce canonical forms that veer from those predicted by MBT, using morphoprosodic interface constraints alone.

In chapter 5, ‘Questions for future research and conclusion’, Downing points out the weaknesses of MBT with a frank discussion, often accompanied by a possible solution. For example, there are many languages that impose a four-syllable maximality condition, and Downing raises the question of what constrains maximality. She provides a few proposals, including one by Orie (Reference Orie1997), which shows parallels with her own revised version of Binarity (239–241).

In conclusion, this book represents an excellent example of three aspects of theoretical innovation. First, it builds on several ideas from current morphophonological theory to create a novel approach. Second, it extends ideas from reduplication theory to a wider domain. And finally, it illustrates how to argue for a new idea, by presenting the types of predictions that can be made along with competing theories. For this reader, the book has readily achieved its stated goal of wanting to engage and inspire readers to do more work in PM that examines and adopts proposals from morphology.