Reviewed by: Morphological length and prosodically defective morphemes by Eva Zimmermann Anthi Revithiadou Morphological length and prosodically defective morphemes. By Eva Zimmermann. Oxford: Oxford University Press, 2017. Pp. xii, 350. ISBN 9780198747321. $99 (Hb). Eva Zimmermann offers an elegant and meticulous monograph on morphological length-manipulation (MLM) phenomena, that is, length alternations bound to morphological contexts. The work is based on the general claim that all nonconcatenative processes—additive and subtractive alike—arise from an ‘enriched notion of affix’ (1) that involves prosodic nodes as primitives. The main argumentation in the book aims at establishing that all morphology is additive, a claim that is empirically supported by an impressively broad typological survey that includes representative data sets for attested morphological length-manipulating patterns culled from sixty-two languages. Furthermore, the author argues that MLM operations are best analyzed in a theoretical framework named prosodically defective morphemes (PDM), which, in addition to morphemic prosodic nodes, includes a rather rich representational apparatus, a constraint-based computational component that regulates the integration of such prosodic nodes into larger structures, and a handful of conditions that impose restrictions on gen. PDM not only succeeds in predicting all attested patterns and systematically excluding imaginable yet unattested ones, but also circumvents several problems encountered by alternative accounts. More importantly, it makes some insightful predictions about, for instance, the possible interaction and coexistence of additive and subtractive operations within the same language. Z sets an ambitious goal: to show that ‘all kinds of subtractive MLM can be derived if one takes into account the full range of prosodic nodes and their possible (defective) integration into a base’ (34). This way she challenges the shared assumption that subtractive phenomena have a processual character and delivers a model that is consistent with an item-and-arrangement (Hockett 1954) morphology framework and the basic tenets of its more sophisticated version, namely distributed morphology (Halle & Marantz 1993). The core idea, defended throughout the book, is that nonconcatenative morphology (i.e. addition and subtraction) is epiphenomenal [End Page 475] and results from the affixation of morphemic prosodic nodes. In order to achieve this goal, Z develops an extended version of generalized nonlinear affixation (Bermúdez-Otero 2012), in which prosodic nodes (μ, σ, Φ) are treated as independent phonological objects related to each other via association lines and not by projection, as standardly assumed in the prosodic hierarchy (Nespor & Vogel 1986). A representative example of additive suffixation is presented in 1a. In Wolof, the reversive suffix /-i/ is lexically specified to trigger gemination (pp. 5–6). Under the PDM framework, gemination is caused by a floating morphemic mora that escorts /-i/ and, more specifically, by its integration into the structure through association to the final root consonant. Subtraction, attested with the causative suffix /-al/ (1b) in the same language (p. 8), is taken to be the surface result of the integration of an underlyingly mora-less vowel. That is, the vowel of the suffix /-al/, which lacks its own mora, takes over the mora of the preceding moraic consonant, yielding the surface effect of degemination. (1) Based on Prince and Smolensky’s (1993) notion of containment (‘every element of the input must be contained in the output’) and following previous work on the topic (e.g. van Oostendorp 2006, Revithiadou 2007), Z develops a comprehensive theory that consists of three key components: colored phonological objects, association relations, and the constraints that regulate their mapping. First, she argues that all phonological objects that are part of the underlying representation of a morpheme share the same information about morphological affiliation: that is, they have the same color (indicated as a subscribed symbol, e.g. V○, C●, μ□, etc.). A phonological object’s morphological color cannot be deleted or changed, although it can be enriched (as explained below in the discussion of color reflection and fusion). Second, remaining faithful to an extended version of containment that requires the whole input (e.g. segments, features, prosodic nodes, and their association relations) to be reconstructable from the output at any time, Z identifies four different types of association relations between phonological objects. Morphological association lines (2a) are underlyingly present and can never be erased; they may...