The book's title, Lexical Relatedness, evokes issues associated with paradigmatic sense relations, such as homonymy, synonymy, antonymy, and so forth, i.e. issues more appropriately ascribed to the realm of lexical semantic studies.Footnote ¹ Although questions such as the elusive partition between polysemy and homonymy are not left untouched by the author, Andrew Spencer goes indeed in another, far-reaching direction, and develops a powerful model of morphology in the tradition of so-called inferential-realizational approaches, as Stump (Reference Stump2001: 3) labels Word-and-Paradigm theories of inflection. Far from being a mere implementation of Stump's (Reference Stump2001) Paradigm Function Morphology, Spencer's book offers a profound and comprehensive view of various morphological phenomena encompassing inflectional and derivational phenomena.

The volume consists of eleven chapters organized in three main parts. In Part I, ‘Lexemes, lexical entries, and lexical relatedness’, the author lays the groundwork of his model, and illustrates his basic assumptions on the nature of the lexicon and on what he calls the within-lexeme vs. between-lexeme types of relatedness (i.e. inflection vs. derivation), together with the empirical grounds for the original research to be developed in Part III.

After a comprehensive overview of the book in the introductory chapter, Chapter 2 establishes the essential sketch of a lexical representation. Specifically, the author describes lexical entries as an ensemble of three independent attributes (or ‘fields’), i.e. form, syntax and semantics, together with a lexemicindex, understood as a key field that unambiguously identifies a lexeme. As far as the syntax field is concerned, a crucial function is assigned to semantic function roles in the argument structure of lexical entries, namely E for verbs; R for nouns; A* for adjectives. According to the author, who explicitly builds on (his own) previous research on this issue (see, in particular, Spencer Reference Spencer, Booij and van Marle1999), such roles might allow one to dispense with traditional lexical categories altogether, once it is accepted ‘that the syntactic phrase structure representations be couched in terms of the argument structure representations of their lexical terminals’ (37). Another interesting issue raised in Chapter 2 concerns the representation of what Spencer defines as ‘degenerate lexical entries’ (43), i.e. entries which lack either morphological forms or syntactic/semantic attribute values.Footnote ²

Chapter 3 provides the core empirical evidence on which the model developed in this book is based. The author presents a puzzling dataset of intermediate types of lexical relatedness, starting with transpositions – action nominalizations, participles and relational adjectives – which are all well-represented in European languages. Examples of relational adjectives and participles are drawn from Hindi-Urdu and Chukchi. The key feature of transpositions consists in a change in the morpho-syntax that is not accompanied by a concomitant semantic change. In this respect, transpositions depart from the more conventional patterns of derivational morphology which typically trigger a semantic change in the base lexeme. Further, some morphosyntactic properties of the base are preserved in transpositions – adjectival participles and event nominals, for instance, keep some of the verb features.

Another thought-provoking set of data is provided by inflectional phenomena. Spencer explores the contextual vs. inherent inflection dichotomy proposed by Booij (Reference Booij, Booij and van Marle1996) and points to its empirical inadequacy, especially in a cross-linguistic perspective. Some of the data which more seriously challenge the existence of a dividing line between inflection and derivation are found in the realm of meaningful inflection, as with semantic cases and, in particular, aspect markers in Slavic. Argument structure operations, meaningless derivation, evaluative morphology, and paradigmatically mixed categories provide other crucial evidence of intermediate types of relatedness and, drawing on data from typologically and genealogically unrelated languages, the author discloses a composite picture of relations among lexemes which traditional approaches to morphology are unable to capture satisfactorily and comprehensively.

The second part of the volume, ‘Paradigmatic organization and the lexicon’, provides an overview of Stump's (Reference Stump2001) Paradigm Function Morphology, which promotes the view that morphological relations, rather than being expressed as rules combining morphemes, should be understood as mathematical functions and governed by principles of default inheritance logic. Here the reader is introduced to some of the technical machinery deployed for capturing the various modes of relatedness probed in Part III of the book. In particular, Chapter 4 illustrates key notions in the inferential-realizational approach of Stump (Reference Stump2001), namely paradigm function, rule of referral, and morphomicstem (the last notion is due to Aronoff Reference Aronoff1994). Other important concepts are the ones of content paradigm and paradigm linkage, which are specifically postulated by Stump to address non-trivial aspects of the morphology–syntax interface. Chapter 5 illustrates the author's novel extension of Stump's (Reference Stump2001) model. Specifically, Spencer introduces the Generalized Paradigm Function (GPF henceforth), intended as a set of functions independently operating over the four attributes in the lexical entry, and other key principles (see, in particular, the Derived Lexical Entry Principle) for the modeling of derivational morphology.

Having provided the empirical grounds and formal machinery, it is in Part III, ‘The factorized lexicon’, that the author addresses the most challenging intermediate types of relatedness, described in Chapter 3 with data from several languages, and here taken as the launch pad for the deployment and gradual refinement of the GPF machinery. In particular, Chapter 6 discusses transpositions, argument nominalizations (e.g. driver, employee), paradigmatically mixed categories (e.g. Russian stolovaja ‘canteen’, a noun inflecting like an adjective), evaluative morphology and meaningless derivation (e.g. under-stand). Chapter 7 grapples with argument structure alternations and incorporates some of the basic ideas and formal modeling of Lexical Functional Grammar. Chapter 8 offers a perspicuous analysis of a much debated class of transpositional phenomena, i.e. deverbal and deadjectival nominalizations. Further instances of transpositions are tackled in Chapter 9. As a general move for capturing the nature of mixed categories, the author attributes a key part to semantic function roles, and allows the lexical representation to superimpose one semantic function role over another; hence, for instance, deverbal nominals retain the original E role in their argument structure, 〈R〈E…〉〉. Chapter 10 provides an assessment of various phenomena of lexical relatedness in Selkup, an Uralic language of the Samoyedic group. This language offers a remarkable empirical base for the GPF model, since it is exceptionally rich in terms of transposition phenomena, which would represent a serious challenge for traditional theories of word formation. Chapter 11 contains some concluding observations.

The contents of the book, though reasonably based on much of the author's previous research, are novel for the most part, with the exception of parts of Chapters 8 and 10, which already appeared elsewhere.

As declared in the very first pages of the book, the author intends to go to the heart of a long-debated, yet still puzzling issue in morphology, i.e. the elusive partition between inflection and derivation. In this attempt, Lexical Relatedness targets one of the most serious shortcomings of inferential-realizational approaches to morphology – the conceptual necessity of a radically distinct approach to inflection on the one hand and derivation on the other – and develops a sophisticated formal apparatus for capturing, with the same technical machinery and conceptual tools, several types of relations among the (inflected) forms of a lexical entry and distinct (i.e. derived) lexical entries.Footnote ³

The existence of intermediate types of relatedness reveals an intricate picture whereby meaningless inflection (i.e. pure agreement phenomena) and standard derivation are indeed exceptional cases, which should be better understood as the extreme points on a scale of intermediate values of lexical relatedness. The author does not deny that the grammar actually draws a line between the complex set of within-lexeme and between-lexeme types of relatedness. This line, however, can be traced neither on the basis of (in)variance of morpho-phonological nor syntactic/semantic values in the lexical entry. For this purpose, the author deploys the lexemic index as the attribute value responsible for discriminating between the two phenomena: derivation only is accompanied by a change in the index, which is tantamount to saying that, trivially, derivation only is responsible for the creation of a new lexeme. Inflection, in contrast, generates forms of a lexeme.

One of the most valuable aspects of this book is the depth and carefulness with which the author tackles an amazing variety of data, both in quantitative and qualitative terms. Even when touching on certain topics in passing, the author deals with the data in a way that reveals a comprehensive and profound knowledge of the phenomena under scrutiny. To give a concrete example, with respect to the puzzling debate on the status of aspectual oppositions in Russian, the author discerningly concludes that aspect is a type of inflection that cannot be properly defined as either inherent or contextual. As a matter of fact, the aspectual choice being highly context-sensitive – as Spencer argues – it is not the marking per se but rather the use of this marking by the speaker that could be judged as more or less contextual or inherent (90).

Given that it is not possible to do anything more than scratch the surface of such a profound and extensive contribution in a short review, I will single out a couple of critical issues, both at the empirical and at the conceptual levels.

A general concern arises from the definition of paradigmatic derivation, i.e. a notion which is vital for delimiting the range of data to which this model applies. The reader is warned, indeed, that the GPF model applies to a subset of the phenomena traditionally understood as derivational, since only productive, active and regular relations between lexemes can form derivational paradigms and, accordingly, can be modeled in this inferential-realizational framework of lexical relations. As an example, the author confronts in Chapter 6 causative verb formation (through -ize and -ify) and noun-to-verb conversion, on the one hand, with person and subject noun formation and (negative) un- prefixation in English, on the other. Let us consider the latter first, and subject noun formation in particular. In the spirit of the Separation Hypothesis (see Beard Reference Beard1995), in which the GPF model is rooted, the author makes an excellent point in separating the concrete forms assumed by these phenomena (-er in driver, -ant/-ent in claimant, -ist in publicist, etc.), which are highly irregular and allomorphic, from the characterization in abstract terms of person noun formation, which is instead a highly regular and predictable phenomenon. The regularity of subject noun formation, abstracting away from its concrete manifestations, makes it available to a GPF analysis, since, in Spencer's words: ‘that process is paradigm-driven in the sense that a base of the right sort will always give rise to a derivative’ (217). However, with respect to causative verb formation and noun-to-verb conversion, it is concluded that these are not good candidates for paradigmatic derivation relations, for reasons tied to their lack of systematic productivity and semantic transparency. In Spencer's view, these phenomena should be better understood as cases of word creation – together with blends, occasionalisms, and ludic word creation in general – and there is no further attempt at assessing the overall nature of these phenomena. The author's arguments are questionable here, in my view, since they are tied to the notoriously problematic notion of morphological productivity, which is not further investigated/formalized throughout the book, and to the semantic opacity associated with these lexemes, a feature imputable – to a certain extent, at least – to most derivational phenomena, even those that do seem to hold a maximally ‘paradigmatic’ nature. Moreover, the reader might be confused by the fact that, in the introductory chapter, causative verb formation through -ize is in fact presented as an instance of derivational paradigm (11). Along the same lines, after stating that conversion is not paradigmatic in English and does not represent the kind of derivational morphology captured by GPFs, at the end of Chapter 6 (272), Spencer does propose a GPF defined over a feature Conv for English noun-to-verb conversions, characterized by a deficient semantic specification.Footnote ⁴

Another issue arises in the analysis of the semantic attribute of (a class of) mixed categories. The GPF model is brilliantly designed to capture intermediate types of lexical relatedness. Thanks to the fact that lexical representations are factorized into four distinct attributes and that functions range independently over them, mixed categories such as action nominals make an excellent case in point, being instances of relations where the form and syntax attributes are changed, with semantics and lexemic index basically left unaltered by the function. In the GPF model, this state of affairs results in the following facts: (i) transpositions of this sort are non-derivational, which is captured in the system by the invariant lexemic-index value; and (ii) while there might be important changes in the syntactic behavior of the transposed lexeme, the overall semantics of the base remains unaltered.Footnote ⁵ However, with respect to action nominals, the author recognizes the systematic addition of some meaning nuances in the transposed entry. For instance, regarding Italian Infinito Sostantivato, the meaning values the author grapples with are the fact and proposition readings (see Zucchi Reference Zucchi1993 as a main reference). The author convincingly argues that such added semantic nuances should not be treated as lexical properties, but should be imputed to the syntactic construction in which the nominal occurs. On the other hand, other types of lexical polysemy, abundantly debated in the literature, such as the event/result polysemy of derived action nominals, are not tackled in the present analysis.Footnote ⁶ I believe, however, that these ‘shifts’, far from being totally idiosyncratic, could have been an intriguing testing ground for the GPF model, and, arguably, their analysis could have resulted in a more explicit assessment of the polysemy of transposed/derived entries and of its formal modeling within this approach to morphology. This minor criticism might be generalized to the way in which the semantic attribute of lexical entries is characterized in the GPF model. As the author himself admits, the semantics deployed is too coarse-grained, and fails to incorporate fine-grained semantic generalizations put forward in much research on derivational word formation, such as Barker (Reference Barker1998) on -ee nominals, and the substantial framework developed by Lieber (Reference Lieber2004). All in all, this is a very understandable shortcoming for a comprehensive analysis and formal apparatus which do not focus on subtle semantic relations, but seek to capture several, understudied types of relation among items, especially those which have been more problematic for non-inferential approaches to morphology.

Overall, Lexical Relatedness is a mine of insightful analyses and well-structured reflections on the nature of several challenging phenomena, which are built on data from ‘near’ (Slavic, Germanic, Romance) and ‘distant’ languages (Uralic and Australian ones), and which are construed within a novel framework of morphology designed to overcome the elusive partition between inflectional and derivational phenomena. As such, Spencer's book represents an invaluable resource for researchers and students interested in many general questions on morphology and the lexicon, and, more generally, in key topics in theoretical linguistics.