With The Semantics of Compounding, Pius ten Hacken, Professor for Translation Studies at the University of Innsbrück, presents a collection of papers with the aim of showing how Jackendoff’s conceptual semantics approach, Lieber’s lexical semantics approach, and Štekauer’s onomasiological approach can advance our understanding of the semantics of compounding.

After Ten Hacken’s introduction, Part I lets the original authors present their frameworks in one chapter each. Part II presents four papers on noun–noun compounds. Pierre J. L. Arnaud presents a categorization of the modification relations in French noun–noun compounds. Zoe Gavriilidou analyzes the semantics of noun–noun combinations in Greek. Ingmarie Mellenius and Maria Rosenberg give an overview of the semantic categories found in novel Swedish compounds in their diary data. Finally, Jesús Fernández-Domínguez considers non-lexicalized English noun–noun compounds. Part III turns to other compound types, with Carola Trips analyzing English phrasal compounds, Barbara Schlücker analyzing German adjective–noun compounds, and Renáta Panocová analyzing English and Russian neoclassical compounds. A conclusion by the editor completes the volume.

In their review of the Oxford Handbook of Compounds (Lieber & Štekauer, 2009), Meir and Aronoff (Reference Meir and Aronoff2011) lament the lack of exchange between different theories and between people working on different languages. The set-up of this volume looks like it was intended to remedy this situation: three specific frameworks are set up for comparison, so that the individual contributions can situate their data against these theories.

What are these three frameworks, then? Jackendoff proposes a system that is able to generate possible meanings of compounds by drawing on generally available components on the one hand and compound specific procedures on the other. The three general components are profiling, choosing an action modality, and cocomposition. Compound-specific are the head principle, which identifies the second noun as the head of English noun–noun compounds, and the argument and modification schemata. While the argument schema is needed for compounds in which the first element is an argument of the second, the modification schema is used for all other compounds. This schema introduces a function linking the two constituents, and Jackendoff provides a list of thirteen basic functions that can be used here. Unsurprisingly, these thirteen functions are reminiscent of basic predicates proposed in other accounts, e.g., Levi’s (Reference Levi1978) Recoverable Deletable Predicates. The generative potential of his system is demonstrated by showing how these different components taken together can be used to establish complex relations between the constituents.

While Jackendoff does not see a principled distinction between semantic and pragmatic knowledge, Lieber’s lexical semantic framework, developed from Lieber (Reference Lieber2004) onwards, keeps these separated. Her framework decomposes word meanings into two basic parts: the semantic skeleton containing syntactically relevant material, and the semantic body containing all meaning aspects that are encyclopaedic in nature. Just as in Jackendoff’s system, there is a clear difference in how compounds in which the first part is an argument of the head are treated as opposed to other types of compounds. In the former case, the correct interpretation is achieved through co-indexation in the skeletons, while in all other cases material from the semantic body is needed. Besides showing how her framework deals with subordinate, coordinate, and attributive compounds, she also explains how it can account for exocentric compounds.

Šteckauer’s approach rests on the assumption that each new complex word results from a specific act of naming. Whereas the two previous approaches are more interested in how the two parts of a compound can account for its meaning, Štekauer looks at the process from the opposite perspective: the concept to be named is already known, and now a name needs to be found. He distinguishes between the onomasiological base, which gives the class of the object to be named, and the onomasiological mark, which specifies the object and consists of the determining and determined constituent. Depending on the correspondence between the onomasiological base and the two parts of the onomasiological marks with morphemes in the compound, he distinguishes eight different onomasiological types.

What do we learn about the differences between the three approaches in the other contributions? The aim of the first contribution by Arnaud on French compounds is purely descriptive, only a few example analyses in Jackendoff’s frameworks are given. In contrast, the contribution by Gavriilidou explicitly aims to apply Lieber’s framework to Greek noun–noun combinations. However, since no independent motivation is given for which aspect of a combination’s meaning is situated in the skeleton and which in the body it is hard to judge the merit of these analyses. No comparison with either Jackendoff’s or Štekauer’s system is provided. Somewhat surprisingly for a book on the semantics of compounds, actual compounds, that is purely morphological compounds (right-headed, phonological words), are explicitly excluded. In Mellenius and Rosenberg’s contribution, Jackendoff’s framework is used as one of several sources of relations used for the classification of the children’s novel compounds, but apart from that, the frameworks play no role. Fernández-Domínguez applies Jackendoff’s and Šteckauer’s framework to English non-lexicalized compounds. The analyses themselves and the comparison of the two approaches remain somewhat unsatisfying, with the author concluding that “the results of this comparison confirm the validity of both systems for the description of compounds semantics, even if with different orientations” (p. 149).

The first two contributions in the third part explicitly aim to present an analysis in Jackendoff’s parallel architecture. However, Carola Trips also considers how her examples would be analyzed in Lieber’s system. While both frameworks seem to work for her data, she argues that cases without a verbal predicate (e.g., a trial-and-error technique) need to be distinguished from cases with verbal predicates (e.g., a make-haste-slowly situation). Whereas the former function similarly to non-phrasal compounds and require an underspecified relation R to be specified, the latter always require a metonymic shift of the phrasal part and the relation between the two constituents is always the IS-A relation. Schlücker’s investigation of German adjective–noun compounds stays firmly within Jackendoff’s analysis. She argues that one of the basic relations he introduces, classification, should actually be seen as the default meaning of nominal compounds, as this meaning is independent of any specific semantic relation between the compound constituents. This view is supported by adjective–noun compounds, which mostly do not need semantic relations in their analysis. The final contribution by Panocová operates within Šteckauer’s approach. However, there is no reason to believe that her main findings are incompatible with the other two frameworks. She argues that there is no principled difference between neoclassical and other compound types. By comparing neoclassical compounds in English and Russian, she argues that English has a dedicated subsystem for neoclassical word formation, whereas in Russian the neoclassical compounds are borrowings.

Ten Hacken’s conclusion considers the treatment of different types of compounds in all three approaches and tries to draw out their similarities and differences. Drawing mostly on Dutch data, he notices that the analyses in Jackendoff’s framework are more specific than those in Lieber’s approach. This is not surprising, as Jackendoff’s presentation format applies to both semantic and pragmatic aspects of meaning alike. In contrast, the more formalized part of Lieber’s approach, the skeleton, is more specific in its inclusion of only those semantic aspects that are relevant to syntax. A similar discrepancy in detail exists between Jackendoff’s and Šteckauer’s presentations, again explainable by the different foci of the approaches. Are the three approaches compatible? As Ten Hacken points out, Lieber’s categorical distinction between skeleton and body is clearly incompatible with Jackendoff’s conceptual semantics approach. The compatibility of Šteckauer’s approach with the other approaches, in contrast, depends on the exact integration of the naming act into these approaches. These final points are illuminating, because they show what would be needed for explicit comparisons of the different approaches. In order to compare Šteckauer to the other two, explicit assumptions about the status and place of naming, or at least about a difference between the perspectives of speaker and hearer in the two frameworks, would be needed. Since neither framework makes this explicit, there is little to compare. And for an evaluation of Lieber’s approach against Jackendoff’s approach, one would need to provide diagnostics for determining whether there is a principled difference between meaning aspects that are semantic in nature and those that are rooted in the wider cognitive system. Since none of the contributions works on this issue, the question of incompatibility cannot arise.

The contributions in Part II and III vary in the empirical bases that are used. Arnaud’s research on French is based on his own collection of 1049 compounds. Hapax legomena were excluded (p. 79). Gavriilidou speaks of corpus data (p. 101), but does not say which corpus was used, or whether all her examples come from the corpus. Melenius and Rosenberg again use their own private corpus, this time based on diary data of three Swedish children. A selection criterion was that the compounds “are considered to be productively formed by the children” (p. 114), but it does not become entirely clear what this means in practice (it seems that forms corresponding to established Swedish compounds where excluded). The same authors seem to use the same data for a different analysis, Rosenberg and Mellenius (Reference Rosenberg and Mellenius2018). This analysis does not reference the current paper at all, but we learn that the children are the children of the authors. Fernández-Domínguez examines a selection of English compounds from the BNC Sampler, that is, a two-million-word corpus. Trips, again on English, uses the full BNC, and bases her analysis on her full dataset. Schlücker, on German, does not say anything about the source of her data. Panocová uses data from dictionaries. The whole volume would have benefited immensely if all authors had been explicit about their data sources. Even better would have been to make the non-public sources available alongside the publication of this volume in electronic form. This is technically trivial, and would allow other researchers to test the data for other patterns of interest, ranging from form-based aspects to the representativeness of the dataset as a whole.

While Ten Hacken is correct in pointing out that labels of compound classes are not so important if understood as pre-theoretical descriptions (p. 212), the usage of labels in this volume is at times confusing. Thus, Arnaud’s relational subordinative compounds correspond to Lieber’s attributive compounds and to the primary nominal compounds in Fernández-Domínguez’s contribution and Ten Hacken’s conclusion. This terminological confusion is not helped by the usage of the term ‘attributive compound’ in Scalise and Bisetto (2009, p. 51) for only those compounds in which the “head is modified by a non-head expressing ‘property’ of the head”, e.g., mushroom cloud ‘cloud in the shape of a mushroom’ as opposed to mushroom soup, which is a subordinate compound of the subcategory ‘ground’ in their classification.

The decision to have the original authors present their frameworks in the first part of the books leads to few new insights. In fact, both Jackendoff and Lieber have already contributed chapters on how compounding is treated in their frameworks to the Handbook mentioned above (Jackendoff, Reference Jackendoff, Lieber and Štekauer2009; Lieber, Reference Lieber, Lieber and Štekauer2009). The go-to version for Jackendoff’s approach to compounding remains his expanded and revised version of that work, Jackendoff (Reference Jackendoff2010), from which the chapter here is mostly excerpted. However, Lieber’s contribution is a very clear distillation of the treatment of compounding in her approach. Šteckauer’s contribution benefits from reading Grzega (Reference Grzega, Lieber and Štekauer2009) beforehand, and presupposes knowledge of the principles governing meaning predictability introduced in Šteckauer (2009), again in the aforementioned Handbook.

In conclusion, the volume presents something of a mixed bag: the overall idea is good, but the execution does not entirely convince. It does show that the Oxford Handbook of Compounding is essential reading for anyone interested in compounds and compound semantics, whereas the contributions in this book, The Semantics of Compounding, are of interest more to the researcher working on the respective object languages than to the linguist wanting to get a first overview of the semantics of compounding.