The lyrics for Dave Frishberg's 1962 jazz song ‘Peel me a grape’ abound in novel instances of (di)transitive argument structure constructions such as French me a fry, polar bear rug me, mink me or New Thunderbird me. We expect such creative language use from poetry and can easily make sense of novel uses of the constructional schema in the conceptual domain of pampering. We do this against the backdrop of more familiar instances such as crush me some ice and chill me some wine (from the same song), occasionally combined with some general cultural knowledge, for instance the fact that Ford brought out a new (third) generation of luxury convertible cars called Thunderbird in the early 1960s. In everyday language use, however, we might be somewhat less ready to accept creativy of this kind. In particular, native speakers of English (but not German!) will find the ditransitive use of explain in the title of Adele Goldberg's most recent book unacceptable, even if they were primed by similar constructions such as tell me the difference. Since the publication of her seminal work on argument structure constructions (ASCs) in 1995, Adele Goldberg has been one of the most prominent proponents of construction grammar. Her new monograph explores the limitations and boundaries of creativity in language use by attempting to answer one of the core questions of linguistics, i.e. ‘How is it that native speakers know to avoid certain expressions while nonetheless using language in creative ways?’ (p. 3). The shorthand for this question throughout the book is the ‘explain-me-this puzzle’.
Readers familiar with Goldberg's work might wonder whether the volume provides a new version of construction grammar, a students’ introduction to the psychology of language, as the preface (p. ix) indicates or, as the layout of the cover and snappy title suggest, a popular science book on human linguistic creativity. The introduction in chapter 1 gives a partial answer to these questions in the new definition of what a construction is. While Goldberg's original definition of ‘construction’ was one that made use of conventions of formal logic and emphasized idiomaticity (Reference Goldberg1995: 4), the second version moved away from formalism and included frequency as an important (additional) defining criterion (Reference Goldberg2006: 5). We are now presented with a definition in the first chapter that is firmly grounded in psycholinguistics: constructions (p. 7) are ‘emergent clusters of lossy memory traces that are aligned within our high- (hyper!) dimensional conceptual space on the basis of shared form, function, and contextual dimensions’. The different dimensions of form and function – core components of earlier Goldbergian definitions of ‘construction’ – are brought into the equation later on: ‘lossy representations of language experience cluster together on various dimensions, including those related to form (phonology, grammatical categories, order, morphology) and function (meaning, information structure, register, genre, dialect)’ (p. 94).
The new definition Goldberg offers in chapter 1 takes a while to unpack even for experts in the field, so despite outward appearance, this is definitely not a popular science book on human linguistic creativity and its boundaries. The text on the jacket offers a rephrasing for ‘emergent clusters of lossy memory traces’ which helps with the unpacking: it describes the process of incrementally developing the construct-i-con (i.e. the readers’ knowledge of language) as ‘recording partially abstracted tokens of language’. The introduction also offers a mnemonic for the cognitive mechanisms that Goldberg takes to be fundamental to our understanding of how language works, namely the cence me (read ‘sense me’) principles (pp. 5–7). The usefulness of the mnemonic for some readers might be reduced somewhat by the different order in which the component parts are spelled out (and summarized in table 1.3): ee stand for the requirement on language to be both expressive and efficient; m is for the lossy memory (i.e. the partial abstraction); c for the alignment of partially abstracted expressions into clusters (i.e. constructions); n is shorthand for the new information that speakers keep integrating into their existing knowledge about language and addresses the fact that constructions are emergent; the second c represents competition in production through multiple activation of existing elements in the speaker's construct-i-con; and the final e is for ‘error-driven learning’ where ‘errors’ are not production errors but rather the absence of expected patterns in the input, which in turn shapes expectations for future encounters with a speaker's language. Chapter 1 also provides a roadmap for the book; in the following, I will briefly summarize the main content of chapters 2–8.
Chapter 2 introduces some basic concepts of the psychology of language using the acquisition of word meaning as an example, on the assumption that the same principles apply in the acquisition of grammatical patterns. Knowledge of word meaning emerges from repeated encounters of words in context, strengthening relevant connections among individual words and word meanings over time while dispreferring others. Chapter 3, ‘Constructions as invitations to form categories’, illustrates how the same mechanisms are at work in building up knowledge about grammatical patterns, with a focus on ASCs. Unlike Goldberg's (Reference Goldberg1995, Reference Goldberg2006) previous book-length treatments of constructions, which only mention variation in passing (e.g. as a factor that has an influence on acceptability judgements), chapter 3 in the volume under review discusses variability more systematically. Goldberg assumes variable word order to be captured in separate word order constructions that combine with ASCs (p. 39), brings phonology and etymology to bear on lexical variation in the double object construction (pp. 40–1), and touches on information structure as being subject to constructional variation (p. 42). Social (register), regional and cross-linguistic variation in ASCs are briefly illustrated in separate subsections in chapter 3, too, but like other aspects of variation, they are explicitly ruled out as possible explanations for the limits of linguistic creativity. How speakers learn to use language creatively from the input that they receive and how this creativity is at the same time constrained is the topic of chapters 4 and 5, which thus form the core argument of the book.
In chapter 4, ‘Creativity: coverage is key’, Goldberg relates how type frequency (rather than text frequency) is an important factor for creatively using an ASC. In a nutshell, the larger the number of different verbs is that a speaker has previously encountered in a particular ASC, the more likely she is to either use a novel verb in the ASC herself or to accept the ASC with a novel verb. Highly frequent constructions with low variability, on the other hand, are less likely to lead to novel uses. Coverage thus acts as a backdrop to productivity. More specifically, ‘[c]overage relates the number of different verbs witnessed in a given construction (a construction's type frequency), the semantic and phonological variability of witnessed types, and the similarity of the coinage to attested types’ (p. 63). In other words, formal surface similarities and speakers’ general pattern matching abilities help in fostering a link between a novel instance of an ASC and a previously encountered exemplar. The chapter includes a subsection that illustrates how coverage can be statistically modelled (i.e. the clustering algorithms from Bayesian models, pp. 70–2). Frishberg's song creates a conceptual domain (pamper x with y), which easily accommodates cashmere/mink/champagne as novel (denominal) transitive verbs but also structurally more complex items such as polar bear rug and New Thunderbird. The song text does this by locally creating what Goldberg refers to as high coverage, i.e. clustering of partially abstract exemplars (p. 62). These verbs are all possible in the ASC because they are novel as verbs. Why their novelty status as members of the word class ‘verb’ is key for their apparently unlimited productivity finds its explanation in the following chapter.
The focus of chapter 5, ‘Competition: statistical preemption’, is on the factors that constrain productivity. Put somewhat simplistically, this boils down to existing constructions in the language that express the same meaning and have a similar distribution, or, as Goldberg (p. 75) phrases it, the fact that ‘native speakers have learned a more conventional way to express the intended message-in-context’. The chapter provides ample support from psycholinguistic experiments that show how existing competing constructions inhibit the extension of a constructional schema to novel contexts. Two subsections of the chapter describe ways of statistically modelling preemption, once in terms of probability (section 5.6) and once with a focus on confidence (section 5.7), with ‘confidence’ taking both frequency and the existence of a structural alternative into account. Importantly, the chapter also aims to explain why native speakers of English can alternate between the double object construction (He told me a story) and the prepositional object construction (He told a story to all the children visiting that afternoon) with a verb like tell but why explain is a communication verb that is barred from this constructional alternation. Specifically, it is the notion of ‘error-driven learning’ that Goldberg introduced as part of the cence me mnemonic and that she sees as playing the key role here: ‘mismatches between the input from the speaker and what had been anticipated by the listener provide an error signal that is used to improve future predictions through a process of error-driven learning’ (p. 91). With respect to constraints on constructional creativity, this means that if speakers repeatedly fail to witness a particular verb in an alternate ASC that it could potentially be used in, it will make them increasingly less likely to actively extend the verb to that construction or to judge it as acceptable in an experiment. Novel members of the category ‘verb’ (such as champagne or mink) will not be constrained by statistical preemption: they are not attested in competing ASCs and are immune to error-driven learning because speakers will not have been able to form predictions on their use by virtue of being novel verbs. They can therefore be judged as ‘unusual’ but not as ‘unacceptable’.
Chapter 6, ‘Age and accessibility effects’, looks at differences between language development in (young) children and adults with respect to their differing abilities to use ASCs in the way that adult native speakers of a language would. With respect to young children, the chapter highlights the complexities involved in making the ‘right’ kind of abstractions as children are initially less likely to see similarities between individual instances of an ASC than adult L2 learners, which leads to more conservative use of ASCs; they are also less likely to see distinctions between ASCs, which results in overextensions of ASCs in production. The difficulties that adult learners encounter largely stem from having a conceptual space that is already defined or ‘warped’ (p. 111), so they are likely to transfer patterns from their first into the target language. For those whose input is limited (e.g. to classroom situations), the possibility to make use of the second type of statistical preemption, i.e. error-driven learning, is much reduced (p. 116).
Chapter 7 reviews alternative/competing approaches to the explain-me-this puzzle, including studies that take a more formal framework as their starting point as well as related (but slightly different) approaches from within usage-based construction grammar. Chapter 8 provides an overview of the main points made in chapters 2–7 (pp. 140–4) and an outlook on some of the directions in which future work on the psychology of language should be taken: widening the scope beyond ASCs and, ideally, beyond the sentence level, drawing further on interdisciplinary work (including anthropology), fostering the link between experimental approaches and modelling, and extending existing frameworks beyond the rather limited number of languages that have been studied (pp. 144–6).
Hypercritical readers might point out that the volume under review mostly elaborates on ideas that are present in Goldberg's earlier work. The notion of partial productivity is already discussed quite extensively in chapter 5 in Goldberg (Reference Goldberg1995), including, in essence, the discussion of coverage and competition (just not by these terms), but with little corpus-based or experimental evidence being available at the time to back up her claims. Similarly, the notion of ‘lossy memory’ or partial abstraction is already present in Goldberg (Reference Goldberg2006) in the notion of ‘selective encoding’ as ‘a partial abstraction over what was encountered’ (p. 46). Almost a quarter of a century after the publication of her first study on ASCs, extensive research in corpus, psycho- and neurolinguistics is now available, and Goldberg draws extensively on these studies, a substantial number having been co-authored by herself. It is this rich body of data-based and experimental research that the book brings to bear on the central concepts that experts in the field will find illuminating. Also, whereas learners and the development of the construct-i-con were discussed in her previous books, the juxtaposition of very young L1 learners and adult L2 learners adds a more nuanced perspective to the issue of partial productivity. And while construction grammar has been taken up in SLA research (see e.g. Ellis Reference Ellis, Hoffmann and Trousdale2013), the specific angle provided in chapter 6 of the present volume provides food for thought to the field, including discussions of transfer effects in World Englishes research.
Finally, does the book ‘deliver’ as ‘an accessible introduction to students, teachers, and researchers’ (p. ix) on the psychology of language? Each of the main chapters ends with a chapter summary. Those of chapters 4, 5 and 6, moreover, explicitly make connections with previous chapters, which definitely serves the purpose of providing an accessible introduction well. Throughout the book, Goldberg supplies very useful figures to visually illustrate the abstract cognitive processes that form the subject matter of the book. Short summary statements such as ‘productivity begets productivity’ (p. 73) contribute to the accessibility of the book in a similar fashion. I am less convinced that the cence me mnemonic is useful in this context (having made frequent cross-reference to table 1.3 that summarizes the underlying mechanisms when it was mentioned in later chapters). If the book were intended as an introduction to undergraduate students, a glossary with key terms and definitions would have been a useful additional feature. But it is more likely to be aimed at more advanced students of linguistics.
