Recent research on grammatical variation phenomena has shown that, in most cases, there is more variation than expected, and, far from being random, a language user's choice of otherwise (near-)semantically equivalent variants is actually governed by a wide array of factors (see, e.g., Fanego et al. Reference Fanego, López-Couso and Pérez Guerra2002; Rohdenburg & Mondorf Reference Rohdenburg and Mondorf2003; Dufter et al. Reference Dufter, Fleischer and Seiler2009). The empirical investigation of these determinants bears the explanatory potential to account for mechanisms underlying language processing, that is, to shed light on processes at work in language production and to assess the degree to which processing efficiency affects the choices language users make in encoding the preverbal message. A case in point focused on in the present study by Arne Lohmann concerns coordinate constructions. These involve the juxtaposition of two linguistic items, which may also occur in alternate orderings, as the binomial in (1) illustrates:

1. (1)

   1. (a) he's like wood and metal (BNC)

   2. (b) . . . metal and wood     (adapted from p. 1)

Earlier research on binomials has predominantly considered irreversible cases such as law and order and odds and ends, which rather have the status of quasi-idioms, and has mainly concentrated on hypothesising factors responsible for the exclusive preference of these frozen orders (see, e.g., Malkiel Reference Malkiel1959). In contrast, Lohmann's corpus-based study is primarily concerned with potentially reversible constructions, which also display preferences but which still allow for considerable variability. What is more, the present book goes beyond the analysis of binomials in that it also takes into account two further representatives of coordinate constructions, namely copulative compounds, as in (2), and the coordination of complex noun phrases, as in (3), whose potential for reversibility has been neglected in previous literature:

1. (2)

   1. (a) As a singer-dancer, however, she hasn't advanced much . . . (COCA)

   2. (b) As a dancer-singer, . . .

2. (3)

   1. (a) . . . the gypsies identify with her nomadic existence and her life as an oppressed servant. (ICE-GB)

   2. (b) . . . with her life as an oppressed servant and her nomadic existence     (examples adapted from p. 1)

From a psycholinguistic perspective, the issue of reversibility is of great interest because the order of reversible binomials, copulative compounds and conjoined complex noun phrases grants insight into the linearisation of linguistic elements and therefore has important implications for the human processing system in general. In the following, I will briefly summarise the main aspects of the book's twelve chapters, before I conclude with a general evaluation of the present study that distinguishes itself by its innovative combination of variationist, psycholinguistic and functional approaches.

The introduction in chapter 1 sets the stage perfectly for the study to follow by motivating the book's primary interest in coordinate constructions, narrowing down the three constructions of relevance and embedding the study into the context of empirically oriented grammatical variation research and processing-related accounts of language production.

Chapter 2 presents two lines of research concerned with coordinate constructions and points out their respective limitations. On the one hand, genuinely linguistic approaches have mainly concentrated on the formulaic character of irreversible binomials, which results from various ordering constraints, while reversible cases have only been considered marginally at best. However, as Lohmann aptly remarks, what these studies largely ignore is that the findings on irreversible binomials bear the potential to extrapolate the constraints underlying reversible cases of ad hoc coordination. On the other hand, coordinate constructions have been of interest to psycholinguistic research, as they allow for observations on an alleged effect of lexical/phonological accessibility in the linearisation process during language production. However, findings from psycholinguistic studies lack conclusive evidence and are therefore in need of empirical support, which the present book seeks to provide with a corpus-based study. Lohmann justifies his choice of methodology by briefly sketching previous approaches to coordinate constructions (section 2.3), that is, introspection and experimental studies such as naturalness judgement tasks, which, again, are mainly restricted to irreversible cases of coordination. In this respect, then, a corpus-based analysis promises a more sensitive assessment of actual language use and its underlying ordering constraints, with the proclaimed aim of ‘bridg[ing] the gap between linguistic and psycholinguistic approaches, thereby hopefully alleviating the empirical uncertainty regarding the influences speakers are subject to when ordering elements’ (p. 30).

Chapter 3 provides a highly comprehensive overview of the variables that previous literature has considered as influencing language users’ ordering choices, eighteen of which are operationalised and empirically investigated in Lohmann's book. The first group of factors comprises pragmatic and semantic variables:

• • information status, i.e. elements representing given information are positioned before those providing new information in the discourse

• • iconic sequencing, i.e. temporal, causal and logical orders of elements natural to the extra-linguistic world are iconically replicated in the order of linguistic elements (e.g. birth and death)

• • extra-linguistic hierarchies, i.e. the first referent tends to be the one of greater significance in the sociocultural sphere (e.g. university and college), or of a higher numerical value (e.g. dollars and cents)

• • conceptual accessibility, i.e. constituents that are cognitively more accessible precede less accessible ones, as in the case of animate before inanimate referents.

A second group of variables involves stress-related factors, with the ordering of constituents being influenced by:

• • a preference for rhythmically alternating sequences of stressed and unstressed syllables (on the principle of Rhythmic Alternation see Schlüter Reference Schlüter2005),

• • the avoidance of ultimate stress on the second constituent, or

• • the accentuation of the second constituent by means of a heavier main syllable.

Following up on the short-before-long principle, the third group of variables concerns the parameters of length and complexity, with the second position of the coordinate construction typically being filled by constituents that are longer (counted in number of syllables or phonemes), morphologically more complex or (as in the case of conjoined phrases) syntactically more complex. Furthermore, phonological and phonetic length is taken to be a decisive factor in the fourth group of variables, which postulates that typically, the second constituent:

• • is characterised by a longer vowel,

• • ends in a voiced consonant, or

• • contains a less obstruent final segment.

Other phonological and phonetic variables form the fifth group of factors, which predict that the second constituent is characterised by a lower vowel or a more back vowel. As for the number of initial consonants or the sonority of the initial element, previous research has found conflicting results. Last but not least, frequency has been taken to be an important factor, with more frequent constituents preceding less frequent ones.

While all of the factors mentioned above have been hypothesised to be at work in language users’ ordering choices in coordinate constructions, previous research has treated these variables in isolation and thus often jumped to monocausal conclusions. Also, conclusive answers as for the synergetic and antagonistic interaction of these determinants are lacking, and it has remained unclear so far to what extent some of the variables discussed in the literature are merely epiphenomenal. In order to remedy these deficits, the present study adopts a multifactorial approach, which is outlined in chapter 4, together with details concerning the corpus databases for the three coordination phenomena under investigation; the coding practices of the respective variables are also detailed. More specifically, the chosen method is a logistic regression analysis, which serves to identify those variables that influence ordering choices significantly and also measures their effect sizes. In order to predict orderings in the data most accurately, the model-building procedure is based on minimal adequate models, which only keep variables that improve predictive accuracy and discard non-significant variables step by step. What is more, Lohmann opts for the inclusion of scalar variables, which allows for more fine-grained analyses than logistic regression models without intercept and thus avoids information loss.

Chapter 5, which marks the beginning of the empirical part of the book, investigates the order of constituents in reversible copulative compounds. The study involves the analysis of 645 types, the large majority of which have been extracted from the Corpus of Contemporary American English (COCA). The results reveal that the order of constituents in nonce-formations is far from being ‘completely free’, as Olsen (Reference Olsen, Booij and van Marle2001: 298) has speculated, but is determined by a number of variables, with iconic sequencing having the largest effect size. Phonological factors, such as rhythmic alternation or the length of the final vowel, as well as length and frequency, also prove to be significant predictors in the model-fitting stage.

The order of constituents in binomials is the topic of chapter 6, which contrasts irreversible coordinates with cases of ad hoc coordination that might also occur in an alternate ordering. It is interesting that the picture emerging from the analysis based on the British National Corpus (BNC) is quite a varied one. Not all of the variables prove to be equally significant for irreversible and reversible binomials; for instance, rhythmic alternation is a significant determinant only with fossilized coordinates, whereas the avoidance of ultimate stress only influences ordering choices with reversible binomials coordinated by the conjunction and. Generally, effect sizes are larger in the case of irreversible expressions. The significance of the individual factors also varies depending on the conjunction used in the coordinate construction. The subtle differences that come to the fore clearly underscore the advantages of a minimal adequate model, which makes the detection of such vacillating influences possible.

Chapter 7 turns to the order of constituents in reversible complex noun phrases and investigates to what extent the factors identified for bare binomials are similarly at work in phrase ordering. This is a particularly thought-provoking question in that psycholinguistic theories such as Stallings et al. (Reference Stallings, MacDonald and O'Seaghdha1998) assume language production processes to involve two stages, with phrase ordering preceding word ordering. This implies that ‘the number of words a phrase contains should be a relevant predictor for order, but not the syllabic length of words making up the phrases’ (Lohmann 2014: 98), a hypothesis tested on the basis of the spoken section of the British component of the International Corpus of English (ICE-GB). Interestingly, this assumption is not borne out by the results, as length in number of syllables does indeed play a significant role, apart from syntactic complexity as well as semantic-pragmatic variables.

Chapter 8 discusses the results obtained from the three case studies, showing that some variables prove to affect ordering universally, while others yield mixed results. Factors that display a general effectiveness are, first and foremost, semantic-pragmatic variables (with one exception: extra-linguistic hierarchies), which also have the largest effect sizes. Furthermore, length in number of syllables and frequency are significant determinants across the board. Rhythmic alternation and syllable weight, on the other hand, only turn out to constrain ordering choices significantly in the case of copulative compounds and irreversible binomials. Against this backdrop, reductive approaches are revisited; these give precedence to factors such as stress alternation or frequency, while they regard other factors as merely epiphenomenal. A very important outcome of the multifactorial analysis is that such reductive attempts cannot account satisfactorily for the actual variation as evidenced in the coordinate constructions under investigation.

Chapter 9 provides a comparative discussion of the results with an emphasis on the distinctions between reversible and irreversible coordinate constructions, which are nicely visualised in a multi-dimensional scaling diagram (p. 129). Reversible and irreversible coordinate constructions display marked differences in that the latter group adheres much more strongly to ordering constraints than the former; also, the to-be-ordered constituents of irreversibles are crucially characterised by sharp contrasts to each other. For this, Lohmann offers an explanation in terms of ‘selection pressures’ (p. 136), which are understood as a tendency to strongly favour one ordering over the other if it complies best with ordering constraints, a hypothesis that may account for the eventual lexicalisation of such coordinate constructions:

As Lohmann further elaborates, the increased ease of processing results from the serialisation of the two constituents in an order that, due to certain activation-boosters, puts first the elements that are easily retrievable during the language production process. For this reason, first constituents of irreversible coordinate constructions can be argued to be easily selected due to their high frequency, high degree of conceptual accessibility, shortness, and the fact that they reflect extra-linguistic hierarchies, among other factors.

The processing perspective is fleshed out in chapter 10, which discusses correlations between the findings obtained from the three case studies and language production models. In this way, ample support is given for the plausibility of psycholinguistic theories that postulate the grammatical encoding phase to proceed in two stages (see Bock & Levelt Reference Bock, Levelt and Gernsbacher1994): a functional stage, which processes conceptual/semantic information, and a positional stage, which activates word forms with the corresponding morphological and phonological information. However, such a processing model with two discrete stages implies that coordinate constructions would not be affected by conceptual factors as they involve the serialisation of two functionally equivalent constituents. This assumption is clearly not borne out by the results. In this respect, a language production model that allows for incremental processing across the two stages is better suited to accommodate the actual results: ‘While the ordering process would still be located in the positional stage, it may be sensitive to the speed of lemma retrieval which itself is contingent upon conceptual characteristics’ (p. 148).

What is more, Lohmann provides persuasive evidence for interactive activation models that are based on an intricate interplay between activation and competition. With two linguistic units competing for first mention, the constituents that can be expected to win out over their rivals are either highly activated, because they are conceptually accessible or highly frequent, or else more quickly activated, because they are short or involve fewer syntactic nodes. The larger effect sizes of conceptual/semantic factors can also be accounted for by the architectural features of a processing model based on the notions of interaction and competition, as these constraints are situated on the higher layer of functional encoding and for that reason yield stronger effects. In turn, phonological or morphological factors, located on the lower level of positional encoding and thus occurring at a later stage, may affect ordering only via feedback between the respective layers. This observation helps to explain why effects of rhythmic alternation, for example, are relatively unstable in the data.

Chapter 11 sketches further avenues of research by addressing two major implications derived for other alternation phenomena in English. Drawing on findings obtained for two well-researched cases of grammatical variation, dative alternation and genitive alternation, Lohmann convincingly shows that these phenomena share striking similarities with the coordinate constructions investigated in his study: here, conceptual accessibility, that is, animacy, and information status are significant variables, together with length as another important predictor. It is also conspicuous that the effects of conceptual/semantic variables outweigh those of morphological and phonological factors, which again ties in with the results of the present study and provides further support for the processing model advocated in chapter 10. Moreover, implications for typological research are pointed out by referring to universal word order preferences, which characteristically place topical, that is, highly activated elements first, resulting in subject-initial word orders in the majority of the world's languages. The processing preferences evidenced in the three coordinate constructions thus also seem to influence ordering choices on a more universal level.

Chapter 12 wraps the book up with a concise summary of the results, once more emphasising that ‘indeed corpus linguistics and psycholinguistics do not need to go separate ways, as the corpus findings can be explained by psycholinguistic theories’ (p. 197). Lohmann's study substantiates this claim in a most impressive way.

Overall, the present book makes a highly valuable contribution to research on coordinate constructions and features one of the first empirical analyses that systematically take the issue of reversibility into account. In this regard, the perception of coordinate constructions as a gradual phenomenon and not as a clear-cut opposition between frozen expressions and more flexible nonce-formations is similar to Mollin's (Reference Mollin2014) corpus-based study on the (ir)reversibility of English binomials, in which she empirically investigates ordering constraints from a predominantly diachronic perspective. What is unique about Lohmann's approach, though, is, on the one hand, his inclusion of two further coordinate constructions, namely copulative compounds and complex binomials, and on the other hand, the discussion of his results within the context of processing-related accounts of language production.

In meticulous detail, Lohmann thoroughly explores a wide array of variables from various levels of linguistic analysis (phonology, morphology, syntax, semantics and pragmatics). In this way, he identifies at least eighteen verifiable factors that influence the order of constituents in coordinate constructions. While previous research on the linearisation of elements in coordinate constructions was largely characterised by inconclusive results, Lohmann is able to reconcile what seems to be contradictory evidence at first sight. Implementing multivariate models, he makes us perceive language variation as a complex interaction between reinforcing and inhibiting factors, which ultimately determines the choice of otherwise functionally equivalent variants. Especially noteworthy is Lohmann's observation that the variables’ effect sizes vary depending on the respective coordinate construction as well as the level of linguistic analysis that the respective variable belongs to, a finding that corroborates the high plausibility of interactive activation models and their specific architecture. In this regard, then, Lohmann's study contributes to theory-building with its immediate relevance to the evaluation and further development of language production models.

Finally, also worthy of mention is Lohmann's excellent expertise in corpus-based methodologies and multifactorial analyses – surely, his study sets a benchmark for future investigations of grammatical variation phenomena.