The stated aims of this monograph are twofold. First, the author seeks to compare and contrast the generative and grammaticalization approaches to morphosyntactic change in as objective a manner as pos-sible. According to Fischer, an objective treatment of both models requires that one adopt the view that form and function play equally important roles in how grammatical systems operate synchronically and change over time. Such a discussion further demands that one refrain from dwelling too much on a particular framework's shortcomings, and that, in diplomatic fashion, one recognizes what each theory does well. The monograph's second goal is to present an alternative model that, in the author's opinion, combines form and function in a unitary way. This third framework is a single mechanism, analogical model of morpho-syntactic change that takes as its point of departure Anttila's (Reference Anttila, Joseph and Janda2003) “analogical grid,” in addition to connectionist frameworks (for example, Pulvermller Reference Pulvermüller2002). According to this model, morphosyntactic structures are stored in associative memory where they are linked to one another based on similarities in form and function. While the author is clear about some of the broader implications of adopting such a view of morphosyntactic changeFischer explicitly rejects what can be charac-terized as an early generative view that language production and change are entirely rule based (p. 136)other consequences are implicit. For example, in opting for an analogy-based approach the author is also rejecting more nuanced (dual-mechanism) models of morphology that allow for the operation of rules alongside an inventory of linked lexical entries, as argued for in works such as Fertig Reference Fertig2000 and Pinker & Prince Reference Pinker, Prince, Lima, Corrigan and Iverson1994). Fischer's data are mostly drawn from English and its historical variants, with occasional references to Dutch and German, as well as other non-Germanic languages.

The book is divided into two larger sections, the first of which examines general questions of morphosyntactic change and discusses how historical linguists have sought to answer them. This first section also presents the formalist and functionalist approaches to accounting for such change, while at the same time providing a sketch of the analogical framework Fischer adopts. In the second part the author looks at three different morphosyntactic phenomena, critically evaluates the ways in which grammaticalization and generative theories have contributed (or failed to contribute) to our understanding of their diachronic develop-ment, and suggests how her analogy-based approach might advance the state of the art.

After a general introductory chapter, chapter 1 of part I (“General issues of morphosyntactic change”) asks the question “What is ‘good practice’ in historical linguistics?” Here, Fischer discusses what a dia-chronic analysis might seek to accomplish as well as the pitfalls that one is sure to encounter when executing such an analysis. For example, the author addresses the question of text type (§1.3), pointing out that the historical linguist's analysis must often rely on data drawn from a limited number of sources that are several times removed from the actual language of investigation (as is the case for corpora and edited texts). Further, when analyzing these data, one should not only bear in mind the limitations of the source but also the synchronic system of which the data are a part, thereby avoiding the temptation to look too much to the grammar of the language's modern descendant (assuming there is one). Hence, in this chapter Fischer advocates a practical, data-driven ap-proach to investigating morphosyntactic change.

Chapter 2 presents the two dominant frameworks in which most linguists cast their analyses of morphosyntactic change: the gram-maticalization and generative models of syntax. Fischer highlights the theories' approach to the relationship between form and function as one of the central ways in which the frameworks differ. In the case of the latter model, form and function are assumed to operate autonomously, whereas grammaticalization theory presumes that function underlies form in every respect. Already during this initial presentation of the two competing frameworks, the author's own view of morphosyntactic change starts to emerge. It becomes clear that Fischer is uncomfortable with several generative ideas, including the notions of an innate biological language faculty (p. 88) and modularity (pp. 87–88), favoring instead the view held by functionalists of all stripes that a grammar is no different from other cognitive systems. Because its task is to compare and contrast grammaticalization on one hand and generative theory on the other, this chapter emphasizes the differences between formalist and functionalist frameworks, pointing out that the formal linguist tends to ignore any functional pressure that external factors may exert on the grammar, whereas the functionalist neglects to take into consideration that form may also effect change in the grammar. In this chapter the author briefly refers to her preferred model, the analogical model, as an ideal way to reconcile the two camps, but gives this framework its fullest treatment in chapter 3.

After the discussion of the formalist and functionalist views on the relationship between form and function, chapter 3 discusses the ways in which the two models deal with language change. Most of this chapter is devoted to comparing and contrasting generative and grammatical-ization views of language change, while the chapter's final section presents an outline of an analogy-based mechanism for language learning. It is in this chapter that the author is the most explicit about her ideas of how and why morphosyntactic change occurs, and, therefore, this review devotes most of its attention to these broader ideas. As seen in the previous chapter, the focus here is on contrasting the former two models and, as a result, those analyses that are situated on the far ends of the formalist-functionalist continuum receive the greatest attention. In this chapter Fischer presents Lightfoot's Principles and Parameters approach to diachronic change (1979, 1981, 1991, and 1999) as the representative for the generative work on syntactic change and takes issue with several assumptions that serve as the foundation for such analyses. First and foremost, Fischer opposes the idea that only internal factors can effect “real,” or as Lightfoot calls it “grammatical” change, which is defined in terms of abrupt parametric shifts. Instead, Fischer argues that one must consider external factors (for example, social networks, communicative need, and pragmatics), as well as the pos-sibility that grammars change gradually, if one hopes to account for the variation that actual data inevitably exhibit. In contrast, the functionalist sees language change as a gradual, evolutionary, and unidirectional process, a view that leads some proponents of grammaticalization theory to treat those characteristics that define grammaticalization as a dia-chronic process and the individual factors that occur on the speaker level as one and the same. This failure to distinguish between the “language” and “speaker” levels or, to use different terminology, social and individual grammars, is closely connected to the conclusion drawn by some that grammaticalization itself is a cause of change. If the uni-directional trajectory as a linguistic phenomenon is as frequently attested as the functionalist claims, it must somehow be a part of the individual's grammar.

Fischer sums up her position on p. 117, where she states that she would like to see grammaticalization linguists maintain a strict division between individual and social grammars and explain the prevalence of the unidirectionality in syntactic change without conflating the two. At the same time, formalists should not ignore that a social grammar can affect an individual's grammar, that is, that external factors can also effect change. The final sections of this chapter, §§3.3–3.5, constitute Fischer's attempt to bring the two opposing frameworks together in a model of morphosyntactic change that recognizes the importance of form and function, and of both internal and external forces. To set up the discussion of her preferred model, the author summarizes those factors that she considers relevant to change in a table on pp. 126–127, which comprises three columns, the first of which lists the “main factors involved in change,” the second, the “sub-factors,” and the third, general “primitive” or “pre-adaptive constraints.” Those constraints listed in the final column are the basic cognitive abilities and social structures that underpin not only morphosyntactic change, but also language evolution and acquisition and include the ability to think iconically (that is, to perceive the differences and similarities in linguistic forms) and indexically (that is, to identify the ways in which signifier and signified are linked). The columns that list “main” and “sub-” factors of change identify a range of diachronic processes that occur on all linguistic levels and include a broad assortment of topics including “phonetic/prosodic changes,” “analogy,” “communicative and social factors,” and “fre-quency.”

The reader will note that the term “factor” is used in a somewhat ambiguous way hereFischer does not distinguish between those items in the table that are catalysts for change and those that are the outcomes of change or simply involved in change in some way. In addition to this point, it is also not clear what exactly the relationship is between the “main factors” and the “sub-factors” listed alongside them. In some instances, the main factor seems to be the underlying cause of change (as is the case for the sub-factors “phonological, morphological, lexical and semantic loss or bleaching,” which could all be seen as effects of the main factor “reduction/erosion, ease of effort, simplification”). In others, the identified main factor seems only to be a more general term that could be used to refer to a number of sub-factors (as is the case with “social factors,” under which “social networks, standardization, age, sex, class,” among others, appear as sub-factors). As a result, the overall effect of the table and the following prose is somewhat confusing. However, several important and valid arguments do clearly emerge from these sections. For example, Fischer clearly links the frequency with which a token is produced to its eventual grammaticalization (that is, repetition leads to reduction, which leads to a change from lexical to grammatical category), something that is lost in those functionalist studies that assume grammaticalization itself to be the force of change. Additionally, Fischer's point that the same factors that underpin language change should also be crucially involved in the processes of language evolution and acquisition is well taken.

The final section of this chapter is devoted to Fischer's sketch of an analogy-based learning mechanism. This model, drawn from work on analogical sets in morphology (Skousen Reference Skousen1989, Baayen Reference Baayen, Bod, Hay and Jannedy2003, Chapman & Skousen Reference Chapman and Skousen2005) and Pulvermller's (Reference Pulvermüller2002) connectionist model, consists of a single mechanism that should be able to account for language production (that is, lexical and syntactic output), acquisition, and change. In order to describe Fischer's model of grammar, we now focus on the author's presentation of how linguistic information is stored and produced and how the L1 and L2 acquisition processes unfold. Beginning with the question of lexical storage, the basic unit of storage in this grammar is the token, which, in turn, is part of a larger token-set. The token-set comprises a number of tokens that are semantically similar (as would be the case with the token apple, which would link with pear, bananas, kiwis, etc.). Each token is also identified by type. In contrast to the token-set, which links together tokens along semantic (or functional) lines, a token's type is based on its formal shape and comprises the set of lexical features that dictate how a token is used. For example, one can have a token of the formal type Noun (that is, Count Noun, Inanimate Noun, Abstract Noun, etc.). In this way, all tokens contained in a lexicon are linked together with any number of other tokens according to similarities in form and function. These tokens and their type can com-bine with others to form even larger groups. If they combine according to semantic similarity, these larger groups are called type-sets. If tokens combine on the syntagmatic level, the larger group is called a construction type, which would control the shape that the larger units of syntax could take. For example, the formal type Noun would be part of a larger construction type that includes a preceding determiner and perhaps an intervening adjective, while at the same time being embedded within larger construction types, prepositional and verb phrases to name but a few.

With regard to the acquisition process, then, linguistic abstractions, including those one finds in morphosyntax, “can be largely built via an initial concrete base (of at first token- and later type-recognition) starting with signs referring to concrete objects and activities in concrete situations, sorting these into sets and generalizing slowly type-by-type from there (p. 140).” As this quote makes clear, within Fischer's model all abstract categories are initially semantically or functionally motivated, whereas within a generative framework abstract categories are innate and provided through UG. Another point of contrast between the analogical and generative views is Fischer's treatment of L1 and L2 acquisition as exactly the same processthe only difference between the two is that L2 acquisition operates with a “far greater corpus of linguistic data than is the case for children (p. 135).” Hence, language learning is a continuous process that is accomplished through the construction of analogical sets whose formation is based on indexical and iconic relations, that is, on the back of formal and functional analogy. In this regard, Fischer's notion of a grammar is far more fluid than the generative notion of a grammar.

A more formal-minded linguist may have several questions about this proposed model of language. For example, how many construction types might one have to acquire in order to produce language? How many token-sets? How many connections between analogical sets would have to be forged? Is this a task that can be accomplished during the relatively brief time it takes for a child to acquire her first language, without the aid of UG? Drawing on Pulvermller, Fischer writes (pp. 136–137) that one hundred “lexical categories” should suffice with a few million connections that link about 100,000 input units together. Given that “most cortical neurons have above 10⁴ synapses, the large number of connections should not constitute a problem (Pulvermller Reference Pulvermüller2002:193) (quoted on p. 137).” These conclusions are not accompanied with much explicit discussion of how the authors arrived at them, and the reader who is unfamiliar with Pulvermller's work will have to take Pulver-mller and Fischer at their word.

In one way, however, Fischer presents a model of syntactic output that seems not unlike a generative view of the same. In fact, syntactic production within the author's model is still dependent on a process of discrete combination, that is, that constituents may be combined with (and swapped out for other like) constituents to create a novel utterance. According to Fischer's description of the constraints of analogical output, tokens stored in the lexicon are listed for construction type, a categorization that is identical to the syntactic categories of the generative framework. These tokens may be combined with one another, but not arbitrarily. As noted on p. 141, analogical extensions, and hence linguistic production, are constrained based on the fact that tokens and types are ordered into sets. For example, in the clause John knows Bill and Mary, one could replace John, Bill, and knows at the token level with similar Proper Noun and verb type tokens, for example Fred, Peter, and sees. One could make a similar exchange on the type-level, swapping out the coordinated phrase NP and NP for another NP (on the same type-level). One cannot, however, exchange tokens with higher NP-types or vice-versa. Hence, an utterance such as *Mary, who John knows Bill and is impossible, as it confuses the token level and the NP-type level and, what is more, Bill and is “not a possible construction-type (p. 141).” This explanation, which relies on multi-leveled, lexically-listed construction types, could easily be recast in generative terms. That is, only syntactic constituents can combine, and be exchanged with one another to form utterances. Of course, within the generative tradition constituents combine according to rules and in an infinite number of ways, whereas Fischer believes that syntactic production comprises more units of prefabricated, fixed phrases than a generativist would assume (pp. 128, 137). Yet this description of linguistic production offered on p. 137 indicates that, even within her single-mechanism model, not all syntactic production is lexically stored; some of it is generated.

The purpose of the case studies presented in the book's final chapters is to illustrate the generative and grammaticalization approaches to three different diachronic phenomena, that is, the development of modals in English, clause fusion, and changes in scope and subjectivity, and show why an analogical approach might be preferred. Fischer's main charge against the two prevailing frameworks is that their assumptions regarding language production and change yield methodologies that can lead to an inadequate or skewed treatment of the data and erroneous conclusions. On the generative side, Lightfoot's theory that grammatical change is abrupt prompts him to ignore the variation that data invariably exhibit. On the other end of the continuum, grammaticalization theory's assumption that change must be unidirectional can lead to a misinterpretation of the data, which is the case when Hopper & Traugott (Reference Hopper and Traugott2003:176) assume that the clause Numerous witnesses heard the bomb explode is “historically later” than We realize that you have to make a profit, as the first clause is “more integrated” than the second. As described in her introductory chapter, these chapters should present the relevant data so that one may evaluate both the formal and functional explanations of these linguistic phenomena, while considering Fischer's alternative at the same time. Fischer's presentation is, at times, unclear, which is to say that if the reader is, for example, unfamiliar with the vast literature on the development of English modals, he or she may have a difficult time following, much less critically evaluating, the entirety of Fischer's discussion. Furthermore, Fischer's treatment of the analogical alternative is not always the most convincing, due mostly to the fact that she does not always present adequate data to illustrate and support her hypotheses, but instead chooses just to outline the argument. The absence of data is particularly frustrating in those cases in which the author is directly discussing the issue of frequency, which she quite rightly highlights in earlier chapters as a central element in grammatical change. For example, when discussing the question of scope increase in the case of the English modals (§6.1), the author makes numerous references to the notions of frequency or “rarity,” that is, the lack of frequency, but never actually shows frequency through data. In failing to do so, Fischer misses an opportunity to properly illustrate one of the most important points raised in her monograph.