This monograph discusses some of the major problems in German syntax from the Lexical Functional Grammar (LFG) perspective codified in Bresnan 2001. Although there are some LFG papers that discuss phenomena in German syntax and there is an older introduction by Berman and Frank (1996), there was until now no book-length work in the LFG framework that focuses uniquely on the syntax of German. This work, although still a rather slender volume, changes this state of affairs.

After an introduction to LFG based on Bresnan 2001 the book provides a treatment of some often, and some less, discussed phenomena and spells out their relevance to LFG theory. Chapter 3 (the first substantial one) translates the topological model of German sentence structure into a phrase structure model with functional projections, and centers on the problem of whether German has an IP along with a CP. Following mainly arguments given by Haider (1991, 1997), Berman adopts a non-IP analysis. She also follows Haider in proposing a flat structure for the clause final verbal complex. In the course of this chapter, Berman assumes, without much discussion, that German follows the endocentric mapping principles of Bresnan 2001 for the projections of C. This means that she assumes that these German c-to-f-structure mappings are configurational. For the projections of V, she also proposes several levels of VP embedding, but does not clearly say which role they play, because she proposes that the grammatical functions in the middle field are identified through lexicocentric function specification (case marking). Various levels of VPs in the middle field are often proposed in GB to account for the different partial VP-fronting possibilities, but LFG does not need this device.

Chapter 4 argues that although German has no subject position it requires a subject function in every sentence. Given the obvious absence of overt subjects in several types of German sentences, that point of view needs to be supported. Berman adapts to LFG a proposal made in GB, namely that the 3rd person marking on the tensed verb in ostensibly subjectless sentences constitutes the subject. In zu-infinitives with anaphoric control, the subject function with a “PRO” value is introduced by a functional equation on the zu and in clauses with functional control via the functional control itself. As the discussion shows, the LFG machinery makes it very easy to introduce categories in the functional structure that do not have a corresponding c-structure. This makes clear arguments for such categories desirable. Berman shows that her proposals can be made work. However, she does not show that they solve any problems, such as the occurrence of nominative phrases in infinitives (for example, einer nach dem anderen ‘one after the other’), in a more elegant or economic way than other alternatives. Still, following GB tradition, she tries to link her treatment of verbal inflection as introducing a subject to a typology of languages with and without overt subjects. She contrasts Scandinavian languages without verb agreement with German to account for the obligatoriness of overt subjects in the former. However, she also classifies Dutch as a language with obligatory overt subjects; but it is well known that this is the wrong generalization for the majority dialect, where the expletive is optional. Moreover, it creates a problem for her account, as the morphology of Dutch is rich enough to allow the subject to be absent in the c-structure. Of course, the optionality of the Dutch expletive also creates a problem for the proposal advocated.

Chapters 5 is likely the most interesting for readers focusing on LFG theory rather than on German syntax because it discusses the introduction of empty c-structure categories (traces), a relative novelty in the theory. Berman follows Bresnan 2001 in proposing a limited set of empty c-structure categories in LFG on the basis of a study of weak crossover phenomena. The arguments follow Bresnan 2001 rather closely for long distance crossover (Bresnan's account in turn is based in part on an earlier paper by Berman). The account relies on empty c-structure categories to handle contrasts such as that between 1 and 2.

The argument is based on the assumption that one needs to refer to linear order to state the exact conditions on the grammaticality of such crossover configurations. Bresnan (2001) motivates the need for empty categories in long distance dependencies by the mapping principles from c-to-f-structure, which state that in languages such as English, GF are mapped from configurationally identified c-structures. Berman assumes the same account for German.

Berman accounts for the difference in grammaticality between German and English sentences, such as those found in 3, by assuming a different status for German local subjects and local topicalization: they are in the domain of lexicocentric function specification (identification through case marking).

The proposal then assumes that there needs to be a trace in the German middle field to account for long distance crossover, but given the lack of crossover effects within the middle field it can be anywhere. The contrast between English and German with respect to object crossover is covered by some not very well-worked out but not implausible assumptions about constraints on argument structure.

Berman discusses the counterproposal of Dalrymple et al. 2001 that does not require traces, but rejects it because it does not rely on the endrocentric mapping principles and the morphological function specification proposed in Bresnan 2001. It seems rather frivolous to accept a major revision of the theory and the whole philosophy of LFG without discussing in detail exactly what damage the Dalrymple et al. 2001 proposal does to the architecture of the c-structure to f-structure mappings. As far as I know, the long distance crossover facts are the only argument for traces in German (and any other language). One cannot reject a counter analysis by simply saying that it would contradict assumptions that are made only to make one's own analysis possible.

The assumption that German is an endocentric (configurational) language in the relevant aspects, and the c-to-f-structure mapping principles defined in Bresnan 2001 make it impossible to account for long distance dependencies without traces, and this is the path that Berman follows without further discussion in chapter 6. In light of these assumptions, she discusses both a functional uncertainty and a cyclic approach with iterating local dependencies without adducing data that would clearly favor one over another.

In chapter 7, “Distribution of Sentential Subjects”, Berman defends two claims: that all embedded tensed clauses in German are at the periphery (left-dislocated or extraposed), and that clauses have the same grammatical functions as nominal arguments, so that one can dispense with the COMP function. I find this chapter rather badly organized and do not understand very well what these two claims have to do with each other. What follows is my best attempt to make sense of it. The claim that sentential arguments have the same grammatical functions as nominal ones is based on a reconsideration of arguments presented in Dalrymple and Lødrop 2000, who propose a mixed analysis where some clauses are COMPs and others are OBJs. Berman argues that some of their arguments rest on the wrong examples, and that when the right examples are chosen it can be argued that that-clauses in German can be OBJs or OBLs, and that none need to be assumed to be COMPs. An argument in favor of COMP-clauses given in Dalrymple and Lødrup 2000 is based on the contrast between 4 and 5.

The assumption here is that OBJs can be topicalized, whereas COMPs cannot. Berman proposes to reanalyze apparent sentential topicalization as left-dislocation with deletion of the resumptive pronoun, which is seen as a case of topic drop. As Berman observes, this analysis is an LFG adaptation of an analysis proposed for Dutch by Koster (1978) and for German by Oppenrieder (1991). As topic drop is only possible with subjects and (accusative) objects, it accounts for the contrasts above.

Note, however, that this treatment of “topicalized” that-clauses does not require the assumed function assignment: left-dislocated elements are not assumed to have the same function as the element that they are anaphorically connected to. In fact they are generally assumed to have only a discourse function and no subcategorized one.¹

This, of course, is no argument against the proposed analysis, only against the way the argument is structured. Things become more puzzling if one looks at the second position where that-clauses can occur: extraposed in the Nachfeld. This is a position where nominal elements cannot occur. As sentences are not morphologically marked, we need to develop new positions to identify them. Berman proposes to right adjoin them to the VP. However, again this is not a position for other subcategorized arguments, so it does not jibe with the proposal that that-clauses have the same function as nominal arguments. Here I would rather argue that this is an argument against this proposal, as maintaining it weakens the mapping theory. What then are the arguments in favor of OBJ and OBL function? They are the ones developed in Dalrymple and Lødrop 2000: alternation with a nominal object, passivization, and coordination. However, passive sentences can be analyzed as involving topicalization/left-dislocation as above. Alternations have to be allowed in general, so the main argument is coordination. Here Berman makes the interesting observation that the (a) versions of the following sentences are grammatical, whereas the (b) versions are not.

However, she does not propose an analysis of this contrast. She suggests that the sentence final versions can be assumed to be extraposed, but if that is the case this would be the first clear violation of the coordinated constituent constraint, and if we are to assume null pronouns in the second constituent of the grammatical clauses it would be nice to have some arguments for them.

To summarize: left-dislocated clauses do not need an OBJ or OBL function, extraposed ones do not pattern with nominal OBJs or OBLs, and so are no argument for a similarity in function. The contrasts in 6 and 7 might form the basis for an argument if identity of function is required for coordination, but they remain unexplained under the analysis given.

The last substantial chapter discusses the occurrence of es with finite clauses. It is too complex to summarize, as it relies on six interacting assumptions, only one of which is specific to this discussion. It is a rather satisfying demonstration of the power of the framework developed in the previous chapters. The book ends with a summary and a list of some open problems.

Overall, Clausal Syntax of German is a more than competent analysis of some major problems in German syntax in the LFG framework. From the discussion above, it is most likely clear that its main quality is not originality, which is not surprising given that this is a reworked doctoral dissertation. The lack of originality is potentially positive because specialists in German syntax could read the book to familiarize themselves with LFG, as it allows one to see clearly what is common to previous analyses and what is specific to LFG. Most of the discussion is clear and coherent, adapted ideas are correctly attributed to their original authors, and problems are clearly flagged. Together with Bresnan 2001, this work constitutes an excellent source for anyone who wants to teach an introduction to LFG with German data.