Word order in West Germanic languages is characterized by the kind of operation Ross (Reference Ross1967) labeled scrambling, which allows constitu-ents that occupy a position in the nachfeld (roughly equivalent to the vP/VP) to appear also in the mittelfeld (roughly equivalent to the TP). The phenomenon has been extensively studied in the framework of generative grammar in the last thirty years, but the issue of its proper characterization is still far from being settled. Putnam's Scrambling and the Survive Principle sheds new light on this long-standing debate in novel and appealing ways.

With the exception of those studies that have treated scrambling as an operation pertaining to non-core modules of the grammar, most analyses have focused on whether the position scrambled elements occupy is derived or non-derived. Putnam aligns himself with those researchers that have defended the former hypothesis, but crucially departs from them in the theoretical framework he adopts, namely the modified version of Chomsky's minimalist program—henceforth MP (Chomsky Reference Chomsky1995, and subsequent work)—known as survive.Footnote ¹ This framework is advanced in Stroik Reference Stroik1999, 2007 and Stroik & Putnam Reference Stroik, Putnam and Grohmann2007. The dif-ferences between the two approaches are dealt with in chapter 1. Putnam defends Survive as an optimal solution to the conflict between those aspects of the MP that seem to be incompatible with the requirements of the strongly derivational model of syntactic computation in Brody Reference Brody1995, 1998, 2002, on the one hand, and the kind of crash-proof syntax Frampton & Gutmann (Reference Frampton, Gutmann, Epstein and Seely2002) have proposed, on the other. In a nutshell, Survive gets rid of operations such as attract and internal merge, together with certain mechanisms orthodox Minimalism resorts to in order to limit the range of syntactic structures in which such operations apply, specifically match and the phase impenetrability condition. Instead, it makes use of a single operation merge in two variants, merge and remerge. Merge concatenates two syntactic objects in the same fashion as standard external merge, that is, the concatenated syntactic objects enter the syntactic derivation directly from the numeration. Remerge—the Survive replacement for standard Internal Merge—creates a concatenation between two syntactic objects, one of which has already undergone Merge but, nevertheless, must still be present in the Numeration, according to the survive principle (see 1).

1. (1) The Survive Principle (based on Stroik Reference Stroik1999:286)

   If Y is a syntactic object (SO) in an XP headed by X, and Y has an unchecked feature [+F] that is incompatible with the feature X, Y remains active in the Numeration.

It is important to note here that Survive Minimalism departs from Chomsky's MP in its rejection of uninterpretable features, even though some sort of feature checking (valuation, deactivation, etc.) is the trigger for Merge in both approaches.Footnote ² From this perspective, the task of the strictly syntactic component is not to make uninterpretable features invis-ible to the interfaces, but rather to make interpretable features appear in the kind of amalgamated configuration that may receive a proper LF and PF interpretation, an operation Putnam (p. 20) calls link! (see 2).

1. (2) Link!

   If XP is a syntactic object (SO) bearing the interpretable feature α, and a syntactic head Y° enters the derivation bearing the interpretable feature α, XP will remerge into the structural position immediately dominating Y° [Spec YP], at which point the matching features will amalgamate and be interpreted at the interfaces.

Equipped with these tools, Putnam interprets West Germanic scram-bling as the product of the amalgamation of the interpretable [+Referential] feature on the scrambled phrase with an identical feature on the head of the RefP that he contends appears in the West Germanic Mittelfeld. The main pieces of evidence for his proposal appear in chapters 2 and 3, where he deals with both traditional and new sets of data. Chapter 2 is devoted to showing that West Germanic scrambling does not quite correspond to any of the standard movement processes in the generative literature (NP movement, wh-movement, and topical-ization), despite the fact that it shares some of their characteristics. Putnam clearly has reasons for dispensing with the “insurmountable task of classifying scrambling within the archaic A vs A-bar movement typo-logy” (p.95); however, not all the properties he ascribes to scrambling structures are equally unquestionable. Thus, he takes parasitic gaps to indicate that scrambling is a process targeting a derived position in West Germanic, but this is still in my view an open issue (see the arguments in Fanselow Reference Fanselow1993 and Haider & Rosengren Reference Haider and Rosengren1998, 2003). The dismissal of headfinalness as a distinctive characteristic of scrambling, as opposed to the other types of movement, is another controversial move. Putnam justifies it already in chapter 1 by appealing to Yiddish, purportedly a VO language in which scrambling is, nevertheless, possible. However, pace Diesing Reference Diesing1997, there is fairly compelling evidence that Yiddish is OV (Vikner Reference Vikner2001, Haider & Rosengren Reference Haider and Rosengren2003). Furthermore, Putnam does not account for the fact that in both German and Dutch scrambling is restricted to elements of head-final projections, namely VPs, APs, and postpositional phrases, as argued by Corver and van Riemsdijk (Reference Corver, van Riemsdijk and Palek1996), Haider & Rosengren (1998, 2003), Chocano (Reference Chocano2007), among others.

Putnam shows in chapter 3 that his analysis of scrambling yields an elegant account of the sets of data previously reviewed. The very inter-esting opening section is, however, devoted to providing sound evidence in support of the legitimacy of the [+Referential] feature as a syntactic feature in languages other than West Germanic (multiple wh-movement in English, axiomatic negative predicates in Xhosa, and DPs in Old Norse).

Putnam first deals with scrambled accusative objects that bind dative anaphors in German. These structures have traditionally resisted a prin-cipled explanation, due to the contrasting ungrammaticality of scrambled datives binding accusative anaphors. Putnam resorts to the notion of anti-locality and the condition on domain exclusivity as formu-lated in prolific domains (Grohmann Reference Grohmann2000, 2003, and subsequent work), and attributes the difference to the fact that both binder and bindee are indeed the same phrase undergoing Merge in two different positions.Footnote ³ Accusative objects are first merged in a position lower than that occupied by datives, and may raise to it. On the contrary, datives cannot lower onto the position associated to accusatives, as shown in 3a,b.Footnote ⁴

1. (3)

   

Putnam's conclusion is that the inability of scrambled datives to bind accusative anaphors in German is unrelated to their status as scrambled constituents and strictly “determined by the initial Merge positions of arguments in the vP” (p. 123). He is certainly right to restrict his analysis to the binding of anaphors. However, as Haider & Rosengren (Reference Haider and Rosengren2003) argue, the predicate appears also to be reflexive in the sense of Reinhart & Reuland (Reference Reinhart and Reuland1993), as shown by examples such as 4, where the dative can bind an anaphor embedded within the accusative object.

1. (4)

A second set of data Putnam revisits in chapter 3 concerns parasitic gaps. His account parallels the proposal for Condition C asymmetries advanced in Stroik & Putnam Reference Stroik, Putnam and Grohmann2007, and resorts to the notion of late num merge. Since, according to the Survive Principle, elements that undergo Merge in the derivation return to the Numeration if any of their features remain unchecked, it is expected that they may form new syn-tactic objects there. Putnam capitalizes on Dutch examples such as 5, where the licenser of a parasitic gap can also bind an anaphor.

1. (5)

After entering the derivation for the checking of its theta-role and Case-features, den rivalen returns to the Numeration due to its unchecked [+Referential] feature. There it undergoes Late Num Merge with the elements that constitute the adverbial clause. The adverbial clause enters the numeration as a unit, but the still unchecked [+Referential] feature makes den rivalen go back again to the Numeration, from where it may be taken for Remerge with the head of the RefP in the Mittelfeld.

Anti-locality and Prolific Domains are also crucial in Putnam's expla-nation of the third set of data studied in chapter 3, namely the freezing effects observed in the scrambling of elements out of already scrambled constituents, and their absence in the case of scrambling out of topical-ized phrases. Thus, in 6a, das Brot undergoes Remerge twice within the same domain (TP), as opposed to 6b, where Remerge takes place first in the TP and then in the CP.

1. (6)

   

This analysis runs into some empirical difficulties due to the gram-maticality of structures such as 7, where both a weak/clitic pronominal (das, es, 's) and a remnant coherent infinitive appear within the same domain.

1. (7)

Putnam's explanation for this kind of example is postponed until chapter 4, which is devoted to the prosodic side of scrambling. He sides with Chocano (Reference Chocano2007) in rejecting previous strictly syntactic accounts (Müller Reference Müller1998), and hypothesizes that “certain phenomena such as per-muted word orders in the middle field and scrambling from coherent infinitives are marked as ungrammatical at the level of prosodic structure rather than by evaluation of syntactic structure alone” (p. 175). The contrast between 6a and 7 is attributed to a requirement that phonological phrases landing in the TP-layer as a result of remnant movement must constitute a single phonological phrase. This is possible in 7, where weak/clitic pronouns undergo phonological restructuring, but impossible with the full NP of 6a. I think, however, that Putnam's solution needs further refinement in view of examples such as 8, which, according to Gärtner and Steinbach (Reference Gärtner and Steinbach2003), constitutes a well-formed instance of I-topicalization in the Mittelfeld (capital letters mark stress).

1. (8)

Putnam's study of the prosodic side of scrambling also includes an interesting analysis of the shibboleth rhematic vir in Afrikaans as the surface manifestation of the interaction of Grohmann's Prolific Domains with the kind of prosodic checking defended in Ackema & Neeleman 2004. Aware that this analysis seems to be tied to a model of multiple spell-out incompatible with the basic tenets of Survive, in the closing section of the chapter Putnam makes some tentative suggestions about the “survival” of Prolific Domains in the Survive model.Footnote ⁵

Despite the empirical problems some of Putnam's specific proposals may encounter—undoubtedly related to the large and complex sets of data any research on West Germanic scrambling must invariably face—I fully agree with his words in the final chapter:

It will certainly prove worthwhile to explore their possibilities in full.