Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-13T15:41:41.504Z Has data issue: false hasContentIssue false
  • English
  • Français

Deriving Coordinate Symmetries: A Phase-Based Approach Integrating Select, Merge, Copy, and Match. By John R. te Velde. (Linguistics Today, 89). Amsterdam: John Benjamins, 2005. Pp. x, 385. Hardcover. €135.

Published online by Cambridge University Press:  10 November 2006

TIMOTHY OSBORNE
TIMOTHY OSBORNE
Affiliation:
Pennsylvania State University
Rights & Permissions [Opens in a new window]

Abstract

An abstract is not available for this content. As you have access to this content, full HTML content is provided on this page. A PDF of this content is also available in through the ‘Save PDF’ action button.

Information

Type
REVIEWS
Information
Journal of Germanic Linguistics , Volume 18 , Issue 4 , December 2006, pp. 321 - 337
Copyright
© 2006 Society for Germanic Linguistics

1. Derivation by Phase.

John te Velde's monograph presents a comprehensive and detailed theory of coordination cast within the Minimalist framework. The data are drawn mainly from English and German, although some Dutch data are included as well. The novelty of te Velde's approach resides mainly in the understanding of the derivation; it builds on Chomsky's (1998, 1999) concept of phase. The derivation proceeds phase-by-phase, rather than sentence-by-sentence. Derivation by phase has each conjunct of a (certain type of) coordinate structure being derived before the derivation of the next conjunct begins. In this manner, a derivational procedure such as Move can apply to a constituent in the one conjunct without having to apply to the parallel constituents in the other conjunct(s). The parallelism that does usually obtain across conjuncts is established by the interaction of the processes Copy and Match with active memory (AM). Once the initial conjunct is derived, its syntactic and semantic features are placed in AM, so that they can be copied to the non-initial conjunct(s) as soon as it/they are derived. The relevant features of a preceding conjunct are always pasted to those of the following conjunct(s).

This approach contrasts with Across-the-Board (ATB) accounts, which assume that parallelism across conjuncts is established by the simultaneous application of a derivational procedure to all the conjuncts in a uniform manner. The utility of derivation by phase is perhaps best illustrated by instances of asymmetric coordination such as in 1 (adapted from p. 82).

The two conjoined clauses in 1 are asymmetric in the following sense. The second conjunct involves verb raising, whereas the first conjunct does not. An ATB account would have difficulties with such cases, since it assumes that movement in each conjunct within a coordinated structure proceeds in a like manner. Any account that focuses on conjunct symmetry is therefore challenged by such cases. In contrast, the derivation by phase approach allows the derivations of the two conjuncts to be distinct. What counts is that core aspects of parallelism obtain; it can suffice if the conjuncts are parallel with respect to certain features. In the case of 1, the conjuncts are parallel insofar as they are both TPs.

2. Central Components.

While derivation by phase is perhaps the most unique and distinguishing characteristic of te Velde's approach, there are a number of central components to the theory. These components are briefly enumerated here.

The approach strives to account for the structure and behavior of coordination via the principles and procedures that are already acknowledged for non-coordinate structures. It maintains that coordination can be accounted for by using the standard derivational processes of the Minimalist Program that already exist for non-coordinate structures, that is, Select, Copy, Merge, and Match. No extra mechanisms or concepts are necessary. This view assumes that the structure of coordination is actually quite similar to the structure of subordination. The traditional stance, in contrast, assumes the two types of structure are quite distinct.

The structure in 2a has the conjuncts in equi-level positions, it represents traditional assumptions about the nature of coordination. In contrast to 2a, something like the structure in 2b has become widely accepted in the generative paradigm. There one sees that the conjuncts are no longer in equi-level positions, but rather the second conjunct is on a level below the level of the first conjunct and is hence subordinate to the first conjunct. Te Velde's approach pursues a variant of 2b.

The variant of 2b that te Velde assumes differs substantially from other proposals in the area. Compare the following structures involving coordinated DPs adapted from pages 26 and 30:

The structure in 3a is that of Munn (1987), Kayne (1994), Johannessen (1996, 1998), and Wilder (1994, 1996. 1997), etc. It sees the coordinator as the head of the entire coordinate structure, projecting up to the root node. Te Velde's structure in 3b, in contrast, grants the coordinator a unique status. DP2 merges (it does not adjoin) to the tail end of DP1 in such a manner that the coordinator does not receive the status of a head. The coordinator in 3b does not project at all; it is therefore a defective lexical item. Te Velde produces numerous arguments demonstrating that 3b should be preferred over 3a.

Concerning the large versus small conjunct debate, te Velde takes the middle ground. On the one hand, he assumes small conjuncts whenever possible, for example, DP+DP, NP+NP, AP+AP, and PP+PP. The nature of the structure in 3b even allows for certain small conjuncts that do not actually qualify as constituents, given traditional assumptions; for example, John gave [Mary a record] and [Bill a book] (p. 202). On the other hand, te Velde assumes large conjuncts when necessary, namely when the unique structure illustrated in 3b no longer allows for strictly small conjuncts. There are three kinds of structures in which ellipsis is necessary: left-edge ellipsis (LEE), right-edge ellipsis (RNR), and conjunct-internal ellipsis (Gapping), as exemplified in 4.

The ellipses must be licensed by a c-commanding element. In the case of LEE, the licensor is the coordinator itself. In the cases of RNR and Gapping, the licensor is a prosodic element. Derivation by phase requires that the ellipses in these cases have all the properties of lexical items except their phonetic features (p. 9, 10, 181, 225, 290, 292). The full lexical items are generated in the course of the derivation, but they experience non-spell-out. In other words, the lexical items are deleted (p. 180f.). Despite his claims to the contrary (pp. 179), this heavy reliance on deletions actually positions te Velde's theory between the large conjunct and small conjunct approaches. I return to this point in section 5.

Te Velde emphasizes that the structure in 3b and derivation by phase with feature matching can together best address the characteristics of coordinate structures. On the one hand, the strong tendency for the conjuncts of coordinate structures to be parallel is accounted for by feature matching. Syntactic parallelism is assured when the features of the initial conjunct are pasted to each of the following conjuncts. Semantic parallelism is then assured by matching in LF. On the other hand, the numerous asymmetries that can obtain between the conjuncts of coordinate structures can be accounted for by the asymmetric structure shown in 3b, partial feature matching, and derivation by phase, whereby the derivation of one phase (≈ conjunct) need not always be exactly parallel to the derivation of the following phase(s).

3. Organization and Presentation.

The book contains five chapters. Chapter 1 is brief; it provides an overview of the theory and of the presentation in the following chapters. Chapter 2 examines non-elliptical instances of asymmetric coordination. Asymmetric structures, such as those in 5a–c, are explored.

Despite the non-parallel categories, these structures are licit. Te Velde assumes that asymmetric conjuncts are licensed by virtue of matching of abstract features. He uses the term COORDINATE FEATURE MATCHING (CFM) (p. 13) to refer to the phenomenon. While the conjuncts (DP+AP) in 5a lack parallelism in syntactic category, they are parallel insofar as they both appear in the same position of the VP and both carry the abstract feature [+COMPL]. The situation is similar in 5b. There the conjuncts again lack parallelism in syntactic category (DP+CP), but they are parallel with respect to the abstract feature [+OBJECTIVE]. The situation in 5c is somewhat different. While the conjuncts there lack parallelism with respect to the position of the finite verb, they are parallel with respect to the abstract feature [+CLAUSE]. The varying verb positions are possible by virtue of derivation by phase. The first phase (= initial conjunct) is derived and spelled-out before the second phase (= second conjunct) is. Multiple spell-out of this sort allows the conjuncts to undergo separate derivations.

Chapter 3 presents the particulars of the structure te Velde assumes for coordination, that is, the structure illustrated in 3b, as well as the mechanics of derivation by phase. The discussion remains with non-elliptical structures, elliptical structures being addressed later in chapters 4 and 5. One of the main goals of the chapter is to convince the reader of the virtue of 3b over 3a. This point is convincingly argued based on data from various areas, for example, subject-verb agreement and feature checking, multiple conjuncts (not just two), and the defective nature of the coordinator. In presenting the mechanics of derivation by phase, the emphasis is placed on the nature of the derivational procedures, that is, Select, Merge, Copy, and Match. The first three procedures occur in “narrow syntax,” whereas the last occurs both in narrow syntax and then again at LF. Finally, the chapter returns to the asymmetries of coordination, examining them again in view of the particular aspects of structure and derivation that were presented earlier in the chapter.

Chapter 4 explores the motivation for positing the existence of gaps, whereby a gap is assumed to have “all the features of a lexical item except its phonetic features” (p. 181); following Chomsky (1995:202–203), “deletion” is assumed to occur “in PF” (p. 180); the phonetic features of the deleted lexical items are not spelled out. After examining the motivation for gaps, the discussion investigates the manner in which the gaps are licensed. Each gap must have a licensor. The gap of LEE is licensed by the coordinator itself, since the coordinator immediately c-commands the gap. The gap of RNR is licensed by the prosodic feature [PROS], which appears in the position that immediately c-commands the gap. And the gap of Gapping is also licensed by the prosodic feature [PROS], but this time [PROS] appears in a position that immediately c-commands C° or T° (p. 224). The structures in 6 show the basic licensing constellations for English.

In addition to the requirement that gaps be licensed in PF, they must also be recovered in LF. The discussion of recovery is mainly concerned with the gaps that can and cannot be recovered. The central claim about recovery is that the licensor must c-command and immediately precede the gap. A gap is not licensed if an overt element intervenes between it and its licensor. This requirement predicts ungrammaticality in cases such as 7.

The sentence in 7a is an attempt at LEE; this attempt fails because the stressed adverb truthfully intervenes between the licensor of the gap, that is, the lexical item and, and the gap itself, that is, e1.1

Examples where the pre-verbal adverb is not stressed constitute exceptions, as in Billi doesn't want to write the short story and definitely ei won't consider a novel. The non-stressed adverb definitely does not block the licensing of the gap, despite te Velde's claim. This sort of data motivates te Velde to modify the claim. He must stipulate that non-stressed adverbs do not constitute blocking categories (p. 206).

The sentence in 7b is an instance of Gapping; it fails because the [PROS] feature cannot license e2 through the blocking element Betty. The last section in chapter 4 presents and examines the derivational specifics of coordinate structures involving ellipsis. Emphasis is placed on identifying the correct order of the derivational processes.

The final chapter, chapter 5, is comparatively brief. It discusses some remaining issues, whereby the focus is on the correct analysis of the pre-verbal projections. Te Velde weighs in on the CP vs. TP debate for German. He argues against the generalized V→C analysis for all V2 clauses in favor of the TP analysis; that is, he argues for a TP analysis of subject-initial V2 clauses. His analysis allows for fine-tuning of the left periphery, whereby additional functional projections are assumed to intervene between CP and TP.

4. Discussion.

Te Velde's monograph represents a comprehensive study of coordination. The data examined and the issues addressed show the challenges and difficulties the phenomenon of coordination presents for syntactic theories. One sees why coordination is one of the most studied areas of syntax and why the area continues to motivate a seemingly unending stream of descriptive and explanatory accounts. Te Velde demonstrates that an approach in terms of derivation by phase is in a position to address many of these challenges.

The basic structure te Velde proposes for coordination illustrated in 3b is indeed more appropriate than the alternative structure in 3a. The alternative in 3a is suited to address the asymmetries of coordination, since it has the conjuncts appearing in inherently unequal positions, that is, in Spec-&P and Comp-&P. It has difficulty, however, accounting for the symmetries of coordination, since there is nothing inherently symmetric about it. In this regard, te Velde's structure in 3b is less asymmetric since the one conjunct is merged onto the tail end of the other. It is hence in a better position to allow for the symmetries of coordination.

There are, however, a number of significant problems with te Velde's theory and the manner in which it is presented in the book. The presentation suffers from redundancy. The book is long, over 300 pages of main text with some 370 notes. The latter appear on the 41 pages that follow the main text. Working through the book hence requires a lot of flipping back and forth. More importantly, key points are repeated more than necessary, the result being that the text is redundant at times. For instance, the arguments for rejecting the structure in 3a in favor of the one in 3b are presented first on pages 25–27. These arguments appear again in detail on pages 89–99, and then thereafter in parts on pages 108, 112, 115–117, 122–123, and 132. The discussion of Ross' (1967) data is another example of redundancy. Ross illustrates that the coordinator forms a constituent with the following conjunct, as opposed to with the preceding conjunct. The relevant data appear first on page 105, then again on pages 129 and 145.

Some of the grammaticality judgments are debatable, for example, 2b (p. 3), 11a (p. 23), 64 (p. 48), 76a′, 77a′ (p. 219), iia,b (p. 344, n. 54), ia (p. 344, n. 58). One area where te Velde's data differ from the literature is with those instances of gapping where an auxiliary verb has gapped without the main verb:

Ross (1970, 1976) and Sag (1976:145) take instances of gapping such as 8a in English to be marginal.2

Sentence 8a actually allows two distinct analyses:

  • i. Peter has kissed (the young bride) and Paul (has) insulted the young bride.
  • ii. Peter has kissed (the young bride) and Paul insulted the young bride.

On the analysis shown in i, gapping does not obtain at all. Te Velde, however, assumes the analysis in ii.

Maling (1972:106–107) and Lechner (2001:697f.) dislike cases such as 8b in German. Te Velde, however, assumes they are fully acceptable. The actual status of such cases can have an impact on the overall theory, since it provides a clue for differentiating between gapping and non-gapping instances of coordination.

The theory's reliance on deletions opens it up to the major criticism facing all large conjunct approaches. Te Velde actually rejects the large conjunct approach in its purest form (p. 179), that is, where every conjunct qualifies as a CP below the surface. At the same time, however, his theory posits three types of ellipsis (LEE, RNR, and Gapping). Ellipsis is understood as the result of deletions, which means that the approach pursues a large conjunct account of these three cases.

The pre-deletion reading in 9a involves two people, whereas the post-deletion reading involves just one; the pre-deletion reading in 9b is nonsensical; and the pre-deletion reading in 9c is truth-functionally distinct from the post-deletion reading. Te Velde does not appropriately address the difficulty such examples pose. Note 25 on page 304 does seem to acknowledge the problem with 9a, suggesting that a “constraint against opacity” is responsible for the contrast. In order to overcome the problem in 9b, the derivation on page 259 must first select a set of paintings as the object of each conjunct, delete these objects, and then replace them with the same set of paintings. The difficulty with 9c is addressed by positing that the prosodic feature necessary for licensing of the gap conflicts with the prosodic feature associated with the negation. These explanations are unconvincing.

Te Velde's theory is not unlike many other theories in this area. Most accounts of coordination within the generative paradigm acknowledge deletions at least to some extent; for example, Kayne 1994, Wilder 1994, 1996, 1997, Johannessen 1996, 1998, and Hartmann 2000. Therefore te Velde's theory cannot be challenged on this point from within the generative paradigm. Concerning those frameworks that are less likely to so readily embrace deletions, however, for example, HPSG, LFG, Categorial Grammar, Word Grammar, and Meaning-to-Text, the heavy reliance on deletions reduces acceptance for the overall theory. To gain greater acceptance, theories of coordination that utilize deletions need to convincingly address in a transparent manner the challenge posed by data such as 9a–c.

The two main arguments te Velde produces to demonstrate the presence of LEE are also unconvincing (pp. 185–187). The first of these involves indefinite subjects:

Perhaps the most natural reading of 10 involves two women, one in England and one in the US. This reading is thus consistent with the pre-deletion structure, a fact that supports the presence of the ellipsis. Te Velde emphasizes this point time and again. The weakness in this reasoning, though, is seen in example 9a above, where the reading involving just one person is the only reading available. Example 11 further illustrates the problem.

Klein (1993) and Hartmann (2000:67ff.) take such sentences as evidence against left-edge ellipsis; that is, the most natural reading of 11 involves only one person, contrary to what the pre-deletion structure predicts.3

Klein and Hartmann do not actually argue against the possibility of ellipsis in such examples, but rather they never even consider LEE a possibility in such cases. Their discussions of 11 focus on the contrast between 11 and i:

Unlike sentence 11, the preferred reading of i involves two people. Hartmann uses this contrast to argue in favor of ellipsis in RNR.

It should be noted as well that examples such as 9a and 11, where just one person is involved, are the rule, whereas examples such as 10, where two persons are involved, are more difficult to generate. Te Velde thus emphasizes the significance of the exception in the debate, and in so doing, deemphasizes the rule.

The second argument te Velde produces in favor of subject gaps in LEE references the V2 principle. Without an ellipsis, the clause in the second conjunct would violate the V2 principle. This principle requires one (and only one) constituent to precede the finite verb. While this argument is valid given the structure te Velde assumes for coordination, that is, the one shown in 3b, its validity relies on that structure (or one similar to it). Since non-initial conjuncts are right-merged onto the tail end of the initial conjunct, the possibility that the two conjuncts might share the subject is not available, for these conjuncts are in hierarchically distinct positions. In this regard, one should acknowledge that the structure traditionally assumed for coordination, that is, the one shown in 2a, can assume the subject is shared, for the conjuncts there are equi-level. The traditional approach to coordination can hence assume the following analysis: Jemand [kam um vier] und [ging um fünf] ‘Someone came at four and left at five’. This analysis has conjoined finite VPs, whereby the finite verbs can share the subject by virtue of their equi-level appearance. In short, te Velde's second argument in favor of LEE arises out of theory-internal necessity.

Te Velde's understanding of Gapping is also problematic. It sees the prosodic feature licensing the gap appearing in the position that immediately c-commands the gap. The examples illustrating this licensing all have the ellipsis appearing in C° or T° (pp. 215, 218, 223), and the claim is made that “gapping occurs clause-internally and licenses only a finite verb and one or two redundant objects immediately following it” (p. 214). In this regard, it is not clear how the theory accounts for the distribution of gaps in cases such as 12a–c.

The gaps in these cases do not appear in C° or T°, and examples 12b and 12c do not involve finite verbs. Example 12c does not even contain a verb.

Another difficulty with te Velde's understanding of gapping occurs due to the prosodic licensing of the gap. The relevant prosodic feature [PROS] is reliant on the contrastive stress placed on the initial remnant, which means [PROS] immediately follows the initial remnant. Given this assumption, no gap may ever precede the initial remnant because it would not be licensed by a c-commanding [PROS]. In this respect, many cases that are arguably instances of gapping cannot be acknowledged as such because there is an unlicensed gap preceding the initial remnant; for example, the cases in 13a,b.

Example 13a is an instance of Stripping; it is ambiguous. Many linguists assume Stripping to be a particular manifestation of Gapping, for example, Williams (1977:112, note 6), Chao (1988), and Lobeck (1995:28). On the reading where me is the logical subject, the gap follows the remnant, as te Velde's theory assumes. On the other reading, however, where me is the logical object, the gap must precede the remnant, which te Velde's theory does not allow.4

Te Velde might analyze such cases in terms of “Late Merge” (p. 122), whereby the second conjunct is merged onto the structure after the derivation of the entire sentence is complete. The question to be addressed in this regard is whether Late Merge and Stripping might actually be different names for the same phenomenon.

Example 13b is also arguably an instance of Gapping because of the flexibility in phonological identity of the finite verb across the conjuncts is characteristic of Gapping (du arbeitest versus er arbeitet).

The same problem facing te Velde's theory is evident in the ambiguity of 14.

This question can be a polar question (Answer yes/no), in which case there is a single intonation curve with rising intonation on tea; or, it can be a constituent question (Answer coffee/tea), in which case there is rising intonation on coffee, a pause, and then falling intonation on tea. The former reading involves constituent coordination, that is, Did Bill drink [coffee] or [tea]?, and is hence non-problematic for te Velde's theory. The latter reading, however, requires gapping, that is, [Did Bill drink coffee] or [tea]? and is a problem since the gap precedes the one remnant tea.

To conclude, the most convincing aspect of te Velde's approach is its emphasis on feature matching. Examples such as 5a–c above demonstrate that parallelism between conjuncts need not be complete, but rather it can suffice if certain core features of the conjuncts are parallel. Syntactic, morphological, as well as semantic features can all play a role. In this regard, I think it is fair to say that the approach represents a move away from the strictly configurational account of coordination to a more functional one, whereby “functional” is understood mainly in terms of feature matching. The book should interest linguists investigating coordination in the Minimalist framework.

References

Chao Wynn. 1988. On ellipsis. New York: Garland Publishing
Chomsky Noam. 1995. The minimalist program. Cambridge, MA: MIT Press.
Chomsky Noam. 1998. Minimalist inquiries: The framework. MIT Occasional Papers in Linguistics 15. Cambridge, MA: MIT Press. (= Step by step, ed. by Roger Martin, David Michaels, and Juan Uriagereka, 2000, pp. 89– 155. Cambridge, MA: MIT Press.)
Chomsky Noam. 1999. Derivation by phase. MIT Occasional Papers in Linguistics 18. Cambridge, MA: MIT Press. (= Ken Hale. A life in language, ed. by M. Kenstowicz, 2001, 1– 52. Cambridge, MA: MIT Press.)
Hartmann Katharina. 2000. Right node, raising, and gapping: Interface conditions on prosodic deletion. Amsterdam: John Benjamins.
Hudson Richard. 1989. Gapping and grammatical relations. Linguistics 25. 57– 94.Google Scholar
Johannessen Janne. 1996. Partial agreement and coordination. Linguistic Inquiry 27. 661– 676.Google Scholar
Johannessen Janne. 1998. Coordination. Oxford: Oxford University Press.
Kayne Richard. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press.
Klein Wolfgang. 1993. Ellipse. Syntax—Ein internationales Handbuch zeitgenössischer Forschung (Syntax—An international handbook of contemporary research), ed. by Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld, and Theo Vennemann, 763– 799. Berlin: Walter de Gruyter.
Lechner Winfried. 2002. Reduced and phrasal comparatives. Natural Language and Linguistic Theory 20. 683– 735.Google Scholar
Lobeck Anne. 1995. Ellipsis: Functional heads, licensing, and identification. Oxford: Oxford University Press.
Maling Joan. 1972. On “gapping and the order of constituents.” Linguistic Inquiry 3. 101– 108.Google Scholar
Munn Alan. 1987. Coordinate structure and X-bar theory. McGill Working Papers in Linguistics 4. 121– 140.Google Scholar
Ross John. 1967. Constraints on variables in syntax. Doctoral dissertation, MIT.
Ross John. 1970. Gapping and the order of constituents. Recent developments in linguistics, ed. by Manfred Bierwisch and Karl Heidolf, 245– 259. New York: Mouton.
Ross John. 1976. Clausematiness. Formal semantics of natural language, ed. by Edward Keenan. Cambridge: Cambridge University Press, 422– 475.
Sag Ivan. 1976. Deletion and logical form. Doctoral dissertation, MIT. (Reproduced by the Indiana University Linguistics Club, 1977).
Wilder Chris. 1994. Coordination, ATB, and ellipsis. Minimalism and Kayne's asymmetry hypothesis (Groninger Arbeiten zur germanistischen Linguistik 37, ed. by Jan-Wouter Zwar), 291– 329. Groningen: Center for Language Cognition.
Wilder Chris. 1996. V2-Effekte: Wortstellungen und Ellipsen. Deutsch-typologisch, ed. by Ewald Lang and Gisela Zifonun, 142– 180. Berlin: Walter de Gruyter.
Wilder Chris. 1997. Some properties of ellipsis in coordination. Studies in universal grammar and typological variation, ed. by Artemis Elexiandou and Tracy A. Hall, 59– 106. Amsterdam: John Benjamins.
Williams Edwin. 1977. Discourse and logical form. Linguistic Inquiry 8. 101– 139.Google Scholar