Date: Tue, 01 Mar 2005 17:51:02 +0100 From: Lucia Grimaldi Subject: Ellipsis in Comparatives
AUTHOR: Lechner, Winfried TITLE: Ellipsis in Comparatives SERIES: Studies in Generative Grammar 72 PUBLISHER: Mouton de Gruyter YEAR: 2004
Lucia Grimaldi, Department of Romance Philology, Freie Universität Berlin, Germany.
The book focuses on two phenomena which supposedly occur in the formation of comparative constructions: comparative deletion (CD) and comparative ellipsis (CE). The author develops an analysis of comparatives which does not require construction-specific deletion processes. He claims that the phenomena generally attributed to CD and CE, stem from movement and from more general deletion processes. All mechanisms involved should be located in syntax. The book is divided into five sections: Introduction, Comparative Deletion, Comparative Ellipsis, Comparative Coordination, and Conclusion.
The first chapter (Introduction, 1-7) includes a short outline of the main claims and introduces the terms "comparative deletion" and "comparative ellipsis". Comparative deletion (CD) denotes a deletion process first described by Bresnan (1973). Assuming that in comparative constructions two degrees are involved, CD obligatorily triggers the deletion of the second degree/adjectival phrase:
(1) Bill is taller than John is [CD] (CD = tall)
Generally a distinction is made between CD and another deletion mechanism, so-called comparative ellipsis (CE) (cf. among others Pinkham 1982), by which phrasal comparatives (PC) are derived from full sentential counterparts. This ellipsis process is optional and applies in addition to CD, deleting further constituents under identity with the matrix clause:
(2) Bill is taller than John [CE] [CD] (CE = is, CD = tall)
In chapter 2 (Comparative Deletion, 9-88), the author tackles the CD phenomenon, discussing the precise nature of the process. In line with numerous approaches, he postulates a comparative structure in which two degree phrases (DegPs) are present, with Deg as a functional category taking AP as an argument (cf. Abney 1987). However, he does not locate the AP in the complement, but rather in the Specifier of the DegP. The than-XP is base-generated in the complement position of the first degree phrase. In the case of a synthetic comparative, the Deg-head, otherwise filled by "more", is phonetically empty. It is characterized by the feature [+comparative], which has to be checked by an AP in a Spec/Head configuration. The comparative structure postulated by the author is the following (p. 22):
(3) Bill is [DegP [AP taller] [Deg' [Deg [+comparative]][XP than John is [DegP [AP [CD]]]]]]
In addition, operator movement is involved in the comparative clause; this is a widely accepted assumption since Chomsky (1977). The empty operator (OP) is raised to SpecCP and binds the Deg-Trace in the complement position of the DegP (p. 38):
(4) Bill is taller than [CP OP(i) [IP John is [DegP [AP [CD]] [Deg' Deg t(i)]]]]
NP-comparatives with attributive modifier (in the following called: attributive comparatives) pose a classical problem in all analyses that assume the than-XP to be generated within the first DegP:
(5) I know younger authors than Peter knows
Under the common assumption that prenominal adjectives are left adjoined to NP, the comparative structure in (3) would yield the following wrong linearization:
(6) *I know younger than Peter knows authors
In order to avoid (6), the than-XP either has to be extraposed to the right, or else another analysis for prenominal attributives has to be chosen. The author takes the second option and, in line with Abney (1987), opts for a DP-structure in which the AP is selected by the D-head, and takes the NP as a complement (p. 29):
(7) [D [AP young [NP authors ]]]
Starting with (3) and (7) as basis the author proposes the following analysis for (5):
(8) I know [DP [DegP [AP younger [NP authors]][Deg' [Deg [+comparative]] [XP than Peter knows [CD]]]]]
To turn back to the question regarding the precise nature of the process described as CD, the author suggests that the AP within the than-XP should be raised to the matrix clause (10), a solution inspired by the Raising Analysis of the relative clause (cf. Bianchi 1999) shown in (9):
(9) [DP the [CP authors(i) [C [IP I know t(i)]]]]
(10) [DP [DegP [better authors](i) than [CP I know [DP [DegP t(i)]]]]] (t = CD = good authors)
CD would thus not be a deletion process but rather a special case of movement, AP-raising. The movement is motivated by the Deg-head of the matrix clause which, as mentioned above, is characterized by the feature [+comparative] that needs to be checked by AP-raising (p. 50). Due to the specific assumptions on AP-structure in attributive comparatives, the NP is subject to pied-piping, thus being also affected by movement. With reference to Poole (1996), the author claims that in AP-raising no chain formation is involved. As a consequence, both copies remain visible for semantics, whereas in the usual movement scenario only one copy from the movement chain is in fact interpreted semantically (p. 87).
With regards to the representational level of CD, the author argues against approaches which assume that the constituent affected by CD is only reconstructed on the level of semantic interpretation. He lists some phenomena in support of the evidence that the CD-constituent is visible on logical form (LF) and thus must be a result of syntactic derivation processes.
In Chapter 3 (Comparative Ellipsis, 89-183) the author presents data in support of the view that reduced comparatives such as (13) and reduced coordinations such as (14) obey similar restrictions, e.g. prohibiting the gapping of a verb with an embedded finite CP (p. 100) (in the following, deleted elements shall be placed in square brackets):
(13) *Lisa promised that her mother will visit Millhouse and Sally [promised that her mother will visit] Otto
(14) *More people promised that their friends will visit Millhouse than [CD] [promised that their friends will visit] Otto
The only difference is that in (13) the second subject is overt whereas it is eliminated by AP-raising in (14). As a consequence, the author opts for a unified analysis of reduced comparatives and reduced coordination. Both are assumed as being derived from full clausal constructions by conjunction reduction (CR) operations, especially gapping, right node raising and across the board movement. This claim, called Conjunction Reduction (CR)- Hypothesis (p. 114) implies that a construction-specific ellipsis rule such as comparative ellipsis can be dispensed with.
The author next turns to phrasal comparatives (PCs) which are defined as comparatives with one (non-verbal) remnant in the than-XP, as opposed to partially reduced comparatives (PRCs), which include more than one remnant in the than-XP (p.89). He presents a large variety of English and German data to support evidence that PCs and PRCs obey the same internal and external restrictions. Since PRCs are uncontroversially analyzed as derived from sentential comparatives, consequently PCs too should be treated as derived structures. This conclusion is clearly formulated in the PC-Hypothesis: "All PCs without explicit standards derive from a clausal source" (p. 93). It follows that all comparatives should be analyzed in the same fashion, i.e. in terms of AP-raising. Additionally, PCs and PRCs are subject to ellipsis operations, previously identified as CR operations.
A further assumption regarding the structure of PCs and PRCs should be mentioned here: the author posits a so-called comparative coordination, according to which comparatives are optionally parsed as coordinate structures (p. 110). This property only affects reduced comparatives and is linked to the application of CR operations. It is also limited to the syntactic derivation: it is not reflected in the semantics of the comparative clause which is assumed to be semantically subordinated (p. 224). Comparative coordination explains data such as (15) in which the subjects of the matrix and the than-clause have different agreement features (p. 162):
(15) weil wir mehr Buecher gekauft haben als Peter [CD] gekauft hatbecause we more books bought have than Peter bought has
While the application of gapping is possible, giving (15)a, right node raising does not seem to be acceptable, since it would lead to the ungrammatical (15)b (deleted elements in square brackets):
(15)a. weil wir mehr Buecher gekauft haben als Peter [[CD] gekauft hat] because we more books bought have than Peter bought has
(15)b. *weil wir mehr Buecher [gekauft haben] als Peter [CD] gekauft hatbecause we more books bought have than Peter bought has
It appears that in comparatives where subjects with different personal features are involved, the remaining verb has to agree with the subject of the matrix clause. The author proposes a solution to this problem by claiming that the height of coordination can differ in reduced comparatives. Assuming a finer grained IP-structure, the coordination may apply at VP/vP, AspP, TP or AgrSP (p. 166). In cases such as (15) the coordination applies at an early stage in the derivation, at the VP-level. This means that there is only one TP available, which contains an auxiliary and the subject of the matrix clause. Therefore the auxiliary obligatorily agrees with the matrix subject. The subject of the than-clause remains in situ and can't have influence on verbal morphology (the case-filter is satisfied by default case). Thus (15)b is ruled out.
The fourth chapter (Comparative Coordination, 185-227) takes a closer look at the assumed coordinate structure for comparatives. The author claims that "Initially, the than-XP is base-generated in the complement position of the degree head, resulting in a subordinate parse" (p. 190). In cases, in which the than-XP is on the right periphery of the whole construction, a particular extraposition operation, so-called than-XP raising (TR), is assumed to have applied. This operation is optional in principle, but becomes obligatory, in the case of a CR operation requiring *Embedding, e.g. gapping. TR, which is only available in comparatives, and not in coordination, explains why comparatives such as (16) can have wide and narrow ellipsis reading in a and b respectively, while coordination cannot, as is shown in (17) (p. 186s.):
(16) John wants to write more plays than Sam [e] [CD] a. [e = wants to write] b. [e = writes/wrote]
(17) John wanted to write plays and Sam [e] poems a. [e = wanted to write] b. *[e = wrote/writes]
The claim is that, when two IPs are involved in the matrix clause, the than- XP is attached at different levels of the structure, i.e. at the higher IP in (16)a and at the lower IP in (16)b. This option is not available in coordination where no TR is involved. In the last part of chapter four, the structure of comparative coordination is modified (p. 225): instead of a three-branch- representation, a binary structure is favoured, this fully in line with "asymmetric" analyses of coordination.
Chapter 5 (Conclusion, p. 229-233) contains a brief summary of the main claims and cites some still unresolved problems. It is followed by the Notes, the Bibliography as well as a Subject and an Author Index.
The book is a synthesis of Lechner (1999) and Lechner (2001) supplemented with some recent considerations. It provides the reader with an impressive variety of interesting data concerning sentential and phrasal comparatives in German and English and also presents some appealing ideas, such as taking up the Raising Analysis of relative clauses and applying it to comparatives (cf. Donati 1997, 2000 for a different elaboration of the same concept). The resolution of comparative deletion as syntactic movement, i.e. AP-raising, solves some problems related with semantic and phonological ellipsis accounts of CD. In contrast to semantic approaches, it does not assume that the CD-site has either empty or no syntactic structure at all, and thus avoids problems related with scope and binding properties of the CD-site. Since only the AP(+NP) is targeted by AP- raising and by subsequent phonological deletion of the lower copy, an identity condition between the two degree heads, generally required by phonological deletion accounts, becomes superfluous.
The PC-Hypothesis, which assumes all comparative complements to be derived from full thematic clauses, has the theoretical advantage of being semantically transparent, whereas the so-called direct analyses, which analyze the comparative complement as a simple prepositional phrase, have to explain how the missing structure can be reconstructed semantically (cf. Kennedy 1999: p.151). Additionally, the PC-Hypothesis reduces the number of phenomena to be dealt with, a desirable consequence from the point of view of economy. The author presents interesting arguments on how to solve classical problems of a derived analysis of phrasal comparatives. These are shown by the following data, in which a clausal basis for the derivation does not seem to be available:
(18) John is older than me [*am] (p. 179)
(19) John(i) couldn't possibly be taller than himself(i) [*is] (p. 180)
(20) She ran faster than the world record [*ran] (p. 182)
The author remarks that the accusative case in (18) is not an exclusive property of PCs, but can also be found in reduced coordination, an additional argument for the CR-Hypothesis, at least for English data:
(21) John is eager to see the movies, and me too (p. 180)
The ungrammaticality of (19) is explained by assuming that a small clause (SC) taking the remnant as a subject is the basis of the derivation. According to this analysis the structure of the than-XP is [-finite] and no binding principle is violated. The same small clause analysis is assumed for (20). But this case is somewhat problematic, since the assumption of an underlying structure implies that CD has occurred. This means that the derivational basis of (20) would be something like (22), which seems problematic for a semantic interpretation:
(22) #She ran faster than [SC the world record [CD fast]]
Furthermore, there is good reason of being critical towards the strong distinction made between the PC-Hypothesis and direct analyses. While the author uses the number of remnant constituents as a discerning criterion between PCs (one remnant) and PRCs (more than one) (p. 89), this is not necessarily the defining criterion for direct analyses. In these theories only a subset of the constructions defined as PCs by the author are considered as base generated (cf. among others Hankamer 1973, Napoli 1983). These include the problematic cases in (18)-(20) and comparatives in which the remnant is compared to the subject of the matrix clause (a more precise definition is not possible in this context). In this more narrow sense, base generated PCs are distinguished from other non-sentential comparatives. Note that this distinction is realized by different particles in some languages (e.g. in Italian). Taking into account this reduced applicability of direct analyses, and considering that the PC-Hypothesis cannot resolve all problematic cases (cf. fn. 145, p. 258), and excludes "subcomparatives, comparatives with explicit standards and small clause comparatives" (cf. fn. 177, p. 261), the distinction between the PC-Hypothesis and direct analyses doesn't seem so clear cut anymore.
A similar remark can be made with regards to the strong CR-Hypothesis: The claim that no construction specific ellipsis operation (such as comparative ellipsis) is required is well motivated. However, the stronger claim that all reduced comparatives can be explained by CR operations seems too radical, a fact which is partly recognized by the author (cf. fn. 105, p. 250) and partly shown by examples with IP- and CP-ellipsis, which violate at least the principle of Isomorphism (deleted elements in square brackets):
(23) John is taller than I expected [IP John to be [CD]]
(24) Jan ist groesser als ich dachte [CP dass Jan [CD] ist] John is taller than I thought that John is
An advantage of the present proposal consists in the assumed DegP- structure. On the one hand, it yields the right linearization without obligatory extraposition of the than-XP (cf. Bresnan 1973). Furthermore, it avoids another well known problem for the analysis of comparatives, namely Left Branch Extraction (cf. Corver 1990). However, there is a problem related to the structure of attributive comparatives. As the author points out, the [AP[NP]]-nexus shows conflicting categorial properties: in order to fulfil the requirements of AP-raising it must be detected as an AP. However, being a verbal argument headed by a DP, it requires the external structure of an NP. As a solution to this conflict, the author introduces a functional category (FP) which takes the NP as a complement and the AP as a specifier. While only the AP is targeted by AP-raising, it is the whole FP, including the NP, which is pied-piped. According to this analysis the category affected by AP-raising is no longer an AP but an FP with the categorial status of NP (p. 45s.). It is not clear how the Spec/Head- configuration required for the checking of the [+comparative]-feature in Deg and for the phonological realization of comparative morphology (p. 23) may be established in a structure of this kind.
One of the merits of this book lies in the idea of treating comparatives as hybrid structures located between subordination and coordination. It allows for an explanation of similarities with relatives (cf. Chomsky 1977; Donati 1997, 2000) as well as with coordinate structures (cf. Napoli 1983). Nonetheless, the argument concerning comparative coordination is somewhat circular. On the one hand the author stipulates a "coordinate structure required in order to provide the syntactic basis for the application of CR" (p. 232s.). In this case the coordinate structure seems to be the prerequisite for CR operations. But in chapter 3, CR operations, specifically across the board (ATB) movement, are taken to be the condition required for the coordinate structure, and the causal chain seems to be inverted: "Whenever comparative formation implicates an ATB movement process such as ATB V2, the structure has to be parsed as a comparative coordination" and, even more explicitly: "... ATB movement triggers a comparative coordination...." (p. 172).
The sequence of arguments is generally clear, with one exception: the structure of comparative coordination, including TR and the asymmetric analysis of the coordinate structure, is introduced only in chapter 4; yet, it would have been helpful in chapter 3 for a better comprehension of intraposed and extraposed PCs (3.3.4) and of the restrictions on comparative coordination (3.4).
Although one of the declared goals of the analysis was to avoid structure dependent mechanisms such as comparative deletion and comparative ellipsis, the author is forced to introduce some "unorthodox" mechanisms, such as movement without chain, switch between subordinate and coordinate parse, and obligatory TR. There seems to be no escape: one either makes a finer grained distinction of the phenomena at stake - a solution partly chosen by the author by excluding subcomparatives from his account -, or alternatively one has to assume construction-specific mechanisms.
In conclusion, despite the problems discussed so far, the book is an important contribution to the comprehension of these complex constructions and should be taken into account for further research.
Abney, Steven (1987): The English Noun Phrase in its Sentential Aspect. Ph.D. diss., Department of Linguistics and Philosophy, MIT.
Bianchi, Valentina (1999): Consequences of Antisymmetry: Headed Relative Clauses. Berlin: Mouton de Gruyter.
Bresnan, Joan (1973): "Syntax of the comparative clause construction in English", Linguistic Inquiry 4, 275-343.
Chomsky, Noam (1977): "On Wh-movement". In: Peter Culicover, Tom Wasow and Adrian Akmajian (eds.): Formal Syntax, 71-132. New York: Academic Press.
Corver, Norbert (1990): The Syntax of Left Branch Extraction. Ph.D. diss., Department of Linguistics, University of Tilburg.
Donati, Caterina (1997): "Comparative clauses as free relatives: A raising analysis", Probus 9, 145-166.
Donati, Caterina (2000): La sintassi della comparazione. Padova: Unipress.
Hankamer, Jorge (1973): "Why There are two than's in English", CLS 9, 179- 191.
Kennedy, Christopher (1999): Projecting the adjective. The syntax and semantics of gradability and comparison. New York/London: Garland.
Lechner, Winfried (1999): Comparatives and DP-structure. Ph.D. diss., University of Massachusetts.
Lechner, Winfried (2001): "Reduced and Phrasal Comparatives", Natural Language and Linguistic Theory 19, 683-735.
Napoli, Donna Jo (1983): "Comparative Ellipsis: A Phrase Structure Analysis", Linguistic Inquiry 14, 675-694.
Pinkham, Jessie (1982): The Formation of Comparative Clauses in French and English. Ph.D. diss., Cambridge MA: Harvard University.
Poole, Geoffrey (1996): "Optional Movement in the Minimalist Program", in: Werner Abraham, Samuel Epstein, Höskuldur Thráinsson and Jan-Wouter Zwart (eds.): Minimal Ideas. Amsterdam/Philadelphia: John Benjamins, 199- 216.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Lucia Grimaldi is a teaching and research assistant for Italian Linguistics at the Freie Universität Berlin. Her field of research (Ph.D. dissertation in progress) covers the syntax of comparative constructions and related phenomena (wh-movement, DP structure, focus, syntax and semantics of degree). She mainly works on Italian and other Romance languages.