All ideas expressed in Thor's Articles and The Passionate Skeptic are entirely those of the author, who has no aim to influence, proselytize or persuade others to a point of view. He is pleased if his writing generates reflection in readers, either for or against the sentiment of the argument.
Australian Journal of Linguistics 7 (1987), 25-42
IN THE COMPLEMENTS OF RAISING CONSTRUCTIONS
Thor May 1987
Abstract : The analysis considers the manner in which a class of matrix verbs, the so-called raising verbs, have been fitted into some generative linguistic models. Taking as a cue the difficulty posed for these models by sentences of the kind, *Linda believes Gary to murder David, the analysis proceeds beyond existing criteria for "raising" to the notion of Relative Tense.
It is found that Relative Tense has a direct bearing on the infinitival complements permitted by raising-to-object verbs and some raising-to-subject verbs. The relevant constraints are formulated for incorporation into Bresnan's Lexical Functional Grammar as the Independent XCOMP Singularity Condition. The IXCSC may be recorded for convenience in the functional structure of LFG as a complex feature. When IXCSC carries a positive marking the functional structure of a sentence, that sentence may only be interpreted if ASPECT (AUX) is also marked as positive.
Note: The author has now moved on from Joan Bresnan's Lexical Functional approach. This is relevant to interpreting the paper as it is presented here, although the actual issues with which the content deals are, I hope, still usefully clarified by the argumentation.
Table of Contents
introduction // 1. existing analyses of raising // 2. a semantic perspective on raising // 3. relative tense // 4. stative & process complements // 5. resultative complements // 6. the verb ":expect" // 7. marking temporal conditions in the grammar // 8. definition of terms // relative tense // thematic feature // independent XCOM singularity condition // application of IXCSC // lexical feature sets: c-structure // 9. postscript: raising-to-subject verbs // 10. conclusion // references // end notes //
This paper explores the ungrammaticality of sentences of the kind:
(1) *Linda believes Gary to murder David.
Gregory Lee (1970:9) recognised a class of such sentences, but was content to characterise matrix verbs like believe as creating "nonagentive contexts". As it happens, the infinitival complement of believe can contain agentive-type verbs subject to special conditions. Moreover, a careful look at the issue suggests generalisations of wider significance to the grammar. The explanation for the ungrammaticality of sentence (1) turns out to be a hitherto unrecognised constraint on the complements of raising verbs. (Throughout this paper1 I will use the term "raising verb" purely as an identifying label without any suggestion of favouring the transformational analysis which gave rise to the description.)
1. EXISTING ANALYSES OF RAISING
"Raising" refers to the constructions depicted by sentences such as (2) and (3): (2) The train seems to be a diesel. (3) The commander imagined his charm to be irresistible. In sentence (2) the train appears to be assigned the semantic and grammatical roles of logical and syntactic Subject by both seem and the infinitival phrase to be a diesel. In sentence (3) his charm is assigned the role of Object by the matrix verb and Subject by the infinitival phrase. The process in both sentences has been considered sufficiently similar to encourage a unified explanation in models of generative grammar, although in this paper important differences will be noted. At this point two principles in the theory of grammatical relations become significant. Firstly, the role of Subject as a grammatical relation (whether "logically" or syntactically conceived) is generally felt to be assigned by a predicate rather than a verb alone. Hence, for example, the traditional configuration where VP is equated with a predicate:
Secondly, it has been argued (Marantz 1981:62; also Chomsky 1981) that for an NP to appear in surface structure it has to bear a grammatical relation with respect to some lexical item (as opposed to a phrase). In the case of Subject, Chomsky and Marantz at least take that "lexical item" to be INFL (the inflection of the matrix verb which signals tense/agreement). These two principles appear to be in conflict in "raised" NPs, and have led to competing explanations.
A look at some of the semantic factors inherent in raising verbs may lead us to a better understanding of the syntax. It can be no accident that verbs classified on ostensibly syntactic criteria almost always fall into recognisable semantic sets. One interesting semantic fact is that in both kinds of raising verbs the formal controller of the action or state in the complement remains nonthematic; (see Bresnan's argument structures above). What this seems to mean ultimately is that the controller is incidental to the implicit semantic role of the matrix verb. Consider those elements of semantic role which seem common to raising-to-object verbs: 1. All imply a certain loose presupposition ("expectation" for the purists) in their complements. The presupposition is that the complement expresses some knowledge or a state of affairs which the [+human] Subject has accepted as true; and 2. The point of using a raising-to-object verb is to mark a link between the matrix Subject as sentient commentator (a kind of speech-act role) and the complement as a proposition about the world (...] rather than the complement as a unsourced description of reality (in which case its controller would be thematic4). Readers may care to test the following list of such verbs for themselves against this suggestion:
(15) Malinda believeS Izzy to BE wringING the chook's neck at this moment.
It is now necessary to digress a little to consider the special case of the verb expect. Expect has often been analysed as a raising-verb (e.g. Bresnan 1982:227; Chomsky 1981:99). However, it has many properties in common with that class of so-called Equi-verbs including prefer, want etc., and other properties which make it unique in the class of Raising verbs including believe, know etc. Firstly, expect may take a PRO8 in the Subject position of its complement: (19) He expects PRO to win. (20) He prefers PRO to win. (21) *He believes PRO to win. Secondly, Bresnan (1982:72; 229) suggests that a special feature of raising-verbs is their capacity to take there (be) as a kind of pleonastic Subject to the complement. Actually, the class of verbs including prefer will generally accept this form as well9: (22) He believed there to have been dark deeds done. (23) He expected there to be dark deeds done. (24) He preferred there to be absolute silence. Sentence (24) differs from (22) and (23) of course in possessing an adjectival complement and in being unable to passivize: (25) There were believed to have been dark deeds done. (26) *There was preferred to be absolute silence. However, expect, while passivizing freely, also accepts adjectival complements of the kind in sentence (24) and, as we would expect from the behaviour noted in sentence (1), is able to delete the constituents, "[...] there to be", in a way that believe cannot: (27) *He believed absolute silence. (28) He expected absolute silence. (29) He preferred absolute silence.
Neither the G & B nor the LFG models provide an obvious way to mark temporal conditions of the kind discussed in this paper (although, as usual, Chomsky seems to leave room for various solutions10). Since the phenomenon applies to a whole class of verbs the attempt seems worthwhile. It has been suggested that the dependency between matrix and lower verbs in particular infinitival sentences finds expression in constituent structure but is controlled by functional considerations. That is, from a constituent viewpoint, an important element in the lower construction, AUX, is not free within what would traditionally be taken as its governing category, but rather is controlled in some sense by the verb of a higher sentence. Nominal dependencies (anaphora, PRO, etc.) are well explored in the literature, but the temporal dependencies in question are essentially verbal. It would seem that the category INFL in G & B, which has a projection of S and governs tense/subject agreement, is concerned with "absolute tense" (c.f. Comrie), and is quite independent of "relative tense". Chomsky's deletion of S and the government of NP2 by the matrix verb therefore contributes nothing to an understanding of the latter's relationship with lower verbal constituents.
The Functional component of LFG seems to offer a better prospect of capturing the necessary dependency conditions than G & B, and this possibility is developed below. But first it will be necessary to define a precise class of verbs since the label "raising-verb" isn't entirely sufficient.
Let relative tense define the relationship between the temporal incidence of a matrix verb and the temporal incidence of any complement which is predicated of that verb. Where RT = b ('before'), then the incidence of the matrix verb anticipates the incidence of its complement; (e.g. hope to : RTb). Where RT = a ('after'), then the incidence of the matrix verb is subsequent to that of its complement; (e.g. turn out to : RTa). Where RT = t ('together'), then the incidence of the matrix verb and its complement are simultaneous; (e.g. watch : RTt). Where RT = i ('independent'), then the incidence of the matrix verb and the incidence of its complement are pragmatically independent. That is, no inference of influence or causation may hold between the matrix and its complement; (e.g. believe: RTi). Many verbs that are RTi will also be typically RTa.
The other essential properties of the class of matrix raising verbs, and of the subordinate verbs to which they become related in various ways, turn on what are often broadly termed "thematic relations" (Agent, Patient, etc.). I wish to express a finer discrimination than that available from terms like Agent, so a new class of thematic feature sets (attached to verbs rather than nouns) will be recognised here. The following thematic features will be needed to describe the behaviour of verbs relative to their referents: (a)[+ Stative]. This feature is normally subcategorized as an inherent property of a verb, although active verbs as participles can be interpreted in a stative manner. It is cautioned that "stative", when applied to verbs like FIND or TURN OUT TO is a little confusing since an inceptive sense also seems to be implied ([...] "the beginning of the state" as it were). However [+ Stative] conveys the required meaning better than [+ Durative], which includes active processes as well. (b)[+ Resultative]. This is also an inherent feature in some verbs, and indicates that the verb describes an outcome which can be objectively verified; (contrast especially with 'psych' verbs). (c)[+ Sgr]. This feature is really contingent upon the particular situation being described and indicates the singular incidence of a verb; (i.e. a single act or event as opposed to a process or state). Some verbs like murder seem to be inherently [+ Sgr] whereas others like hit would mark the feature with a particular morphological environment. (+Sgr] would normally co-occur with [+Resultative]; (take the small 'r', Sgr, as a mnemonic to suggest this). We distinguish [Sgr) of course from the Sg./P1. marking on nouns.
The lexical feature sets, c-structure (constituent structure) and f-structure (functional structure) with its singularity violation would appear as follows for sentence (1):
Lexical feature sets:c-structure:
Note: IXCSC has been recorded for convenient reference in f-structure as a complex feature. Where IXCSC carries a positive marking in f-structure, ASPECT must also carry a positive marking if the sentence is to be interpreted.
So-called raising-to-subject verbs are less consistent in their properties than raising-to-object verbs, but a number of them also exclude infinitival complements without temporal marking. Included in the general class are seem, appear, happen, be likely, be certain, grow to, come to, turn out to, prove to and strike ... as.. Of these, the first four can take complements without temporal marking while grow to and come to cannot take a complement implying a singular event under any circumstances. The complement of strike ...as11 should also have a durative or stative sense, although it is a bit fuzzy about this for some speakers; (see, for example, sentence (39)). Thus: (32) The client seemed to agree. (33) The client appeared to agree. (34) The client happened to agree. (35) The client was likely to agree. (36) The client was certain to agree. (37) The client *grew to agree/*has grown to agree/grew to accept (STATIVE) the situation. (38) He struck me as *daring/having dared much to win little. (39)?? He struck me as having beaten the hell out of the opposition. (40) Marco proved to *succeed/have succeeded in the quest. (41) Gondor turned out to *cause/have caused the trouble. The notion of COMMENTATOR does not assist in capturing a disjunction between the temporality of matrix and subordinate verbs here. Only strike ...as contains an obligatory COMMENTATOR (as OBJ)while seem and appear take optional OBL COMMENTATORs. The speech function reference of all the other raising-to-subject verbs is arbitrary. It may be that because the senses of seem and appear are normally contemporaneous (RTt) with the action of the subordinate verb, they establish sufficient temporal context for the whole complex sentence. Formally, the feature (RTt) excludes them from the XCOMP Singularity Condition. Note also that these two verbs may be paraphrased as adverbial properties of the subordinate verb (seemingly, apparently). Happen to also has a contemporaneous sense, although its adverbial paraphrase would have to be something less transparent like incidentally. Be likely and be certain, on the other hand, are predictive of the complement (RTb), and an event which has not yet happened may not require aspectual contextualization. In fact we know this to be the case in English since there is no future tense, and a morphologically unmarked sentence such as (42) (42) I leave on Thursday. would normally take a future interpretation.
I have already excluded grow to, come to and strike ... as from the class of verbs allowing complements to express singular acts of results. This leaves prove to and turn out to. These verbs are generally reflective. That is, they may recall the occurrence of a prior state, result or action in their complements, but without imposing any pragmatic temporal dependencies on the lower verb. Prove to and turn out to therefore carry the features [RTa/i]. They share these essential properties with those uses of raising-to-object verbs which entail complements of similarly independent reference, and through the IXCSC similarly require aspectual marking in the complement.
It has been found that Relative Tense has a direct bearing on the infinitival complements permitted by raising-to-object verbs and some raising-to-subject verbs. The relevant constraints have been formulated for incorporation into Bresnan's Lexical Functional Grammar as the Independent XCOMP Singularity Condition. The IXCSC may be recorded for convenience in the functional structure of LFG as a complex feature. When IXCSC carries a positive marking the functional structure of a sentence, that sentence may only be interpreted if ASPECT (AUX) is also marked as positive.
Akmajian, A. & Heny F.1975. An Introduction to the Principles of Transformational Syntax. Cambridge, Mass.: MIT Press.
Bresnan, J. (ed.) 1982. The Mental Representation of Grammatical Relations. Cambridge, Mass.: MIT Press.
Chomsky N.1981. Lectures on Government and Binding. Dorecht: Foris.
Comrie, B.1976. Aspect. Cambridge: Cambridge University Press.
Fillmore, C. 1986. "The Case for Case" In Bach, E. & R. Harms, (eds.), Linguistic Theory. New York: Holt Rinehart & Winston.
Gruber, J.1976. Lexical Structures in Syntax and Semantics. Amsterdam: North-Holland.
Heinamaki O. 1978. Semantics of English Temporal Connectives. Indiana University Linguistics Club Mimeo.
Jackendoff, R.1972. Semantic Interpretation in Generative Grammar. Cambridge, Mass.: MIT Press.
Lee, Gregory 1970. Subjects and Agents. PhD Dissertation, Ohio State University.
Marantz, A.1981. On the Nature of Grammatical Relations, PhD Dissertation, MIT.
Postal, P.1974. On Raising. Cambridge, Mass.: MIT Press.
1 Thanks to Ray Cattell, Peter Peterson and George Horn, University of Newcastle, N.S.W., for comments on earlier drafts of this paper. [return]
2 Chomsky (1981:36) uses 'argument' in a slightly different way to include non-thematic elements and even trace. [return]
3 One of the most comprehensive analyses of raising is to be found in Postal (1974). However, it should be noted that perceptions of the problem have changed somewhat. Whereas "raising" might have earlier been seen as a "rule of the grammar", it is more likely now to be decomposed into sub-systems of principles and constraints. This is a general trend. Chomsky (1981:7) describes some other concepts analogously: "[...] The notions of 'passive' 'relativization' etc. can be reconstructed as processes of a more general nature, with a functional role in the grammar, but they are not 'rules of grammar ." See also G & B where Chomsky suggests that a 'rule of raising would violate the principle of 0-marking. It would also be contrary to the Projection Principle. [return]
4 An example of a sentence with a thematic controller in the complement would be (i): (i) Mark forced the dog to wag its tail. The relevant points to note are that:
5 'Identity' is used here in the formal sense employed by Bresnan in Lexical Functional Grammar (1982:376). In this usage a Subject NP and its coreferent reflexive, for example, do not share absolute identity. [return]
6 Heinamaki (1978:3' after McTaggart, from Prior 1967) has also recognized a notion of relative tense which he calls "B-Series". In B-Series time each position is earlier than some and later than others and this relationship is "permanent". In the A-series time of PAST PRESENT and FUTURE things change, so that something which was FUTURE becomes PRESENT and then PAST. Heinamaki believes that "there doesn't seem to be any instances of pure B-type expression in natural language". He turns to temporal adjuncts for evidence of B-series. However, in this article I am concerned with ' B-series" phenomena that are inherent properties of matrix verbs. [return]
7 Chomsky (1981:55) argues that such participles are not true adjectives but rather neutralised verb-adjectives with the feature structure [+V] [return]
8 Trace in the Government & Binding model (Chomsky,1981:56,156). [return]
9 Chomsky (1981:26) uses this fact in arguing against Bresnan's analysis of infinitival complements as VP = COMP VP1 [return]
10 See for example the discussion in G & B (1981:54) where it is suggested that if COMP be empty it might be possible to permit direct relations between matrix verbs and the INFL of the embedded clause. [return]
11 strike ...as possesses a number of odd characteristics. See Postal (1974) for a discussion of some of these. [return]
Verbs of Result in the Complements of Raising Constructions (c) Thor May 1987
Department of Literature & Language, University of the South Pacific P.O. Box 1168, Suva, FIJI
|©2005 Thor May|