TIPPING THE SCALES: THE SYNTAX OF SCALARITY IN THE COMPLEMENT OF SEEM

Transcription

1 Syntax 5:3, December 2002, TIPPING THE SCALES: THE SYNTAX OF SCALARITY IN THE COMPLEMENT OF SEEM Ora Matushansky Abstract. This paper argues for a syntactic and semantic distinction between the verb seem that takes propositional complements (i.e., CP and IP) and the verb seem that takes nonpropositional complements. The latter takes a smaller sized complement (in terms of the presence of functional structure), has a perceptual rather than epistemic interpretation, and imposes a scalarity-related restriction on its complement, which will be formalized as a selectional requirement of a DegP complement. The notion of scalarity, heretofore applicable only to adjectives, is extended to PPs, such as out of her mind, and nouns, such as fool. DegP is projected either if a predicate is scalar (has a degree argument slot) or to function as a landing site for QR of degree, which will be shown to function as the licensing mechanism for complements of seem. 1. Introduction This paper argues for a syntactic and semantic distinction between the verb seem that takes an IP or CP complement (the epistemic seem) and the one that takes a small-clause complement (the perception seem). The perception seem differs from the epistemic seem in a number of ways, including a scalaritybased restriction on its complement. Depending on whether the complement of seem is large (a CP or an IP) or small (a small clause with an AP, NP, or PP predicate), its semantics changes. The small complement of seem is subject to a scalarity-related restriction, which does not seem to play a role with the large complement of seem. I argue that the complement must contain a DegP and formalize it as an uninterpretable [degree] feature on seem. As a result, I am able to offer an explanation for such syntactic effects as the obligatory interpretation of the subject of the small-clause seem in its surface position and the locality constraint on whether an AP in a DP small clause can license the small clause as a legitimate complement of seem. * The first footnote people are, in counter-alphabetical order: Tim Stowell, Philippe Schlenker, David Pesetsky, Tania Ionin, Sabine Iatridou, Irene Heim, Martin Hackl, Danny Fox, Kai von Fintel, Michel DeGraff, Noam Chomsky, Daniel Büring, and Rajesh Bhatt. I m also very grateful to the two anonymous reviewers and the very nonanonymous Tim Stowell (again!) for their extremely helpful comments. Had it not been for the request to reduce the number of footnotes, each page would have borne a reference to these people. Thanks also are due to the audiences at the UPRESA seminar (Paris, January 2000), the Conference on (Preferably) Non-lexical Semantics (Université de Paris 7, May 2000), UCLA syntax/semantics seminar (February 2001), and WCCFL 20 (February 2001), who provided many comments on different versions of this work. Everything is subject to future research.. Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA

2 220 Ora Matushansky The paper is organized as follows. I first address the issues of the complements of seem. I show that when the complement of seem is an IP or a CP, seem is interpreted as an epistemic modal, as opposed to the perception interpretation it gets when its complement is a small clause. I then examine the restrictions on the small-clause complement of seem and show that scalar APs, DPs, and PPs can always appear there. The first indications that just scalarity is not enough come from section 4, where I give the semantic definition of scalarity and show how it is encoded in syntax. Section 5 shows that scalarity does not permit us to describe the conditions on the complement of seem because it incorrectly excludes all cases of degree quantification and modification, such as better. This argument is further supported by the fact that a structural ap, np, or pp layer separates the predicate from seem. I argue that the constraint on the complement of seem must be formalized as a [degree] feature on the verb, forcing QR of a DegP from its complement. In section 6, I demonstrate that modification changes the ability of a DP to appear in the complement of seem, and I argue that the ability of an adjective to license a DP in the complement of seem depends on the adjective s type and its structural position within the DP. Section 7 is the conclusion. 2. Two Seems Better Than One? Leaving aside clauses (CPs), the seem + complement construction has very different properties depending on whether the complement is an IP or a small clause. I show that the category of the complement of seem influences both its interpretation and its syntactic properties. 2.1 Complements In this subsection I argue for two verbs seem: the epistemic one that takes a large (IP or CP) complement and the perception one that takes a small complement, at the core of which is an NP, AP, or PP. Following Stowell (1981, 1983), I take this small complement to be a small clause. VPs, including gerunds and passive participles, are not possible complements of seem Possible complements Languages differ with respect to possible complements of their equivalents of seem. In English, seem can appear with an (extended) AP, PP, or NP, as in (1), with an IP, as in (2), or with two types of CP, as in (3). (1) a. Yolanda seems stupid. b. Belinda seems out of her mind. c. Orlando seems an idiot.

3 Tipping the Scales 221 (2) a. Miranda seems to be stupid/an idiot/out of her mind. b. Amanda seems to believe that money grows on trees. (3) a. Clarinda seems like she is in a bad mood. b. It seems that Fernando is sick. Depending on the complement, seem shows different syntactic and semantic properties Like-complements 1 Despite apparent similarity between (3a) and (3b), the two structures are strikingly different. When the complement of seem is a like-cp, as in (3a), seem can assign a h-role to its subject but does not have to, as shown by (4a). On the other hand, with a that-cp complement, seem cannot assign a h-role to its subject, as shown by (4b). (4) a. It seems like Clarinda is in a bad mood. b. *Fernando seems that he is sick. The construction in (3a) and (4a) is not a raising construction, given that the matrix subject can be connected with nouns other than the subject of the embedded clause, as shown by (5), though the presence of a pronoun seems near-obligatory. (5) a. She seems like her arms go on forever. b.?/?? Hermione seems like her teacher/mr. Snape is giving her a hard time. c.??/* Hermione seems like Crookshanks is shedding. Importantly, to the extent like-cases are grammatical, they require direct visual perception (i.e., seem is roughly synonymous with look). This is not true for that-cases, which are interpreted as epistemic judgments (i.e., It seems to me that P means I conclude that P holds). The complement of like can be an IP or a DP, which places it in the same category as than or for. Its behavior suggests that it is not impossible for seem to assign a h-role to its subject. On the other hand, the epistemic seem does not assign a h-role to its subject and appears only with temporal complements (an IP or a that-cp). 1 CPs with as if and as though behave similarly, although their distribution is somewhat more restricted. For a discussion of seem like/as if/as though, see Schwarzschild 1986.

4 222 Ora Matushansky Reconstruction Stowell (1981, 1983) proposed that nominal, adjectival, and prepositional complements to seem are small clauses containing a subject position. Williams (1983), in his argument against small clauses, notes that reconstruction to the subject position of a small clause is impossible. Whereas (6a) is ambiguous, depending on whether the quantifier takes wide scope (in its surface position and above seem) or narrow scope (below seem and presumably in the subject s h-position), (6b) is not. (6) a. A musketeer i seems [t i to be [t i sick]] b. A musketeer i seems [t i sick] Whereas to utter (6b) one needs to have a specific musketeer in mind (e.g., D Artagnan), this is not true for (6b) or, to put it differently, (7) is a possible interpretation of (6a) but not (6b). (7) It seems that a musketeer is sick. One could argue that the effect may not be due to the properties of the matrix verb, since it is not restricted to seem. Non-IP complements of consider also resist reconstruction: (8) a. At least one woman is considered to be sick. b. At least one woman is considered sick. Whereas in (8a) the subject can scope either above or below the intensional verb, in (8b) it has wide scope only. The effect is therefore triggered not by the verb seem but by the fact that its complement is a small clause, as opposed to an extended VP (an IP or a CP). On the other hand, as Noam Chomsky (p.c.) notes, if the same copular infinitive as in (8a) appears with believe, as in (9), reconstruction is also impossible. (9) At least one woman is believed to be sick. If true, this suggests that reconstruction can be constrained by the choice of the main-clause verb. Further arguments for this proposal are found in the distribution of there There-insertion Another difference between the two types of complements that seem can appear with is the possibility of there-insertion, which is related to the basegeneration site of the surface subject. In (10a) the subject is raised from the

5 AP-internal specifier to check the EPP feature of the infinitival T and then to the specifier of seem. Alternatively, as in (10b), the expletive there can be inserted in the specifier of the infinitival IP and then raised to the main clause. (10) a. A squire seems [ IP t i to [ VP be [ SC t i sick]]] b. There i seems [ IP t i to [ VP be [ SC a squire sick]]] c. *There seems a squire sick. d. There is a squire sick (next door). Tipping the Scales 223 Williams (1983) notes that no such option is available when the complement of seem is not an infinitive. We can now link the failure of matrix there-insertion with small-clause complements to the inability of the subject of a small clause to reconstruct to the base position, noted in section If for some reason the subject cannot be interpreted in the base position, both the failure to reconstruct and the failure of there-insertion are due to the same cause. To answer this question we should now ask why (10c) contrasts with (10d). In (10d) the matrix copula be takes a small-clause complement (see Stowell 1981). If the subject position of a small clause is not a possible interpretation position, why is there-insertion possible in (10b)? In view of its existence, it becomes difficult to argue for the impossibility of reconstructing the subject of a small clause to the base position in the general case h-assignment The inability of a small-clause subject to be interpreted in its base position can be due to two reasons. On the one hand, it can be that the subject of seem cannot be interpreted in the scope of the intensional verb because seem takes not a small clause but a predicate, as in (11a). If seem is not a raising verb, its subject cannot be interpreted below it. The other alternative, shown in (11b), is that something special about seem prevents its subject from reconstructing to its base position. I assume that the base-generation site of a small-clause subject is the specifier of a functional category (ap, np, and pp for AP, DP, and PP predicates, respectively). I leave it open if small clauses contain more structure than this.

6 224 Ora Matushansky (11) a. VP subject VN V 0 AP/DP/PP seem b. VP subject VN V 0 seem t i ap/np/pp predicate The analysis in (11a) has the advantage of explaining the contrasts in (6) and (10), whereas the analysis in (11b) needs an additional component explaining why a quantified subject of the small clause embedded under seem cannot be interpreted in situ. On the other hand, the analysis in (11a) makes all similarity between the two verbs totally accidental, by making one of them a transitive of sorts, assigning a (theme) h-role to its subject and taking a predicate as its argument. This by itself places the AP/DP/PP seem into a category of its own, while the IP/CP seem is merely a raising verb, and gives the two verbs not only a very different syntax but also very different semantics. The fact that the two are realized by the same lexical entry becomes totally mysterious. Furthermore, examples like (12a) (due to an anonymous reviewer) provide another reason to reject the analysis in (11a) and to conclude that the AP/DP/ PP seem is a raising verb. (12) a. There seems likely to be a problem in this sentence. b. *A problem seems likely to be in this sentence. If seem here takes an adjectival complement and assigns a h-role to its subject, what is there associated with in example (12a)? Note also that thereinsertion is obligatory (see (12b)). Other cases where the subject of seem does not appear to have an obvious h-role are also incompatible with (11a): (13) a. It seems cloudy today. (weather it) b. It seems obvious that the answer lies elsewhere. (expletive it) I therefore continue to assume that seem always takes a propositional complement, despite the impossibility of interpreting the subject of the AP/

7 Tipping the Scales 225 DP/PP seem in situ. Nevertheless, the differences go further: the two seem s have different interpretations. 2.2 Perception and Epistemicity In this subsection I argue that syntactic properties of seem are directly correlated with its interpretation: whereas the IP/CP seem is interpreted as an epistemic judgment, the AP/DP/PP seem is interpreted as a perception verb. Example (14a) cannot be felicitously used if perceptual evidence is unavailable for example, one cannot enter a room, look at Kleenexes and medicine bottles strewn all over the floor, and utter (14a). On the other hand, (14b) can be felicitously uttered in the same circumstances. (14) a. The squire seems sick. b. The squire seems to be sick. c. It seems that the squire is sick. So, (14a) implies perception of the subject s condition by the experiencer (I perceive that P holds), whereas (14b) is an epistemic deduction (from what I see I conclude that P holds). In other words, (14b) is (truth-conditionally and) pragmatically identical to (14c), but (14a) is not. Although the absence of the (14c) reading in (14a) could be attributed to the fact that the subject of the AP/DP/PP seem is base generated above seem, the difference in the meaning of the verb itself cannot be explained so easily. 2 The opposition between the epistemic and the perceptual meaning and the corresponding syntactic difference can be found elsewhere. Thus, Iatridou (1990) notes that the syntax of the verb sound changes depending on its interpretation and suggests that There are two verbs sound: one that depends on auditory perception (SOUND 1 ) and one that is a raising predicate (SOUND 2 ) (p. 560). The auditory perception SOUND 1 can only be used when the subject is physically present (or audible, as when reporting a phone conversation). It requires a stage-level predicate, with which SOUND 2 is only marginally compatible. This is why (15a), where the predicate (wonderful) cannot report a direct perception (unless of course Daniel is a singer), only has one interpretation (SOUND 1 ), in which it is roughly synonymous with (15b). (15) a. Daniel sounds wonderful. b. It sounds like Daniel is wonderful. 2 I find it next to impossible to formulate the difference between an epistemically derived conclusion and a perceptually derived one. My intuition is that in the latter case some empathy is required. The exact pragmatics of epistemicity being beyond the scope of this paper, I only wish to remark that, while the perceptual seem requires some sort of perception (including metaphorical cases like This law seems/looks/feels unfair), the epistemic one is not excluded there. In other words, the distribution of epistemic seem encompasses that of perceptual seem.

8 226 Ora Matushansky An intentional adverb like deliberately, being only compatible with a h- marked subject, can also disambiguate the two SOUNDs (Iatridou s example (38)): (16) John deliberately/consciously sounds happy (so that you wouldn t worry). Example (16) can only mean that John is making happy sounds; it cannot be intended as an epistemic judgment. The structure that Iatridou assigns to the two SOUNDs is given here (her (31) (32)): (17) a. John [sound happy] (SOUND 1 ) b. John i [sound [t i happy] (SOUND 2 ) Whereas the raising SOUND 2 has an epistemic interpretation and is naturally incompatible with subject-oriented adverbs, the h-assigning SOUND 1 relies on direct auditory perception and requires stage-level complements. The argument about sound extends to seem, which also relies on direct perception. Unlike SOUND 1, the AP/DP/PP seem does not seem to require stage-level predicates, but this can be due to the fact that it relies on all perceptual evidence rather than on only visual one. The difference in interpretation is then enough to argue for two lexical entries, a perception seem and an epistemic one. 3 To summarize, there are three major differences between the epistemic and the perception seem. Two of them reconstruction to a position below the intensional verb seem and the availability of expletives may be due to the structure of the complement of seem or its size. The remaining one, the difference in meaning that accompanies the different syntax, provides the first argument for postulating separate lexical entries for the epistemic and the perception seem Other Perception Verbs English has exceptionally many perception verbs, most of which can take a complement of some sort or another. To begin with, I wish to exclude those uses that seem to have little to do with my main concern. 3 A subject-oriented adverb can appear only with the AP/DP/PP seem or with both kinds, depending on the dialect, as long as the complement is not a that-cp, and with similar interpretations. 4 It is, of course, possible that one meaning is derived from the other, but then it becomes more than likely that it is the syntax (e.g., the type of complement) that determines the interpretation. A possible objection to this is the fact that even a small complement can have an epistemic reading (section 3.2.5).

9 Tipping the Scales Middles and transitives Most but not all perception verbs in English that have a middle form (verbs like reek and stink do not) have corresponding transitive forms: 5 (18) a. Susie smelled wonderful/tasted even better/felt surprised/etc. b. Susie smelled the smoke/tasted the apple/felt surprise/etc. Another use of perception verbs is when they appear with a resemblance complement: (19) a. The cake tastes/smells of chocolate. b. The apartment reeks/stinks of stale marijuana smoke. Only taste, smell, and their synonyms appear in this configuration. Given that seem does not have either the transitive or the resemblance use, I do not consider them here Complements All perception verbs can appear with a like-dp or a like-cp (as if and as though are also in this category): (20) a. The cake smells like a garbage can/like it s made of chocolate. b. The wine tastes like water/like it has gone sour. c. The pillow feels like a brick/like it s made of soft down. d. D Artagnan sounds like a musketeer/like he is hoarse. e. Grandmother looks like a lemon/like she might be grumpy. f. It seems/feels like it s/??it is going to rain. The choice between a that-clause and a like-clause as a complement of seem also appears to depend on the speech register. It should be noted that the cases in (20), where a perception verb takes a like-complement, should be distinguished from the following cases, where the like-dp is probably an adjunct: (21) a. John laughs like a horse. b. John laughs. c. *John laughs like a horse to me. The most important differences between (20) and (21) are the fact that in (21a) the like-dp can be omitted, as in (21b), and the absence of the experiencer h-role, as in (21c) (see the next section). 5 An interesting question is what comes first: the middle or the transitive form. Given that at least one perception verb (look) does not have a transitive form (though it does have a pseudotransitive variant, look at), presumably the middle form is basic.

10 228 Ora Matushansky I will exclude the like configuration from consideration, while noting that it establishes an additional point of similarity between seem and other perception verbs. Another such point is that none of these perception verbs permit an infinitival complement: (22) a. *The cake smells to be made of chocolate. b. *The wine tastes to be/have gone sour. c. *The pillow feels to be made of soft down. d. *D Artagnan sounds to be hoarse. e. *Grandmother looks to be grumpy. On the other hand, small complements are fine: (23) a. The cake smells wonderful. b. The wine tastes sour. c. The pillow feels soft. d. Mary sounds hoarse. (SOUND 1 ) e. Grandmother looks grumpy. I conclude that the restrictions on the complement of perception verbs strongly suggest a unification between them and seem Experiencer argument and intensionality Another point of similarity between seem and the other perception verbs is the fact that all of them take an experiencer argument: (24) a. Cannibals taste bitter to me. b. D Artagnan sounds hoarse to me. c. The pillow feels soft to me. d. Mary sounds hoarse to me. e. Grandmother looks grumpy to me. f. The patient seems insane to me. Modifying the sentences in (24) by a subject-oriented adverb poses no problem, but the experiencer argument cannot be projected then. This latter constraint is probably pragmatic in nature. 6 The verb appear is probably not part of the group, given that it does not allow the complementizer like in most dialects, and does allow an infinitival complement. I will not pursue the matter further.

11 (25) a. Cannibals intentionally taste bitter (*to me). b. Cannibals intentionally smell awful (*to me). c. A good doctor intentionally feels soft (*to me). d. D Artagnan intentionally sounds hoarse (*to me). e. Grandmother intentionally looks grumpy (*to me). f. The patient intentionally seems insane (*to me). Tipping the Scales 229 On the one hand, the possibility of projecting a subject-oriented adverb suggests that seem and other perception verbs assign a h-role to their subject position. Although one could argue that it is just a matter of coercion, similar to what happens in the passive sentence in (26a), passives (as opposed to middles, as in (26b)) require some h-role assignment. (26) a. The ship was intentionally sunk to collect insurance. b. *The ship intentionally sank to collect insurance. It is clear that the intentionality in perception verbs does not refer to the experiencer, but the matter is sufficiently marginal to leave it here, as there are a sufficient number of points of similarity between seem and other perception verbs Summary From the paradigms in (20) (25) I conclude that English perception verbs form a natural class. They all disallow an infinitival complement and appear to assign a theme h-role to their subject position. They are all compatible with a like-complement. They all take an experiencer argument and can appear with subject-oriented adverbs. In other words, they behave like the perception seem does. These similarities provide additional support for my assumption that the epistemic and the perception seem are two different lexical entries. To see that such ambiguity is not exceptional, consider the behavior of the verb feel. Like seem, feel can take a CP complement with the complementizer that, as well as a like-phrase. However, the meaning of the verb changes. Whereas in (20) the subject is the theme, in (27a) it is the experiencer, as shown by the ungrammaticality of an additional experiencer argument. (27) a. The doctor feels (*to me) that he might explode (*to me). b. The doctor feels angry (*to me). [* if the subject is interpreted as the experiencer] Adding the experiencer argument renders (27a) not ungrammatical but pragmatically strange, as if by touching the doctor one can feel that he might explode. In other words, the standard meaning of (27a) resembles that of (27b).

12 230 Ora Matushansky The same question surfaces: how is this new meaning of feel related to the old one? The parallel between feel and seem is not full (e.g., both meanings of seem are compatible with an experiencer argument, and in neither meaning can feel take an infinitival complement). The important point, however, is that the change in verbal syntax is accompanied by the change in meaning. This is unsurprising if we are dealing with a case of lexical ambiguity. If such is indeed the case, the fact that seem permits infinitival complements is treated on a par with the fact that feel can take the complementizer that with a special meaning. Both facts, although interesting in themselves, are irrelevant here. One seem (like one feel) is a perception verb taking a small-clause complement. 7 The other seem takes an IP or a CP and is interpreted epistemically. Table 1 summarizes the differences between the perception and the epistemic seem, regardless of whether these differences come from the verb itself or from the syntax that it imposes. In what follows, I am concerned only with the perception seem specifically, with the scalarity restriction that it imposes on its complement. 3. Scalarity in the Complement of Seem In this section, I show that seem can always combine with a predicate that is lexically scalar (a notion to be made precise in section 4). I will show that adjectives are felicitous in the complement of seem to the extent to which they can be interpreted as scalar. Nouns behave in a similar fashion, the difference between the two syntactic classes being the possibility of coercion: whereas, in the appropriate context, adjectives can be coerced into being scalar, nouns require an overt degree operator. Finally, I address the question of scalar PPs and show that the same constraints apply. The primary generalization to be made in this section is formulated in (I), and I will show later that the real constraint on the complement of seem has to do with the projection of a DegP. (I) Scalarity in the complement of seem (to be adjusted) The complement of perception seem must contain a scalar predicate. Table 1. Two lexical entries for seem Perception verb Epistemic modal Interpretation Perceptual Deductive Complements Small clause IP or CP h-assignment to subject position No No like-complements Yes No Experiencer Yes Yes 7 I am leaving like-dps and like-cps out of the picture for the moment.

13 Tipping the Scales 231 An interesting question is why verbs (gerunds, participles, or infinitives without to) are not possible complements of seem (or of any other perception verb) although Bolinger (1972), Doetjes (1997), and Sapir (1944), among others, have argued that VPs can be scalar. I do not deal with this question here. 3.1 Adjectival Small Clauses Bolinger (1972) argues that if the small-clause complement of seem is adjectival (i.e., its predicate is an [extended] AP), the adjective must be scalar (p. 77, judgments from Bolinger): (28) a. The music seems nice/*choral. b. The problem seems insoluble/*mathematical. In reality, the starred examples in (28) are usually considered grammatical though slightly odd. As Bolinger notes, a degree modifier (an intensifier, in his terms) makes them perfect: (29) a. The music seems almost choral. b. The problem seems pretty much mathematical. Attempting to combine a nonscalar adjective with a degree modifier in (29) results in a type mismatch. The repair strategy used in this case is a typeshifting operation called scalarity coercion, which changes the meaning of the nonscalar adjectives in (29) so that they can be interpreted as scalar. There are, however, some adjectives that cannot be coerced, as the contrast in (30) shows. (30) a. This number seems large. b. *This number seems third (in a sequence). c. This number seems to be third (in a sequence). Intuitively, the property of being third is binary: either something is third or it is not. Binary properties cannot be scalar because they cannot be present to a greater or lesser degree. Dimension properties like size or height, on the other hand, can, and this difference is reflected in the felicity of an adjective with seem. Additional support for this intuition comes from environments where scalarity is essential, such as exclamatives, comparatives, and modification by degree modifiers such as utterly, so, and so forth: (31) a. This number is the largest/*most third. b. This number is larger/*more third than that one. c. This number is so large/*third.

14 232 Ora Matushansky Bolinger notes that participles are possible in the complement of seem only as long as they are scalar (or, I would add, can be coerced into such interpretation). The only participles that can appear in the small-clause complement of seem are so-called adjectival passives and gerunds (for a discussion of adjectival passives, see Levin & Rappaport 1986, among others): (32) a. The floor seemed washed (*by Yolanda s husband (every day)). b. The children seem amusing/*dancing. Note that even such an extremely binary adjective as third can nevertheless be interpreted as scalar with enough effort. This is evidence for the restriction rather than against it, especially because it disappears if the complement of seem is an IP (or a CP), as in (30c) (repeated here). (30) c. This number seems to be third (in a sequence). Whatever semantics be has (if any), scalarity is not part of it. It is possible then to be sure that the grammaticality of (30c) cannot be due to the semantic impact of be-insertion, given that be has no semantic impact (in the relevant respect, at least). The only other difference between the two examples is in the meaning of seem: it is a perception verb in (30b) and an epistemic modal in (30c). Thus, we see another difference between the epistemic seem and the perception one: the latter but not the former imposes a scalarity-related restriction on the AP in its small-clause complement. In the following subsection I show that this constraint also applies to NPs. 3.2 Nominal Small Clauses The ability of seem to appear with a nominal small clause (a small clause with a DP predicate) is mostly a property of British English. Some dialects of American English allow it too, or distinguish the degrees of (un)grammaticality of different kinds of NPs in the complement. At first glance, nominal complements of seem and its adjectival complements are restricted in a different way. Bolinger (1972:78) demonstrates that only predicative degree nouns are permitted in the complement of seem: (33) a. What he writes seems nonsense. b. *What he writes seems adventure. c. What he writes seems to be adventure. Bolinger (1972:17) convincingly demonstrates that the class of intensifiers ( any device that scales a quality, such as perfect or rather) is open and, moreover, that almost any adjective or noun can be used as an

15 intensifier. This is why the only criterion for deciding whether a noun is a predicative degree noun is by checking its sensitivity to scalar environments. 8 For [+human] nouns, the generalization seems to be that only those expressing personal evaluation (i.e., those that can be used as epithets) are scalar (i.e., predicative degree nouns). However, as Ruwet (1982) demonstrates, the class of epithets is also open. However imprecise the definition, it works, because the nouns possible in the complement of the perception seem are exactly predicative degree nouns. As before, this restriction disappears if be is inserted (cf. (33c)). (34) a. Merlin seems a fool/a bastard. b. *Merlin seems a wizard/a man/arthur s friend. c. Merlin seems to be a wizard/a man/arthur s friend. Tipping the Scales 233 The goal of this subsection is to establish that the restriction on the nominal complement of seem is related to scalarity. To do so, I will demonstrate that other environments sensitive to scalarity consistently single out these nouns Degree adjectives English, like many other languages, has a group of adjectives, traditionally included in the group of nonstandard or subsective adjectives, that make reference to the extent to which the subject possesses the quality expressed by the predicate. A typical example is utter: 9 (35) a. Alice is an utter/complete/perfect idiot. b. *Merlin is an utter/complete/perfect wizard/briton/arthur s friend. Nouns compatible with these adjectives are possible complements of the perception seem, and those that cannot be complements of the perception seem also cannot be modified by these adjectives. An ambiguous noun like freak can indicate mental abnormality, in which case it is scalar, or physical abnormality, and in which case it is not scalar. Only the former meaning survives under modification by utter: (36) a. Sam is an utter freak he eats herring with strawberry preserves. b. #Sam is an utter freak not only does he have breasts, he has three of them. 8 It should be noted that, for nonscalar nouns, Bolinger (1972) leaves it unclear how to distinguish between a nondegree noun and a nonpredicative degree noun. 9 Compiling a full list of such adjectives is beyond the scope of this paper, but I find it interesting that all of them seem to refer to the high degree, rather than to the middle or low one, although there exist degree adverbs, such as barely, expressing the notion.

16 234 Ora Matushansky The reference to extents implicit in the meaning of these adjectives indicates that they refer to the degree variable of the head noun. For this to happen, the head noun has to be scalar Such and quite In this subsection I introduce two modifiers that refer to the extent to which an individual (the subject) possesses the property expressed by the (nominal) predicate. Both have equivalents in adjectival modification. I argue that the interpretation of these constructions depends on the inherent scalarity of the noun. The quite a construction is interpreted differently, depending on the noun it modifies. If the predicate noun expresses personal evaluation, the quite a construction is interpreted as indicating a high degree to which the predicate holds: (37) a. Adam is quite a genius. b. John is quite a fool. c. John is quite a fool yesterday he made the king laugh for three hours. The examples in (37a,b) mean that Adam is a genius and John, a fool to a high degree. This gradable interpretation is not available if the noun does not express personal evaluation. In that case, it is interpreted as if there is a covert adjective, similar in meaning to outstanding or nontypical: (38) a. William is quite a doctor last time he performed a surgery, he operated on four patients at once! b. This is quite a table! How much did you pay for it? In examples like (38) quite does not apply to the noun. (38a) does not mean that William is a doctor to a high degree (whatever that might mean) but rather that he is an exceptional doctor. This interpretation is not available when the predicate is Bolinger s predicative degree noun (or if the predicate noun is modified by an adjective, because then such is preferentially treated as applying to the adjective). Thus, examples (37a,b), where the predicate is a noun expressing personal evaluation, cannot mean that Adam is an exceptional genius and John is an exceptional fool. This latter interpretation becomes available only if fool is taken to mean a jester (cf. (37c)). To see that the contrast is indeed due to scalarity, consider the interpretation of scalar and nonscalar adjectives in the same context: (39) a. This house is quite old/charming/red. b. *This house is quite wooden.

17 The degree modifier quite distinguishes scalar and nonscalar predicates, and this is also what happens in such a construction. Here also, the meaning of scalar nouns differs from that of nonscalar ones (see Bolinger 1972:60 77): (40) a. You re such a linguist. You re such a typical linguist. b. You re such a fool. You re an utter fool. The adjectival equivalent of this construction is so-modification, and the restrictions are the same. If the adjectival predicate is nonscalar, somodification is infelicitous: (41) a. This house is so old/charming/red that every tourist photographs it. b. #This house is so wooden that every tourist photographs it. Whereas (41a) is perfectly grammatical, (41b) is only interpretable if the property under discussion is not being made of wood but rather having many of the typical properties associated with being wooden the meaning shift that is typical of scalarity coercion. As with nonscalar nouns, the notion of a prototype or a stereotype comes into play, and the link seems clear: in the context of so-modification, scalar adjectives behave in the same manner as nouns assumed to be scalar do in such a construction. To be sure, the interpretation that nonscalar nouns receive in this environment (a typical X) is also available to scalar nouns, but it is more difficult to obtain and is not the default one. It should be noted that the presence of degree modifiers such and quite, if felicitous with a given nonscalar noun (preferably as an exclamative), make it a possible complement of seem, just like other degree modifiers do with adjectival small clauses. (42) a. You d seem such a linguist! b. He sure seems quite a linguist! Tipping the Scales 235 To summarize, in this subsection I examined two constructions involving scalarity such a/so and quite (a). Both behave identically with respect to adjectives and nouns expressing personal evaluation. I conclude that nouns expressing personal evaluation are indeed scalar and their behavior in the complement of seem is only another manifestation of this status How questions Bolinger (1972:78) notes that the interrogative how exhibits the same restrictions as seem: it is the interrogative replacement for degree adjectives and for predicative degree nouns, contrasting with what :

18 236 Ora Matushansky (43) How is the new book he wrote? a. It s fine/*descriptive. b. It s nonsense/*adventure/*description. (44) How is his mother? Angelic/an angel/*a seamstress. According to Bolinger, the starred examples become acceptable if what, which has nothing to do with degrees, replaces how Cross-categorial comparison: adjectives versus nouns I have shown that the scalarity-related restriction imposed by seem on its adjectival complement extends to nominal complements. Only predicative degree nouns are treated as scalar. That the distinction is indeed semantically based is shown by the behavior of these nouns in scalarity-sensitive contexts and by their argument structure. The fact that adjectives such as utter can only be combined with predicative degree nouns is taken to indicate that these nouns are scalar. Just like scalar adjectives, predicative degree nouns seem to be vague that is, it is relatively unambiguous whether someone is a doctor or not, but deciding whether someone is or is not a fool is much more a matter of opinion. The use of these nouns as epithets (as in that idiot!) further supports the analogy, as does the interpretation of these nouns when combined with quite and such. I have shown that these contexts are indeed related to scalarity by analyzing what happens to scalar and nonscalar adjectives under similar degree modification. It would seem that predicative degree nouns are also scalar in other languages, notably in French and Russian, and the same may be true of scalar PPs discussed in section 3.3. If the semantics of inherent scalarity is universal, then we must ask why predicative degree nouns, and no other kind, should be conceived as scalar, although it can be shown that some sort of prototype exists for all categories. 10 I have also shown that there exists a major difference between nominal and adjectival complements of seem. Whereas most nonscalar adjectives can appear in the complement of seem in the right context, only predicative degree nouns can appear there. Nouns that are barred from the complement of seem often have the same meaning as adjectives permitted there: (45) a. Lucy is/seems Spanish. b. Lucy is/*seems a Spaniard. 10 Even concepts like ODD NUMBER, which clearly do have definitions, often have prototype structure as well. The number is 3 is a better odd number than The discovery of the massive presence of prototypicality effects in all sorts of mental processes is one of the success stories of cognitive science (Fodor 1998:93).

19 Tipping the Scales 237 Adjectives such as Spanish are not scalar, but they can receive a scalar interpretation in an appropriate context, such as modification by a degree operator or in the complement of seem. Felicitousness of such interpretation is pragmatically determined (e.g., prime is barely coercible, but Spanish is perfect). As Bresnan (1973), among others, notes, nouns can be coerced into scalarity in the context of a degree operator (e.g., more of a man). The question is why nouns cannot be coerced into being scalar in the small-clause complement of seem in the same way adjectives are. A possible answer is that they can, but other considerations interfere. I return to this issue in section 6.3. The final puzzle of this section is why languages differ with respect to whether they treat all nonverbal predicative categories the same. French and most dialects of American English allow adjectival and prepositional small clauses in the complement of seem, while disallowing nominal complements. However, some but not all English dialects (including British English) also allow nominal small clauses with predicative degree noun predicates. In the next section, I briefly consider a variety of English where the nominal complement of seem appears to be even less constrained Counterexamples As an anonymous reviewer points out, nineteenth-century literature and modern written media contain numerous sentences of the following sort: (46) a. Certainly it would seem a break from a Thatcher revolution that has stressed harsh economic realities, value for money, and the new brutalism of monetarism. Wall Street Journal Corpus b. It became so monotonous as to seem a part of the quietness. Brown Corpus c.... they seemed an eruption of bears from Labrador. Melville, Moby Dick d.... it seemed to her her duty to do so. Thackeray, Vanity Fair These cases clearly contradict the generalization expressed previously that only scalar nouns can be complements of seem. Two questions must be answered: are these sentences grammatical in modern English, and why are they grammatical if the generalization above is correct? It is clear that the modern reader is not shocked with Thackeray s grammar and that editors of the Wall Street Journal can understand examples like (46a,b). The fact remains, however, that when asked to judge similar sentences out of context, most English speakers consider them ungrammatical. To explain this, I draw on the fact that a language does not consist of a single dialect, and English is no exception, and that each native speaker of English has the ability, however it is expressed, to understand some standard dialects.

20 238 Ora Matushansky It is then necessary to ask why the constraints on seem are loosened in the nineteenth-century literary style and in modern journalism, and what the difference is. To begin with, in modern Italian, nonscalar DPs can be complements of seem (judgments by Elena Guerzoni and Marco Nicolis): (47) (Me) sembra un dottore/idiota. me-dat seems INDEF doctor/idiot S/he seems (to me) to be a doctor/idiot. What distinguishes English from Italian (and other varieties of English)? Suppose the Italian seem is not a perception verb but rather, as the English seem with an IP complement, means It can be deduced from perceptual evidence that P holds. The constraint then becomes a pragmatic one of perceptual or epistemic uncertainty. What evidence is there that the meaning proposed for Italian seem is not that of English seem? In other words, how do we know that the constraint on the complement of seem in English is not that of perceptual or epistemic uncertainty, as suggested by an anonymous reviewer? Such an analysis would link the ungrammaticality of precise measure phrases in the adjectival complement of seem in (48a) versus (48b) to the fact that the precision is too high for the embedded proposition to be epistemically or perceptually at issue. (48) a. *Thumbelina seems two inches tall. b. Thumbelina seems about two inches tall. c. Thumbelina seems two inches taller than expected. One reason to reject this analysis has to do with the fact that precision expressed by the measure phrase does not seem to be at issue when the measure phrase is part of a comparative. Furthermore, consider a situation where the normal way of measurement is in fairly large units, as in (49). Using smaller, larger, or different units of measurement, as in (49a), or a vague plural, as in (49b), which should have influenced the grammaticality if it truly relied on precision of perception, has no effect here. (49) a. *Mount Everest seems 8848 m/8.8 km/29,028 feet/8,848,000 mm high. b. *Mount Everest seems thousands of meters high. An additional reason to doubt a purely pragmatic explanation is the difference made by be-insertion. Why is there a difference between the ungrammatical (34b) and the grammatical (34c), repeated here, when we believe that be does not make any semantic contribution, its role being purely syntactic?

21 (34) b. *Merlin seems a wizard/a man/arthur s friend. c. Merlin seems to be a wizard/a man/arthur s friend. Tipping the Scales 239 Given how dubious it is that a copular infinitive is more epistemically or perceptually uncertain than a small clause, some sort of a conceptual difference between the perception seem and the epistemic seem must be postulated. The question, then, is how to formalize it. The fact that Italian appears to differ from modern English and resemble Victorian English or today s written English makes it even less likely that the problem can be solved by a mere appeal to pragmatics. Finally, it is unclear how a pragmatics-based theory can explain the difference between the behavior of APs and DPs with respect to scalarity coercion in the complement of seem, given that it is difficult to imagine how the veridicality of nonscalar adjectives can be more easily perceived as being at issue than nonscalar nouns Prepositional Complements Corver (1997) notes that degree operators fall into two classes. Degree operators of class 1, such as very or as, appear only with adjectives. Degree operators of class 2, such as more or enough, attach to syntactic classes other than APs. Doetjes, Neeleman, and van de Koot (1998:fn. 3) point out that the distinction is more fine-grained, once PPs are considered. 12 Whereas most PPs can only appear with class 2 operators, as in (50), their example (11), certain PPs but not others can appear with class 1 (ex. (51a,b), their (12a,b), vs. (51c,d)). (50) a. He is [more [ PP on drugs]] than any of his friends. b. He is [less [ PP into syntax]] than he was before. c. He is [enough [ PP over the limit]] to be arrested. (51) a. *He is [too [ PP under scrutiny]] to be elected at this time. b. *He is [as [ PP over the limit]] as Bill. c. He is [too [ PP out of his mind]] to be elected at this time. d. He is [as [ PP in love]] as Bill. PPs that can combine with degree operators of class 1 (henceforth, degree PPs ) are also possible complements of seem (as well as of consider; see Doetjes et al. 1998:fn. 3): 11 Additional syntactic constraints on the scalar licenser of seem discussed in sections 6.2 and 6.4 are also quite inexplicable from a purely pragmatic point of view. 12 Doetjes (1997) and Doetjes et al. (1998) use these contrasts to argue for a syntactic difference between the two classes: class-1 degree operators are heads, c-selecting their complement, and class-2 items are phrases, either adjoined or merged in the specifier position.

22 240 Ora Matushansky (52) a. Leila seems down in the mouth/out of sorts/out of it. b. *Leila seems in the room/out of the house. As expected, example (52b) becomes grammatical if to be is inserted before the PP: (52) b 0. Leila seems to be in the room/out of the house. Most PPs in this class describe emotional or mental states, but Irene Heim (p.c.) notes that certain nonidiomatic positional PPs are also possible complements of seem, as illustrated in (53). 13 The felicity of the expression appears to depend on how detectable the property is: (53) a. The singer seems (?)out of luck/ out of synch. b. The picture seems out of focus. c. The ball seems off balance. d. The wheels seem (?)out of alignment/*in alignment. The scalarity of degree PPs can be confirmed by the same diagnostics as that of predicative degree nouns Scalarity diagnostics Like predicative degree nouns and scalar adjectives, degree PPs can be modified by quite, so, and utterly. Nonscalar (e.g., locational) PPs may not, resulting in cases like (54b), (55b), and (56b), which are at best marginal: 14 (54) a. Red was so out of his mind that he needed a rest. b. *Red was so out of the hotel that he needed a rest. (55) a. Red might be quite out of his mind. b. *Red might be quite out of the hotel. (56) a. Red might be utterly out of his mind. b. *Red might be utterly out of the hotel. The issue is complicated by the fact that spatial PPs resemble dimension adjectives in their ability to take measure phrases. Unlike dimension adjectives, however, they cannot appear in the complement of seem: 13 Because my analysis does not hinge on the noncompositionality of idiomatic PPs, I will not deal with the locus of scalarity in idiomatic PPs. 14 For unclear reasons, prepositional so-modification requires a complement or strong focal stress.