2.1 Inappropriateness and adjectival selection

On the surface, nominal AICs are identical to a construction I call the clausal attributive-with-infinitive construction (clausal AIC). The terms ‘nominal’ and ‘clausal’ reflect a basic semantic intuition about what the attributive adjective modifies: the adjacent noun or the gapped infinitival clause that follows it.Footnote ² Nominal AICs may be identified and distinguished from clausal AICs on both syntactic and semantic grounds. The basic contrast between the two constructions is illustrated in (2).

1. (2)

   1. (a) Middlemarch is a long book to assign. (nominal AIC)

   2. (b) Middlemarch is a bad book to assign. (clausal AIC)

As can be seen from these examples, both types of AIC have a surface syntax in which a noun is flanked by an attributive adjective to its left and a gapped infinitival clause to its right. Two major characteristics of nominal AICs set them apart from clausal AICs, making them readily identifiable.

First, nominal AICs have an interpretation of unexpectedness or inappropriateness associated with them. In (2a), we have the sense that Middlemarch is unexpectedly or inappropriately long for the purpose at hand, i.e., for an act of assigning. The purpose in question is always expressed by the gapped infinitival clause in a nominal AIC. Clausal AICs lack this interpretation. Any sense of inappropriateness associated with (2b) is due to the meaning of the adjective bad; it is not a general interpretive feature of the construction. Consider, for the sake of comparison, the clausal AIC Middlemarch is a good book to assign; here, there is no sense that Middlemarch is unexpectedly or inappropriately good. Nominal AICs, by contrast, always have the inappropriateness interpretation, even with adjectives that typically lack pejorative connotations (see, e.g., (5a) below).Footnote ³

Second, nominal AICs may be formed with adjectives that do not independently select infinitival-clause arguments, while clausal AICs may only be formed with adjectives that are able to select such arguments. Evidence comes from a difference in the acceptability of impersonal paraphrases involving an extraposed infinitival-clause subject. The contrast between nominal-AIC long and clausal-AIC bad is shown in (3) and (4); a second pair of examples, with a nominal AIC formed with well-made and a clausal AIC formed with easy, is shown in (5) and (6).Footnote ⁴

1. (3)

   1. (a) Middlemarch is a long book to assign.

   2. (b) *It is long to assign Middlemarch.

1. (4)

   1. (a) Middlemarch is a bad book to assign.

   2. (b) It is bad to assign Middlemarch.

1. (5)

   1. (a) That is a well-made car to sell for scrap.

   2. (b) *It is well-made to sell that car for scrap.

1. (6)

   1. (a) That is an easy car to drive on hills.

   2. (b) It is easy to drive that car on hills.

Nominal AICs thus permit a wider range of adjectives than clausal AICs, including adjectives that are not independently able to select infinitival-clause arguments.Footnote ⁵ Moreover, when adjectives like good and easy occur in nominal AICs, they fail to have impersonal, extraposed-subject paraphrases of the type seen in (4b) and (6b), and instead take on the inappropriateness interpretation that is characteristic of nominal AICs:

1. (7)

   1. (a) [Context: Someone has been looking for a parking space for a long time and is running late.]

      That is a good parking space to pass up.

      ≠ It is good to pass up that parking space.

   2. (b) [Context: Someone becomes exhausted from a very modest amount of exercise.]

      That is an easy workout to get exhausted from.

      ≠ It is easy to get exhausted from that workout.

Together with the data in (3) and (5), the examples in (7) show that the infinitival clause of a nominal AIC is not an argument of the attributive adjective. The adjective is either generally unable to take such an argument, or it fails to do so in that particular case, as evidenced by the lack of an extraposed-subject paraphrase.

Direct modification of the noun

There is ample evidence that the attributive adjective in a nominal AIC modifies the following noun, and not the infinitival clause as in a clausal AIC. First, as discussed above, nominal AICs differ from clausal AICs in their ability to host adjectives that do not independently select infinitival arguments. The unacceptability of the impersonal paraphrases in (3b) and (5b) suggests that we cannot adopt an analysis in which the attributive adjective modifies the infinitival clause and not the noun (i.e., the type of analysis we might pursue for clausal AICs; for specific proposals on clausal AICs, see Berman Reference Berman1974a, Flickinger & Nerbonne Reference Flickinger and Nerbonne1992, Dubinsky Reference Dubinsky, Gelderen and Samiian1998, and Huddleston & Pullum Reference Huddleston and Pullum2002). Rather, we must assume that the attributive adjective modifies the adjacent noun directly, as in an ordinary DP.

One source of support for this idea is the fact that attributive adjectives in nominal AICs display selectional restrictions observed with ordinary attributive adjectives. Consider adjectives that describe a mental state or attribute, like smart. Such adjectives require that the nouns they modify denote sentient (most likely human) beings. Notably, it is impossible to construct nominal AICs in which such adjectival selectional restrictions are violated. In clausal AICs, by contrast, such restrictions may be freely ignored. In (8a) below, the adjective smart is followed by the noun sofa, which fails to satisfy the adjective's selectional requirements; as a result, the only available interpretation of the sentence is a clausal-AIC interpretation, with the adjective modifying the infinitival clause, not the noun.Footnote ⁶ In (8b), by contrast, where the selectional restrictions of the adjective are met, both nominal- and clausal-AIC interpretations are available. In this case, the advisability of hiring Bob differs depending on which interpretation one chooses: on the nominal-AIC reading, Bob is perhaps too crafty to be a trustworthy accountant; on the clausal-AIC reading, hiring Bob is a crafty move and thus good.

1. (8)

   1. (a) That is a smart sofa to buy.

   2. (b) Bob is a crafty person to hire as your accountant.

The fact that nominal-AIC interpretations are available only when the noun meets the selectional restrictions of the adjective suggests strongly that nominal-AIC adjectives modify the following noun directly, i.e., that whatever syntactic and semantic relationship holds between attributive adjectives and nouns in ordinary DPs holds between them in nominal AICs as well. Further evidence in favor of this conclusion comes from patterns of entailment. Unlike clausal AICs, nominal AICs fail to support entailment into supersets of noun denotations. In this respect, nominal-AIC DPs behave like ordinary predicative DPs in which an attributive adjective modifies the following noun, while clausal AICs behave as if the attributive adjective does not modify the noun at all. The contrast is shown in (9) and (10).Footnote ⁷

1. (9) Nominal AIC

   1. (a) That is a big sparrow to see in this area. ↛ That is a big bird to see in this area.

   2. (b) That is a big sparrow. ↛ That is a big bird.

1. (10) Clausal AIC

   1. (a) That is a good novel to read.→That is a good book to read.

   2. (b) That is a novel.→That is a book.

The behavioral parallels between attributive adjectives in nominal AICs and those in ordinary DPs suggest that they bear the same relationship to the immediately following noun in both cases. Moreover, it is quite different from the adjective–noun relationship found in clausal AICs.

Infinitival relative or degree complement?

One of the most compelling characteristics of nominal AICs, from both a descriptive and a theoretical perspective, is their apparent hybrid nature: nominal-AIC infinitival clauses seem to have properties of both relative clauses and degree complement clauses. This observation applies equally to their syntax and to their semantics. Consider the examples in (11).

1. (11)

   1. (a) Middlemarch is a long book. (positive)

   2. (b) Middlemarch is a longer book than Pnin is. (comparative)

   3. (c) Middlemarch is a long book to assign. (nominal AIC)

First, and perhaps most obviously, we can observe that the morphological form of the adjective in the nominal AIC in (11c) is the same as that in the ordinary positive in (11a). In both cases, we see the unmarked, morphologically basic form of the adjective, in contrast to the comparative form in (11b), which bears overt degree morphology in the form of the -er inflection.Footnote ⁸ The nominal AIC, however, also contains an infinitival clause that appears to function analogously to the comparative than clause in (11b); both clauses provide information about the relevant standard to which Middlemarch's length is compared.Footnote ⁹ The infinitival clause has no counterpart in the ordinary positive in (11a). The nominal AIC thus seems to be part positive, part comparative; in particular, the infinitival clause appears to have properties of both an ordinary relative clause and a degree complement clause such as a than or as clause.

Two additional facts suggest connections between nominal AICs and degree constructions. First, nominal AICs may be formed only with gradable adjectives, a property they share with comparatives and other degree constructions. Adjectives that denote absolute, categorical properties, like American or dead, are infelicitous in nominal AICs, just as they are in comparatives:

1. (12)

   1. (a) That is a dead bug.

   2. (b) #That is a deader/more dead bug than this one is.

   3. (c) #That is a dead bug to leave on your windowsill.

Example (12c) lacks the inappropriateness reading that is characteristic of nominal AICs. The bug in question cannot be dead to an inappropriate degree, as the requisite scale of values simply does not exist for the adjective dead: something is either dead or it is not. This conclusion is reinforced by the unacceptability of the comparative in (12b). Nominal AICs, and the inappropriateness reading that characterizes them, thus require a gradable adjective in order to be felicitous.

Second, the inappropriateness reading of a nominal AIC is lost when a standard of comparison is provided by an overt measure phrase. Consider the examples in (13).

1. (13)

   1. (a) Middlemarch is a 700-page-long book to assign.

   2. (b) That is a 3-year-old car to drive across the country.

Neither sentence in (13) has the inappropriateness reading. (13a) tells us that Middlemarch is 700 pages long, and suggests that one ought to assign it – i.e., it has what one might call an exhortative reading – but there is no implication that the book is inappropriately long for the purpose at hand. The facts are analogous for (13b). We lose the signature interpretive characteristic of nominal AICs – inappropriateness – precisely when a standard of length or age is specified overtly by a measure phrase. This is consistent with an analysis in which the infinitival clause in a nominal AIC specifies a standard of comparison. On such a view, the measure phrases in (13) preempt the infinitival clauses' ability to provide a standard, with the result that the nominal-AIC reading is unavailable. Note in this connection that when measure phrases occur in ordinary comparatives, they fail to denote a standard of comparison, indicating instead an interval by which the subject differs from the standard expressed by the than clause:Footnote ¹⁰

1. (14)

   1. (a) Middlemarch is 500 pages longer than Pnin is.

   2. (b) My car is 3 years younger than yours is.

The behavior of measure phrases in nominal AICs once again points to the hybrid nature of the construction. The loss of the inappropriateness reading in such examples suggests a standard-providing role for the infinitival clause in ordinary nominal AICs; if the infinitival clause provides a standard, then it looks much like a degree complement clause. The measure phrase, however, has a different semantics in nominal AICs than it does in comparatives: in the former it provides a standard of comparison, while in the latter it indicates the difference between the standard and another value. I take up the analysis of nominal-AIC infinitival clauses in detail in Section 3.

Nominal AICs and too

I conclude the overview of the basic properties of nominal AICs by pointing out some important differences between nominal AICs and sentences with too. At first sight, nominal AICs seem to be syntactically and semantically very similar to attributive too constructions, as shown in (15).

1. (15)

   1. (a) Middlemarch is a long book to assign.

   2. (b) Middlemarch is too long a book to assign.

Both sentence types express the thought that Middlemarch is inappropriately long for the purpose of assigning a book. Moreover, in both cases, the relevant standard of length is expressed by a gapped infinitival clause; this is in contrast to what we see with comparatives and equatives, where the standard is expressed by a finite clause. While the nominal AIC in (15a) may be less forceful in its assertion of inappropriateness than the too construction in (15b), in both cases there is a sense that the relevant standard is exceeded. Given the very close syntactic and semantic correspondences between the two constructions, we may ask whether nominal AICs are simply a variant of the attributive too construction.

Further consideration shows that the answer must be no. To begin, I note three important syntactic differences between nominal AICs and attributive too constructions. First, there is an obvious difference in the surface position of the attributive adjective in the two sentence types. In nominal AICs, the adjective occupies its customary place to the left of the noun. With attributive too, however, the adjective is pied-piped with too to a position to the left of the determiner. This position has been identified as the specifier of a functional projection above DP by Kennedy & Merchant (Reference Kennedy and Merchant2000), who propose it as the landing site for how-plus-adjective strings in attributive comparatives. Note that in both attributive too constructions and attributive comparative questions, the head of this functional projection may optionally be lexicalized as of, as shown in (16). No such functional projection is involved in the derivation of nominal AICs, as shown in (17).

1. (16)

   1. (a) Middlemarch is too long (of) a book to assign.

   2. (b) How long (of) a book is Middlemarch?

1. (17) Middlemarch is (*of) a long book to assign.

Second, the adjectival complex in a too construction may occur after the noun, a configuration that is impossible in nominal AICs, as shown in (18).Footnote ¹¹

1. (18)

   1. (a) *Middlemarch is a book long to assign.

   2. (b) Middlemarch is a book too long to assign.

Third, nominal AICs differ syntactically from attributive too constructions in their failure to license resumptive pronouns in the gap position of the infinitival clause. Consider the contrast in (19).

1. (19)

   1. (a) *Middlemarch is a long book to assign it.

   2. (b) Middlemarch is too long a book to assign it.

There are likewise important semantic distinctions between nominal AICs and attributive too. In particular, these constructions differ markedly with respect to the semantic status of their infinitival clauses. In nominal AICs, the truth of the infinitival clause is presupposed, while in too constructions its falsity is generally entailed. The contrast is particularly salient in the past tense. Consider (20).

1. (20)

   1. (a) Middlemarch was a long book to assign.

   2. (b) Middlemarch was too long a book to assign (it).

From the nominal AIC in (20a), we conclude that Middlemarch was in fact assigned. Consider the incongruity of the following conjunction: #Middlemarch was a long book to assign, and in fact we didn't assign it. Negating or questioning the nominal AIC does not alter the conclusion that the book was assigned, as shown in the infelicitous dialogues in (21). The content of the infinitival clause is thus presupposed true.

1. (21)

   1. (a)

      A: Middlemarch was not a long book to assign.

      B: #Then why didn't you assign it?

   2. (b)

      A: Was Middlemarch a long book to assign?

      B: #Yes, so it's a good thing you didn't assign it.

The too construction in (20b), by contrast, generally gives rise to an entailment that the book was not assigned.Footnote ¹² Consider the infelicity of the following: #Middlemarch was too long a book to assign (it), but we assigned it. Under negation and questioning, the conclusion that the book was not assigned vanishes, as shown in the perfectly felicitous dialogues in (22). This is the hallmark of semantic content that is entailed, not presupposed.

1. (22)

   1. (a)

      A: Middlemarch was not too long a book to assign (it).

      B: Is that why you assigned it?

   2. (b)

      A: Was Middlemarch too long a book to assign (it)?

      B: No, and it's a good thing you assigned it.

These data indicate a fundamental semantic difference between nominal AICs and too constructions. While both indicate some amount of inappropriateness, it is clear from the examples in (20) and (21) that nominal-AIC inappropriateness is not so great as to rule out the event or state denoted by the infinitival clause. On the contrary, it is presupposed true. In too constructions, by contrast, the inappropriateness is such that the infinitival-clause content is entailed to be false: in (20b), the length of Middlemarch is so great that one cannot assign it. Nominal AICs thus cannot be seen as simple variants of too constructions. The semantic divide between the two is far wider than the simple comparison in (15) suggests.

Nominal AICs thus differ in important respects from too constructions, both semantically and syntactically. While I will not pursue an analysis of attributive too constructions here, I hope to have shown that nominal AICs are sufficiently different from them as to merit independent consideration.

3.1 The distribution of nominal-AIC DPs

Distributional data support the analysis of nominal-AIC infinitivals as infinitival relatives. Nominal-AIC DPs are limited in their syntactic distribution, occurring only in the predicative position of copular clauses. This restriction is shared by other DPs that contain infinitival relatives, but not by attributive comparative DPs. The restriction follows naturally on the relative-clause analysis of nominal AICs proposed here, but not on the degree complement analysis, offering support for the former.

It is no accident that all examples of nominal AICs discussed so far take the form of copular clauses. The nominal-AIC DPs of interest – i.e., the DPs that contain an attributive adjective and an infinitival clause – always occur in the predicative position of a copular clause. They cannot be construed referentially, and thus are unacceptable in the argument positions of non-copular clauses, where DPs are typically found; this is shown in (24).

1. (24)

   1. (a) #Bob is reading a long book to assign.

   2. (b) #Susan went to a violent movie to take her nephew to.

   3. (c) #A tall person to hire is standing over there.

The restricted distribution of nominal-AIC DPs is paralleled by that of indefinite DPs that contain infinitival relatives. The examples in (25) differ from those in (24) only in their lack of attributive adjectives. Meanwhile, such DPs are perfectly fine in the predicative position of a copular clause, as shown in (26).

1. (25)

   1. (a) #Bob is reading a book to assign.

   2. (b) #Susan went to a movie to take her nephew to.

   3. (c) #A person to hire is standing over there.

1. (26)

   1. (a) Middlemarch is a book to assign.

   2. (b) That is a movie for Susan to take her nephew to.

   3. (c) Bob is a person to hire.

The parallel distribution of nominal-AIC DPs and ordinary indefinite DPs with infinitival relatives is unsurprising on the relative-clause analysis of nominal AICs. On this analysis, the infinitival clause is simply an infinitival relative adjoined to NP, and so whatever distributional restrictions exist for such DPs on independent grounds should also hold for nominal AICs.Footnote ¹⁶ These distributional facts would be difficult to account for on the degree complement analysis. As shown in (27), indefinite attributive comparative DPs are felicitous in non-predicative positions, in sharp contrast to the nominal-AIC DPs in (25).

1. (27)

   1. (a) Bob is reading a longer book than Steve assigned.

   2. (b) Susan went to a more violent movie than she can take her nephew to.

   3. (c) A taller person than we've ever hired is standing over there.

If nominal-AIC DPs had the structure of attributive comparative DPs, then it would be difficult to explain why the latter are permitted in non-predicative positions while the former are not. The restricted distribution of nominal-AIC DPs thus both supports the relative-clause analysis and casts doubt on the degree complement analysis.

It must be stressed that the restriction to predicative position is operative only for DPs that contain infinitival relatives. In particular, it does not hold for constructions with infinitival purpose clauses. These are often indistinguishable on the surface from nominal AICs or other infinitival-relative–containing DPs. Consider the examples in (28).

1. (28)

   1. (a) Bob found a book to assign.

   2. (b) Susan heard about a violent movie to take her nephew to.

On the face of it, the examples in (28) appear to violate the distributional restriction discussed above. As discussed by Berman (Reference Berman1974b), however, the infinitival clauses in (28) are not infinitival relatives adjoined to NP, but rather purpose infinitivals adjoined at the VP level (for detailed discussion of purpose clauses, see Jones Reference Jones1991). To begin, the examples in (28) lack the exhortative reading observed with DPs that contain infinitival relatives, as in (13) above: (28a) can be uttered felicitously in a situation where one's attitude towards assigning the book in question is neutral or even negative. Furthermore, consider that wh-movement from the direct object positions in (28) must strand the infinitival clause:

1. (29)

   1. (a) Which book did Bob find to assign?

   2. (b) #Which book to assign did Bob find?Footnote ¹⁷

The pattern in (29) makes sense only if we assume that the infinitival clause is adjoined not within DP, but at a higher level, such as VP. If it were adjoined within DP, it would be able to undergo wh-movement via pied-piping; for comparison, consider the finite relative clause in (30).

1. (30)

   1. (a) Bob found a book that Susan had recommended.

   2. (b) Which book that Susan had recommended did Bob find?

Note further that the inappropriateness interpretation that is characteristic of nominal AICs is absent from the purpose infinitival sentence (28b). The sentence does not mean that Susan heard about a movie that was inappropriately violent for taking her nephew to; rather, it implies that she had been intending to take her nephew to a violent movie all along, and expresses no speaker commitment one way or the other as to the appropriateness of taking one's nephew to violent movies. The purpose infinitival interpretation arises most naturally with predicates of discovery or introduction like find, bring, and hear about. The interpretation is far less salient with a verb like go, which is why (24b) and (25b) are infelicitous: without the purpose infinitival parse, the only remaining interpretation is one in which the infinitival clause adjoins to NP as an infinitival relative, and this runs afoul of the predicative-position restriction. Once we control for the purpose infinitival reading and its different syntactic structure, we see that the predicative-position restriction on DPs with infinitival relatives remains valid.

Finally, we must note another apparent exception to the predicative-position restriction: specificational copular clauses (Higgins Reference Higgins1979). In such clauses, the predicative DP is found not in the ordinary post-copular position, but rather in subject position. The somewhat degraded acceptability of nominal-AIC DPs in specificational copular clauses, shown in (31), results not from any lack of predicativeness on the part of the DP, but rather from the fact that specificational-clause subjects are topics (Mikkelsen Reference Mikkelsen2005), and nominal-AIC DPs make poor topics. For example, they fail to occur with the definite article in either predicational or specificational copular clauses, as shown in (32).

1. (31)

   1. (a) #A long book to assign is Middlemarch.

   2. (b) #An old man for them to hire is Bob.

1. (32)

   1. (a) #Susan is the young sibling to invite.Footnote ¹⁸

   2. (b) #The young sibling to invite is Susan.

We thus see that nominal-AIC DPs are subject to a distributional restriction observed to hold independently for DPs that contain infinitival relatives. Attributive comparative DPs, meanwhile, are not restricted in the same way. These facts support the relative-clause analysis of nominal AICs and cast doubt on the degree complement analysis.

3.2 Scope of quantifiers in the infinitival clause

Nominal AICs fail to exhibit scopal interactions that are characteristic of ordinary comparatives; such interactions are not expected on the relative-clause analysis. It is well known that quantificational DPs within a finite comparative than clause may – and in many cases must – take wide scope with respect to the comparative degree morpheme itself. Quantificational DPs in nominal-AIC infinitival clauses, by contrast, are always interpreted within the scope of the positive degree morpheme (absent contrastive emphasis on the DP in question; see below). The degree complement analysis of nominal AICs fails to predict the absence of the wide-scope reading for these quantificational DPs, while the absence of this reading follows naturally on the relative-clause analysis.

The examples in (33) illustrate the phenomenon in question (Larson Reference Larson1988, Schwarzschild & Wilkinson Reference Schwarzschild and Wilkinson2002, Heim Reference Heim2006). In each case, a quantificational element in the comparative clause scopes outside the comparative degree head; i.e., the ‘greater than’ relation between degrees denoted by -er lies within the scope of every girl or exactly two girls at LF. The authors cited here focus on predicative comparatives; the examples in (34) show that the same phenomenon occurs in attributive comparatives. (Logical translations are approximate, and are meant to show the scopal relations; in particular, these representations do not capture the presuppositions in (34) that all of, or at least two of, John's friends are men; for discussion, see Bresnan Reference Bresnan1975.)

1. (33)

   1. (a) John is taller than every girl is.

      ∀x[girl(x)→John's height>x's height]

   2. (b) John is taller than exactly two girls are.

      |λx.girl(x)∧John's height>x's height|=2

1. (34)

   1. (a) John is a taller man than all of his friends are.

      man(John)∧∀x[friend(x)→John's height>x's height]

   2. (b) John is a taller man than exactly two of his friends are.

      man(John)∧|λx.friend(x)∧John's height>x's height|=2

Example (33a) means that John is taller than the tallest girl. This intuition is captured by the semantic representation sketched for (33a), which results from the scopal relationship every girl≻-er. With the opposite scope, -er≻every girl, the sentence would mean that John is taller than the height that all girls have, i.e., the height of the shortest girl; this is clearly not an available reading of the sentence. The other examples are interpreted analogously.

The wide scope of the quantificational DPs in (33) and (34) is surprising, given the syntactic structure of the sentences. Ordinarily, such quantificational elements scope no higher than the clause that immediately contains them. Schwarzschild & Wilkinson (Reference Schwarzschild and Wilkinson2002) and Heim (Reference Heim2006) offer different but related solutions to the problem, the basic thrust of which is that it is not the quantificational elements themselves that raise to take wide scope, but the than clauses that contain them. With the entire than clause raised to a position above the comparative morpheme, any quantificational DPs within that clause take the comparative operator in their scope at LF.

On the degree complement analysis of nominal AICs, we would be led to expect the same scopal behavior for quantificational DPs in nominal AICs, as on this analysis nominal-AIC infinitival clauses merge in the same syntactic position as comparative than clauses. The examples in (35) below demonstrate that this is clearly not the case. Note that, as above, the semantic representations in (35) are incomplete, and are intended to show the scopal relationships between the positive morpheme and quantificational elements within the infinitival clause. I use the expression max(λd … to-read(a d-long book)…) as a place-holder to indicate the maximal degree of length – i.e., the standard of length – for what one can reasonably be expected to read; for more on maximality in standards of comparison, see von Stechow (Reference Stechow1984) and Rullmann (Reference Rullmann1995). The symbol >! indicates the ‘significantly exceeds’ relation proposed for positives by Fara (Reference Fara2000). A detailed compositional semantics for nominal AICs will be given in Section 4.

1. (35)

   1. (a) Middlemarch is a long book for every student to read.

      M's length >! max(λd.∀x[student(x)→to-read(a d-long book)(x)])

      ‘Middlemarch is longer than what one can reasonably expect all of the students to read.’

   2. (b) Middlemarch is a long book for more than two students to read.

      M's length >! max(λd.|λx.student(x)∧to-read(a d-long book)(x)|>2)

      ‘Middlemarch is longer than what one can reasonably expect (any group of) more than two students to read.’

These examples must be interpreted with surface scope for their infinitival clauses, or more precisely for the quantificational DPs within their infinitival clauses. The sentence in (35a) means that Middlemarch is longer than the standard for what one can reasonably expect all of the students to read; i.e., for the student who can read the least, it exceeds the length of the longest book that one can reasonably expect that student to read. Unlike the reading that would arise if every student took wide scope with respect to the positive morpheme, it does not rule out the possibility that there are students for whom it is perfectly reasonable to expect that they can read Middlemarch; it simply says that not all students fall into that category. The sentence in (35b) is interpreted similarly. It means that, of all the students, one cannot reasonably expect more than two of them to read Middlemarch. This is because the standard of comparison refers to the maximal degree of length d such that one can reasonably expect (any group of) more than two students to read a d-long book. Crucially, the sentence does not mean that, for some group of more than two students, one cannot reasonably expect them to read Middlemarch; this weaker reading would result from the quantificational DP more than two students taking wide scope with respect to the positive morpheme.

The fact that quantificational DPs in nominal-AIC infinitival clauses can only be interpreted with surface scope, and not with wide scope with respect to the positive degree morpheme, is a serious problem for the degree complement analysis of nominal AICs. If nominal-AIC infinitival clauses are structurally analogous to finite than clauses – that is, if they merge as complements of Deg⁰ – then they should participate in the kinds of scopal interactions observed for comparative than clauses. On the relative-clause analysis of nominal AICs, by contrast, the restriction to surface scope is a natural consequence of the merge position of the infinitival clause. This clause is interpretable in its base position, adjoined to NP, and therefore is not subject to QR for interpretation.Footnote ¹⁹

There is, however, at least one exception to the generalization that quantificational DPs in nominal AICs always take surface scope. With focal emphasis, individual DPs may take wide scope with respect to the positive degree morpheme. Compare example (36) to (35a) above; emphasis is indicated in (36) by capitalization.

1. (36) Middlemarch is a long book for EVERY student to read.

   ∀x[student(x)→M's length >! max(λd.to-read(a d-long book)(x))]

   ‘Middlemarch is longer than what one can reasonably expect any of the students to read.’

Focal emphasis on the determiner every in (36) allows the DP every student to take wide scope with respect to the positive morpheme. On the face of it, this would appear to violate the generalization seen above, according to which nominal-AIC infinitival clauses (and the quantificational elements contained within them) always take narrow scope with respect to the positive morpheme. Two facts, however, suggest that this reading is not evidence of a comparative-like structure for nominal AICs. First, the force of the default modality associated with the infinitival relative (indicated above with the notation to-read) is the same in both (35a) and (36). If the entire infinitival clause were raised to a position above the positive morpheme, then we would expect all quantificational elements within – i.e., both the DP every student and the default modal – to outscope the degree head. The fact that the modal interpretation is the same in both examples, however, suggests that only the DP every student actually undergoes a scopal change with respect to the positive morpheme. Second, the wide-scope reading for every student seems to require the phonological emphasis indicated above, which in turn suggests that its wide scope is the result of an association with focus. Wide-scope readings for quantificational DPs in ordinary comparatives like (33) and (34), by contrast, require no such focal emphasis.

The scopal relationship between nominal-AIC infinitival clauses and the positive morpheme thus argues against the degree complement analysis of nominal AICs. In combination with the facts discussed in the previous subsection, this weighs heavily in favor of the relative-clause analysis.

3.3 Relative-clause structure and idiom interpretation

Finally, evidence from the interpretation of idioms in nominal AICs suggests that we must adopt the relative-clause analysis of their syntactic structure. Nominal AICs may involve displaced idiom chunks; specifically, certain idiom chunks may serve as the head of a nominal-AIC relative clause. On standard generative assumptions about the syntax of idioms, this means that a head-internal analysis of such relative clauses must be available. For the relative-clause analysis sketched above, this is a straightforward amendment; for the degree complement analysis, it is a structural impossibility. As I show in this section, the difference in the external syntax of the infinitival clause on the two analyses constrains the possible analyses of its internal syntax, with the idiom data favoring the relative-clause analysis.

Relative clauses have been analyzed as both head-external and head-internal in the generative syntactic literature. The locus classicus of the head-external analysis is Chomsky (Reference Chomsky, Culicover, Wasow and Akmajian1977). On this analysis, a null operator merges in the position of the relative-clause gap and undergoes A′-movement to SpecCP. The head of the relative clause merges outside the relative clause, and the relative clause right-adjoins to this NP. Head-internal analyses come in at least two flavors; of primary interest here is the ‘raising’ analysis of restrictive relatives, on which the relative-clause head originates in the gap position and raises to SpecCP; the relative-clause CP itself serves as the complement of a determiner.Footnote ²⁰Vergnaud (Reference Vergnaud1974) is recognized as the originator of this analysis by Kayne (Reference Kayne1994) and Bianchi (Reference Bianchi1999), who note its indispensability for dealing with idiom-chunk relative-clause heads (as well as its compatibility with Kayne's linear correspondence axiom, the centerpiece of his theory of antisymmetry).

The head-external and raising analyses of the internal syntax of restrictive relatives are sketched in (37), for the relative clause a book that I assigned.Footnote ²¹

1. (37)

   1. (a) Head-external analysis

      

   2. (b) Raising analysis

      

It is immediately apparent that the head-external and raising analyses do not fit equally well into the relative-clause and degree complement analyses of the syntax of nominal AICs. In the relative-clause analysis shown above in (23a), the infinitival clause and the NP book form a single NP constituent. The relative-clause analysis of nominal AICs is thus compatible with both the head-external and the raising analyses outlined above (although the CP is adjoined to NP in that tree, nothing would prevent the head noun from merging inside the relative clause and raising to SpecCP).Footnote ²² The degree complement analysis, sketched in (23c), is compatible only with the head-external analysis. The infinitival clause merges as the complement of Deg⁰ and undergoes extraposition to the right edge of DP. It is thus never in the local configuration with the head noun that is required in order to implement the raising analysis.

With these restrictions in mind, let us now turn to the idiom data. Nominal AICs in which an idiom chunk serves as the head of the infinitival relative clause are indeed grammatical, as shown in (38). If Bianchi (Reference Bianchi1999: 43–45) is correct that idiom-chunk relative-clause heads require the raising analysis of relative clauses, then these data provide strong support for the relative-clause analysis of nominal AICs.

1. (38)

   1. (a) That is a delicate nerve to touch.

   2. (b) That is a big albatross to have around your neck.

   3. (c) That is a hard bargain for you to drive with someone in such dire need.Footnote ²³

The VPs touch a nerve, have an albatross around [one's] neck, and drive a hard bargain have non-compositional, idiomatic meanings which are commonly thought to arise only when the verb and its arguments are merged into the derivation together.Footnote ²⁴ On the raising analysis, the idiom-chunk head of the relative clause originates in its required base position within the idiomatic VP before raising to SpecCP. The head-external analysis does not allow such a derivation, and thus incorrectly predicts that idiom chunks should be impossible as relative-clause heads. Under these assumptions, and given the acceptability of the idiom-chunk nominal AICs in (38), we must choose the relative-clause analysis of nominal AICs, as it is the only one compatible with a raising analysis for relative clauses. The idiom-chunk nominal AICs cast serious doubt on the viability of the degree complement analysis sketched in (23c), as this analysis does not allow the raising structure required for idiom interpretation.

Note that this does not mean that the raising analysis must be employed for all nominal-AIC infinitival clauses. As mentioned above, the relative-clause analysis of nominal AICs is compatible with both the head-external and raising approaches to relative-clause–internal syntax. The relative-clause analysis of nominal AICs is likewise compatible with many different positions of adjunction for the relative clause itself, as discussed at the beginning of Section 3 and in footnote 16. Moreover, a one-size-fits-all approach to relative clauses has been argued to be inappropriate by Hulsey & Sauerland (Reference Hulsey and Sauerland2006), who propose that at least the matching and raising structures must be available. With the exception of the idiom cases, the nominal-AIC data do not compel a choice among the various structural possibilities in particular examples, and it is not clear to me that it is necessary or appropriate to designate one of them as the unique structure for nominal-AIC infinitival clauses. In proposing the relative-clause analysis, I thus advocate not a single structure for nominal-AIC infinitivals, but a family of related relative-clause structures.

4.1 The semantic framework

For my implementation, I adopt the degree-based semantics for comparatives developed by Kennedy (Reference Kennedy1999). Kennedy's framework differs from many other degree-based approaches in that it is non-quantificational, taking the meanings of Deg⁰ heads like more/-er to involve simple comparison of degrees, rather than quantification over sets of degrees. Such an approach is satisfactory for nominal AICs, as there are no scopal interactions between the positive Deg⁰ head pos and other quantificational elements that might require an analysis involving degree quantification (as discussed above in Section 3.2).

The semantic analysis of gradability and comparison benefits greatly from the addition of degrees to the semantic ontology, and degree-based analyses have been used for many years (Cresswell Reference Cresswell and Partee1976, Hellan Reference Hellan1981, von Stechow Reference Stechow1984); a particularly useful discussion of the benefits of degree-based analyses is found in Chapter 1 of Kennedy (Reference Kennedy1999). Degrees are introduced as a solution to the problem of gradable predicates, which cannot be modeled via simple set membership like ordinary first-order properties. Whereas set membership is a binary relation that allows for no comparison between members – an element is either a member of the set or not, and is no more or less a member than any other member – gradable predicates require greater semantic flexibility. One person may be taller than another even while both, or neither, count as tall. Degree-based analyses handle such phenomena by having gradable predicates (such as the adjectives found in nominal AICs) relate the subjects they are predicated of to (sets of) degrees on a semantic scale.Footnote ²⁵ For example, the gradable predicate tall relates its subject to a degree on the scale of height, or vertical extent. Comparatives like taller then effect a comparison of (sets of) degrees on that scale. To count as tall simpliciter is to correspond to a degree on the scale that exceeds a standard degree: the standard of comparison. A scenario in which A is taller than B but both A and B are tall is thus easily captured in a degree-based analysis.

In Kennedy's (Reference Kennedy1999) framework, gradable adjectives are modeled as functions, not from individuals to truth values like ordinary first-order properties (type 〈e,t〉), but from individuals to degrees (type 〈e,d〉). Kennedy calls these measure functions. Application of the gradable predicate tall to the argument Bob thus yields the degree of Bob's height. Deg⁰ heads like comparative more/-er and positive pos, meanwhile, denote functions that specify a relationship between that degree and another. The sentence Bob is taller than Susan is thus has the interpretation shown in (39). Gradable adjectives in the positive degree, as in the sentence Bob is tall, specify that the subject's degree (e.g., of height) exceeds a contextually defined standard value, labeled d _STND in (40). (In fact, as proposed by Fara (Reference Fara2000) and Kennedy (Reference Kennedy2007), it must exceed the standard by a ‘significant’ amount; this is indicated by the symbol >! in (40).)

1. (39) 〚Bob is taller than Susan is〛=tall(Bob)>tall(Susan)

1. (40) 〚Bob is tall〛=tall(Bob) >! d _STND

With this background on the Kennedy-style measure function approach to gradable adjectives, we are now ready to move on to the analysis of nominal AICs.

4.2 Composing nominal AICs

In this section I propose a compositional analysis of the predicative DP in a nominal AIC. I will assume, for simplicity, that the meaning of the larger sentence containing such a DP is derived by composing the matrix subject meaning directly with that of the predicative DP, which denotes a property of individuals (i.e., I follow Heim & Kratzer (Reference Heim and Kratzer1998: 62) in proposing that, in predicative DPs, the indefinite article denotes the identity function on properties). For reference, a type-adorned tree representing the predicative DP is shown in (41). Sister nodes compose by function application, with two exceptions: (i) the NP book and the infinitival CP, and (ii) the AP and its NP sister, both of which pairs combine by intersection.

1. (41)

   

4.3 Inappropriateness from modal standards

In this section I discuss the modality of the infinitival clause in nominal AICs and its role in producing the interpretation of inappropriateness associated with the construction. I propose that the infinitival is associated with a bouletic modality, adopting a well-established approach to the semantics of infinitival relatives (Kratzer Reference Kratzer, Eikmeyer and Rieser1981, Reference Kratzer, Stechow and Wunderlich1991; Hackl & Nissenbaum Reference Hackl and Nissenbaum2003). This modal interpretation, in combination with the comparison relation expressed by pos, gives rise to the inappropriateness reading.

Recall from our earlier discussion that one of the most salient interpretive features of nominal AICs is the sense of unexpectedness or inappropriateness associated with the construction. When we say that Middlemarch is a long book to assign, we understand that it is unexpected or inappropriate for Middlemarch to be assigned; Middlemarch is unexpectedly or inappropriately long for present purposes, namely for the purpose of assigning a book. One of the benefits of the semantic analysis of nominal AICs presented in this section is that it allows us to pinpoint the source of the inappropriateness: it is the presence of a ‘significantly greater than’ relation between the reference value and a standard value that refers to what is ‘to be done’. In examples above, I have abbreviated the infinitival meaning with the symbol to-assign. In this section, I propose a modal analysis of the infinitival clause that allows us to describe the standard of comparison in a precise way, and thereby to derive the inappropriateness interpretation.

The infinitival clause in a nominal AIC tells us something about what is likely, reasonable, permissible, or desirable given the facts of the world. It thus involves a circumstantial modal base with a bouletic modal ordering source (Kratzer Reference Kratzer, Eikmeyer and Rieser1981, Reference Kratzer, Stechow and Wunderlich1991; Hackl & Nissenbaum Reference Hackl and Nissenbaum2003).Footnote ²⁹ With a properly intensionalized implementation of our semantic analysis of nominal AICs, we can capture this meaning precisely. I propose that the infinitival clause lies within the scope of a bouletic modal, which specifies a modal ordering source, boul; when supplied with a world argument, it specifies a set of propositions – in this case, those consistent with what is likely, reasonable, permissible, or desirable – and imposes a partial ordering on the set of accessible worlds (the modal base) based on their compatibility with that set of propositions. For the proposition p that results from saturating the infinitival relative clause, this modal tells us that p is significantly more compatible with the bouletic ideal than the relevant standard. Formally, this is implemented by stating that the most boul(w)-compatible world in which p is true is significantly more compatible with the ideal in w than the standard for most-boul(w)-compatible worlds in which other possible propositions resulting from saturation of the infinitival relative clause are true.

Our place-holder expression to-assign, shown in (49a), can now be rewritten as in (49b). As the intensional interpretation in (49b) involves a comparison of best worlds with respect to the ordering source boul(w), I have abstracted out a ‘best world’ function, , in order to make the truth conditions in (49b) a bit clearer.Footnote ³⁰ (The symbol <! indicates the ‘is significantly more compatible with’ relation between worlds with respect to a given ordering source, indicated by the subscript; cf. the use of the symbol ⩽ in Kratzer 1981.)

1. (49)

   1. (a) Extensional

      〚[_CPPRO _arbto assign]〛=λx _e.to-assign(x)(a)

   2. (b) Intensional

      〚[_CPPRO _arbto assign]〛=

      λw _sλx _e.(w)(assign)(x)(a)<!_BOUL(w)s^w(λy _e.(w)(assign)(y)(a))

Paraphrased in English, the truth conditions in (49b) state that the infinitival CP to assign denotes the function f such that f(w)(x)=1 iff the world most consistent with boul(w) in which one assigns x is significantly more consistent with boul(w) than the standard for most-consistent-with-boul(w) worlds in which one assigns something. Note that this is precisely the interpretation we find for exhortative infinitival relatives, as well: in the sentence Middlemarch is a book to assign, we understand that assigning Middlemarch is significantly more consistent with the bouletic ideal than the standard for assigning books (leaving open the possibility, of course, that it may be even better to assign some other book). There is thus independent support for adopting the infinitival-relative interpretation shown in (49b).

The modalized expression shown in (49b) takes the place of to-assign in the composition of the nominal AIC. Importantly, this means that the contextually salient NP denotation that serves as the value for R (the domain restrictor for the standard-identification function s, discussed above) now contains a modal component. The full NP denotation – i.e., the intersection of the lower NP denotation and the adjoined infinitival CP denotation – is shown in (50).

1. (50) 〚book [Op _i [PRO _arbto assign t _i]]〛=λw _sλx _e.book(w)(x) ∧

   (w)(assign)(x)(a)<!_BOUL(w)s^w(λy _e.(w)(assign)(y)(a))

With this predicate as the domain restriction, the standard of comparison for our nominal AIC is modalized. In Middlemarch is a long book to assign, the standard of comparison is a standard of length for books that are assigned in worlds consistent with what is likely to be the case, what is typically the case, what might reasonably be the case, what one wants to be the case, or other situations specified by a bouletic modal ordering source. Note in addition that the base world for the modal is the same as the world of evaluation for the nominal predicate, book. This allows us to capture the fact that the matrix subject is asserted to be a member of the set denoted by the NP in the base world (in our default example, the actual world), a welcome consequence. If it is true in the actual world that Middlemarch is a long book to assign, then it must be true in the actual world that Middlemarch is a book.

By putting the modalized standard of comparison together with the ‘significantly exceeds’ comparison relation expressed by pos, we derive the inappropriateness reading of nominal AICs in a straightforward way. The interpretation of Middlemarch is a long book to assign – derived by integrating the modal analysis of the infinitival relative with the semantics for attributive positives developed in Section 4.2 – states that the length of Middlemarch significantly exceeds the standard of length for books that are likely to be assigned, are typically assigned, might reasonably be assigned, etc. This, I claim, is the source of the inappropriateness reading. To have a standard of comparison modalized in the way described here, and to assert that an entity significantly exceeds that standard, is to say that there is something unusual, unexpected, or inappropriate about the entity in question (or, more specifically, about its length or whatever other gradable property is involved) for the purpose at hand. Note that, depending on the context and the particular adjective and infinitive chosen, the interpretation in question may be closer to unexpectedness than to actual inappropriateness: an anonymous JL referee suggests the example She's a young woman to be elected Senator, which seems to convey surprise or unexpectedness rather than disapprobation. Given a suitably broad interpretation of the infinitival-relative modality (see footnote 29) and of the modalized standards that result from it, it should not be surprising to find just this sort of variation along the unexpectedness–inappropriateness spectrum. In all such cases, the interpretation in question arises naturally from the interaction between the comparison relation expressed by pos and the modalized domain restriction on the argument of the standard-identification function s.

Furthermore, we now have an explanation for why the inappropriateness reading disappears in the presence of an overt measure phrase, as shown above in (13). In such cases, the standard of comparison is provided directly by the measure phrase; the standard-identification function s plays no role in its determination.Footnote ³¹ The sentence Middlemarch is a 700-page-long book to assign thus means that Middlemarch is (at least) 700 pages long, and that it is a book that one should assign. With s out of the picture, the infinitival clause cannot perform the domain restriction that influences the standard of comparison and thus cannot give rise to the inappropriateness reading. Instead, it gets the exhortative reading, just like an ordinary infinitival relative adjunct. The disappearance of the inappropriateness reading in the presence of an overt measure phrase strongly supports the analysis developed here, in which inappropriateness is derived from the effect of the infinitival clause on the standard of comparison.

As an additional note on the composition of the infinitival clause, it is possible that the bouletic modal ordering source is a modal default associated with the complementizer for that heads the infinitival clause. First, we may note that for is clearly a complementizer in nominal AICs, and not a preposition, despite its absence from certain examples: infinitival relatives generally, including those in nominal AICs, omit for when there is no overt subject, but otherwise must have it. Its status as a complementizer is confirmed by its ability to introduce thematically deficient expletive subjects, as in examples like This is a small room for there to be so many chairs in. It has long been noted that the complementizer for has a distinct meaning associated with it that is semantically close to the modal meaning outlined here. Kiparsky & Kiparsky (Reference Kiparsky, Kiparsky, Bierwisch and Heidolph1970: 169) identify the quality associated with for as ‘emotivity’, writing, ‘[e]motive complements are those to which the speaker expresses a subjective, emotional, or evaluative reaction’. In their formulation, for complements are concerned with ‘the subjective value of a proposition rather than knowledge about it or its truth value’. Bresnan (Reference Bresnan1972: 84) writes that for complements can ‘be interpreted as describing unrealized states of affairs, both future and hypothetical’, and notes that this aspect of the meaning of the complementizer for has been recognized at least since the work of Jespersen. The modal interpretation associated with nominal-AIC infinitival clauses is very much of a piece with the overall semantic characterization of for offered by these authors, particularly with that of Bresnan. There is thus some independent support for the proposal that nominal-AIC infinitival clauses, which are headed by for, are associated with bouletic modality.

I have shown in this section that the inappropriateness interpretation of nominal AICs arises straightforwardly from the interaction between the ‘significantly exceeds’ relation denoted by pos and the modalization of the standard of comparison. A book that significantly exceeds the standard of length for books that get assigned in worlds conforming to what is likely to be the case, what is typically the case, what might reasonably be the case, what one wants to be the case, and so on, is a book that one ought not assign. I have shown here that this interpretive characteristic of nominal AICs, which on first inspection seems puzzling and idiosyncratic, follows completely naturally from the interaction of two independent factors: the meaning of pos and the modality of the infinitival relative clause. The fact that our analysis captures the inappropriateness interpretation in an independently motivated and compositional way lends strong support to the approach to nominal AICs developed here.

4.4 Modality beyond the standard

The modality associated with the infinitival relative clause shows up not only in the computation of the standard of comparison in nominal AICs, but also as part of the main nominal predication of the sentence. I suggest in this section that the modality may help to explain an important semantic difference between nominal AICs and attributive too constructions discussed above in Section 2.4.

Recall from the truth conditions for the nominal AIC Middlemarch is a long book to assign, repeated below in (51), that the intersective/conjoined NP-plus-CP denotation appears in two places:

1. (51) 〚poslong book to assign〛(〚Middlemarch〛)=

   long(Middlemarch) >! s(λx _e: book(x) ∧ to-assign(x)(a).long(x))∧

   book(Middlemarch) ∧ to-assign(Middlemarch)(a)

Here we have a conjunction of the NP meaning (book) and the infinitival-clause meaning, discussed at length in the previous section. Thus, when we say that Middlemarch is a long book to assign, apart from the comparison relation, we assert both that Middlemarch is a book and that it is something that gets assigned in worlds consistent with what is likely, reasonable, permissible, etc.

At first sight, this appears to pose a semantic problem for our analysis. How can we reconcile the inappropriateness interpretation that so saliently characterizes nominal AICs with the assertion that the matrix subject is something that is likely to be, typically is, or might reasonably be assigned? Doesn't this fly in the face of the intuition that if Middlemarch is a long book to assign, it is too long to assign?

I suggest that there is in fact no contradiction here, and that the semantic interpretation we derive for the complete sentence might actually help us to make sense of an important difference between nominal AICs and attributive too. The modalized infinitival clause is a component of both the domain restriction on the standard of comparison and the main sentence predication. It is thus simultaneously the case that the subject's degree of length significantly exceeds the standard of length for books that likely/typically/reasonably get assigned and that the subject is a book that likely/typically/reasonably gets assigned. The subject surpasses a standard of length for a category of things to which it itself belongs. Notice that there is nothing wrong with this schema in general: there is no contradiction whatsoever in something's exceeding the standard of length for books while itself being a book. Obviously, it is the modal meaning component that makes this situation appear contradictory: how can something exceed a standard for what is reasonable and still itself be reasonable? I suggest that as long as we accept that the standard value is not a maximum limit on what is likely/typical/reasonable, the threat of contradiction dissipates. The standard could be an average value for books that likely/typically/reasonably get assigned, or it could be higher or lower than the mathematical average (assuming that such a figure could be calculated to begin with). All that is required in order for the present analysis to go through is that the standard not be a value so high that anything higher is out of the ‘reasonable’ category. Moreover, as we have seen above in Section 2.4, nominal AICs in the past tense give rise to a strong inference that the proposition expressed by the infinitival clause is true, i.e., that the corresponding event actually took place. This is in contrast to what we find with attributive too, where we get the opposite inference:

1. (52)

   1. (a) Middlemarch was a long book to assign.

      inference: Middlemarch was assigned.

   2. (b) Middlemarch was too long a book to assign (it).

      inference: Middlemarch was not assigned.

The assertion that Middlemarch belongs to the category of books that are to be assigned, which is a component of the meaning of the nominal AIC on the analysis advanced here, appears to receive empirical support from the data in (52). I leave a more detailed semantic analysis of this past-tense paradigm to future research, but for now I believe that having the infinitival modality involved in the main sentence predication does not yield a semantically incorrect meaning for nominal AICs.