Navigation – Plan du site

Unclear Predication in the <Evaluative Adjective +Infinitive> Construction

Christopher DESURMONT


Cet article examine l’emploi d’adjectifs évaluatifs (Evaluative adjectives: EAs) comme brave, clever, crazy, foolish, silly, wise. La construction < NP + be + EA + toV> illustrée ci-dessus a été considérée comme un cas de « prédication épisodique » (Stage-level predication). Stowell (1991) par exemple propose une « coercion analysis » selon laquelle la proposition infinitive transformerait la nature du jugement adjectival: d’absolu et non borné, il deviendrait relatif. Cependant, Kertz (2006), prenant appui sur une batterie de « S-level tests », a démontré que cette construction est un cas de prédication d’individu (Individual level predication) et non de prédication épisodique (S-level); ce qui implique que la présence d’une infinitive ne modifierait pas fondamentalement la nature de la prédication. Cependant, considérer la nature de la prédication dans cette construction comme étant clairement I-level est manifestement contre-intuitif . Il semblerait que nous ayons affaire à ce que l’on pourrait appeler une « prédication incertaine » (unclear predication). Ce point de vue est ici défendu en impliquant les paramètres suivants: 1) le statut syntaxique de l’enchâssée, 2) son statut sémantique, 3) le préconstruit à l’origine du jugement adjectival.

Haut de page

Texte intégral

1. Introduction

  • 1  Here is a list of Evaluative Adjectives given by Barker (2002): brave, careless, clever, (in)consi (...)

1The starting point of this paper was Kertz’ (2006) “Adjunct Control analysis” of the “EA + toV” construction, an analysis in which the infinitive clause is seen as an AP adjunct (so neither a subcategorized complement, nor a VP-adjunct), and above all, an analysis demonstrating the I-level nature of the predication (Individual level). This analysis counters Stowell’s (1991) “coercion analysis” that says that combined with an infinitival clause (IC), the evaluative adjective1 is coerced to receive an S-Level reading (Stage Level). Kertz’s demonstration is based on a battery of classic S-Level tests going back to the 70’s.

2This paper aims at showing that although the I-Level diagnosis is quite convincing, all in all, the Derived EA cxn (NP + be + EA + toV) might be a case of “unclear predication”.

2. Two conflicting interpretations

3For Stowell (1991), although evaluative adjectives like silly or stupid are typical Individual-Level predicates (ILP), i.e. property-denoting, they are “coerced” to receive the Stage-level interpretation (S-level) when combined with an infinitive clause (IC). For Barker (2002), Feynman is stupid to dance like that (2b) does not entail that Feynman is stupid (2a). For Landau (2006), It was clever of John to leave the party (13b), signifies a property that is ‘temporally bound’. All of these statements imply a relativized or eventive reading of the EA toV construction.

4However, in so far as the impersonal It + be + EA + of NP + toV construction illustrated in (5) has more or less the same meaning as the personal cxn in (2), it might be useful here to quote Yagi (2006) according to whom this impersonal cxn does not receive a relativized reading:

3 - It is silly of him to believe her (10b)

4 - He is silly to believe her.

5‘The preposition of as in (10b) is used to specify who is silly, in other words, (10b) is a statement about him, not the event as a whole.’

6If this is so, and if (5) does have more or less the same meaning as (2), both utterances would appear to be ILPs, not SLPs (I-Level, not Stage-Level).

5 - It would be crazy OF YOU to sell today.

7One possible argument in favour of the I-Level interpretation might come from the IC’s presuppositional content. Sentence (4) for instance presupposes that “he believes her” is true, and the truth of this proposition belongs to the truth conditions of the adjectival predication. Because of its presuppositional status, the IC’s propositional content is not open to debate unlike the predication in the matrix clause; and this difference in status implies a presupposion boundary between the IC and the adjective that seems to give some credit to the ILP interpretation. However, as will be claimed in part 7, not all ICs in the personal “NP be EA toV” construction have a clear presuppositional status.

3. Applying the classic adjuncthood tests

8Stowell (1991) and Kertz (2006) both argue in favour of an adjunct analysis of the Derived EA’s IC, and Kertz claims that the EA’s infinitival is AP-adjoined, which implies a lower attachment site than for VP-adjuncts.

9To begin with, let us briefly apply the classic adjuncthood tests (summed up in Radford 1988, or Dowty 2006 for instance). The aim here is to show that these tests are generally inapplicable to the EA’s IC, and this may be partly due to the fact that the infinitive clause in this construction is not VP-adjoined.

    • 2  Disposition Adjectives.

    An adjunct is often said to be an optional element, while a complement is an obligatory element. This criteria is applicable to the EA/DA2 distinction: Jane is clever versus *Jane is eager.

  •  An adjunct ‘modifies’ the meaning of its head, while a complement ‘completes’ its head’s meaning. In Jane is clever to avoid answering such questions for instance, the infinitive clause does not complete the meaning of the EA, but it does introduce a restrictive clause which some would claim turns the absolute (intersective) reading into a relative (subsective) reading, i.e turns an ILP into a SLP.

    • 3  The temporal interpretation of the EA’s IC depends on the form of the main verb, and on whether or (...)

     The same adjunct combined with different heads affects their meaning in the “same” way semantically. (e.g. walk slowly vs. write slowly), but the same complement can have more radically different effects with different heads (e.g. manage to leave vs. refuse to leave). The EA’s infinitival does not affect the meaning of the EA, and the EA itself does not affect the meaning of the embedded infinitive clause except from a temporal point of view3.

  •  Adjuncts can be moved, while complements in English follow the governing head. The EA’s IC cannot be moved (*To worry, Jane is silly); yet as will be shown, other tests demonstrate that the EA’s infinitive clause is not a complement.

  • The Derived EA cxn does not allow recursive modification as is possible for locative adjuncts: *She is silly to worry to behave like this versus She has been working in the garden all afternoon. So this test cannot be called upon to demonstrate the adjunct status of the EA’s infinitive clause.

4. Comparing Evaluative and Dynamic Adjectives

10Evaluative Adjectives (EAs) and Dynamic Adjectives (DAs) have been usefully compared in the relevant literature because the EA and the DA constructions are the same and because the DA cxn is clearly eventive (SLP).

11EA and DAs differ in several ways. DAs (eager, keen, ready, reluctant, willing, unwilling, ...) are subcategorized with a clausal complement, EAs are not.

12EAs can stand alone in predicative position, DAs cannot and so require a “toV” complement; i.e, there are “basic EAs”, but no “basic DAs”:

6 - She’s clever. Basic EA

7 - She’s clever to pretend not to understand. Derived EA

8 - ?? She’s eager (to do what?)

9 - She’s eager to come DA cxn

13The complement nature of the “toV” constituent right of the DA (eager) is suggested by a comparison with a typical verbal complement:

10 - They want to come. Complement of Tr. V

11 - *To come, they want.

14The close syntactic link between V (want) and its complement would explain why fronting of the toV constituent is disallowed; and similarly, fronting of the DA’s complement is disallowed:

12 - *To come, she’s eager.

15However, fronting of an EA’s IC is equally disallowed (*To worry, she’s silly).

16Kertz (2006) suggests that this may be due to the low level at which adjunction takes place in the ”EA toV” cxn: the EA’s IC is claimed to be an AP-adjunct, not a VP-adjunct.

  • 4  ISLAND CONDITION (on Movement Rules)

17The adjunct analysis of the Derived EA cxn (Stowell, 1991) relies partly on evidence from patterns in wh-movement. The EA’s infinitival is affected by what looks like an adjunct-island effect4 in question formation, unlike the DA’s infinitival: ‘In interrogatives where the degree of the adjective is questioned, the infinitival clause in an evaluative control structure cannot prepose, but must remain in situ’:

13 - How stupid was John − to leave town

14 - *How stupid to leave town was John − (Stowell 1991, ex. 45-46)

18‘In contrast, the clausal complement of eager can prepose along with the adjective as a single constituent’:

15- -How eager was John − to leave town.

16 - How eager to leave town was John?

5. Adjunction sites (Kertz, 2006)

19Let us now quote the beginning of Kertz’ (2006) demonstration in favour of the claim that the EA’s IC is AP-adjoined. This passage is based on a comparison between two different types of modifiers to the right of evaluative adjectives, a PP-with-gerund modifier and an infinitival modifier.

  • 5  Nominalized DAs are followed by an infinitive complement, but nominalized EAs are modified by a ge (...)

20‘The infinitive control construction examined so far can alternate with a gerund construction similar to the one observed in nominalizations 5. In that construction, a gerund is embedded in a PP modifier. I propose a difference in adjunction site for the two structures, with the infinitival adjoining at AP, and the gerund at VP. The contrast is shown in (35)-(36):

17 - The government was [AP smart [CP to bring the trial to Houston]. (35)

18 - The government [VP was smart [PP in bringing the trial to Houston]. (36)

  • 6  However, Kertz suggests that should there be a difference in the degree of acceptability between ( (...)

21Kertz then compares the following sentences to show that in question formation, extraction is only fully acceptable out of the complement of the DA (39)6:

19 - # Where was the government smart in bringing the trial? VP adjunct (37)

20 - ? Where was the government smart to bring the trial. AP adjunct (38)

21 - Where was the government eager to bring the trial? Complement of DA (39)

22Evidence that the gerund adjoins at VP comes from parallels with other verbal predicates, as in (40), while evidence for the infinitival’s lower AP-attachment comes from examples like (41)-(42), which demonstrate that the gerund modifier, like other VP modifiers, can prepose to a sentence-initial position. The AP-adjoined infinitival, however, cannot prepose:

22 - The government [VP made a good decision] [PP in bringing the trial to Houston]. (40)

23 - In bringing the trial to Houston, the government was smart. (41)

24 - *To bring the trial to Houston, the government was smart. (42)

6. Applying classic ‘Stage-Level’ tests

  • 7  “Stowell’s analysis of the adj-to-inf-cxn, however, cannot be maintained; as effectively demonstra (...)

23The <EA toV> cxn raises what has been called the entailment puzzle, which, applied to the sentence Jane is silly to worry for instance means that this utterance does not entail that #Jane is silly. Apparently, what is being said here is that Jane is only silly in as much as she behaves like that, or for the limited duration of her regrettable behaviour, which implies a relativized reading. But the trouble with the relativized interpretation is that the derived EA cxn does not pass the classic S-level tests, as demonstrated in Kertz’ (2006) “Adjunct Control analysis”7. This analysis is based on a battery of well-established S-level tests going back to the 70s, some of which will now be recalled.

24● One test (Kratzer 1995) is based on the observation that the adverb OFTEN quantifies over events and so is only compatible with eventive predicates like disposition adjectives:

25 - He is often eager to discuss. DA

26 - *He is often crazy to say that. EA

25Knowing that the DA cxn is eventive (SLP), ungrammaticality in (26) suggests that this EA cxn on the contrary does not imply S-Level predication.

26However, the “often-test” obviously cannot apply if the infinitive clause denotes a unique event as in (27) which presupposes that “you (have) found us”:

27 - CMJ 525 ‘Your mother,’ she said to Anna, ‘thinks you are soclever to have found us.’

28 - Presupposition in the EA’s IC: You have found us. (unique event)

27● Another test (Milsark 1977) says that I-level predicates do not felicitously combine with an existential there subject; and indeed, (29) fails this “existential test”, while the DA cxn passes it as in (30):

29 - *There were peopleclever to find us. EA

30 - There were people eager to find us. DA

28● Another S-level test says that I-level predicates cannot embed under verbs of direct perception, and we see this in (32). By way of comparison, (33) shows that the DA construction can, as expected if this construction is essentially eventive (S-level):

31 - BP1 498 ‘Young lady, you areimpudent to ask, but since the matter seems important to you I'm prepared to say he won't be back much before Christmas.

32 - *We all saw the Young lady impudent to ask. EA

33 - We all saw the Young lady eager to ask. DA

7. Presupposition

29The Derived EA’s IC may have presuppositional status; thus, (34) presupposes (35):

34 - HA02081 It was silly PRO to worry.

35 - Presupposition: SOMEONE has been worrying.

30This is confirmed by the incoherence of sentence (36) in which the second conjunct contradicts the presuppositional content of the first conjunct:

36 - ?? It was silly to worry but no one has been worrying.

31With negation, the adjectival predicate is still factive, so (37) presupposes (38):

37 - AK9 1863 It's not terribly smart PRO to admit that you vote Tory.

38 - Someone admits they vote Tory.

32But (39) is another competing interpretation:

39 - Someone might be thinking of admitting that they will vote Tory.

33One may then wonder if there is any real presupposition when several implications are possible. When there is presupposition as in (34), there is what one might call a “presupposition boundary” between the embedded clause from the adjective and, as suggested above, this seems to give credit to the I-level interpretation. Presupposition in the embedded clause implies that Event time (temporal location of the event denoted in the EA’s IC) and Speech time (when the judgment is being pronounced) do not coincide.

34When the verb be of the matrix clause is in the conditional mood rather than in the indicative, presupposition disappears, and the event becomes hypothetical:

40 - FEE1239 Well, you know, you'd be silly to spend all that money, now wouldn't you?

35Clearly, (40) does not presuppose that #You have spent/are spending all that money, but implies this:

41 - You are thinking of spending all that money

36So what the speaker believes to be true (41) is merely implied in (40); it cannot be said to have presuppositional status in any strict sense of the word. This is shown by the semantic acceptability of (42):

42 - They would be silly to spend all that money but nothing proves that they have ever considered doing so.

8. Preconstruction

37Analyzing the impersonal “IT be EA PRO toV” construction as illustrated in (43), one may consider applying an “expectation operator” borrowed from “expectation modal logic”, as suggested by Oshima (2008).

43 - CS0 223 It is silly PRO to exaggerate differences when the similarities are also significant.

44 - In view of the normal course of events, ANY X who exaggerates differences when the similarities are also significant is silly.

38The empty PRO subject of the infinitive in (43) is not syntactically controlled; it is either a case of arbitrary control ( for anyone to ...), or a case of situational control (eg. for you to …).

39An interpretation as in (44) implies the existence of a preconstructed value-system. The predication made at the moment of speech appears to be essentially property-denoting (I-level), and is being applied to ANY X who has the property of “exaggerating differences when the similarities are also significant” (and who therefore identifies with the empty PRO subject). This predication however applies only “in view of the normal course of events”. This in turn would imply that in circumstances not belonging to the “normal course of events”, the judgment no longer applies. The judgment is mediated since any one particular person would need to identify with this ANY X class to qualify as a real target of the adjectival predication. This kind of interpretation may be applicable to the personal “NP be EA toV” construction.

45 - Youi are silly PROi to worry.

46 - (?) In view of the normal course of events, ANY X who worries in the present circumstances (previously mentioned or understood) is silly, and YOU have been identifed as belonging to this class.

9. Conclusions

40The “Adjunct Control Analysis” of the Derived EA construction (Kertz, 2006) demonstrates a low attachment site for the EA’s infinitival clause, apparently AP-adjunction, and pleads in favour of an Individual-Level interpretation of the adjectival predication (I-Level), i.e. essentially property-denoting rather than eventive. However, a strictly I-Level interpretation remains somewhat counter-intuitive. One is tempted to talk in terms of “unclear predication”, possibly as the effect of a complex combination of AP-adjunction, presupposition and preconstruction. The unclear nature of the predication between a relativized and an absolute reading (S-Level and I-Level) would then imply that the speaker can pronounce a judgment without clearly committing themselves.

41In the Derived EA construction, the infinitive clause (IC) is grammatically optional, whereas it is obligatory with a DA. The DA’s IC is a complement, while the EA’s IC is an adjunct, presumably AP-adjoined. The eventive content of the EA’s IC can be presupposed, while the DA is not a presupposition trigger. The close syntactic link between the IC (if it is indeed AP-adjoined) and the evaluative adjective pleads in favour of a relativized reading of the adjectival predication; conversely, a presuppositional boundary between the EA and the IC pleads in favour of an autonomous I-Level reading of the adjective. The sum of the above conflicting observations seems to indicate that the "EA toV" construction is not a clear case of Individual-Level predication, even if it passes none of the classic Stage-Level tests.

Haut de page


BARKER, C., (2002), The dynamics of vagueness, Linguistics and Philosophy, 25 (1): 1-36.

DOWTY, D., (2066), The Dual Analysis of Adjuncts/Complements in Categorial Grammar.

KERTZ, L., (2006), Evaluative Adjectives: An Adjunct Control Analysis. Proceedings of the 25th West Coast Conference on Formal Linguistics, ed. Donald Baumer, David Montero, and Michael Scanlon, 229-235.

KRATZER, A., (1995), Stage-Level and Individual-Level-Predicates, in Carlson/Pelletier, eds., The Generic Book, 125-175. University of Chicago Press.

LANDAU, I., (2006), Ways of Being Rude.

OSHIMA, D., (2008), Stereotypes, Desires and Constructions, Gronn, Atle (ed.): Proceedings of SuB12, Oslo: ILSO 2008, 470-484.

MILSARK, G., (1977), Toward an Explanation of Certain Peculiarities of the Existential Construction in English, Linguistic Analysis 3, 1-29.

RADFORD, A., (1988), Transformational Grammar. A First Course. CUP.

STOWELL, T., (1991), The Alignment of Arguments in Adjective Phrases. In: Rothstein, Suzan, (ed.), Perspectives on Phrase Structure: Heads and Licensing, Syntax and Semantics 25, Academic Press, 105-135.

YAGI, K., (2006), Patterning of English Adjectives – For the Improvement of Learners' Dictionaries, Asian Bilingualism and the Dictionary. Proceedings of The Second International Congress, Konsei University, Seoul, Korea: 112-118.

Haut de page


1  Here is a list of Evaluative Adjectives given by Barker (2002): brave, careless, clever, (in)considerate, courageous, cowardly, crazy, cruel, dumb, evil, foolish, impudent, (un)kind, (un)lucky, mean, naughty, nice, (im)polite, (im)prudent, right, silly, smart, stupid, wicked, (un)wise, wrong.

2  Disposition Adjectives.

3  The temporal interpretation of the EA’s IC depends on the form of the main verb, and on whether or not the adjectival predicate acts as a presupposition trigger.

4  ISLAND CONDITION (on Movement Rules)

Subjects and Adjuncts are islands; Complements are not.

The “Island Condition” recalled in Radford (1988: 487) implies that “constituents can be extracted out of Complement Phrases, but not out of Subject/Adjunct phrases”.

5  Nominalized DAs are followed by an infinitive complement, but nominalized EAs are modified by a gerund construction:
H8T 442 But they had also demonstrated, by their eagerness to deflect him from it, that his was no fool’s errand. J7B 1050 Vaisey J went on to say that the employers had shown “the utmost carelessness in allowing this defendant to get knowledge of what they knew was a very valuable secret…”

6  However, Kertz suggests that should there be a difference in the degree of acceptability between (37) and (38), this might be due to a lower attachment site for the infinitival modifier, though the suggestion is said to be speculative.

7  “Stowell’s analysis of the adj-to-inf-cxn, however, cannot be maintained; as effectively demonstrated by Kertz (2006), a number of diagnostics unanimously indicate that the adjective in the P1 slot of the “adj-to-inf-cxn” is individual-level, rather than stage-level.” (Oshima 2008)

Haut de page

Pour citer cet article

Référence électronique

Christopher DESURMONT, « Unclear Predication in the <Evaluative Adjective +Infinitive> Construction », E-rea [En ligne], 9.2 | 2012, mis en ligne le 15 mars 2012, consulté le 18 janvier 2017. URL : ; DOI : 10.4000/erea.2368

Haut de page


Christopher DESURMONT

Université Lille 3, UMR 8163, STL
Christopher Desurmont est maître de conférences en linguistique anglaise à l’Université Lille 3. Après une thèse sur l’ellipse en anglais contemporain (Paris 4, 1999), ses travaux de recherche sur la grammaire de l’anglais ont porté sur la détermination nominale (this/that, Ø, one); la catégorie de l’adjectif (ordres séquentiels, coordination, propriétés positionnelles, modalisations adverbiales, questions d’interprétation); les structures « adjectif + toV »; les figures de l’hypallage; la représentation des procès (locution just about); try and V / try toV.

Haut de page
  • Logo Laboratoire d’Études et de Recherche sur le Monde Anglophone
  • Logo DOAJ - Directory of Open Access Journals