Skip to navigation – Site map

A posteriori modality, implicative modality by abduction. A case study: he must have been drunk to have said that

Geneviève Girard-Gillet


The construction “he must have been drunk to have said that” has not been analysed in depth, even if it has been mentioned by P. Larreya, G. Furmaniak, E. Gilbert, among other linguists working on modalisation. It is particular in that it expresses, with an infinitival proposition, an event that took place before Speech Time, and, with a modalized proposition, the cause that, according to the speaker, can explain the occurrence. We are enlarging here the concept of a posteriori modalization, defined by Larreya (2000, 2009), to include cases in which the event that triggers the modal stance of the speaker is explicitly mentioned and not presupposed. This construction is highly constrained as it requires the presence of an epistemic modal form. It differs from constructions that have been more consistently studied, such as “John was stupid to refuse the job” (Stowell, Kertz, Desurmont). It is the nature of the cause-consequence link that enables one to construe a possible interpretation, and we consider that the inference is made by abduction (Pierce 1974, Desclés 2000). This explains that it is rare and highly expressive, since an abductive inference only yields plausible conclusions, which the speaker must ascertain as its own. The construction functions as a whole (syntactically, semantically, phonologically) and builds up its own interpretation, through the reconciliation of jarring viewpoints by the speaker. It expresses some kind of humour, irony even, and always denotes the speaker’s astonishment at what happened and at the reason why a particular individual took part in the event in question.

Top of page

Author's notes

We wish to thank Paul Larreya and Christopher Desurmont for discussing this construction with us and helping us with their comments. Any misinterpretations and shortcomings are of course my own. We also wish to thank two anonymous readers for their remarks and suggestions. Any remaining misinterpretations and shortcomings are of course my own.

Full text


  • 1 Furmaniak 2004, ex (82) p. 274.

1The construction we propose to study in this article is made up of a modalized proposition he must have been drunk, and of an infinitival proposition to have said that. The interpretation of the epistemic modal has been analysed in depth in various studies, but the interesting point here is, for us, the relationship between the modalized proposition and the infinitive, as the latter encodes an actual event (the state of affairs), which is held to be true. This construction has been noted by P. Larreya, G. Furmaniak and E. Gilbert, but no detailed analysis seems to exist in the literature. Huddleston and Pullum (2002) do not mention this use of an infinitive in their grammar. P. Larreya (2000) posits that you have to be mad to do that is a case of implicative modality, G. Furmaniak (2004) considers that you must be 18 to work here1 is an ambiguous sentence, out of a context, as it can refer to an obligation (root modality) or to a deduction (epistemic modality), and E. Gilbert (2001) compares he must have a computerised brain to have absorbed all that information with an if-sentence. In order to understand what motivates this construction, we shall first describe the respective role of the modalized proposition and of the infinitive, then we shall analyse the occurrences we have gathered, before discussing, in the third part, the semantic data that can account for the use of such a construction by the speaker.

1. Description of the construction

1.1. The role of the modal in the construction

  • 2 See also his contribution in this volume.

2The modal in this construction is interpreted as an epistemic modal. For P. Larreya, epistemic modality consists in the attribution of a truth-value to a proposition (or to the situation that constitutes the referent of that proposition). Within epistemic modality, P. Larreya (2004) distinguishes between problematic and implicative modality2. The sentence he must be tired is a case of problematic modality, and the sentence you have to be mad to do that is a case of implicative modality. Implicative modality consists in linking two propositions by a relation of implication: A implies B. A is the antecedent of the implication and B the consequent. In: you have to be mad to do that the antecedent is X does that and the consequent is X is mad. The meaning can be glossed as [[you do that] IMPLY [you are mad], or being mad is a necessary condition for doing that. In an a posteriori modalization the modal judgment bears on a situation whose reality (or unreality) is known (Furmaniak and Larreya, to be published).

3The construction under study differs from cases of problematic modalization, as regards the antecedent of the modalization. The problematic modality is constructed by the speaker in a given situation, but the situation that leads to the modal stance, the antecedent, is not always given. The speaker can say John must be tired, if he noticed that John is always late, but he will not feel it necessary to encode explicitly I think that John must be tired, because I noticed that he is often late. Delechelle (1989: 412) considers that the explicative because does not express the cause of the modalized sentence, but gives the reason why the speaker feels justified in saying that John is surely tired. E. Sweetser (1990: 61) agrees with him, and posits that I can say You must have been home last night, if “the available (direct) evidence compels me to the conclusion that you were home.” But the evidence seems to be rarely expressed in case of a problematic modality. In the construction we wish to account for the evidence IS always mentioned, and it is mentioned with an infinitival clause.

4With the implicative modality, P. Larreya acknowledges that the antecedent of the implication is mentioned, and it is in most cases abductive (the cause-consequence relation is reversed): the speaker infers the cause (X is mad) from the consequence (X does that). Paradoxically it is the consequence (not the cause) which is the antecedent of the relation: X does that → X is mad. For Larreya, it is a strong relation, and it involves absolute truth-values (TRUE and FALSE). It is this characteristic that clearly differentiates it from problematic modality. He adds that the problem with implicative modality is that it does not make sense to say that X does that is a sufficient condition for [X is mad]. What is understood is I believe/ I say that, which means: ‘X doing that is a sufficient condition for me to believe that X is mad’. We do not entirely agree with him here, because we do not think that we have what he calls a ‘strong relation’. An inference by abduction does not lead to a general law, but to a plausible hypothesis, unlike inferences by induction (see Pierce 1974, Desclés 2000). In the case of an induction, we start with already observed correlations, and we generalize the correlations to formulate a general law:


{P (ai) & Q (ai) ;
i : 1, ... n}
(∀x) P(x) => Q (x)

5In the case of an abduction, we start with the acknowledgement of a given state of affair, q, we know, from a general law, that p causes q, and we conclude that it is plausible that we have p:


(∀x) P(x) => Q (x)
plausible (P (a))

6The plausibility of the cause that leads to the state of affairs explains, paradoxically, why a modalized proposition is compulsory, and even only one expressing a high degree of certainty, namely, must or certainly. We did not find any hit in the Corpus of Contemporary American English, henceforth referred to as COCA for [he might have been drunk to V], or [he may have been drunk to V]. There are no occurrences for [he may be crazy to V], whereas there are 8 hits for [he must be crazy to V]. Our hypothesis is that the speaker resorts to an element expressing a high degree of certainty to counterbalance the fact that the inference by abduction does not guarantee a logical cause-consequence relation in the situation in question. In John must have been drunk to have said that, the state of affair in the reality space (Fauconnier 1984) may be explained by the inebriety of John, but other causes may be just as logical; John might have been exhausted, depressed, angry. To make himself understood, and to force the addressee to accept his interpretation, the speaker feels it necessary to eliminate any possible doubt on his part, as to the causal state of mind leading to the noticed event, hence the use of must or certainly.

(1) It's a way to hold Jack near me, to still feel his tolerant affection. An invisible shirt of comfort that I can still wear. I must be crazy to admit this. (COCA, FIC, 1992, Twelve Golden Threads, by Aliske Webb)

(2) Bellandra replied in a surprisingly gentle manner. Socrates thought that she must have been drunk to be so friendly like that. (COCA, FIC, 2000, Walkin’ the Dog, by Walter Mosley)

  • 3 The fact that there are very few occurrences in the COCA for this construction compelled us to reso (...)

7We looked for other occurrences on Google3, and we drew the same conclusion. Google gives 4 examples for [might have been drunk to V], none for may, whereas there are over 400 000 hits for [must have been drunk to V]. What is clear is that the construction is not possible without a modalized clause:

(4) *He was drunk to have invaded Irak.

8This sentence can be compared with:

(5) He was stupid to refuse the job.

which is felicitous without or with a modal:

(6) He must have been stupid to refuse the job.

9The difference between drunk and stupid is that drunk expresses a state that can be objectively noticed, because of the behavior induced by the quantity of alcohol in the individual’s blood, whereas stupid refers to the lack of judgment that the speaker thinks is the characteristic of the individual in question. Needless to say that the criteria to acknowledge a stupid behaviour are difficult to define and vary from one person to another. The attribution of this property is then highly subjective. But this semantic difference does not actually operate, because drunk, which is thought to be objective cannot appear, in our construction, as an assertion attributing the property to the individual considered. For a discussion of the interpretation of evaluative adjectives, see Desurmont (2008, 2012). The solution is to be found in the distinction between how an event and an individual can be evaluated. An event can be deemed stupid, intelligent, but not *drunk, and the assessment of the event is metaphorically transferred to the individual concerned, who is thought to have been stupid, intelligent: it was stupid of him to refuse the job, it was intelligent of him to refuse the job, but *it was drunk of him to refuse the job.

10This construction, exemplified in (5), has often been studied in the literature, the debate being whether the property was only present when the character refused the job (state-level interpretation, in Stowell 1991) or whether the stupidity was an inherent characteristic (individual-level interpretation in Kertz 2006) of his. We shall tackle this question when we analyse the semantic interpretation of the whole construction, in relation with the adjectives present.

1.2. The role of the infinitive

11The other important point is the presence of an infinitive to refer to the actual event which is the antecedent of the modalized proposition. In the following sentence:

(7) He [G.Bush] must have been drunk to have invaded Iraq.

it is common knowledge that Bush invaded Irak, or at least that he declared war on the country. This point is not debatable, and the infinitive provides the factual context that will contribute to interpreting the sentence correctly. The cause-consequence relation is: I know, as everybody does, that Bush invaded Irak; hence I want to find a reason for his making this decision, and the one I choose among many other possible reasons is that he must have been drunk, even if I am not sure that this reason is the right one. In this example, it is clear, of course, that the speaker suggests this possible explanation – tongue in cheek – as it is common knowledge that Bush had a drinking problem, but the context alone cannot account for the construction. We shall discuss this point in part 3.3.

12This construction is not the only construction including an infinitive which is interpreted as encoding an actual event. An infinitive does not carry any tense marker, and its temporal interpretation depends entirely on the tensed proposition it is associated with. Most studies on infinitives (Duffley 1992, 2007, Egan 2006, 2008, Deschamps 1997, 2006) focus on complement infinitives, such as John wants to succeed, or Mary wishes to visit Greece, and the prospective interpretation is only due to the semantic features of want and of wish, since they both imply a future realisation of the content of the infinitive. Other authors (Boulonnais 2004, 2008, Girard-Gillet 2008, 2011) consider that it is the function of the infinitive in the sentence that must be taken into account, along with other semantic parameters, to construe the relevant temporal and aspectual interpretation, as well as its truth-value. In the following example,

(8) ‘You have some scars on your left lung.’
He was meant to understand something, but if that was a flicker of irritation on the medical officier’s face, he needed more clues. To be considered dim pushed him even more into thoughtlessness (S. Sillitoe,
The Open Door, p 8),

the infinitive, to be considered dim, which is the propositional subject, posits that the character did feel that he was considered dim by the medical officer.

13In the construction we are dealing with, the infinitive is neither the subject nor the complement of any lexical verb, but is interpreted as the basis of the speaker’s judgment as regards the circumstances of the event he is interested in. The infinitive does not function as a cause adverbial, as the following paraphrasis sounds strange:

(9) ? He must have been drunk, since/as he invaded Irak.

even if the cause-consequence relationship is more or less the same. A since-sentence presupposes (Delechelle) the existence of the actual event, namely that he invaded Irak, just as the infinitive does. The infinitive is not a purpose clause either:

(10) ? He must have been drunk in order to invade Irak.

14If we follow Whelpton (2000)’s description of infinitival clauses, the construction is neither a Rationale Clause – such as:

(11) John1 bought his friends some champagne [(in order) PRO1 to show his generosity].

nor a Purpose Clause,

(12) John1 bought his friends2 some champagne3 [PRO2 to take e3 to the party].

or a Telic clause,

(13) John1 bought his friends some champagne [only PRO1 to discover that they didn’t drink].

even if the telic clause denotes an acknowledged state of affairs, John discovered that they didn’t drink.

15Another modification, with a time adverbial, is not any better, as it would modify the temporal chain of events:

(14) He must have been drunk when he invaded Irak.

and would encode the simultaneity of the invasion and of his state of drunkenness.

16The infinitive is part and parcel of the whole construction and it derives its interpretation from the structure itself. The construction is not compositional; the infinitival clause is not embedded in a matrix clause, and it is hardly possible to say that it functions as an adjunct. This suggests that we are dealing with a construction that has to be interpreted as a whole, following Croft (2001: 28), who considers that ‘any node with a unique, idiosyncratic morphological, syntactic, lexical, semantic, pragmatic, or discourse-functional properties is a construction’. This construction is indeed syntactically, semantically and phonologically constrained:

17syntactically: the linear order is compulsory, or at least favoured, since a fronted infinitive sounds awkward:

(15) ? To have said that he must have been drunk.

18semantically: only some types of adjectives are possible since,

(16) He was pleased to have said that.

suggests a different cause-consequence relation: the infinitival clause [he-be-pleased] must be interpreted as the consequence of the event [he-say-that], and not as its cause. This issue will be addressed in part 3.

19phonologically: the prosody requires a pause between the modalized proposition and the infinitive, whereas there is no pause in he was pleased to have said that. The adjective is stressed and the tonic stress is a rise-fall to enhance the speaker’s expressivity.

2. The data

  • 4 We wish to thank Vincent Hugou for helping us to collect the data.

20It was very difficult to collect data as the construction seems to be quite rare. In the last twelve novels we have read, we have not found one single occurrence. The Corpus Of Contemporary American English did not prove as useful as we hoped. The request [must have been all Adj to V] yielded 1728 hits, and out of the first 200 occurrences we analysed only 13 occurrences were relevant,4 8 with crazy, 3 with insane, 1 with sick, 1 with drunk, after eliminating those that construed a different meaning. Here are a few of them:

(17) Kosnik was driving here to work on his deathday and was killed in a car crash. We don't allow our employees to work on their deathdays, much less encourage it. He must have been crazy to take such a risk. (FIC, 2001, Happy Deathday, by Robert Scherrer)

(18) As he waited for the time-comp to bring them back to normal space, he decided he must have been insane to even think of going to Halmarin-IV. (FIC, 1011) Starfire Down, by Lee McKeone)

(19) Bellandra replied in a surprisingly gentle manner. Socrates thought that she must have been drunk to be so friendly like that. (FIC, 2000, Walkin’ the Dog, by Walter Mosley)

21Among the eliminated ones, were the following:

(20) When international reporters showed up to interview Frederica or Lal, they must have been surprised to find the newspaper's embattled journalists engaged in a fierce contretemps. (ACAD 2011 AmerScholar, Living Dangerously, by Michael Hardy)

(21) The chroniclers must have been careful to hide it. (FIC, 1992, The Ring of Winter, James Lowder.)

22It is the semantic features of the adjectives, surprised, careful, that seem to trigger the different construal. We shall address this issue in the next part, keeping in mind that it is necessary to rule out rigid semantic features (Sweetser 1990: 14, Culioli’s concept of notional domain 1992).

23Out of the request [must be all Adj to V] we retrieved 5 occurrences, among which:

(22) The old lady said vehemently. ‘The wretched creature must be demented to keep on raking it up. It is finished!’ (FIC, 1993, Farriers’Lane, by Anne Perry)

(23) I'm not completely okay with the idea, you understandwe must be nuts to let an amateur take this onbut we're going to do it. (FIC, 1996, Lost Angel, by Marilyn Wallace)

  • 5 We started our search with adjectives describing negatively-oriented physical, psychological or men (...)

24The lack of a high number of occurrences, as no hits were found for silly, barmy, foolish, daft, mad, depressed, to V,5 in the COCA, compelled us to resort to Google, and mainly Google books, whenever possible, to obtain examples with these adjectives in order to be able to deal with the question of their semantic type. We are aware of the fact that the search with chosen adjectives constitutes a bias, but if the results turn out to be as expected, they can be considered as evidence of the interest of our hypothesis. We mean that our hypothesis can then predict how the sentences are interpreted. After checking that the occurrences came from native speakers, we retrieved, among others:

(24) Your friend must have been silly to advertise it in the local paper, that's asking for trouble.

(25) I didn't like this place one bit and must have been barmy to have come here.

(26) Well, he must be barmy to want you to stay with him. (

(27) But what? Rozalin sighed. She must be foolish to believe in dreams. They are simply memories and thoughts mixed up and arranged during one's deepest sleep. (

(28) It was as if she was judging Vidia, thinking she must be daft to have even thought of asking that.

(29) Surely he must be daft to torture himself so? At last, he moaned again and went still. She rewarded him for his efforts with a sigh of sympathy.

(30) And suddenly it occurred to him he must have been mad to marry. (

(31) It seems that the general consensus is that he must have been depressed to do what he did.

25Other adjectives can appear in this construction, while construing a similar semantic interpretation. This suggests that it is the speaker’s stance as regards the event he/she noticed that triggers the construction.

(32) So we started off and the rain came on. I persisted however and told Max of the many beauties of he moor, which he could not quite see. [...] He certainly must have been fond of me to have endured it and preserve his air of enjoying himself. (A. Christie, An Autobiography, p. 428)

(33) John and I went to Paris on birthday money he received from a relative. He must have been fond of me to spend that money. He let me gave all the Banana milkshakes I wanted. (Paul McCartney)

(34) Despite the cold, he must have fallen asleep for a while, for the sun had deserted him and when he’d looked out he’d seen his mother removing the chainail from the rocker. He must have been very tired to have slept through that noise. (C. Robb, The Guilt of Innocents)

(35) He realized that he must have been very tired to have slept so long, he did feel refreshed. He had no problem going through passport control and customs. (H.G. Poyner, Mary, Touched by a Ring)

(36) Having spent a lifetime with people who tell me I must be old-fashioned to care about etiquette, could, if I were not so polite, turn around and say, ‘Nyah nyah nyah nyah nyah, you're the one who is old-fashioned if you think that etiquette is old-fashioned – you obviously don't spend time on the Internet.’

26We can hardly consider that tired, or old-fashioned are as negatively-connoted as mad, or drunk, and fond is certainly positivily-evaluated. Therefore we can wonder why they can appear in this construction with the same interpretation of surprise and unexpectedness.

3. The semantic interpretation of the whole construction

3.1. Previous analyses

  • 6 Ex (77) p. 262 in his Ph. D. Dissertation

27Some previous studies need to be referred to, before we develop our tentative interpretation of the construction. Furmaniak (2004) analyses the following example:6

(37) ‘You must be stark raving mad’, Paula burst out as Tweed drove them back to their hotel. ‘Giving all that information to that tart’.

and following P.S. Tregidgo (1982: 78), who posits a DEMAND relationship, he paraphrases (37) as follows:

  • 7 Tregidgo says that ‘when we say that something must be true, we mean that this conclusion is demand (...)

(37’) [you gave all that information ] DEMAND [you-be-stark raving].7

28The evidence demands that Paula should state that Tweed is mad, and the evidence is given afterwards in an ING-clause. Such a construction does exist, but is rare. Google gives only 4 hits for he must have been drunk V-ING.

(38) He must have been drunk doing something like that. (​.../​status/​491315393769005056)

29E. Gilbert (2001) compares the construction with an if-constative sentence:

(39) He must have a computerised brain to have absorbed all that information.

(40) Boy, you must have had some really bad experiences with women to make you so distrustful.

and says that whether the content of the proposition takes up the form of a hypothesis or of an infinitive, in each case there is a property that constitutes the notion that is construed by the predicative relation containing must. It is very close to what J. Piaget calls “significant implications” where pq is necessary.

3.2. Comparison with slightly different constructions

30English speakers have similar constructions as their disposal to qualify someone’s attitude, and in particular the construction mentioned above: John was stupid to refuse the job. Unlike the construction we are dealing with here, this construction has often been analysed and it is time to sum up T. Stowell, L. Kertz and C. Desurmont’s viewpoints on the interpretations construed. The difficulty lies in the existence of two possible readings for evaluative adjectives, the Stage-Level interpretation and the Individual-Level interpretation. T. Stowell (1991) considers that the presence of the infinitival leads to a relativised interpretation of the adjective in: Hoffman is smart to cut a deal, and does not entail that Hoffman is smart. The infinitival clause is realized syntactically as an adjunct, and not as a complement, and T. Stowell suggests that the infinitival coerces a Stage-Level (eventive) reading. L. Kertz (2006) argues against this interpretation and opts for an Individual-Level interpretation, after proposing several syntactic tests. As for C. Desurmont, he considers that an Individual-Level interpretation is counter-intuitive, since the presence of the infinitival clause does not make sense, if it is does not limit the scope of the property defined. He prefers to talk of ‘unclear predications’, meaning by that that speakers can pronounce a judgment without clearly committing themselves. We agree with him, and will go so far as to say that when the speaker says that John was stupid to refuse the job, not only does the speaker accuse him of acting stupidly on this particular occasion, but he also considers that John is prone to act stupidly; otherwise he would not have used such an adjective. In other words, the accusation may have been partly motivated by what the speaker generally thinks of John.

3.3. The interpretation of the adjectives

  • 8 We cannot give a frequency analysis, but only a tendency, even if Mair (2007) showed in his analysi (...)

31To what extent does the discussion of evaluative adjectives help us for our construction? In our examples, not all adjectives are evaluative, but the question remains whether we have a State-Level or an Individual Level interpretation. The very frequent presence of drunk in the modalized proposition (2 300 000, occurrences) in Google vs. a more limited use of mad (230 000 examples)8 or of daft (94 000) or of silly (300 000) tend to show that what is highlighted is a momentary state affecting an individual, and a state that the speaker posits as debatable. We must keep in mind that drunk cannot appear in the following constructions encoding situations, unlike silly or daft:

(41) It was silly of him to take offence at what she had said. (

(42) Ashley, for his part, kept reassuring his mother that it was daft of him to have left home. (

(43) *It was drunk of him to invade Irak. / *To invade Irak was drunk.

32What is then the semantic link that unites the infinitive and the modalized proposition? It is here that the abductive inference plays its role, in the evaluation of the subjective stance adopted by the speaker. In most examples, the situation (state of affairs) noted by the speaker is a bizarre, strange, unexpected situation, at least for him (speaker’s belief space), when he considers the individual taking part in it: to take a risk (17) to go to Halmarin IV (18), to rake up something (22), to advertise in a local paper (24), to believe in dreams (27). As the speaker cannot understand why the individual he is interested in acted as he did, he imagines a strange reason, a reason that can only explain a momentary lack of soundness. Let us remember that the abductive inference is not based on a necessary law but on what the speaker considers a possible law, in his own mental construct, unlike a deduction. As we tend to think that a state of madness can lead to foolish actions, we are prone to conclude, from what we consider a senseless process, that it was caused by a deranged mind. But this does not rule out other plausible causes which might explain what actually happened. The logical/illogical link is close to a fallacy – that is the use of an incorrect argument that results in a lack of validity. To come back to Larreya’s explanation, the speaker infers the cause (X is mad/drunk) from the consequence (X did that). Paradoxically it is the consequent (not the cause) which is the antecedent of the relation: X does that → X must be mad/foolish/drunk.

33But resorting to someone’s negatively-oriented state of mind to explain a baffling event is not entirely satisfactory, since example (32) runs counter to such an interpretation. Let us note in passing that *he was fond of me to have endured it is not grammatical. A. Christie is indeed happy that Max, whom she will soon marry, is enjoying their walk in the moors even if it’s raining. Why does she resort to this particular construction? We think that she first considered that walking in the rain is always unpleasant and that Max should have asked to go back home; but then she realizes that far from feeling bored, he enjoys himself. An explanation is then required to reconcile these two jarring situations. The solution she comes up with is that he must be fond of her. Only someone fond of her would agree to go on walking in the rain. But she could have imagined other reasons, if she had not been in love with Max. This means that the situational context may sometimes play a role, and it does play a part here for the choice of the property ascribed to Max. The construction is derived from the taking into account of the following stages:

  • stage 1: real world level (Sweetzer 1990), reality space (Fauconnier 1984): walking in the rain;

  • stage 2: epistemic level1: everybody thinks that walking in the rain is unpleasant;

  • stage 3: real world level: Max enjoys walking in the rain;

  • stage 4: epistemic level2: a clash exists between stage 2 and stage 3;

  • stage 5: epistemic level3: a possible explanation to reconcile them is necessary;

  • stage 6: epistemic level4: Max must be fond of me.

34We understand with this example that it is the comparison of the various semantic data contained in the construction that, a posteriori, gives rise to the reader’s interpretation. But although the context does play a role here, it is not always so. Example (30), John must have been mad to marry suggests that marrying is a negative process for the speaker (speaker’s belief space), because of the negative mental state, mad, which is the speaker’s explanation, even if we do not know what his beliefs about wedded life are. In the speaker’s mental space, it is presupposed that one should avoid marrying. If we compare (a) Andrew must have been happy to marry Jane, and (b) Paul must have been mad to marry Mary, we understand that marrying Jane is supposed to be rather satisfactory, whereas marrying Mary is not. The differences between these two interpretations stem from two opposite beliefs: marrying Jane is thought to be a positive decision, and marrying Mary is considered a mistake. It is the interaction between social/cultural/individual beliefs concerning a given event and the positive or negative content of the adjective that makes the sentence understandable as a highly appreciative stance on the part of the speaker.

3.4. The adjective-infinitival proposition link

35Since we rejected some examples when collecting our data, we think that our decisions call for some kind of justification. In example (20) they must have been surprised to find the journalists engaged in a fierce contretemps, the adjective surprised is a reaction-oriented adjective. One feels surprised after experiencing or witnessing a certain situation/process/event. Therefore the cause-consequence relation in example (20) is the following: [they find the journalists engaged] → [they felt surprised]. Conversely, in example (21) they must have been careful to hide it, the adjective careful is action-oriented and the interpretation is that they acted in a sensible way. The following sentence, which is closer to those we have been analysing so far, as regards the syntactic construction itself, does not yield the same humorous interpretation:

(44) I guess he must be intelligent to have attained his lofty position, but his lack of empathy and conceit are overwhelming.

36The speaker acknowledges that the man was successful in his career, and that his position is one that everybody would envy in their shared social/cultural background. No feeling of amazement is felt and the most logical explanation of his success is that he is intelligent. No shade of irony is lurking here behind the words describing the speaker’s stance. It is the assessment of the semantic features of the evaluative adjective – however complex it may be – in relation with what the content of the infinitive denotes, as far as the reality space is concerned, that disambiguates the construction for the hearer, as the two following examples suggest:

(45) How old is your son Karen, he must have been worried to call your Ex.

(46) I realise that bill is not the current Minister of Local Government's bill, but he must have been worried to inherit a bill that was based on myths. (​en-nz/​pb/​debates/​debates/​speeches/​50HansS_20120918_00000930/​king-annette-environment-canterbury-temporary-commissioners)

37In (45) we understand that it was because Karen was worried that he called her Ex, since a worried state of mind prompts some kind of action to get some help, whereas in (46) it was because the character inherited a bill based on myths, and thus a fraud, that the character got worried.


38This particular construction can only be understood as a whole, the two propositions being entirely inter-related. Neither of them can be understood alone. The reading of the modalized proposition depends entirely on how the content of the infinitive is considered in the speaker’s mental construct. The infinitive does encode an actual fact (a situation that is held to be true), which cannot be questioned. It is the basis (antecedent) of the speaker’s modal stance about the mental state of the individual he is characterizing, positively or negatively. This mental state seems to be the focus of the construction the speaker is using, and if the speaker is resorting to this particular construction it is to highlight his surprise at what he witnessed. The unexpected action/event/process suggests a bizarre cause, which the speaker endeavours to account for, either by a haphazard reason – occurrences with such adjectives as mad, foolish, drunk – or by a well-thought out one in the case of fond in A. Christie’s example.

Top of page


Boas, Hans. A Constructional Approach to Resultatives. Standford: CSLI, 2003.

Bolinger, Dwight. Meaning and Form. London & New York: Longman, 1977.

Boulonnais, Dominique. « TO et les infinitives : l’hypothèse de la transcendance prépositionnelle. » CIEREC, 116. 55-90, 2004.

Boulonnais, Dominique. « Les prépositions TO et FOR : grammaticalisation et subjectification. » Anglophonia 24 (2008). Toulouse : Presse Universitaire du Mirail. 47-97

Coates, Jennifer. The Semantics of the Modal Auxiliaries. London & Camberra: Croom Helm, 1983.

Cotte, Pierre. Le système des auxiliaires modaux dans le système verbal de l’anglais contemporain. Thèse pour le doctorat d’Etat, Université de Grenoble III, 1988.

Croft, William. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: Oxford University Press, 2001.

Croft, William and D. Alan Cruse. Cognitive Linguistics. Cambridge: Cambridge University Press, 2004.

Culioli, Antoine. La théorie d’Antoine Culioli. Ouverures et incidences. HLD. Paris : Ophrys, 1992.

Declerck, Renaat. Tense in English. Its Structure and Use in Discourse. London & New York: Routledge, 2004.

Deléchelle, Gérard. Expression de la cause en anglais contemporain : étude de quelques connecteurs et opérations. Thèse d’état, Université de Paris 3, 1989.

Delmas, Claude. (à paraître). « Gestion énonciative, espaces mentaux et genre », in Enonciation et neurosciences cognitives, Groupe d’Etudes et de Recherches pour le Français Langue Internationale, n°11.

Depraetere, Ilse and Susan Reed.  “Mood and Modality in English.” In Aarts, B. & A. McMahon (eds.) The Handbook of English Linguistics. Malden, MA: Blackwell Publishing, 2006. 269-90.

Deschamps, Alain. « Traitement des paramètres des complétives ». In M-L. Groussier et C. Rivière (éds.) La Notion. Paris : Ophrys. 60-74, 1997.

Deschamps, Alain. « Pour un traitement formel énonciatif des compléments de verbe à forme non-finie ». In C. Delmas (éd.). Complétude, cognition, construction linguistique. Paris : Presses de la Sorbonne Nouvelle. 135-146, 2006.

Desclés, Jean-Pierre. “Abduction and Non-observability. Some Examples from Language Science and the Cognitive Science.” In Agazzi, E. and M. Pauri (eds.). The Reality of the Unobservable. Observability, Unobservability and their Impact on the Issue of Scientific Realism. Springer, 87-122, 2000.

Desurmont, Christopher. « L’interprétation de l’adjectif qualitatif. » In G. Girard-Gillet (éd.), L’envers du décor. Presse universitaire d’Avignon. 74-94, 2008.

Desurmont, Christopher. Unclear Predication in the  “Evaluative Adjective + Infinitive Construction”, E-rea [En ligne], 9.2. URL :, 2012.

Dufaye, Lionel. « Les modaux et la négation en anglais contemporain », in Cahiers de Recherche, n° spécial, Gap : Ophrys, 2001.

Duffley, Patrick. The English Infinitive. London: Longman, 1992.

Duffley, Patrick. “A Natural-Language Semantics Approach to Infinitival and Gerund-Participial Complementation in English.” In Anglophonia n° 22 (2007). Toulouse : Presses du Mirail. 55-67.

Egan, Thomas.  “The Classification of Non-Finite Complement Constructions in English”. In D. Lebaud, C. Paulin, K. Ploog (eds.). Constructions verbales et production du sens. Presses Universitaires de Franche-Comté, 2006. 187-199.

Egan, Thomas. Non-finite Complementation. A Usage-Based Study of Infinitive and –ing Clauses in English. Amsterdam and New York: Rodopi, 2008.

Fauconnier, Gilles. Espaces mentaux, aspects de la construction du sens dans les langues naturelles. Paris : Editions de Minuit, 1984.

Furmaniak, Grégory. Influence des marqueurs aspectuels sur la construction et l’interprétation de la modalité dans les énoncés en MUST, Ph. D. Dissertation, Université Sorbonne Nouvelle Paris 3, 2004.

Furmaniak, Grégory et Paul Larreya. (to be published). “On the Uses of Would in Epistemic Contexts”. Cambridge: Cambridge Scholar. 105-124.

Gilbert, Eric. « MAY, MUST, CAN et les opérations énonciatives ». Paris : Ophrys, 1987.

Gilbert, Eric. « Vers une analyse unitaire des modalités. May, must, can, will, shall ». In Bouscaren, J., A. Deschamps, & L. Dufaye (éds). Modalité et opérations énonciatives Cahiers de recherche, T. 8. Gap : Ophrys, 2001. 23-99.

Girard-Gillet, Geneviève. « Les infinitives problématiques : l’exemple de cease to », in Anglophonia, n° 24 (2008), Toulouse : Presses du Mirail. 99-115.

Girard-Gillet, Geneviève. “Is TO a Cohesion Marker? ” in RANAM, n° 20 (2011). Université de Strasbourg. 47-66.

Huddleston, Rodney and Geoffrey K. Pullum. The Cambridge Grammar of the English Language. Cambridge. UK: Cambridge University Press, 2002.

Hugou, Vincent. Productivité et émergence du sens : l’exemple de la construction (all) X-ED OUT dans un corpus de blogs et de forums de discussion, Ph. D. Dissertation, Université Sorbonne Nouvelle Paris 3, 2013.

Kertz, Laura. “Evaluative Predicates: An Adjunct Control Analysis”. In Proceedings of the 25th West Coast Conference on Formal Linguistics, edited by Donald Baumer, David Montero, and Michael Scanlon. Somerville, MA: Cascadilla Proceedings Project, 2006. 229-235.

Larreya, Paul. Le possible et le nécessaire, modalités et auxiliaires en anglais britannique. Paris : Nathan, 1984.

Larreya, Paul. « Connaissance, inférence et modalités épistémiques dans le système verbal de l’anglais ». In Pauchard, J. (éd.) La modalité et les modaux en diachronie et en synchronie (domaine anglais). Reims : Presses universitaires de Reims, 2000. 175-199.

Larreya, Paul. “Types of Modality and Types of Modalisation”. Poster. Second International Conference on Modality in English. Université de Pau, septembre 2004.

Larreya, Paul. “Towards a Typology of Modality in Language”. In Salkie, R., P.Busuttil, J. van der Auwera (eds.) Modality in English: Theory and Description. Berlin and New York: Mouton de Gruyter, 2009. 9-30.

Mair, Christian. “Change and Variation in Present-day English: Integrating the Analysis of Closed Corpora and Web-based Monitoring”. In Hundt, M, N. Hesselhaufand Biewer, C. (eds.) Corpus Linguistics and the Web, Amsterdam: Rodopi, 2007. 233-47.

Pierce, Charles Sanders. Collected Papers, vol I-VI. The Belknap Press of Harward University Press, 1974.

Stowell, Tim. “The Alignment of Arguments in Adjective Phrases ”. In Rothstein, S. (ed.) Perspectives on Phrase Structure: Heads and Licensing, Syntax and Semantics 25 (1991), Academic Press. 05-135.

Sweetser, Eve,. From Etymology to Pragmatics – Metaphorical and Cultural Aspects of Semantic Structure. Cambridge Studies in Linguistics 54. Cambridge: Cambridge University Press, 1990.

Sweetser, Eve. “Meta-metaphorical Conditionals”. In M. Shibatani, S. A. Thompson (eds.) Grammatical Constructions. Oxford: Oxford University Press, 1999. 221-233.

Whelpton, Matthew. The Syntax and Semantics of Infinitives of Result in English. Oxford, UK: University of Oxford dissertation, 1995.

Whelpton, Matthew. “Elucidation of a Telic Infinitive”. Journal of Linguistics 37 (2001). 313-337.

Top of page


1 Furmaniak 2004, ex (82) p. 274.

2 See also his contribution in this volume.

3 The fact that there are very few occurrences in the COCA for this construction compelled us to resort to Google books to collect more data. We made sure that the occurrences came from native speakers. Even if no frequency analysis is possible, we think, as Boas does (2003), that Google can help one to confirm one’s hypothesis.

4 We wish to thank Vincent Hugou for helping us to collect the data.

5 We started our search with adjectives describing negatively-oriented physical, psychological or mental states, because the first instance of this construction that caught our ears was indeed he must have been drunk to have said that. Other adjectives are possible as we shall show later on.

6 Ex (77) p. 262 in his Ph. D. Dissertation

7 Tregidgo says that ‘when we say that something must be true, we mean that this conclusion is demanded of us.’

8 We cannot give a frequency analysis, but only a tendency, even if Mair (2007) showed in his analysis of different from/than/to that the proportions were similar in the corpora studied and on the Web.

Top of page


Electronic reference

Geneviève Girard-Gillet, « A posteriori modality, implicative modality by abduction. A case study: he must have been drunk to have said that », Anglophonia [Online], 19 | 2015, Online since 15 July 2015, connection on 27 June 2017. URL : ; DOI : 10.4000/anglophonia.497

Top of page

About the author

Geneviève Girard-Gillet

Sorbonne Nouvelle Paris 3 USPC
EA 4398

By this author

Top of page


Licence Creative Commons
Anglophonia – French Journal of English Linguistics est mis à disposition selon les termes de la licence Creative Commons Attribution - Pas d'Utilisation Commerciale - Pas de Modification 4.0 International.

Top of page
  • Logo Presses universitaires du Midi