Skip to navigation – Site map

The mirative and scalar meaning of even

Olivier Simonin


The analysis I develop here draws on the concept of mirativity (DeLancey 1987) and on Kay’s (1990) insight into the scalar nature of even. I show that even calls for the reconstruction of an implicit scale, on which the element highlighted by this adverbial marker stands out as the most (or more) surprising in the set considered. This non-referential, non-propositional meaning of even is likely to trigger a series of inferences, whose status is duly discussed, and to strengthen or weaken other derived assumptions, and even to guide their derivation itself.

Top of page

Full text

I would like to thank Paul Larreya for sparking my interest in the marker even and encouraging me to write on this topic. This article is dedicated to him.

1. Introduction

1There is a long-standing linguistic and logical tradition (see, e.g., Stalnaker 1974, Karttunen and Peters 1979, Giannakidou 2007) that refuses to grant adverbial even truth-conditional meaning. I use the term adverbial to distinguish its use in, say, I even went to her party, from others in which it is either an adjective (The ground is even) or part of some complex conjunctive phrase (even though, even if). In spite of this failure to be assigned truth-conditional meaning, even turns out to involve some kind of semantic accommodation in so far as it contributes to modifying or qualifying propositional meaning.

2The account outlined here is very much in the line of post-Gricean pragmatics, even though it purports to be eclectic and draws on analyses from other theoretical horizons. The corpus selected for this study is ICE-GB, the one-million word stratified, parsed and annotated corpus of British English (Nelson et al. 2002). I will also discuss examples taken from the literature, and on one occasion offer acceptable sentences with several occurrences of even for lack of authentic examples (in my corpus or elsewhere).

3I first critically review previous research, discussing several relevant semantic and pragmatic concepts in the process, before presenting the various syntactic contexts in which even occurs in ICE-GB and building on the idea that it conjures up an implicit scale that has to be (at least partially) reconstructed by the hearer. Lastly, I introduce the concept of mirativity and show in what sense even is a mirative marker, to provide an almost exhaustive and detailed account that can predict the inferences (or, more precisely, implicatures) that can be drawn from its use.

2. Review of previous research on even

4To understand the theoretical debate of even, it is necessary to grasp several logical, semantic and pragmatic concepts that it draws on. I start with presuppositions and entailments, present the accounts relying on them, then discuss how the notion of scales has been applied to explain the meaning of the marker, and finally comment on several attempts invoking pre-construction, noteworthiness or focus to cover the uses of even.

2.1. Presuppositions or entailments: two accounts

  • 1 In French, the distinction inherited from Ducrot is between présupposés and sous-entendus (which ar (...)

5In linguistic pragmatics, presuppositions are routinely contrasted with implicatures, typically characterized by their context sensitivity (Ducrot 1969, Ducrot 1987: 46, Kerbrat-Orecchioni 1998: 39, Potts 2014).1 In the analytic, Anglo-Saxon tradition, it is customary to distinguish pragmatic presuppositions from semantic ones (pace Huddleston and Pullum, who choose to label semantic presupposition “pragmatic presupposition”: 2002, 40-41). In a series of papers (see Stalnaker 1998 for a summary), Stalnaker argues that pragmatic presuppositions include “the preconditions for linguistic interaction (for example, the mutual public knowledge that we are speaking the same language), the norms of turn-taking in dialogue, and more particularized information about conversational plans and goals” (Potts 2014: 170). By contrast, Reboul and Moeschler (1994: 229) provide the following logical definition for semantic presupposition:

  • 2 My translation of “Une proposition P présuppose une proposition Q si, et seulement si, (i) dans tou (...)

A proposition P presupposes a proposition Q if, and only if, (i) in all the situations in which P is true, Q is true and (ii) in all the situations in which P is false, Q is true.2

  • 3 I follow standard practice in using simple inverted commas to indicate logical propositions.

6To illustrate: if I say “Peter has stopped drinking”, I assert the proposition that corresponds to the utterance, but to do so I have to presuppose ‘Peter used to drink’.3 Even if it is untrue that Peter has stopped drinking (or that he used to drink), the truth of the (semantically) presupposed proposition is taken for granted. On the other hand, with “Peter has bought a Ferrari”, I also assert the proposition that corresponds to the utterance, and by so doing I additionally convey ‘Peter has bought an expensive sports car’ or, simply, ‘Peter has bought a car’. Yet Peter may have bought a car, or even an expensive sports car, without it being a Ferrari. The latter would then be false (Peter has not bought a Ferrari) while what is entailed by buying a Ferrari (i.e. buying a car, possibly the same type) may be true. Yet, both may be false: if Peter hasn’t bought a Ferrari, perhaps he has bought no car at all. In other words, contrary to (semantic) presuppositions, what is entailed may be true or false if the entailing proposition is false. That is why entailed propositions are not regarded as semantic (or pragmatic) presuppositions, but as entailments (Jaszczolt 2002: 83).

7The first two influential publications on even (Fillmore 1965, Horn 1969) seek to account for its meaning in terms of some entailment or presupposition triggered by the mere use of the adverb, which is otherwise presented as devoid of any meaning.

8Fillmore, in the course of a general discussion of “entailment rules”, comments on the semantic difference between (1) and (2):


She reads Sanskrit


She even reads Sanskrit

He writes (1965: 68):

The second tells us the same as the first, but it adds that this fact is somehow surprising. The contribution of the word “even” cannot be explained, it seems to me, by assigning it semantic features of the usual kind. I would say that the sentence “She even reads Sanskrit is to be ‘factored’, so to speak, into two sentences, namely “She reads Sanskrit” and “One would expect that she does not read Sanskrit.”

9In other words, even conveys an expectation that is the opposite of what is imparted in the clause in which it appears (and since the expectation is entailed, its falsity has no bearing on the truth value of what is asserted).

The problem with this conception, as pointed out by Kay (1990: 83-84), is that it is both too broad and too narrow. Here are his two examples:


The boss wasn’t there and we still worked hard.


Everyone is remarking on Mary’s improvement. Last week she beat the number ten player and this week, as everyone expected, she even beat the number three player.

  • 4 Garrido Médina (1993) provides a lengthier counter-argument.

In (3), following Fillmore’s account, substituting even for still would generate an infelicitous utterance, since the expectation ‘We did not work hard’ would be violated. The counter-expectancy condition is therefore not sufficient to explain the meaning of even. In addition, this condition is not even necessary since, according to Kay (1990: 84), even sometimes fails to communicate counter-expectancy, as in (4), which explicitly states that everyone expected Mary to win her match against the number three player. This is debatable however. Fillmore’s condition could be saved by postulating a variety of possible epistemic expectations (i.e. expected situations or outcomes) that are contextually reconstructed, and at least one of which is actually violated. A likely background assumption for (4) is that everyone would expect Mary to be defeated under normal circumstances (not just under the present circumstances). This is presumably that expectation that is actually defeated.4

10Horn relies on presuppositions as he proposes a parallel account of only and even. He argues that the utterances in which they appear convey an assertion and a presupposition. To take up Horn’s example with even (1969: 105-106):


Even Muriel voted for Hubert.
Assertion: ‘Muriel voted for Hubert.’
Presupposition: ‘Not only Muriel voted for Hubert.’

  • 5 Yet Kay (1990: 86) argues that it is far from obvious that Muriel’s voting for Hubert is more infor (...)

11Note that the falsity of the presupposition (or its truth) cannot affect the truth value of what is asserted. Postulating this type of presuppositional content for utterances like (5) might then seem intuitively appealing,5 yet one major objection can be raised against the account: it fails to cover cases in which even does not have scope over a noun phrase (NP) and, for instance, occurs before a verb (6) or a comparative form (7), to take up the two other main contexts of occurrence considered by Leroux (2012):


I hadn’t even thought about it you know [that it was duck and not pork]. (ICE-GB: <s1a-039 043>)


Replying to a message is even easier than sending. (ICE-GB: s2a-028 081)

  • 6 The position of even after hadn’t is canonical here, as it comes after the auxiliary, even though i (...)
  • 7 Situations or states-of-affairs are used to refer to events or states denoted by verbs and their ex (...)

12In (6) even has scope over the verb phrase (VP) hadn’t thought about it6 or read Sanskrit in (2). Despite Horn’s failure to provide suggestions to extend his analysis, predicates or VPs can be opposed to other potential predicates or VPs, representing situations or states-of-affairs7 that are presumably either more readily achieved, such as read French or read German for (2) or, in negative contexts, more demanding, like pondered the issue or discuss the matter for (6). One could then argue that ‘She does not only read Sanskrit’ and ‘I had not only failed to think about it’ should be considered as presuppositions for (2) and (6). Yet when even has scope over a comparative form, this kind of presupposition does not seem to be available: (7) does not in any way presuppose or communicate ‘Sending is not only easier’.

13Just as entailments proceed from the meaning of a linguistic unit, (semantic) presuppositions are ordinarily triggered by a linguistic form (Abbot 2006: 5). In “Peter stopped drinking”, the verb “stop” acts as a trigger for the presupposition ‘Peter used to drink’. Yet triggers like stop have a full-fledged meaning of their own, and not just a presuppositional one. Similarly, an entailment arises in virtue of the meaning of a linguistic form, which cannot be reduced to it. Generally speaking, it is debatable whether forms that only have entailed or presuppositional content should be postulated. Put differently, what is treated as a semantic presupposition or an entailment could in fact simply be seen as deriving from the overall meaning of even. This meaning remains to be fully elucidated.

2.2. Even and scales

  • 8 His characterization of présupposé excludes truth value considerations: the falsity of a présupposé(...)
  • 9 Ducrot’s example and the présupposé he postulates are, respectively: Même Jacques est venu and La v (...)

14Linguists have also attempted to account for the meaning of even in scalar terms, proposing that the use of even calls for the representation of a scale. Ducrot implicitly invokes a scale of significance as he suggests that Even Jacques came conveys, as a présupposé (a much broader term than semantic presupposition in his sense),8Jacques’s coming is more significant than anyone else’s’ (1972: 105).9 However, although this type of proposition could also apply to most cases in which even has scope over a predicate or a VP (and not just an NP like Jacques), as in (2) and (6), details have to be filled in to determine what it should be for utterances like (4) (‘To beat the number three player is more significant than anything else she’s done so far’?), not to mention those with comparative forms, like (7).

  • 10 For Karttunen and Peters, the two propositions are conventionally implicated: they are not presuppo (...)

15Karttunen and Peters (1979: 12) obviously postulate another scale, which is not grounded on significance but on (least) likelihood, when they write that (8) implicates two propositions:10


Even Bill likes Mary
‘Other people besides Bill like Mary’
‘Bill is the least likely person to like Mary’

Yet the notion of least likelihood is not always relevant, or at least not at first sight – witness Fauconnier’s (1976: 262) example, given as (9), and (10):

  • 11 My translation of “Georges a bu un peu de vin, un peu de cognac, un peu de rhum, un peu de calva et (...)


Georges has drunk a little wine, a little cognac, a little rum, a little calvados and even a little armagnac.11


They were obviously bored, one of them even getting down on his knees to look under the car, grateful for the exercise. (ICE-GB: <w2f-012 018>)

16The speaker of (9) can be understood here as not suggesting that armagnac is the least likely alcoholic beverage to be consumed, but that Georges has had too much alcohol, with all the various drinks combined. Neither does (10) imply that the person who went down on his knees was the least likely to do it – and even less so that there were others who did the same thing (although they have presumably displayed their boredom in other ways). More plausibly here, it is rather the situation of getting on one’s knees that is most unlikely. In addition, when even has scope over comparative forms, as in (7), lesser likelihood could be resorted to instead of least likelihood, but the result is not wholly convincing: whether Replying to a message is even easier than sending implies ‘Replying to a message is unlikely to be easier than sending one’ is highly debatable. (7) could simply be used by the speaker to reassure the hearer that replying is not as hard as the latter may think, the comparison with sending being incidental, to illustrate the easiness of replying, in which case the notion of likelihood fails to be an essential component of the utterance meaning.

  • 12 Fauconnier also noted that even scales are often probabilistic (1976: 262).

17Fauconnier (1976: 32) developed the idea that even has a superlative function in so far as it highlights the end of a scale – a notion that is akin to least likelihood, and on which Karttunen and Peters have obviously drawn.12 In (8), Bill would be at the lowest end of a scale ranking people who like Mary. Yet, even has a comparative function rather than a superlative one if (7) and, as Kay (1990: 89) has shown, it does not always bring an end-of-scale focus.


Not only did Mary win her first round match, she even made it to the semi-finals.

As Mary does not reach the final, she does not find herself at the very end of the performance scale. Put differently, on a scale of tennismanship, the top rung is the final (of the tournament considered), and it is not reached. There might then be a series of elements positioned on the scale (that may be reconstructed or left implicit), or just two (wine or armagnac in He has drunk wine and even armagnac, let’s say, just as with comparatives), and ends of scales do not need be reached, although they tend to be.

18Kay himself treats even as a scalar operator (1990: 62). Its scalar nature, which is now taken for granted by most scholars, is conveniently and concisely described by Giannakidou (2007: 40):

  • 13 The “even phrase” corresponds to the adverb even itself and what it has scope over syntactically.

There is a set of alternatives to x, the even phrase, that the context makes salient… and … these alternatives are ranked on a scale.13

19This characterization does not exhaust the meaning of even, and Kay additionally argues that (a) the clause in which it occurs expresses a proposition which, compared to some alternative propositions(s), is more informative (i.e. the latter can be inferred when knowing the former), and that (b) the set of alternative propositions “are part of the shared background of speaker and hearer at the time of occurrence.” (1990: 63) In other words, they pertain to the common ground (Clark 1992: 3), which means that they are pragmatically presupposed (see 2.1). Although claim (a) is perhaps difficult to challenge, claim (b) appears to be much less tenable, even when taking into account accommodation by which a speaker makes as if propositions or assumptions were part of a shared background. The first question is whether what has just been uttered can be considered as pertaining to shared background information: in (11), the hearer presumably learns that Mary won her first round match at the beginning of the utterance, while in (10), the first part of the utterance states that everyone (in the group discussed) was bored, the second part showing how one of them demonstrated he was especially so. What is more, when fleshing out the context for an utterance like (8), Even Bill likes Mary (changing its form slightly), it is obvious that the speaker does not at all assume that ‘Mary is likeable’ or ‘that people like Mary in general’ belongs to the common ground, since the hearer has just explicitly rejected the suggestion beforehand:


A. I don’t find this Mary very friendly.
B. Well, in fact, even Bill likes her.

In spite of this, the notion of even being a scalar operator has been very influential, and Kay’s contribution is often discussed, if not taken as a starting point for further discussion (see, e.g., Giannakidou 2007).

2.3. Enunciative (and related) accounts

20French enunciative linguists have tended to underline the unexpectedness of the situation expressed in the utterance containing even (see, e.g., Leroux 2012) positing that “the negative value [for that situation] is pre-constructed but, contrary to expectation, the positive value is then assigned again.” (Ranger and Méry 2007: 8). This is very close to Zeevat’s suggestion that in the utterance Even Bill is in Paris, “Bill is in Paris while being weakly presupposed to be not there unlike others.” (2009: 138) The weakly presupposed expectation “can be much weaker than the belief of the interlocutor (or the common ground, or a second speaker) and can even be vanishingly weak: a mere suggestion or what somebody might think.” (2009: 122) With such a broad acceptation, the account appears to be unassailable. Consider (13) however:


And we can all see that this tax is virtually impossible to collect from some people so that those who do pay are being asked to pay even more to cover the cost of defaulters. (ICE-GB: s2b-030 117)

It would perhaps be difficult to regard ‘Those who pay are NOT being asked to pay even more’ as pre-constructed or (weakly) presupposed: it seems that the opposite proposition, rather, can be inferred from the beginning of (13), though it is still worth stating explicitly.

21Blandine Pennec (2016: 195) does not use the notions of preconstruction or weak presupposition, arguing that:

  • 14 My translation of “Tout en homogénéisant les différentes occurrences, even souligne également le ca (...)

While presenting the various tokens of a set as homogeneous, even underlines the (initial) unexpectedness and thereby noteworthiness of the positive truth value attributed to the proposition involving the “extreme” token. This token therefore finds itself under some emphasis.14

This analysis is consonant with the etymology of even, which primarily meant ‘equal’, ‘like’ and ‘level’ in Old English (Hall 1960: 100), suggesting a levelling of all the tokens considered with respect to the relevant situation. In (8), Bill is on a par with all the other tokens with respect to liking Mary (which is surprising in his case).

22The noteworthiness mentioned by Pennec dovetails neatly with Close’s classification of even as focus adverb, highlighting one element within a clause (1975: 287-288). This focalization on the “extreme” token is correlated by the fact (observed by Kay 1990: 86) that whereas an utterance like Bill likes even Mary can be used as a conversation opener, Even Bill likes Mary would be much less acceptable – if at all. This is because it is hardly feasible to start a conversation with an element in strong focus without having provided a topic beforehand, to recall or establish some common ground. In any event, the connection between unexpectedness and noteworthiness as well as focalization appears to be very relevant to the meaning of even.

3. Scope and scales

23It is now time to provide a systematic account of the syntactic contexts in which the adverb even can occur, relying on the ICE-GB corpus, and to relate those contexts to the scope of even and the notion of a scale that it conjures up.

3.1. The scope of even

  • 15 That is, Prepositional Phrases (PPs) that complement a head noun.

24The potential syntactic scope of even is much wider than a cursory glance at the literature might lead one to think. The corpus ICE-GB, although restricted in size (one million words) and limited to British English, gives a globally adequate picture of the distribution of the adverb even in this variety of English, with 613 occurrences for which I have found 10 different types of syntactic scopes. Even can have scope over (a) constituents that are verbal arguments – subjects, objects, predicative complements (14-16) – (b) adverbial constituents (17-18), (c) VPs or predicates (19), (d) whole clauses (20), (e) comparative phrases (21-22), (f) attributive adjectives (23), (g) constituents that complement a predicative adjective or phrase (24), (h) noun complement PPs15 (25) and, lastly, NPs within a PP that either (i) complements another NP (26) or (j) is used with a concessive meaning (27). For the sake of convenience, (10) has been reproduced as (20) and (7) as (21), and the phrases on which even has scope have been emphasized with bold type:

  • 16 The PP could also be parsed as adverbial, and not as an NP complement (influence here). However, On (...)


Even a ten percent saving on estimated costs of manual conversion can show savings of the order of tens of millions of pounds. (ICE-GB: <w2a-032 046>)


… I could not see them running even a medium sized multinational. (ICE-GB: <w2b-012 077>)


They [Dahl’s books portraying children battling against evil adults] were sometimes said to be bizarre grotesque even cruel (ICE-GB: < s2b-011 012>)


[You probably need someone’s help to learn to sing well] And probably even when ones writing songs and polishing up the way youd use the idioms and all that (ICE-GB: <s1a-033 089>)


It uhm takes an enormous amount of storage space even with compression techniques. (ICE-GB: <s1b-076 195>)


I mean its assumed if you work for the Royal Shakespeare Company that youve read the play or seen the play or even collaborated on a previous production (ICE-GB: <s1b-023 057>)


They were obviously bored, one of them even getting down on his knees to look under the car, grateful for the exercise. (ICE-GB: <w2f-012 018>)


Replying to a message is even easier than sending. (ICE-GB: s2a-028 081)


And [it is] even less commonly known that we outbid the Lord Berkley of Berkley Square fame at that time for her hand (ICE-GB: <s2a-045 084>).


The light flared and diminished, casting a flickering light across the streets and lending to the houses and pedestrians a faintly unreal or even theatrical quality. (ICE-GB: <w2b-006 106>)


I dont think that room is worth fifteen an hour or even seventeen an hour by the time you put VAT on top (ICE-GB: <s1a-083 122>).


He really had no time for personal research even on the theoretical side... (ICE-GB: <s2a-041 110>)


Vivid TV coverage of battlefield events and of bodies being brought home could have a remarkably restraining influence on even the most gung-ho generals and bellicose politicians (ICE-GB: <w2c-003 089>).16


But beyond even the fault lines of Europe the litany of ministers she has lost – rehearsed with feline skill in the Heseltine text – renders her unfit for the task ahead. (ICE-GB: <w2e-003 019>)

In (14), a ten percent saving is the smallest percentage considered by the speaker for the sake of his argument and it still leads to impressive savings, which higher percentages would ensure as well, but with higher returns. The smallest percentage considered on the scale is contrasted with the other ones with respect to the likelihood of leading to massive savings. The NP (even) a ten percent saving on estimated costs of manual conversion is a subject – and so a verbal argument. Similarly, in (15), the object a medium sized multinational denotes an entity that is in a mid-position on the size scale. It is contrasted with bigger ones, which could not be run either by the people meant by the speaker. In (16), cruelty is the property that trumps the others (weirdness and grotesqueness) in terms of its scaring power, regarding whatever the subject of copulative be refers to.

25With an adverbial scope, even underscores that circumstances in which the situation is not as likely to occur still lead to its occurrence, opposing them to others that are presumably more conducive to the situation: when learning to sing, hiring a specialist is very valuable, while it may appear less important when trying to improve lyrics (17). With the help of compression techniques, storage problems should be reduced, in contradistinction with cases in which they are not used but, in the context of (18), they do not make it possible to avoid having an enormous space taken up.

26With a predicative or a VP scope, types of situations involving the subject (i.e. its referent) are set off against one another, with one being presented as the most extreme: having collaborated on a previous production of a play in (19), and not just having read it or seen it performed.

27With a clausal scope, a general type of situation is envisaged, and more precisely, illustrative situations involving different agents or subject referents are set in opposition, the most (or more) extreme or telling of which being the one introduced with even. In (20), the person who goes on his knees to look under the car evinces much boredom for the speaker, being more conspicuous than other potential signs (sighing, eye-rolling…) that the hearer is left to imagine.

28With comparative phrases, only two rungs are considered. In (21), on a scale of easiness, the two elements that are ranked are: replying to a message and sending one. As Fillmore noted (1965: 68) for a similar utterance, the element serving for comparative purposes already exhibits the property denoted by the adjective (the action of sending a message is deemed easy), but that property reaches a higher degree with the other element (replying to a message). The comparative form may be the suffix -er, and the adverb more or less – as in (22), where the relevant property is denoted by an adverb and not an adjective.

29Even appears to be used in conjunction with attributive adjectives only in written, rather formal genres. Its use is parallel to that of predicative adjectives, as a copular relation seems to hold between adjective and noun: in (23), a faintly unreal or even theatrical quality is semantically equivalent to a quality that was faintly unreal or even theatrical.

30With complements of predicative constituents, even seems to work along similar lines to when it occurs with verbal arguments like subjects and objects: I don’t think that room is worth fifteen an hour or even seventeen an hour in (24) can be rewritten as I don’t think that room amounts to fifteen an hour or even seventeen an hour, with amounts to replacing is worth, so that in the new sentence even has scope over (part of) an indirect object. Even though scopes differ in both sentences, the semantic relation between the element which even points out and the situation designated through the predicative constituent in the original utterance and the verb in the rewritten sentence, remains unchanged. The syntax is different, but the relevant semantic relation is the same.

31A similar analysis holds for noun complement PPs: in (25), He really had no time for personal research even on the theoretical side can be paraphrased He really had no time to research by himself even on the theoretical side, with even having scope on an adverbial element here, and not a verbal argument. Yet even has the same semantic impact in both cases.

32For NPs within a PP complementing another NP, the same reasoning prevails: in (26), Vivid TV coverage… could have a remarkably restraining influence on even the most gung-ho generals… could be rewritten as Vivid TV coverage… could remarkably influence (in a restraining way) even the most gung-ho generals…, with the NP the most gung-ho generals… becoming a direct object. Note also that in the paraphrase, even appears just before the NP, which suggests that its natural scope is the NP – and even occurs just before the NP in (26). Placing even just before on and the PP would have been a marked use (if deemed acceptable).

33Let us now consider NPs within a concessive PP which complement a preposition like behind, beyond or without, which conveys (with its NP complement) an adverbial meaning. In (27), the topic being Mrs. Thatcher, the PP beyond even the fault lines of Europe is more or less equivalent to when discounting even the fault lines of Europe, or rather even when discounting the fault lines of Europe. What is referred to as the fault lines of Europe is opposed to other potential political liabilities for Mrs. Thatcher’s upcoming task. They are also serious, but not as much. The most serious handicap, which tops all the others (the fault lines of Europe being the next one on the scale), is introduced just after the PP: the litany of ministers she has lost. As it creates a sense of suspense (and of outbidding), this use of even contributes to the concessive meaning of the PP, already conveyed by but (which could have been omitted): under the cloak of rhetorical magnanimity, the speaker discounts a very serious argument to consider only the one that is superior to it.

  • 17 This also includes entities that NPs refer to when the NP is found within an indirect or a secondar (...)

34To close this section on the scope of even, let us relate syntax (syntactic scope) and semantics: the constituent on which even has scope denotes the element on which even bears semantically, whether it is an individual extralinguistic entity (with NPs),17 a trait (with AdjPs), a given circumstantial qualification (with adverbial constituents), a degree (with comparatives) or a (type of) situation (with predicates and clauses).

3.2. Reconstructing the implicit scale

35Building on Kay’s insight into the scalar nature of even as well as its etymology (see 2.3), I will now defend another scalar account, leaving the discussion of issues concerning implicitness and the precise status (presupposed, entailed, implicated…) of the meaning communicated by even until later.

36We have seen that the element on which even bears is not necessarily found at the end of a scale. It can be situated in mid-scale in some instances (15), or quite high (or low) but not at the very end (4). This element is the highest or higher one considered by the speaker, and is opposed to a set of other elements – or sometimes just another element, as in occurrences with comparatives – that do not rank as high. With even having scope over an NP within a concessive PP (27), the speaker can put aside an element still higher up on the scale, and take into consideration the next in line, to set it off against all the others below. Sometimes, there are clear textual clues and even explicit indications about the element(s) with a less high value on the scale, as in (16), (19), (21), (23) and (24). Beyond those, the set of elements serving as a foil is simply pointed to by the speaker’s use of even. The reader is left to reconstruct inferentially, or at least envisage, that set of elements drawing on the contextual (linguistic, situational) information available or shared knowledge (common ground or general, socio-cultural assumptions). It is essential to recognize that in some cases at least, this process is not straightforward. An illustration is provided by (9), which is reproduced here as (28):


Georges has drunk a little wine, a little cognac, a little rum, a little calvados and even a little armagnac.

37The relevant set is not made up of all the different drinks listed, or the interpretation would have been different, with armagnac being the most confidential drink to enjoy as a connoisseur, for instance. This demonstrates the importance of the situational context, which interacts with shared assumptions or knowledge about Georges and the occasion on which he drank. In addition, the alcoholic consumptions making up the various elements on the implicit scale (for the number of alcohols drunk, let’s say) are also reconstructed relying on the linguistic information provided by the utterance. The elements are understood to be the sets of alcohols consumed, as Georges is considered: drinking (just) wine; drinking wine and cognac; drinking wine, cognac and rum; drinking all those plus calvados; drinking all the alcohols mentioned (including armagnac). This brings us now to the question of the property that structures the scale.

  • 18 This appears to be frequent when even has scope over adjectives or AdjPs. Pennec (to appear) argues (...)
  • 19 And some adverbials: She went swimming regularly, even daily. Here, the property is frequency of oc (...)

38The property according to which the elements are ranked on the scale has to be reconstructed in the same way. For comparatives, this is the degree of the trait designated by the adjective (or circumstantial quality for an adverbial constituent). With adjectives, the property is a quality inferred from the adjective itself and, if it comes at the end of a list,18 those it co-occurs with: it might be the worrying power conveyed by the adjectives in (16), or a sense of estrangement in (23). The same can generally be said of predicative complements:19


It’s even at a higher level than that though (ICE-GB: <s1a-005 189>)

39The speaker responds to a comment to the effect that regular computer users do not understand how difficult it is for laypeople to learn to use them well. The property here is the seriousness of the problem. With predicates, the property is an implicit quality that is evinced by situations like the one denoted by the predicate: in (19), the coordinated predicates communicate various degrees of knowledge of a Shakespeare play. With clauses, in the first analysis, the property is the aptness of specific situations – like the one expressed in the clause – to exemplify tellingly another situation that has just been evoked: in (20), this other situation corresponds to the fact that the members of the group in question are really bored, which is clearly stated at the beginning of the utterance. The property could also be argued to be, just as with predicates or VPs, a quality evinced by the situation expressed by the whole clause (boredom here), but which has just been explicitly pointed out previously. It would then appear that with clauses, predicates and VPs, the structuring property for the scale is a quality perfectly illustrated by the situation on which even bears, however that quality is reconstructed inferentially. Lastly, with NPs that are verbal arguments (or PPs for indirect and prepositional secondary objects), with adverbial adjuncts as well as with NPs within a concessive PP, the property is an effective and actual compatibility with the (type of) situation designated by the main verb (except for an utterance like (28), which has just been tackled, or cases in which there is no main verb, as in (17), and the situation has to be inferred). The compatibility is effective and actual in so far as the utterance asserts the relevant situation: (18) states that still a lot of storage space is needed, in spite of compression techniques.

In some instances, the reference to the situation pointed out by the main verb turns out to be somewhat loose: we have seen that in (14), can show savings of the order of tens millions of pounds stands for potentially huge savings, whatever the precise amount.

  • 20 Anderson himself writes: “Kuroda observes that, while even can appear in any of several positions i (...)

40These considerations now make it possible to appraise a partially true observation on even that dates back to Anderson, who himself built on a point previously made by Kuroda:20 “the limitation on occurrence to once per sentence” (1972: 905), by which is meant clause. Here is one of Anderson’s examples (1972: 894):


*Even the president even enjoys Skrunkies for breakfast.

41With one even having scope over a verbal argument (subject NP) and the other one over the predicate, two scales need to be reconstructed by the hearer. The property that structures the first one is the compatibility of individuals with enjoying Skrunkies for breakfast. The property structuring the second one is an implicit quality evinced by enjoying Skrunkies for breakfast. Reconstructing the two scales and identifying the second property is too tall an order. It would have been possible to find the relevant quality if the president had not been set against other individuals defined by their same particular breakfast preference: The president (does not go in for luxury, he) even enjoys Skrunkies for breakfast can be interpreted readily. The individuals in question can be characterized by their enjoying Skrunkies, but not by their even enjoying Skrunkies. Much the same reasoning prevails for *Even the president enjoys even Skrunkies for breakfast. The set of individuals contrasted with the president cannot be defined by enjoying even Skrunkies, with a potentially open list of alternatives that can hardly be circumscribed. Even though I have found no double occurrence of even within a clause in the ICE-GB corpus, or in any other authentic sources for that matter, I will offer three examples (whose acceptability I have checked with several native informants) that demonstrate that the limitation of one occurrence of even per clause is not a rule:


Even the president enjoys cereal, even Skrunkies, for breakfast.


[Our small group was in a great shape, but not James who still managed to run very fast, as usual.] Even John went even quicker than James.


[A group of children has received a visit and all of them have behaved extremely well] Even John has been described as kind, gentle, even generous.

42(31) could in fact be regarded as syntactically made up of two clauses telescoped into one, as it can be paraphrased Even the president enjoys cereal for breakfast, he even enjoys Skrunkies. It may not be too serious a counterexample. (32) is more complex and harder to interpret, but it is possible to picture a group of individuals that went even quicker than James, because the quickness (the property structuring the second scale) is given a precise reference point (James’s). In (33), the list of adjectives ending with the one introduced by even suggest that the property meant is some kind of goodness, for which generosity is deemed to be the best indicator. The group meant is characterized by that property and a minimal level of goodness, which John presumably stands for. When interpretation is not unduly problematic, clauses with more than one occurrence of even can then be fully acceptable.

4. Mirativity, inferences and focalization

43I now wish to show that the speaker marks the positioning on the scale of the element on which even bears as surprising: the property structuring the scale reaches a surprisingly high (or low) value. In this sense, even is a mirative marker. I first introduce the notion of mirativity and explain how even is mirative, before I analyze the nature of inferences that can be drawn from its use and finally relate the notion of mirativity with focalization.

4.1. Mirativity and even

44Discussions of mirativity and its full recognition as a grammatical, typological category are grounded in the work of DeLancey, who demonstrates that it is distinct from evidentiality and defines its fundamental function as (1997: 33):

mark[ing] sentences which report information which is new or surprising to the speaker, regardless of whether the information source is first- or second-hand.

45Building on previous work, he postulates the existence of mirativity to account for languages that grammatically code the expression of surprise, like Tibetan, while suggesting that the notion of mirativity could be used to account for some phenomena in English as well as in other languages, where the category is “covert” (1997: 48), i.e. not grammaticalized, but semantic. DeLancey’s doublefold characterization of mirativity, including surprising or simply new information, is problematic in so far as surprising information implies informative newness, but the latter does not imply the former. To prevent the category from being stretched beyond any linguistic usefulness, it is necessary to exclude information that is only and merely new from the purview of mirativity. De Haan characterizes mirativity “as the marking of unexpected information, information that somehow shocks or surprises the speaker,” (2012: 1038) while Peterson argues that all previous concepts invoked to define the category boil down to surprise (2013: 5):

All of the other semantic characteristics of mirativity commonly used to define it in the literature, such as unexpectedness, suddenness, and new information, are entailed by surprise.

46Peterson, among others, focuses on mirativity as a semantic rather than a typological category. From a semantic perspective, it is important to note that mirativity is connected to the domain of epistemic modality, since being surprised by the occurrence of a fact usually implies having previously considered its probability of occurring very low. This could go some way towards explaining why some languages use the same markers for evidentiality and mirativity (see, e.g., DeLancey 1997 and De Haan 2012).

47Zeevat draws on the semantic conception of mirativity to describe the implicit expectations conveyed by only and a few other markers. In the wake of Horn’s parallel account of only and even, he suggests (2009: 180) that both markers are mirative as they “express surprise at a large size of a quantity (even), surprise at the small scale of a quantity (only).” In Even Bill is in Paris, the group of others who are in Paris is felt to be surprisingly numerous, while in Only Bill is in Paris, what is surprising is that none other than Bill is in Paris (in the group meant). We have seen the limitations of this type of account when reviewing Horn’s treatment of even (see 2.1). Nonetheless, the notion of mirativity appears to be fully relevant to an analysis of its meaning.

48What is intrinsically mirative in even is not the vast number of other elements on the scale, but the positioning on the scale of the element on which even bears, as the structuring property of the scale reaches a surprisingly high (or low) value. This explains why when the property is an effective and actual compatibility with the (type of) situation designated by the main verb, the value is low on the scale: with such a low value, it is remarkable that the situation still occurs. Now consider (34), in which even has scope over a PP introducing the agent:


The work we do is insufficiently understood or appreciated by the public at large or even by our colleagues. (ICE-GB: <s2b-042 007>)

49The potential risk for the speaker’s work to be insufficiently understood or appreciated is fairly high with the public, but presumably much lower with colleagues. However, that work is not valued by colleagues, as stated in (34), and the combination of a very low a priori risk and the actual occurrence of the situation expressed makes the latter very surprising. When the property is not an effective and actual compatibility, but a degree or quality, what is surprising is the height reached by the element on which even bears. In (35), the relevant property is the difficulty to leave the bus:


The layout of buses makes a quick getaway by trouble-makers difficult or even impossible. (ICE-GB: <w2d-009 089>)

50Impossibility is the highest value that can be assigned, and that this value is apparently reached must come as a surprise to the hearer, if not to the speaker as well. With even, surprise can then arise in two different ways, as a result of the “scalar” position assigned to the element on which it bears. Obviously, the positions assigned to elements that can be found in the vicinity of the most surprising one may also share in the sense of surprise coded by even.

51I have intentionally referred to surprise and not unexpectedness when discussing the mirative meaning of even. There are occurrences with comparatives in which the higher degree is given as both surprising and unexpected, as in (36):


And we can all see that this tax is virtually impossible to collect from some people so that those who do pay are being asked to pay even more to cover the cost of defaulters. (ICE-GB: s2b-030 117)

52Nonetheless, there might be no expectations altogether in some instances. For this reason, the semantic conception of mirativity that is adopted here is limited to the notion of surprise (to the speaker – and potentially to the hearer too). Surprise can be feigned by the speaker, who then adopts the point of view the hearer is assumed to hold. This is perhaps even clearer with (7), reproduced here as (37):


Replying to a message is even easier than sending. (ICE-GB: s2a-028 081)

53What will presumably come as a surprise to the hearer is not the fact that replying is even easier than sending, but that the easiness of replying is higher than an already very high easiness which is incidentally illustrated by the act of sending a message.

4.2 The meaning of even and inferences

54The view that I take in this paper is that even calls for the reconstruction of an implicit scale on which the element it bears on occupies a surprising position. Even is a mirative marker that requires hearers to draw inferences to determine the nature of the scale: its structuring property and the different elements taken into consideration and ranked on it (although all those elements do not need to be accessible to the hearer or given in full to represent the scale). It highlights the most outstanding element on the scale, whose positioning on it is surprising in itself. Even does not convey propositional meaning per se, but calls for determination in a way that is reminiscent of deictic and anaphoric constituents, so that an implicit scale can be adequately reconstructed. In this sense it can be said to carry instructional meaning, as it triggers a series of necessary inferences to that effect.

55Let us consider again the logical propositions that have been argued to be communicated by its use in the literature, to show that their derivation and its contextual constraints can be explained in the light of the meaning postulated for even.

  • 21 This observation suggests that, for even, mirativity only partly overlaps with concession: there ar (...)

56Fillmore’s (1964) claim that even conveys an expectation that is the opposite of what is imparted in the clause is almost always valid (and the claim has been taken up under various forms, as in Ranger and Méry 2007, Zeevat 2009, Leroux 2012). Surprise and unexpectedness are very close notions. Surprise usually comes as a consequence of something unexpected. But we have seen that the former can appear without the latter (36). Moreover, surprise arises from the positioning of one element on the scale, which is the most remarkable (the others might be as well, but to a lesser degree), whereas the unexpectedness envisaged by Fillmore relates to a situation. When there is no contradictory contextual indication – apparently with comparatives only, as in (36) – the mirative component of even translates into counter-expectancy for what is expressed in (sometimes part of) the clause. With comparatives, the surprise of having a degree that is higher than another one does not guarantee the prior and a priori unexpectedness of the situation connected to it. In that case, the marker even conveys highly frequently the proposition ‘One would normally expect the opposite of the (relevant) situation’ – though not always. The proposition should be regarded as implicated since it can be cancelled.21

57Horn’s (1969) notion that Even X did Y semantically presupposes ‘Not only X did Y’ can be maintained when even bears on individual entities (which can be contrasted to others on a scale), but it would have to be substantially modified to account for other cases, and especially so when even has scope over a comparative form. Yet it remains true that the scale conjured up by even always includes at least two elements, so that the use of even entails the presence of another element at least. Similarly, Ducrot’s (1972) analysis and suggestion that the utterance Even Jacques came communicates the proposition ‘Jacques’s coming is more significant than anyone else’s’ derives from the fact that if one element is the most surprising on the scale, then the situation relating to it is the most (or more) significant when the structuring property of the scale is actual and effective compatibility with that type of situation. When even bears on an individual entity, the highest (or higher) significance of the situation in which it is involved is entailed, along with the (inferable) proposition that expresses it. Much the same could be said of Karttunen and Peter’s (1979) treatment in terms of unlikeliness. With individual entities (on which even bears), the element that ranks lowest among the set considered for the scale is the least effectively compatible with the situation, and as such it is the most unlikely one (in the set under consideration, which is a significant constraint). Furthermore, as the relevant propositions discussed in those three accounts are felicitously communicated owing to the fact that even bears on individual entities, this raises the possibility that the inference it expresses is implicated since its derivation is contingent on this type of context, so that the inference – which is an implicature – can be cancelled, given another context.

58Further implicatures can also regularly be drawn from utterances in which even occurs. (26) is reproduced here as (38):


Vivid TV coverage of battlefield events and of bodies being brought home could have a remarkably restraining influence on even the most gung-ho generals and bellicose politicians (ICE-GB: <w2c-003 089>).

59If even the most gung-ho generals and bellicose politicians could be restrained by watching the TV coverage mentioned, it is a fair assumption that it is extremely bloody and sobering, much more so than one is naturally led to expect. It is worthwhile to compare (38) with the same utterance without even (Vivid TV coverage… could have a remarkably restraining influence on even the most gung-ho generals), which would generate a similar, weaker assumption. However, even contributes to drawing attention to the paradox between being one of the most unlikely types to be affected by war coverage and being actually potentially shaken by it. In other words, its use strengthens the assumption that would have been derived without it. In relevance theory, it is customary to regard the meaning of some discourse markers and connectives as procedural, i.e. bringing constraints on implicatures, guiding their generation and reinforcing or weakening some assumptions (Blakemore 1992, Wilson and Sperber 1993). Yet it appears that constraints on implicatures result from the meaning of the marker, as demonstrated for (38). The procedural component of even is then a consequence of its overall meaning, which pertains to mirativity.

4.3. Focalization (and end focus)

60The mirativity or sense of surprise expressed by even is correlated with the highlighting of, or focalization on, the element on which it bears (see 2.3). This sense of surprise is additionally emphasized by leaving the marker occur at the end of the phrase it has scope over, and thus enjoy end focus. As a result, the element on which it bears undergoes further focalization. ICE-GB contains 11 occurrences of what I will call final even. Here are two of them:


[Is there] anything else you’ve particularly enjoyed in your years so far not necessarily at College even that you might want to use and think about for career purposes [?] (ICE-GB: <s1a-035 162>)


The rate of unemployment in various parts of the United Kingdom which were most affected by closures in the steel industry are now lower even than the national average in this country. (ICE-GB: <s1b-055 087>)

61The scope of even is the PP at College in (39), and the comparative lower in (40). Even could have occurred in its canonical position in both utterances – with even at College and even lower. The sense of surprise communicated by even would then have been less marked or emphasized. The conditions for felicitous occurrences of final even remain to be fully elucidated, but they appear to be more widely attested in oral genres (8 occurrences in the oral component of ICE-GB out of a total of 11 for the whole corpus, whose written component is 40%).

5. Conclusion

62Even is a mirative marker that calls for the reconstruction of an implicit scale on which the element it highlights is (the most) surprising, in the position it is found. Its structuring property and the elements considered and ranked on it need to be determined inferentially. In the wake of this reconstruction, the hearer can derive further (cancellable) inferences: implicatures like ‘One would have expected the opposite’. The element that even highlights is pinpointed by the constituent over which it has scope. The potential syntactic categories for this constituent are more numerous than has been previously assumed, and even sometimes occurs at the end of it, with an extra emphasis on the sense of surprise conveyed. More research is nonetheless needed to account for final even fully, among other marked uses. Other questions have also been left aside, which deserve future investigation within the framework that I have proposed. First is the vexed and complex issue of the co-occurrence of negative markers with even, on which much ink has been spilled (see, e.g., Wilkinson 1996, Nakanishi 2012). Then are cross-linguistic considerations: even and French même, for instance, are far from being perfectly equivalent (Leroux 2012). Even more beautiful would usually be translated as encore plus belle, to take up a phrase made famous by the tale of Snow White. The potential competition (in some contexts) between even and some paradigmatic alternatives like still (Barker 1991, Michaelis 1993) also needs to be reexamined. Lastly, the relation between the mirative scalar marker even and even if, even though as well as frequent combinations like and even (Pennec 2016), or even, among others, should equally be addressed.

Top of page


ICE-GB Corpus. The British Component of the International Corpus of English. London: University College London (Survey of English Usage), 1998.

Abbott, Barbara. “Where have the presuppositions gone?” in Betty J. Birner & Gregory Ward (eds). Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn. Amsterdam: John Benjamins, 2006: 1-20.

Anderson, Stephen R. “How to get even.” Language 48: 4 (1972): 893-906.

Bach, Kent. “The myth of conventional implicatures.” Linguistics and Philosophy 22: 4 (1999): 327-366.

Barker, Stephen. “Even, Still and Counterfactuals.” Linguistics and Philosophy 14: 1 (1991): 1-38.

Blakemore, Diane. Relevance and Linguistic Meaning. Cambridge: Cambridge University Press, 1992.

Clark, Herbert. H. Arenas of Language Use. Chicago: The University of Chicago Press, 1992.

Close, Reginald A. A Reference Grammar for Students of English. Harlow: Longman, 1975.

De Haan, Ferdinand. 2012. “Evidentiality and mirativity.” in Robert I. Binnick (ed.) The Oxford Handbook of Tense and Aspect. Oxford: OUP, 2012: 1020-1045.

DeLancey, Scott. “Mirativity: The grammatical marking of unexpected information.” Linguistic Typology 1 (1987): 33-52.

Ducrot, Oswald. “Présupposés et sous-entendus.” Langue Française 4 (1969) : 30-43.

Ducrot, Oswald. Dire et ne pas dire. Paris : Hermann, 1972.

Ducrot, Oswald. Le dire et le dit. Paris : Minuit, 1987.

Fauconnier, Gilles. Étude de certains aspects logiques et grammaticaux de la quantification et de l’anaphore en français et en anglais. Paris: Champion, 1976.

Fillmore, Charles J. Entailment Rules in Semantic Theory. Project on Linguistic Analysis n°10. Ohio State University, 1965.

Garrido Médina, Joaquín C. “Connecteurs et opérateurs : même.” in Gerolt Hilty (ed.) Actes du XXe Congrès International de Linguistique et Philologie Romanes 1. Tübingen: Francke, 1993: 247-258.

Giannakidou, Anastasia. “The landscape of EVEN.” Natural Language and Linguistic Theory 25: 1 (2007), 39-81.

Hall, John Richard Clark. A Concise Anglo-Saxon Dictionary. 4th edition. Toronto: Toronto University Press, 1960.

Horn, Laurence R. “A presuppositional analysis of only and even.” Chicago Linguistics Society 5 (1969), 98-107.

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

Jaszczolt, Katarzyna M. Semantics and Pragmatics: Meaning in Language and Discourse. London: Pearson Education, 2002.

Karttunen, Laurie and Stanley Peters. “Conventional Implicature” in Choon-Kyu Oh and David A. Dinneen (eds.), Syntax and Semantics 11: Presupposition. New York: the Academic Press, 1979: 1-56.

Kay, Paul. “Even.” Linguistic and Philosophy 13 (1990): 59-111.

Kerbrat-Orecchioni, 1998 [1986]. L’implicite. 2nd edition. Paris: Armand Colin.

Leroux, Agnès. “Even et quelques traductions en français : la référence à l’inattendu.” LINX [online] (2012): 66-67.

Michaelis, Laura A. “‘Continuity’ within three Scalar Models: The Polysemy of Adverbial StillJournal of Semantics 10 (1993): 193-237.

Nakanishi, Kimoko. “The scope of even and quantifier raising.” Natural Language Semantics 20: 2 (2012): 115-136.

Nelson, Gerald, Wallis, Sean, and Bas Aarts. Exploring Natural Language: Working with the British Component of the International corpus of English. Amsterdam: John Benjamins, 2002.

Pennec, Blandine. “L’effet d’intensification lié à l’emploi de la locution AND EVEN en anglais contemporain” Faits de Langues, Varia 48. Bruxelles : Peter Lang editions, 2016 : 177-201.

Peterson, Tyler. “Rethinking mirativity: the expression and implication of surprise.” available on 1/11/2016 at:, 2013.

Potts, Christopher. “Presupposition and Implicature.” in Shalom Lappin and Chris Fox (eds). The Handbook of Contemporary Semantic Theory. Chichester: John Wiley & Sons, Ltd, 2014: 168-202.

Ranger, Graham, and Renaud Méry. “On Subject-Operator Inversion in English” in Christine Copy & Lucie Gournay (eds.), Points de vue sur l’inversion. Paris: Ophrys, 2007: 3-22.

Reboul, Anne and Jacques Moeschler. Dictionnaire encyclopédique de la pragmatique. Paris : Seuil, 1994.

Simonin, Olivier. “Typologie des contenus implicites.” Etudes de Stylistique Anglaise 5 (2012) : 27-42.

Stalnaker, Robert C. “Pragmatic Presuppositions”, in Milton K. Munitz and Peter K. Under (eds). Semantics and Philosophy. New York: New York University Press, 1974: 197-213.

Stalnaker, Robert C. “On the representation of context.” Journal of Logic, Language, and Information 7: 1 (1998): 3-19.

Wilkinson, Karina. “The Scope of evenNatural Language Semantics 4: 3 (1996): 193-215.

Wilson, Deirdre and Dan Sperber. “Linguistic Form and Relevance.” Lingua 90 (1993): 1-25.

Zeevat, Henk. “Only as a mirative particle”. Sprache und Datenverarbeitung 33 (2009): 179-196.

Top of page


1 In French, the distinction inherited from Ducrot is between présupposés and sous-entendus (which are implicatures).

2 My translation of “Une proposition P présuppose une proposition Q si, et seulement si, (i) dans toutes les situations où P est vraie, Q est vraie et (ii) dans toutes les situations où P est fausse, Q est vraie.”

3 I follow standard practice in using simple inverted commas to indicate logical propositions.

4 Garrido Médina (1993) provides a lengthier counter-argument.

5 Yet Kay (1990: 86) argues that it is far from obvious that Muriel’s voting for Hubert is more informative than anyone else’s voting for her. He obviously takes for granted that what is presupposed is always less informative than what is asserted – an assumption which could be challenged, especially from the hearer’s standpoint.

6 The position of even after hadn’t is canonical here, as it comes after the auxiliary, even though it includes the auxiliary and the negative n’t in its scope as well. In contrast, They even hadn’t thought about it is informationally marked, drawing more attention to the adverb and its meaning. According to my informants, the difference might be that the focalization involved on even is not as strong as in occurrences of final even (see 4.3.), or that there might be an additional emotional meaning conveyed in the latter case – such as annoyance in (6).

7 Situations or states-of-affairs are used to refer to events or states denoted by verbs and their expansions or clauses. For simplicity’s sake, I will from now on use the word situation only.

8 His characterization of présupposé excludes truth value considerations: the falsity of a présupposé in an utterance does not necessarily entail that of the asserted proposition (Ducrot 1972: 100-101).

9 Ducrot’s example and the présupposé he postulates are, respectively: Même Jacques est venu and La venue de Jacques est plus significative que la venue de quiconque.

10 For Karttunen and Peters, the two propositions are conventionally implicated: they are not presupposed because, they argue, their falsity does not affect the truth value of the utterance as a whole. Besides, as they are not asserted, they are assumed to be implicated, though conventionally – i.e. automatically, in virtue of the meaning of even, without drawing on Grice's maxims (or the implicature would have been conversational, the only other type that he considered in his framework). This makes conventional implicatures quite akin to entailments, a point that the authors fail to discuss. The notion of conventional implicatures has however undergone some serious criticism in the literature (Wilson and Sperber 1993, Bach 1999).

11 My translation of “Georges a bu un peu de vin, un peu de cognac, un peu de rhum, un peu de calva et même un peu d’armagnac.”

12 Fauconnier also noted that even scales are often probabilistic (1976: 262).

13 The “even phrase” corresponds to the adverb even itself and what it has scope over syntactically.

14 My translation of “Tout en homogénéisant les différentes occurrences, even souligne également le caractère (initialement) inattendu et donc remarquable de la validation par l’occurrence « extrême ». Cette occurrence se trouve ainsi mise en relief.”

15 That is, Prepositional Phrases (PPs) that complement a head noun.

16 The PP could also be parsed as adverbial, and not as an NP complement (influence here). However, On even the most gung-ho generals and bellicose politicians, vivid TV coverage of battlefield events and of bodies being brought home could have a remarkably restraining influence is not quite felicitous, thus arguing against an adverbial interpretation of the PP on even the most gung-ho generals.

17 This also includes entities that NPs refer to when the NP is found within an indirect or a secondary object PP (e.g. ICE-GB <w2a-010 018>: Princes were conditioned to believe in the justificatory, and even cathartic nature of war as a positive force in human affairs.) or within some other PPs that are reducible to a verbal (not copular) argument analysis, as in (26). What even semantically bears on is then denoted by part of the PP: the NP which complements the head preposition.

18 This appears to be frequent when even has scope over adjectives or AdjPs. Pennec (to appear) argues that and even should be regarded as a locution since it can be attributed properties that emerge from the combination of and and even. While this may be true with instances of coordination and juxtaposition ending with even more generally, the analysis suggested here for even still holds.

19 And some adverbials: She went swimming regularly, even daily. Here, the property is frequency of occurrence for the situation. Some predicative complements do not follow the same general pattern. This is for instance the case with predicative complements that are part of a cleft or pseudo-cleft construction: It's not the actual story or even the people that attract me to write about something (ICE-GB: <s1b-048 109>). The sentence can be rewritten The actual story or even the people do not attract me to write about something, and the semantic interpretation for even is the same as when it has scope over a subject NP, in the rewritten sentence.

20 Anderson himself writes: “Kuroda observes that, while even can appear in any of several positions in the sentence, it can generally appear in only one of these positions per sentence.” (1972: 893) The references to Kuroda are his 1965 MIT PhD dissertation (Kuroda, Sige-Yuki. Generative Grammatical Studies in the Japanese Language. Doctoral dissertation, Cambridge, Ms.: MIT. 1965) and a 1969 paper (Kuroda, Sige-Yuki. “Attachment transformations” in D. Reibel and S. Schane (eds). Modern Studies in English. Englewood Cliffs, N-J.: Prentice-Hall, 1969: 331-51).

21 This observation suggests that, for even, mirativity only partly overlaps with concession: there are circumstances in which even does not convey expectation of the opposite and no concessive meaning.

Top of page


Electronic reference

Olivier Simonin, « The mirative and scalar meaning of even », Anglophonia [Online], 26 | 2018, Online since 01 November 2018, connection on 21 July 2019. URL : ; DOI : 10.4000/anglophonia.1775

Top of page

About the author

Olivier Simonin

CRESEM, Université de Perpignan – Via Domitia

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
  • Logo DOAJ - Directory of Open Access Journals
  • OpenEdition Journals