Available online at www.sciencedirect.

com

ScienceDirect
Journal of Pragmatics 103 (2016) 1--14
www.elsevier.com/locate/pragma

What’s a distinct or alternative?§

Mira Ariel
Tel Aviv University, Israel
Received 19 March 2016; received in revised form 26 April 2016; accepted 6 July 2016
Available online 25 July 2016

Abstract
The alternatives introduced by or constructions are typically distinct from each other. Hurford’s (1974) constraint dictates that disjuncts
must not entail each other, which defines Distinctness as mutual nonentailment. In agreement with Simons (2001), I first argue that this
constraint is too strong and too semantic, but relying on the Relevance-theoretic concept of contextual adjustment (Carston, 2002), I call
for a more radical pragmatic shift. I include a discussion of Equivalence or constructions, where the alternatives are not referentially
distinct, arguing that again, it is pragmatic considerations that determine whether to impose Distinctness or not. The take-home message
is that pragmatics governs the variable application of the Distinctness constraint.
© 2016 Elsevier B.V. All rights reserved.

Keywords: Hurford’s constraint; Contextual adjustment; Sperber & Wilson relevance; Explicature; Privileged Interactional Interpretation

1. Introduction

It seems self-evident that or constructions introduce multiple and distinct alternatives. Surely, the syntax of the
construction requires that at least two disjuncts be included, and the semantics, as well as the discourse relevance of or
constructions, should make it necessary for multiple alternatives to be introduced, such that each disjunct stand for a
distinct option (Hurford, 1974; Simons, 2001; Singh, 2008). Indeed, consider (1):

1. MARILYN: do you want half --

. . . half of one of these,
or more than that. (SBC: 003)

Trivially, each disjunct here comes with its distinct linguistic form and meaning: half of one of these has a different linguistic
representation from more than that. Given the different linguistic forms and meanings, the two alternatives above naturally
denote two distinct referents as well. Finally, adopting a Relevance-theoretic approach, each of the alternatives carries its
own distinct (but mutually relevant) contextual implications with respect to a single topic (akin to Simons, 2001).1 In the

§
The research here reported was funded by the Israel Science Foundation (grant 431/15). I would like to thank the JoP reviewers for their
constructive comments.
E-mail address: mariel@post.tau.ac.il.
1
I prefer the Relevance-theoretic concept of ‘contextual implications’ over Simons’ ‘distinct answers to a Question Under Discussion (QUD)’,
first, because it more straightforwardly applies to nonassertions. But more importantly, we ultimately need to define multiple types of
distinctnesses, because different or constructions require different subsets of such distinctnesses (see Section 4 and Ariel, 2015a). Distinctness
in contextual implications is the strongest of these constraints and entails the others.

http://dx.doi.org/10.1016/j.pragma.2016.07.001
0378-2166/© 2016 Elsevier B.V. All rights reserved.
2 M. Ariel / Journal of Pragmatics 103 (2016) 1--14

case of (1), Marilyn will perform different actions, depending on the alternative chosen by her addressees (as is made
clear in the discourse, she’ll either cook the whole fish, or else only some of it and make ceviche from the leftovers).2
But which of the above distinctnesses is the crucial distinctness which defines felicitous or constructions? And even
more basically, what does it mean for an or construction to introduce multiple alternatives? The goal of this paper is to
problematize the requirements for distinct (multiple) alternatives within or constructions, initially based on the Santa
Barbara Corpus of spoken American English (Du Bois et al., 2000--2005) (SBC, henceforth). But the analysis here
presented is not restricted to SBC, because it is not large enough to include all cases pertinent to the Distinctness issue.
Examples were therefore also collected from web searches, as well as from miscellaneous personal encounters. Section
2 argues for a pragmaticized approach to Hurford’s constraint. I further demonstrate the crucial role of pragmatics in
establishing Distinctness in Section 3. Section 4 introduces a set of or constructions where referential Distinctness is
absent. I conclude with Section 5.

2. Pragmaticizing Hurford’s constraint

Consider the following example:

3. I do not want these messages at all, but the box doesn’t give me that option. So I could (1) get the messages
or (2) get the messages. What kind of choice is that?
(John C. Dvorak http://www.pcmag.com/article2/0,2704,2251579,00.asp)

Examples such as (3) are very rare, and achieve their special (’inevitability’) effect (see Ariel, 2015a) precisely by violating
a very basic condition placed on or constructions, namely that they should introduce multiple, distinct alternatives. Section
2 discusses the nature of such a constraint. I start off with Hurford’s (1974:410) purely semantic constraint, but extending
Simons’ (2001) analysis, I propose that the ‘Distinctness’ constraint is thoroughly pragmatic.
Hurford’s (1974:410) constraint, originally part of an argument for the ambiguity of natural language or, is a prime
candidate for articulating a rather strong condition of ‘Distinctness’ of alternatives for or constructions: ‘‘The joining of two
sentences by or is unacceptable if one sentence entails the other’’.3 Indeed, if they do, the or construction would seem to be
pointless, given that disjunctions are supposed to be about options profiled as genuine alternatives to each other (as in 1).
The typical unacceptable examples which motivated Hurford’s constraint are not X or X identity cases as in (3), but rather,
semantically different alternatives where one of the disjuncts entails the other (in 4, if John is a Californian, then he’s an
American; if there’s dirt in the fuel line, then there’s something in there)4:

4. a. ?? John is an American or a Californian (= Hurford’s ex. 15).

b. ?? John is a Californian or an American.
c. ?? Either there’s dirt in the fuel line or there’s something in the fuel line (=Simons’ 2001:ex. 23).

Indeed, note that B in (5) cannot properly answer A. Instead, he informs her of what she must not have known, that
‘Zinfandel’ entails ‘red wine’:

5. A: Would you like red one?

Or Zinfandel?
B: Zinfandel IS red wine (Feb. 24, 2013).

The ban against an entailing relation between the disjuncts naturally accounts for the unacceptability of (4), as does
Simons’ (2001) ‘Non-Vacuity Principle’, which specifies that the or construction as a whole must express a different
proposition from any one of its disjuncts (see also Katzir and Singh, 2013 anti-redundancy constraint).
But Hurford’s constraint is too strong, argues Simons. She points to the different felicity of 6(A), where the second
disjunct entails the first, as a response to Q1 (it is infelicitous) as opposed to a response to Q2 (it is felicitous)5:

2
In addition, I account for the relevance between the alternatives by reference to a ‘Higher Level Category Constraint’, namely, by imposing a
condition that all the alternatives must be construable as members of a single, possibly ad hoc, context-relevant, higher-level category. But most
likely this is essentially the same as the requirement for a single QUD.
3
But we need the constraint for nonclausal disjunctions as well, of course, and I will assume it applies to such cases.
4
An initial  indicates an unattested example.
5
I slightly modified Simons’ original example to ensure that the second disjunct entails the first one.
 M. Ariel / Journal of Pragmatics 103 (2016) 1--14 3

6. Q1: What kind of car does Jane drive?

Q2 : How are they going to move all their stuff?
A: Well, either Jane has a truck, or she has an old truck and George has a
Station wagon. (Either way, they can get everything into their own cars).

According to Simons, the difference stems from the fact that A’s disjuncts provide two distinct answers to a single
contextually relevant question in the sequence of Q2 + A, but not so in the sequence of Q1 + A. In other words, an
entailment relation does not automatically rule out a felicitous or construction. What determines the acceptability of the
construction according to Simons is whether or not the disjuncts introduce distinct answers to a single question. A does
that when preceded by Q2, but not when preceded by Q1.
While Simons recognizes the role of the ad hoc context, her ‘‘rescue’’ strategy hinges on finding a contextually relevant
Question Under Discussion (QUD) where the disjuncts manifesting an entailment relation nonetheless constitute distinct
responses. But the (non)entailment relations themselves are defined over purely semantic representations, which is why
she considers as unacceptable in any context her ex. 23(=4c) above). The next set of examples will show a deeper role for
pragmatics in establishing the felicity of or constructions, beyond framing the discourse topic (or QUD).
Gazdar (1979) pointed to a different set of counter-examples to Hurford’s constraint (and in effect to Simons’ constraint
as well). The examples in (7) are perfectly acceptable, despite the entailment relation:

7. a. Many of them know it [the Koran---MA], most or all of it, by heart. . .

(Originally Hebrew, The Islamic Museum, Jerusalem, December 2003).
b. He wants to hear it from Beckman or Rosenshine or both of them (LSAC).6

Examples such as 7(a) are cases where the semantic meaning commonly attributed to quantifier most does not preclude
‘all’, which means that the second disjunct entails the first. Similarly for (b), where or is taken to be semantically ‘inclusive’.
Gazdar’s solution is that Hurford’s constraint is suspended when a scalar implicature (‘not all’ in (a), ‘not both in (b)) can
block the entailment. Others have proposed not to suspend Hurford’s constraint, but instead, to explain how the disjuncts
here do not maintain an entailing relation after all. We can see (7) as offering distinct alternatives (distinct answers to a
single question, according to Simons) if we integrate the scalar implicatures into the semantic meanings: ‘most but less
than all’ and ‘all’ in (a), ‘Beckman (alone)’, ‘Rosenshine (alone)’ or ‘both Beckman and Rosenshine’ in (b). Neo-Griceans
(Horn, 2006; Levinson, 2000) have then set aside certain cases of scalar implicatures as ‘‘special’’, ‘‘retroactive’’ or
‘‘intrusive’’ implicature cases. In such cases, the scalar (pragmatic) implicatures (’not all’ in a, ‘not both’ in b) participate in
the truth-conditional content of the proposition. Hurford’s constraint can then apply to these pragmatically strengthened
meanings, which account for the acceptability of (7). According to Relevance theory, pragmatic inferences routinely
‘‘intrude’’ on semantics (Carston, 1990, 2002; Sperber and Wilson, 1986/1995). Scalar ‘‘implicatures’’ often constitute part
of the explicature, namely the truth-bearing representation, so according to them there is actually nothing special about
(7).
Other solutions to Gazdar’s counter-examples are semantic, rather than pragmatic. I have argued for different
analyses for the linguistic meanings of scalar quantifiers and of or. On my analysis, scalar quantifiers such as most in
(7) linguistically encode an upper-bounded quantity (larger than 50%) (see Ariel, 2004, 2015b), and or constructions are
neither discoursally ‘exclusive’ by default, nor linguistically ‘inclusive’ (Ariel and Mauri, 2015). Chierchia et al. (2009) have
proposed a different account by reference to a grammatical Exhaustification operator which derives scalar ‘‘implicatures’’
as part of the compositional meaning. The upshot of these two proposals is that there is no entailment relation between the
disjuncts, and this is so independently of pragmatics.
But these independently motivated linguistic accounts are not available for other cases, where rather ad hoc scalar
implicatures (Hirschberg, 1991) are called for in order to view the disjuncts as offering distinct alternatives. Note that the
attested examples in 8(a--e) virtually form minimal pairs with (4):

8. a. Is there anyone who lives in an Arab country or Lebanon?

(www2.dailyroxette.com/node/6543.html)
b. A pope from the Middle East or Lebanon--someone able to speak the language of the Muslim, Islamic, or Hindu
faith
(www.myfoxchicago.com/. . ./cardinal-george-popes-
resignation)

6
LSAC, The Longman Spoken American English Corpus, is a 5-million word corpus.
4 M. Ariel / Journal of Pragmatics 103 (2016) 1--14

c. Being neither from the Middle East or from Israel, I would actually have to side with Palestine.
(https://ca.answers.yahoo.com/question/index?qid=20090122122840AAR7ZVF)
d. I was never tempted by the calls of Israeli prime-ministers to leave Europe or Francei, wherei I live, ‘‘before it’s too
late’’. . . They tell me to leave a continent which. . ., and in France. . . (Haaretz, March 20, 2015)7
e. LINDSEY: and he’s got some bones,
. . . or something in there. (SBC: 018).8

Thus, Arab country must be ‘‘strengthened’’ into ‘an Arab country other than Lebanon’, and the Middle East must be
strengthened to ‘the Middle East exclusive of Israel’. These interpretations depend on quite ad hoc scalar implicatures. In
light of such examples, one could simply reject Hurford’s constraint, which is Pott’s (2013) position. But we still need to
account for the unacceptability of (4) and (5): We need some Distinctness constraint.
In fact, Relevance theory, I propose, offers a natural solution for the problematic cases above, maintaining Hurford’s
constraint, without recourse to any special mechanism. The Relevance-theoretic claim is that bare semantic meanings
are routinely adjusted in context in order to convey the speaker-intended ad hoc concepts (see especially Carston, 2002).
This occurs as part of the process of developing a relevant explicature (the speaker-intended truth-bearing proposition)
out of the under-specified linguistic meaning. For example, the first something in the subway sign If you see something say
something is narrowed to an explicated ‘something alarming’. (9) is a classical example from Deirdre Wilson, where it is
the pragmatically inferred causal interpretation between the conjuncts that creates a distinction between the two
otherwise semantically identical disjuncts:

9. It’s always the same at parties: either I get drunk and no-one will talk to me or no-one will talk to me and I
get drunk (Wilson and Sperber, 1998: ex. 3).

While the semantic meanings of the alternatives in (9) are not distinct, the developed explicatures (’I get drunk and
therefore no-one will talk to me’, ‘no-one will talk to me and therefore I get drunk’) are distinct. (10) is an attested example
where the speaker is trying to elicit a choice between two alternatives from the addressee. The alternatives must then be
distinct, despite the fact that they seem identical:

10. Interviewer: How are you?

Interviewee: How are you is a tricky question.
How are you?
or how are you.
(Originally Hebrew, Reshet Bet, Nov. 20, 2014).

Once we provide different explicatures to the how are you’s (one as a formal greeting, the other as a genuine question) the
two disjuncts stand for different alternatives.
Indeed, narrowing the meaning of linguistic expressions is essential for the examples we considered above. Going
back to (8), Arab country (in a) and the Middle East (in b and c) must be narrowed to ‘an Arab country other than Lebanon’
(a) and ‘the Middle East other than Lebanon/Israel’ (b,c).9 Europe must be narrowed into ‘Europe exclusive of France’ in
(d). In (e) something is narrowed to ‘some object other than bones (that the dog under discussion must have swallowed)’.
Consider in this connection the English translations of Samuel 2, 17: 9 (11b,c), where the Hebrew version literally
translates as (a):

11. a. . . . he is hidden now in one of the caves or one of the places

b. . . . he is hidden in a cave or some other place (New International Version)
c. . . . he is hid now in some pit, or in some other place (King James Version).

It seems that the translators chose to explicitly assert what was originally left to inference.10 Once we perform these
contextual narrowings, there is no entailment relation between the disjuncts. So, it looks like Hurford’s constraint, as well
as Simons’ ‘Non-Vacuity Principle’, may be maintained, provided we apply it to the pragmatically enriched explicature,
rather than to the bare semantic meaning.

7
The leaving under discussion in (d) is for Israel, which is why leaving France necessarily entails leaving Europe.
8
Note that this or something, where something is accented, is not a ‘Hedged X’ constructional case (discussed in Ariel, 2015a).
9
See Levy et al. (2014) for a similar interpretation of roses and flowers as ‘roses and flowers other than roses’.
10
I thank Itai Kupershmidt for sharing with me these translations.
 M. Ariel / Journal of Pragmatics 103 (2016) 1--14 5

Next, the need to take into account implicit interpretations in order to view an or construction as appropriate is not
restricted to explicatures or to ‘‘scalar implicatures’’, be they of whatever status. Consider the phone message in
(12), where in order to count as a relevant disjunction, the second disjunct must be enriched into something like ‘are you
sleeping there’, and the first alternative must incorporate a Background assumption (Searle, 1980) that the addressee is at
the same time awake, so that the reality behind the first disjunct is something like ‘are you there and awake’. Only in this
way is the entailment blocked:

12. JILL: .. (TSK) Hey cutie pie,

are you there?
. . . (H) Or are you sleeping. (John Du Bois p.c: Cutiepie)11

These additions make sure that the two disjuncts introduce alternatives that are relevant to each other and do not entail
each other. Background assumptions are not scalar implicatures, however, and they don’t form part of the explicature (nor
of ‘what is said’). They are, however, meant, and very much context-dependent. So, if it is the case, for example, that
whenever Jill’s addressee is ‘‘there’’ he is sleeping, the or construction in (12) will be infelicitous, because we won’t have
two distinct alternatives. An alternative way to account for the felicity of (12) is by reference to Particularized
conversational implicatures (PCIs) (based on the explicit or construction in conjunction with the Background assump-
tions). On this account, Jill’s are you there (+ an added contextual assumption ‘and awake’)? serves as a basis for deriving
‘can you pick up the phone?’ as a PCI, and are you sleeping (+ ‘there’) gives rise to a PCI of ‘are you not able to pick up the
phone?’. Either way, it’s clear that uncontroversially pragmatic inferencing is implicated in producing the necessary
Distinctness between the alternatives, thus ensuring the felicity of or constructions.
13(a) is a more radical implicature case. What JDB actually wanted to know was who was the original babysitter to be
hired, as opposed to the second babysitter (’her sister’) who was later on introduced by the first one. The real alternatives
here are then as in the strongly implicated (b)12:

13. a. JDB: Yuka is the sister of Mika?

Or Mika is the sister of Yuka? (22 February 2001).
b. Is Yuka the nonoriginal babysitter?
Or is Mika the nonoriginal babysitter?

While the crucial representation for assessing the felicity of or constructions with respect to the Distinctness of its
alternatives is most often the explicature, on some occasions, it is the strong PCI(s) associated with the utterance. Cases
where meaning levels other than explicatures constituted the most relevant speaker-intended contribution in discourse
motivated my proposed Privileged Interactional Interpretation representation (PII) (Ariel, 2002). PII stands for the
discourse-relevant contribution of the speaker, namely, the basis for judging what she says as true and as relevant
(see Jaszczolt, 2005 for a similar, Primary Meaning). Still, typically it is the explicature that counts as PII.

3. It’s even more pragmatic: rescues are not guaranteed

Examples (8)--(13) demonstrate the crucial role of pragmatic inferencing in ‘‘rescuing’’ or constructions by rendering
the alternatives relevant to each other, as well as distinct enough. All the examples were ‘‘success stories’’, where
pragmatic inferencing applied and rescued what seemed to be nondistinct or nonrelevant alternatives. The goal of Section
3 is to show that pragmatic rescues cannot be freely generated just in order to rescue an or construction. They must be
independently pragmatically plausible, i.e., functional in developing the Relevant PII.
Singh (2008) has already argued that not all Hurford constraint violations are ‘‘rescuable’’, based on minimal pairs such
as (14):

14. a.  John ate some of the cookies or he ate all of them (Singh’s ex. 5).
b. #John ate all of the cookies or he ate some of them (Singh’s ex. 11).

11
This recording is from a voicemail message that precedes the conversation known as Cutie Pie (SBC: 028).
12
Incidentally, note that examples 8--13 show that Hurford’s Constraint does not provide evidence for the grammatical (as opposed to
pragmatic) account of scalar implicatures (Chierchia et al., 2009) since unquestionably pragmatic processes (i.e., inferences that do not involve
Exhaustification) may ‘‘rescue’’ the constraint.
6 M. Ariel / Journal of Pragmatics 103 (2016) 1--14

In order for the alternatives in (14) to be distinct from each other, some must be narrowed to ‘some but not all’. This is
equally true for (a) as it is for (b). But since he finds 14(b) unacceptable, Singh argues for an asymmetry between disjuncts,
such that only early disjuncts can be strengthened in order to avoid a Hurford constraint violation. I don’t accept this
judgment.13 14(b) seems perfectly acceptable, for example, against a background where we’ve just found out that the
cookies we had, all gone now, were poisonous, so we need to find out who to rush to the emergency room. If John ate all
the cookies, it’s only him. If he only ate some of the cookies, we need to find out who else ate the cookies. In any case, here
are a few attested all or some examples:

15. a. You can burn all to DVD’s or some it’s your choice.
(http://www.bleedingcool.com/2014/01/04)
b. A categorical statement is an assertion or a denial that all or some
members of the subject class are included in the predicate class.
(en.wikipedia.org/wiki/Categorical_proposition)
c. ALL OR SOME Cleaning Services is a company that lets you decide
what needs to be cleaned from the floor to the ceiling and everywhere. . . (eauclaire.craigslist.org)

And here are nonquantifier examples where the second disjunct must undergo contextual adjustment in order to constitute
a ‘distinct’ alternative from the first one:

16. a. They thought he was from Lebanon or an Arab country, (LSAC)

b. JO: . . . You know they take old pants,
that were . . . suit pants or,
. . . or nice pants?
And then they’re.. all spotted,
and they wear em anyway? (SBC: 059)
c. JIM: . . . when we’re going to,
(H) uh,
.. say blow up the first atomic bomb,
(H) and and and,
we don’t bother,
to ask the American people,
or the people of the world,
if w- if if we mind, (SBC: 017)

It is the subsequent disjuncts that must be narrowed in (16): ‘An Arab country other than Lebanon’ in (a), ‘nice pants other
than suit pants’ in (b), ‘people of the world other than Americans’ in (c). Clearly, then, subsequent disjuncts are narrowable
too.
Singh (2008) makes an even bolder move to maximize Distinctness and to restrict our ability to ‘‘rescue’’ or
constructions. Based on the unacceptability of (17) he proposes a stronger version of Hurford’s constraint:

17. #John ate some of the cookies or he ate at least three of them (Singh’s ex. 18).

It’s not enough for the alternatives to be distinct in that neither one entails the other, argues Singh. The two disjuncts must
be mutually inconsistent for a felicitous or construction. (17) is unacceptable, since even when some is enriched into
‘some but not all’ it is still consistent with ‘at least three’, he argued.
This is too strong a constraint against ‘‘rescues’’, however, in view of the attested examples in (18):

18. a. As I’m sure most of you . . . or some of you . . . or at least three of you know, this Sunday, Feb. 9, is
National Bagel Day.
(http://www.nashvillescene.com/bites/archives/2014/02/07).

13
A Google search (Dec. 5, 2014) for some or all and all or some revealed that the former was indeed much more common (1,370,000 hits), but
the latter seems far from unacceptable, with 235,000 hits. A comparison between coffee or coffee with milk and coffee with milk or coffee (in both
English and Hebrew) again showed more of the former, but this search yielded vanishingly few occurrences for both orders. In the overwhelming
majority coffee was modified (typically, as black coffee) in both orderings. So at best, the two comparisons show a nonabsolute preference for
placing the pragmatically adjusted term first. Now, of course, Google searches can support conclusions only tentatively, but the examples cited in
(15) are perfectly acceptable.
 M. Ariel / Journal of Pragmatics 103 (2016) 1--14 7

b. I didn’t write any of the songs that will appear in the book, but some of them, or at least three of them,
I rewrote the lyrics.
(http://kicsterash.deviantart.com/art/Soul-Speaking-210174055)

What can the difference be between the unacceptable (or less acceptable) (17) and the obviously acceptable (18)?14 I
propose that when eating cookies is concerned some cookies and at least three cookies are simply not distinct enough
under normal circumstances, because both expressions are most likely similarly pragmatically narrowed (to an explicated
‘3--5 cookies or so’). But when the total relevant number is relatively large, the difference between the narrowed some and
the narrowed at least three may be significant enough to matter. This is what we see in (18), where you refers to many
people. The determining factor is that the difference between or construction alternatives, as defined by their contribution
to the relevant Privileged Interactional Interpretation (here, their explicatures), must matter to the discourse point at hand.
This goes back to Simons’ (2001) insight in arguing that appropriate alternatives must constitute distinct answers to a
single QUD. Whatever difference may be derived between the alternatives in (17) is not discourse-relevant enough. The
differences in (18) are.15 The fact that the disjuncts are equally consistent with each other in all three cases has no effect
on their acceptability.
But the question remains, why is it that interlocutors can’t impose a contextually adjusted interpretation for either some
or for at least three in (17), thus rescuing the or construction? I am not in a position to provide a complete account of this
issue, but I propose that the explanation lies in the fact that contextual adjustments cannot be freely generated with the
sole goal of rescuing the or construction. The pragmatic narrowing required must make sense in the general scheme of
things as well. So, in fact, I agree with Singh that the contextual adjustments needed in order to rescue or constructions are
not totally unconstrained. But I offer a different explanation for the restriction. It seems that the degree to which pragmatic
narrowing is contextually available for the relevant disjuncts plays a crucial role in rendering distinct potentially entailing
disjuncts, and hence the difference between (17) and (18), and possibly between (4) and (16) as well.
Here are two cases where such a rescue narrowing is indeed independently contextually supported:

19. a. MARIE: .. people that are that low there’s,

.. they have asthma and stuff like that.
((2 LINES OMITTED))
LISA: Yeah.
Or emphysema. (SBC: 036).
b. S: I do not know if they sold or advertised, but we didn’t sell
anything, no apartment and a week ago suddenly ((people))
came in, bought. . . (Originally Hebrew, Lotan: 6).

Asthma and stuff like that can in principle include emphysema, because like asthma, emphysema too involves a
malfunctioning of the lungs, which causes breathing difficulties. So the first disjunct in 19(a) must be narrowed to ‘asthma
and other respiratory illnesses exclusive of emphysema’ for the or construction to be felicitous. This is indeed a
reasonable case of contextual narrowing, because emphysema is an extremely severe illness, and therefore not
necessarily included under the contextually adjusted concept of ‘illnesses like asthma’. The same is true for 19(b),
where ‘sold’ and ‘advertised’ (apartments) are mutually consistent with each other. The contextually relevant narrowing is
‘sold or only advertised (but did not sell)’.16
Here are two other contextually supported rescues:

20. a. RICKIE: but you couldn’t they couldn’t see nothing,

or anybody, (SBC: 008)
b. LINDSEY: if he doesn’t get better,
. . . or if he gets worse,
.. be sure and let us know. (SBC: 018)

14
See also the question of the first witch in Shakespeare’s Macbeth (Act 1, scene 1): ‘‘When shall we three meet again? In thunder, lightning,
or in rain?’’
15
Simons (2001) also makes the complementary point that nonentailing alternatives (e.g., ??Either she drives a Subaru station wagon, or
George drives a Toyota and she drives a Subaru) may nonetheless fail to provide distinct answers (distinct contextual implications on my
account), which is why they are infelicitous.
16
Note that narrowing sold to ‘only sold but did not advertise’ would equally ‘‘rescue’’ the disjunction, but makes no sense here as a PII.
8 M. Ariel / Journal of Pragmatics 103 (2016) 1--14

Nothing must here be narrowed to ‘no inanimate entities’ and doesn’t get better to ‘remains in the same condition’ in order
for these or constructions to express distinct alternatives. But again, the rescues in (20) rely on independently plausible
contextual adjustments.
However, here’s a famous experimental probe, where narrowing the linguistic meaning seems unavailable. Presented
with 16 flowers, eight of which were primulas, and asked to respond to a Choice or construction, Are there more flowers or
more primulas? older children chose ‘‘more flowers’’ more often than younger children did (Inhelder and Piaget, 1964).
This result was interpreted as pointing to only a gradual maturation into understanding logical inclusion relations (that
‘primulas’ are included under ‘flowers’). But note that the experimental sentence is at the same time a test for whether or
not ‘flowers’ here is or is not reasonably narrowed to ‘flowers other than primulas’. If it is, then there aren’t more ‘‘flowers’’
than primulas (because there are 8 primulas and 8 nonprimula flowers), but if it isn’t, then there are more flowers than
primulas (because there are 16 flowers, including the 8 primulas). It is the older children, then, that rejected the narrowed
interpretation more often, which, assuming that ‘‘the older the wiser’’, means that this is not a context where such a
pragmatic enrichment is appropriate. I am guessing that response times here were quite long, reflecting the fact that on the
one hand, narrowing ‘flowers’ to ‘flowers other than primulas’ is not highly available, but at the same time not narrowing it
creates a Hurford’s constraint violation. But I expect that Is there more bread or more toast (under parallel circumstances,
where there are 8 slices of toasted bread and 8 slices of nontoasted bread), should elicit far less ‘‘more bread’’ responses
(and possibly faster too), just because the contrast between the narrowed ‘nontoasted bread’ as opposed to ‘toasted
bread’ (as well as ‘English muffins’) is culturally more readily available, as can be seen in the following17:

21. KRISTIN: it would also be something kind of easy to do with toast,

.. or bread, ((2 LINES OMITTED))
or English muffins, (SBC: 041).

The next pair of examples shows that what assumptions addressees attribute to the speakers may determine the
likelihood of a rescuing contextual adjustment:

22. a. I want to play for a team in either Europe or Spain

(http://en.wikipedia.org/wiki/Alon_Mizrahi)
b. And anyway the problems of LB ((Lehman Brothers -- M.A)) did not come from either Europe or
Spain. . . (http://sustainable-spain.com/2012/05/)

22(a) was uttered by an Israeli soccer player, (b) by the Spanish minister of economy and competitiveness. While (b) was
rescued (into ‘all of Europe’ or ‘only Spain’), (a) was not, although it could have been, by restricting Europe to ‘Europe
exclusive of Spain’. In fact, the (a) speaker immediately became a laughing stock, and the disjunction (as a collocation)
stands for ignorance to this day (the original quote is from 1997). Once again, narrowing for the sake of rescuing an or
construction is not guaranteed.18
So far, we have seen that once we apply Hurford’s constraint to a contextually plausible enriched representation we
can at least preserve the insight behind it, namely, that the alternatives introduced in an or construction are distinct. The
pragmatic adjustments we applied were the same for the most part, namely, we excluded the entailing disjunct (e.g.,
Lebanon in 8) from the entailed one (Arab country or Middle East in 8). In the following examples, however, the contextual
adjustment is different (and see again 22b):

23. a. MILES: . . . But that’s what she said.

.. Now I don’t know if she meant the Bay area or San Francisco, (SBC: 002)
b. The role of those concerned with ageing in Lebanon or the Middle East is to provide communities
and concerned professionals with the knowledge and skills to. . .
(arabhealthmagazine.com/geriatrics-and-gerontology-in-the-mi. . .)

17
I’m wondering what the results might be for more cows/dogs or more bulls/female dogs, namely whether it is easier to narrow cow to ‘female
cow’ and dog to ‘male dog’. But perhaps these cases do not involve narrowing, but rather, choosing the contextually more appropriate meaning for
lexical items that are by now ambiguous between neutral and specific sexes.
18
And see again (5), where B chose not to rescue A’s ‘red one or Zinfandel’ to ‘red wine other than Zinfandel or Zinfandel’ or to ‘red wine or White
Zinfandel’, most likely because he assessed that A simply did not know that Zinfandel is a type of red wine. It’s quite possible that a rescue would
have been performed had A been considered knowledgeable about wines.
 M. Ariel / Journal of Pragmatics 103 (2016) 1--14 9

The entailed disjunct (the Bay area in a, the Middle East in b) must here resist narrowing to e.g., ‘the Bay area except for San
Francisco’. Instead, it is explicated as ‘the whole Bay area’, which is how it is no longer entailed by ‘San Francisco’. The
person cited in 23(a) talked about an alarmingly high rate of HIV positive males, so most likely the statistics she is asking
about is for either the whole Bay area (San Francisco included) or for San Francisco alone (and not for the Bay area
exclusive of San Francisco or San Francisco). The same is true for (b), which is different from (8) and 16(a), where a person
may only come from a single country in the Middle East. So, like (22), (23) too shows that rescues are context-sensitive.
Note that 23(a & b) require the same rescue procedure as Inhelder and Piaget’s ‘‘Are there more flowers [of all types,
M.A] or more primulas?’’ The reason why (23) are much more natural than the flower question is that the explicated ‘Bay
area in general’ makes much more sense in its context than ‘flowers of all types’ does in its context. Once enriched into
‘flowers of all types’ the experimental question becomes too trivial to be felicitous. Indeed, the flower question is a ‘‘trick’’
question, while there’s nothing tricky or special about the constructions in (23).
This is the crux of the matter. Hurford and Singh (see also Katzir and Singh, 2013) did get something right: The
alternatives in or constructions, particularly in cases such as (1), must be distinct. But strictly semantic solutions cannot
account for all the data. Distinctness between alternatives cannot be determined based on semantic representations
alone, and it cannot be measured by the absence of an entailing relation (Hurford) or incompatibility (Singh), even when
applied to pragmatically enriched representations. Rather, distinctness between alternatives is determined relative to the
specific context in a number of ways. First, what the discourse topic is determines whether the disjuncts express distinct or
nondistinct alternatives, which is why the same or construction may be acceptable in one context, but not in another
(22, and 30 below) (see originally Simons, 2001). Second, I have argued that it is the pragmatically derived Privileged
Interactional Interpretation, most often the explicature, that is taken into account when (non)Distinctness is established.
Moreover, the narrowing of the disjuncts into the relevant explicatures is also context-dependent. This is why on some
occasions no contextual adjustments are available in the specific context (despite the theoretically open option to narrow
one of the disjuncts). We can’t assume a general instruction to automatically mobilize pragmatics in order to create a
distinction between alternatives that are not semantically distinct enough. Rescues must at the same time also be
psychologically accessible and pragmatically compatible with the specific context. Singh’s (2008) more radical semantic
(inconsistency) and processing (asymmetry) constraints had to be rejected because they were too strong, but I believe
that the motivation behind his proposals is exactly right, namely, to ensure that or constructions introduce distinct
alternatives with distinct contextual implications regarding a single QUD (Simons, 2001).19

4. Equivalence or constructions

Section 4 further demonstrates the crucial role of contextual considerations involved in the decision whether to narrow
disjuncts into Distinctness or not. I here discuss Equivalence or constructions, where the addressee is specifically
expected to not enrich the disjuncts into referentially distinct alternatives. In other words, Hurford-based rescues must
here be avoided. Once again, it is pragmatic considerations that determine whether a rescue is performed or not, in which
case an Equivalence or construction results.

4.1. Introducing Equivalence or constructions

According to the OED, an ‘‘otherwise called’’ or has been attested since 1357. Horn (1985, 1989:379--380) has
distinguished between logical and nonlogical (metalinguistic) uses of or constructions, and Ball (1986) similarly
distinguishes between propositional and nonpropositional alternatives. (24), where the second disjunct is referentially
identical to the first one, demonstrate one type of a metalinguistic or20:

24. a. BRAD: So you are gonna want,

. . . (H) or desire a pretty.. decent,
. . . uh,
. . . copy of your original tape. (SBC: 016)
b. BEN: .. (H) As concrete hardens,
or cures as we call it,
(H) it produces heat. (SBC: 038)

19
In fact, in Ariel (2015a) I argue that inconsistency (pragmatically construed) is very often characteristic of specific or constructions (Choice,
Indifference).
20
Burton-Roberts (1994) and Haspelmath (2007), however, view such ors as ordinary ors. Indeed the same alternativity relation holds here,
except that it operates over linguistic, rather than referential alternatives.
10 M. Ariel / Journal of Pragmatics 103 (2016) 1--14

c. RICHARD: . . . Even if she goes out with other men,

or dates other men, (SBC: 047)
d. BRAD: (H) dubbing,
.. or copying, ((1 LINE OMITTED))
from one tape deck to the other. (SBC: 016)

It is only in terms of the specific linguistic formulations that the two disjuncts constitute distinct alternatives to each other
here. Indeed, since the disjuncts amount to a single alternative any one of them can be omitted with no effect on the truth-
conditional content of the utterance.21 Blakemore (2007) treats such appositional or constructions as reformulations, but,
following Ball (1986) I prefer to define them as Equivalence or-constructions, capturing the idea that the two disjuncts are
equivalent to each other in their functioning in the given context. Jakobson (1960) has called attention to utterances where
the speaker ‘‘projects the principle of equivalence from the axis of selection into the axis of combination’’ (p. 358, original
emphases). Now, most or constructions actually combine alternatives which in effect stand in a paradigmatic opposition,
instead of selecting one of them. But that’s often because the speaker is not in a position to select one of the alternatives
(in 1, for example, the speaker is asking her addressees to make the choice). In Equivalence or constructions the speaker
refrains from making a choice between alternatives that she could have in fact made. For example, Brad could have
chosen only want in 24(a).
Equivalence or constructions are not very frequent, but I did find 32 such cases in SBC (3%). Note that the alternatives
in Equivalence or constructions are not necessarily semantically identical, but they more often than not denote a single
discourse entity in the relevant context (Blakemore, 2007). Note the singular agreement in (25):

25. BRAD: . . . the,

. . . dual,
. . . or double cassette deck,
.. (H) that Luxman offers,
(H) is basically,
. . . a.. um,
. . . (TSK) double version.. of the kay one ten. (SBC: 016)

Crucially, the alternatives cannot contribute to different contextual effects (as they do in example 1). For example, it is the
fact that Brad’s addressee in 24(a) is interested in a decent copy of the original tape that is relevant for what tape recorder
Brad is going to suggest to sell her. Whatever objective differences between want and desire there may be (the latter
stands for a stronger feeling), they have no bearing on Brad’s point. This is the hallmark of Equivalence or constructions.
As argued above, what I count as a communicated (PII) alternative is the most relevant pragmatically enriched
interpretation, rather than the bare linguistic meaning. This is why the following counts as an Equivalence or case:

26. GRANT: . . . when is the city going to present,

say for instance,
. . . its overall plan,
for conducting the city.
So everyone can see,
.. their seat,
.. in the auditorium,
.. or their.. place around the table,
. . . and how long will they have to wait for their turn for development, (SBC: 026).

Their seat in the auditorium and their place around the table are not at all semantic paraphrases of each other the way
want and desire and goes out with other men and dates other men are. But both are used metaphorically to convey the
same concept here, namely, ‘where they stand in terms of the city priorities for development’. Presumably, having ‘‘a good
seat in the auditorium’’ just like having ‘‘a central place around the table’’ means that they’re in ‘good standing’, namely,
they will not have to wait long for the development relevant to them to take place. So, it is the contextually developed
interpretation that is relevant for Equivalence or constructions.

21
Incidentally, not every ‘or’ expression can be mobilized for a metalinguistic use. Italian o, for example can be used metalinguistically, but not
oppure, while ovvero can only be used metalinguistically (Haspelmath, 2007).
 M. Ariel / Journal of Pragmatics 103 (2016) 1--14 11

In the examples in (27) it’s not even the case that the two alternatives in each example are similarly adjusted into the
contextually relevant explicature. While only a single referent is intended by the speakers, it is differently construed in each
of the disjuncts:

27. a. GRANT: . . . let’s say optimistically,

.. there are some good ideas.
. . . some oversight,
or something that can be corrected,
. . . hopefully that’s . . . of minor proportion. (SBC: 026).
b. FOSTER: (H) he is going to.. detest (H) anything.. in theology,
(H) which smacks of (H) human effort,
. . . or human work,
. . . sufficient . . . to please God, (SBC: 025)

These alternative ways of referring to one and the same referent might make it easier for the addressees to get the
speaker’s message. The same is true for the following example, where again it’s not the case that the two disjuncts merely
use different names for one and the same phenomenon:

28. BEN: . . . Six point six million tons of concrete.

. . . That’s enough concrete to build a two-lane highway (H) from New
York City to San Francisco.
. . . Or,
put another way,
(H) a four-foot wide sidewalk from the North Pole to the South Pole. (SBC: 038).

Still, the two potential referents are exchangeable (note the put another way), because they make an identical discourse
point (the highway and the sidewalk each vividly illustrate what ‘six point six million tons of concrete’ amounts to).
Now, why do speakers use Equivalence or constructions? Why refer to a single alternative in more than one way or
otherwise be redundant? Why violate Hurford’s constraint? Different motivations lie behind Equivalence ors, two of which
are attested in SBC. The more frequent function is exemplified in 24(b), where the speaker wants to use a technical term
which he assumes to be unfamiliar to his addressees. Speakers explain fancy (e.g., auburn) or technical (cures,
penstocks) terms in 24/32 (75%) of the Equivalence or cases. In some of these cases (5, 20.8%) the speakers even
explicitly state that this is what they’re doing, as in 24(b) (‘‘as we call it’’). Another motivation, not attested in SBC, is that the
speaker is not sure which of the alternative names is (more) appropriate, and hence uses both to make sure she ‘‘covers
all her bases’’:

29. a. There is a place to buy soda or pop as we say in Midwest.

www.tripadvisor.com . . . Kalaupapa National Historical Park
b. Obviously, soda or pop is bad.
www.scarymommy.com/message-board/index.php?p=/

A minority of the cases (7, 21.9%) are used as in 24(a), where the second disjunct seems quite redundant. These
rhetorical repetition cases nicely exemplify Jakobson’s (1960) choice to define the poetic function in language by
reference to a sequential combination of a set of equivalents (see the quote above). Such, ‘‘evaluative’’ Equivalence or
constructions (to borrow a concept from Labov, 1972) highlight the utterance or parts of it (see especially 26 and 28), but
they may serve a much more mundane function of holding the floor for speakers to gain time (24c, perhaps).
Note that the order of the disjuncts in Equivalence or constructions is virtually always reversible.22 So, instead of
hardens or cures (24b), Ben could have said cures or hardens. Indeed, unlike Ben, Brad starts with the technical term
(dubbing or copying, 24d). This is due to an important aspect of Equivalence or constructions, namely, that the speaker
attributes equal prominence and relevance to the listed alternatives. Not only is the meaning of e.g., hardens synonymous
with that of cures in (24b), the two linguistic variants are also discoursally equivalent and substitutable for each other as far
as the speaker is concerned. Blakemore (2007), however, who seems to only address the majority, technical-term
Equivalence cases, treats them as parentheticals. Does that mean that the or-prefaced disjunct is discoursally
nondominant? First, note that in the spoken SBC, parenthetical status is true for only 9/32 of the cases (28%), (and for

22
Contra current assumptions, reversibility is only true of a third of nonEquivalence or constructions.
12 M. Ariel / Journal of Pragmatics 103 (2016) 1--14

36% of the type of cases analyzed by Blakemore). While the syntax of the initial utterance is indeed ‘‘disrupted’’ by an extra
disjunct, the prosody tells quite a different story in SBC. Judging by the Intonation Unit structuring, the subsequent disjunct
is no less part of the ‘‘hosting’’ utterance than the first disjunct is in the majority of the cases (e.g., 24a,c,d).23

4.2. Equivalence or constructions and Hurford rescues

Now that we have familiarized ourselves with Equivalence or constructions, we need to tackle a problem raised by the
very solution I offered for the Distinctness constraint in the previous sections. Recall that rescue interpretations impose a
distinction between the disjuncts, so that we can view the or construction as offering distinct alternatives. What’s to
prevent such an interpretative procedure for Equivalence or constructions? After all, there virtually always is some
difference between synonyms. So, why aren’t interlocutors imposing some distinctness on the disjuncts in order to comply
with Hurford’s constraint? Because referential distinctness is not equally important for different or constructions.
Consider the synonymous pair stronger or more powerful. Do such disjuncts constitute a single alternative (an
Equivalence or construction), or should we contextually adjust the two terms differently (strong is associated more with
physical power, powerful more with abstract power)?:

30. a. The word ‘‘stronger’’ or more powerful, does not mean giving a similar drug in quantity but in quality
(books.google.com/books?isbn = 078731076X)
b. to do the right thing -- not based on who’s right, older, bigger, stronger, or more powerful, but based on
what’s right. (books.google.com/books?isbn=1599550458)
c. Is Percocet a stronger, or more powerful pain killer than Tramacet (www.medschat.com/Discuss/
Percocet-vs-Tramacet)
d. What is stronger or more powerful, the bishop or the rook . . .
www.answerbag.com Questions Games Chess More Chess
Does the bengal tiger have more powerful or stronger forearms/forelimb than african lion?
(answers.yahoo.com . . . Science & Mathematics Zoology)

30(a) is quite obviously an Equivalence case where stronger and more powerful are not construed as distinct (note the
singular word). (b), however, treats these same attributes as distinct alternatives, perhaps along the lines mentioned
above. (c), (d) and (e) are less clear. They can certainly be taken as Equivalence cases (especially c, with the single a
determiner), but they may also be analyzed as semi-repair cases (Ariel, 2015a; Lerner and Kitzinger, 2015). Note that pills
and bishops and rooks are not physically strong, although they have power. The Bengal tiger forearms are perhaps more
appropriately described by the more physical strong. Repair or constructions present the ‘‘winning’’ alternative in
subsequent position, so the ordering of stronger and more powerful here may be significant.24 This is why we need
to take into account the speaker’s intention in the specific context. Stronger and more powerful do not denote different
concepts in (a), but they definitely do in (b), where they stand for different causes for doing the right thing. (c-e) seem to me
to be intermediate cases, which is what would be expected under an assumption that Repair or constructions have
evolved out of Equivalence or constructions. In the intermediate cases the two terms are equivalent, but asymmetric,
because the second one is preferred.
In sum, while the discourse motivations behind different Equivalence or constructions vary, all such cases share the
‘equivalence’ aspect, namely, that the two alternatives (overwhelmingly) denote a single referent, that they are equally
profiled, that they are mutually substitutable, and that any one of them can be omitted without affecting the speaker’s basic
meaning. Most importantly, while they need not be semantically synonymous, they must not contribute toward distinct
contextual implications. There are various cues one can use in order to identify an or construction as an Equivalence one
in addition to synonymy: syntactic noncomplexity, nonparaphrasability of the or with either. . . or (Ball, 1986), intonation
structure, omissibility of any one of the disjuncts and lexical rarity of one of the disjuncts.25 But the ultimate criterion is the
reading associated with the disjuncts. It is the contribution toward a single discourse point that ensures that the
alternatives are construed as referentially identical. Clearly, then, Equivalence or constructions violate the ‘Distinctness’
constraint in a very deep way: They don’t introduce multiple alternatives which are referentially distinct, they don’t

23
In fact, in 1 case it is the first disjunct that occurs in a separate IU, while the second occurs with part of the predication in the same IU.
24
Such a claim requires an extensive corpus examination. There were very many more stronger or more powerful (535,000) than more powerful
or stronger (26,100) Google hits (search 6.11.2014). But this ordering is also motivated by the ‘short before long’ principle.
25
Note that while the majority of or constructions conjoin syntactic phrases rather than clauses, for Equivalence or construction a [phrase or
phrase] structure is remarkably more frequent (over 90%) than for the rest (over 70%). It should be interesting to compare the amount of branching
nodes for the alternatives in Equivalence versus other or constructions. I expect there to be less of it for Equivalence cases.
 M. Ariel / Journal of Pragmatics 103 (2016) 1--14 13

contribute to different contextual implications. And to achieve that, they must resist potentially available ‘‘rescues’’, which,
we have argued, are absolutely necessary for many nonEquivalence cases.

5. Conclusions and future research

The main point of this paper has been the crucial role of contextually relevant inferences in ensuring the Distinctness
between or alternatives. Purely semantic nonentailment or incompatibility constraints (Hurford, 1974; Simons, 2001;
Singh, 2008) were shown to be too strong. At the same time, pragmatic enrichments are not freely available to ‘‘rescue’’
just any or construction. Rather, the pragmatic processes involved must be Sperber and Wilson-Relevant in that they are
not too effortful on the one hand (recall the primula experiment), and in that they must contribute to a contextually relevant
Privileged Interactional Interpretation on the other hand. But the story is even more complicated than that. We have seen
that in a specialized subset of or constructions, Equivalence cases, contextual narrowing to ensure Distinctness must
actually be avoided. This decision too is pragmatic in nature.
Moreover, thoroughly pragmaticizing Hurford’s constraint into a contextually relevant Distinctness constraint
applicable to all but Equivalence or constructions is not enough either. Current analyses of or constructions are much too
focused on what is taken to be their prevalent, exclusive use, which seems to require a maximal degree of Distinctness.
The or constructions discussed in Sections 1--3 were carefully selected in that all require referential Distinctness. But
Distinctness is not a matter of either or. An examination of the discourse profiles of 12 different readings associated with or
constructions makes it clear that they in fact require different degrees and different types of Distinctnesses (Ariel, 2015a).
A family of Distinctness types is needed instead. While stereotypical or constructions meet all Distinctness criteria, many
others only meet some of them (or even none of them), which is why we need more than one Distinctness constraint. As
we have seen here, Equivalence or constructions only meet the Formal Distinctnesses criterion (the alternatives contain
distinct linguistic expressions) and the Syntactic Distinctness criterion (they must comprise of multiple syntactic units).
And the or construction in (3) does not even exhibit Formal Distinctness. The final conclusion must then be that
Distinctness is differently applied to different or constructions. And even when pragmatic rescues are available in
principle, their application is not guaranteed. It is the discourse function of the specific or construction in the specific
context that determines whether a Hurford rescue is or is not intended by the speaker.

References

Ariel, Mira, 2002. Privileged interactional interpretations. J. Pragmat. 34, 1003--1044.

Ariel, Mira, 2004. Most. Language 80, 658--706.
Ariel, Mira, 2015a. Distinctnesses: The Difference an or Construction Makes Tel Aviv University (unpublished manuscript).
Ariel, Mira, 2015b. Doubling Up: Two Upper Bounds for Scalars. Linguistics 53.
Ariel, Mira, Mauri, Caterina, 2015. Why Use or?Tel Aviv University, University of Pavia (unpublished manuscript).
Ball, Catherine, 1986. Metalinguistic disjunction. Penn Rev. Linguist. 10, 1--15.
Blakemore, Diane, 2007. ‘Or’-parentheticals, ‘that is’-parentheticals and the pragmatics of reformulation. J. Linguist. 43, 311--339.
Burton-Roberts, Noel, 1994. Apposition. In: Asher, R.E., Simpson, J.M.Y. (Eds.), Encyclopeia of Language and Linguistics, vol. I. Pergamon,
Oxford, pp. 184--187.
Carston, Robyn, 1990. Quantity Maxims and Generalised Implicature, vol.2. UCL Working Papers in Linguisticspp. 1--31 , (reprinted in Lingua
96:213--244).
Carston, Robyn, 2002. Thoughts and Utterances: The Pragmatics of Explicit Communication. Blackwell, Oxford.
Chierchia, Gennaro, Fox, Danny, Spector, Benjamin, 2009. Hurford’s constraint and the theory of scalar implicatures: Evidence for embedded
implicatures. In: Egré, Paul, Magri, Giorgio (Eds.), Presuppositions and Implicatures, Proceedings of the MIT-Paris Workshop, MIT Working
Papers in Linguistics 60.
Du Bois, John W., Chafe, Wallace L., Meyer, Charles, Thompson, Sandra A., Englebretson, Robert, Martey, Nii, 2000--2005. Santa Barbara
Corpus of Spoken American English, Parts 1--4. Linguistic Data Consortium, Philadelphia.
Gazdar, Gerald, 1979. Pragmatics: Implicature, Presupposition, and Logical Form. Academic Press, New York.
Haspelmath, Martin, 2007. Coordination. In: Shopen, Timothy (Ed.), Language Typology and Syntactic Description, vol. II: Complex Construc-
tions. 2nd ed. Cambridge University Press, Cambridge, pp. 1--51.
Hirschberg, Julia Bell, 1991. A Theory of Scalar Implicature. Garland, New York.
Horn, Laurence R., 1985. Metalinguistic negation and pragmatic ambiguity. Language 61, 121--174.
Horn, Laurence R., 1989. A Natural History of Negation. University of Chicago Press, Chicago.
Horn, Laurence R., 2006. The Border Wars: a neo-Gricean perspective. In: Turner, Ken, von Heusinger, Klaus (Eds.), Where Semantics Meets
Pragmatics. Elsevier, London, pp. 21--48.
Hurford, J., 1974. Exclusive or inclusive disjunction. Found. Lang. 11, 409--411.
Inhelder, Bärbel, Piaget, Jean, 1964. The Early Growth of Logic in the Child. Routledge.
Jakobson, Roman, 1960. Closing statement: linguistics and poetics. In: Sebeok, Thomas A. (Ed.), Style in Language. MIT and John Wiley & sons,
New York, pp. 350--377.
Jaszczolt, K.M., 2005. Default Semantics: Foundations of a Compositional Theory of Acts of Communication. Oxford University Press, Oxford.
Katzir, Roni, Singh, Raj, 2013. Hurford disjunctions: embedded exhaustification and structural economy. Sinn Bedeut. 18, 201--216.
14 M. Ariel / Journal of Pragmatics 103 (2016) 1--14

Labov, William, 1972. Language in the Inner City: Studies in the Black English Vernacular (Conduct and Communication 3). University of
Pennsylvania Press, Philadelphia.
Lerner, Gene Howard, Kitzinger, Celia, 2015. Or-Prefacing in the Organization of Self-Initiated Repair. Res. Lang. Soc. Interact. 48, 58--78.
Levinson, Stephen C., 2000. Presumptive Meanings: The Theory of Generalized Conversational Implicature. MIT Press, Cambridge, Mass.
Levy, Roger, Bergen, Leon, Goodman, Noah, 2014. Roses and flowers: an informativeness implicature in probabilistic pragmatics. In: Paper
Presented at SALT, p. 24.
OED Online.
Potts, Christopher, 2013. Conversational Implicature: Interacting with Grammar Stanford University (unpublished manuscript).
Searle, John R., 1980. The background of meaning. In: Searle, John R., Kiefer, Ferenc, Bierwisch, Manfred (Eds.), Speech Act Theory and
Pragmatics (Synthese Language Library 10). Reidel, Boston, pp. 221--232.
Simons, Mandy, 2001. Disjunction and alternativeness. Linguist. Philos. 24, 597--619.
Singh, Raj, 2008. On the interpretation of disjunction: asymmetric, incremental, and eager for inconsistency. Linguist. Philos. 31, 245--360.
Sperber, Dan, Wilson, Deirdre, 1986/1995. Relevance. Blackwell, Oxford.
Wilson, Deirdre, Sperber, Dan, 1998. Pragmatics and time. In: Carston, Robyn, Uchida, Seiji (Eds.), Relevance Theory: Applications and
Implications. John Benjamins, Amsterdam, pp. 1--22.

Mira Ariel is professor of Linguistics at Tel Aviv University. Her research adopts a cognitive approach to language, focussing on the semantics/
pragmatics interface.