Western University

From the SelectedWorks of Robert J. Stainton

2019

Quasi-Factives and Cognitive Efficiency

Axel Barcelo Aspeitia
Robert J. Stainton

Available at: https://works.bepress.com/robertstainton/143/

53

3 Quasi-Factives and Cognitive Efficiency

Axel Barceló and Robert J. Stainton

3.1 Introduction: Motivation and Game Plan

The term ‘quasi-factive’ is our own coinage, hence readers can be forgiven for
not knowing what our topic is. To indicate that right off, it’s best to start with
examples:
(3.1) It isn’t widely known that Obama has been arrested
(3.2) Please do not acknowledge at the press conference that
Obama has been arrested
Both very strongly suggest something which is false, namely that Obama has
been arrested. Yet the first sentence, a declarative, is strictly speaking true. And
the second, an imperative, expresses an order that is wholly appropriate given
the actual facts.
Such sentences are interesting for several reasons. They may seem to call
for an analysis in terms of ‘weak logico-semantic support’: patently, there is
less than full-on encoded entailment of the complement in (3.1), but maybe
something less strict is at work. They raise the familiar issue of how much
linguistic convention contributes to in-context meaning, versus how much is
owed to general-purpose cognitive and pragmatic factors. For instance, does
(3.1) have a compositionally fixed standing linguistic content such that it is
weakly logico-semantically implied that Obama has been arrested? Or is such
an alleged ‘reading’ really just a pragmatic effect of a univocal sentence? In
addition to these long-standing issues in philosophy of language and linguistic
pragmatics, we are motivated by something social and ethical. The (seeming)
mismatch between what such sentences mean in the language versus how they
are standardly used can give rise to what Mitch Green [p.c.] has labelled ‘verbal
sleight of hand’, wherein a hearer may be bamboozled into embracing a belief
without explicit semantic endorsement by the speaker. That is, such sentences
work a kind of magic. As in a magic trick – or, maybe better, a ‘con job’ – there
is a distracter, namely the complex embedding structure. There is an illusion
created, viz., that the speaker has logico-semantically committed himself to the
truth of a claim. Finally, and to anticipate, in creating the illusion, the trickster
53

9781108418638_pi-312.indd 53 03-Jan-19 11:04:30 AM

54

54 Axel Barceló and Robert J. Stainton

exploits a natural psychological tendency in humans:  their perfectly appro-

priate quest for cognitive efficiency.
So much for motivation. Our discussion will proceed as follows. Moving
beyond examples, we offer a fuller description of the phenomenon:  the salient
characteristics of quasi-factives, and various sub-kinds. Then a sketch of a
cognitive-pragmatic explanation of why we hear them as we do. We end with a
major concern.

3.2 Description
A first crucial step is to sort out terminology. Recognising that usage varies widely,
we will simply stipulate: a factive verb is one such that, in the right linguistic
context, the complement of that verb is analytically entailed by the truth of the
matrix sentence. ‘Know’ is the paradigm example, but other factive verbs include
‘inform’, ‘recall’, ‘recognise’, ‘regret’, etc. A factive sentence, on the other hand,
is one in which the linguistic context is right, so that the complement of the fac-
tive verb actually is analytically entailed. Thus (3.3) and (3.4) both contain factive
verbs and are factive sentences:
(3.3) Juanito knows that Ottawa is the capital of Canada
(3.4) Rifat informed her mother that Ottawa is the capital of Canada
Crucially, as we use the terminology, there can be non-factive sentences which
contain factive verbs. (3.5) is not a factive sentence, nor is (3.6):
(3.5) Juanito knows whether Regina is the capital of Canada
(3.6) If Rifat purchases a house and a condo, she will regret
buying a condo
Quasi-factives fall within this genus of cases. Our species is especially
interesting because it is hard not to hear our sentences as factive.1 Said out of
the blue, (3.1) and (3.2) are understood as very strongly suggesting the truth of
the complement. Indeed, the hearer might easily report the speaker of (3.1) and
(3.2) as having ‘said’ that Obama has been arrested; relatedly, the truth of the
complement may naturally be heard as ‘main point’.2 All of this presumably

1
Does the term ‘quasi-factive’ apply to a kind of sentence type or to (potential) utterances of a
certain sort? Are quasi-factives, like factive sentences, united by a properly logico-semantic set
of features, or are they a ‘pragmatic kind’? In this introductory description, we leave such issues
open, so as not to beg too many questions.
2
In the earliest discussion of the facts that we are aware of, Hooper (1975: 97ff.) recognises,
albeit using different terminology, that the complement of ‘I’m not surprised that the dog bit the
mailman’ may easily be heard not only as ‘said’ but as ‘main point’ – with ‘I’m not surprised
that’ being understood as merely parenthetical. For additional early discussion, see Green
(1976). Simons (2007) addresses in compelling detail similar sorts of phenomena.

9781108418638_pi-312.indd 54 03-Jan-19 11:04:30 AM

55

3 Quasi-Factives and Cognitive Efficiency 55

has something to do with the fact that suggesting the truth of the comple-
ment is part of the usual usage of (3.1) and (3.2) – as contrasted with one-off
particularised conversational implicatures. A final and seminal feature of our
quasi-factives is this: if one stops and pays attention, one notices that (3.1) and
(3.2) are not factive sentences in our sense. More than that, in the case of (3.1),
the falsehood of the complement entails the truth of the whole. Thus, though it
sounds humorous, there is no contradiction in saying:
(3.7) It isn’t widely known that Obama has been arrested, because
he hasn’t been arrested!
(3.8) Please do not acknowledge at the press conference that Obama
has been arrested, because he hasn’t been arrested!3
There are numerous sub-varieties of our quasi-factives (Barceló & Stainton
2016). The factive verbs may involve propositional attitudes or discourse. The
propositional attitude verbs, in their turn, divide into cognitive/perceptual ones
(such as ‘forget’, ‘notice’, ‘know’, ‘recall’, ‘recognise’, ‘realise’ and ‘see’) and
emotional/evaluative ones (e.g., ‘amuse’, ‘be proud’, ‘bother’, ‘care’, ‘excite’
and ‘find it odd/surprising/terrific/tragic/unacceptable/wonderful’). Examples
of the discourse sub-variety, beyond ‘acknowledge’, include ‘admit’, ‘confide’,
‘confirm’ and ‘reveal’. (Some speakers hear ‘confess’, ‘note’ and ‘report’ as
discourse-reporting factive verbs too.)
There are also various syntactic configurations which yield quasi-factive
sentences. The verbs can take clausal complements – including finite ones (as
in our original examples), but also infinitival clauses and small clauses:
(3.9) I1 can’t recall [S e1 being shot]
(3.10) I1 don’t regret [S e1 leaving]
(3.11) No one saw [SC him nude]
And the complements don’t have to be clausal:
(3.12) I didn’t recognise [NP Clarice]
(3.13) Have you been informed about [NP Carson’s forced resignation]?
(3.14) I am not pleased about [NP the recent imprisonment of the Governor]
Also, the cancelling negative environment need not be a plain-old negative
particle, as in (3.1) and (3.2). Equally effective is a negative lexical entry or a
negating quantifier phrase:

3
See Carston (1998a) for a more developed argument supporting the claim that sentences like
(7) are not contradictory.

9781108418638_pi-312.indd 55 03-Jan-19 11:04:30 AM

56

56 Axel Barceló and Robert J. Stainton

(3.15) It’s seldom reported that Queen Elizabeth was born in Germany

(3.16) No one noticed the recent imprisonment of the Governor

(3.17) I refuse to recognise that it doesn’t bother you that Clarice is in jail
More controversial as negating contexts are non-declarative moods as in (3.13)
above and (3.18) and (3.19) below, and the antecedent of a hypothetical condi-
tional, as in (3.20) and (3.21):4
(3.18) Is it widely known that Obama has been arrested?
(3.19) Did her husband express surprise that Clarice has been arrested?
(3.20) If Chang realises that Bob has been arrested, he will cry.
(3.21) I would be remiss if I neglected to thank Daniel for his feedback.

3.3 Our Cognitive-Pragmatic Explanation

We call our examples ‘quasi-factives’ precisely because they include factive
verbs and are heard, in the usual case, as requiring (even as part of ‘what is
said’ or the ‘main point’) the truth of the complement; yet, on the other hand,
because of the negative environment, careful reflection establishes that there is
no such entailment, and that cancellation is possible. These are sentences, then,
which are not genuinely factive, but where some sort of ‘quasi’ is in play. We
turn now to an explanation of how they work this magic. That explanation has
a logical bit and a psychological bit.
With factive verbs, as a matter of analyticity, there is always the conjunctive
claim that the agent stands in at least one relation to a proposition, and that said
proposition (however expressed syntactically) is true.5 For instance, and sim-
plifying, ‘Rob knows that Obama is in jail’ entails the conjunction {[(Rob
believes that Obama is in jail) & (Rob has good justification for believing that
Obama is in jail) & (Rob’s justification and belief about Obama being in jail are
connected in the right way) & …] & (Obama is in jail)}. As a result, a negative
embedding environment induces the negation of an (implicitly) conjunctive

4
The generalisation seems to be that negative contexts, in the sense at issue, are those which
license negative polarity items (NPIs). We won’t explore that observation here. As an interesting
aside, notice that quasi-factives can embed within each other, as in (3.17). The effect of the
complement seeming-to-be-said is retained even in such cases.
5
We set aside in this brief paper a complication noted by Karttunen (1971: 56ff), namely that
the complement of the factive verb may consist of an open sentence, one bound by a quantifier
higher up in the matrix sentence. His example is [Some1 senators regret that [they1 voted for
the SST]]. In such cases, it isn’t strictly the case that the complement considered in isolation is
guaranteed to be true by the factive nature of the verb.

9781108418638_pi-312.indd 56 03-Jan-19 11:04:30 AM

57

3 Quasi-Factives and Cognitive Efficiency 57

claim. Let’s put things schematically, and then work through an example. A fac-
tive sentence of the form AΦ(p) entails {[AΦ1(p) & AΦ2(p) & AΦ3(p) & …] & p}.
Hence ~AΦ(p), a schema for our quasi-factives, entails ~{[AΦ1(p) & AΦ2(p) &
AΦ3(p) & …] & p}. Now, the negation of a conjunction is logically equivalent
to a disjunction of negations: De Morgan’s law states that ~(q & r) is equiva-
lent to (~q v ~r). Applying this, ~{[AΦ1(p) & AΦ2(p) & AΦ3(p) & …] & p}
is equivalent to {~[(AΦ1(p) & AΦ2(p) & AΦ3(p) & …] v ~p}. Applying the
law a second time, to the first complex disjunct, we get as the entailment of a
quasi-factive:
(3.22) [~AΦ1(p) v ~AΦ2(p) v ~AΦ3(p) v …] v ~p
Applying this abstract machinery to an example, (3.23) below entails the
negation of a conjunctive claim:  ~{[(Rob believes that Obama is in jail) &
(Rob has good justification for believing that Obama is in jail) & (Rob’s jus-
tification and belief about Obama being in jail are connected in the right way)
& …] & (Obama is in jail)}.
(3.23) Rob doesn’t know that Obama is in jail
This is equivalent to {~[(Rob believes that Obama is in jail) & (Rob has good
justification for believing that Obama is in jail) & (Rob’s justification and belief
about Obama being in jail are connected in the right way) & …] v ~(Obama
is in jail)}. And that, in turn, leaves us with two potential truth-makers for
(3.23): [~(Rob believes that Obama is in jail) v ~(Rob has good justification
for believing that Obama is in jail) v ~(Rob’s justification and belief about
Obama being in jail are connected in the right way) v …] on the one hand and
~(Obama is in jail) on the other.6
The psychological bit is even more complex. We will therefore lay it out in
two steps. Importantly, the two (major) negative disjuncts differ, and along two
axes. In the cases that work most effectively, the complement p is surprising.
This makes the negation, ~p, unsurprising:  that the Queen was not born in
Germany, e.g., is hardly newsworthy. In comparison, the negation of the rela-
tional bit, i.e., of the first (major) disjunct in (3.22), will be more noteworthy.
Suppose Elizabeth the Second was born in Germany; it would be surprising if
this curious and important fact were seldom reported. The first contrast between

6
A logico-metaphysical subtlety merits mentioning, one suggested by Soames’ (1977: 276) criti-
cism of Wilson (1975). Strictly speaking, there are many potential truth-makers for a sentence
like (3.23): e.g., ~[(Obama is in jail) v (Rob has good justification for believing that Obama is in
jail)] would render the sentence true. We hold, however, that such logical combinations are not
among the most psychologically accessible potential truth-makers – any more than ~[(Obama
is in jail) v (32=11)] is, though it too is sufficient for the truth of ‘Rob doesn’t know that Obama
is in jail’. Thus, it is the psychological aspect of our story, not the logical bit, that leaves only
two salient options.

9781108418638_pi-312.indd 57 03-Jan-19 11:04:30 AM

58

58 Axel Barceló and Robert J. Stainton

the two disjuncts, then, pertains to the amount of ‘cognitive pay-off’: there is

little of it, if the complement p is being negated; but there is rather more if the
relation-stating disjunct [AΦ1(p) & AΦ2(p) & AΦ3(p) & …] is denied. A second
difference is this. A quasi-factive ~AΦ(p) is not a cost-effective choice in terms
of conveying that the complement is false, since there is a far less cognitively
taxing option:  namely, saying ~p. In contrast, there isn’t a readily available
alternative for phrasing [~AΦ1(p) v ~AΦ2(p) v ~AΦ3(p) v …].
Returning to (3.23), that Obama is not in jail is unsurprising. Besides, there
is a much more straightforward way of getting it across: saying ‘Obama is not
in jail’. In contrast, the relational bit is surprising: it would top all the headlines
around the world if Obama were in jail, and thus Rob would be expected to
know. And, as Gettier cases teach us, there is no straightforward way of getting
across just the relational bit [~(Rob believes that Obama is in jail) v ~(Rob has
good justification for believing that Obama is in jail) v ~(Rob’s justification
and belief about Obama being in jail are connected in the right way) v …].
(Hence those final ellipses.) The upshot:  it would be twice-over more cog-
nitively efficient for the speaker to intend the negation-of-the-relations truth-
maker. So, for pragmatic reasons, the hearer ignores the not-in-jail disjunct.7
A second step in our psychological story is required because our explan-
andum is that the complement p is heard as ‘said’ in quasi-factives. But the
foregoing only explains why ~p is not conveyed. These are patently not the
same:  e.g., if someone utters ‘It’s raining’, they will almost certainly not
convey that the Queen was not born in Germany – yet they won’t be heard as
saying that she was.
We thus need to revisit a simplification. As the formulation in (3.22)
actually makes clear, what we have sometimes referred to as the relational
disjunct is actually disjunctive content-wise. Hence, as far as the logico-
semantic content of a quasi-factive sentence goes, any of ~AΦ1(p), ~AΦ2(p),
~AΦ3(p), etc., could be intended. But none is very psychologically access-
ible, especially in the presence of a very manifest concept, namely ~Φ(p). It
is altogether natural, given that both negation and Φ are maximally salient,
for a hearer of a quasi-factive sentence ~AΦ(p) to therefore construe ‘the
relational disjunct’ as A~Φ(p). Finally, [~AΦ1(p) v ~AΦ2(p) v ~AΦ3(p) v …]
is not factive, nor are any of its disjuncts. However, given that the ‘relational
verb’ Φ is factive, so is ~Φ. Hence the sentence which the hearer ‘leaps to’
as more psychologically accessible, A~Φ(p), is factive too. Thus, in terms
of logico-semantic relations, ~AΦ(p) does not entail A~Φ(p); nor is it even
weakly logico-semantically supported. Nonetheless, the move from the

7
In effect, this is the inverse phenomenon of what Grice (1981) called ‘conversational
tailoring’:  instead of ignoring the implicit conjuncts that are already part of the common
ground, here we ignore the implicit disjuncts.

9781108418638_pi-312.indd 58 03-Jan-19 11:04:30 AM

59

3 Quasi-Factives and Cognitive Efficiency 59

former to the latter is explicable because of our natural inclination towards

cognitive efficiency.
Let us continue with (3.23) as our worked example. Psychological step one
explains why the proposition THAT OBAMA IS NOT IN JAIL gets ignored.
But how does OBAMA IS IN JAIL get to be ‘said’? ‘Know’ is an extremely
complex verb. The quasi-factive sentence (3.23) will thus provide an extremely
complex relational disjunct: [~(Rob believes that Obama is in jail) v ~(Rob has
good justification for believing that Obama is in jail) v ~(Rob’s justification
and belief about Obama being in jail are connected in the right way) v …].
Any negated disjunct (and, as per footnote 6, any subset of them), is a poten-
tial truth-maker for our sentence. And none is factive. However, these are not
nearly as cognitively salient as the proposition ROB FAILS-TO-KNOW THAT
OBAMA IS IN JAIL, because the verb ‘know’ has actually been spoken, as
has a negating particle. Hence the most cognitively salient content, barring
special contexts, will be this – and it may, in the right circumstances, be heard
as ‘said’.8 Finally, ‘fail-to-know’ functions like a factive verb. Hence this
recovered-content entails that Obama is in jail. And content which is entailed
by the ‘said’ is often itself heard as ‘said’, and even as ‘main point’. Voilà.
That is our positive explanation. We haven’t space here to address alter-
native accounts, except to mention one competing syntactic approach. It is
crucial to distinguish our position from it. We have no objection to the obser-
vation that quasi-factives are reminiscent of scope-of-negation ambiguities. It
is perfectly appropriate, e.g., to describe the effect of (3.1) in terms of hearing
it as expressing the proposition OBAMA HAS BEEN ARRESTED, BUT
THAT ISN’T WIDELY KNOWN (or more exactly, albeit much less colloqui-
ally, as IT IS WIDELY THE CASE THAT PEOPLE FAIL-TO-KNOW THAT
OBAMA HAS BEEN ARRESTED), this contrasting with the wide-scope
negation of the whole proposition, as in IT IS NOT THE CASE THAT IT IS
WIDELY KNOWN THAT OBAMA HAS BEEN ARRESTED. It is a mistake,
however, to move from this to the idea that one should explain the ‘ambiguity
of quasi-factives’ in the same way one explains, e.g., the contrasting readings
of ‘Rifat doesn’t like cats and dogs’. In particular, we reject the postulation of

8
Speaking of ‘special contexts’, it may seem problematic for our account that there is a kind
of case where the ‘quasi-factive effect’ does not show up. Karttunen (1971: 63) noted them in
passing: he pointed out that ‘John didn’t regret that he had not told the truth’ need not inevitably
suggest the complement’s veracity, e.g., when the issue of whether John regretted not telling
the truth has been explicitly raised. In fact, far from posing an objection, we think such cases
provide additional support for our account. It explains these exceptions quite nicely. If someone
out-and-out asks ‘Does Rob know that Obama is in jail?’ then answering with (3.23) need not
be heard as quasi-factive because the mental representation /rob knows that obama is in jail/ is
already activated; the necessary cognitive processing has already taken place. So, a correction
with the fully spelled-out sentence is no longer as unduly demanding, and the pay-off is higher
because a salient issue has been wholly and explicitly addressed.

9781108418638_pi-312.indd 59 03-Jan-19 11:04:30 AM

60

60 Axel Barceló and Robert J. Stainton

multiple trees underlying (3.1), such that the hearer is tricked by an equivoca-
tion between them. (For instance, some theorists might postulate an underlying
tree where the complement clause has been fronted, and another in which it
remains in situ; others might posit one tree which has [neg] being the sister of
the verb, and another which has [neg] as the sister to the matrix sentence. In
both cases, the former tree actually would be factive, while the latter would
not  – thence the verbal sleight of hand.) Our account agrees on how quasi-
factive sentences are heard (i.e., as suggestive of a scope-of-negation ambi-
guity), but we do not explain the phenomenon in any such way. To phrase
our proposal another way, the hearer accesses a narrow-negation content in
processing the utterance, and for cognitive-pragmatic reasons. (Compare: the
sentence [S That is not a bachelor] is surely not syntactically ambiguous among
[[neg][A male]], [[neg][A adult]], [[neg][A human]] and [[neg][A unmarried]].)

3.4 Connection to Deirdre Wilson’s Work

In this section, we underscore three points of contact with Deirdre Wilson’s
ground-breaking work. This should clarify still further our view, and it will
afford an opportunity to hark back to the initial triad of issues in philosophy
of language.
Wilson (1975) was a pioneer in her treatment of presupposition. She criticised
the then-dominant view that it involved some sort of weak logico-semantic
supporting relation. To explain, in (3.3) and (3.4) we encountered genuine entail-
ment:  both sentences logically require that Ottawa be the capital of Canada.
Similarly, the truth of (3.24) logico-semantically guarantees the truth of (3.26):
(3.24) Jane has stopped smoking
(3.25) If Jane has stopped smoking, Mauricio will be thrilled
(3.26) Jane has smoked at some time in the past
But something less potent seems to be present in the relationship between
(3.25) and (3.26):  the former seems to merely provide weak, defeasible,
logico-semantic support; there seems to exist merely an encoded suggestion
of the truth of (3.26). Such appearances notwithstanding, Wilson rejected this
whole picture: ‘presuppositional analysis has no place in semantics’ (Wilson
1975: xii). This isn’t to deny that there is some ‘suggesting’ going on in a use
of (3.25), but it is the speaker who suggests that Jane once smoked; and she
does so by means of Grice-style pragmatics. Here we find a first connection
with our treatment of quasi-factives. By definition, quasi-factive sentences do
not entail the complement. One might suppose, however, that they ‘weakly
logically support it’. With Wilson, we think to the contrary that a speaker of a
quasi-factive sentence will ‘suggest things’ – but these do not follow logically,

9781108418638_pi-312.indd 60 03-Jan-19 11:04:30 AM

61

3 Quasi-Factives and Cognitive Efficiency 61

not even ‘weakly’, from the sentence she uses. We would equally echo two of
her arguments for this approach. First, the class of quasi-factive constructions
is heterogeneous in terms of grammatical structure. Along with Wilson, we
think that’s because what they really have in common is something about
which of their potential truth-makers are pragmatically appropriate  – rather
than anything about shared syntactic form. Second, our mutual pragmatics-
oriented approach explains why the ‘support’ we find in quasi-factives is so
weak that it allows for cancellation:  it’s because there isn’t logico-semantic
support after all. Cancellation is just what one expects from a pragmatic phe-
nomenon: the hearer draws a non-monotonic inference based on available evi-
dence; the speaker then provides additional evidence, explicitly ruling out the
inference’s conclusion via cancellation; what had seemed the most plausible
hypothesis about speaker meaning thereby gets rejected.
Closely related is Wilson’s Grice-inspired position on the role of conventional
linguistic meaning in fixing what people naturally ‘hear’ as the meaning. In
many cases, people ‘hear’ as part of sentence-meaning what is actually supplied
pragmatically. Our view on quasi-factives dovetails with her stance. We do
not believe that a sentence like (3.1) has, as a matter of its standing linguistic
meaning, a ‘reading’ on which it (weakly) requires the truth of the proposition
OBAMA HAS BEEN ARRESTED. Syntax and compositional semantics do not
yield any such encoded ‘reading’. Instead, cognitive-pragmatic factors make it
such that people tend to interpret typical utterances this way. Connecting even
more directly, Wilson noticed forty years ago that (3.27), a paradigm example
of a quasi-factive sentence, does not logically entail (3.28):
(3.27) Mary doesn’t regret that her grandmother was trampled by
an antelope
(3.28) Mary’s grandmother was trampled by an antelope
She held that the first sentence is logically equivalent to:
(3.29) Either the grandmother was trampled and Mary doesn’t regret it, or
the grandmother wasn’t trampled.
Hence (3.27), she says, is simply true if no trampling occurred. This, of course,
represents the heart of our view.9

9
In fact, what Wilson is implicitly proposing here differs slightly. The way that she phrases
example (3.29) suggests that she takes the logical equivalent of the quasi-factive sentence
~AΦ(p) to be {[(p & (~AΦ1(p) v ~AΦ2(p) v ~AΦ3(p) v …)] v ~p}. This contrasts with our pre-
sent account because p appears as a conjunct within the first (major) disjunct. In this respect, the
view seems closer to a position we pursued in Barceló and Stainton (2016: 73ff.). An advantage
of this approach is that it explains directly why p is heard as said, once the second (major) dis-
junct has been set aside by the hearer on the grounds of cognitive inefficiency. There is then no
need for our ‘second psychological step’.

9781108418638_pi-312.indd 61 03-Jan-19 11:04:30 AM

62

62 Axel Barceló and Robert J. Stainton

A final point of contact pertains to Wilson’s later work. In order to keep

our ideas accessible to non-specialists, we haven’t spelled them out using the
rich resources of relevance theory proper. However, the psychological story
we tell about quasi-factives co-opts its notion of a cognitive cost–benefit
analysis. Specifically, the reason why one of the logical options is typically
ignored, according to us, has to do with (i) the comparative improvement of
the cognitive environment given the processing cost of the complex linguistic
stimulus and (ii)  the much higher manifestness of the ‘wrong’ factive rela-
tional verb (in bold), A~Φ(p), as opposed to the actually-entailed non-factive
[~AΦ1(p) v ~AΦ2(p) v ~AΦ3(p) v …].

3.5 A Major Challenge

Both our description of the phenomenon and our explanation of it are incom-
plete and rough and ready, and hence open to a range of objections. This is
inevitable in such a brief treatment. There is, regardless, one major challenge
that is too important to postpone till a future paper. It arises because of phe-
nomena which look very similar to quasi-factives in terms of surface effects,
but where our logico-psychological mechanisms seemingly don’t apply.
In the first class of cases, including (3.30)–(3.37) below, the structure and
elements of the sentences are similar to quasi-factives: there is a propositional
attitude or discourse verb, and some kind of negative environment. And the
sentence is typically heard as factive.
(3.30) Don’t tell anyone that Kathy has been arrested
(3.31) No one is denying that Rob is a serial killer
(3.32) Has it been announced that Mick Jagger was elected President
of the United States?
(3.33) Was Clarke ever questioned about Jenna’s murder?
(3.34) It’s very hard to believe that Obama is dead
(3.35) No one imagined that Pluto was inhabited
(3.36) Rogelio had no idea that Pluto was larger than the Earth
(3.37) Who would have suspected that Trump was born in Russia?
However, none of these sentences contains a factive verb: ‘announce’, ‘deny’,
‘question’ and ‘tell’ are discourse verbs, but not factive ones; ‘believe’, ‘have
an idea that’, ‘imagine’ and ‘suspect’ are propositional-attitude verbs, but not
factive ones. Indeed, some of these verbs suggest that their complement is
false: e.g., ‘deny that’, ‘have an idea that’ and ‘imagine that’.

9781108418638_pi-312.indd 62 03-Jan-19 11:04:30 AM

63

3 Quasi-Factives and Cognitive Efficiency 63

There is, then, seemingly no complex conjunction analytically entailed

in the corresponding positive sentences; hence the negative-environment
versions do not entail a pair of negative disjuncts, either of which would
render the whole negated sentence true (with one of these disjuncts being
ignored for psychological reasons). In Barceló and Stainton (2016), we
addressed this by suggesting, tentatively, that our logico-psychological
account could be extended to such cases. One might speculate, e.g., that
‘No one imagined’ in (3.35) is naturally heard as ‘No one even imagined’.
The latter, in its turn, (sort of?) entails that no one knew… And then our
story would apply. Or again, because of a principle of charity, or because
of human’s automatic trust in testimony, maybe ‘Don’t tell anyone that p’ is
heard as strongly suggesting that p… And then, once again, the negation-of-
a-conjunction mechanism applies. (So implicitly committed were we to this
strategy that we applied the label ‘quasi-factive’ to cases like (3.30)–(3.37)
when explaining the phenomenon.) At this point, we are even less sanguine
about assimilating such examples to our logico-psychological apparatus.
There are issues about the plausibility of the specific stories: e.g., how does
‘even’ get read into (3.35), and why is the negative (3.30) heard as making
the arrest more believable than the positive ‘Tell someone that Kathy was
arrested’? But even if such minutiae could be dealt with, there’s a larger
worry, namely that both (3.1)–(3.2)/(3.9)–(3.21)/(3.23)/(3.27) on the one
hand and (3.30)–(3.37) on the other seem to belong to a still larger family of
cases. And there’s simply no hope of applying our exact mechanism to that
overarching family. There is no negation in (3.38) and (3.39), and no dis-
course or propositional-attitude verb in (3.40). And yet, the ‘verbal sleight of
hand’ clearly arises.
(3.38) Axel hopes to return to space
(3.39) I’m delighted to say that Paul Simon has been elected President
of the United States!
(3.40) If Neil Young were healthy, he would be touring
Happily, we can at least gesture at two avenues of escape consistent with our
overarching linguistico-philosophical commitments. Possibly, there simply
is no genuine underlying kind at work hereabouts, in terms of mechanism –
not even one that underlies both (3.1)–(3.2)/(3.9)–(3.21)/(3.23)/(3.27) on the
one hand, i.e., genuine quasi-factives, and (3.30)–(3.37) on the other. There
are merely shared superficial effects. If this is correct, quasi-factives per se
would remain a proper explanatory kind, characterised by the specific logico-
psychological mechanisms we have proposed. (To give an analogy, just as
the existence of sea creatures which swim and feed in superficially similar
ways is no threat to cartilaginous fish being a unique and interesting kind, it

9781108418638_pi-312.indd 63 03-Jan-19 11:04:30 AM

64

64 Axel Barceló and Robert J. Stainton

is no threat to the reality and specialness of quasi-factives that there are other
similar means to ‘con’ hearers.) The other avenue, the one we presently incline
towards, is that all these cases are united by more than the superficial simi-
larities:  there is something shared at the level of psychological processing.
Specifically, the general cognitive efficiency story can be told in all cases – it’s
only the specifics of the logical bit which will vary. (Again, see Barceló &
Stainton 2016.)

3.6 Concluding Remarks

We conclude by reiterating what we have and have not done.
We described a sub-class of non-factive sentences containing factive verbs.
What’s peculiar about them is that they are, as it were, heard as factive – and
hence lend themselves to ‘verbal sleight of hand’. We then offered a posi-
tive account of how this surprising effect is achieved. Logically speaking, the
sentences analytically entail a pair of negative disjuncts – one about the relation
(e.g., Rob not believing, not having justification, etc.) and one about the prop-
ositional relatum (e.g., THAT OBAMA ISN’T IN JAIL). Psychologically, the
latter disjunct would violate a requirement of cognitive efficiency; hence some-
thing maximally salient and intimately connected conceptually to the former
is seized upon. But this latched-upon content (here, the content FAILS-TO-
KNOW) entails the truth of the embedded proposition: if Rob fails-to-know
p, then p is true. (Compare the English ‘misperceive’, the French ‘ignore’ and
the Spanish ‘desconoce’.) In many contextual circumstances, then, the com-
plement may be heard as ‘said’ and even as ‘main point’ – though as a matter
of linguistic convention it isn’t in fact entailed, nor is it even ‘weakly logico-
semantically supported’.
We haven’t attempted to explore alternative explanations, and we haven’t
filled in the details of our description and explanation (e.g., how the varied
negating contexts apply in the sub-cases). Nor have we offered anything like
a full response to the major concern about generality. Nonetheless, we hope
our discussion will contribute to progress on the three philosophical issues
mentioned at the outset.

Acknowledgements
Drafts of this paper were presented at:  the Mutual Belief in Pragmatics
Conference, Shan’xi University; Linguistic Talks at Western, in London,
Ontario; a philosophy of language discussion group at Universidad de
los Andes; and the Faculty of Linguistics, Philology and Phonetics at the
University of Oxford. We are grateful to the organisers and to all those pre-
sent. We must especially single out Santiago Amaya, Ash Asudeh, Tomás

9781108418638_pi-312.indd 64 03-Jan-19 11:04:30 AM

65

3 Quasi-Factives and Cognitive Efficiency 65

Barrero, Lesley Brown, Mitch Green, Mandy Simons and Elizabeth Allyn
Smith for very fruitful suggestions. Thanks also to Adam Sennet, for encour-
agement and tutorials on presupposition; to David Bourget, Robyn Carston,
Angela Mendelovici and Chris Viger for detailed comments on the penulti-
mate draft; and to the editors and an anonymous referee for helpful feedback.
Special thanks are owed to our dear friend and philosophical co-conspirator
Ángeles Eraña Lagos, to whom we dedicate our work on this project. The
research behind this paper was partially funded by DGAPA Project ‘Términos
numéricos e implicatura escalar’ IN401115 and by a grant from the Social
Sciences and Humanities Research Council of Canada.
Our many and varied debts to Deirdre Wilson, sine qua non, go without
saying; but they are implied throughout.

9781108418638_pi-312.indd 65 03-Jan-19 11:04:30 AM