Moderate vs. Radical Pragmatics

1
Moderate vs. Radical Pragmatics

• Anne Bezuidenhout
• Trondheim, September 18-22, 2006

2
Recanatis Continuum

• Proto-literalism
• Eternalism
• Conventionalism
• Minimalism
• Syncretism
• Quasi-contextualism
• Pragmatic composition
• Wrong-format view (WF)
• Meaning eliminativism (ME)

Where does Predelli fit in? Maybe hes a
conventionalist? Maybe nowhere?
3
Cappelen Lepores Syncretism

• There is a strict semantics-pragmatics divide.
• Every indexical-free sentence expresses a
  complete proposition.
  
• The truth conditions of such sentences can be
  given disquotationally e.g., John is ready iff
  John is ready.
  
• Semantics is pure, uncontaminated by pragmatics,
  except as regards the small (well-behaved) class
  of indexical expressions that belong to the Basic
  Set.
• To determine the referents of indexicals, we may
  have to rely on contextual knowledge, such as
  knowledge of speaker intentions.
  
• Full blown contextual knowledge is only relevant
  to understanding what is said/stated/asserted and
  other such speech acts.
  
• Understanding speech act content is a thoroughly
  pragmatic process, but for this reason it cannot
  be systematized.

4
The Mirror Opposite A Generic Contextualism

• There is pragmatic intrusion into
  truth-conditional content truth-conditional
  pragmatics.
  
• There are sentences that are semantically
  incomplete they express only propositional
  radicals.
  
• The disquotational schema does not give us real
  semantic knowledge, and if it appears to, this is
  because the object and meta-languages are both
  our own.
• There is a wider class of context-sensitive
  expressions than just those in CLs Basic Set.
  
• Besides the sort of bottom-up pragmatic
  processes needed to assign contextual values to
  indexicals, top-down pragmatic processes are
  needed too. (e.g., free enrichment, loosening,
  transfer).
• CL have collapsed the distinction between
  locutionary and illocutionary acts. Saying is a
  locutionary act.
  
• Pragmatic processes are not unsystematic.

5
Who Subscribes to GC?
This doesnt describe any one persons view.
Many who deny they are contextualists subscribe
to some of the points on the list (e.g., Kent
Bach believes that some sentences are
semantically incomplete). One can be a contextual
ist and agree that all sentences-in-context
express complete propositions! (Recanati in some
moods, Predelli(?)) Some of these points dont re
quire one to be very radical. E.g., one could
widen the Basic Set only a little, and do this
only on principled grounds. E.g. Stanley (2000),
Taylor (2001), Sarah-Jane Leslie (2006).
One can make more or less use of the notion of
free enrichment. One can even make no use of free
enrichment and still be a contextualist! E.g.,
Corazza Dokic (2006). Points (3), (6) and (7) h
owever are ones that I insist upon and will
assume henceforth.
6
The Contextualist Strawman?
It certainly does seem that CL have set up a
Strawman as their opposition. They think they can
do this as they have an argument that so long as
one grants even the tiniest part of the
contextualist story one is launched down a
slippery slope into a radical contextualist
hell. This is a place where no constraints hold a
t all, where even the meanings of words are
unstable from one moment to the next, where no
communication is possible, where each person is
trapped inside his own private, solipsistic
world, etc. One is allegedly launched down this s
lide because the sorts of context-shifting
arguments and arguments from incompleteness that
are appealed to by some contextualists can be
applied across the board to show that every
sentence is incomplete. To avoid such a hell, CL
suggest a series of three tests that can be
wielded to show that only expressions in the
Basic Set are context-sensitive.
7
Is There a Slippery Slope?
The quick answer is No. There are many
principled stopping places along the way
One can apply tests to determine whether there
are hidden argument slots/hidden variables.
Grammatical tests (Stanley) Semantic tests (Taylo
r) Pragmatic tests (Recanati) One interesting re
cent paper uses one of CLs own tests for
context-sensitivity to widen the class of
context-sensitive expressions! Leslie (2006). See
Leslies questionnaire. Besides, where did this
radical contextualist hell come from? There is
no contextualist I know of, even someone as
radical as Travis, who thinks there are no
meaning constraints whatsoever. (Even Recanatis
ME has constraints of past usage in ones
linguistic community).
(The slippery slope rhetoric used by CL is
reminiscent of the rhetoric used by Fodor
Lepore to argue against meaning holism)
8
Aims for Remainder of this Lecture
Will assume that there are principled stopping
places. Interested in defending the quasi-context
ualist (QC) and pragmatic composition (PC) views.
Reminder as to what these views entail
QC There is a stone lion in the courtyard
PC John heard the piano Will look at one sort
of strategy used by Recanati for deciding whether
we need to appeal to saturation or free
enrichment. End by raising some questions for fur
ther discussion The psychological role of minima
l propositions (Lecture 2) The coherence of pragm
atic (enriched) composition (Lecture 4)
9
Unarticulated constituents
Consider the following sentence
(1) It is raining Eons ago, Perry claimed tha
t, in an appropriate context, an utterance of (1)
might express the proposition that It is raining
in Paris. This proposition therefore has
unarticulated constituents. A similar view was
advocated by Recanati and by relevance theorists
such as Sperber Wilson and Carston, who argued
that the contextually determined location that a
hearer recovers when processing an utterance of
(1) is the result of free enrichment, where this
is understood to be an optional pragmatic process
(in contrast to such obligatory pragmatic
processes as saturation).
10
Three Waves of Opposition
Wave 1 The Standard View hidden variable/hidd
en argument slot strategy Recanati's response Th
e Weatherman Example Wave 2a Optional Variable
Strategy Recanatis response The Negative Weath
erman Example Wave 2b Broad Location Strategy
Recanati's response Hoisted by their own
petards! Wave 3 Weatherman-type cases are ubi
quitous Recanatis response Meaning shifts
11
The Standard View
Stanley (2000) makes use of examples such as
Wherever he goes, it rains to argue for an
implicit variable that can be bound by an overt
quantifier or supplied with a contextual value
when it occurs free, as in (1).
Taylor (2001) appeals to semantic considerations
about theta roles to argue for an implicit
location argument slot in (1).
Note that when these implicit variables/
arguments are not explicitly quantified over
their contextual values will be specific.
Suggests the following test If we can come up
with a context in which (1) is to be understood
simply as It is raining somewhere, then there is
no hidden variable slot. Overt variables (e.g., p
ronouns) dont have such indefinite/non-specific
interpretations. E.g., He is bald doesnt have
an interpretation according to which it means
that someone is bald. So covert ones shouldnt
either.
Why He collects local newspapers is not a
counterexample.
12
The Weatherman Example
imagine a situation in which rain has become
extremely rare and important, and rain detectors
have been disposed all over the territory
(whatever the territory possibly the whole
Earth). In the imagined scenario, each detector
triggers an alarm bell in the Monitoring Room
when it detects rain. There is a single bell the
location of the triggering detector is indicated
by a light on a board in the Monitoring Room.
After weeks of total drought, the bell eventually
rings in the Monitoring Room. Hearing it, the
weatherman on duty in the adjacent room shouts
Its raining! His utterance is true iff it is
raining (at the time of utterance) in some place
or other.
13
Lesson to be Learned
There is no location variable or argument slot in
the predicate rains. When a location is
recovered this involves free enrichment.
In the Weatherman case no free enrichment occurs
and the proposition expressed by (1) is simply
that It is raining (punkt), a proposition
Recanati (2006) represents as follows
(2) ?e ?t Present (t) ? Time(t,e) ? Raining(e)
Note that Recanati does not represent it thus
(3) ?e ?t ?l Present (t) ? Time(t,e) ?
Raining(e) ? Location(l,e) Claims we can appeal
to the metaphysical fact that events always take
place somewhere to infer (3) from (2).
Problems? No free enrichment? (Note also the irre
levance of rain on Titan, a point acknowledged by
Recanati).
14
Free Enrichment Cases
When (1) is uttered in a context in which the
place of the rain event is relevant, the location
will be recovered by free enrichment and become a
part of the full utterance content.
Such free enrichment is either a matter of adding
an event description under the scope of the event
quantifier or a matter of restricting the domain
of the event quantifier So when free enrichment
occurs, the basic LF in (2) results in one or
other of the following modified LFs
(4) ?e ?t Present (t) ? Time(t,e) ? Raining(e) ?
Location(Paris,e) (5) (?e Location(Paris,e))(?
t Present (t) ? Time(t,e) ? Raining(e)
Recanati wishes to remain agnostic as to which of
these accounts is to be preferred.
I will come back to a discussion of the notion of
a modified logical form.
15
2nd Wave Optional Variables
Answer to Weatherman case Even though overt vari
ables must be supplied with specific contextual
values when they occur free, covert variables are
special. They can be supplied with either a speci
fic or a non-specific value. The latter is what
happens in the Weatherman Example.
The covert variable undergoes existential
closure, yielding (3) above as the underlying
form.
16
Recanatis Response to OVS
Negative Weatherman Case Imagine a scenario in w
hich the absence of rain has become extremely
rare and important (it rains almost everywhere
and everytime). All over the territory detectors
have been disposed, which trigger an alarm bell
in the Monitoring Room when they detect the
absence of rain. There is a single bell the
location of the triggering detector is indicated
by a light on a board in the Monitoring Room.
After weeks of flood, the bell eventually rings
in the Monitoring Room. Hearing it, the
weatherman on duty in the adjacent room shouts
Its not raining! Here we do not get the wide
scope reading There is somewhere that it is not
raining. But this reading is predicted if there i
s an optional location variable available for
existential closure. To rule out the impossible r
eading, the OVS must say that the covert
existential always takes narrow scope with
respect to any other scope bearing elements in
the sentence. But this seems ad hoc (although the
re is more to be said here).
17
2nd Wave Broad Locations
Contrary to Recanatis claim, we do supply a
specific value for the implicit location variable
in the Weatherman case, namely on Earth.
Of course, we dont get the meaning everywhere on
Earth, but rather just somewhere on Earth.
So the specified location has to be understood as
a location in the broad sense.
Event e is located at place l in the broad sense
iff there is some sub-location l of l and e is
located at l in the narrow sense.
18
Recanatis Response to BLS
Those who posit hidden variables/argument slots
for rains (e.g., Taylor) want to contrast
rains with predicates like dance which
allegedly do not have location argument slots.
But this contrast is lost if we appeal to broad
locations, because one can equally say that
dance has a location slot but that it is
usually filled by reference to a broad location.
Defenders of BLS are likely to protest that they
do not have to analyze dance in this way.
But then neither do they have to analyze rains
in this way. The BLS backfires on the defenders o
f the strategy!
19
3rd Wave Ubiquity of Weatherman cases
Recanati has allegedly given us a test to
determine when we need to posit hidden variables/
argument slots and when we must appeal to free
enrichment. We test to see whether or not we can
generate a non-specific reading for a putative
argument position. If we can, we do not have an
implicit argument position. But with sufficient i
ngenuity we can always get such non-specific
readings, even for predicates where all parties
to the debate agree that there are implicit
argument slots, e.g., predicates such as
finish, arrive or notice
(6) John has finished reading the book.
(7) Mary has arrived at Heathrow airport.
(8) Bill noticed my new shoes/ that I had new
shoes.
20
Semi-Coma Example
Consider a scenario with a patient who has been
in a semi-coma, and a technician in another room
is reading the output of an EEG or whatever it is
that measures brain activity in various areas of
the brain. A trained technician could know when
brain activity signals noticing, and since for
the semi-coma patient, the fact that hes
noticing (something) is all thats important, one
might imagine the technician being able to shout
Hes noticing! without being in any position to
know or say what it is the patient is noticing.
(Cited in Recanati (2006), from an anonymous
referee)
21
Recanatis Response to Ubiquity Objection
These are not cases where free enrichment occurs.
Rather they are cases of pragmatically induced
meaning shifts. Recanati posits the notion of a v
ariadic function that can either decrease or
increase the adicity of a predicate. In the
former case we have a recessive function and in
the latter an expansive one. In some cases these
shifts are lexicalized (e.g., transitive and
intransitive readings of eat). In some cases
the shifts are pragmatically induced.
In the Semi-Coma example there is a pragmatically
induced recessive meaning shift, since one of the
lexically specified argument roles in the
predicate notice has been existentially
quantified.
22
Standard View Vindicated?
Having acknowledged the idea of recessive meaning
shifts, Recanati is now forced to agree that his
hidden argument slot opponents can appeal to
this notion too! Advocates of SV can say that ra
ins has a location argument slot, but in the
Weatherman case a pragmatically induced recessive
meaning shift occurs, and the denotation of
rains is shifted from the property in (9) to
the one in (10) (9) ?l ?e Raining(e) ? Location
(l,e) (10) ?e ?l Raining(e) ? Location(l,e)
23
Truth-Conditional Pragmatics
Recanati agrees that at present we have to admit
that free enrichment and recessive meaning shift
accounts of rains are equally able to account
for the data. We have no neutral way at present f
or deciding what the encoded meaning of rains
is. Recanati remains (relatively) sanguine about
this stalemate, since he thinks that either way
the idea of a truth-conditional pragmatics has
been vindicated. This is because both free enrich
ment and pragmatically induced recessive meaning
shifts are optional pragmatic processes, showing
that utterance content is influenced in a
top-down manner by pragmatic information.
24
Expansive Meaning Shifts
Having introduced this machinery of recessive and
expansive meaning shifts, Recanati notes that one
can appeal to this machinery in implementing the
free enrichment idea. Take the case of (1), utter
ed in a situation in which rain in Paris is
relevant. Then one can say that the location-less
predicate rains undergoes a pragmatically
induced expansive meaning shift.
The denotation of the predicate shifts from the
property in (11) to the one in (12), and at the
same time a value for the location argument slot
is supplied, yielding the interpretation It is
raining in Paris (11) ?e Raining(e) (12) ?l ?
e Raining(e) ? Location(l,e)
25
Advantages of Meaning-Shift Implementation
If we go for this implementation, we avoid
commitment to unarticulated constituents, which
have been taken to cause problems for a
compositional account of utterance content.
Meaning shifts are local pragmatic processes that
act on lexical items to yield pragmatically
shifted meanings. In the example on the previous
slide, we would get the modulated meaning
rains-in-Paris. This modulated meaning is what is
submitted to the compositional process, not the
lexicalized meaning. Problems? (Taken up in Lectu
re 4) (Note Different story when we go for the i
mplicit domain restriction version of free
enrichment. Have to allow unarticulated
constituents of utterance content, i.e., of what
Recanati calls the Austinian or global
proposition).
26
Modulated Meanings vs. Unarticulated Constituents
Are there any considerations that would drive us
one way or the other in particular cases?
Recanati admits that in the case of (1) it is a
toss-up. But he is adamant that in some cases
only the meaning shift explanation is plausible,
Consider (13) Ive eaten a full breakfast today
. Suppose (13) is uttered in response to an invi
tation to join someone for breakfast on the day
of speaking, and that the enriched proposition
recovered includes the content indicated by the
words in the braces. Recanati would say that an e
xpansive meaning shift has occurred, and that in
this context eat means eat a full breakfast.
What is the method that decides these cases? An
appeal to intuitions?
Ex. of lexicalized meaning shift to drink has
as one of its lexicalized meanings to drink
alcohol.
27
Context-Sensitivity without Free Enrichment?
Several people have recently argued that
context-sensitivity of the sort exhibited by (1)
can be accounted for without an appeal to free
enrichment. I will discuss two such suggestions
(i) The thought without representation idea
proposed by Corazza Dokic (2006)
(ii) The free generation of variables idea
proposed by Marti (2006)
28
Thought Without Representation
Suppose speaker and hearer are in Paris when the
speaker utters (1) intending to convey that it is
raining in Paris. The utterance situation s impli
citly restricts the location of the rain event to
the location of s. But since speaker and hearer
are in that situation, there is no need for them
to construct a mental representation of the
enriched content It is raining in Paris.
The environment that the speaker and hearer are
embedded in together with the minimal contents in
their heads support the enriched content.
Problem Expressions or embedded clauses may need
to be interpreted relative to situations other
than the utterance situation. So we cant always
get away with just minimal propositions.
29
Lekta vs. Global Propositions
Recanati (forthcoming) suggests a view very
similar to the one suggested by Corazza Dokic.
He distinguishes two levels of content
(A) the content of a sentence-in-context, which
he calls a lekton. (Lekta appear to be identical
to minimal propositions). (B) the content of an u
tterance, which he calls a global proposition.
(Such propositions are arrived at via free
enrichment of lekta). Now, if you say Its rainin
g in your situation (say in Paris) and I say
Its not raining in my situation (say in the
Bahamas) then even though our lekta express
contradictory contents, we are not disagreeing.
We only disagree if were talking about the same
situation. But Recanati also says (and here is wh
ere he seems to agree with CD) that when were
in the same situation, lekta can go proxy for
complete utterance contents and we can assert,
communicate, and disagree about these minimal
contents.
30
The Poker Game
If we can assert and disagree about lekta, they
must be truth evaluable. Recanati does indeed cla
im that we can evaluate the truth of both
sentences-in-context and utterances. However,
there is no guarantee theyll coincide in
truth-value Recanati cites an example from Barwi
se Etchemendy (1987) Jon is watching a poker g
ame and says Claire has a good hand now, which
is true iff Claire has a good hand in the poker
game that Jon is watching right then. But suppose
Jon made a mistake and Claire is not one of the
players in the game he is watching. However, she
is playing in a bridge game across town and in
that game she does indeed have a good hand.
On Recanatis dual content view, what Jon said
was false, since his utterance is evaluated in
the poker-game situation, and in that situation
Claire does not have a good hand. But the
sentence-in-context is true (in the actual
world), since Claire does have a good hand
(somewhere in the world) at the time of Jons
utterance.
31
Problems with Minimal Propositions
Strictly, Recanati should say that the
sentence-in-context is true because Claire has a
good hand of some sort somewhere in the world at
the time of Jons utterance (since it is virtue
of having a good bridge hand, not a good poker
hand, that she has a good hand).
In other words, it is not only the location which
is unspecified in the lekton, but also the type
of hand (and probably also the standard of
goodness by which the hand is judged if Claire
is playing against novices wholl make lots of
unforced errors, she probably doesnt need as
strong a hand as shed need in an international
bridge tournament). Either way, it is not clear t
o me that what is evaluated is the lekton, as
opposed to a general proposition that is derived
from the lekton by free enrichment of a sort. The
lekton that Claire has a good hand punkt at the
time of utterance is a propositional radical.
Does an appeal to metaphysics help here as it
allegedly did in the case of (1)? Is there a
metaphysics of good hands?
32
Free Generation of Variables
Marti (2006) argues against free enrichment.
Instead, she posits the idea that a variable slot
is optionally generated in the syntax.
When it is, it must be supplied with a contextual
value in the ordinary way. When it is not, we get
the sort of non-specific interpretation that we
do in the Weatherman case. But what determines wh
ether the slot is generated? It had better not be
pragmatic considerations, or the view will
collapse into the free enrichment view on its
expansive meaning shift implementation.
In fact, Marti denies that pragmatics triggers
anything. She writes Whether one of the B-variab
les is generated in the syntax or not is left
completely free, just because adjuncts generally
are not necessary. The system tries out different
derivations, and only those that comply with all
the principles of grammar, including Gricean
principles, are successful. (p. 150)
33
Unencapsulated Syntax?
It is not entirely clear what the view is, but
Ill assume that Marti is not saying that
syntactic processing is unencapsulated and
directly influenced by wide pragmatic
considerations of the Gricean sort.
Rather, she is claiming that the output of the
(encapsulated) syntactic system will be multiple
candidate LFs and then wide pragmatic processes
will operate to prune the list.
This would amount to some sort of ambiguity
theory. Sentences like (1) are at least two ways
ambiguous. Of course, it is an empirical question
how language processing works, but this does on
the face of it seems like an inefficient use of
processing resources. Why bother to generate
multiple LFs when a single underspecified LF will
do the trick? Why should we even consider the var
iable slot option in the Weatherman case just to
rule it out? See Sperber Wilson (1997), Carston
(2002) for discussion of the benefits of
underspecified LFs.
34
Modified Logical Forms
Earlier I noted that Recanati (2006) posits that
the basic LF (2) undergoes free enrichment to
yield the modified LF (4) i.e, if one goes for th
e adding-a-conjunct idea if one goes for the
domain restriction idea, the modified LF will be
(5). But does (4) have any sort of psychological
reality? Weve already seen that Corazza Dokic
think not only basic LFs need be represented.
Recanati (2006) is inclined to say that (4) is
simply the theoreticians way of representing the
intuitive truth-conditions of an utterance of (1)
in the Paris situation. It does not represent
another layer of syntactic representation.
35
LFs and the Language of Thought
Recanati attributes the view that modified LFs
are another layer of syntactic representation to
Sperber Wilson and to Jackendoff.
He does this on the grounds that these people all
subscribe to the view that representations in
thought are representations in the language of
thought (which according to Fodor has its own
syntax). But it is very uncharitable to take Sp
erber Wilson to be saying that representations
of enriched propositions are syntactic
representations in the same sense that basic LFs
are. Basic LFs are the output at the conceptual-i
ntentional interface of the syntactic system
proper. Representations of enriched propositions
on the other hand are constituted out of ad hoc
concepts, and are the output of local pragmatic
processes which take basic LFs as their input.
I will come back to a discussion of Jackendoffs
views in lecture 4.
36
Mental Architectures
Weve had hints of various different pictures of
the mental architecture underlying our pragmatic
competence. I personally am very interested in qu
estions about the psychological processing of
utterances and about the psychological
organization of the language system. I take up
some of these issues in more detail in lecture
2. Ill end with a few diagrams of the mental arc
hitecture of the language system that seem to be
suggested by (a) Corazza Dokic (b) Marti (c)
Sperber Wilson (Positions (a) and (c) can be t
aken to be expressions of the rival
representationalist and anti-representationalist
positions that have been debated since at least
the 1980s. See Kempson Meyer-Viol (2004) for a
description of this debate and reasons for
favoring the representationalist side).
37
Situationalism
38
Ambiguism
39
Pragmatic Modulation