Plug on Griceian SolipsismS

by J. L. Speranza
for the Grice Club.

In the study, "Autism as a disproof of Grice on Meaning", M. Plug (of King's College London) thinks Grice is wrong, or "Griceanly wrong", to be more correct.

After all, Plug argues, Gluer and Pagin (2003) have argued that the existence of a subset of speakers on the autistic spectrum, a group of speakers that have a
sophisticated level of language but have problems with attributing
beliefs and thoughts to other people and themselves, present a counterexample
to Grice's analysis (1) of "non-natural meaning" in terms of
higher order thoughts (e.g., beliefs about beliefs).2

However, Grice takes this type of meaning to be related to communication, and it has been argued (Reboul, 2006) that the communicative problems that this group of speakers have indicates that these speakers can't exemplify this particular type of meaning, and that therefore these utterers do not pose a counterexample to his analysis.

Plug argues that the way Reboul mounts her attack rests on a misunderstanding of
Gluer and Pagin's argument (and that it is therefore not successful).

Nevertheless, Plug believes that Gluer and Pagin do not make a su cient
case that the meaning of autistic speakers is a matter of nonnatural
meaning, and so their argument, as it stands, is left open to attacks
of a similar kind.

To ensure the claim that (a subset of) autistic
utterers are a counterexample to Grice's analysis an assessment is
required of the intuitions we have concerning the applicability of the

(1 See for example Grice (1989a)) (2. Gluer and Pagin (2003, pp. 25f) note that Stephen Laurence made this suggestion before they did, albeit without working it out at length, in his Laurence (1998, pp. 209f))

term "meaning" in di erent circumstances (the intuitions that Grice
refers to in his analysing process), and additional data on the grammatical
ability of the subset of autistic speakers who are not (yet)
able to attribute beliefs. I believe that with this addition, Gluer and
Pagin's argument still stands.

Autistic spectrum disorder, which incorporates both Autism and
Asperger Syndrome (3), is a neuro-developmental disorder,4 currently
diagnosed on the basis of selective qualitative impairments in the areas
of communicative, social and imaginative abilities, accompanied
by patterns of restricted interests and repetitive or stereotyped behaviour.

The spectrum character of the disorder consists in the fact
that expression of these impairments (and also further symptoms not
currently incorporated in the diagnosis) varies signi cantly between
individuals and also within individuals over time.

Plug focuses on
linguistic ability and the ability to attribute representational mental
states such as beliefs.

About 20% of children on the autistic spectrum do
not develop functional language.6

At the other end of the spectrum
there are children whose linguistic development may be delayed,

7

but follows roughly the same path as the linguistic development of
typically developing children, when we concentrate on formal (note 8) or

(3. Asperger Syndrome will be classed as part of the Autistic Spectrum in the
forthcoming Diagnostic and Statistical Manual of Mental disorders (DSM-V) of
the American Psychiatric Organisation, although this is not the classi cation the
current Handbook adheres to (the DSM-IV: APA, 2000). (4. As evidenced by brain dysfunction and structural and functional di erences in regions of the brain (Surian et al., 1996). 5. According to DSM-IV: (APA, 2000). 6. Tager-Flusberg et al. (2005). In the past, the proportion of verbal versus nonverbal individuals on the autistic spectrum was thought to be 50/50 (Bryson et al., 1988). 7. Or not|children with Asperger's do not have a speci c language delay (APA, 2000). 8. In particular: children on the autistic spectrum that develop functional language
show the same learning curve in syntactic development as typically developing
children (the same range, and the same developmental ordering of grammatical
structures). Generally sentences are grammatically intact. There are also
no speci c problems with mastering questions, active passive transformations,
negation, and clausal complement construction, for example. See Tager-Flusberg
et al. (1990) for an overview. For a minority of autistic children, problems with
semantic aspects of language. However, these High-Functioning
Autistic or Asperger's children remain challenged in `pragmatic' aspects
of communication, even as adults:10

They may have di culty
sustaining a conversation (as shown by abrupt terminations or shift
in topics, and an inability to give and receive conversational cues, for
example), they may have di culty taking into account information
that a conversational partner can be expected to have and not to
have, are very literal minded, and may not understand jokes, irony
and sarcasm, except through a conscious workaround.11

Both the pragmatic problems children on the autistic spectrum
exhibit, and the impairments in the use12 and understanding13 of
mental state terms such as \belief," \think," \know" and \pretend"
past tense constructions are found, but it is hypothesised that these children form
a speci c subgroup that have SLI (\Speci c Language Impairment") co-morbid
with their autism (Tager-Flusberg and Joseph, 2003).

It is now thought that semantic organization is not signi cantly di erent from
the semantic organization of typically developing children (Paul et al., 2007), although studies in this area have been few.

For example; autistic children use
semantic groupings in typical ways to categorize and retrieve words (Ungerer
and Sigman, 1987; Lewis and Boucher, 1988). They are able to match a typical
exemplar with a less typical exemplar (Tager-Flusberg, 1985), and they show
the same pattern of overextension and underextension in relation to a prototype,
as typically developing children do (Minshew et al., 2002).

See also Toichi and
Kamio (2001) for a more subtle priming study showing similar semantic relationships
between words.

Vocabulary building also follows a roughly similar path as vocabulary building
in typically developing children: autistic preschoolers imitate words in advance
of naming things, their word comprehension is in advance of word production,
and gesture production is used as a bridge between the two (Charman et al.,
2003).

Di erent word types, such as common nouns, predicates, and so on, are
acquired in the same proportion to total vocabulary size as is the case with
TD's (Charman et al., 2003). On the whole, the order of emergence of predicate
categories in ASD children follows roughly the path as in typically developing
children (Peralejo, 2008).

(10 For an overview of the research, see Tager-Flusberg et al. (2005). (11. See in particular Happ e (1993) and Surian et al. (1996) for autistic children's problems with certain kinds of pragmatic inferences. 12 e.g., Tager-Flusberg (1992). 13 e.g., the use of `to know that such and such' without the understanding that
this implies that such and such is true, or the use of \to believe that such and
such," without understanding that this doesn't imply that such and such is true,
or is false (Dennis et al., 2001).


(a notable exception to the fact that semantic development in autism
compared to typically developing children is overall similar) 14

suggest
that autistic children have problems with attributing mental
states. These problems are well documented,

15

although the underlying
reasons for the impairments are being debated, e.g., whether it
is due to a conceptual de cit|no possession of the concept of belief
or thought

16

or mainly due to profound performance factors.17 One
of the most widely researched impairments is an impairment to attribute
false beliefs.

False Belief Tasks are designed to show whether
a child is able to attribute a false belief to themselves or another
person when this is demanded by context.

18

If a child is able to, she

(14. There are other exceptions, for example, the use of a ective terms, problems
with the use of deictic term (this,that), personal pronouns (\you want candy")
and the use of articles: this can be shown alongside correct usage of pronouns).
For an overview of the research, (Tager-Flusberg et al., 2005). (15. To name just a few indications: autistic children show impairments in areas where belief attribution is arguably required, such as the understanding of deception.
e.g., (Baron-Cohen, 1992) and surprise (Baron-Cohen et al., 1993), complex
causes (such as beliefs) of emotion (Baron-Cohen, 1991) spontaneous pretend play
(Lewis and Boucher, 1988) and the appearance/reality distinction (Baron-Cohen,
1989). There are also impairments in areas that are likely to be precursors to the
ability to attribute beliefs, such as joint attention behaviours, (Charman et al.,
2000), the understanding that something looks di erent from one viewpoint than
from another (Hamilton et al., 2009), and the understanding that seeing implies
knowing and not seeing `ignorance' (Baron-Cohen and Goodhart, 1994). (16. Some versions of the \Theory of Mind" De cit Account of autistic spectrum
disorders may posit this: they may for instance say that there is no \concept"
because the module responsible for attributing beliefs has not yet matured or is
faulty. 17. Some versions of the Executive Function De cit account of Autistic Spectrum
Disorder may posit this: they may claim that there is (or there might) be a
concept of belief, but that it can't be employed because it would require, for instance,
too much control over one's attention span or ask too much of the memory
system, or the calculations required to reason with it could be too di cult. (Note
that executive function problems may also be thought to impair the formation of
a concept of belief, depending on one views on concept development.) (18. A typical False Belief Task involves presenting a situation to the child being
tested where one of the protagonists is not privy to some information that the
child is privy to. For instance, Sally leaves her marble in a certain location and
then leaves the room. While she is away, Ann moves her marble from the rst
location to a second location. The child who witnesses these events is then asked,
when Sally returns, where will she look for her marble? Success on the task
passes the task, and this is taken to be unambiguous proof that the
child really is attributing a mental state, a belief, to the other person
rather than just base her predictions of other people's behaviour on
the basis of the state of the world or on the basis of her own belief
about the world). (19 Failure to attribute false beliefs does not by itself
indicate that mental states are not being attributed|theoretically
it's possible that children might be failing the task because they mistakenly
attribute a true belief;

20

but there is no evidence that might
make this plausible as a general rule.

21

And so, when one interprets
depends on whether the child will predict the Sally's behaviour correctly|on
the basis of the belief Sally has (that the marble is in the rst location) which
is di erent from the child's own belief, who has witnessed the moving of the
marble and so knows that its in the second location instead. The classic False
Belief Task was devised by Wimmer and Perner (1983), taking up a suggestion
by Dennett (1978). See for the rst variant of this task administered toautistic
children (Baron-Cohen et al., 1985).

19

Some may think that a prediction of other's behaviour based on the state
of the world is also a kind of content attribution. However, such an attribution
is done without an implicit or explicit understanding of the representational
character of mental states. Grice's analysis of meaning requires not just that
one attributes content, but that one intends to produce a belief in the audience
(and even more than that): and this requires that one implicitly or explicitly
understands the representational character of belief.

20

Pug's approach di ers slightly from Gluer and Pagin, who in part seem to use a
philosophical argument to establish the credibility of False Belief Task failure to
delineate the subset of autistic speakers: \[...]the problem is not that there is only a problem with attributing false beliefs. The problem is that understanding what
it is to have a false belief is an essential part of understanding what it is to have a belief at all." (Gluer and Pagin, 2003, p. 37) and: \to understand the di erence between being true and being believed to be true, one must understand that a
belief can be false, and this understanding is manifested by means of the ability
to ascribe beliefs one takes to be false" (Gluer and Pagin, 2003, p. 27).

This may
be true, but it doesn't help to pinpoint an impairment in belief attribution (if
that is what Gluer and Pagin mean here) because what is required is knowing
when this understanding is not available (not when it is). False belief test failure
in itself only suggests an impairment in explicit False Belief attribution, not an
impairment in the understanding that beliefs are representational.

21

At least, not until False Belief Tasks are almost passed.

There is evidence
that `Diverse Belief' tasks (tasks that test for the understanding that di erent
people may have di erent beliefs about the same situation at the same time|and
therefore supposedly test for the understanding that beliefs are representational|
without having to attribute a false belief) are passed just before passing False
False Belief Task failing from within the wider context just outlined
above, one can assume that failing such tasks means (near enough)22
that beliefs are not yet being attributed.23; 24

Gluer and Pagin use the data on False Belief Task performance to
home in on a subset of autistic speakers who speak on a sophisticated
level, and yet are not able to attribute mental states to other people.
Such a group of speakers are likely to be a counterexample to Grice's
analysis of meaning, given that Grice's analysis posits that in order
to mean something, a speaker must intend to produce beliefs in their
audience:

Grice's aim is to analyse the ordinary English concept of "meaning," in all its forms.

In order to be capable of setting up an analysis, one must already be capable of applying the concept in particular cases.

A Belief tests in autistic children (Peterson et al., 2005), and one therefore has to
be careful to interpret the transition from failing to passing False Belief Tasks as
exactly pinpointing an understanding of the representational character of mental
states.

But there is no general evidence that would suggest an explicit or implicit
understanding of the representational character of mental states in autistic children
well before passing these tests.

For instance, simplified False Belief Tasks are not passed before standard False Belief Tasks (Surian and Leslie, 1999), and nonverbal (or implicit) False Belief Tasks are not passed before standard False Belief Tasks. (Senju et al., 2009, ress).

Also, False Belief Task failure should be interpreted in the context of the other evidence for impairments in the attribution of beliefs (see footnote 15), and lastly, there is speci c evidence that spontaneous (untrained) passing of standard False Belief Task is correlated with enhanced pragmatic abilities (Eisenmajer and Prior, 1991), enhanced social skills in a naturalistic setting (Astington and Jenkins, 1995; Frith et al., 1994) and increased
use and better understanding of mental state terms (Ziatas et al., 1998).

Although many impairments remain even after passing, on a more subtle level.

22 With the caveat of footnote 21 kept in mind.

23

Plug presents a more thorough defence that False Belief Task failure can be so
interpreted in the PhD dissertation, King's College (Plug, 2010).

24

Autistic children are not alone in having problems with the attribution of
false beliefs at a certain point in development. Typically developing children
are unable to pass False Belief Tasks before the age of around 3.5. However,
typically developing children are not likely to pose a problem for Grice's analysis
of meaning because on the one hand, their language is not yet very accomplished
(and is of a much lower level than the language of the autistic children considered
in this paper). And on the other hand, it is far less clear in the case of typically
developing children that False Belief Task failure is indicative of problems with
the attribution of mental states such as belief.


conceptual analysis utilises one's intuitions concerning these cases to
arrive at necessary and su cient conditions for the concept to apply.

Some of our intuitions regarding use are obvious.

Grice seems to
believe this is the case with a distinction between two ways in which
we use the term "meaning" in our ordinary talk.

25

According to
him it is reasonably clear, intuitively,

26

that we can distinguish our use of "mean" in cases we might call "natural meaning;"
a sense which is closely related to something being a natural sign
for something else (e.g., \clouds mean rain"), from our uses of the
term in cases we might call \nonnatural meaning," (e.g., \ `the cat
is on the mat' means that the cat is on the mat")which is \related to
communication."

27

It is the second type of meaning that he subjects to a more careful conceptual analysis.

Plug discusses some of the intuitions regarding use that Grice rejects while analysing nonnatural meaning later.

The end result of the analysing process, is that the most fundamental type of non-natural meaning, utterer's meaning, has the following necessary and
suffcient conditions:

For an utterer U, an utterance x, a response R and time
t:

U means something (that p) at t by uttering x, i for some audience
A, U uttered x intending thereby

1. A to produce a particular response R.

2. A to think (recognize) that U intends 1.

3. A to ful ll 1. on the basis of the ful llment of 2.28

In the case of the making of statements (rather than orderings
for example)and in our discussion we will limit ourselves to Grice's
analysis of statements

29

the response the speaker intends to produce
in the audience must be a particular belief (the content of which

25 See Grice (1989a, p. 214) and Grice (1989b, p. 291).

26 Grice (1989b, p. 291).

27 Ibid.

28 This is roughly the formulation of Grice's initial analysis (Grice, 1989c, p. 92).
29

Because the threat the existence of certain autistic speakers pose to Grice's
program is easiest to present by focusing on his analysis of statements, and also


supplies the content of the statement).

30

Grice is aware that speakers seem to speak quite automatically, and are normally not aware that they possess the higher-order intentions (intentions about beliefs
about intentions about beliefs) he proposes they possess.

Therefore, Grice must take these intentions to be unconscious, tacit, or
implicit.

Nevertheless, speakers must have these intentions.

The intentions must be attributable to the speaker.31

We have seen that autistic children who do not yet pass False Belief
Tasks, are not able to attribute beliefs,

32

and therefore in their
turn can't be attributed the intentions required: they can't intend
to produce a particular belief in the audience on the basis of the
audience's recognising that they intend to produce this belief in the
audience.

33

Nevertheless, Gluer and Pagin argue that a subset of
because the analysis of statements is central to any analysis of meaningful speech.

Plug sets out the threat to his analysis of orderings etc. elsewhere (Plug, 2010).

Such
an argument follows similar lines, but must also refer to the di culty a subset of
autistic speakers have with second order intentions.

30

According to Grice: \to ask what U meant is to ask for the intended e ect"
(Grice, 1989a, p. 220).

In later writings, the intended e ect is suggested to be a
particular thought, an activated belief, a belief about a belief of the speaker, or
an activated belief about the belief of the speaker, and so on (for an overview,
see Neale, 1992, p. 38).

There have been other variations on the formulation
of the analysis (e.g., in later writings, the self-referential nature of the earlier
formulation is replaced) but in all the proposed re nements at the core of the
analysis remains the intended production of a thought or belief.

31

There is some debate about what Grice believes the exact conditions to be
under which one can attribute an intention to a speaker (or in other words, the
conditions under which a speaker possesses the intention).

One may for instance
interpret Grice as supposing that it is not necessary that meaningful speakers
must be able to form explicit intentions of the kind mentioned at the time at which
something is meant, as long as it is possible to form these intentions at a later date and it is also recognised or in some exceptional cases \decided" (Grice, 1989a,
p. 222) at that later date that these type of intentions were formed implicitly
at the earlier date.

If this reading of Grice is correct, additional empirical data
on autistic children is required for Gluer and Pagin's argument to go through. Plug
discusses this in the thesis (Plug, 2010).

33

The formation of such an intention also requires higher order thought: it
requires thinking about the audience's thinking process, because Grice believes
that the audience's recognition of the intention of the speaker is not just a cause
for the audience but also a reason (Grice, 1989a, p. 221), and Grice believes one
can't in general \form intentions to achieve results which one sees no chance of
these children who do not pass False Belief Tasks, have a \sophisticated"

34

linguistic ability. They cite a large metastudy by Francesca
Happ e that has shown that that some autistic children require the
verbal age of a typically developing 9 to 11-year-old35 in order to pass
these tasks (this is of course with IQ and other possible in
uences

partialled out).

36

Gluer and Pagin believe that it would be absurd to
deny, on the basis of their linguistic level, that these speakers mean
something

37

when they speak38 and that they are communicators.

39

Are Gluer and Pagin correct to infer that autistic speakers mean
something, from the data on linguistic development and ability that
they put forward?

It is clear that autistic speakers mean something, in our ordinary
sense of \meaning" (before re
ecting on which types of meaning
should be distinguished in our ordinary use). For instance, because
vocabulary measures test for the correct usage of words,40 it can
achieving" (Grice, 1989c, p. 98).

34

Gluer and Pagin (2003, p. 47).

35

To put it like this is slightly misleading, because the grammar and vocabulary
measures are typed on large random groups that may also include some speakers
with autism. But it is true near enough; these language measures have been
checked extensively for reliability.

36Happ e (1995).

37

Gluer and Pagin phrase it di erently: they say that it would be absurd to
deny that these speakers are language users.

Because Grice does not analyse what it is to be a language user, but what it is to mean something, Gluer and Pagin must take the ability to speak meaningfully (in the nonnatural sense) to be entailed by being a language user.

To avoid this issue Plug focuses directly
on whether, from the data presented, it can be concluded that autistic speakers
mean something (in the nonnatural sense).

38

Plug colloquially refers to autistic utterers meaning something
with their words," or meaning something \when they speak" but Plug does not want
to suggest with this that Grice thought that speakers mean things only when they
use words and sentences (that is, that Speaker Meaning is exhausted by linguistic
meaning) or that Word Meaning is more fundamental than Speaker Meaning.

It
is only that in this paper, we happen to assess the fact that autistic speakers
mean something on the basis of linguistic competence.

39

Gluer and Pagin are aware that the claim that autistic speakers are communicators
requires more argument than referencing linguistic ability. I will discuss
their argument below.

40

On the Peabody Picture Vocabulary Test - III (PPVT) (Dunn and Dunn,

1997) or its British equivalent, the British Picture Vocabulary Scale (BPVS)
hardly be maintained that these words as spoken by them have no
content.

But it is a di erent question whether this sense of meaning is the sense that Grice analyses.

Grice distinguishes two di erent uses of the term \meaning" in our day to
day speech: \natural meaning" and \nonnatural meaning," which
is \related to communication."

41

To make this distinction doesn't require a lot of re
ection:

Grice talks about it as a "reasonably
clear intuitive distinction",

42

and he offers a couple of "recognition
tests"43

that allow one to see with what type of meaning one is dealing
with, if in doubt. These are, rstly, that in the case of nonnatural
meaning, unlike in the case of natural meaning, there is always someone
or somebody meaning something.

Secondly, that in the case of
nonnatural meaning the `something' that is meant is \nonfactive":
what Grice means with this is that it would not be absurd to say
that someone means something, but that it is not the case (or false).

Lastly, what is meant can be paraphrased using quotation marks in
the case of nonnatural meaning, but this can't be done with natural
meaning.

44

If some speakers would use their words in the right context, but
not implicitly or explicitly understand that their words are representations,
and that therefore sentences containing them could be
either true, or false, it would not be unlikely that their \meaning
something" would mainly consist in a correlation between the words
and things in the world.

If this was the case with our autistic speakers,
it could be argued that they only mean something in the natural
sense.

45

The data that Gluer and Pagin present, does not rule out
(Dunn et al., 1982), the two vocabulary tests referred to, lexical comprehension
is tested by presenting an auditory word and asking the participant to point to
or state the number of the corresponding picture (depicting objects, actions, etc.)
from an array of four (numbered) pictures.

41Grice (1989b, p. 291).

42Ibid.

43Ibid. See also Grice (1989a, p. 214).

44

There is one more recognitions test: In the case of natural meaning (and not
nonnatural meaning), what is meant can be restated beginning with the phrase
\the fact that. . . "

45

An argument of that sort would have to deal with conflicting intuitions: for
instance, the tension between our intuition that this is a case of \someone meaning


this possibility (the only studies cited that correlate verbal ability
with False Belief Task performance are standard vocabulary measures,
which only test object-\meaning" correlations)

46

but Gluer
and Pagin are aware that it must be ruled out, rstly, because they
present vocabulary performance data in the context of other studies
that suggest that linguistic development is overall comparable to the
development of typically developing children (although delayed).

47

Secondly, they cite additional data suggesting that there are autistic
speakers that understand negation and a rmation.

48

And thirdly:
they cite additional data that vocabulary age in autistic speakers
is in general correlated with grammar age

49

in short, they aim to
argue that when we are talking about a group of autistic children
with a vocabulary age of 8, say, this group will on average have a
general verbal age of 8 (and that this includes an understanding of
negation, etc).

However, the problem with these three sets of data is that they
concern the linguistic development of autistic speakers in general.

The subset of autistic speakers we are concerned with, a set of sophisticated
speakers who do not (yet) pass False Belief Tasks, could
be a group of speakers with an exceptional linguistic development.50

In particular, from the data it couldn't be ruled that despite their
something" and our intuition that words are close to signs for those who do not
understand the representational character of words, suggesting that the meaning
involved is natural.

46

e.g., the large metastudy by Happ e (1995) only investigates correlations between
British Picture Vocabulary Scale level and False Belief Task performance.

47

Gluer and Pagin (2003, p. 29).

48

Gluer and Pagin (2003, p. 33).

49Ibid.

50

Note that the subset of autistic speakers that is considered to be a counterexample
to Grice's analysis is not a diagnostic subgroup in autism.

This is why studies that investigate the correlations between grammar and verbal ability can't be expected to test this speci c subset for alternative language pro les. For example, the study that Gluer and Pagin refer to (Jarrold et al., 1997) does not do so.

Also, the subset is relatively small, and so a divergent language pro le
of language attainment in the subset would still be compatible with a relatively
uniform pro le of language attainment for autistic children in general.

Lastly,
the fact that linguistic development in general is similar, although delayed, does
not help because we want to home in on a speci c timeframe: linguistic ability
before passing False Belief Tasks.


high vocabulary age, the speakers in our subset do not yet use negation
and a rmation correctly, since precisely those children who do
not yet pass False Belief Tasks could be de cient in this understanding.
And so, there could be an exception to the rule that vocabulary
age matches grammar age precisely in those children who have not
yet passed False Belief Tasks.51

To rule out that the children in the subset mean something in the
natural sense, it is necessary to present data on grammar development
and ability in relation to False Belief Task performance.

Such data exist.

There are children with a grammar level of typically developing
6 to 9-year-old, that do not (yet) pass False Belief Tasks.52

This level of grammar does not di er in fundamentals, from adult grammar,

53

and it means speci cally that negation and embedded sentences have already been mastered (as these are in place in typically developing children from six years onwards).54

Such a high level of grammar suggests that it is understood that words and sentences
are representational, and that sentences can be true or false.

Granted, if the only contexts in which the terms \true" and \false"
were used correctly were simple contexts, this would not suggest representational
understanding (because these words could be referring
to other concepts than truth and falsity). But sustained cor-

51

There are children with autism with exceptional rote memory skills.

Therefore, children who have excellent vocabulary age before passing False Belief Task (rather than the reverse) may have acquired a large vocabulary \associatively"
by rote memory learning.

52

See Tager-Flusberg and Sullivan (1994); Fisher et al. (2005).

Note that because
of the fact that Diverse Belief tasks are passed a little before False Belief
Tasks (see footnote 21), it is best to go with a conservative choice in delineating
the subset and considers autistic speakers that do not pass False Belief Tasks
with a grammar age of 7, rather than 9, as a putative counterexample.

As far as Plug is aware, there are no studies (yet) that test Diverse Belief correlations with linguistic ability.

53

Grammar does not stop developing after age 6 or 7, but the grammatical
constructions that are learned in middle childhood (between 6 and 10) are often
exceptions, and are relatively infrequent, involving access to more than one type
of grammatical knowledge (Goodluck, 1991, p. 98).

54

In typically developing children, mature forms of negation appear after the
age of four. After the age of ve, there is an increase in the amount of complex
sentences (e.g., embedded sentences) produced relative to simple sentences
(Tager-Flusberg et al., 2005, p. 338).


rect usage of the terms in complex situations, and especially correct
usage in context of talk about talk (about what is said or about
the sentences uttered) rules out that these children are using the
term `false' to mean something other than false, and shows that
the representational character of sentences is understood.

Therefore,
autistic children with a grammar level of a typically developing 6
to 9 year old understand the representational character of language;
(because they must have already mastered negation and embedded
sentences).

To corroborate this, there is also speci c evidence that
the use of negation in combination with communication verbs (such
as \saying") is mastered before false belief tests are passed, in some
autistic children.55

It is clear, then, that no case can be made that
these children must mean something with their words in the natural
sense, when they speak.

The speakers in our subset do not mean something in the natural
sense with their words, but from this it cannot be inferred without argument
that they mean something nonnaturally:

Grice distinguishes,
on immediate intuitive grounds, two senses of meaning in our day to
day use of the term, but this doesn't mean that more careful re
ection on our use of the term would not distinguish a third (or even
more) type(s) of meaning.

A third type of meaning might be implicit in our use, of which we might not be immediately aware.

Careful reflection might bring it out.

That this is not a ridiculous proposition becomes clear when one realises that Grice o ered at least some reason to suppose that nonnatural meaning might be partly characterised on the basis of a notion of communication.

This reason is not a good reason and Plug argues against it.

But Plug thinks it cannot be dismissed out of hand, and that therefore an argument against this possibility should be added to Gluer and Pagin's argument.

What reason may one have to suppose that nonnatural meaning
should be partly characterised on the basis of a notion of communication?

First of all, as noted above, Grice sees it as intuitively clear that
one of two uses of meaning is \related to communication" (namely,
nonnatural meaning), and it is clear that on the analysis of the most
fundamental type of nonnatural meaning (\Speaker Meaning"), a

55

Tager-Flusberg and Joseph (2003).


speaker meaning something with her words entails that the speaker
intends to communicate with the hearer (in an intuitive sense of communication).

Also, Grice's systematic reflection on the intuitions
regarding the use of the term \meaning" while analysing, suggests
that the idea of communication was kept in mind during this process
(as we will see below).

So, it would not be farfetched to believe that
nonnatural meaning is partly characterised on the basis of a notion
of communication.

This possibility needs to be further evaluated,
especially because autistic speakers, and certainly speakers that do
not yet pass false belief tests, have trouble with communication, including
linguistic communication (as evidenced by their pragmatic
problems, for instance).

Therefore, the autistic speakers in our subset
may not mean anything nonnaturally if doing so requires one to
communicate, or requires having a sophisticated concept of communication,
or requires having the intention to communicate.

One person who criticises Gluer and Pagin's argument on the
basis of the communicative problems autistic speakers are known to
have, is Ann Reboul (2006), of GRICE, Group pour la recherche de la comprehension elementaire.

Reboul argues that autistic speakers'
communicative problems show that they do not mean (in Reboul's
words: \do not assert"

56)

in the same way as typically developing
speakers.

Behind Reboul's argument, is the fact that she takes Gluer
and Pagin to be arguing that \[h]igher order intentions [...] do not
play a role in the process [of linguistic communication]."

57

Reboul believes that Gluer and Pagin look at particular instances of assertion
(something that is arguably both a matter of meaning something and
of communication),

58

and aim to show that that particular instance
could have taken place without the speaker possessing the required
higher order intentions.

So, Reboul suggests, their argument is analogous
to \other epiphenomenal"59 arguments such as Chalmers': in
his argument for the epiphenomenal character of qualia, Chalmers

56

Reboul (2006, p. 592).

57

Reboul (2006, p. 590) does not properly explain why she interprets Gluer
and Pagin's argument in this manner.

But Reboul observes it right after noting that
Gluer and Pagin do not claim that typically developing speakers do not \have
higher order intentions when they communicate" (Reboul, 2006, p. 590).

58Reboul (2006, p. 588).

59Reboul (2006, p. 591).

invites people to conceive of \twin earth," which is a world that is
micro-physical identical to ours, and to conceive of our counterparts
in it, who are behaviourally indistinguishable from us but who do not
experience qualia.

That such a world is conceivable, and therefore
possible, implies that qualia are not causally e ective in the production
of behaviour, or so the argument goes very roughly.

However|
Reboul's argument continues|because Chalmers describes a hypothetical
situation, he can stipulate that the behaviour in the two
instances (on earth and on Twin Earth) is the same.

Whereas, because
Gluer and Pagin are describing a real situation, they have to
argue that the particular instances of assertion are similar.

But, focusing
just on syntactic form, a sentence of that form may mean
one thing coming out of the mouth of one person and another thing
coming out the mouth of another.

60

To assess similarity, it becomes
necessary, then, to refer to content, she says: the question is whether
a sentence of a particular form would ful l the same role in communication
when spoken by autistic speakers as it would have done if
it had been spoken by typically developing speakers.

But the communicative
problems show that for autistic speakers, assertions do
not play the same role, and therefore, Gluer and Pagin's argument
doesn't go through.

Plug agrees with Reboul only that Gluer and Pagin o er a conceptual
\possibility" that happens to be an actuality (the conceptual possibility
would have been a su cient counterexample), and that they
therefore can't stipulate that what is meant by autistic speakers is
meant \in the same way" as what is meant by typically developing
children|it requires argument that they mean in the same way.

But
it is not necessary, for Gluer and Pagin's argument, that there is a
similarity between particular assertions that autistic speakers make
and typically developing children (would) make.

That these assertions
have the same content, play the same role, or whatever|what
matters is that in both types of cases something is meant (in the
nonnatural sense).

Also, nothing Gluer and Pagin say entails that
higher order intentions do not play a role in communication (they

60

Reboul (2006, p. 592).


address Grice's analysis of meaning only61).

Nor do they imply that
higher order intentions could not play a role in the process whereby
a speakers comes to mean something, or even in the act of a speaker
\meaning something".

62

it is just that nonnatural meaning cannot
be analysed in the way proposed, if the counterexample is successful.

So, it is not that particular contents of the speech of autistic
speakers and typically developing speakers need to be the same, but
they need to be of the same type.

Communicative problems may
well show that particular contents are dissimilar, but we have yet to
establish whether communicative problems have a bearing on what
type of meaning autistic speakers exhibit.

Gluer and Pagin themselves are sensitive to the suggestion that
communication is relevant to meaning something nonnaturally.

They
describe Grice's distinguishing between natural and nonnatural meaning
in the following way:

In an effort to separate out nonnatural
meaning from mere causal regularities Grice correctly turns to the
intention to communicate [...]."63

However, they say, even if an intention
to communicate (partly) characterizes meaning something
nonnaturally, this does not mean that autistic speakers don't mean
nonnaturally.

Although the autistic speakers in the subset do not
possess the same concept of communication as typically developing
speakers, they nevertheless can distinguish communicative situations
on the basis of perceptual similarity, Gluer and Pagin argue, and so
may want to create new situations of that perceptual kind.

64

And,
autistic speakers do indeed initiate conversations,

65

and so it seems
that they do intent to communicate in the more minimal sense just
outlined.

****************************************************************
However, Grice's notion of communication is not a minimal notion.
*****************************************************************

When Grice discusses his theory of communication, he talks about the goals of communication (such as the giving and receiving of in

61

Although they sometimes write as if they are also addressing his theory of
communication, see for instance Gluer and Pagin (2003, p. 47).

62

A conceptual analysis of \swimming" may not refer to ns or arms, and yet
ns or arms may play a role in a particular act of swimming.

63

Gluer and Pagin (2003, p. 47).

64

Gluer and Pagin (2003, p. 48).

65

Gluer and Pagin (2003, p. 47).

formation, in uencing and being in uenced by others) as goals that
are recognised by each participant in a communicative situation as
a shared goal: involving a joint purpose, a mutually accepted direction.

66

What reason is there to assume that it is a minimal notion of communication that is related to nonnatural meaning, rather than Grice's own more substantial notion?

Plug believes that whether and
which notion of communication plays a part in the analysis of nonnatural
meaning can only be settled by taking into account the intuitions
concerning our use of the verb "mean" that come to the
fore when one is analysing meaning.

Since Grice does offer an analysis, we can use the references he makes to intuitions during his analysis as a guide to see whether nonnatural meaning is to be characterised in terms of communication and if so, whether the intuition is reliable.

Firstly, although as we saw it is \intuitively clear," according
to Grice, that nonnatural meaning is related to communication, by
itself this doesn't imply that meaning is characterised in terms of
communication (just that it is related to it).

Secondly, because the
distinction between natural meaning and nonnatural meaning is so
intuitively clear, Grice is able to identify the \recognition tests" almost
immediately, before subjecting the intuitions regarding the use
of the verb "mean" to a more systematic analysis.

As can be
seen from the list mentioned above, none of these tests for nonnatural
meaning refer to the intention to communicate, or explicitly use
the concept of communication, or implicitly describe communication.

So, Grice doesn't refer (before commencing on a more systematic
reflection) to any intuitions that might imply that nonnatural
meaning is to be characterised in terms of communication.

Now we
need to establish whether he does while engaged in the more careful
conceptual analysis (and what this entails about the correct characterisation
of meaning).

During the analysing process, Grice systematically
tests our intuitions regarding the use of the term `meaning'
(in the nonnatural sense), by offering specific hypothetical scenarios
and asking in each case whether one would or wouldn't want to apply

66

See Grice (1989d, p. 28) for Grice's view that talk is adapted to serve a
particular purpose, Grice (1989d, p. 26) for Grice's idea that participants in a
talk exchange recognise a common purpose or mutually accepted direction, and

Grice (1989d, p. 30) for what these purposes are.


the term \meaning" (in the nonnatural sense).

All in the service of
moving towards a general characterization of one's use of the term
\meaning."67

Looking at Grice's analysing process, it's clear that Grice doesn't explicitly refer to communication, but his choice of starting point does suggest that the idea of communication (I take it that this idea always involves the idea of others as receivers, so to speak) is on his mind.

Grice starts out asking whether intending to cause a belief in someone is sufficient to mean something.68

Counterexamples show that it isn't.

But the condition is retained as a necessary condition in the analysis of meaning (as related to the making of statements and the like, not orderings).

However, Grice offers no systematic investigation into the necessity of this condition.

He has offered no crucial hypothetical scenarios that suggest (directly) that we would indeed be inclined to withhold the term \meaning" in every case
where the condition is not ful lled.

It could be of course, that the necessity of the condition can
be (indirectly) deduced from our intuitions regarding the use of the
term \meaning" that are being highlighted by the hypothetical circumstances
that Grice does consider.

***********THE HANDKERCHIEF SCENARIO*******************

For example, Grice nds he
wants to withhold applying the term \meaning" in a case where
someone leaves a handkerchief of a particular person near a scene of
a murder (in order to produce in the mind of the police the belief
that that person is the murderer).

*************THE PHOTOGRAPH SCENARIO*******************

Or, when someone shows a husband
a photograph of his wife having an a air (in order to produce
in the husband the belief that his wife is having an a air).


****************THE DRAWING SCENARIO*****************

On the
other hand, Grice nds he doesn't feel reluctant to withhold the term
\meaning" in a case where someone shows the husband a drawing of
his wife having an a air (in order to etc.).

Grice nds that what prevents
him from applying the term \meaning" in the rst two cases,
is that the belief would have been produced in the audience anyway,
even if there had not been an accompanying intention by the speaker
(if the photograph had been lying around by accident, for instance).


This leads him to add as a condition that the belief production in the

67

Grice (1989a).

68

Grice (1989a, p. 217).


audience must proceed via the recognition that the speaker intends
to produce that belief in them.69

However, our intuitions to withhold applying the term \meaning"
in the rst two counterexamples, can also be explained in another
way.

Plug suggests that if it were possible for a speaker to present
a photograph as being true (which requires the possibility that it
could be false), or the handkerchief as being true, then we would not
feel inclined to withhold the application of the term \meaning" in
these cases.

In such a case it would not be possible for the handkerchief
or the photograph to be causing beliefs in the audience just by
lying around|they would only do so if presented as true.

Plug is not
o ering this as a step in an analysis of Plug's own (Plug is not suggesting
all matters of meaning nonnaturally are a matter of presenting as
true or as false), but simply as an observation that, although Grice
seem to have been led by the idea of communication in his analysis,
it is not necessary to interpret our intuitions concerning the use of
the term \meaning" as involving an idea of communication.

So, neither in making the intuitive distinction between natural
meaning and nonnatural meaning, and o ering recognition tests, nor
in the more careful analysis of nonnatural meaning, have intuitions
regarding the use of the term `meaning' become apparent that would
suggest that meaning is to be characterised in terms of communication.

Plug believes it can therefore be concluded that nonnatural meaning
is not to be characterised in terms of communication.

Should another type of meaning be distinguished in our use of
the term \meaning," apart from the natural sense, the nonnatural
sense, and the now rejected \nonnatural sense characterised in terms
of communication"?

Since nonnatural meaning is related to communication
in general, there was some reason to think that nonnatural
meaning should be characterised in terms of a notion of communication.

But it is unlikely that another type of meaning should be
distinguished in our use of the term `meaning' that has so far completely
avoided detection (and, incidentally, it is unlikely that Grice

69

Grice (1989a, pp. 217{219). Plug intentionally passes over a step in Grice's argument: namely that the second counterexample shows that the belief production must proceed via the intention recognition and that the rst counterexample just shows that there must be intention recognition.

These details, Plug thinks, do not affect the argument.

himself would think this is a possibility, given that he proposed that
one should not multiply senses beyond necessity):

70

because if the
autistic speakers somehow did not t comfortably in either the \natural"
meaning or the \nonnatural" meaning category, there may be
a reason to suspect that a third distinction was implicit in our use,
applicable to the autistic speakers in our subset, but this is not the
case.

All the recognition tests that have been o ered with regard to
nonnatural meaning, comfortably t the autistic speakers.

We have established that, without being able to communicate in
Grice's substantial sense, or form the intention to so communicate
(whether minimal or substantial), autistic speakers can understand
that sentences that are uttered can be true or false, and can themselves
utter sentences that they understand to be either true or false.

It is just as di cult to establish whether autistic speakers are in fact
presenting sentences as true, or as false, rather than just mouthing,
or using words purely \associatively," as it is for any speaker|this
depends on context.

But if they do, it can be concluded that they, as
speakers, mean something, that what is meant can be paraphrased
in a sentence using quotation marks, and that it would not be absurd
to say things like \she (the autistic speaker) means that the cat is
on the mat, but the cat is not on the mat."

Plug argues that Reboul's contention that the communicative
impairments of the autistic speakers in our subset suggest that their
particular assertions do not have the same meaning or do not play
the same role in communication as syntactically identical sentences
would have had, if they had been uttered by typically developing
speakers, is not relevant to Gluer and Pagin's argument.

And this is because
Gluer and Pagin aim to o er a counterexample to a Grice's analysis
of a meaning type (not an argument that higher order intentions play
no role in the formation of particular instances of meaning).

However,
Plug argues that the idea that autistic speaker's communicative
impairments suggest that autistic speakers may not mean
in the nonnatural sense, in other words, may not mean in the sense
of meaning that Grice analyses, needs to be dealt with.

It is not obvious
that a speaker with a vocabulary age of between 9 and 11 means

70

Grice (1989e, p. 47).

something nonnaturally, and therefore not obvious without further
argument that autistic speakers pose a counterexample to Grice's
analysis of nonnatural meaning.

A subset of autistic speakers show
advanced grammatical ability before being able to attribute beliefs,
an ability that does not di er signi cantly from adult grammar, and
where in particular correct use (in context) of negation and a rmation
are in place, in embedded sentences such as \She said that x, but
it is not true that x."

Plug argues that this rules out the hypothesis
that the speakers in the subset mean in the natural sense.

The
same grammatical data also show that the autistic speakers' type of
meaning perfectly passes the recognition tests, proposed by Grice to
help one decide with which type of meaning one was dealing, if in
doubt.

No other intuitions, suggesting either that nonnatural meaning
should be characterised in terms of communication, or that some
other type of meaning should be distinguished alongside natural and
nonnatural meaning, have been revealed by contemplating Grice's
analysing process, and I have therefore concluded that the subset of
autistic speakers do indeed mean nonnaturally, and therefore form
a counterexample to Grice's account.

And so, we can conclude that
Gluer and Pagin's argument still stands. Or not.

REFERENCES

APA.

Diagnostic and statistical manual of mental disorders: DSM-IV-TR. American Psychiatric Association, Washington, 4th edition.

Astington, J. W. and Jenkins, J. M.

"Theory of mind development and social understanding: connections between emotion
and understanding in development, 9

Baron-Cohen, S.

"Are autistic children "behaviorists"? an examination
of their mental-physical and appearance-reality distinctions.

Journal of Autism, 19

--- Do people with Autism understand what
causes emotion? Child Development, 62

--- Out of sight or out of mind? Another look at
deception in autism. Journal of Child Psychology and Psychiatry,
33

--- and Goodhart, F. The \seeing-leads-toknowing"
de cit in autism: The Pratt and Bryant probe. British
Journal of Developmental Psychology, 12

--- Leslie, A. M., and Frith, U.
Does the
autistic child have a \theory of mind"? Cognition, 21(1):37{46.

--- Spitz, A., and Cross, P. (1993). Can children with
autism recognize surprise. Cognition and Emotion, 7

Bryson, S. E., Clark, B. S., and Smith, I. M.

First report of a Canadian epidemiological study of autistic syndromes.

Journal of Child Psychology 29

Charman, T., Baron-Cohen, S., Swettenham, J., Baird, G., Cox, A.,
and Drew, A.

Testing joint attention, imitation, and play as infancy precursors to language and theory of mind. Cognitive Development, 15

Charman, T., Drew, A., Baird, C., and Baird, G.

Measuring early language development in preschool children with autism spectrum
disorder using the MacArthur Communicative Development Inventory (Infant Form). Journal of Child Language, 30.

Dennett, D. C.

Beliefs about beliefs. Behavioral and Brain Sciences, 1

Dennis, M., Lazenby, A., and Lockyer, L.

Inferential language in high-function children with autism.

Journal of Autism 31

Dunn, L. M. and Dunn, L. M.

Examiner's manual for the
PPVT-III peabody picture vocabulary test. American Guidance
Service, Circle Pines, 3rd edition.

Dunn, L. M., Dunn, L. M., Whetton, C., and Pintilie, D. (1982).
British Picture Vocabulary Test. NFER-Nelson, London.

Eisenmajer, R. and Prior, M.

Cognitive linguistic correlates
of \theory of mind" ability in autistic children. British Journal of
Developmental Psychology, 9(2):351{364.

Fisher, N., Happ e, F., and Dunn, J.

The relationship between
vocabulary, grammar, and false belief task performance in children
with autistic spectrum disorders and children with moderate
learning di culties. Journal of Child Psychology and Psychiatry,
46

Frith, U., Happ e, F., and Siddons, F.

Autism and theory of
mind in everyday life. Social Development, 3

Gluer, K. and Pagin, P. (2003). Meaning theory and autistic speakers.
Mind and Language, 18

Goodluck, H. (1991). Language acquisition: A linguistic introduction.
Blackwell Publishers, Oxford.

Grice, H. P. 1938. Privation and meaning. Harborne, Warwickshire.
-- 1941. Personal identity. Mind.
-- 1948. Meaning, repr. in WoW
-- 1949. Disposition and intention.
-- 1961. The causal theory of perception.
-- 1965. Some remarks about the senses.
-- 1966. Logic and conversation: the Oxford lectures
-- 1967. Utterer's meaning and intentions: the fifth William James lecture
-- 1982. Meaning revisited, in WoW
-- 1989. Studies in the Ways of Words, chapter Meaning,
-- 1991. The conception of value.
-- 2001. Aspects of reason.

Hamilton, A. F. C., Brindley, R., and Frith, U. (2009). Visual perspective
taking impairment in children with autistic spectrum disorder.
Cognition, 113(1):37{44.

Happ e, F. Communicative competence and theory of mind
in autism: A test of relevance theory. Cognition, 48(2):101{119.

Happ e, F.

The role of age and verbal ability in the theory
of mind task performance of subjects with autism. Cognition,
66(3):843{855.

Jarrold, C., Boucher, J., and Russell, J. (1997).

Language pro les
in children with autism: Theoretical and methodological implications.
Autism, 1(1):57{76.

Laurence, S. (1998).

Convention-based semantics and the development
of language. In Carruthers, P. and Boucher, J., editors,
Language and Thought. Interdisciplinary Themes, pages 201{218.
Cambridge University Press, Cambridge.

Lewis, V. and Boucher, J. (1988).

Spontaneous, instructed and
elicited play in relatively able autistic children. British Journal
of Developmental Psychology, 6(4):325{339.

Minshew, N. J., Meyer, J., and Goldstein, G. (2002).

Abstract reasoning
in autism: A dissociation between concept formation and
concept identi cation. Neuropsychology, 16(3):327{334.

Neale, S. (1992).

Grice and the philosophy of language. Linguistics
and Philosophy, 15(5):509{559.

Paul, R., Chawarska, K., Klin, A., and Volkmar, F.

Dissociations
in the development of early communication in autism
spectrum disorders. In Paul, R., editor, Language disorders from
a developmental perspective: Essays in honor of Robin Chapman,
pages 163{194. Lawrence Erlbaum Associates, Inc., Mahwah.

Peralejo, J. (2008). A longitudinal study of lexical development in
young children with autism spectrum disorders. Master's thesis,
University of British Columbia.

Peterson, C. C., Wellman, H. M., and Liu, D. (2005).

Steps in theoryof-
mind development for children with deafness or autism. Child
development, 76(2):502{517.

Plug, M. [Autistic Utterers as a Counterexample to Grice on Meaning]. PhD thesis, King's College London.

Reboul, A. (2006). HOT theories of meaning: The link between
language and theory of mind. Mind & Language, 21

Senju, A., Southgate, V., Miura, Y., Matsui, T., Hasegawa, T., Tojo,
Y., Osanai, H., and Csibra, G. (In Press). Absence of spontaneous
action anticipation by false belief attribution in children
with autism spectrum disorder. Development and Psychopathology.

Senju, A., Southgate, V., White, S., and Frith, U. (2009). Mindblind
eyes: An absence of spontaneous theory of mind in Asperger
Syndrome. Science, 325(5942):883{885.

Surian, L., Baron-Cohen, S., and Van der Lely, H. (1996). Are children
with autism deaf to Gricean maxims? Cognitive Neuropsychiatry,
1(1):55{72.

Surian, L. and Leslie, A. M. (1999). Competence and performance
in false belief understanding: A comparison of autistic and normal
3-year-old children. British Journal of Developmental Psychology,
17(1):141{155.

Tager-Flusberg, H. (1985). Basic level and superordinate level categorization
by autistic, mentally retarded and normal children. Journal
of Experimental Child Psychology, 40(3):450{469.

Tager-Flusberg, H. (1992). Autistic children's talk about psychological
states: De cits in the early acquisition of a theory of mind.
Child Development, 63(1):161{172.

Tager-Flusberg, H., Calkins, S., Nolin, T., Baumberger, T., Anderson,
M., and Chadwick-Dias, A. (1990). A longitudinal study
of language acquisition in autistic and down syndrome children.
Journal of Autism and Developmental Disorders, 20(1):1{21.

Tager-Flusberg, H. and Joseph, R. (2003). Identifying neurocognitive
phenotypes in autism. Philosophical Transactions of the Royal
Society of London - Series B: Biological Sciences, 358(1430):303{
314.

Tager-Flusberg, H., Paul, R., and Lord, C. (2005). Language and
communication in autism. In Volkmar, F. R., editor, Handbook
of autism and pervasive developmental disorders, volume 1: Diagnosis,
development, neurobiology, and behavior, pages 335{364.
John Wiley & Sons, Hoboken, 3rd edition.

Tager-Flusberg, H. and Sullivan, K. (1994). Predicting and explaining
behavior: A comparison of autistic, mentally retarded and
normal children. Journal of Child Psychology and Psychiatry,
35(6):1059{1075.

Toichi, M. and Kamio, Y. (2001). Verbal association for simple
common words in high-functioning autism. Journal of Autism
and Developmental Disorders, 31(5):483{490.

Ungerer, J. A. and Sigman, M. (1987). Categorization skills and
receptive language development in autistic children. Journal of
Autism and Developmental Disorders, 17(1):3{116.

Wimmer, H. and Perner, J. (1983). Beliefs about beliefs: Representation
and constraining function of wrong beliefs in young children's
understanding of deception. Cognition, 13(1):103{128.

Ziatas, K., Durkin, K., and Pratt, C. (1998). Belief term development
in children with autism, Asperger Syndrome, speci c language
impairment, and normal development: Links to theory of
mind development. Journal of Child Psychology & Psychiatry &
Allied Disciplines, 39(5):755{763.