Asp Metaforas

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Eur Arch Psychiatry Clin Neurosci (2013) 263 (Suppl 2):S177–S187

DOI 10.1007/s00406-013-0453-9

ORIGINAL PAPER

Automatic metaphor processing in adults with Asperger


syndrome: a metaphor interference effect task
Ismene Hermann • Verena Haser • Ludger Tebartz van
Elst • Dieter Ebert • Daniel Müller-Feldmeth •
Andreas Riedel • Lars Konieczny

Received: 31 August 2013 / Accepted: 16 September 2013 / Published online: 1 October 2013
Ó Springer-Verlag Berlin Heidelberg 2013

Abstract This paper investigates automatic processing of previous findings on automatic metaphor processing in
novel metaphors in adults with Asperger Syndrome (AS) typically developing individuals.
and typically developing controls. We present an experi-
ment combining a semantic judgment task and a recogni- Keywords Asperger syndrome  Nonliteral
tion task. Four types of sentences were compared: Literally language  Novel metaphors  Metaphor
true high-typical sentences, literally true low-typical sen- comprehension  Metaphor interference effect 
tences, apt metaphors, and scrambled metaphors (literally Autism spectrum disorders (ASD)  Pragmatic
false sentences which are not readily interpretable as language
metaphors). Participants were asked to make rapid deci-
sions about the literal truth of such sentences. The results
revealed that AS and control participants showed signifi- Introduction
cantly slower RTs for metaphors than for scrambled met-
aphors and made more mistakes in apt metaphoric Nonliteral language processing in ASD and AS
sentences than in scrambled metaphors. At the same time,
there was higher recognition of apt metaphors compared Nonliteral language comprehension in everyday commu-
with scrambled metaphors. The findings indicate intact nication does not seem to pose particular difficulties for
automatic metaphor processing in AS and replicate typical language users [1]. However, for individuals diag-
nosed with autism spectrum disorders (ASD), including
individuals with Asperger Syndrome (AS), understanding
nonliteral language is often reported to be challenging.
Ismene Hermann and Verena Haser have contributed equally to the
paper. Metaphors, irony, and idioms seem to cause problems both
in children [2–4] and adults [5–7] with ASD, though a
Andreas Riedel and Lars Konieczny are senior authors who have number of recent papers argue that deficits are not specific
contributed equally to the paper. to autism and that not all persons with ASD experience
difficulties (e.g., [8, 9]).
I. Hermann  L. T. van Elst  D. Ebert  A. Riedel (&)
Department of Psychiatry and Psychotherapy, University Although there is no language delay described in the
Medical Center Freiburg, Hauptstraße 5, 79104 Freiburg, diagnostic criteria of AS [10], the difficulties in nonliteral
Germany language comprehension have been described for AS as
e-mail: andreas.riedel@uniklinik-freiburg.de
well: Impairment of novel metaphor comprehension in AS
V. Haser has been observed in 10–41-year-old individuals with AS
Department of English Linguistics, University of Freiburg, [6]. Martin and McDonald [11] observed poor performance
Freiburg, Germany of young adults with AS in the interpretation of nonliteral
utterances or ironic jokes. Furthermore, neural processes
D. Müller-Feldmeth  L. Konieczny
Center for Cognitive Science, University of Freiburg, Freiburg, during semantic integration of metaphors and right hemi-
Germany sphere contribution to metaphor processing in individuals

123
S178 Eur Arch Psychiatry Clin Neurosci (2013) 263 (Suppl 2):S177–S187

with AS seem to be different from typically developing [20, 21] and the incompatibility of the initial interpretation
participants [12, 13]. with the surrounding context allow interpreters to recog-
Nonliteral language impairment in ASD is often con- nize that this literal interpretation is inadequate. This leads
strued as evidence of a more general deficit in pragmatic them to revise their initial interpretation in a final stage
language understanding, which in turn is frequently involving the construction of an alternative, nonliteral,
attributed to theory-of-mind (ToM) deficits [2, 14], weak interpretation [18].
central coherence (WCC) [15], and executive dysfunc- The traditional view of metaphor interpretation has been
tioning [16]. Pragmatic ability refers to the use and com- challenged by a number of experiments (see, e.g., [18, 19,
prehension of language in context, which involves 22]). For example, reading metaphors does not necessarily
parameters such as the linguistic context of utterances take longer than reading literal expressions, especially in
(previous utterances), social variables (e.g., the speakers’ cases where sufficient context is provided [23]. Glucksberg
relationship toward each other), the role of common ground et al. [24] showed that automatic processing of metaphor-
(background assumptions shared by interlocutors), the role ical meanings interferes with literal false decisions in a
of interlocutors’ intentions, and many further aspects of semantic judgment task. They concluded that metaphors
social context [17]. are processed even in situations where metaphor processing
A number of scholars offer arguments in favor of the is neither requested nor helpful. This experiment intro-
view that impairment of metaphor processing in AS duced the metaphor interference task (see Sect. ‘‘The
cannot exclusively be ascribed to pragmatic impairment in metaphor interference task’’) as a method for studying
the sense of deficient integration of contextual informa- automatic metaphor processing; the task was also used in
tion. Thus, Gold et al. [12] have compared reaction pat- several later studies [25–28].
terns for metaphoric and non-metaphoric expressions
using event-related potentials (ERP). The authors used The metaphor interference task
two-word expressions to examine metaphors regardless of
context. They found a different pattern of linguistic Glucksberg et al. [24] used a sentence-verification proce-
information processing between individuals with AS and dure, which we have adopted, with slight modifications, in
control subjects: In comparison with the control group, our experiment. They compared five types of sentences:
AS participants showed significantly larger N400s for (a) True high-typical sentences, e.g., Some birds are rob-
novel metaphors. For literal and unrelated word pairs, on ins. (b) True low-typical sentences, e.g., Some birds are
the other hand, no differences in N400s were found. The penguins. (c) Standard false sentences, e.g., Some birds are
authors conclude that in addition to the pragmatic deficits apples. (d) Metaphors, e.g., Some jobs are jails, Some flutes
described in AS, metaphor processing during the stage of are birds. Sentences of type (d) were literally false cate-
semantic integration is disrupted. Another study by Gold gory-membership statements, but they were readily inter-
and Faust [13], using the divided field paradigm to pretable if taken nonliterally. (e) Scrambled metaphors,
examine right hemisphere recruitment during novel met- e.g., Some jobs are birds, Some flutes are jails; these sen-
aphor processing, showed lack of right hemisphere tences were also literally false, but were not readily inter-
advantage in the AS group, supporting the hypothesis that pretable. The scrambled metaphors were produced by re-
the neural processes during metaphor processing in AS distributing topic and vehicle terms of the apt metaphors
are different from typically developing controls. These used (i.e., category (d)); in other words, two apt metaphors
studies are compatible with the idea of an impairment in Some jobs (topic) are jails (vehicle) and Some flutes (topic)
the early automatic stage of metaphor processing, which are birds (vehicle) yield the scrambled metaphors Some
might represent a further factor disrupting metaphor jobs are birds and Some flutes are jails).
interpretation, in addition to the above-mentioned prag- For each sentence, participants had to decide whether it
matic factors. was literally true or false. Glucksberg et al. [24] found that
participants had longer response latencies for those types of
Models of metaphor comprehension literally false statements that were easily interpretable as
metaphors (i.e., type (d) statements), when compared to
The so-called multi-stage model of metaphor interpretation those statements which did not intuitively make sense as
espoused by various philosophers and psychologists (see metaphors (i.e., type (e) statements). The authors ascribed
[18, 19] for overviews) postulates that literal meaning has this effect to the unintended processing of the metaphoric
priority over metaphorical meaning. Metaphor interpreta- meanings of type (d) sentences, which produced a conflict
tion is construed as a multi-stage process. In the first stage, in truth-value judgments: Though these statements are false
metaphors and other nonliteral expressions are interpreted when construed literally, they are true on a nonliteral
literally. The patent violation of conversational maxims reading. This conflict, called the metaphor interference

123
Eur Arch Psychiatry Clin Neurosci (2013) 263 (Suppl 2):S177–S187 S179

effect (MIE), slowed down the responses to metaphoric Table 1 Descriptive statistics for the Asperger Syndrome and control
items. Glucksberg et al. [24] also applied a recall task, participants: mean age; mean scores of the Raven’s Standard Pro-
gressive Matrices test (SPM) and the Multiple-Choice-Word-Test-B
which revealed a higher capacity to recall metaphoric items
(MWT-B)
compared with scrambled metaphors. The authors ascribed
this effect to a deeper processing of metaphors. These Mean (SD) p
experiments suggest that in some cases at least, the pro- Asperger Controls
cessing of metaphors is automatic and non-optional. Later
Age 42.4 (9.5) 35.1 (12.5) 0.045
publications by Glucksberg and others provide further
SPM 116.0 (11.7) 117.7 (9.5) 0.777
evidence for automatic metaphor processing, supporting a
MWT-B 123.1 (13.2) 123.1 (11.8) 0.920
model of metaphor comprehension that is opposed to the
traditional account. In Glucksberg’s model, metaphorical
and literal interpretations are processed in parallel, which
Method
contrasts with the traditional account according to which
the metaphorical meaning is accessed only if the literal
Participants
meaning is recognized as deficient (see [19] for an
overview).
Twenty persons diagnosed with AS (13 men, 7 women)
and 20 typically developing persons (11 men, 9 women)
The present study: do people with AS process participated in this experiment. All participants were right-
metaphors automatically? handed. There was no significant difference in gender
distribution between groups, v2 = 0.417, p = 0.748. Age
As described in ‘‘Nonliteral language processing in ASD ranged between 22 and 68 years. The t test revealed a
and AS’’, several studies report anomalous metaphor pro- slightly significant group difference in age (Table 1). Since
cessing in AS. Our hypothesis was that individuals diag- individual IQ-differences have been described to have an
nosed with AS do not process metaphors automatically and impact on the performance in semantic decision tasks [28],
hence will not display the MIE in a metaphor interference two IQ tests were carried out: the short version of Raven’s
task. Standard Progressive Matrices Test [29] and the Multiple-
Following Glucksberg et al. [24], in the present study we Choice-Word-Test-B [30]. AS and control groups did not
carried out a metaphor interference task in individuals with differ significantly in mean IQ (Table 1).
AS and healthy controls. Our main goal was to examine the At the Freiburg Center for Autism Research, the
MIE in both groups by comparing reaction times (RTs) and clinical diagnosis of autism spectrum disorders and AS is
error rates for sentence verification of (apt) metaphors and established as a consensus diagnosis of a multiprofes-
scrambled metaphors. Based on the results of Glucksberg sional team following the recommendations of the NICE
et al. [24] and subsequent studies, we expected a significant guidelines [31]. The clinical diagnosis includes a thor-
MIE in control participants. We expected poor automatic ough history of the patient following the above princi-
metaphor processing in AS—and thus no MIE. Error rates ples, a history of carers and behavioral observations in a
for metaphors should not substantially differ from error diagnostic process that usually takes several sessions of 2
rates for scrambled metaphors in the AS group, while in the or more hours. Psychometric tools obtained in a routine
control group, we expected higher error rates for apt met- procedure prior to clinical assessment comprise the Aut-
aphors than for scrambled metaphors. Furthermore, we did ism Spectrum Quotient (AQ) [32], the Empathy Quotient
not expect a higher memory for apt metaphors compared (EQ) [33], the Australian Scale for Asperger Syndrome
with scrambled metaphors in the AS group in a subsequent [34], the Social Responsiveness Scale [35], the Bermond-
recall task, while in the control group, we expected a Vorst Alexithymia Questionnaire [36], the Adult Asperger
replication of previous findings showing higher memory Assessment [37], the Wender-Utah-Rating-Scale [38], and
for apt metaphors compared with scrambled metaphors. the Beck Depression Inventory [39]. Additionally,
Overall, we did not expect participants with AS to process instruments like the Autism Diagnostic Interview-Revised
readily interpretable novel metaphors differently from (ADI-R) [40] and the Autism Diagnostic Observation
scrambled metaphors in an experimental design where the Schedule (ADOS) [41], as well as additional neuropsy-
potential occurrence of metaphoric sentences is not men- chological tests assessing executional and theory-of-mind
tioned and metaphor comprehension is not requested. To capacities, are obtained in selected and unclear cases.
the best of our knowledge, this is the first experiment to use The final consensus diagnosis is made by all persons
the metaphor interference task in German language. involved in the diagnostic process, which will invariably

123
S180 Eur Arch Psychiatry Clin Neurosci (2013) 263 (Suppl 2):S177–S187

include at least two experienced consultant psychiatrists scrambled metaphors) were rated for aptness and novelty,
or psychologists. based on a scale ranging from 1 (bad) to 7 (apt), and 1
In the present study, control participants completed AQ (familiar) to 7 (unfamiliar), respectively. All items that were
and EQ prior to the experimental session. In the AQ, all but used in the experiment had a mean rating of 6.0 or higher for
one participants with AS scored above 26, which has been novelty. We only used items as scrambled metaphors whose
suggested a suitable cut-off to distinguish general popula- mean rating did not exceed 2.0; we only used items as apt
tion from ASD population [42]. In the control group, all metaphors whose mean aptness rating averaged 6.0 or
participants scored below 26 in the AQ. The EQ scores higher. We used an analogous scheme for goodness (proto-
showed the reverse picture—high scores in the control typicality) ratings for the literally true sentences.
group (M = 42.6; SD = 12.3) and low scores in the AS The test items were distributed over eight lists by means
group (M = 18.9; SD = 12.0). of a Latin square rotation scheme: Topics and vehicles
All participants signed a statement of agreement prior to appeared only once in a list, and each list contained an
the experimental session. Control participants were paid for equal amount of items in each condition. Each list con-
their participation, AS participants received a compensa- sisted of 40 items. Every participant was presented with
tion for travel expenses. one list, and the lists were successively rotated between
participants: participant 1 was presented with list no. 1,
Apparatus and materials participant 2 was presented with list no. 2, etc. The order of
item presentation was randomized for each participant.
Stimuli
Passive recall task
The stimulus material consisted of 320 sentences of the
pattern ‘‘Some X are Y’’. Materials were constructed Participants performed a passive recall task. Therefore, we
according to a 2 9 2 design, comprising the experimental constructed four passive recall forms, each one consisting
factors Literalness and Goodness plus the between-subjects of two complementary test lists. The items were listed in
factor Group. The factor Literalness indicates whether a random order. Participants received the form which mat-
sentence is literally true or false. The factor Goodness ched their test lists.
indicates whether a literally true sentence is of high typi-
cality and thus easier to verify (high Goodness) or of low Self-assessment questionnaire
typicality (low Goodness). In the literally false condition, a
readily understandable metaphor would be of high good- Subsequent to the experiment, participants were asked to
ness and a scrambled metaphor of low Goodness (Table 2; answer some questions about their personal ability to
for more examples see section ‘‘Appendix’’). comprehend nonliteral language. For this purpose, we
Items were balanced between conditions according to constructed a questionnaire based on previous accounts of
word frequency. On average, word length did not differ common difficulties in the field of language and commu-
between apt metaphors and scrambled metaphors; scrambled nication in AS. This enabled us to collect information about
metaphors were constructed by re-pairing topic and vehicle the participants’ subjective perspective on their ability to
terms of different apt metaphors, yielding literally false interpret nonliteral language. Participants were presented
expressions that were difficult to interpret metaphorically. with several statements such as ‘‘To me, nonliteral language
Goodness and novelty ratings were completed by 10 is confusing’’ or ‘‘I have to reflect deeply in order to
undergraduate students who did not participate in the understand nonliteral language’’. Participants were sup-
experiment. Literally false sentences (i.e., metaphors and posed to state to what degree they agree with each statement
(I agree/I rather agree/I rather disagree/I disagree). In the
Table 2 Stimulus types (rough translations from German; predicate present article, results of this questionnaire will only be
noun phrases correspond to single-word expressions in German)
briefly touched upon, as the experimental groups were too
Goodness Literalness small to permit extensive questionnaire analysis.
True False
Procedure
Good Literally true, high typical Metaphors
Example Some lawyers are juvenile Some lawyers are scale Participants were tested individually. There were three
magistrates rulers
response keys available: One to the left hand index or
Bad Literally true, low typical Scrambled metaphors
middle finger, one to the right hand index or middle finger,
Example Some lawyers are tax Some lawyers are days and one to the right hand thumb: the latter initiated the
advisors of spring
session. Of the other two response keys, the left one was

123
Eur Arch Psychiatry Clin Neurosci (2013) 263 (Suppl 2):S177–S187 S181

designated ‘‘true’’ and the right one was designated that there was no difference concerning the MIE between
‘‘false’’. groups (t = 1.12, SE = 0.045). As three-way interactions
The participants were given a brief instruction on the are difficult to interpret, we conducted a separate analysis
screen. They were told that they would be presented with of literally false sentences: In literally false sentences, the
various sentences and that their task was to decide as main effect of Goodness indicates the difference in RTs for
quickly and accurately as possible whether each sentence metaphors and scrambled metaphors. In line with the
was literally true or false and that they would have to press metaphor interference hypothesis, the main effect of
the left or right key depending on their decision. The Goodness was highly significant in literally false sentences
sentences were presented word by word via rapid serial (t = 7.6; SE = 0.02), irrespective of Group (Good-
visual presentation. 50 ms per letter were added to the ness 9 Group: t = -0.6; SE = 0.03).
basic presentation time of 500 ms per word. At the As we found a significant group difference in mean age,
beginning of each trial, a fixation cross appeared in the we added the factor Age to the linear mixed model in
middle of the screen and persisted for 1,000 ms. Prior to another analysis. Results show that there was no significant
the presentation of the 40 test items, participants were main effect of Age (t = 0.52; SE = 0.01) and the previ-
presented with 20 practice items so that they could get used ously obtained main effects remained the same.
to the procedure. In the practice session, participants were For the analysis of error rates, we conducted a logistic
given feedback. A time-out after 2,000 ms was imple- regression model with the dichotomized response variable, the
mented to ensure that participants would answer sponta- fixed factors Literalness, Goodness, and Group and the ran-
neously. The unannounced passive recall task was carried dom factors Item and Participant. Figure 2 shows the pro-
out 15 min after the completion of the semantic judgment portion of correct answers for group and condition. There was
task. a significant effect of Literalness due to the very low error rates
in the scrambled condition (z value = -2.404,
Pr([|z|) = 0.016). A highly significant interaction of Literal-
Results ness 9 Goodness (z value = 3.790, Pr([|z|) \ 0.001) shows
that, in both groups, fewer mistakes were made in the
Reaction times and error rates scrambled condition and the high-typical condition than in the
(apt) metaphor and low-typical condition. There was no sig-
RTs were log-transformed and outliers were removed on nificant group effect (z value = -1.229, Pr([|z|) = 0.219) or
the basis of a linear mixed model (see below) prior to the three-way interaction of Literalness 9 Goodness 9 Group
analysis. Data points which exceeded the twofold standard (z value = -0.244, Pr([|z|) = 0.807).
error of the estimated RTs for each participant and condi-
tion were not included in the analysis. 1.79 % of all cases Passive recall task and self-assessment
were removed as outliers. We fitted a linear mixed model
with the fixed factors Literalness, Goodness, and Group Recall analysis was performed using a logistic regression
and the random factors Participant and Item on our RT model. In the literally false conditions, fewer items were
data. In the analysis of RTs, only correct answers were memorized than in the literally true conditions
included. (z value = 8.961, Pr([|z|) \ 0.001) (Fig. 3). This effect
Figure 1 illustrates mean RTs for both groups in the seems to arise from poor memory for scrambled metaphors.
various conditions. We found no significant main effects of This is confirmed by the highly significant effect of
Literalness (t = -0.01; SE = 0.011), Goodness (t = Goodness (z value = 3.623, Pr([|z|) \ 0.001) which varies
-0.46; SE = 0.011) or Group (t = -1.31, SE = 0.069). as a function of Literalness (Literalness 9 Goodness: z
As illustrated by Fig. 1, there was a significant interaction value = -4.189, Pr([|z|) \ 0.001). There was no sub-
of Literalness and Goodness (t = -10.50, SE = 0.023): In stantial group effect (z value = 1.690, Pr([|z|) = 0.091)
literally false sentences, Goodness is associated with and no significant three-way interaction of Literal-
slower RTs, contrary to its effect in literally true sentences. ness 9 Goodness 9 Group (z value = -1.091,
This result for literally false sentences is crucial and Pr([|z|) = 0.275), which suggests no difference in recall of
expected, given the metaphor interference hypothesis. We metaphors and scrambled metaphors in AS and controls
also observed a significant interaction of Literalness and after the semantic judgment task.
Group (t = 2.15, SE = 0.023) due to longer RTs in liter- As announced in Sect. ‘‘Self-assessment questionnaire,’’
ally false sentences in the AS group. There was no three- the subject of self-assessment will only briefly be broached
way interaction of Literalness 9 Goodness 9 Group, at this point. Merely two examples will be brought up to
which, in line with the impression given by Fig. 1, suggests show that our participants indeed differed in terms of their

123
S182 Eur Arch Psychiatry Clin Neurosci (2013) 263 (Suppl 2):S177–S187

Fig. 1 RTs as a function of


Goodness (good/bad) 9 Group
(AS/Control (C)) for literally
true and literally false sentences
(error bars ± 2 SE)

subjective assessment of their nonliteral language ability: within an experimental setting with relatively impover-
In the AS group, 13 of 20 participants stated that they were ished context information. In terms of automatic metaphor
confused by figurative utterances made in conversations. comprehension, participants diagnosed with AS, who are
Among controls, only 1 of 20 participants agreed with this often reported to have severe difficulties understanding
statement. On the other hand, only 3 of 20 AS participants nonliteral meanings (e.g. [2, 5–7]), did not differ from
stated that they understood figurative expressions intui- healthy controls in our study: For both groups, response
tively, whereas 16 of 20 controls agreed with this latencies for rejecting apt metaphorically true statements
statement. were significantly longer compared with response latencies
for statements that did not intuitively make sense (scram-
bled metaphors). Furthermore, both experimental groups
Discussion made more mistakes in assessing apt metaphors as literally
false compared with scrambled metaphors. This finding
This experiment was the first to examine whether indi- supports the metaphor interference account, indicating that
viduals diagnosed with AS show a metaphor interference a semantic decision is more difficult in (apt) metaphoric
effect, which is seen as an indicator for automatic metaphor sentences, since the metaphorically true interpretations of
processing. Somewhat surprisingly given previous these expressions interfere with their literal falsity. In line
research, the results revealed no difference in metaphor with previous research mentioned above, we thus interpret
interference between AS and control participants. This slower response times for apt as opposed to scrambled
indicates that the early automatic stage of metaphor pro- metaphors as indicating participants’ difficulties suppress-
cessing is unimpaired in AS. ing automatic activation of metaphorical meanings that
RT analysis showed that both groups took longer to make intuitive sense. Another relevant result is that AS
judge literally false but metaphorically true sentences (apt participants and typically developing controls were both
metaphors) than scrambled metaphors (literally false sen- able to remember more metaphors than scrambled
tences which do not comprise a readily understandable metaphors.
meaning). In line with Glucksberg et al. [24], we found These findings contrast with the self-assessments we
evidence for automatic metaphor comprehension, even collected subsequent to the experiment. In line with much

123
Eur Arch Psychiatry Clin Neurosci (2013) 263 (Suppl 2):S177–S187 S183

Fig. 2 Proportion of correct


answers as a function of
Goodness (good/bad) 9 Group
(AS/Control (C)) for literally
true and literally false sentences
(error bars ± 2 SE)

of the previously cited research, these self-assessments processing in AS may be less impaired than expected.
suggest that adults diagnosed with AS do, at least subjec- There is some independent support for this argument. For
tively, experience difficulties interpreting nonliteral one thing, as already mentioned, previous findings in this
language. area of research are inconsistent. In fact, there is growing
Taken together, these results indicate that despite the evidence for what can be interpreted as (at least partially)
frequently reported nonliteral language impairment of intact nonliteral language comprehension in individuals
individuals with AS, automatic processing of novel meta- with AS. For instance, Giora et al. [43] found that young
phoric expressions seems to be intact. adults with AS are not biased toward the literal in a
Thus, our hypothesis that metaphor processing in adults meaningfulness decision task comprising literal and meta-
with AS would be less automatic than in typically devel- phorical meanings. The findings of Pijnacker et al. [44]
oping individuals could not be confirmed. indicate that children with ASD are not impaired in
Our data are limited by the fact that there was a sig- deriving scalar implicature, with a potential differentiation
nificant group effect in age, which, however, did not turn between AS and autistic disorder: participants with AS
out to have a significant effect in the analysis of RTs. Note performed better than participants with high-functioning
as well that there was a significant interaction of Literal- autism. Also, Gold and Faust [13], for example, did not
ness and Group in that AS participants took longer to find a difference in the behavioral data (reaction times and
assess literally false sentences in general. This is another error rates) for young adults with AS and typically devel-
interesting finding, though the key result for our purposes is oping participants in a semantic judgment task involving
that AS participants took significantly longer to assess apt literal and metaphorical meanings. Gold et al. [12] repli-
metaphors than scrambled metaphors and hence show a cated this finding. In their review article, Kasari and
metaphor interference effect (regardless of the fact that Rotheram-Fuller [45] observe that AS and high-functioning
overall response times for literally false sentences are autism are frequently either grouped together in terms of
slower than for controls). ASD or else only HFA participants are examined. The
The presence of the MIE in individuals with AS could above-mentioned study by Pijnacker et al. [44] shows that
be interpreted in terms of evidence that metaphor HFA and AS might differ in their nonliteral language

123
S184 Eur Arch Psychiatry Clin Neurosci (2013) 263 (Suppl 2):S177–S187

Fig. 3 Proportion of
recognized items as a function
of Goodness (good/
bad) 9 Group (AS/Control (C))
for literally true and literally
false sentences (error bars ± 2
SE)

ability. This has implications for how to interpret the per- be more idiosyncratic than interpretations proposed by
tinent literature, as studies on language processing in pure controls [46]. Hence, one avenue for future research is to
AS populations are relatively rare. Taken together, these investigate in even greater detail and with regard to many
findings seem to indicate that metaphor processing espe- different linguistic types of metaphors whether individuals
cially in AS is less impaired than often assumed. with AS derive interpretations of metaphors that are com-
Setting aside the previously mentioned important debate parable to those of typically developing controls.
concerning the nature and extent of nonliteral language Another aspect that might contribute to impaired meta-
impairments in (at least some) persons with AS, our study phor processing in some persons with AS is reported dif-
suggests that possible differences in metaphor processing ficulties in ignoring irrelevant information. According to
between individuals with AS and controls do not affect the inhibitory deficit hypothesis, inability to suppress
early automatic stages of metaphor processing. Differences irrelevant information during metaphor processing results
in metaphor processing might still be found with regard to in greater difficulties in metaphor comprehension.
the quality of figurative interpretations or the extent to Gernsbacher et al. [47] conducted a semantic decision task
which linguistic or social context is accessed for interpre- involving semantic priming and found that ‘‘interpreting a
tation. But in light of previously mentioned research metaphor involves both enhancing the attributes that are
summarized in [8], these possibilities should be considered relevant to the metaphorical interpretation and suppressing
with some caution. the attributes that are not relevant to the metaphorical
We did not ask participants to explain their understanding interpretation.’’ Older people seem to suppress irrelevant
of the expressions. As observed by Gold and Faust [12], it information less effectively than younger ones [48]. In an
might be the case that participants with AS do correctly judge fMRI study on the neural basis of a simple response inhi-
an expression as metaphoric, but still do not properly bition task and an inhibition task involving working
understand it, arriving at interpretations which are at odds memory, Kana et al. [49] found atypical and less syn-
with the speaker/writer’s original intentions. It might be chronized inhibitory function in autism, while there were
objected that there is generally more than one way of inter- no differences at the behavioral level. They suppose that in
preting (especially novel) metaphors, yet metaphor inter- autism as compared with controls, inhibitory processes are
pretations suggested by persons with AS have been argued to less automatic and more strategic in nature. Therefore,

123
Eur Arch Psychiatry Clin Neurosci (2013) 263 (Suppl 2):S177–S187 S185

complex inhibitory processes such as the suppression of due to Prof. Sam Glucksberg for sending us his materials for inspi-
irrelevant information could play a role in impaired figu- ration and to Thomas Fangmeier for his expert advice in neuropsy-
chological testing and his assistance in recruiting participants. This
rative language processing in (some persons with) AS. research received no specific grant from any funding agency in the
An alternative explanation for AS’ reported impairment public, commercial, or not-for-profit sectors.
of metaphor processing in everyday life could relate to
deficits with regard to the adequate use of contextual Conflict of interest The authors declare that they have no conflict
of interest.
information in order to find the most appropriate meaning
of a metaphoric expression. A number of studies are con- This article is part of the supplement ‘‘Bridging the gap between
sistent with the possibility that linguistic context is pro- Neurobiology and Psychosocial Medicine’’. This supplement was not
cessed in an unusual way by individuals with ASD (see sponsored by outside commercial interests. It was funded by the
German Association for Psychiatry and Psychotherapy (DGPPN).
[50] for an overview), though again it is unclear whether
autism per se is responsible for impairments in contextual
integration (e.g., [8]). In the present study, we examined
metaphor processing based on relatively impoverished Appendix
stimuli, providing no linguistic or extralinguistic context
for our stimuli sentences. Both groups show evidence of See Table 3.
automatic metaphor processing in this setting. However,
metaphors in ordinary language can only be fully under- Table 3 Examples of the four types of sentences
stood by integrating the expressions into a larger context. It Literally true, Literally true, Metaphors Scrambled
might be interesting to modify the experiment in such a high-typical low-typical metaphors
way as to require participants to use contextual information sentences sentences
for processing the sentences.
Some lawyers are Some lawyers Some lawyers Some lawyers
This study was motivated by the aim to replicate the juvenile are tax are scale are days of
MIE in the German language and in adults with AS. In the magistrates advisors rulers spring
original MI-task by Glucksberg et al. [24], only behavioral Some throats are Some throats Some throats Some throats
data were assessed. More recent studies on metaphor pro- human jaws are predator are sewer are scale
cessing demonstrated a discrepancy between behavioral jaws tunnels rulers
and neural data (e.g. [7, 12]). For example, evidence from Some Some Some Some
confessions are confessions confessions confessions
an ERP study in which participants were supposed to judge accounts of are are deep sea are coma vigil
the meaningfulness of metaphoric and literal expressions crimes statements voyages (plural in
suggests that metaphor processing in AS might be achieved of faith German)
in the same time but not with the same efficiency or ease as Some types of Some types of Some types of Some types of
in typically developing individuals [12]. This has impli- support (single support are support are support are
word in non-cash days of hummingbirds
cations for the interpretability of behavioral data: Even if German) are benefits spring
RTs and error rates are identical, there still might be a cash offerings
covert group difference, which can only be revealed by the
Note that the sentences were translated from German to English.
analysis of neurophysiological data. Given these facts as Predicate noun phrases correspond to single-word expressions in
well as the limitations of basically every psycholinguistic German
method, it seems important to complement our findings and
methodology by future studies addressing these issues from
a neurophysiological perspective.
In summary, the present study demonstrates a striking
similarity in automatic metaphor processing in AS and References
control individuals. Neither RTs nor error rates indicate
1. Gentner D, Bowdle B (2001) Convention, form, and figurative
group differences in this task. Although a sizeable number language processing. Metaphor Symb 16:223–247. doi:10.1207/
of previous studies on this topic—as well as self-reports of S15327868MS1603&4_6
individuals diagnosed with AS—suggest that persons with 2. Dennis M, Lazenby AL, Lockyer L (2001) Inferential language in
AS often experience difficulties in understanding novel high-function children with autism. J Autism Dev Disord
31:47–54. doi:10.1023/A:1005661613288
metaphors, at least this early stage of metaphor processing 3. Rundblad G, Annaz D (2010) The atypical development of
seems to be unimpaired. metaphor and metonymy comprehension in children with autism.
Autism 14:29–46. doi:10.1177/1362361309340667
Acknowledgments We wish to thank all participants who were 4. Melogno S, D’Ardia C, Pinto MA, Levi G (2012) Explaining
willing to donate their time and effort to this study. Special thanks are metaphors in high-functioning autism spectrum disorder children:

123
S186 Eur Arch Psychiatry Clin Neurosci (2013) 263 (Suppl 2):S177–S187

a brief report. Res Autism Spectr Disord 6:683–689. doi:10.1016/ 24. Glucksberg S, Gildea P, Bookin H (1982) On understanding
j.rasd.2011.09.005 nonliteral speech: can people ignore metaphors? J Verbal
5. Ozonoff S, Miller JN (1996) An exploration of right-hemisphere Learning Verbal Behav 21:85–98. doi:10.1016/S0022-5371
contributions to the pragmatic impairments of autism. Brain Lang (82)90467-4
52:411–434. doi:10.1006/brln.1996.0022 25. Keysar B (1989) On the functional equivalence of literal and
6. Gunter HL, Ghaziuddin M, Ellis HD (2002) Asperger syndrome: metaphorical interpretations in discourse. J Mem Lang
tests of right hemisphere functioning and interhemispheric com- 28:375–385. doi:10.1016/0749-596X(89)90017-X
munication. J Autism Dev Disord 32:263–281. doi:10.1023/A: 26. Pierce RS, Maclaren R, Chiappe DL (2010) The role of working
1016326701439 memory in the metaphor interference effect. Psychon Bull Rev
7. Strandburg RJ, Marsh JT, Brown WS, Asarnow RF, Guthrie D, 17:400–404. doi:10.3758/PBR.17.3.400
Higa J (1993) Event-related potentials in high-functioning adult 27. Wolff P, Gentner D (2000) Evidence for role-neutral initial
autistics: linguistic and nonlinguistic visual information pro- processing of metaphors. J Exp Psychol Learn Mem Cogn
cessing tasks. Neuropsychologia 31:413–434. doi:10.1016/0028- 26:529–541
3932(93)90058-8 28. Kazmerski VA, Blasko DG, Dessalegn BG (2003) ERP and
8. Gernsbacher MA, Pripas-Kapit SR (2012) Who’s missing the behavioral evidence of individual differences in metaphor com-
point? A commentary on claims that autistic persons have a prehension. Mem Cognit 31:673–689. doi:10.3758/BF03196107
specific deficit in figurative language comprehension. Metaphor 29. Raven JC (1988) Standard progressive matrices. Beltz, Weinheim
Symb 27:93–105. doi:10.1080/10926488.2012.656255 30. Lehrl S, Triebig G, Fischer B (1995) Multiple choice vocabulary
9. Norbury CF (2005) The relationship between theory of mind and test MWT as a valid and short test to estimate premorbid intel-
metaphor: evidence from children with language impairment and ligence. Acta Neurol Scand 91:335–345. doi:10.1111/j.1600-
autistic spectrum disorder. Br J Dev Psychol 23:383–399. doi:10. 0404.1995.tb07018.x
1348/026151005X26732 31. Pilling S, Baron-Cohen S, Megnin-Viggars O, Lee R, Taylor C
10. American Psychiatric Association (1994) Diagnostic and statis- (2012) Recognition, referral, diagnosis, and management of
tical manual of mental disorders, 4th Edn, Text Revision (DSM- adults with autism: summary of NICE guidance. BMJ 344:e4082.
IV-TR) doi:10.1136/bmj.e4082
11. Martin I, McDonald S (2004) An exploration of causes of non- 32. Baron-Cohen S, Wheelwright S, Skinner R, Martin J, Clubley E
literal language problems in individuals with Asperger syndrome. (2001) The autism-spectrum quotient (AQ): evidence from As-
J Autism Dev Disord 34:311–328. doi:10.1023/B:JADD. perger syndrome/high-functioning autism, males and females,
0000029553.52889.15 scientists and mathematicians. J Autism Dev Disord 31:5–17.
12. Gold R, Faust M, Goldstein A (2010) Semantic integration during doi:10.1023/A:1005653411471
metaphor comprehension in asperger syndrome. Brain Lang 33. Baron-Cohen S, Wheelwright S (2004) The empathy quotient: an
113:124–134. doi:10.1016/j.bandl.2010.03.002 investigation of adults with Asperger syndrome or high func-
13. Gold R, Faust M (2010) Right hemisphere dysfunction and tioning autism, and normal sex differences. J Autism Dev Disord
metaphor comprehension in young adults with asperger syn- 34:163–175. doi:10.1023/B:JADD.0000022607.19833.00
drome. J Autism Dev Disord 40:800–811. doi:10.1007/s10803- 34. Garnett M, Attwood T (1998) The Australian scale for asperger’s
009-0930-1 syndrome. In: Attwood T (ed) Asperger’s syndrome: a guide for
14. Happé FG (1993) Communicative competence and theory of parents and professionals. Jessica Kingsley, London, pp 16–20
mind in autism: a test of relevance theory. Cognition 48:101–119. 35. Constantino JN, Davis SA, Todd RD, Schindler MK, Gross MM,
doi:10.1016/0010-0277(93)90026-R Brophy SL, Metzger LM, Shoushtari CS, Splinter R, Reich W
15. Jolliffe T, Baron-Cohen S (1999) A test of central coherence (2003) Validation of a brief quantitative measure of autistic traits:
theory: linguistic processing in high-functioning adults with comparison of the social responsiveness scale with the autism
autism or Asperger syndrome: is local coherence impaired? diagnostic interview-revised. J Autism Dev Disord 33:427–433.
Cognition 71:149–185. doi:10.1016/S0010-0277(99)00022-0 doi:10.1023/A:1025014929212
16. Lam YG, Yeung SSS (2012) Towards a convergent account of 36. Vorst HC, Bermond B (2001) Validity and reliability of the
pragmatic language deficits in children with high-functioning aut- bermond–vorst alexithymia questionnaire. Pers Individ Dif
ism: depicting the phenotype using the pragmatic rating scale. Res 30:413–434. doi:10.1016/S0191-8869(00)00033-7
Autism Spectr Disord 6:792–797. doi:10.1016/j.rasd.2011.08.004 37. Baron-Cohen S, Wheelwright S, Robinson J, Woodbury-Smith M
17. Allan K, Jaszczolt KM (2012) The Cambridge handbook of (2005) The adult asperger assessment (AAA): a diagnostic
pragmatics. Cambridge University Press, Cambridge method. J Autism Dev Disord 35:807–819. doi:10.1007/s10803-
18. Gibbs RW, Colston H (2006) Figurative language. In: Traxler M, 005-0026-5
Gernsbacher MA (eds) Handbook of psycholinguistics, 2nd edn. 38. Retz-Junginger P, Retz W, Blocher D, Weijers HG, Trott GE,
Elsevier, Amsterdam, pp 835–860 Wender PH, Rössler M (2002) Wender Utah rating scale. The
19. Glucksberg S (2008) How metaphors create categories: quickly. short-version for the assessment of the attention-deficit hyper-
In: Gibbs RW (ed) The Cambridge handbook of metaphor and activity disorder in adults. Nervenarzt 73:830–838. doi:10.1007/
thought. Cambridge University Press, Cambridge, pp 67–83 s00115-001-1215-x
20. Grice HP (1989) Studies in the way of words. Harvard University 39. Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J (1961) An
Press, Cambridge inventory for measuring depression. Arch Gen Psychiatry
21. Searle JR (1993 [1979]) Metaphor. In: Ortony A (ed) Metaphor 4:561–571. doi:10.1001/archpsyc.1961.01710120031004
and thought, 2nd edn. Cambridge University Press, Cambridge, 40. Lord C, Rutter M, Le Couteur A (1994) Autism diagnostic
pp 83–111 interview-revised: a revised version of a diagnostic interview for
22. Glucksberg S (2001) Understanding figurative language: from caregivers of individuals with possible pervasive developmental
metaphor to idioms (Oxford Psychology Series). Oxford Uni- disorders. J Autism Dev Disord 24:659–685. doi:10.1007/
versity Press, New York BF02172145
23. Inhoff AW, Lima SD, Carroll PJ (1984) Contextual effects on 41. Lord C, Risi S, Lambrecht L, Cook EH, Leventhal BL, DiLavore
metaphor comprehension in reading. Mem Cognit 12:558–567. PC, Pickles A, Rutter M (2000) The autism diagnostic observa-
doi:10.3758/BF03213344 tion schedule-generic: a standard measure of social and

123
Eur Arch Psychiatry Clin Neurosci (2013) 263 (Suppl 2):S177–S187 S187

communication deficits associated with the spectrum of autism. 46. MacKay G, Shaw A (2004) A comparative study of figurative
J Autism Dev Disord 30:205–223. doi:10.1023/A:1005592 language in children with autistic spectrum disorders. Child Lang
401947 Teach Ther 20:13–32. doi:10.1191/0265659004ct261oa
42. Woodbury-Smith MR, Robinson J, Wheelwright S, Baron-Cohen 47. Gernsbacher MA, Keysar B, Robertson RRW, Werner NK (2001)
S (2005) Screening adults for Asperger syndrome using the AQ: a The role of suppression and enhancement in understanding
preliminary study of its diagnostic validity in clinical practice. metaphors. J Mem Lang 45:433–450. doi:10.1006/jmla2000.2782
J Autism Dev Disord 35:331–335. doi:10.1007/s10803-005- 48. Newsome MR, Glucksberg S (2002) Older adults filter irrelevant
3300-7 information during metaphor comprehension. Exp Aging Res
43. Giora R, Gazal O, Goldstein I, Fein O, Stringaris A (2012) Sal- 28:253–267. doi:10.1080/03610730290080317
ience and context: interpretation of metaphorical and literal lan- 49. Kana RK, Keller TA, Minshew NJ, Just MA (2007) Inhibitory
guage by young adults diagnosed with asperger’s syndrome. control in high-functioning autism: decreased activation and
Metaphor Symb 27:22–54. doi:10.1080/10926488.2012.638823 under connectivity in inhibition networks. Biol Psychiatry
44. Pijnacker J, Hagoort P, Buitelaar J, Teunisse J-P, Geurts B (2009) 62:198–206. doi:10.1016/j.biopsych.2006.08.004
Pragmatic inferences in high-functioning adults with autism and 50. Braeutigam S, Swithenby SJ, Bailey AJ (2008) Contextual inte-
asperger syndrome. J Autism Dev Disord 39:607–618. doi:10. gration the unusual way: a magnetoencephalographic study of
1007/s10803-008-0661-8 responses to semantic violation in individuals with autism spec-
45. Kasari C, Rotheram-Fuller E (2005) Current trends in psycho- trum disorders. Eur J Neurosci 27:1026–1036. doi:10.1111/j.
logical research on children with high-functioning autism and 1460-9568.2008.06064.x
Asperger disorder. Curr Opin Psychiatry 18:497–501. doi:10.
1097/01.yco.0000179486.47144.61

123

You might also like