Recent empirical findings suggest a significant influence of emotion on memory processes. Surprisingly, although emotion-processing difficulties appear to be a hallmark feature in autism spectrum disorders (ASD), their impact on higher-level cognitive functions, such as memory, has not been directly studied in this population. The aim of this study was to address this issue by assessing whether the emotional valence of visual scenes affects recall skills in high-functioning individuals with ASD. To this purpose, their recall performance of neutral and emotional pictures was compared with that of typically developing adults (control group). Results revealed that while typically developing individuals showed enhanced recall skills for negative relative to positive and neutral pictures, individuals with ASD recalled the neutral pictures as well as the emotional ones. Findings of this study thus point to reduced influence of emotion on memory processes in ASD than in typically developing individuals, possibly owing to amygdala dysfunctions.
Besides impairments in communication and social cognition, coexisting with stereotypic behaviors [Diagnostic and Statistical Manual of Mental Disorders—Fourth Edition (DSM-IV); APA, 1994], individuals with autism spectrum disorders (ASD) also show persistent difficulties in emotional information processing [e.g., Boraston, Blakemore, Chilvers & Skuse, 2007]. Such difficulties have been found in a wide range of tasks, including matching and labelling facial expressions of emotion [e.g., Hobson, 1993; Tantam, Monaghan, Nicholson, & Stirling, 1989, respectively], matching gestures, postures, context and facial expressions on the basis of emotion [e.g., Hobson, Ouston, & Lee, 1988], interpreting emotional gestures [Hubert et al., 2007; Moore, Hobson, & Lee, 1997], and matching expressions of emotion to verbal and pictorial labels [e.g., Lindner & Rosen, 2006]. However, the impact of these emotion-processing deficits on overall cognitive functioning in ASD remains elusive.
In typical development, several behavioral and neuroimaging studies have shown that emotion can have a significant impact on memory [e.g., Kensinger & Corkin, 2003; Ochsner, 2000; Sharot, Verfaellie, & Yonelinas, 2007]. Although memory is a complex system including several components (e.g., short-term working memory, long-term procedural and semantic memory), most of the effects of emotion on memory that have been reported in the literature concern episodic memory. There is indeed considerable evidence that emotionally colored events are more likely to be recalled than neutral events [e.g., Kensinger, Garoff-Eaton, & Schacter, 2007] and that emotional stimuli are recalled more accurately than neutral stimuli [e.g., Cahill, 2000; Cahill et al., 1996]. For instance, several studies have shown that information with a negative emotional valence enhances word- [e.g., Kensinger & Corkin, 2003; Kensinger et al., 2007; Medford et al., 2005] and visual-scene recognition [Anderson, Yamaguchi, Grabski, & Lacka, 2006]. Although several factors can influence how well a stimulus is recalled, it has been argued that emotion can play an important role on memory by influencing the amount of attention allocated to the stimulus during encoding [Fox, Russo, Bowles, & Dutton, 2001; Ohman, Flykt, & Esteves, 2001]. Interestingly this interpretation holds not only for episodic but also for short-term memory tasks [e.g., Reinecke, Rinck, & Becker, 2006].
Given the well-known poor interest in emotional information in individuals with ASD, the question of whether they, in turn, fail to show this facilitative impact of emotion on memory is of major interest. Though few studies have focused on this question, the existing results suggest that emotional information has a reduced influence on cognition in individuals with ASD relative to typical controls. For instance, Beversdorf et al.  have shown that recall skills are less enhanced by the presence of emotional content in high-functioning individuals with ASD than in typically developing individuals. More recently, Kamio, Wolf, and Fein  have shown that individuals with ASD fail to show affective priming effects in both subliminal and supraliminal exposure conditions, also indicating an absence of impact of the emotional content of stimuli on cognitive processes in ASD.
Here, we propose to directly test the hypothesis of whether the emotional content of a stimulus has a reduced impact on memory skills in adults with ASD. More precisely, we hypothesize that recall skills in this group will be impervious to the emotional content of stimuli, in contrast to that of typically developing individuals for which negative stimuli should enhance recall performance [Cahill et al., 1996]. We thus used a memory task pinpointing the recall of emotionally laden (negative or positive) or neutral visual scenes.
Two experimental groups participated in this study. The first group included 15 adults with ASD (12 males) aged from 17 to 55 years (M=29, SD=11.5). These participants were diagnosed by a psychiatrist according to the DSM-IV [APA, 1994] and/or the Autism Diagnostic Interview—Revised [Lord, Rutter, & Le Couteur, 1994] criteria for autism. In order to further confirm the diagnosis, parents of all the subjects were asked to answer a screening questionnaire for ASD [Ehlers, Gillberg, & Wing, 1999] during a semi-directive interview. Mean IQ scores, assessed with the Wechsler  Adult Intelligence Scale—III, ranged from 79 to 150 (M=106, SD=24). All participants were found to have a level of functioning corresponding to their chronological age, as reflected by their level of education and daily autonomy (total or partial).
As mean IQ scores for the ASD group were within the normal range, participants with ASD were individually matched to typically developing individuals (N=15) on the basis of gender (12 males), chronological age (M=25.4, SD=3.4 years), and socio-educational status. None of the control participants had history of physical, psychiatric, or neurological disorders, as confirmed by the same psychiatrist that diagnosed the ASD participants.
At the time of testing, none of the participants in this study had known associated medical disorders and visual examination was found to be normal for all participants.
Informed consent was obtained for all subjects before participation.
Stimuli used in this study consisted of colored pictures taken from the International Affective Picture System (IAPS) database [Lang, Bradley, & Cuthbert, 2005]. A total of 486 images were selected according to the emotional valence and arousal rating criteria of the IAPS. These pictures were either negative (mean of valence=3.08, SD=0.56), positive (mean of valence=7.01, SD=0.58), or neutral (mean of valence=5.05, SD=0.63). The same number of pictures was used in each condition (N=162). Pictures were matched across condition in terms of arousal (mean arousal=4.7, SD=0.9; see Table I), complexity, luminance, and the number of characters displayed. Images subtended approximately 18° × 18° of visual angle at a viewing distance of 60 cm.
Table I. Mean and Standard Deviation (SD) Scores for Valence and Arousal of Pictures Used in Each Condition
Note that scores for each picture were taken from the IAPS norms. IAPS, International Affective Picture System database.
Pictures were divided in two sets of 324 photographs each. The “encoding” set included 108 negative, 108 neutral, and 108 positive pictures. The “recalling” set included 162 “new” images (i.e., not included on the “encoding” set; 54 negative, 54 neutral, and 54 positive) and 162 “old” images taken from the first “encoding” set (54 negative, 54 neutral, and 54 positive).
The experiment was divided into three blocks, each containing an encoding and a recalling phase. During the encoding phase, 18 images (6 positive, 6 neutral, and 6 negative) were presented successively (one per second) for 150 msec at the center of the screen. Then the word “recognition” appeared for 5 sec at the center of the screen to indicate the beginning of the recalling phase. Following this, nine new images and nine images already seen during the encoding phase were presented successively for 1 sec at the center of the screen, separated by a blank screen of 500 msec. Participants were asked to decide whether they had seen the picture during the encoding phase and to press the left button of a computer mouse if their response was positive (“yes”).
Responses were classified as “hits” (correct responses) when participants pressed the left mouse button after seeing an image presented during the encoding phase. Responses were classified as “false alarms” when participants pressed the left button in response to an image that had not been presented in the encoding phase. The number of false alarms was subtracted from the number of hits, resulting in a single accuracy score (AC) for each participant in each condition. Mean AC were first analyzed using a two-way analysis of variance (ANOVA), including group (ASD/control) as a between-subjects factor and valence (positive/negative/neutral) as within-group factor. We also conducted one-way ANOVAs in order to determine the main effect of valence for each group separately. Tukey tests were used for all post hoc comparisons.
In order to determine whether age exerted an influence on AC for each condition, we conducted correlation analyses using Pearson r tests for each group separately. For the ASD group, similar analyses were conducted using IQ as a factor.
Results of the two-way ANOVA revealed that neither the main effect of GROUP nor of valence reached significance (both Ps>0.05). However, the group × valence interaction was found significant (F(2, 56)=3.02, P<0.05; see Fig. 1).
In order to further track these results we conducted one-way-ANOVAs for the control and the ASD groups separately. Results for the control group showed a significant effect of valence (F(1, 28)=6.29, P=0.006), with negative images (M=72.8, SD=15.4) being better recalled than both positive (M=67.2, SD=19.4, P=0.02) and neutral (M=66.7, SD=13.9, P=0.009) images. By contrast, no such difference as a function of the emotional valence of stimuli was found for the ASD group (F(1, 28)=1.29, P>0.29). Rather, this group recognized neutral images (M=70.9, SD=12.1) as well as emotional ones (positive valence: M=64.6, SD=13.9; negative valence: M=65.9, SD=21.4).
Finally, correlation analyses revealed that age did not influence the performance of the ASD or the control group (all Ps>0.05). Also, no significant correlation was found between IQ and performance for the ASD group (all Ps>0.05).
The aim of this study was to investigate the influence of emotional information on a memory task in individuals with ASD relative to typically developing individuals.
This study revealed two main findings. First, typically developing participants recalled negative images better than neutral ones. This finding is consistent with that of previous studies showing a facilitating effect of negatively laden stimuli on memory performances [Anderson et al., 2006; Kensinger & Corkin, 2003; Medford et al., 2005]. By contrast, positively laden pictures did not give rise to better scores than neutral pictures. According to Brandt, Sünram-Lea, and Qualtrough , this stronger enhancement effect observed for negative than for positive stimuli could be owing to a difference in arousal. However, our data do not support this hypothesis as the images included in these two categories of stimuli were chosen to be equivalent in terms of arousal (see Methods section). This facilitating effect of negatively laden stimuli on performance is also commonly thought to be mediated by attention [Fox et al., 2001; Ohman et al., 2001; Phelps, 2004]. More attention given to emotional features of the scene would lead to better encoding; hence, when participants are asked to keep in mind both neutral or emotional pictures, related cerebral activity is different according to the valence of the picture [Perlstein, Elbert, & Stenger, 2002].
Second, and most importantly for the purpose of this study, this impact of negative pictures on recognition performance was not found for the ASD group. More precisely, this group's performance was not affected by the emotional content of pictures, positive or negative. Memory performance of the ASD group was thus found to be impervious to the emotional value of images. This result suggests that emotion does not influence recall performance in individuals with ASD.
Another interesting finding of this study is that the overall level of performance of the ASD group was similar to that of controls. In other words, individuals with ASD performed as accurately as controls when asked to recognize not only neutral, but also both positive and negative visual scenes. This suggests that the lack of memory enhancement by emotional images cannot be imputed to general memory impairments in ASD. Moreover, this assumption is consistent with previous studies showing intact recall skills in ASD [e.g., Williams, Goldstein, & Minshew, 2006]. What could thus explain the lack of impact of emotional content on recall skills in ASD?
As stated above, the enhancement of memory performances for emotional stimuli reported in typically developing groups is thought to result from increased attention allocated to these stimuli. In line with this idea, previous studies suggesting that individuals with ASD do not allocate particular attention to emotional relative to neutral stimuli [e.g., Begeer, Rieffe, Terwogt & Stockmann, 2006] provide ground for a first interpretation. Indeed, there is little doubt nowadays that individuals with ASD deal with emotional information in an atypical fashion. They certainly do not process emotional stimuli, such as faces and social scenes, like controls [e.g., Spezio, Adolphs, Hurley, & Piven, 2007] and even appear selectively impaired in remembering these stimuli [e.g., Williams, Goldstein, & Minshew, 2005].
Reversely, over-reactivity to the emotional features of perceived pictures could be an alternative hypothesis. One may argue that if ASD individuals are not able to control their own emotions in face of emotionally laden pictures, these latest may rather have a noxious effect on their recall abilities. However, our findings clearly establish that individuals with ASD recalled both neutral and emotional conditions as accurately, rendering this latest explanation unlikely.
Taken together, findings of this study suggest an atypical link between emotion and cognition in ASD. The specific link between emotion and a sophisticated function such as working memory has been scarcely explored in ASD, and studies on typical development have addressed this question in children only [for a review, see Pessoa, 2008]. In this respect, although behavioral in essence, our study provides results that can feed the literature on brain abnormalities in autism. Several studies have pinpointed the amygdaloid complex as a potential candidate underlying the impact of emotional information on memory skills in typically developing individuals [Cahill, 2000; Hamann, 2001; McGaugh, 2004; Paz, Pelletier, Bauer, & Paré, 2006]. As the amygdala has a central role in processing threat-related emotional expressions [e.g., Morris et al., 1996], it is not surprising that the emotional enhancement of episodic memory is greater for negative than for positive or neutral stimuli in typically developing individuals. Importantly, the amygdala has been suggested as a major site of cerebral dysfunction in the autistic brain [e.g., Baron-Cohen et al., 1999]. More precisely, many neuroimaging studies have reported abnormal activation of the amygdala in tasks involving facial-emotion recognition [Ashwin, Baron-Cohen, Wheelwright, O'Riordan, & Bullmore, 2006; Schultz, 2005] or memory for faces [e.g., Howard et al., 2000] in adults with ASD in comparison with typically developing adults. If one considers the emotion/cognition link tested in our study, it is plausible to evoke an amygdala dysfunction as the underlying neural basis of the ASD group's performance. However, it should be emphasized that it is probably the functional interactions between the amygdala—as a limbic/emotional structure—and other cerebral regions such as the dorsolateral prefrontal cortex involved in cognition and/or attention that are dysfunctional. Be it grounded on hyper- or hypo-reactivity to the perception of emotional content in pictures, if the processing is wrong then it will affect the whole brain network engaged in emotional information encoding. In turn, abnormal amount of attention will be allocated to these emotional stimuli—via abnormal gaze behavior, for example [Dalton et al., 2005]—leading to a lack of privileged processing and encoding of the emotionally informative parts of the picture. Although encoding is well performed by a potentially intact brain memory system, emotional valence may not modulate it in ASD. To be fully explored and verified, this hypothesis would need further investigation using neuroimaging methods.
Findings of this study should be interpreted with caution owing to the relatively small sample size and the absence of direct IQ-based groups matching. It is possible that encoding and recall performances are decreased in subjects with lower IQ levels. This is, however, unlikely as executive IQ is generally within the normal range in ASD subjects, and because our results show similar recall performances in both populations.
Findings of this study revealed that recall skills in ASD are atypically affected by the emotional content of visual scenes. These findings are not only consistent with previous reports showing that individuals with ASD atypically process facial and gestural expressions of emotion [e.g., Hill, Berthoz, & Frith, 2004; Spezio et al., 2007], but also show that this extends to the processing of emotional content of visual scenes. More importantly, this study establishes that atypical processing of emotional content in ASD has a reduced impact on higher-level cognitive processes such as memory. Taken together, these findings provide further theoretical background for the development of neutral-oriented remediation programs.
We thank all individuals who participated in this study, as well as their family. A. Santos was supported by a grant from the FCT-MCTES (Portugal, SFRH/BD/18820/2004) to conduct this study.