a neuropsychological perspective
Yifat Faran and Dorit Ben Shalom
Introduction
In this chapter it is argued that the socioemotional and memory profiles in autism may be consistent with an abnormality in two systems both of which involve an interaction between the limbic system and the medial prefrontal cortex. The first system, which is probably critical for a diagnosis of autism, includes the pathway from the amygdala to the medial prefrontal cortex. This system is involved in the generation of basic and complex feelings, and in the successful performance of theory of mind tasks. The other system is the one including the pathway from the hippocampus to the medial pre- frontal cortex (including the medial orbital prefrontal cortex). This system is arguably necessary for episodic memory. It is also perhaps involved in some of the ‘coherence’ impaired in ‘weak coherence’. In the case of both of these systems, it is suggested that autism involves an impairment in the interaction between the limbic system and the medial prefrontal cortex.
The chapter is organized as follows. First there is a brief review of findings on memory in autism, using Tulving’s (1995) analysis of differ- ent memory systems as a framework (for a fuller review, see Ben Shalom, 2003). This brief review focuses in particular on the episodic memory impairment in autism, and its likely brain bases. There follows a detailed account of what is known about emotion processing in autism, and the likely brain bases of the processing of emotions and feelings. The chapter closes with a brief discussion of the possible parallel between memory impairment and emotion-processing impairments in autism.
Intact and impaired memory systems in autism
Behavioural findings
In his 1995 review, Tulving distinguishes five types of memory systems: the procedural memory system, working memory, the perceptual
representation system, semantic memory and episodic memory (see Gardiner, this volume, Chapter 1). Procedural memory includes non- declarative motor and cognitive skills, as well as simple conditioning. This type of memory will not be dealt with in this chapter. Likewise, the chapter does not deal with working memory, whose subsystems hold ‘on- line’ information during the execution of cognitive tasks (see Poirier & Martin, this volume, Chapter 12). The chapter focuses on the last three memory systems on Tulving’s list: the perceptual representation system, semantic memory and episodic memory. Of these three memory systems, the perceptual representation system is clearly the most basic – a limbic (probably rhinal cortex) representation of ‘raw’, nondeclarative, percep- tual and cognitive information. Declarative information is derived from the perceptual representation system through at least two different path- ways (Tulving & Markowitsch, 1998; Aggleton & Brown, 1999). One, from the rhinal cortex to the inferior frontal gyrus derives fact information (semantic memory); the other, from the rhinal cortex to the hippocampus and then the medial prefrontal cortex (including the medial orbital pre- frontal cortex), derives event information (episodic memory).
The perceptual representation system is probably unimodal, and is sensitive to transformations that semantic and episodic memory are not sensitive to (such as the orientation of three-dimensional objects). Items in the perceptual representation system are rich in context, as are items in episodic memory. The within-item context in the perceptual representa- tion system is, however, unimodal, as are the core items themselves. In addition, whereas context and item are bound together in the perceptual memory system, contextual information is factored out of individual items in episodic memory and can be used to create organization between differ- ent items. Items in semantic memory are, by contrast, relatively context free. Semantic and episodic memory differ in that episodic memory has to do with conscious recollection of previous experiences of events, whereas semantic memory is concerned with the acquisition and use of the knowl- edge of what is, or could be, in the world (Tulving & Markowitsch, 1998). It has been argued in an earlier paper (Ben Shalom, 2003) that the perceptual representation system and the semantic memory system are relatively intact in people with high-functioning autism (HFA – used here to include people with a diagnosis of Asperger syndrome), whereas the episodic memory system is selectively impaired. This hypothesis, which would probably be accepted by most of the authors contributing to the present volume, gives rise to the following three predictions (Ben Shalom, 2003):
1. Memory tasks that do not depend on episodic memory but on the perceptual representation system or semantic memory, such as
rote memory and memory for facts, will show intact performance in autism.
2. Memory tasks that depend on episodic memory, and that cannot be compensated for using compensatory strategies involving either the perceptual representation system or semantic memory, will show impaired performance in autism. Such tasks would include tests of source memory, temporal order memory and memory for personally experienced events.
3. Memory tasks that depend partly or wholly on episodic memory but which can be compensated for by the perceptual representation sys- tem or semantic memory, such as free recall and recognition, will show a mixed performance in participants with autism. The level of per- formance may depend on many factors, including the test methods used and the cognitive level of the participants tested. Qualitatively, there should be signs of the use of atypical strategies.
The successful use of compensatory strategies, at least by more able individuals with autism, may derive from the fact that autism is a devel- opmental disorder. These types of compensations may be harder to acquire when a memory disorder is acquired in adulthood.
The evidence reviewed in Ben Shalom (2003) and reported in other chapters in the present volume seems largely compatible with these predictions, as summarized below:
1. Performance on memory tasks that do not depend on episodic mem- ory but on the perceptual representation system or semantic memory, such as rote memory and memory for facts, is intact in people with HFA. People with low-functioning autism (LFA) rely less on seman- tic memory and more on the perceptual memory system. Memory in people with LFA is famously rote, restricted and inflexible (e.g. O’Connor & Hermelin, 1988), all hallmarks of the perceptual repre- sentation system (Tulving & Schacter, 1990). In fact, both Toichi and Kamio (2002) and Mottron, Morasse and Belleville (2001) have reported that despite comparable levels of performance, individuals with high-functioning autism or Asperger syndrome show an equal memory gain from semantic as compared to perceptual or phonolog- ical processing. Mottron and Burack (2001) interpret these findings in terms of an enhancement of low-level perceptual processing in autism.
2. Memory tasks that are episodic in nature, and which cannot be compensated for using strategies involving semantic memory or the perceptual representation system, are universally impaired in partic- ipants with autism. In particular, source memory and temporal order memory have been shown to be impaired even in high-functioning
individuals (Bennetto, Pennington & Rogers, 1996). Deficits in the organization of information in memory, arguably related to episodic learning, are also consistently found in individuals with HFA (Minshew & Goldstein, 1993; Bennetto, Pennington & Rogers, 1996; Renner, Klinger & Klinger, 2000). Memory for personally experienced events has been shown to be impaired in people with LFA (Boucher, 1981a; Boucher & Lewis, 1989) as well as those with HFA (Klein, Chan & Loftus, 1999; Millward et al., 2000).
3. Memory tasks that depend partly or wholly on episodic memory, but for which it is possible to compensate using the perceptual represen- tation system or semantic memory, show a mixed pattern of perform- ance in participants with autism. The level of performance may depend on the cognitive level of the participants tested. Thus, in low-functioning children with autism, Boucher and Warrington (1976) found delayed free recall to be impaired relative to both normal and ability-matched comparison groups. Recognition, often argued to be less dependent on episodic memory than is free recall, was also impaired, but to a lesser degree. By contrast, Boucher (1981b) found immediate free recall to be intact in low-functioning children. In higher-functioning individuals, delayed free recall and recognition are unimpaired or close to unimpaired relative to ability-matched controls (Ameli et al., 1988; Minshew & Goldstein, 1993; Bennetto, Pennington & Rogers, 1996; Renner, Klinger & Klinger, 2000). The difference between immediate and delayed free recall in low- functioning autism can be explained if the perceptual representation system can be used in immediate but not delayed recall. The difference between delayed recall in LFA as compared to that in HFA might be explained if a higher IQ facilitates the existence and efficacy of semantic compensatory strategies (perhaps similar to the effects of cognitive ability on the ability to perform tasks that require empathy in participants with autism, Yirmiya et al., 1992).
It was also predicted that there should be qualitative signs of the use of atypical strategies. Reduced or absent primacy effects in the face of intact recency effects have been reported in both high-functioning (Renner, Klinger & Klinger, 2000) and low-functioning autism (Boucher, 1981b). Performance in high-functioning autism is reduced on the late but not the early trials of the California Verbal Learning Test (Minshew & Goldstein, 1993; Bennetto, Pennington & Rogers, 1996). Finally, despite similar levels of performance, participants with HFA (as used here to include Asperger syndrome) show more ‘know’ responses and fewer ‘remember’ responses in recognition compared to a comparison group (Bowler, Gardiner & Grice, 2000).
A neuropsychological perspective
A general role for the hippocampus and prefrontal cortex in episodic memory is well established (e.g. Tulving & Markowitsch, 1998; Aggleton & Brown, 1999). Whilst the above memory profile in autism is largely consistent with a selective deficit in episodic memory, it does not allow localization of the anatomical abnormality to either the hippocam- pus or the prefrontal cortex. All three predictions are consistent with abnormalities in either of these locations, while specific comparisons with other clinical groups sometimes point to the one and sometimes to the other.
One possible cause of this ambiguity is a variation in the location of the abnormality among individuals with autism (perhaps contributing to some of the variation between the autopsy findings reported by Kemper and Bauman (1993), who found abnormalities in the hippocampal region in a postmortem study in people with autism, and by Bailey et al., 1998, who did not find such abnormalities in the majority of their cases).
Intact and impaired emotion-processing systems in autism
Terminology
Turning to affective processing in autism, two different distinctions about affective processing have been developed within different areas of psycho- logy. One, arising within neuropsychology (e.g. LeDoux, 1996), distin- guishes between physiological emotions such as being in the physiological state of fear: for example, the way our body responds when we see a snake (fast heart beats, cold sweat, etc.); and cognitive feelings such as feeling afraid: for example, explaining to ourselves and to others that we felt afraid because we saw a snake.
Another distinction, arising within social psychology, distinguishes between basic emotions (feeling afraid, angry, happy, sad or disgusted; in our example feeling afraid of the snake) and more complex mental states, including the more properly ‘social’ emotions (e.g. feeling embar- rassed, guilty, proud or ashamed; for example, being embarrassed that you were afraid of such a small snake).
The fact that there are two theories that have different terminologies makes it hard to formulate hypotheses and test their predictions. To preserve both distinctions the following terminology will be adopted in the present chapter. Physiological emotions will denote only the physio- logical signs of emotions (i.e. heart pounding, cold sweat, etc.). Basic
feelings will refer to the cognitive counterparts of these physiological emotions (i.e. feeling afraid, angry, happy, sad or disgusted). More com- plex, ‘social’ feeling will be called complex feelings (e.g. feeling guilty, embarrassed, proud or ashamed). Going back to the example given above, when we see the snake and our heart beats faster and we are covered with cold sweat this is a physiological emotion; when we are conscious of these body reactions as feeling afraid this is a basic feeling; after a while when we get embarrassed that we reacted in panic to such a small snake this is a complex feeling. That is, complex feelings must include the understanding of other people’s reaction to our feelings. Thus, the result is a three-way distinction within affective processing (physiological emotions vs basic feelings vs complex feelings), instead of a pair of two-way distinctions. Armed with this less ambiguous termino- logy, we can describe two major hypotheses concerning affective process- ing in autism. These are presented next, followed by a review of the relevant current evidence and finally a neuropsychological perspective on the findings.
Hypotheses
Whilst many different hypotheses about affective processing in autism are in principle possible, it seems current debate centres around the following two hypotheses. One approach, revolving around the work of Baron- Cohen and his colleagues, makes the following three predictions con- cerning the three-way distinction outlined above: physiological emotions should be intact in autism; basic feelings should be intact, at least in high- functioning autism; complex feelings, on the other hand, should be impaired even in high-functioning autism. This latter prediction stems from the assumption that people with autism cannot appreciate another person’s point of view. Going back to our example, people with autism will have all the physical signs of fear when they see a snake and they will also feel afraid; however, they are not able to be embarrassed about feeling afraid, because they cannot appreciate how they may appear to another person.
A different approach, expressed in Ben Shalom (2000) and, perhaps, in Grossman et al. (2000), makes the following three predictions: physio- logical emotions should be intact in autism; all feelings, including basic and complex alike, should be atypical even in high-functioning autism (but not necessarily in PDD-NOS (pervasive developmental disorder – not otherwise specified)). Instances of intact performance by high-functioning individuals on tests of basic feelings should be achieved by using atypical compensation strategies, rather than by
using hard-wired brain mechanisms. Thus, for our example, people with autism will experience the normal physiological signs of fear when they see the snake; but if they say to themselves or to someone else that they are afraid, this reflects a compensation they have acquired either for themselves, or through instruction: the feeling (as opposed to the raw emotion) of fear is not something a person with autism will automatically experience. The difference between these two different approaches stems from differences in the assumptions they make about the locus of impairment in autism. While the first approach traces the disorder to an impaired processing of complex mental states, the second hypothesis claims that all feelings (as opposed to emotions) are processed atypically by people with autism. Thus, while both approaches would agree that both social and affective processing may be atypical in autism, they differ with respect to whether the primary impairment is social or rather cognitive-affective.
Current evidence relating to the two hypotheses Physiological emotions
Not much is known about physiological emotions in autism. Three recent studies are all suggestive, but none of them are completely convincing.
Willemsen-Swinkels et al. (2000) compared heart rate changes in chil- dren with autism spectrum disorders and typically developing children during separation and reunion with a parent. Children with autism spec- trum disorders and disorganized attachment showed an increase in heart rate during separation, and a decrease during reunion, a pattern familiar from younger typically developing infants (e.g. Spangler & Grossman, 1993). The number of children in this part of the study was however very small. Blair (1999) showed pictures of distress, threatening pictures and neutral pictures to children with LFA, children with mental retardation without autism and typically developing children, matched on verbal mental age. He measured electrodermal responses following presentation of the pictures. Like the children with mental retardation and the nor- mally developing children, the children with autism showed larger elec- trodermal responses to the pictures depicting distress than to the neutral pictures. Unlike the children with mental retardation and the typically developing children, the children with autism did not show larger electro- dermal responses to the threatening pictures than to the pictures with neutral affective content. It is not entirely clear why this is the case. Ben Shalom et al. (2006) showed pleasant, unpleasant and neutral pictures to children with HFA and a typically developing comparison group. They measured electrodermal responses following presentation of the pictures.
All children showed larger electrodermal responses to the unpleasant pictures than to the pleasant ones. The main effect of picture type, however, was only marginally significant.
Complex feelings
In contrast, a lot is known about complex feelings in autism, which have been extensively studied by Baron-Cohen and his colleagues, in partic- ular. For example, Baron-Cohen, Spitz and Cross (1993) matched chil- dren with autism, children with mental retardation and typically developing children on verbal mental age. They tested the ability to recognize the feelings of people from faces. The children with autism were impaired relative to both comparison groups in recognizing the complex feeling of surprise. Baron-Cohen, Wheelwright and Jolliffe (1997) compared adults with high-functioning autism and an IQ- matched normal comparison group on the ability to recognize complex feelings from photographs of faces. The participants with autism were impaired relative to the comparison group, and this impairment was even more pronounced in a condition in which the stimuli included the eye region alone. This finding may suggest that when participants with autism did manage to perform the task with the faces, they were using a strategy different from that of participants in the comparison group. Baron-Cohen et al. (1999) compared adults with HFA with typical adults on their ability to recognize complex feeling from the same type of stimuli of the eye region as the ones in Baron-Cohen, Wheelwright and Jolliffe (1997). The participants with autism performed less well than the normal adults. They also showed less activation of the amygdala during the task, as measured by functional magnetic resonance imaging. Concerning individual complex feelings, Capps, Yirmiya and Sigman (1992) com- pared children with high-functioning autism and verbal and nonverbal ability-matched typically developing children on the ability to talk about complex feelings of pride and embarrassment. The children with autism had more difficulty than children in the comparison groups in talking about these complex emotions. Similarly, Hillier and Allinson (2002) compared children with LFA to children with mental retardation and typically devel- oping children, matched for verbal and mental age. They tested the ability to understand scenarios describing embarrassing situations. The children with autism were impaired relative to the typically developing children on scenarios describing empathic embarrassment.
These kinds of results were given a comprehensive theoretical treat- ment in the theory of mind hypothesis of autism (Baron-Cohen, 1995). According to this hypothesis, people with autism have difficulty in ascrib- ing mental states to themselves and to others.
Basic feelings
The third, and perhaps most important line of evidence in deciding between the two hypotheses outlined above concerns basic feelings in autism. In contrast to physiological emotions, a large number of studies have been conducted about basic feelings in autism. Their results are for