Module 3_Discussion 3_Emotional Processing and Autism_Social Cognitive Neuroscience

Module 3

Emotional Processing and Autism

A. Historical Perspectives On The Emotions

The classic science fiction depiction of androids such as C3PO in Star Wars and

Data in Star Trek: Next Generation is of superhuman intelligent beings able to speak

many languages and store vast amounts of information. Nevertheless, such intelligence

does not enable them to fully understand the eccentric behaviors of their human

colleagues who constantly place themselves in danger, fall in love, and tell jokes. These

androids lack emotions. Reading between the lines of these popular depictions we might

conclude various things. We might conclude that emotions are ‘what makes us human,

force us to make illogical decisions, and that we could do without them if redesigned

from scratch. Needless to say, these conclusions are incorrect. Emotional processes have

a long evolutionary history and are by no means unique to humans. What may ‘make us

human’ is our ability to consciously reflect on our emotions and share them socially via

our language and culture, but not our emotions per se. If a new organism were redesigned

from scratch it would still be helpful to have early warning routines for danger and fast-

acting mechanisms that prepare it to fight or flee. It would still be helpful to devote

greater attention to stimuli that are necessary for survival. These are all considered

functions of emotions. Finally, emotions do sometimes lead to decisions that may not

have occurred via more deliberative reasoning but this, by itself, does not make them

illogical. For instance, in cooperative games with another person we often make decisions

based on social values of fairness rather than maximizing individual financial gain (e.g.

Sanfey, Rilling, Aaronson, Nystron, & Cohen, 2003). This is not necessarily illogical –

there may be good survival reasons, honed by evolution, that promote such cooperation.

For humans, many social stimuli and situations are rewarding (e.g. imitation,

cooperation) or punishing (e.g. social exclusion). As such, both social stimuli and non-

social stimuli are likely to have been selected as having survival value in our evolutionary

past.

Why are some stimuli associated with emotions and others are not? The standard

answer to this question is that some stimuli are more important than others (e.g. because

they enhance or threaten survival chances). Emotions are one way of tagging these

stimuli to ensure that they receive priority treatment and are responded to appropriately.

Broadly speaking, they can be tagged in one of two ways: either as something that is to

be sought (i.e. a rewarding stimulus) or avoided (i.e. a punishing stimulus). As such,

many theories closely tie emotions with the concept of motivation. For example, Rolls

(2005) defines emotions as states elicited by rewards and punishers, whereas motivation

is defined as states in which rewards are sought and punishers are avoided. Importantly,

emotions are not just tied to stimuli but also to predicted stimuli. Thus, the omission of an

expected reward can lead to emotions (e.g. anger), as can omissions of expected

punishment (e.g. relief). For humans, we can make the further claim that we like rewards

and we dislike punishers, and that we are motivated to seek the things we like and avoid

the things we dislike. (For animals, we tend to avoid the terms like/dislike and adopt

more neutral terminology such as seek/avoid because we cannot know their subjective

feelings.) Although we may be born with a core set of basic likes and dislikes (e.g. we

like sweet things and dislike pain), it is possible to arbitrarily learn new emotional

associations by pairing neutral stimuli with emotive responses. We may come to be afraid

of flying in airplanes, or we may come to like certain painful stimuli (e.g. eating chilies,

fetishes). As such, emotional learning is a highly flexible system that is not limited to

stimuli in our evolutionary past and extends beyond stimuli with obvious survival value.

In 1872, Charles Darwin published The Expression of the Emotions in Man and

Animals (Darwin, 1872/1965). For much of this work Darwin was concerned with

documenting the outward manifestations of emotions – expressions – in which animals

produce facial and bodily gestures that characterize a particular emotion such as fear,

anger, or happiness. Darwin noted how many expressions are conserved across species:

anger involves a direct gaze with mouth opened and teeth visible, and so on. This is

shown in Figure 4.1. He claimed that such expressions are innate ‘that is, have not been

learnt by the individual’. Moreover, such expressions enable one animal to interpret the

emotional state of another animal – for example, whether an animal is likely to attack, or

is likely to welcome a sexual advance. For Darwin, an emotional expression was a true

reflection of an inner state: ‘They reveal the thoughts and intentions of others more

clearly than do words, which may be falsified.’ Darwin’s contribution was to provide

preliminary evidence as to how emotions may be conserved across species. His reliance

on expressions resonates with some contemporary approaches, such as Ekman’s attempts

to define ‘basic’ emotions from cross-cultural comparisons of facial expressions (e.g.

Ekman, Friesen, & Ellsworth, 1972). This is covered in detail later in the chapter. More

recent research has elucidated the functional origins of some of these expressions; for

instance; a posed fear expression increases the visual field and nasal volume and leads to

faster eye movements (adaptive for detecting danger), whereas a disgust expression has

an opposite effective (adaptive for avoiding contaminants).

For Freud, our minds could be divided into three different kinds of mechanisms:

the id, the ego, and the super-ego. The id was concerned with representing our ‘primitive’

urges that connect us to non-human ancestry. It includes motivations to meet our basic

emotional needs for sex, food, warmth, and so on. The id was concerned with

unconscious motivations, but these ideas would sometimes be consciously accessible via

the ego, which operates according to reason rather than passion. The super-ego, by

contrast, represents the ideal self, such as our cultural norms and our aspirations. For

Freud (and many of his clients), there was a perceived conflict between the superego, for

which sexual behavior was tightly regulated by cultural norms, and the id, for which

more unbridled sexual impulses were considered as desirable.

One of the founding fathers of psychology, William James, proposed a theory of

emotion that placed the somatic (i.e. bodily) response of the perceiver at its center

(James, 1884). This theory later became known as the James–Lange theory of emotion.

According to this theory, it is the self-perception of bodily changes that produces

emotional experience. Thus, changes in bodily state precede the emotional experience

rather than the other way around. We feel sad because we cry, rather than we cry because

we feel sad: see Figure 4.2. This perspective seems somewhat radical compared to the

contemporary point of view elaborated thus far. For instance, it raises the question of

what type of processing leads to the change in bodily states and whether or not this early

process could itself be construed as a part of the emotion. Changes in the body are

mediated by the autonomic nervous system (ANS), a set of nerves located in the body

that controls activity of the internal organs. There is good empirical evidence to suggest

that changes in somatic state, in themselves, are not sufficient to produce an emotion.

Schacter and Singer (1962) injected participants with epinephrine (also termed

adrenaline), a drug that induces autonomic and visceral changes. They found that the

presence of the drug by itself did not lead to self-reported experiences of emotion,

contrary to the James–Lange theory. However, in the presence of an appropriate

cognitive setting (e.g. an angry or happy man enters the room), the participants did self-

report an emotion. A cognitive setting, but without epinephrine, produced less intense

emotional ratings. This study suggests that bodily experiences do not create emotions

(contrary to the James–Lange theory) but they can enhance conscious emotional

experiences.

The Cannon–Bard theory of emotions that emerged in the 1920s argued that

bodily feedback could not account for the differences between the emotions (Cannon,

1927). According to this view, the emotions could be accounted for solely within the

brain, and bodily responses occur after the emotion itself. The Cannon–Bard theory was

inspired by neurobiology. Earlier research had noted that animals still exhibit emotional

expressions (e.g. of rage) after removal of the cortex. This was considered surprising

given that it was known that cortical motor regions are needed to initiate most other

movements (Fritsch & Hitzig, 1870). In a series of lesion studies, Cannon and Bard

concluded that the hypothalamus is the centerpiece of emotions. They believed that the

hypothalamus received and evaluated sensory inputs in terms of emotional content, and

then sent signals to the autonomic system (to induce the bodily feelings discussed by

James) and to the cortex (giving rise to conscious experiences of emotion). Although it

has not stood the test of time (e.g. the hypothalamus is not a central nexus of emotions,

although it does regulate bodily homeostasis), the theory was important historically in

providing an alternative to the James–Lange theory and also for the development of

another important theory: namely the Papez circuit and the limbic brain hypothesis.

Papez (1937) drew upon the work of Cannon and Bard in arguing that the

hypothalamus was a key part of emotional processing, but extended this into a circuit of

other regions that included the regions of the cingulate cortex, hippocampus,

hypothalamus, and anterior nucleus of the thalamus. Papez argued that the feeling of

emotions originated in the subcortical Papez circuit, which was hypothesized to be

involved in visceral regulation. A second circuit, involving the cortex, was assumed to

involve a deliberative analysis that retrieved memory associations about the stimulus. The

work of MacLean (1949) extended this idea to incorporate regions such as the amygdala

and orbitofrontal cortex, which he termed the ‘limbic brain’. The different regions were

hypothesized to work together to produce an integrated ‘emotional brain’. There are a

number of reasons why these earlier neurobiological views are no longer endorsed by

contemporary cognitive neuroscience. First, some of the key regions of the Papez circuit

can no longer be considered to carry out functions that relate primarily to the emotions.

For example, the role of the hippocampus in memory was not appreciated until the 1950s

(e.g. Scoville & Milner, 1957). Second, contemporary research places greater emphasis

on different types of emotion (e.g. fear versus disgust). Each basic emotion may form

part of its own circuit, and different parts of the circuit may make different cognitive

contributions.

B. Different Categories of Emotion in The Brain

One of the most influential ethnographic studies of the emotions concluded that

there are six basic emotions that are independent of culture (Ekman & Friesen, 1976;

Ekman et al., 1972): happy, sad, disgust, anger, fear, and surprise (shown in Figure 4.3).

This study was based on comparisons of the way that facial expressions are categorized

and posed across diverse cultures. Ekman (1992) considers other characteristics for

classifying an emotion as ‘basic’ aside from universal facial expressions, such as: each

emotion having its own specific neural basis; each emotion having evolved to deal with

different survival problems; and occurring automatically. The list of emotions is not

considered closed. For instance, Ekman (1992) considers adding embarrassment, awe,

and excitement as basic emotions and dropping surprise. Johnson-Laird and Oatley

(1992) examined the words that we have for emotions and came up with a list of five

basic emotions that overlap closely with Ekman’s but does not contain surprise. What of

other candidate emotions? One possibility is that different candidate emotions are

different shades of the same basic emotion. For example, happiness might include

amusement, relief, pride, satisfaction, and excitement (Ekman, 1992). Another possibility

is to consider some emotions as being comprised of two or more basic emotions. Plutchik

(1980) offers a detailed account along these lines. He proposes eight basic emotions

(surprise, sadness, disgust, anger, anticipation, joy, acceptance, fear) that may be

combined in various ways: for example, joy + fear = guilt and fear + surprise = alarm. A

third possibility is that some emotions should be construed in terms of a basic emotion(s)

plus a non-emotional cognitive appraisal. These cognitive + emotional blends might be

needed to account for complex emotions such as jealousy, pride, embarrassment, and

guilt. Such emotions might involve attribution of mental states that imply awareness of

another person’s attitude to oneself, or awareness of oneself in relation to other people.

As such, they have been referred to as moral emotions.

Not all models of emotion assume that some emotions are more basic. Three

different accounts along these lines are considered here: that of Ortony and colleagues

(Ortony, Clore, & Collins, 1988; Ortony & Turner, 1990); that of Rolls (2005); and that

of Barrett (2006; Lindquist & Barrett, 2012). Ortony and colleagues argue that all

emotions are appraisals based on a valenced reaction (i.e. positive vs. negative) to a given

stimulus and event. The range of emotions is limited by the range of appraisals that one

can deploy, rather than consisting of some pre-determined number. These appraisals can

occur unconsciously as well as consciously. For example, an emotion such as ‘shame’

would be an outcome of various appraisals such as: it has a negative valence; it refers to

the action of people; and it is self-focused. This is illustrated in Figure 4.4. Although this

theory is not couched in terms of neuroscience, it is not hard to imagine how such a

model would translate. It may, for instance, involve a small set of regions involved in

generating the ‘valenced reaction’ that interact with a more distributed set of regions

involved in different kinds of appraisal. Indeed, we see examples of this kind of model in

the theories presented below.

Finally, the theory of Barrett and colleagues (Barrett & Wager, 2006; Barrett,

2006; Lindquist & Barrett, 2012) assumes that all emotions tap into a system termed core

affect that is organized along two dimensions: pleasant–unpleasant and high/low arousal.

The latter is also termed activation. This is illustrated in Figure 4.5. Evidence that

emotional experience can be classified along these two dimensions comes from studies

employing factor analysis of current mood ratings (Yik, Russell, & Barrett, 1999). This

study found that all subjective moods fall somewhere within this twodimensional space.

In biological terms, this is linked to bodily feelings of emotion and linked to limbic

structures such as medial temporal lobes, cingulate and orbitofrontal cortex (Lindquist &

Barrett, 2012). This echoes the older ideas of Papez and Maclean. The novel aspect of the

model is the idea that categories of emotion are constructed (and can be differentiated

from each other) because they tap the core affect system in somewhat different ways and

because they are linked to certain kinds of information processed outside of the core

affect system, including executive control (for regulating and appraising emotions),

language (for categorizing and labeling), theory of mind (for conceptualizing emotions in

terms of other agents), and so on.

The amygdala (from the Latin word for almond) is a small mass of gray matter

that lies buried in the tip of the left and right temporal lobes as shown in Figure 4.6. It lies

to the front of the hippocampus and, like the hippocampus, is believed to be important for

memory – particularly for the emotional content of memories (Richardson, Strange, &

Dolan, 2004) and for learning whether a particular stimulus/ response is rewarded or

punished (Gaffan, 1992). In monkeys, bilateral lesions of the amygdala have been

observed to produce a complex array of behaviors that have been termed the Kluver–

Bucy syndrome (Kluver & Bucy, 1939; Weiskrantz, 1956). These behaviors include an

unusual tameness and emotional blunting, a tendency to examine objects with the mouth,

and dietary changes. This is explained in terms of objects losing their learned emotional

value. The monkeys typically also lose their social standing (see Figure 4.7). In humans,

the effects of amygdala lesions are not as profound (Paul, Corsello, Tranel, & Adolphs,

2010). This may reflect either a greater cortical influence on emotional and social

behavior or the fact that the earlier monkey studies are likely to have produced lesions

extending beyond the amygdala.

Some researchers have argued that the ability to detect threat is so important,

evolutionarily, that it may occur rapidly and without conscious awareness (LeDoux,

1996). Ohman, Flykt, and Esteves (2001) report that people are faster at detecting fear-

related stimuli such as snakes and spiders amongst flowers and mushrooms than the other

way around. When spiders or snakes are presented subliminally to people with spider or

snake phobias, then participants do not report seeing the stimulus but show a skin

conductance response indicative of emotional processing (Ohman & Soares, 1994). In

these experiments, arachnophobics show the response to spiders, not snakes; and

ophidiophobics show a response to snakes but not spiders. In terms of neural pathways, it

is generally believed that there is a fast subcortical route from the thalamus to the

amygdala and a slow route to the amygdala via the visual cortical pathways (Adolphs,

2002; Morris, Ohmann, & Dolan, 1999). This is illustrated in Figure 4.9. Functional

imaging studies suggest that the amygdala is indeed activated by unconscious fearful

expressions, presented too briefly to be consciously seen (Morris et al., 1999). This is

consistent with a subcortical/fast route to the amygdala, although it is to be noted that the

temporal resolution of fMRI does not enable any direct conclusions to be drawn about

processing speed.

Activation of a fear response by the amygdala may trigger changes elsewhere in

the brain that enable the threat to be evaluated and responded to, if necessary. There are

connections from the amygdala to the autonomic system (LeDoux et al., 1988). These

may help prepare the body for fight and flight by increasing the heart and breathing rates.

The anterior cingulate is believed to be involved in this process (Critchley et al., 2003),

and it too is selectively activated by fear relative to happiness (Morris et al., 1996). In

addition, there is a strong relationship between the level of fear and increases in activity

in regions of the visual cortex (Morris et al., 1998). Thus, the detection of potential threat

by the amygdala may trigger more detailed perceptual processing of the threatening

stimulus, enabling further evaluation. Other, more frontal, regions may also be important

for deciding whether to act on this information. In conclusion, although the amygdala

may be essential for the evaluation of potential danger, its role should be construed in

terms of its influence upon a wider circuit of emotional processing.

The most convincing evidence for a specialized role of the amygdala in fear

comes from functional imaging studies that compare fear expressions with other

emotional expressions, and studies of human patients with damage to the amygdala who

show relatively selective deficits in recognizing fear. However, the conclusion that the

amygdala is specialized for fear may still be premature. First, the interpretation of these

findings hinges on the assumption that the stimuli were appropriately matched. If fear-

related stimuli are simply more difficult or more arousing (e.g. because a fearful face has

more survival value than happy or sad faces) then the data could be explained without

assuming specialization for fear. However, evidence that damage to other regions does

not selectively affect fear (e.g. insula lesions and disgust) speaks against this more

general account. Second, the amygdala might be specialized for some other process that

just happens to be more relevant for fear. For instance, it has been suggested that

selective impairments in fear may arise because of a failure to attend closely to the eyes

(Adolphs et al., 2005). However, evidence that the amygdala is involved in fear in other

domains (music, speech) speaks against this account.

The insula is a small region of cortex buried beneath the temporal lobes (it

literally means ‘island’), as shown in Figure 4.11. It is involved in various aspects of

bodily perception, including important roles in pain perception and taste perception. The

word disgust literally means ‘bad taste’ and this category of emotion may be

evolutionarily related to contamination and disease through ingestion. Patients with

Huntington’s disease can show selective impairments in recognizing facial expressions of

disgust (Sprengelmeyer et al., 1997) and relative impairments in vocal expressions of

disgust (Sprengelmeyer et al., 1996). Huntington’s disease is a genetic disorder with

symptoms arising in mid-adulthood and including excessive movements, cognitive

decline, and structural atrophy in the brain, particularly in regions such as the basal

ganglia. However, the degree of the disgust-related impairments in this group correlates

with the amount of damage in the insula (Kipps, Duggins, McCusker, & Calder, 2007).

Selective lesions resulting from brain injury to the insula can affect disgust perception

more than recognition of other facial expressions (Calder et al., 2000). In healthy

participants undergoing fMRI, facial expressions of disgust activate this region but not

the amygdala (Phillips et al., 1997). Feeling disgust oneself and seeing someone else

disgusted activates the same region of the insula.

According to a ‘basic emotion’ viewpoint, a separate neural substrate for disgust

may have evolved to deal with one particular situation – contamination. This may also

explain why disgust has its particular anatomical location, close to the primary gustatory

cortex involved in early cortical processing of taste. However, we use the word ‘disgust’

in at least one other context, namely to refer to social behavior that violates moral

conventions. Disgusting behavior is said, metaphorically, to ‘leave a bad taste in the

mouth’. But is there more to this than metaphor? Some have argued that moral disgust

has evolved out of non-social, contamination-related disgust (e.g. Tybur, Lieberman, &

Griskevicius, 2009). Moral disgust also results in activity in the insula (Moll, Zahn, de

Oliveira-Souza, Krueger, & Grafman, 2005) and moral disgust is associated with subtle

oral facial expressions characteristic of disgust more generally.

A selective deficit in recognizing anger has been reported following damage to

the ventral striatal region of the basal ganglia (Calder, Keane, Lawrence, & Manes,

2004). The dopamine system in this region has been linked to the production of

aggressive displays in rats (van Erp & Miczek, 2000), as well as in reward-based

motivation more generally. In this latter context anger/aggression could be construed as a

motivated behavior to obtain or defend rewards. Functional imaging studies of anger also

show activity in several regions linked to emotional processing, and are not suggestive of

a unique neural signature. For example, one study insulted participants and then asked

them to ruminate on the insult in the scanner (Denson, Pedersen, Ronquillo, & Nandy,

2009). Activity in the anterior cingulate correlated with self-reported anger, and initial

activity in this region, the insula, and hippocampus predicted the degree of rumination.

C. How Emotions Interface with Other Cognitive Processes

In order to serve these functions they must necessarily be linked to other cognitive

processes outside of the network of regions that are primarily involved in emotion and

motivation. This includes cognitive systems involved in representing our thoughts and

beliefs, the memory system, and the attention and perceptual system. The sections below

consider each of these. It is worth noting that although these systems can be considered as

functional, in the broadest sense, they can become maladaptive in certain cases. For

instance, affective disorders such as depression and anxiety need to be understood not

only in terms of affect (negative mood) but also in terms of the interplay between affect

and other aspects of cognition. They tend to be accompanied by certain styles of thinking

(e.g. catastrophizing) and attentional biases (e.g. to negative or threatening information)

that maintain the negative affect and lock the individual into a vicious cycle.

Although emotional processing is often considered to be automatic, it is possible

nevertheless to exert a degree of control in terms of whether to act upon this information

or how to interpret it. This involves an interplay between lateral prefrontal cortex (PFC)

involved in cognitive control and ‘executive functions’, the ventromedial/orbital parts of

the PFC (involved in emotional experience and contextualizing emotions), and regions

such as the amygdala. This general field is subsumed within the topic of emotion

regulation. This can be studied experimentally using fMRI by presenting participants

with an affective stimulus and then instructing them to think consciously about it in

different ways. For instance, they may try to suppress it or put a negative/positive spin on

it. The latter mechanism also termed reappraisal (for a review see Ochsner, Silvers, &

Buhle, 2012). Ochsner, Bunge, Gross, and Gabrieli (2002) presented negative images

(e.g. of someone in traction in a hospital) to participants in one of two conditions: either

passively viewing them or a ‘cognitive’ condition in which they were instructed to

reappraise each image ‘so that it no longer elicited a negative response’.

Their analysis revealed a trade-off between activity in the lateral PFC (high when

reappraising) and the ventromedial PFC and amygdala (high during passive looking).

When participants are asked to reappraise the stimulus negatively (i.e. making it worse

than it looks), then this also engenders a similar network in the lateral PFC but tends not

to dampen activity in the ventromedial PFC and amygdala (Ochsner et al., 2004). Similar

results are found when comparing passive viewing of negative images and explicitly

describing/labeling the images (Hariri, Bookheimer, & Mazziotta, 2000; Lieberman et al.,

2007). The EEG-based event-related potential (ERP) method enables the time course of

emotion regulation to be studied. Using this method it is generally found that regulatory

effects emerge from around 300 ms onwards, i.e. after the perceptual stage of processing

(Hajcak, MacNamara, & Olvet, 2010). In terms of brain stimulation, excitatory (anodal)

tDCS over left lateral PFC reduced the rated emotionality of negative pictures (Pena-

Gomez, Vidal-Pineiro, Clemente, Pascual-Leone, & Bartres-Faz, 2011). This was

interpreted as increased emotion regulation induced by PFC stimulation.

Heatherton and Wagner (2011) propose a model of emotion regulation and its

failures. They construe it as a balance between top-down processes (e.g. in the lateral

PFC) and affective processes linked to reward (such as nucleus accumbens) and

emotional salience (such as the amygdala). This is illustrated in Figure 4.24. In this

context, the lateral PFC may act to maintain one’s long-term goals (do not take drugs,

stick to the diet) whereas other regions may signal contradictory behavior (take me, eat

me). Self-regulatory failure takes place either when the prefrontal system is compromised

(e.g. mental fatigue or ‘resource depletion’) or the external cues are particularly strong

(e.g. after a period of abstinence or ‘lapse activated consumption’). For instance, when

cocaine addicts attempt to suppress their craving this results in reduced activity in

orbitofrontal cortex and nucleus accumbens and this reduction is correlated with

increased fMRI activity in lateral PFC (Volkow et al., 2010). The model also applies to

social stimuli: for instance, when one tries to avoid acting on racist stereotypes, or to

suppress (or reappraise) one’s anger. In one study, fMRI activity in the lateral PFC was

found to be inversely correlated with amygdala activity to viewing racial outgroup

members, but only when the stimuli were presented for long enough to elicit reappraisal.

Attention is the process by which certain information is selected for further

processing and other information is discarded. Attention is needed to avoid sensory

overload. The brain does not have the capacity to process fully all the information it

receives. Nor would it be efficient for it to do so. Modern theories of attention emphasize

that selection can take place at multiple levels in the cognitive system (Desimone &

Duncan, 1995). At the top level, selection may occur due to task-relevance and goals (i.e.

current priorities). At the bottom level, selection may occur due to perceptual salience:

for example, a red object amongst green objects will stand out. Emotional and social

stimuli may also be made to ‘stand out’ not due to the perceptual salience but due to

specialized pathways for processing this kind of stimuli; for instance as in LeDoux’s

(1996) proposed fast route to sensory cortices via the amygdala (Figure 4.9). However,

top-down mechanisms are likely to be important too. For instance, attention is drawn

faster to emotional items than to neutral items when these also constitute the targets to be

searched.

Pourtois, Schettino, and Vuilleumier (2013) offer a neural model of how

emotional stimuli capture attention. In general, the perceptual representations of attended

objects tend to be activated more (in fMRI) than those that are unattended (Kanwisher &

Wojciulik, 2000). For instance, attending to a face and ignoring a house will activate

face-selective regions more than place-selective regions, but attending to a house will

reverse that pattern. This boost of activation may relate to increased awareness of the

attended stimuli and prioritize it for further processing. Emotional stimuli (e.g. facial

expressions relative to neutral ones) tend to be linked to greater activity in perceptual

representations whether attended or unattended (Vuilleumier, Armony, Driver, & Dolan,

2001). Pourtois et al. (2013) argue that this reflects an additional boost from the

amygdala that serves to prioritize these stimuli.

Memory is not a single system in the brain but is rather a constellation of different

mechanisms. When the lay public use the term ‘memory’ they are generally referring to

what psychologists term episodic memory: i.e. they are memories of specific events that

occurred in a particular time and place. They are consciously reportable memories (and

hence fall under the umbrella terms of declarative memory or explicit memory) and are

often imbued with affective features and visuo-spatial imagery. It can be contrasted with

other forms of memory such as semantic memory – this is information that is consciously

known but is not re-experienced as an event (e.g. the capital of France). In addition, there

are various unconscious forms of remembering including procedural memory (e.g. skills

such as riding a bike) and also various forms of conditioned associations (e.g. tone-shock

associations) that have been discussed already.

One consistent finding in the literature is that episodic memories for emotional

stimuli (e.g. affective images presented experimentally, or events from one’s own life)

tend to be better remembered than neutral ones. This has been examined in imaging

experiments using the subsequent memory paradigm. The participant is shown a series of

stimuli (e.g. images) whilst scanned. They are then asked which stimuli they remember,

and the experimenter can then go back and determine what, in the brain, distinguished

remembered from forgotten trials when they were initially encountered. The amygdala

has been consistently implicated in the memory advantage for emotional stimuli.

Activation of the amygdala and hippocampus is correlated during learning of emotional

scenes that are subsequently better remembered (Dolcos, LaBar, & Cabeza, 2004). In

addition to coupling between the amygdala and hippocampus, there may also be a boost

to object-based representations in the visual ventral stream that is consistent with its

known role in attention (Kensinger, Garoff-Eaton, & Schacter, 2007). Reappraisal of

negative stimuli at encoding also boosts their subsequent memory, relative to suppression

or passive viewing, and this boost is linked to the amygdala, hippocampus, and lateral

PFC (Hayes et al., 2010). Patients with amygdala lesions show impaired memory for the

emotional details of scenes but not for other details of scenes.

The studies cited above use emotional stimuli but didn’t contrast social (i.e.

involving other people) and non-social (e.g. a snake, a gun). When social stimuli are used

there tends to be a greater involvement of the ventro-medial and orbital PFC. Mitchell,

Macrae, and Banaji (2004) asked participants to encode memories of people either

socially (form an impression of them) or non-socially (remember the order). Subsequent

memory affects were found in the medial PFC for social encoding and in the

hippocampus for non-social encoding. Tsukiura and Cabeza (2008) compared memory

for face–name associations with smiling or neutral faces. Smiling face–name pairs were

better remembered and this was linked to increased coupling between orbitofrontal cortex

and hippocampus during successful encoding and retrieval. Finally, Summerfield,

Hassabis, and Maguire (2009) contrasted memories for events that happened to

themselves versus other people and found medial prefrontal regions discriminated self

from other during memory recall.

D. Empathy as A Multi-Faceted Concept

If you see someone yawning do you yawn too? Most people probably do to some

extent. Some behavior, such as laughing and yawning, is socially contagious. But can any

wider significance be attached to such findings? One study of contagious yawning in

chimpanzees speculates that ‘contagious yawning in chimpanzees provides further

evidence that these apes possess advanced self-awareness and empathic abilities’

(Anderson, Myowa-Yamakoshi, & Matsuzawa, 2004). Another study, this time on

humans, administered tests requiring reasoning about the mental states of other people

(e.g. beliefs, knowledge) as well as measuring yawn contagion, and concluded that

‘contagious yawning may be associated with empathic aspects of mental state attribution’

(Platek, Critton, Myers, & Gallup, 2003). Of course, there is unlikely to be anything

special about yawning itself. There might be a general tendency to simulate the behavior

of others on ourselves (internally in our minds and brains) even if we do not overtly

reproduce it (as observable behavior on our bodies). Thus, we may understand others by

creating a similar response in our brain to that found in the other person’s brain.

Contagious yawning, under this account, is one extreme example of this more general

and, normally, more subtle tendency. This chapter will attempt to unpick these claims and

place them alongside traditional concepts in social and cognitive psychology, such as

empathy and theory of mind. The chapter will also consider how these processes may be

disrupted after brain injury and in people with autism.

The word empathy is relatively modern, being little more than 100 years old. It

was coined by Titchener (1909) from the German word einfühlung (Lipps, 1903) and

originally referred to putting oneself in someone else’s situation (literally ‘feeling into’).

This would also go under the contemporary name of perspective taking. This section will

first consider the various different ways in which the term empathy is used today. This

reveals potentially important distinctions that theories of empathy need to explain. If one

starts with the working definition of empathy introduced above (‘putting oneself in

someone else’s situation’) it is clear that there are subtle, but potentially crucial, different

ways in which this could be understood.

The first three scenarios differ with respect to whether the knowledge/feeling is

the same in self and other. Knowing about another person’s internal state need not

necessarily imply that the observer shares that state. This important consideration lies at

the heart of some tests of theory of mind, specifically false belief tasks, but they are

relevant to some conceptions of empathy too. The second sense in which empathy is used

(‘adopting the posture or matching the neural response of an observed other’) is the one

most closely linked with mirror systems, imitation, and contagion (emotional contagion,

yawning contagion, etc.). For example, one might feel personal distress in response to

someone else’s suffering. The third sense in which the term empathy may be used differs

from the second in that the person’s response is not matched. For instance, one might feel

a sense of pity to another’s situation or sympathy or compassion towards someone who is

suffering. These reactions are directed outwards (other-oriented) rather than being self-

oriented (as in personal distress), and the response of the perceiver does not match that of

the other person (Singer & Klimecki, 2014). The fourth and fifth notions of empathy

relate more directly to the idea of perspective taking, but they differ in whether they are

selforiented versus other-oriented. The fourth scenario (‘imagining how I would feel/

react in that situation’) could be construed as a shallow attempt to empathize, in which

the level of success is dependent on self–other similarity rather than a true understanding

of the other.

Given these somewhat different conceptions of empathy, it is not surprising that

there is no single agreed-upon measure of empathy. Theory-of-mind tests, discussed in

detail below, normally involve assessments based on linguistic reasoning of the sort: ‘If

X believes Y then how will he/she behave in situation Z?’ Others use neural or bodily

responses to seeing others in pain, for example, as a measure of empathy (e.g. Bufalari,

Aprile, Avenanti, Di Russo, & Aglioti, 2007; Jackson, Meltzoff, & Decety, 2005). Of

course, this presupposes a certain idea of what empathy is (i.e. that it can be measured

solely in physiological ways). There are various questionnaire measures of empathy, such

as the Interpersonal Reactivity Index (IRI; Davis, 1980) and the Empathy Quotient (EQ;

Baron-Cohen & Wheelwright, 2004), which touch upon some of the distinctions

discussed above. For example, the IRI contains separate subscales such as personal

distress (items such as ‘I tend to lose control during emergencies’), perspective taking

(items such as ‘Before criticizing somebody, I try to imagine how I would feel if I were

in their place’), and empathic concern (items such as ‘I often have tender, concerned

feelings for people less fortunate than me’). One current trend is to incorporate

questionnaire measures in functional imaging experiments. For example, watching

someone drinking a pleasant or disgusting drink may activate the gustatory (taste) regions

of the perceiver Moreover, the extent to which this occurs may be greater in those people

who report higher empathy on questionnaire measures (Jabbi et al., 2007).

Findings such as these are often used to argue that the different concepts of

empathy are related or, at least, share a common core (perhaps based upon simulation).

Finally, one could potentially measure the ability to accurately empathize (i.e. to

accurately state what another person is thinking or feeling) rather than the extent to which

the person may report the motivation to empathize (i.e. most questionnaire measures) or

to simulate that state themselves (which need not be linked to the ability to consciously

report that state). An example of such a test is shown in Figure 6.3. The ‘reading the mind

in the eyes’ test requires participants to match expressions in the eye region of faces to

labels denoting mental states such as bored, sorry, or interested (Baron-Cohen,

Wheelwright, Hill, Raste, & Plumb, 2001). Another test requires two participants to work

together in a scenario that is video recorded. Each participant can then watch it back and

report their own internal states as well as attempting to infer that of the other participant,

thus enabling the experimenter to cross-reference the responses together in order to infer

empathic accuracy (e.g. Ickes, 1993; Ickes, Gesn, & Graham, 2000). Although women

tend to score higher on questionnaire measures of empathy, this gender bias is reduced (if

not eliminated) when measures of empathic accuracy are used (Ickes et al., 2000). This,

again, points to the need to distinguish between different conceptions of empathy such as

a disposition to empathize (i.e. deliberately attempt to perspective take), which may be

tapped by questionnaires, and an ability to empathize which may be tapped by

performance-based measures.

Most simulation theories of empathy are based on the notion of perception–action

coupling, i.e. the link between seeing actions on other people and reproducing those

actions on one’s own motor system. In some cases the action might be literally

reproduced (direct imitation), or be reproduced in a more subtle form (e.g. contractions of

facial musculature that can be detected by electromyography, EMG), or reproduced

solely ‘in the head’ of the perceiver (i.e. activation of the motor system as detected by

fMRI). As discussed in Chapter 3, the candidate neural mechanism for this perception-to-

action coupling is the mirror neuron system. Mirror neurons respond both when an

animal performs an action and when it observes another performing the same (or similar)

action. They act as a neural bridge between self and other, and it has been suggested that

their capacity to support imitation provides the foundation for some aspects of social

cognition such as empathy (e.g. Iacoboni, 2009). A link between imitation and empathy

receives some support from social psychology in studies examining the Chameleon

Effect in which there is a spontaneous mimicry of gestures during positive inter-personal

exchanges. These studies generally use unintentional imitation in which the participant

engages in a task with another person (a confederate) and the extent to which the

participant imitates the confederate is measured. The participant is unaware of the true

nature of the study (i.e. that his/her imitative behavior is being assessed). Participants

who imitate more (based on blind scoring of their actions) whilst performing a

cooperative task with a confederate tend to rate themselves as higher in trait empathy

(Chartrand & Bargh, 1999). When the confederate deliberately imitates the participant in

a cooperative task, then he/she is liked more by the participant than in a control condition

in which imitation is avoided (Chartrand & Bargh, 1999). Van Baaren, Holland,

Kawakami, and van Knippenberg (2004) showed that being imitated increases the

chances of helping behavior when a confederate drops something. However, the effects

are quite general. The person who has been imitated is not just more likely to help the

imitator but they are more likely to help others too. It also increases the amount of money

that the participant opts to donate to charity at the end of the experiment.

The activity of the human motor system can be assessed using the method of

motor evoked potentials (MEPs). Stimulation of the brain using TMS causes the

peripheral muscles to produce neuroelectrical signals known as motor evoked potentials

(MEPs) – see Figure 6.4. These can be measured by electrodes attached to the skin using

the principle of electromyography (EMG). TMS is applied over the primary motor cortex

whilst the participant is at rest and the TMS threshold is found below which EMG

responses in the muscles can no longer be reliably elicited. Asking the participant to

perform a voluntary movement or simply observing the action of another person

increases cortical excitability, defined as an increased MEP when this threshold level of

TMS is applied (Fadiga, Fogassi, Pavesi, & Rizzolatti, 1995). This is specific to the

action observed such that observing an arm action (flexing at the elbow) facilitates MEPs

on the biceps but not the hand muscles and observing a hand action (writing) facilitates

MEPs in the hand but not the biceps (Strafella & Paus, 2000). Observing another person

in pain, such as an injection applied to the hand, results in cortical inhibition as revealed

by reduced MEPs (Avenanti, Bueti, Galati, & Aglioti, 2005). Although such results may

seem distantly related to the notion of empathy, they provide further evidence for the

notion of mirroring. Moreover, MEPs are modulated by social factors such as

questionnaire measures of empathy (Avenanti, Minio-Paluello, Bufalari, & Aglioti,

2009), the race of the hand observed (Avenanti, Sirigu, & Aglioti, 2010), and priming to

think of oneself as having more or less power over others.

Some theories of empathy propose a variety of different interacting mechanisms

of which simulation is only one. In such models, simulation may either be a junior or

senior partner. As noted above, watching someone in pain activates certain parts of our

own pain circuitry. This offers clear support for simulation theories. However, our beliefs

about the person in pain can modulate or override this mechanism. Singer et al. (2006)

had participants in an fMRI scanner play a game with someone who plays fairly (a

‘Goodie’) and someone else who plays unfairly (a ‘Baddie’). Mild electric shocks were

then delivered to the Goodie and Baddie (who, of course, were only actors but the

participant did not know this). Participants empathically activated their own pain regions

(such as anterior insula) when watching the Goodie receive the electric shock (see Figure

6.5). However, this response was attenuated when they saw the Baddie receiving the

shock. In fact, male participants often activated their pleasure and reward circuits (such

as the nucleus accumbens) when watching the Baddie receive the shock, which is the

exact opposite of simulation theory. This brain activity correlated with their reported

desire for revenge, which suggests that although simulation may tend to operate

automatically it is not protected from our higher order beliefs.

Other studies support this view. Although doctors may be expected to show

empathy for their patients, it would be unhelpful for them to experience personal distress

when performing painful procedures. Indeed acupuncturists show less activity, measured

by fMRI, in the pain network (including the anterior insula and anterior cingulate) when

watching needles inserted into someone, relative to controls (Cheng et al., 2007). Lamm,

Batson, and Decety (2007) found that activity in these painrelated regions, induced by

watching painful facial expressions induced by medical treatment, was modulated by the

observer’s beliefs about whether the treatment was successful or not (more activity in

pain-processing regions when less successful). It was also related to whether the

participants were instructed to imagine the feelings of the patient or to imagine

themselves in that situation (more activity in pain-processing regions when imagining

self). This suggests that the tendency to simulate is moderated by cognitive control (e.g.

based on our beliefs) and also our efforts to take different perspectives.

Models of empathy based solely on the notion of perception–action coupling or

affective sharing have been shown to be lacking, but may nevertheless be one aspect of

empathy. So what do alternative models of empathy look like? This section will consider

three kinds of models: those that conceptualize empathy as an interaction, or trade-off,

between mirroring and mentalizing; those that make a distinction between affective

versus cognitive empathy; and those that make a distinction between emotion sharing

versus emotion regulation. Zaki and Ochsner (2012) consider empathy as a product of

two kinds of mechanism – mirroring versus mentalizing. The extent to which one

mechanism may dominate over the other is assumed to be dynamic and may depend on

what the expected outcome is. Merely observing another person in a decontextualized

setting may bias towards mirroring. Deciding whether to act prosocially (or otherwise)

towards another person may involve some interplay between the two mechanisms, and

may differ from person to person. According to this framework, in the experiment of

Singer et al. (2006) the tendency to simulate another’s pain would be part of the

mirroring system, and the representation of the other’s intentions (to play fairly or

unfairly) would be part of the mentalizing (or theory of mind) system.

An alternative distinction to mirroring versus mentalizing is the proposed

distinction between cognitive empathy and affective empathy (e.g. Baron-Cohen &

Wheelwright, 2004; Shamay-Tsoory, Aharon-Peretz, & Perry, 2009) In theory, the

cognitive/affective distinction can be regarded as separate to the mirroring/mentalizing

distinction. One can attribute mental states to others that are either affective in nature

(e.g. ‘John is angry’) or non-affective in nature (e.g. ‘John thinks X’). Similarly,

mirroring can either be affective in nature (emotional contagion) or non-affective in

nature (imitation/mimicry). In practice, the extent to which the mirroring/ mentalizing

and affective/cognitive distinctions are related or separate remains a matter of debate. For

instance in the model of Shamay-Tsoory (2011) the mirroring mechanism is assumed to

be common for both actions and emotions, but separate mentalizing components are

postulated for affective and non-affective mental states. However, others argue that

mirror neurons should be understood solely in terms of motor actions.

E. Theory of Mind and Reasoning About Mental States

The term ‘theory of mind’ derived originally from research on primate cognition.

Premack and Woodruff (1978) conducted a number of studies on a chimpanzee to see if it

understood an experimenter’s intentions. For example, the chimp might point to a picture

of a key when an experimenter was locked in a cage, the inference being ‘he wants to get

out’. A number of criticisms were leveled at the study. For instance, it may reflect

knowledge of object associations (e.g. between lock and key) rather than mental states. In

a reply to the article, Dennett (1978) suggested that one way of testing for theory of mind

would be to consider false beliefs, in which someone else may hold a mental state (e.g. a

belief) that differs from one’s own belief and from the current state of reality. In

developmental psychology, the paradigmatic false belief test is the object transfer task,

such as the Sally–Anne task shown in Figure 6.8 (Baron-Cohen, Leslie, & Frith, 1985;

Wimmer & Perner, 1983). Sally puts a marble in a basket so that Anne can see. Sally then

leaves the room, and Anne moves the marble to a box. When Sally enters the room, the

participant is asked ‘Where will Sally look for the marble?’ or ‘Where does Sally think

the marble is?’ A correct answer (‘In the basket’) is typically taken to indicate the

presence of a theory of mind. An incorrect answer is potentially more problematic to

interpret. It could imply a lack of theory of mind. However, one also has to rule out other

factors such as language comprehension difficulties or a failure to inhibit a more

dominant response (one’s own belief).

False beliefs are harder to accommodate within simulation theories because one’s

own belief is at odds with that attributed to the other person. This cannot be done by

straightforward simulation involving shared self–other representations. It requires taking

one’s own mental states ‘offline’ and creating a hypothetical scenario different to current

reality. So-called meta-representation and pretense is often regarded as a hallmark of

theory-of-mind ability (Leslie, 1987). Social psychologists use the term attribution to

refer to the process of inferring the causes of people’s behavior. The philosopher Dennett

(1983) uses his own term of intentional stance to refer to our tendency to explain

behavior in terms of mental states, which could otherwise be considered synonymous

with mentalizing or theory of mind. However, Dennett (1983) has a particularly useful

way of describing different levels of intentionality that might be used to account for

behavior. For example, an observer might have to evoke zero-order intentionality to

explain the behavior of an object, first-order intentionality to explain the behavior of

some animals, and second-order intentionality to explain some human behavior.

Domain specificity is linked to the notion of modularity (Fodor, 1983). A

cognitive mechanism, or brain region, can be said to be domain specific if it is

specialized to process only one kind of information. Thus, a domain-specific theory-of-

mind mechanism would be a process that is specialized for attributing mental states

(Leslie, 1987). There are two dominant lines of evidence that have been brought to bear

on this. First, there is the question of whether there is a specific region of the brain that

responds to reasoning about mental states but not other kinds of things. It is possible that

such a mechanism could be distributed in several locations, or that only one of the

regions in that network is truly domain specific. Second, one can look to see if there are

specific impairments in mental state attribution but not in other domains. Most evidence

related to this question has come from the developmental condition of autism (e.g. Baron-

Cohen, 1995b), but other lines of research have addressed this question from the

perspective of acquired brain damage.

Historically, explanations of theory of mind have fallen into two camps that are

termed theory-theory and simulation theory. Theory-theory argues that we store, as

explicit knowledge, a set of principles relating to mental states and how these states

govern behavior (e.g. Gopnik & Wellman, 1992). In this sense, the ‘theory’ in theory of

mind is like a mental rulebook for understanding others. This can be contrasted with

simulation theory, which in one form would argue that perceptual-motor systems (rather

than thinking and theorizing) are all that is needed for understanding others (e.g. Gallese

& Goldman, 1998). When phrased in this way, it is reasonable to say that theory-theory

makes more domain-specific assumptions whereas simulation theory can be considered a

domain-general account. However, one needs to be cautious in dividing explanations into

black and white dichotomies. For example, some versions of simulation theory argue that

we do reason about mental states (rather than it being solely an outcome of perceptual-

motor processes) but these versions are distinguished from theory-theory by making the

claim that our own mental states form the foundation for understanding others.

Whilst the idea of a domain-specific mechanism for theory of mind is

controversial, the idea that theory of mind requires basic competency in a number of

domain-general mechanisms such as executive functions is not controversial, and a basic

competency in language is required for many tasks. Language ability in typically

developing children predicts success on a false belief task independently of age (Dunn &

Brophy, 2005), and deaf children whose parents are non-native signers are delayed in

passing such a task (Peterson & Siegal, 1995). This suggests that language is important

for the development of theory of mind. Language may serve several functions: both a

social, communicative role and also the acquisition of semantic knowledge of mental

state words such as ‘want’ and ‘think’. For example, children have to learn that these

words denote concepts that are privately held (Wellman & Lagattuta, 2000). However,

once a normal theory of mind is established it may not be dependent solely on language.

Evidence for this assertion comes from braindamaged patients with acquired aphasia.

Apperly, Samson, Carroll, Hussain, and Humphreys (2006) report a single case study of a

man with left hemisphere stroke who was impaired in many aspects of language,

including syntax comprehension, but showed no impairments on non-verbal tests of

theory of mind, including secondorder inferences (X thinks that Y thinks).

Evidence for the neural basis of theory of mind has come from two main sources:

functional imaging studies of normal participants and behavioral studies of patients with

brain lesions. Numerous tasks have been used, including directly inferring mental states

from stories (e.g. Fletcher et al., 1995), from cartoons (e.g. Gallagher et al., 2000), or

when interacting with another person (e.g. McCabe, Houser, Ryan, Smith, & Trouard,

2001a). A review and meta-analysis of the functional imaging literature was provided by

Frith and Frith (2003), who identified three key regions involved in mentalizing: the

temporal poles, the temporo-parietal junction, and the medial prefrontal cortex. More

recent reviews reveal a similar general pattern but with some important differences

depending on the nature of the theory-of-mind test that is used.

Brain damage to the temporal poles is a feature of the degenerative disorder

known as semantic dementia (Mummery et al., 2000). Patients with semantic dementia

lose their conceptual knowledge of words, objects, and people and show difficulties in

language comprehension and production. However, there is little evidence from these

patients that social concepts are selectively impaired. In general, although the temporal

poles are important for theory of mind, there is no convincing support that it is domain

specific for this kind of information and they are likely to provide more domain-general

input in the form of social knowledge. Frith and Frith (2003) reported that this region,

shown in Figure 6.11, is activated in all functional imaging tasks of mentalizing to that

date, and it still features prominently in a more recent meta-analysis (Schurz et al., 2014).

Saxe (2006) argues that a sub-region of this area is involved in ‘uniquely human’ aspects

of social cognition. This region lies in front of, but extends into, the ventral region of the

anterior cingulate, labeled by Bush et al. (2000) as the affective division. Functional

imaging studies reliably show that this region responds more to: thinking about people

rather than thinking about other entities such as computers or dogs (e.g. Mitchell, Banaji,

& Macrae, 2005a; Mitchell, Heatherton, & Macrae, 2002); thinking about the minds of

people rather than thinking about their other attributes, such as their physical

characteristics (Mitchell et al., 2005b); thinking about the minds of people who are

similar to ourselves (Mitchell et al., 2005b); and thinking about social groups who are

humanized relative to dehumanized.

Can a generic function be ascribed to this region? If so, how does it relate to

theory of mind? Amodio and Frith (2006) argue that the function of this region is in

reflecting on feelings and intentions, which they label a ‘meeting of minds’. One

intriguing finding concerning this region is that it can be activated when a person believes

they are playing a computer game against another person relative to when they think they

are playing against a computer (Rilling, Sanfey, Aronson, Nystrom, & Cohen, 2004).

Even though the situation is physically identical (the participant always played the

computer), the act of cooperating with another person/mind engenders activity in this

region. A more recent explanation of the function of this region is similar to, but different

from, that of Amodio and Devine (2006). Krueger, Barbey, and Grafman (2009) argue

that the function of this region is to bind together different kinds of information (actions,

agents, goals, objects, beliefs) to create what they term a ‘social event’. They note that

within this region some sub-regions respond more when participants make judgments

about themselves and also about others who are considered to be similar to themselves

(this is discussed in detail in Chapter 9). This suggests that this region is not attributing

mental states per se, but is considering the self in relation to others (e.g. when playing a

game against a human rather than a computer). It is also consistent with some versions of

simulation theory in which participants understand others by reflecting on how we would

react in this scenario (e.g. Mitchell et al., 2005b). The notion of creating internal social

events could also explain some of the findings of the role of this region in linking ideas in

story comprehension.

The TPJ region was previously highlighted in the discussion on empathy because

it responds more when participants are asked to imagine how someone else would feel

relative to how they would feel (e.g. Ruby & Decety, 2004). Patients with brain lesions in

this region fail theory-of-mind tasks that cannot be accounted for by difficulties in body

perception (Samson, Apperly, Chiavarino, & Humphreys, 2004). Saxe and Kanwisher

(2003) found activity in this region, on the right, when comparing false belief tasks

(requiring mentalizing) with false photograph tasks (not requiring mentalizing but

entailing a conflict with reality). A false photograph may involve taking a picture of an

apple on the tree, and then the apple falling down. In this scenario, there is a conflict

between reality and a representation of reality. The result was also found when the false

photograph involved people and actions, consistent with a role in mentalizing beyond any

role in action/person perception. The region responds to false beliefs more than false

maps or signs, which differ in an important way from a false photograph in that they are

designed to represent current reality.

Saxe and colleagues do not dismiss the fact that this region has a role to play in

recognizing people and actions, but they claim that there may be different sub-regions

within it, with one subregion specialized for the attribution of mental states as shown in

Figure 6.12 (Scholz, Triantafyllou, Whitfield-Gabrieli, Brown, & Saxe, 2009). Moreover,

Saxe (2006) argues that it is uniquely human in doing so. It is important to note that this

region is not specialized for false belief per se. It responds to true beliefs and other types

of mental states (Saxe & Wexler, 2005). In other words, it responds to attributions of

first-order intentionality as well as higher order intentionality (in Dennett’s terms). Saxe

and Powell (2006) have shown that this region responds to attribution of contentful

mental states (such as thoughts and beliefs) rather than subjective states (such as hunger

or tiredness). This suggests that it may have a role over and above ‘thinking about

others’. However, it is important to mention that one should be cautious in making strong

claims about relative differences in BOLD signal. The differences can reflect different

functional specialization (Saxe’s claim), but they can also reflect the different difficulty

of tasks, and the attention or strategy deployed to solve them. Other accounts propose

more general functions to the TPJ that are relevant to theory of mind but without

assuming it represents mental states as such. It may serve a general functioning of

orienting of attention to a stimulus (Corbetta & Shulman, 2002; Mitchell, 2008) that, in

social terms, may include orienting attention to other people and away from the self.

F. Explaining Autism

Autism has been formally defined as ‘persistent deficits in social communication

and social interaction across multiple contexts’ (DSM-V, American Psychiatric

Association, 2013). It is a severe developmental condition that is evident before 3 years

of age and lasts throughout life. There are a number of difficulties in diagnosing autism.

First, it is defined according to behavior because no specific biological markers are

known, although there are some known associations (see ‘Biological markers for autism’

box). Second, the profile and severity may be modified during the course of development.

It can be influenced by external factors (e.g. education, temperament) and may be

accompanied by other disorders (e.g. attention deficit and hyperactivity disorder,

psychiatric disorders). As such, autism is now viewed as a spectrum of conditions

spanning all degrees of severity. It is currently believed to affect 1.2% of the childhood

population, and is three times as common in males (Baird et al., 2006). Asperger’s

syndrome falls within this spectrum, and is often considered a special sub-group. The

diagnosis of Asperger’s syndrome requires that there is no significant delay in early

language and cognitive development, although the term is also used to denote people with

autism who fall within the normal range of intelligence. Learning disability, defined as an

IQ lower than 70, is present in around half of all cases of autism.

Much of the behavioral data has been obtained from high-functioning individuals

in an attempt to isolate a specific core of deficits. On a purely theoretical level, one

reason why researchers have been interested in the study of autism is the belief that it

might reveal something fundamental about social interactions more generally. One

candidate deficit is the ability to represent mental states, or theory of mind (e.g. Baron-

Cohen, 1995b; Fodor, 1992). The first empirical evidence in favor of this hypothesis

came with the development of a test of false belief devised by Wimmer and Perner

(1983) and tested on autistic children by Baron-Cohen et al. (1985) as the Sally–Anne

task (described above). Autistic children tend to fail the task whereas normally

developing children (from 4 years on) pass the test, as do control participants with

learning disability matched in IQ to the autistic children. The erroneous reply is not due

to a failure of memory, because the children can remember the initial location. It is as if

they fail to understand that Sally has a belief that differs from physical reality – that is, a

failure to represent mental states. This has also been called ‘mind-blindness’ (Baron-

Cohen, 1995b). Autistic children are still impaired when the false belief was initially their

own. For example, in one task, the child initially expects to find candy in a candy packet

and is surprised to find a pencil, but when asked what other people will think is in the

packet the child replies ‘pencil’.

Passing false belief tasks requires the ability to form meta-representations (i.e.

representations of representations: in this instance, beliefs about beliefs). It was originally

suggested that a failure of meta-representation may account for impaired theory of mind

in autism (Baron-Cohen et al., 1985). However, other studies suggest that autistic people

can form meta-representations in order to reason about false photographs in which the

information depicted on the photograph differs from current reality (Leekam & Perner,

1991). If their deficit really is related to mental state representations rather than physical

representations, then this offers support for the domain-specific account. A number of

other studies have pointed to selective difficulties in mentalizing compared to carefully

controlled conditions. For example, people with autism can sequence behavioral pictures

but not mentalistic pictures (Baron-Cohen, Leslie, & Frith, 1986); they are good at

sabotage but not deception – they tend to think that everyone tells the truth (Sodian &

Frith, 1992); and they tend to use desire and emotion words but not belief and idea words

(Tager-Flusberg, 1992). In all instances, the performance of people with autism is

compared to mental-age controls to establish that the effects are related to autism and not

to general level of functioning.

Finally, it may be necessary to make a distinction between implicit mentalizing

(intuitive, reflexive) and more explicit forms of mentalizing (based on reasoning), as

occurs in some two-system models of theory of mind (e.g. Apperly, 2011). Whilst

explicit theory of mind tends to be measured by overt predictions of behavior (as in the

Sally–Anne task), the former may be measured by non-declarative means (e.g.

monitoring of eye movements). For example, some high-functioning people with autism

pass standard theory-of-mind measures but may still lack an intuitive understanding of

others and may still show abnormal performance on other measures (e.g. eye movements

to a location consistent with a false belief; Senju, Southgate, White, & Frith, 2009). By

contrast, children under the age of 4 years show some implicit understanding of false

beliefs (based on the same measure) despite failing on explicit measures.

As discussed throughout the chapter, empathy can be measured in many ways and

can almost certainly be broken down in to different kinds of cognitive and neural

mechanisms. As such, we need to think carefully about how we answer a question such

as ‘Do people with autism lack empathy?’. On questionnaire measures of empathy,

people with autism tend to score lower, implying that they do have less empathy (e.g.

Baron-Cohen & Wheelwright, 2004). However, most of these measures ask about high-

level aspects of empathy (e.g. such as thinking about the feelings of others), that most

likely tap the mentalizing network, rather than more feeling-based aspects of empathy.

Frith (2012) argues that an impairment of mentalizing need not reflect an inability to

resonate with the feelings of others. When observing other people in pain, people with

autism show a similar brain response in the pain matrix (Hadjikhani et al., 2014). They

show normal spontaneous facial mimicry of emotional expressions measured with EMG

(Deschamps, Coppes, Kenemans, Schutter, & Matthys, 2015), and normal modulation of

motor-evoked potentials (MEPs) in response to TMS when viewing hand actions,

including during social situations.

Many people with autism do, however, have problems in labeling their own

feelings – a symptom that is known as alexithymia (literally ‘no words for feelings’)

(Hill, Berthoz, & Frith, 2004). For instance, people with alexithymia tend to agree with

statements such as ‘I don’t know if I am feeling tired or angry’. This could be interpreted

as a difficulty in reflecting on one’s own mental states (an emotional feeling is a mental

state) as well as those of other people. Several studies have contrasted brain responses to

emotional or social stimuli according to whether participants have autism, alexithymia,

both autism and alexithymia, or neither. Silani et al. (2008) showed images with different

levels of emotional content, and participants rated how unpleasant it made them feel or

how colorful the images were. All participants showed activity in the amygdala linked to

their level of emotional arousal (i.e. no evidence of impairment in basic emotional

intensity). Whereas alexithymia was linked to reduced activity in the anterior insula

(assumed to reflect poor awareness of specific bodily based feelings), autism was linked

to reduced activity in regions implicated in mentalizing when rating their emotions.

The mentalizing or theory-of-mind account of autism has not been without its

critics. These criticisms generally take two forms: that other explanations can account for

the data without postulating a difficulty in mentalizing (e.g. Russell, 1997); or that a

difficulty with mentalizing is necessary but insufficient to explain all of the available

evidence (e.g. Frith, 1989). A number of studies have argued that the primary deficit in

autism is one of executive functioning (Hughes, Russell, & Robbins, 1994; Ozonoff,

Pennington, & Rogers, 1991; Russell, 1997). Executive functions refer to control

processes that are needed to coordinate the operation of more specialized components of

the brain, thus enabling us to switch attention from one task to another, to give priority to

certain kinds of information, or to develop novel solutions, which would include

inhibiting familiar solutions (e.g. Goldberg, 2001). For example, the incorrect answer

might be chosen on false belief tasks because of a failure to suppress the strongly

activated ‘physical reality’ alternative. Some patients with brain damage in prefrontal

regions do this when given false belief tasks (Samson, 2009). However, it is not clear that

this explanation can account for all the studies relating to mentalizing (e.g. picture

sequencing). Moreover, high-functioning autistic people often have normal executive

functions (e.g. Baron-Cohen, Wheelwright, Stone, & Rutherford, 1999) and early brain

lesions can selectively disrupt theory-of-mind abilities without impairing executive

functions.

Baron-Cohen (2002, 2009) argues that the characteristics of all individuals can be

classified according to two dimensions: ‘empathizing’ and ‘systemizing’. Empathizing

allows one to predict a person’s behavior and to care about how others feel. Systemizing

requires an understanding of lawful, rule-based systems and requires an attention to

detail. Males tend to have a brain type that is biased towards systemizing (S > E) and

females tend to have a brain type that is biased towards empathizing (E > S). However,

not all men and women have the ‘male type’ and ‘female type’, respectively. Autistic

people appear to have an extreme male type (S >> E), characterized by a lack of

empathizing (which would account for the mentalizing difficulties) and a high degree of

systemizing (which would account for their preserved abilities and unusual interests).

Questionnaire studies suggest that these distinctions hold true (Baron-Cohen, Richler,

Bisarya, Gurunathan, & Wheelwright, 2003; Baron-Cohen & Wheelwright, 2004).

However, these distinctions are very broad, and it should be borne in mind that, for

example, people with autism may only be impaired on particular kinds of measures

relating to empathy. How does the extreme male brain hypothesis relate to other theories

of autism? Baron-Cohen (2002, 2009) regards this explanation as an extension of the

earlier mind-blindness theory, which has the advantage of being able to incorporate

additional data. Specifically, it accounts for some of the non-social differences found in

autism, and it offers an explanation for why autism is more common in men (i.e. because

men are more likely to have S > E type brains). However, there are at least two ways in

which these different ideas (mind blindness vs extreme male brain) could be related: that

an inability to engage with others (due to a theory-of-mind deficit) leads to systemizing

as a kind of compensatory strategy; or that an unusual interest or ability in systemizing

leads to a lack of interest and understanding of social behavior. A third possibility is that

both are true – that whatever it is that causes high systemizing also causes low

empathizing. Possible mechanisms include fetal testosterone levels (e.g. Auyeung et al.,

2009) or sex-related genetic differences (e.g. Creswell & Skuse, 1999). Although the

extreme male brain theory predicts an autistic advantage for understanding systems, it

differs from the weak central coherence theory by not making predictions about a

difference between local versus global information.

The broken mirror theory of autism argues that the social difficulties linked to

autism are a consequence of mirror system dysfunction (Iacoboni & Dapretto, 2006;

Oberman & Ramachandran, 2007; Ramachandran & Oberman, 2006; Rizzolatti &

Fabbri-Destro, 2010). Hadjikhani, Joseph, Snyder, and Tager-Flusberg (2006) examined,

using structural MRI, the anatomical differences between the brains of autistic

individuals and matched controls. The autistic individuals had reduced gray matter in

several regions linked to the mirror system, including the inferior frontal gyrus (Broca’s

region), the inferior parietal lobule, and the superior temporal sulcus. Although these

were not the only regions where differences were found, the degree of thinning in these

regions correlated with autistic symptom severity. EEG and fMRI data also suggest

differences in mirror system functioning during certain tasks. Oberman et al. (2005) used

EEG to record mu waves over the motor cortex of high-functioning autistic children and

controls. Mu waves occur at a particular frequency (8–13 Hz) and are greatest when

participants are doing nothing. However, when they perform an action there is a decrease

in the number of mu waves, a phenomenon termed mu suppression. Importantly, in

typical controls mu suppression also occurs when people observe actions and, as such, it

has been regarded by some as a measure of mirror system activity (e.g. Pineda, 2005).

Oberman et al. (2005) found that the autistic children failed to show as much mu

suppression as controls during action observation (watching someone else make a pincer

movement) but did so in the control condition of action execution (they themselves make

a pincer movement).

In general, the criticism takes two forms. First, it does not account for all the

unusual behavior found in autism (e.g. embedded figures; interest in systems).

Defendants of the theory argue that it is not trying to explain all the features of autism

(i.e. it is not a theory of autism but a theory of certain characteristics of autism). The

second general criticism surrounds the extent to which empathy and imitation are linked

to mirror systems. Certain forms of emotional empathy appear normal in autism

(Hadjikhani et al., 2014) and the extent to which the mirror system supports emotional

empathy is not clear (de Vignemont & Singer, 2006). A core deficit elsewhere (e.g. in

representing mental states) could nevertheless affect the functioning of the mirror system

and perhaps even lead to structural changes within that system. Heyes (2010) argues that

the properties of mirror neurons may be learned as a result of social interactions. Thus,

impoverished social interactions may cause mirror system dysfunction, as well as vice

versa.

For many years the dominant explanation of autism has been a failure to represent

the mental states of others. This has been termed mind blindness and has tended to be

regarded as a failure to develop a theory of mind (although not necessarily with

commitment to the idea that this exists as a domain-specific module). Other theories,

such as weak central coherence theory and extreme male brain theory, maintain this basic

idea but adopt a wider perspective in order to explain other features of autism. The most

significant challenge to this idea previously came from the notion of executive

dysfunction in autism, but now comes in the form of broken mirror theory. There is good

evidence of mirror neuron dysfunction in autism, but it is less clear whether this

dysfunction is a core feature of autism or a by-product of other deficits – given that

mirror systems in general are modulated by beliefs, social knowledge, and cognitive

control.

Module 3_Discussion 3_Emotional Processing and Autism_Social Cognitive Neuroscience

Students also viewed