Jump to ContentJump to Main Navigation
The Measure of MadnessPhilosophy of Mind, Cognitive Neuroscience, and Delusional Thought$

Philip Gerrans

Print publication date: 2014

Print ISBN-13: 9780262027557

Published to MIT Press Scholarship Online: September 2015

DOI: 10.7551/mitpress/9780262027557.001.0001

Show Summary Details
Page of

PRINTED FROM MIT PRESS SCHOLARSHIP ONLINE (www.mitpress.universitypressscholarship.com). (c) Copyright The MIT Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in MITSO for personal use (for details see http://www.mitpress.universitypressscholarship.com/page/privacy-policy). Subscriber: null; date: 21 June 2018

Dreaming, Default Thinking, and Delusion

Dreaming, Default Thinking, and Delusion

(p.89) 5 Dreaming, Default Thinking, and Delusion
The Measure of Madness

Philip Gerrans

The MIT Press

Abstract and Keywords

This chapter examines the nature of default processing, with particular emphasis on its outputs in different modes: dreaming, mental time travel, and delusion. It explains how the similarity between dreaming and delusion results from the fact that they are both states characterized by activity in the default mode network unsupervised by decontextualized processes. It offers a mechanistic and a cognitive explanation of similarities and dissimilarities between delusions and dreams. The mechanistic explanation adapts the AIM (Activation, Information, Mode) model of dreams and delusions developed by Hobson and colleagues, whereas the cognitive explanation shows how automatic and controlled processes degrade in different ways when unsupervised in virtue of differences in cognitive architecture. The cognitive explanation distinguishes feature binding from context binding. The chapter considers how the phenomenology of dreaming, mental time travel, and delusion is affected by neurobiological processes and suggests that delusion is a mixed mode of cognition that accounts for its belief-like and non-belief-like properties. Finally, it presents a case study of delusions of misidentification.

Keywords:   default processing, dreaming, mental time travel, delusion, default mode network, dreams, feature binding, context binding, cognition, misidentification

In this chapter we pursue the idea that the often-remarked, but not well-explained, similarity between dreaming and delusion results from the fact that they are both states characterized by activity in the default system unsupervised by decontextualized processes. We give a mechanistic and a cognitive explanation of similarities and dissimilarities between delusions and dreams. The mechanistic explanation adapts the AIM (Activation, Information, Mode) model of dreams and delusions developed by Hobson and collaborators. The cognitive explanation shows how automatic and controlled processes degrade in different ways when unsupervised in virtue of differences in cognitive architecture.

That explanation distinguishes feature and context binding. Binding here refers to the integration of elements of a representation. Feature binding refers to the construction of an integrated perceptual or quasi-perceptual representation and contextual binding to the organization of such representations into a coherent metacognitive structure such as a narrative or theoretical explanation. Feature binding is a relatively modular process implemented in localized neural circuits whose processing properties are fairly rigid. In general, feature binding is a (p.90) task performed by weight-based processing systems that automatically assemble representational elements according to procedures standardized by evolution or in development. Because these processes have fairly rigid architectures, they tend not to disintegrate unless their neural substrates are damaged. Context binding is intrinsically a more unstable process because it is by its nature flexible and open-ended. It requires active maintenance of transient distributed circuits and constant top-down supervision.

One form of context binding relevant to delusions is the binding of simulations into the narratives and narrative fragments of mental time travel. Delusions are a form of context binding in which subjects produce a default thought subjectively adequate to experience. That thought is not the result of an attempt to confirm an empirical hypothesis. It is an untested (because unsupervised) default thought. We develop these ideas by discussing an example of delusion-like phenomena in dreams—the experience of hyperfamiliarity—and suggest that it results from a similar feature-binding anomaly to one that occurs in a waking state in the Fregoli delusion. In dreams, however, this anomaly does not lead to delusion because in dreaming the default network is not configured to produce a narrative context for such anomalous experiences.

5.1 Dreaming and the Default Mode Network

The default system evolved as the first stage in the transformation of automatism to agency. It allows humans to rehearse alternatives before committing to action. These rehearsals typically take a narrative form. The production and consumption of these narratives is in turn supervised by a higher level of decontextualized (p.91) cognitive processing. This supervision can take the form of testing narrative elements for consistency and veridicality or evaluation of competing narratives for accuracy or utility.

When dorsolateral systems required to represent high-level goals and evaluate default narratives against them are inactive, the default network reverts to its default state: the production of subjectively adequate narratives. When there is no goal or end point for the narrative, it tends to degrade into the random association of default representations. This is sometimes put in terms of the absence of attentional control in default thinking. A simulation process that starts with a specific goal may meander as top-down attentional control wanes. It is for this reason that dreaming represents a state of unsupervised default processing. As well as the absence of top-down control, the default processing in dreaming is characterized by the absence of organizing input from the sensory periphery, which leaves the default net-work entirely at the mercy of endogenous activation by subcortical inputs:

Dreams can be seen as a unique and more fully developed form of mind-wandering, and therefore as the quintessential cognitive simulation. They are the quintessential cognitive simulation not only because they have elaborate storylines that are often enacted with exquisite sensory, motor, and cognitive involvement, with some dreams unfolding over a period of several minutes to half an hour or more. There is also the striking fact that they are usually experienced as real while they are happening. (Domhoff 2011, 1172)

This is one reason that people have drawn a parallel between dreaming and delusion. John Nash, for example, said of his delusional states: “It's kind of like a dream. In a dream it's typical not to be rational” (Nash n.d.). While this is correct, it slightly mischaracterizes the phenomenon. Some dream reports include (p.92) episodes of coherent thought, though such episodes are transient and not coherent with other episodes of the dream. Equally, some delusions and dream episodes can include episodes of rational transition between thoughts that are bizarre and false. Paranoid delusions, for example, form coherent and sometimes logical narratives given the thoughts that trigger them.

For this reason, the similarity between dreams and delusions is sometimes put in terms of absent or compromised “reality testing” (Bentall 2004). For example, in 1949 Paul Federn wrote that “the basis of sanity is correct and automatic recognition of [the] breach between subjective mental experiences in the world and the knowledge of the status of the world as it actually exists” (Federn 1953, 229).

“Reality testing” is a slight term of art and clearly can include evaluating propositions for truth or experience for veridicality. But John Nash, while remaining capable of high-level mathematical reasoning, never thought to question the reality of his delusion that the Pope and CIA were part of a conspiracy against him. So the term really refers to the suspension or absence, partial or complete, of the ability to detect and resolve inconsistency between the information represented in an episode of thought and the background knowledge about the world, normally available to be brought to bear in alert, waking cognition.

One reason for taking the cognitive approach to the phenomenon is that it avoids a controversy over the exact nature of reality testing. Sometimes it is tested in paradigms that in effect ask the subject to reason about possible causes of experience. A version of this paradigm is employed in source-monitoring experiments that compare the ability of delusional and nondelusional subjects to estimate the probability that an experience has endogenous or exogenous origin (Mitchell and Johnson 2000; (p.93) Moritz, Woodward, and Ruff 2003). Naturally these paradigms are congenial to the Bayesian model, which treats source monitoring as the detection of prediction error.

While remaining neutral about the implementation of Bayesian algorithms, we can note that there is clear evidence for the following:

  1. (i) A role for dorsolateral circuitry in the supervision of default thinking

  2. (ii) Hypoactivity or lesion of that circuitry in conditions such as delusion

  3. (iii) Deactivation of that circuitry in dreams

  4. (iv) Activity or hyperactivity of the default circuitry in both delusions and dreams

Two well-known dream theorists explain the resultant phenomenology as follows, also providing an explanation of the personal-level consequences:

Much of wake-learned knowledge is not accessible or even applicable to the largely self-contained dream narrative. This alleged inability to exchange information is buttressed by experiments which have shown that there is a decrease in DLPFC with orbital prefrontal exchange specifically during rapid eye movement (REM) stage when dreaming is pronounced and a decrease in information exchange of DLPFC with parts of the occipital, parietal, and temporal lobes. (Kahn and Hobson 2003, 65)

Kahn, Pace-Schott, and Hobson make a similar point:

When the DLPFC is in poor communication with these areas as in REM sleep, the ability to perform logical inference, to recall accurately and to discern whether a premise is fact or fiction may very well be compromised. (2002, 46)

The most extreme (and congenial to the thesis developed here) attempt is by Gottesmann, who explicitly situates his (p.94) neurobiological account of dreaming and schizophrenic psychosis in a framework that treats the mind as a control hierarchy with decontextualized supervision at the top. He explains the properties of the dream state in terms of the absence of top-down control due to the deactivation of the dorsolaterally regulated supervisory systems in combination with the absence of perceptual and sensory input:

The dorsolateral prefrontal deactivation observed both during REM sleep and in schizophrenia seems to suppress or decrease its own functions, including the loss or decrease of reflectiveness, and at the same time disinhibits older subcortical structures and corresponding functions, with the exaggeration of accumbens' and amygdala nuclei's own processes: in our case, the appearance of hallucinations, delusions, bizarre thought processes, and affective disturbances. (Gottesmann 2006, 1113)

Thus there is consensus that what we have called decontextualized supervision is absent in dreams due to the deactivation of the dorsolateral prefrontal circuitry on which it depends: “REM sleep may constitute a state of generalized brain activity with the specific exclusion of executive systems that normally participate in the highest order analysis and integration of neural information” (Braun et al. 1997, 1190).

We first give a version of the neurochemical account of dreaming (the AIM model) and delusion before marrying it to the binding theory of cognitive coherence. Although the AIM model has generated controversial interpretations (for example, the relationship between REM sleep, dream experience, and activity in specific brainstem structures), we rely here only on uncontroversial theses about the mechanisms involved in context binding. We then discuss a concrete example, using the Fregoli delusion, of the way in which the account can work to explain similarities between dreams and delusions.

(p.95) 5.2 The AIM Model

The AIM model of dreaming is based on the fact that changes in patterns of activation across the brain alter the flow of information within and between components of the control hierarchy. The intrinsic cognitive properties of these components are preserved through transitions from mode to mode. What changes are the interactions between these components. Different modes (REM sleep, non-REM [NREM] sleep, alert waking, daydreaming) reflect variations in cognitive coherence or the way in which the components of the hierarchy are integrated.

The AIM model also points out that transitions between modes of waking cognition (from alert, focused attention to default mode, for example) are comparatively subtle instances of larger-scale modal transitions of which the most important is between sleeping and waking. Within sleep, the important modal transition is between NREM and REM sleep. Figure 5.1 integrates the AIM model with the discussion of controlled and automatic processing in previous chapters.

These transitions between modes are essentially a consequence of the balance between cholinergic (acetylcholine) and aminergic (5-hydroxytryptamine [5-HT]/serotonin, norepinephrine, and dopamine) regulation of the brain by reticular activating systems that project throughout the brain from the brainstem.

Very roughly, aminergic regulation is required for wakeful exploration of the environment and goal-directed behavior and cognition. Amines assist the prefrontal cortical structures to communicate with posterior ones and to build and actively maintain transient distributed networks necessary for controlled processing. During active waking, characterized by slow-wave firing patterns across the brain maintained by tonic levels of


Dreaming, Default Thinking, and Delusion

Figure 5.1

5-HT in particular, the noradrenergic, serotoninergic, and cholinergic systems are firing together. During slow-wave (NREM) sleep, these three systems reduce their discharge, while during REM sleep, the noradrenergic and 5-HT systems shut off completely, as the cholinergic system resumes its firing. However, the dopamine system projecting from the nucleus accumbens remains active even during REM sleep.

(p.97) Serotonin seems to enable the construction of stable patterns of activation across widely distributed neural circuits, enabling global integration of different systems for particular tasks involved in wakeful exploratory activity. If 5-HT levels are reduced in rats, their foraging, sexual, and social activity is reduced (Allman 1999). Even in animals with almost no brain, serotonin neurons are involved in orienting the animal toward nutrient sources and controlling digestion. This suggests that the regular delivery of serotonin across the brain keeps the organism in a state of wakeful exploration, coordinating perceptual, motor, and cognitive activity to enable life-supporting activities. What constitutes such an activity and the neural circuitry that supports it varies from organism to organism, so a monolithic representational interpretation of the role of serotonin across all neural circuitry is impossible. This is why Hobson introduced the notion of cognitive mode. It describes the global functional integration and coordination of cognitive subsystems when the mind is in exploratory mode.

Acetylcholine plays a role in maintaining patterns of activation across localized assemblies rather than the global patterns maintained by the serotonin system. It is thus an antagonist to 5-HT, adaptively disrupting stable global patterns of activity and detaching local assemblies from global integration. One feature of cholinergic regulation is that when it increases in wakeful stressful episodes, prefrontal activation is reduced. High levels thus produce a reversion to automatic mode. This has consequences for the understanding of many developmental disorders, and stress disorders manifest in episodes of disinhibition. As with other neurotransmitters, a one-to-one correlation with a cognitive function has not been proposed. Rather, the balance with other neurotransmitters and the density and type of (p.98) receptors in areas to which it projects determine how it influences cognition. However, we can note that when choline levels are high, relative to amines, they produce lack of prefrontally based executive supervision and reversion to automaticity.

The norepinephrine system plays a crucial role in poising the system for defensive action (“fight or flight”). Circuits it innervates control alertness, vigilance, and agonistic behavior, releasing appropriate hormones and priming visceral and somatic systems for rapid response.

Finally, as we saw previously, the dopamine system is a crucial part of the salience system. It enables an organism to target attention and cognitive effort on relevant stimuli by accentuating activation in circuits that represent salient information.

At any given moment, the effect of any neurotransmitter is highly context-dependent. It depends on density and type of receptors, mode of delivery and current activation level, the neurochemical state of the circuit it innervates, and the representational architecture of those circuits. For this reason, no monolithic interpretation of the role of a neurotransmitter is possible. Dopamine, for example, plays roles in learning and memory as well as motor control (evidenced by Parkinson's symptoms caused by low levels of dopamine). Serotonin causes blood vessels to contract as well as influencing a wide variety of prefrontally based functions.

The global state of the brain (sleep, wake, explore, withdraw, automatic, controlled) is, nonetheless, regulated by fundamental neurochemistry because that chemistry synchronizes global activation patterns at different ranges and time scales.

When the balance of cholines and amines distributed across brain regions by the reticular activating systems changes in favor of cholines, prefrontal activity—and with it the capacity for meta-cognitive responses to experience—subsides. Simultaneously, (p.99) motor expression is inhibited, and early stages of perceptual and sensory processing are shut down. Alert wakefulness and REM sleep are the ends of a cycle with NREM sleep constituting an intermediate stage, neurochemically and cognitively (see figure 5.1). In NREM sleep, there is no perceptual input, and metacognitive supervision is reduced. Volitional control is largely absent as a consequence of prefrontal deactivation. Consequently, standard routines or associative repertoires tend to be replayed, often accompanied by negative affect since the emotional systems remain active, but there is no top-down integration or supervision of these routines.

When serotonin is at its lowest level and the brain is cholinergically regulated, automatic processes continue without being organized externally by the environment through perceptual inputs or internally under volitional control. This is REM sleep.

In REM dreams, some cognitive processes such as late states of perceptual processing remain relatively intact, producing the characteristic stream of imagery. Others such as logical argumentation, volitional control, and planning are absent or reduced. Thus, in REM dreams we experience cognitive fragments such as images and sensations juxtaposed incongruously in vignettes and scenarios rather than coherently organized in narratives or explanations (Schwartz and Maquet 2002; Solms 2007; Röhrenbach and Landis 1995; Hobson 1999; Dawson and Conduit 2011; Domhoff 2005; Revonsuo 2006; Revonsuo and Salmivalli 1995; Revonsuo and Tarkko 2002).

5.3 Feature Binding and the Fregoli Delusion

The phenomenon of hyperfamiliar experiences seems common to both dreams and some delusions of misidentification. In delusions of misidentification, people report seeing “doubles,” (p.100) impostors, people in disguise, and people changing appearance and identity (Breen et al. 2000; DePauw and Szulecka 1988; Ellis, Whitley, and Luauté 1994; Spier 1992; Stone and Young 1997). The phenomenology of these delusions can be explained in terms of abnormalities of feature binding, combined with abnormalities in context binding. The abnormality of feature binding produces a representation in which elements normally bound together such as face, name, autonomic response to a familiar person, and identity may dissociate. We mentioned one such delusion, the Capgras delusion (the delusion that a familiar person has been replaced by an impostor or lookalike), maintained for a whole family, in the introduction. In another delusion of misidentification, the Fregoli delusion, patients report being followed by a familiar person in disguise (Courbon and Fail 1927; DePauw, Szulecka, and Poltock 1987; Ellis and Szulecka 1996; Ellis, Whitley, and Luauté 1994; Eva and Perry 1993; Joseph and O'Leary 1987; Wolff and McKenzie 1994; Wright, Young, and Hellawell 1993). An explanation based on the cognitive architecture of face recognition is that the patient sees a stranger but has a strong affective response characteristic of seeing a familiar person. This incongruity is produced within dedicated circuitry that recognizes faces and matches them to identifying information and initiates appropriate affective responses. The patient then has the experience of seeing a stranger but feeling as if the stranger is familiar to her. The patient then deals with this incongruity by producing the delusion “I am being followed by a familiar person in disguise.”

Recently cognitive theorists have taken a unified approach to these phenomena, noting that these Fregolilike phenomena, in which strangers are felt to be familiar, also occur in dreams. These cases of hyperfamiliarity in delusion and in dreaming are (p.101) explained in terms of similar abnormalities of feature binding. The differences are explained in terms of characteristic differences in context binding in each state.

To explain feature binding, these theorists use a cognitive theory of face recognition developed to explain disorders of face recognition known as prosopagnosia and extended initially to explain the Capgras delusion (as well as other delusions of misidentification such as the Fregoli). Prosopagnosia comes in different forms, but the relevant disorder for our purposes occurs when the subject can recognize a face qua face and recognize facial features but is unable to determine whether a seen face is familiar to her. Such patients will produce identical responses when presented with both novel and previously seen faces (Bauer 1984, 1986; Bruyer 1991; Ellis, Young, and Koenken 1993; Landis et al. 1986; Sergent et al. 1992; Shraberg and Weitzel 1979). In associative prosopagnosia, the subject has no explicit (“overt,” as it is called in the face recognition literature) awareness that she has seen a face before. Nonetheless in some cases, lack of explicit recognition is accompanied by preserved implicit recognition (called “covert recognition” in the face recognition literature). Covert recognition comes in two forms: The first I will call behavioral covert recognition (BCR). The second is skin conductance response (SCR). These covert forms usually co-occur, but it is useful to distinguish them.

In the behavioral case, covert recognition is evidenced by things like Stroop-type interference effects, response time priming, and a paradigm called true face relearning. Interference effects were shown in a task in which a prosopagnosic patient was asked to classify names presented in a list (e.g., politician or actor). His performance was affected by the presentation of a (p.102) distractor face from the wrong semantic category (e.g., George Clooney presented alongside David Cameron) (De Haan, Young, and Newcombe 1987). The point of these interference cases is that the distractor face could only interfere with performance if identified. In response time priming, prosopagnosic patients are asked to judge familiarity of names, which typically presents no problem (their problem is with faces not names). However, their response time is faster if they have previously been presented with the named face. This should not be the case if they are genuinely unable to recognize faces. In true face relearning, patients are presented with pairs of faces and names. In half the cases the names are incorrect. The patients are then re-presented with the pairs after an interval and asked to name the face. The fact that their performance is better in the “true” case is evidence of covert or implicit recognition (Schweinberger and Burton 2003; Bruyer 1991; Sperber and Spinnler 2003; Bruyer et al. 1983).

An example of covert SCR was provided by Bauer, who tested the galvanic skin responses of an overtly prosopagnosic patient, LF, to photos of familiar and unfamiliar faces. LF could not spontaneously name a single face and guessed names at chance level from a list of five, but covertly responded to more than 60% of familiar faces (Bauer 1984, 1986).

It is possible that SCR is simply a way of testing for BCR, that is to say that it is another indication of activation in the same neural system responsible for true face name relearning and response time priming. In fact, this is one way to interpret Bauer's early work in the area.

There is some evidence, however, to suggest that for the processing of faces, SCR and BCR depend on processing in distinct pathways. BCR is indicative of early stages of processing in the pathways culminating in face recognition. SCR is indicative of (p.103) the subject's autonomic response to the face consequent on “early” activation in this pathway.

As Young puts it, “it is inadequate to think of it [prosopagnosia] as simply involving loss of recognitional mechanisms. Instead, at least some recognition does take place. What has been lost is awareness of recognition” (quoted in Schweinberger and Burton 2003, 284; my italics). A model that explains how this can be the case is now a basis for explanation of the Capgras delusion.

The standard account of delusions of misidentification is that autonomic response is “mismatched” to a seen face. In the Capgras delusion, the autonomic response to a familiar face is absent. This makes the Capgras delusion in effect a double dissociation with prosopagnosia with preserved SCR. This is shown in figure 5.2 by the locations of the relevant lesions. In Capgras, the lesion is to pathways connecting the face recognition units and the amygdala. In prosopagnosia, the lesion is to pathways leading from the FRU to later, more explicit forms of recognition.

In the Capgras delusion, the patient sees a familiar person but does not have the characteristic autonomic affective response. The delusion is an attempt at context binding, which accounts for the familiar appearance and the absence of affective response. Doubles or impostors appear familiar but do not evoke affective responses.

The same model can also explain the Fregoli delusion involving cases where perception of strangers evokes an autonomic response characteristic of familiars due to hyperactivity in pathways that trigger the autonomic response to familiars. A theoretically economical explanation of the relationship between Fregoli and Capgras would equate them to pathological instances of déjà and jamais vu for faces instead of places. Déjà vu for place


Dreaming, Default Thinking, and Delusion

Figure 5.2 Basic architecture of the face processing system: from perception to identification.

involves a feeling of familiarity for a place that is not recognized as familiar. Jamais vu is the reverse: a feeling of unfamiliarity simultaneous with recognition that the place is in fact familiar. This analogy has been explicitly drawn a number of times (O'Connor and Moulin 2010).

The same model has been invoked to explain dream experiences of hyperfamiliarity. Revonsuo and Tarkko (2002) analyzed 592 dream reports that contained bizarre dream characters (people who appear in dreams). They divide bizarreness into two (p.105) types that correspond to relational (e.g., lives in UK) and intrinsic (facial appearance, familiarity, and identity) properties of the character. Among the latter, which they hypothesize is essentially a consequence of feature binding in the face recognition system, they include “the emotional component signaling the familiarity or unfamiliarity of the person” (2002, 7). Interestingly and in accordance with the hypothesis that the modularity of the cognitive process that produces it corresponds to the degree of representational disintegration in dreams, relational bizarreness was more frequent than intrinsic. Thus characters encountered in unusual or impossible settings or behaving uncharacteristically are more frequently reported than characters with unusual appearance. The latter account for only 2.2 percent of bizarre reports, while the former account for more than 50 percent. Bizarreness of familiarity accounts for 5.7 percent.

This is consistent with the standard model in which face perception and recognition is highly modular, but the link between a face and semantic knowledge about a person is not. Thus mismatches between face and semantic knowledge are quite possible if the relevant circuits are not linked by perceptually driven synchronization of activity. The link between face recognition and autonomic response is less modularized than face recognition per se, but far more modular than the linking of semantic knowledge and familiar faces. Consequently, as the authors note, inappropriate feelings of familiarity are one of the most common reports of internal bizarreness. “We may interpret the abundance of pseudo-familiar persons in dreams as over-activation of face recognition units: they fire even when the face percept does not correspond to any of the descriptions in the store. This creates the inappropriate feeling of familiarity” (Revonsuo and Tarkko 2002, 17).

(p.106) If the standard model is taken as a guide, it may be more accurate to suggest that a feeling of familiarity is a consequence of activity in the FRU. That activity in the face recognition unit propagates to the amygdala, which initiates an autonomic response that is “felt” as familiarity.

The authors explain erroneous familiarity in terms of the relationship between the face recognition unit and the autonomic response. If an unfamiliar face is mistakenly represented as familiar, an autonomic response is generated. A similar effect could be produced by endogenous activation of subcortical structures that generate autonomic response. Since limbic systems are active in dreams but not driven by perceptual inputs, this is another possible cause of inappropriate feelings of familiarity. In either case, the subject could have an autonomic response to a face that is not familiar. Schwartz and Maquet (2002) note in their discussion of dream experience that it resembles the phenomenology of other neuropsychological disorders such as achromatopsia, polyopsia, and microtopsia. All these cases can be explained in terms of abnormal patterns of activity produced when later processing in the ventral visual processing stream is driven by subcortical afferents rather than its usual inputs from early vision. The consequence is abnormalities of feature binding in systems that, because automatic, retain a high degree of cognitive coherence. Rather than disintegrate completely, perceptual processes assemble their components differently.

Thus, hyperfamiliarity is a relatively frequent occurrence in dreams. As we saw in waking states, the delusion that one is being followed by a familiar person in disguise explains why feelings of familiarity are evoked by someone who is not recognized as familiar. In dreams, of course, no such explanation is forthcoming.

(p.107) Slightly different Fregoli-like phenomena have been remarked on by Schwartz and Maquet (2002; Röhrenbach and Landis 1995). They use the following examples: “I had a talk with your colleague but she looked differently, much younger, like someone I went to school with, perhaps a 13-year-old girl.” In another case, a subject reported, “I recognize A's sister. I am surprised by her beard. She looks much more like a man, with a beard and a big nose.” Schwartz and Maquet describe these as “Fregoli-like” phenomena generated by activation in the facial fusiform area. Once again, it is plausible to see these cases as candidates for delusional rationalization. If the sister were disguised as a man, it would explain her masculine appearance.

An interesting feature of this type of Fregoli-like report is that the mismatch here is between facial appearance and identity (“your sister”) rather than facial appearance and autonomic response. In this sense, the phenomenon perhaps resembles the phenomenon described by Revonsuo and Tarkko as appearing infrequently in dreams: “cases of impostor relatives; persons we have never met in the real world but who we accept as our ‘sisters,’ ‘brothers,’ or ‘cousins’ or ‘uncles’ in the dream” (2002, 15–16).

Such cases are in some ways the converse of the Capgras delusion. Very interestingly, there are almost no reports of Capgraslike phenomena in dreams. This suggests to dream theorists that the patterns of activity in the facial fusiform area, which produce the Capgras delusion, do not arise in dreams: “Thus we should expect the pattern of brain activity to be different in normal REM sleep and Capgras patients” (Revonsuo and Tarkko 2002, 18).

One likely reason is that identity and appearance are more rigidly bound than affective response to processing within the (p.108) face recognition system. Mismatches between downstream affective response and appearance are more likely since the link is intrinsically more flexible.

The hypothesis of dream theorists about these phenomena of hyperfamiliarity or mismatch between appearance and identity in dreams is that they depend on similarities between patterns of brain activity in REM sleep and the delusional waking state. Precisely why Fregoli-like phenomena but not Capgras phenomena arise must depend on the nuances of circuitry and physiology.

5.4 Context Binding in Dreams and Delusions

At one point in his discussion of dreaming, Allan Hobson says that the ventromedial prefrontal cortex vainly tries to “impose a plot” on the flux of experience but is doomed to fail because it is not in communication with other structures such as DLPFC required to provide coherence and consistency.

In REM sleep, the dorsolateral prefrontal cortex “remains conspicuously deactivated” (Hobson 1999, 691) and the mind is dominated by highly salient imagery and emotional sensations. The resources to respond to inconsistencies and gaps in the narrative or to represent an overall goal for the narrative are absent due to dorsolateral deactivation, and the hyperassociation of salient imagery continues unbroken.

This selective recruitment of dorsolateral areas for decontextualized and ventromedial for essentially indexical thought is a feature of waking cognition replayed in dreaming. In studies of dreaming in prefrontal lesion patients (many of whom experience intense and disturbing dreams), Mark Solms (2000) found that dreams were suppressed in ventromedial patients and patients with lesions to sensory association areas, which process (p.109) inputs from “late” perceptual and sensory areas—in other words, those components of the default network that produce and associate perceptual and sensory imagery. In contrast, patients with dorsolateral lesions who exhibit the characteristic deficits in decontextualized problem solving had normal dreams.

The similarity between dream experience and delusion can thus be explained by the fact that they share important properties of default thinking, although delusions occur in a waking state. Namely, they involve the endogenous activation of components of a network that associates a stream of personally relevant experiences in the absence of actual perceptual input or supervision by decontextualized thought. This captures Hobson's (1999) idea that the ventromedial prefrontal cortex “tries to impose a plot.” However, there is no final reel in dreaming. It ends when the aminergic systems reactivate and their antagonists, the cholinergic systems, reduce their firing rates according to their own diurnal imperatives.

5.5 Dorsolateral Deactivation in Dreams and Delusions

We have already seen that delusions are also characterized by reduced or absent activity in dorsolateral prefrontal areas relative to the default network in experiments in which schizophrenic patients did not produce the normal “task-induced deactivation” of the default network in a working memory task. Thus, relative hypoactivity in the DLPFC is a feature of some delusions. An even more dramatic example of the association between reduced DLPFC activity and delusion, especially relevant to the cases discussed here, is provided by cases of lesion to the DLPFC associated with delusions of misidentification. Papageorgiou et al. studied nine patients with delusions of misidentification (p.110) (Capgras, Fregoli, intermetamorphosis, with some patients suffering more than one of these delusions) (Papageorgiou et al. 2003, 2005). The 2005 study was an event-related potential (ERP) focusing on the P300 ERP component. Compared to controls, the delusional patients group showed significant reductions in P300 amplitude in the right frontal hemisphere. If we accept the standard hypothesis that abnormal feature binding in the face recognition system is part of the causal chain leading to delusions of misidentification, then these cases are instances of abnormal feature binding occurring in a mind with reduced DLPFC function.

The conclusion that hypoactivity of the right hemisphere is involved in delusion is consistent with other neuropsychological studies. For example, in a group of thirty-three patients with Alzheimer's disease—of whom eighteen had a content-specific delusion concerning place, person, or object—single positron emission tomography revealed hypoperfusion in the right frontal hemisphere in the delusional groups compared to the fifteen nondelusional sufferers (Coltheart 2007).

5.6 Are Delusions Dreams?

Delusions are not dreams, but as in dreams, the balance of activity between default and decontextualized processing has changed. In some dreams, the default system churns out highly salient simulations triggered by activity in automatic feature-binding systems unmoored from the environment. The incongruities and inconsistencies of both feature and context binding are not detected or resolved. Thus narrative incoherence as well as (a lesser degree of) feature-binding incoherence is characteristic of REM dreams.

(p.111) In delusions, context binding is more intact. Subjectively adequate, coherent default narratives, or elements of them, are triggered by salient experiences. What seems to be missing in delusion is the ability to subject that narrative to decontextualized evaluation.

In the next chapter we turn to the question of how to characterize that lack of supervision. (p.112)