Word counts: short abstract – 105; abstract – 252;
main text – 13,676; figure captions – 457

Hallucinations in schizophrenia, sensory impairment and brain disease: a unifying model

 

 

Dr. Ralf-Peter Behrendt, MRCPsych, Specialist Registrar in Old Age Psychiatry

Dr. Claire Young, MRCPsych, Consultant in Old Age Psychiatry

 

 

The Longley Centre

Norwood Grange Drive

Sheffield, S5 7JT

UK

Tel: +44-114-271 6310 (switchboard)

E-mail: clairey@chsheff-tr.trent.nhs.uk, rp.behrendt@btinternet.com

 

Short abstract

What we perceive is the product of an intrinsic process and not a reflection of external physical reality. This notion agrees with recent advances in understanding the thalamocortical system. Perception may be subserved by assemblies of resonant gamma oscillations in thalamocortical networks that form intrinsically in response to cholinergic arousal, with afferent sensory input to the thalamus only playing a modifying or constraining role. In conditions that are accompanied by hallucinations, the relative impact of sensory input on self-organisation of thalamocortical gamma activity may be reduced, thus enabling underconstrained perception. This perspective may allow the integration of a wide range of findings relating to hallucinations.

Abstract

Based on recent insight into the thalamocortical system and its role in perception and conscious experience, a unified pathophysiological framework for hallucinations in neurological and psychiatric conditions is proposed, which integrates previously unrelated neurobiological and psychological findings. Gamma-frequency rhythms of discharge activity from thalamic and cortical neurons are facilitated by cholinergic arousal and resonate in networks of thalamocortical circuits, thereby transiently forming assemblies of coherent gamma oscillations under constraints of afferent sensory input and prefrontal attentional mechanisms. If perception is based on synchronisation of intrinsic gamma activity in the thalamocortical system, then sensory input to specific thalamic nuclei may merely play a constraining role. Hallucinations can be regarded as underconstrained perceptions that arise when the impact of sensory input on activation of thalamocortical circuits and synchronisation of thalamocortical gamma activity is reduced. In conditions that are accompanied by hallucinations, factors such as cortical hyperexcitability, cortical attentional mechanisms, hyperarousal, increased noise in specific thalamic nuclei and random sensory input to specific thalamic nuclei may to a varying degree contribute to underconstrained activation of thalamocortical circuits. The reticular thalamic nucleus plays an important role in suppressing random activity of relay cells in specific thalamic nuclei and its dysfunction may be implicated in the biological vulnerability to hallucinations in schizophrenia. Combined with general activation during cholinergic arousal, this leads to excessive disinhibition in specific thalamic nuclei, which may allow cortical attentional mechanisms to recruit thalamic relay cells into resonant assemblies of gamma oscillations regardless of their actual sensory input, thereby producing an underconstrained perceptual experience.

Key words: Charles Bonnet syndrome, gamma oscillations, hallucinations, Lewy-body dementia, perception, schizophrenia, thalamocortical system

1. Introduction

Gestalt psychologists in the first half of the last century argued that perception cannot be broken down into patterns of sensory stimulation and is not a derivative of the richness of stimulation from the external world (e.g., Koehler 1940). Instead, they maintained that perceptual experience is an active achievement of the nervous system. More recently, Llinas and Pare (1991) suggested that conscious perception is subserved by intrinsic activity in thalamocortical circuits, involving cortical pyramidal neurons and relay cells in specific thalamic nuclei. They pointed out that most of the connectivity in thalamocortical circuits is geared to the generation of internal functional modes, which can principally operate in the presence or absence of sensory input; only a minor part of thalamocortical connectivity is devoted to the transfer of sensory input. Cells in thalamocortical circuits are intrinsically active and sensory input may only modulate their activity. Llinas and Pare (1991) viewed consciousness as a closed-loop property of the thalamocortical system, and not a by-product of sensory input. Accordingly, they regarded wakefulness and paradoxical sleep as fundamentally equivalent states. The main difference between perception in wakefulness and dream imagery in paradoxical sleep would lie in the weight given to sensory afferents. In the state of wakefulness, but not in paradoxical sleep, the intrinsic functional mode underlying consciousness is modulated by sensory input (Llinas & Pare 1991).

Figure 1: According to transcendental idealism, a crucial distinction has to be made between the world that we perceive around us and the external physical world with which we interact. What we perceive as being around us is not the external physical world; instead, it is a part of our mind that is projected outside. In order to be adaptive, the subjective image of the world has to be constrained by external physical reality, and this is where sensory input plays its role. Thus, in the state of normal wakefulness, there is a relationship between the external physical world and the world that we perceive, but it is not the physical world that we see.

In their implication for the relationship between conscious experience and physical reality, these views are consistent with the philosophical position of transcendental idealism (Kant): The world that we see around us is internally created and a fundamentally subjective experience that in the state of normal wakefulness is merely constrained by external physical reality (Figure 1). Transcendental idealism predicts that normal perception, dream imagery and hallucinations are principally manifestations of the same internal process. They differ only with respect to the degree to which they are constrained by physical reality represented by sensory input. Thus, hallucinations can be conceptualised as perceptual experiences in the state of wakefulness that are underconstrained by sensory input (and it is suggested that in schizophrenia and some organic conditions this can be caused by peripheral sensory impairment or increased random neural activity in specific thalamic nuclei). Otherwise, there should be no difference: hallucinations arise in the focus of attention, just like any other perception, and they should involve activity in the same physiological systems that subserve normal perception. In hallucinations, attentional factors determine the content of conscious experience in a manner that is unrestricted by external sensory stimulation  (although in some organic conditions focal cortical hyperexcitability may substitute prefrontal and limbic input to the cortex mediating attention).

In contrast, psychopathology usually adopts a philosophical position of realism (in form of dualism or materialism), which assumes that the world that we perceive around us is an objective reality. The world is thought to exist independently of ourselves and not to be a product of our mind (Hamilton 1974). Accordingly, hallucinations are defined as false perceptions that arise in the absence of an external object or event. They are thought to differ from true perceptions in that they come from within the person’s mind as opposed to from outside the mind (Hamilton, 1974). The realist approach to hallucinations is intuitive and practical, however it suggests that hallucinations differ fundamentally from normal perception with respect to their source and mechanism of generation. Here, we suggest that in order to better understand the nature of hallucinations and integrate accumulating data pertaining to these phenomena we need a shift in paradigm from regarding the world around us as an objective reality to recognising it as a fundamentally subjective experience. In other words, normal perception, dreaming and hallucinations are equivalent, because even normal perception in wakefulness is fundamentally a state of hallucinations, one however that is constrained by external physical reality. The adaptive state of wakefulness certainly does depend on changes taking place in the outside world, but we do not see, hear, feel or smell physical reality itself, instead physical reality constrains the internal and fundamentally subjective process of perception, which is all that is necessary to ensure its adaptivity. From an evolutionary perspective, perception did not develop to copy the external world; the world that we perceive around us is as complex and differentiated as is necessary for the organism’s adaptive interaction with the external world in pursuit of its physiological needs.

In the following sections, we outline specific and non-specific factors that influence intrinsic thalamocortical activity, such as attentional mechanisms, sensory input and cholinergic control. A review of the thalamocortical system as it subserves perception enables us to relate the proposed general perspective on hallucinations to a specific pathophysiological model. Then, we show how this model relates to clinical and neurobiological findings in conditions that are associated with hallucinations, including schizophrenia, visual and hearing impairment as well as some cortical lesions and neurodegenerative disorders. What emerges is a theme of disruption of sensory constraints and determination by attentional factors, supporting the specific pathophysiological model and the general approach to hallucinations.

1.1. Resonance in thalamocortical networks

Projection neurons in specific and non-specific thalamic nuclei and inhibitory neurons in the adjacent reticular thalamic nucleus form neuronal circuits with interneurons and pyramidal neurons and in the cerebral cortex. In specific thalamocortical circuits, thalamic relay cells send axons to interneurons in layer IV of the cortex, which in turn connect to pyramidal neurons in cortical layer VI. Pyramidal neurons send glutamatergic projections back to the thalamus. These corticofugal  (corticothalamic) projections exert a direct excitatory influence on thalamic relay cells as well as an indirect inhibitory influence that is mediated by the reticular thalamic nucleus (reviewed in Llinas & Ribary 1993). In non-specific thalamocortical circuits, neurons in intralaminar thalamic nuclei project to layer I of the cortex; pyramidal cells in cortical layers V and VI project back to intralaminar nuclei both directly and indirectly via collaterals to the reticular thalamic nucleus (reviewed in Llinas & Ribary 1993). The reticular thalamic nucleus, which forms a sheet along the outer surface of the thalamus, plays an important part in thalamocortical connectivity. It consists of GABAergic inhibitory neurons that project to all other thalamic nuclei in a topographically organised manner and that receive collateral terminals from both thalamocortical and re-entrant corticofugal axons passing through the nucleus (reviewed in Saper 2000). Reticular thalamic neurons establish synaptic connections predominantly with dendrites of thalamic projection neurons and to a lesser extent with inhibitory interneurons in thalamic nuclei (Liu et al. 1995).

Thalamic projection neurons and neurons in the reticular thalamic nucleus can be in one of two electrophysiological response modes: a tonic firing mode, in which cells are partly depolarised and respond to afferent stimulation with firing of single action potentials, and a burst-firing mode, in which cells are hyperpolarized and respond with bursts of action potentials. In the tonic-firing mode, thalamic relay cells generate action potentials in a manner that is related to afferent sensory input, whereas in burst-firing mode sensory information is not transmitted effectively (McCormick & Feeser 1990). During wakefulness, thalamic relay neurons are predominantly in tonic mode; burst-firing mode becomes more prevalent in states of inattentiveness and drowsiness and predominates in slow-wave sleep.

Partial membrane depolarisation in thalamic neurons not only enables tonic firing of action potentials but also produces subthreshold gamma (around 40 Hz) oscillations of membrane potential (Steriade et al. 1991; Steriade et al. 1993). With further depolarisation, membrane potential oscillations can give rise to spikes or spike-bursts of action potentials that recur at gamma rhythms (Steriade et al. 1993). Subthreshold membrane potential oscillations that depend on partial depolarisation were also demonstrated in cortical neurons (Steriade et al. 1996; Nunez et al. 1992). They may predispose cortical and thalamic neurons to fire at gamma frequencies and synchronously in response to sensory input during wakefulness or internal input during paradoxical sleep (Steriade et al. 1996).

Rhythmic discharges from thalamic or cortical neurons can entrain oscillatory activity in connected neurons, whereby resonance occurs at a preferred frequency of synaptic input. Synchronised firing of several neurons will elicit temporally overlapping excitatory postsynaptic potentials in other cells and increase their chance of firing too. Thus, “single cell oscillators” and the conduction time of the intervening pathways can resonate to generate “large functional states” in the thalamocortical system (Llinas & Ribary 1993). Reverberating activity in local assemblies of interconnected thalamic and cortical neurons can manifest in gamma oscillations of magnetic or electrical field potentials recorded over the neocortex (Ribary et al. 1991).

Neocortical gamma oscillations of electrical or magnetic field potentials are more likely to occur in states of increased alertness and focused attention (Bouyer et al. 1981; Herculano-Houzel et al. 1999) and are also characteristic of paradoxical sleep (Llinas & Pare 1991; Llinas & Ribary 1993). Stimulation of cholinergic nuclei in the brainstem enhances neocortical 40-Hz oscillations in the electroencephalogram (Steriade et al. 1991; Curro Dossi et al. 1991) and facilitates their synchronisation in response to sensory stimulation (Munk et al. 1996; Herculano-Houzel et al. 1999). Electroencephalographic activation is mediated by acetylcholine that is released in the thalamus from cholinergic projections where it acts on muscarinic receptors (Steriade et al. 1991) to induce delayed and prolonged membrane depolarisation in thalamic projection neurons (Curro Dossi et al. 1991), thus enabling gamma discharge activity – most strongly in intralaminar thalamic nuclei (Steriade et al. 1993; Steriade & Amzica 1996)).

Neocortical gamma oscillations during wakefulness or paradoxical sleep show a coherent rostrocaudal phase shift from the frontal to the occipital pole of the hemisphere (Ribary et al. 1991). Rostrocaudal sweeps of cortical activation may be caused by internal waves of neural activity in intralaminar thalamic nuclei (Llinas & Ribary 1993). Intralaminar thalamic nuclei are organised as a circular mass and project to superficial layers of all neocortical areas in a spatially continuous manner (Llinas & Ribary 1993). Neurons in these nuclei have a particularly strong intrinsic 40-Hz rhythmicity that may entrain oscillatory discharge activities in cortical neurons (Steriade et al. 1993). By distributing gamma rhythms over the neocortex, intralaminar thalamic nuclei can facilitate the synchronisation of gamma reverberations in specific thalamocortical circuits that are activated by sensory input and attentional mechanisms. Llinas & Ribary (1993) suggested that conscious experience might be based on coherent 40-Hz co-activation of specific and non-specific thalamocortical circuits. While the content of consciousness may lie in specific thalamocortical circuits, non-specific thalamocortical circuits may ensure the temporary binding of activated specific thalamocortical circuits towards the creation of a unitary conscious experience (Llinas & Pare 1991; Llinas & Ribary 1993).

Neocortical 40-Hz oscillations have been recorded simultaneously with normal perception (Joliot et al. 1994) and hallucinations (Baldeweg et al. 1998) and are thought to underlie conscious experience in dreaming (Llinas & Ribary 1993; Amzica & Steriade 1996). Neocortical 40-Hz oscillations recorded magnetoencephalographically during paradoxical sleep are similar in distribution, phase shift and amplitude to those recorded during wakefulness (Llinas & Ribary 1993). During wakefulness, sensory stimulation can reset and enhance 40-Hz oscillatory activity recorded from the neocortex (Ribary et al. 1991). Such resetting is not observed during paradoxical sleep when random bursts of 40-Hz oscillations occur in a manner unrelated to sensory stimulation (Llinas & Ribary 1993). This is thought to represent the central difference between wakefulness and paradoxical sleep; neocortical 40-Hz oscillations and conscious experience are generated during both wakefulness and paradoxical sleep, however during paradoxical sleep the external world is mostly excluded from conscious experience (Llinas & Ribary 1993).

1.2. Non-specific regulation of thalamocortical activity

Thalamic relay cells and reticular thalamic neurons are regulated non-specifically by cholinergic, noradrenergic and serotonergic systems ascending from the brainstem. During wakefulness, brainstem regulatory systems globally facilitate or inhibit fast oscillatory and resonance capabilities of thalamic neurons and modulate their responsiveness to afferent sensory input. Thus, afferents from brainstem neurotransmitter centres adjust the impact of sensory information on resonance in the thalamocortical system, or, in other words, regulate the capacity of specific thalamic nuclei to transmit sensory information to the cortex. The neuromodulatory effects of serotonergic, noradrenergic and cholinergic input on spontaneous and evoked activity of thalamic relay cells have been studied mostly in the dorsal lateral geniculate nucleus of the cat. Noradrenaline released from fibres originating in the locus coeruleus produces a delayed enhancement of spontaneous firing in lateral geniculate relay cells and enhances their responsiveness to afferent synaptic excitation (Rogawski & Aghajanian 1980). Serotonin released from terminals of dorsal raphe nucleus neurons induces a delayed and prolonged suppression of spontaneous firing in lateral geniculate relay cells (Kayama et al. 1989). It also suppresses responses of lateral geniculate neurons to weak retinal stimulation (Kemp et al. 1982). Serotonergic suppression of relay cell activity is associated with augmentation of slow waves in the electroencephalogram (EEG) (Kayama et al. 1989).

The laterodorsal tegmental nucleus and the pedunculopontine nucleus in the mesopontine region of the brainstem are the main cholinergic nuclei that project to the thalamus. Mesopontine cholinergic neurons provide high concentrations of acetylcholine to the thalamus during both wakefulness and paradoxical sleep and much less so during slow-wave sleep (Williams et al. 1994). Acetylcholine released in the thalamus plays a crucial role in electroencephalographic activation during wakefulness and the generation of paradoxical sleep. In the lateral geniculate nucleus, acetylcholine exerts a facilitatory influence over the transfer of visual information. Mediated by a muscarinic receptor mechanism, cholinergic activation during arousal facilitates visually evoked responses (McCormick & Pape 1988) but also enhances spontaneous discharge activity of geniculate relay cells (Francesconi et al. 1988). In thalamic relay cells and cortical neurons, cholinergic activation induces sustained muscarinic depolarisation characterised by subthreshold oscillations of membrane potential, which enables tonic firing of action potentials at 40-Hz rhythms and predisposes neurons to participate in reverberations of gamma oscillations (Curro Dossi et al. 1991; Steriade & Amzica 1996).

Acetylcholine activates thalamic relay cells in the lateral geniculate nucleus both directly and indirectly, the indirect effect being mediated by activation of muscarinic receptors on local GABAergic inhibitory neurons (McCormick & Pape 1988; Francesconi et al. 1988). By mediating a reduction in the release of GABA in specific thalamic nuclei, muscarinic receptors on interneurons play an important role in increasing the efficacy of signal transmission in states of arousal and increased attention (Carden & Bickford 1999).

The reticular nucleus is among the thalamic nuclei with the highest density of cholinergic input (Heckers et al. 1992). Nicotinic receptors, particularly those with the alpha-7 subunit, are concentrated on reticular thalamic neurons (Spurden et al. 1997; Quik et al. 2000; Agulhon et al. 1999). Cholinergic input from the brainstem inhibits spontaneous activity of GABAergic neurons in the reticular nucleus, contributing to disinhibition of thalamic relay cells (Murphy et al. 1994). However, in response to certain patterns of sensory stimulation, reticular thalamic neurons can mediate inhibition of thalamic relay cells during arousal (Murphy et al. 1994). Stimulus-specific inhibition of thalamic relay cells may be due to activation of presynaptic nicotinic receptors on GABAergic terminals such as those from reticular thalamic neurons (Lena & Changeux 1997). Thus, while inhibition of GABAergic neurons in the thalamus mediated by muscarinic receptor activation may contribute to the global increase of relay cell activity during arousal, nicotinic facilitation of GABAergic transmission may at the same time improve the signal-to-noise ratio of thalamic activity (Lena & Changeux 1997).

The reticular thalamic nucleus assists in organising activity in specific thalamic nuclei according to characteristics of sensory input. In the auditory modality for example, the reticular nucleus participates in time-dependent analysis of the auditory input, with different neurons of the auditory part of the reticular nucleus being sensitive to different latencies of stimulus presentation (Villa 1990). Dysfunction of the reticular thalamic nucleus would lead to loss sensory-specific inhibition in specific thalamic nuclei. This may manifest particularly at times of arousal when thalamic relay cells exhibit increased spontaneous activity. Then, random activity may predominate over stimulus-specific activity and relay cells may become recruited into thalamocortical reverberations without receiving adequate sensory input.

1.3. Attentional modulation of thalamocortical activity

Figure 2: Perception may be a manifestation of intrinsic self-organisation of neural activity. Fast oscillatory activity in the thalamocortical system that is facilitated by cholinergic arousal self-organises into coherent assemblies under constraints of attentional mechanisms and sensory input. If sensory constraints are weak, then attentional mechanisms may become the dominant modulatory influence on thalamocortical self-organisation and hallucinations may arise.

The pattern of gamma oscillations that underlies conscious perception results from thalamocortical self-organisation and does not derive from purposeful sensory information processing (Figure 2). Following activation by arousal mechanisms, populations of neurons synchronise their gamma-frequency discharge activity through reciprocal interaction via re-entrant loops, while the thalamocortical system as a whole converges towards a state of transient stability, an attractor state that is determined by the current constellation of external and endogenous constraints imposed on the system (Varela et al. 2001). Sensory input imposes specific patterns of depolarisation on specific thalamic nuclei. At the same time, endogenous activity from the prefrontal cortex and limbic system, reflecting attention, current behavioural goals and recent memories, modulates cortical activity in primary and secondary sensory areas (reviewed in Varela et al. 2001). Prefrontal cortex neurons can sustain their activity for short periods of time despite ongoing behaviour and changes in sensory stimulation (Fuster 1991; Bodner et al. 1996), which may allow them to constrain self-organisation of gamma oscillatory activity in posterior sensory areas via long-distance corticocortical projections.

Neuroimaging studies show that attention modulates regional cerebral activity in primary and secondary sensory areas of the neocortex (O'Leary et al. 1996; Woodruff et al. 1996; Shulman et al. 1997), which are thought to be concerned with different stages of analysis of sensory information. Attention-dependent activity changes (despite identical sensory stimulation) can be seen even in parts of sensory systems that are believed to subserve the earliest stages of sensory processing, such as parts of the lateral geniculate nucleus of the thalamus and the retinotopically organised striate cortex (Vanduffel et al. 2000). It appears that activity in early sensory systems depends to a large extent on attentional factors and does not just reflect the pattern of external sensory stimulation. This challenges the notion implicit in cognitive or information-processing accounts of perception that sensory information is analysed to create a meaningful representation of the world from which attentional mechanisms then select relevant stimuli. Instead of being derived from hierarchically processed sensory information, perception appears to be created in the very focus of attention.

Corticofugal control of relay cells

Sherman and Koch (1986) suggested that sensory relay plays only a minor role in the activity of thalamocortical circuits compared with re-entrant activity from the cortex and reverberating activity. Numerically the largest input to thalamic nuclei does not derive from sensory organs but from layer VI of the cerebral cortex. Cortical neurons establish abundant excitatory synaptic connections with both thalamic relay cells and neurons in the reticular thalamic nucleus. In the lateral geniculate nucleus, only 10-20% of synapses on thalamic relay cells stem from the retina, about one third of synapses are from inhibitory terminals of local interneurons or neurons in the perigeniculate section of the reticular thalamic nucleus, and roughly half of all synapses are from neurons in cortical layer VI (reviewed in Sherman & Koch 1986). Attentional modulation of thalamocortical transmission may be a major function of these corticofugal projections (Sherman & Koch 1986; Montero 2000).

Through corticofugal projections, the cortex can adjust thalamic sensory-related activity and control the emerging pattern of synchronised gamma oscillations that underlies perception. Steriade (1997) reported that cortical pyramidal neurons discharging at 30-40 Hz rhythms are effective in synchronizing gamma oscillations in thalamocortical networks. Enhanced excitability of cortical pyramidal cells, which can be induced by noradrenaline or acetylcholine, may increase the spatiotemporal coherence of oscillatory activity in the thalamus (Destexhe et al. 1999). As proposed by Destexhe et al. (1999), a more excitable cortex may generate a more powerful feedback onto the thalamus, resulting in highly coherent oscillations.

The excitatory effect of cortical input to thalamic relay cells may be similar to that of afferent sensory input (Destexhe 2000). If thalamic relay cells are in tonic firing mode, corticothalamic excitatory postsynaptic potentials may contribute to depolarisation of relay cells, alongside accumulating sensory-induced excitatory potentials, until the firing threshold is reached. In this manner, corticofugal input may complement or predict afferent sensory input (Destexhe 2000). It is therefore conceivable that strong corticofugal input in combination with increased spontaneous activity pre-existing for various reasons can induce fast rhythmic discharges in relay cells despite insufficient or absent sensory input to these cells.

To resonate with cortical pyramidal neurons discharging at gamma rhythms, thalamic relay cells have to be in tonic firing mode. Sherman and Koch (1986) suggested that, as part of some attentional mechanisms, corticothalamic input might modulate thalamocortical sensory transmission by altering the electrophysiological response mode of thalamic relay cells. McCormick and von Krosigk (1992) showed that corticothalamic glutamatergic input can induce slow depolarisation in thalamic relay cells through activation of metabotropic (as opposed to ionotropic) glutamate receptors. This switches the response mode of relay cells from burst firing to tonic firing and facilitates thalamocortical transmission (McCormick & von Krosigk 1992).

Reticular thalamic nucleus

The reticular thalamic nucleus has been recognised to play a role in attentional mechanisms (e.g., Brunia 1993; Montero 1997). McAlonan et al. (2000) found attention-dependent activation in sectors of the reticular thalamic nucleus in rats despite identical sensory input. Montero (2000) reported that the visual sector of the reticular thalamic nucleus in rats was activated by attentional exploration of a new environment and that this activation depended on corticofugal inputs from the primary visual cortex. According to Montero’s (2000) hypothesis, a focus of attention in primary sensory cortex generates a column of increased thalamocortical sensory transmission by corticofugal glutamatergic activation of thalamic relay cells in conjunction with input to the reticular thalamic nucleus mediating the inhibition of surrounding relay cells. Thus, to a large extent the pattern of thalamic relay cell activity depends on cortical feedback. Instead of being ‘transmitted’ to the cortex, sensory input may only play an adjuvant role in thalamocortical activation.

Destexhe (2000) considered that attentional mechanisms might involve control of the electrophysiological response mode of reticular thalamic neurons. Their hyperpolarisation for instance can be achieved through muscarinic receptor mechanisms. Once reticular thalamic neurons are hyperpolarized, glutamatergic input from the cortex can trigger in these neurons a burst of action potentials. This would lead to hyperpolarisation of connected thalamic relay cells (Destexhe 2000), switching their response mode to burst firing too and preventing them from generating action potentials in accordance with incoming sensory and corticothalamic excitatory postsynaptic potentials.

Nucleus basalis of Meynert

Acetylcholine is involved not only in global brain activation during arousal, but also mediates attentional mechanisms based on the nucleus basalis of Meynert. The nucleus basalis represents the sole source of cholinergic input to the cerebral cortex. It receives terminals from the limbic system and the cholinergic peduncluopontine and laterodorsal tegmental nuclei (among others) and projects to all cortical areas. The nucleus basalis also sends cholinergic input to the reticular thalamic nucleus. Arousal is associated with increased tonic firing of nucleus basalis neurons and the release of acetylcholine to the cortex (reviewed in Smythies 1997). Activation of muscarinic cholinergic receptors on cortical neurons leads to increased neuronal excitability and facilitation of synaptic transmission from thalamic projections (Metherate & Ashe 1993), thus enhancing cortical sensory-evoked activity.

The nucleus basalis is involved in shifting attention to environmental stimuli that are behaviourally significant, e.g. in predicting a reward (Wenk 1997). For this purpose, the nucleus basalis receives information about the behavioural significance and reinforcement value of stimuli via afferents from the limbic system. Cholinergic projections from the nucleus basalis, in turn, modulate cortical excitability appropriately in order to facilitate perception of the significant stimulus (Wenk 1997). Apart from sending direct excitatory projections to the cortex, the nucleus basalis may modulate thalamocortical activity indirectly via projections to the reticular thalamic nucleus. Unlike most other thalamic nuclei, the reticular nucleus receives a substantial cholinergic innervation from the basal forebrain (Heckers et al. 1992).

Figure 3: Gamma oscillations in thalamocortical circuits depend on the depolarisation of thalamic and cortical cells. Attentional mechanisms activate neurons in cortical sensory areas and via corticothalamic projections contribute to depolarisation of thalamic relay cells. Sensory input complements activation of thalamic relay cells. Globally, the level of fast oscillatory activity (and therefore perceptual productivity) is regulated by input from brainstem neurotranmitter systems.

To briefly summarise, cholinergic input from the brainstem enhances evoked and spontaneous activity of thalamic relay cells and facilitates fast rhythmic discharges and their synchronisation in thalamocortical networks. Attentional mechanisms based on prefrontal cortex, limbic system and nucleus basalis provide specific patterns of activation to cortical neurons in sensory areas and indirectly participate in activation of thalamic relay cells. Activation of thalamic relay cells is balanced by general inhibitory input form serotonergic brainstem centres and sensory- and attention-specific inhibition form the reticular thalamic nucleus (Figure 3). A disturbance in mechanisms that maintain inhibition in specific thalamic nuclei may cause increased levels of noise in thalamic activity, particularly at times of arousal. This would allow oscillatory input from cortical pyramidal neurons to establish resonance with thalamic relay cells more easily – regardless of the pattern of sensory input to specific thalamic nuclei. Alternatively or in addition, pathological activation of thalamocortical circuits may be caused by more powerful corticofugal input from cortical foci of hyperexcitability or under conditions of general cortical hyperexcitability.

Normally, sustained thalamocortical activation and perception may ensue if the pattern of thalamocortical circuits that is pre-activated by prefrontal attentional mechanisms is matched by the pattern of sensory input to specific thalamic nuclei. In hallucinations, attentional mechanisms alone may produce sustained assemblies of thalamocortical activation, regardless of the pattern of sensory input. In sensory imagery, which is a non-substantial and fleeting experience, attentional mechanisms may initiate patterns of thalamocortical resonance that would normally lead to perception but cannot be sustained in the absence of supporting sensory information. Indeed, activity in early sensory processing areas during sensory imagery resembled that observed during selective attention (reviewed in Frith & Dolan 1997), again emphasising the importance of attention in the modulation of sensory systems.

2. Schizophrenia

Schizophrenia is characterised by episodes of hallucinations and other psychotic symptoms in clear consciousness that are usually accompanied by lack of insight and occur in the absence of a primary mood disturbance or identifiable brain disease. Hallucinations in the auditory modality and particularly verbal hallucinations appear to prevail in schizophrenia and affective psychoses. Nevertheless, visual hallucinations are not uncommon in schizophrenia; in contrast to some organic conditions, they typically occur without prodromata and in a psychological setting of intense affect (Asaad & Shapiro 1986). Tactile, kinaesthetic, olfactory and gustatory hallucinations are also reported in schizophrenia. Asaad and Shapiro (1986) suggested that the development of hallucinations in mental illness might represent a final common pathway involving biological vulnerability and psychological influences. While the biological vulnerability to hallucinations is likely to be indiscriminate of perceptual modality (affecting either all modalities or affecting them randomly), it may be that psychological influences can explain the apparent predominance of verbal hallucinations in mental illness. First, human communication and interpersonal relationships are largely mediated by language, and verbal hallucinations may reflect social experiences or fulfil defensive functions in people with enduring social anxiety and problems in relating to others. Verbal hallucinations may be employed ‘unconsciously’ to project social fears, confirm suspicions or fulfil social desires, bypassing open interaction with the social environment. Second, subvocal speech may provide a mechanism to unconsciously modulate and maintain the experience of verbal hallucinations once they have started to develop.

In schizophrenia, the development of hallucinations and possibly other psychotic symptoms may represent the outcome of a general biological predisposition towards hallucinations that interacts with psychological distress or anxiety arising from different constellations of personality problems, limited social coping skills and current interpersonal conflicts or social problems. Gruzelier (1999) proposed that schizophrenia might be a partial disorder of consciousness involving dysregulation in specific and non-specific thalamocortical systems. It is argued here that failure of sensory input to modulate intrinsic thalamocortical activity may be the core biological disturbance in schizophrenia that predisposes to hallucinations at times of arousal and heightened attention. In the first instance such failure could manifest in relative uncoupling of perception from sensory input across modalities, that is, if perception were generally subserved by reverberating activity in thalamocortical circuits. The olfactory system, which departs from organisational arrangements common to other sensory systems, may not be an exception in this respect since olfactory pathways between thalamus and orbitofrontal cortex appear to be critical for the perception and discrimination of odours (reviewed in Buck 2000).

Antipsychotic drugs are effective in controlling hallucinations, almost regardless of the underlying aetiology. Their effectiveness is established in mental disorders, such as schizophrenia, late paraphrenia and severe depression or mania, but also in psycho-organic syndromes and drug-induced psychoses. Although psycho-organic and substance-induced psychoses classically present with visual hallucination, this may nevertheless indicate (i) that all hallucinations result from a common pathophysiological mechanism regardless of their aetiology and (ii) that other symptoms of acute psychosis, such as disorders of the possession of thought and disturbances of self-experience, may share a common mechanism with hallucinations or be secondary to hallucinations. The effectiveness of antipsychotic medication across a variety of psychotic symptoms and across aetiological conditions testifies against there being a specific mechanism for hallucinations in schizophrenia. Cognitive or neuropsychological theories of hallucinations in schizophrenia tend to focus on verbal hallucinations and neglect the fact that hallucinations in schizophrenia can occur in other modalities. What would be desirable is a more general theory of hallucinations that provides a common framework for understanding content, form and meaning of these experiences, regardless of modality or aetiology.

Neuropsychological and cognitive theorists regard auditory verbal hallucinations as self-generated mental events, such as inner speech, thoughts, retrieved memories or verbal images, that are mistaken for external events (i.e. misattributed to an external origin) because they arise without intention and/or are experienced as alien to the self (e.g., David 1994; reviewed in Behrendt 1998). Consequently, the cause of hallucinations is attributed to a disorder of a hypothetical mechanism that controls or ‘monitors’ the corresponding self-generated mental phenomenon. Unfortunately, cognitive theories do not show convincingly why and how internally generated mental events can acquire the substantiality, richness and clarity that characterise normal perception. As long as perception is conceptualised as an experience that derives from external objects or events and as long as these external objects or events are required to be absent in hallucinations, we have to look for mental phenomena outside the normal process of sensation and perception that could become a source of hallucinations. Cognitive theories tend to suggest that the absence of a correspondence to external reality, which is thought to exclusively characterise self-generated mental events, has to be complemented by a conviction of their external origin to yield hallucinations. However, it is doubtful that thoughts, inner speech, verbal images, or retrieved memories can be transformed into experiences with perceptual qualities just by virtue of their misattribution to an external origin.

Once we recognise that both hallucinations and normal perceptions are fundamentally subjective experiences that are externalised into a virtual space surrounding oneself, we can eliminates questions about how hallucinations acquire perceptual qualities. Hallucinations and normal perceptions would differ only with respect to the degree to which they are constrained by external physical reality and we can now move on from hypothesising about sources of hallucinations and mechanisms of ‘inner’ mental phenomena to considering factors that lead to a disruption of sensory constraints normally imposed on the process of perception. Functional neuroimaging studies of actively hallucinating patients did confirm that verbal hallucinations involve activity in cortical areas that are normally concerned with perception of external speech (David et al., 1996; Woodruff et al., 1997; Lennox et al., 2000), which is consistent with the notion of underconstrained perception. The model predicts that further similarities will be found in brain activation; e.g., with the appropriate technology, it should be possible to detect correlated patterns of activity in auditory sensory cortex and auditory sections of the thalamus during both normal speech perception and verbal hallucinations.

2.1. Impaired response synchronisation

In visual backward masking tasks, detection of a briefly presented target stimulus is prevented by a mask stimulus that is presented shortly after the target (at interstimulus intervals of less than 100 msec). Compared to normal subjects, target identification by schizophrenic patients is prevented more easily by presentation of an early masking stimulus. Visual backward masking deficits were also demonstrated in schizophrenic patients who were in clinical remission (Green et al. 1999) and unaffected siblings of schizophrenic patients (Green et al. 1997), suggesting that these deficits are a marker of predisposition to schizophrenia rather than the presence of active illness (Green et al. 1999). Green et al. (1999) related visual backward masking deficits in patients with schizophrenia to failure to establish cortical oscillations in the gamma range in response to sensory stimulation. Kwon et al. (1999) found delays in entrainment of the electroencephalogram to 40-Hz auditory stimulation in patients with schizophrenia, which the authors interpreted as failure to entrain intrinsic gamma-frequency oscillators. These findings are consistent with the prediction by Llinas and Ribary (1993) that impaired resetting of 40-Hz oscillations by sensory stimulation characterises conditions that are accompanied by hallucinations.

The auditory evoked potential P50, which may be a subcomponent of the synchronised gamma response to sensory stimulation (Clementz et al. 1997; Basar et al. 1987, 1991), tends to be reduced in patients with schizophrenia. As pointed out by Gruzelier (1999), P50 amplitude reduction in patients with schizophrenia is associated with the presence of auditory hallucinations, although this is controversial. In experiments that involve repeated presentation of paired auditory stimuli and averaging of the electroencephalographic responses, most normal subjects show an amplitude reduction of mid-latency evoked-potential components such as P50 in response to the second stimulus (S2) as compared to their response to the first stimulus (S1). This amplitude reduction of the P50 response to S2 is absent in most schizophrenic patients, as expressed in an increase of their P50 S2/S1 amplitude ratio (so-called ‘gating ratio’). Lack of suppression of S2 P50 in auditory paired-stimulus paradigms can be found both in acutely psychotic and medicated clinically stable patients (Freedman et al. 1983; Adler et al. 1990) and is also present in many relatives of schizophrenic patients (Adler et al. 1992; Clementz et al. 1998). There is some controversy as to whether the increase in the P50 S2/S1 ratio is associated with perceptual abnormalities among patients with schizophrenia (Jin et al. 1998; Light & Braff 2000), however there is no doubt that this abnormality is associated with a vulnerability to schizophrenia, which – to a large extent – may represent a vulnerability to hallucinations (Asaad & Shapiro 1986).

Clementz et al. (1997) suggested that amplitude suppression of the P50 response to the second stimulus S2 that is normally observed in auditory paired-stimulus paradigms might be a proxy for suppression of the gamma-band response to S2. While at short interstimulus intervals synchronisation of the electroencephalogram is impaired normally only in response to S2, in patients with schizophrenia response synchronisation is impaired also in response to S1 (Zouridakis et al. 1997), which may explain their reduced S1 P50 amplitudes and increased P50 S2/S1 amplitude ratios. Evoked potentials are averaged from electroencephalographic recordings of many individual trials and the amplitudes of evoked-potential components, such as P50, can therefore be influenced by the temporal variability of the evoked response. Jin et al. (1997) and Patterson et al. (2000) showed that the lack of relative suppression of S2 P50 (i.e., increase in P50 S2/S1 amplitude ratio) in patients with schizophrenia was related to increased temporal variability of the P50 response to S1. Recognising that neuronal synchrony can be affected when the rate of background firing is too high, Patterson et al. (2000) hypothesised that increased temporal variability of S1 P50 in schizophrenic patients may be the result of “erratic neuronal firing” in a “hyperactive nervous system.” Increased random neuronal activity in schizophrenia could mask stimulus-specific activity, leading to deficient synchronisation in response to sensory stimulation or greater temporal variability of the synchronised response with consequences of reduced mid-latency evoked potentials and increased S2/S1 P50 ratios.

It may be added here that despite also showing an increase in their P50 S2/S1 amplitude ratios, clinically unaffected relatives of schizophrenic patients did not show a marked reduction in their P50 amplitudes in response to S1 (Clementz et al. 1998). Therefore, increased temporal variability of evoked potentials to S1 is unlikely to explain the increase of the S2/S1 P50 ratio in schizophrenic patients’ relatives. Even in patients with schizophrenia, increased temporal variability may not provide the sole explanation for increased P50 S2/S1 ratios. Although Patterson et al. (2000) showed that correction for temporal variability eliminated the significant difference in P50 S2/S1 ratios between patients with schizophrenia and control subjects, there still appears to have been a trend towards greater S2 P50 amplitudes in patients with schizophrenia (3.92 vs. 3.08 mV). It could be hypothesised that the increase in S2 P50 in relatives of schizophrenic patients is a manifestation of the same process that with greater severity leads to amplitude reduction of mid-latency evoked potentials, including reduction of S1 P50. At lower levels of neural noise – as may be the case in relatives of schizophrenic patients –, sensory input (S1) may have a reduced impact on neural activity in specific thalamic nuclei and evoke less sustained thalamocortical synchronisation, which would result in less phase opposition at S2 and therefore increased S2 P50 amplitudes and S2/S1 ratios. At higher levels of neural noise – as may be the case in patients with schizophrenia, changes in afferent sensory input to thalamic nuclei may be of similar magnitude to fluctuations of noise, which would render the impact of sensory input on thalamocortical activity unpredictable. As a result, the latency variability of stimulus-induced thalamocortical responses would increase, leading to amplitude reduction of the averaged S1 P50 (and a further increase in the S2/S1 ratio). This is consonant with suggestions that the increased P50 S2/S1 ratio (in relatives partly due to increased S2 P50) represents a vulnerability factor for schizophrenia and additional factors such as reduction of auditory-evoked potentials (including reduction in S1 P50 – not found in relatives) are involved in active schizophrenia (Adler et al. 1990).

2.2. Implication of the reticular thalamic nucleus

Administration of nicotine transiently restored amplitude suppression of P50 in response to S2 in patients with schizophrenia (Adler et al. 1993) and relatives of schizophrenic patients (Adler et al. 1992), which implicated nicotinic cholinergic receptors in schizophrenia. S2 P50 amplitude suppression in schizophrenic patients also normalised after brief periods of slow-wave sleep (Griffith et al. 1993), suggesting that nicotinic receptors might undergo abnormally rapid desensitisation during cholinergic arousal and re-sensitise only after a period of absence of cholinergic stimulation (Griffith et al. 1998). Genetic linkage analysis established that the increase in the P50 S2/S1 amplitude ratio in patients with schizophrenia and their relatives was linked to a polymorphic marker at chromosome locus 15q13-14, which is the site encoding the alpha-7 subunit of the nicotinic cholinergic receptor (Freedman et al. 1997). Altered expression or function of the alpha-7 nicotinic receptor may therefore be responsible for failure to suppress the auditory-evoked P50 response to the second of paired auditory stimuli in patients with schizophrenia and their relatives.

Nicotinic cholinergic receptors with the alpha-7 subunit are particularly concentrated in the reticular thalamic nucleus (Spurden et al. 1997; Quik et al. 2000; Agulhon et al. 1999). Interestingly, expression of alpha-7 nicotinic receptors was moderately reduced in the reticular thalamic nucleus in post-mortem tissue from patients with schizophrenia (Court et al. 1999). Activation of nicotinic receptors on terminals from reticular thalamic neurons facilitates GABAergic transmission in the thalamus, which may contribute to an increase in the signal-to-noise ratio of neural activity in specific thalamic nuclei during arousal (Lena & Changeux 1997). Rapid desensitisation and/or reduced expression of alpha-7 nicotinic receptors on reticular thalamic neurons would therefore result in decreased stimulus- or attention-specific inhibition and increased random activity in specific thalamic nuclei. This is consistent with the hypothesis by Patterson et al. (2000) that greater temporal variability of auditory evoked responses reflects erratic neuronal activity in schizophrenia.

In electroencephalographic recordings from a patient with recurrent somatic hallucinations, Baldeweg et al. (1998) observed gamma oscillations that occurred simultaneously with hallucinations. It appears that, on the one hand, sustained patterns of thalamocortical gamma resonance can occur in the absence of sensory input and give rise to hallucinations; on the other hand – as indicated above –, thalamocortical gamma rhythms are less modifiable by external sensory input in schizophrenia. The notion of increased random activity or noise in specific thalamic nuclei may explain this apparent paradox. Increased noise in specific thalamic nuclei could both mask changes in sensory input and, particularly at times of arousal, facilitate the recruitment of thalamic relay cells by cortical attentional mechanisms into assemblies of coherently activated thalamocortical circuits regardless of their sensory input. Thus, not only would an impaired signal-to-noise ratio predispose to hallucinations, it would also dampen the impact of sensory input on thalamocortical activity and perception. More intensive or prolonged sensory stimulation would be necessary to induce or modulate patterns of coherent thalamocortical oscillations, with the consequence of reduced perceptual responsiveness to changes in sensory input. This may manifest in elevated thresholds for tone discrimination (Rabinowicz et al. 2000) and reduced auditory acuity (Mathew et al. 1993) that were demonstrated in patients with schizophrenia.

2.3. Dopaminergic hyperactivity

Lack of amplitude suppression of P50 in response to the second of paired auditory stimuli, or increased P50 S2/S1 amplitude ratio, appears to be more related to a predisposition to schizophrenia rather than the presence of active illness. Waldo et al. (1994) suggested that an increase in the P50 S2/S1 amplitude ratio might be a necessary factor but not in itself sufficient to cause schizophrenia. For schizophrenia to become clinically manifest, a pre-existing increase in the P50 S2/S1 ratio may have to be complemented by other abnormalities, such as diminished hippocampal volume or increased dopamine metabolism (Waldo et al. 1994).

Dopamine activates D2 and D4 dopamine receptors on GABAergic neurons in the reticular thalamic nucleus (Khan et al. 1998). D2 and D4 are metabotropic receptors that are negatively coupled to adenylate cyclase and their activation on reticular thalamic neurons – among other effects – may suppress activation or expression of glutamic acid decarboxylase, which is the rate-limiting enzyme in the synthesis of GABA. This would reduce the synthesis of GABA and reduce the release of GABA to specific thalamic nuclei during arousal. D2 and D4 receptors are common targets for antipsychotic drugs. By blocking these receptors, different antipsychotic drugs were consistently found to increase the expression of glutamic acid decarboxylase in Sprague Dawley rats, particularly in the reticular thalamic nucleus (Sakai et al. 2001). This would restore the release of GABA to specific thalamic nuclei. As a result, chronic administration of antipsychotic drugs would increase the level of inhibition in specific thalamic nuclei, which was indeed demonstrated for haloperidol (Lukhanina 1989). Clozapine, which is known to be particularly effective in the treatment of symptoms of schizophrenia, is characterized by a high affinity for the D4 receptor. Interestingly, D4 receptors are expressed particularly on GABAergic neurons (including those in the reticular thalamic nucleus) suggesting that the antipsychotic effect of clozapine, and antipsychotics in general, may be achieved by modulation of GABAergic transmission (Mrzljak et al., 1996).

The hypothesis that emerges is that while dopaminergic hyperactivity results in excessive noise in specific thalamic nuclei and thus impaired thalamocortical response synchronisation to sensory stimulation, antipsychotic agents may reverse this process. Indeed, the use of indirectly acting dopamine agonists, such as cocaine or amphetamine, was associated with amplitude reduction of the auditory-evoked potential P50 (Boutros et al. 1993) and failure of relative amplitude reduction of auditory P50 (Light et al. 1999) in response to the second of paired stimuli. In patients with schizophrenia, the amplitude reduction of evoked potentials that is observed during exacerbation of schizophrenia may similarly be mediated by excessive dopamine (Adler et al. 1990). Treatment with antipsychotics, on the other hand, can normalise reduced P50 amplitudes in patients with schizophrenia (Boutros et al. 1993). However, conventional antipsychotics cannot reduce the increased P50 S2/S1 amplitude ratio in schizophrenia (reviewed in Gruzelier 1999), which is a more enduring abnormality that is independent of clinical state (Adler et al. 1990).

A constitutionally elevated P50 S2/S1 amplitude ratio in schizophrenia, which appears to be related to rapid desensitisation of nicotinic receptors on reticular thalamic neurons, may indicate increased baseline levels of random thalamic activity. Additional factors, such as dopaminergic hyperactivity, which can manifest in reduced amplitudes of mid-latency evoked potentials, may contribute to further random disinhibition of thalamic activity, until eventually perception becomes underconstrained by sensory stimulation and psychosis emerges. To suppress hallucinations, antipsychotic drugs may only have to reverse excessive dopaminergic inhibition of reticular thalamic neurons and restore an appropriate release of GABA to specific thalamic nuclei. In addition, antipsychotic drugs may prevent pathological activation of thalamocortical circuits by blocking D2 or D4 receptors on inhibitory cortical interneurons, thus restoring the release of GABA onto cortical pyramidal neurons (Sharp et al. 2001).

A similar mechanism to the one proposed for dopaminergic hyperactivity and exogenous dopamine agonists (i.e. inhibition of reticular thalamic neurons, reduction of the release of GABA onto thalamic relay cells, disinhibition of thalamic relay cells and pathological activation of thalamocortical circuits) was suggested by Tomitaka et al. (2000) and Sharp et al. (2001) to underlie the capacity of non-competitive NMDA (N-methyl-D-aspartate) receptor antagonists, such as phencyclidine and ketamine, to cause psychosis in humans. Particularly in schizophrenia, excessive disinhibition of thalamic relay cells alone may not be enough to produce psychotic symptoms. Corticofugal attentional mechanisms may be involved in the formation of coherent assemblies of pathologically activated thalamocortical circuits by providing additional specific depolarisation to cortical and thalamic neurons. Under conditions of excessive disinhibition of thalamic relay cells, corticofugal attentional mechanisms may recruit thalamic relay cells into temporarily sustained assemblies of thalamocortical circuits in a manner that is unrestricted by the pattern of sensory input. Such assemblies of pathologically activated thalamocortical circuits may underlie hallucinations and other psychotic symptoms.

2.4. Hyperarousal

Tonic electrodermal hyperactivity was regarded a state indicator of acute psychosis, since tonic electrodermal activity in patients with schizophrenia was abnormally elevated during psychotic states but not during remission (Dawson et al. 1994). Tonic electrodermal hyperactivity may even precede psychotic relapse (Dawson et al. 1992). Moreover, in schizophrenic patients with intermittent hallucinations the onset of hallucinatory periods was associated with an increase in the rate of spontaneous fluctuations of skin conductance (Cooklin et al. 1983). Tonic electrodermal activity is a measure of autonomic arousal and these findings may indicate that hyperarousal contributes to the development of psychosis and the production of hallucinations.

Central cholinergic activation during arousal results in electroencephalographic activation with an excess of fast activity. Electroencephalographic recordings from patients with schizophrenia tend to show increased beta activity, particularly in postcentral regions, and less alpha activity, but also excessive slow-wave activity, particularly in frontal areas (Morihisa et al. 1983). On its own, the excess of fast activity in the EEG would indicate hyperarousal (Gruzelier 1999). Among schizophrenic patients, excessive fast beta activity and less alpha and slow-wave activity in the resting EEG were associated with more florid psychotic symptoms and better response to neuroleptic treatment (Itil et al. 1975), further supporting an association between hyperarousal and acute psychosis. Acute psychosis may itself contribute to hyperarousal, but hyperarousal may also be the result of excessive stress and anxiety in schizophrenia or it may indicate that cholinergic brainstem centres are excessively responsive. The number of neurons in the pedunculopontine nucleus was shown to be increased in most schizophrenic patients, which suggested that increased cholinergic output from the midbrain reticular formation can overdrive the thalamus to produce schizophrenic symptoms such as hallucinations (Garcia-Rill et al. 1995).

Figure 4: Neurobiological factors suggested to be involved in the generation of hallucinations in schizophrenia: Disruption of sensory constraints may be due to disturbed function of the reticular thalamic nucleus, leading to lack of specific inhibition in specific thalamic nuclei, or peripheral sensory disorder. Reticular thalamic nucleus dysfunction in turn may be caused by nicotinic receptor abnormalities or dopaminergic hyperactivity. During arousal, relay cells in specific thalamic nuclei may be activated by attentional mechanisms alone and induced to participate in reverberations of gamma activity that underlie hallucinations.

In a biopsychosocial model of schizophrenia, environmental stressors such as life events and ongoing social stress can precipitate psychotic episodes by interacting with pre-existing biological vulnerability factors (Nuechterlein & Dawson 1984). Premorbid limitations in social competence and coping skills would influence the likelihood of life events and social problems and thereby determine the extent to which the individual’s biological vulnerability is stressed. Asaad and Shapiro (1986) predicted that the neurobiological basis for the vulnerability to hallucinations is also the basis for the vulnerability to schizophrenia. The biological vulnerability to hallucinations may be given by excessive random activity in specific thalamic nuclei, which can be caused by reticular thalamic nucleus dysfunction or random sensory input from disordered sensory organs (Figure 4). Cholinergic arousal that accompanies psychological stress and anxiety may mediate between environmental stressors and acute psychosis in a predisposed individual by further increasing thalamic background activity to a point where sustained pathological activation of thalamocortical circuits becomes possible. Dopaminergic hyperactivity may also play a mediating role. Once hallucinations have started to occur, a vicious circle of anxiety and psychotic defences may develop, with accompanying hyperarousal maintaining thalamic random background activity at a level compatible with ongoing hallucinations throughout the psychotic episode.

Based on the proposed neuroanatomical model of perception one should expect the emergence of further evidence in schizophrenia that shows a dysbalance between the impact of sensory input on activity in specific thalamic nuclei and the general level of arousal. Either thalamocortical systems are activated by arousal mechanism normally and sensory input is less effective in consistently modulating activity in thalamic relay cells, or sensory input modulates activity in the thalamus normally but hyperactive arousal mechanisms excessively facilitate thalamocortical gamma oscillations and thus perceptual productivity. Both mechanisms may play a role in schizophrenia.

2.5. Psychological factors

One of the benefits of an approach that conceptually unifies normal perception and hallucinations is that hallucinations can be recognised as arising in the focus of attention just like any other perception. This means that, once the process of perception loses its external constraints, unconscious and conscious attentional factors become the main determinants of content and form of perceptual experience. Contemporary cognitive theories of verbal hallucinations fail to explain why hallucinatory voices tend to resemble voices of particular people rather than the hallucinator’s own voice and why hallucinatory voices characteristically make statements in the second- or third- person grammatical form rather than in the first person. These facts are not easily reconcilable with theories that hypothesise that hallucinations derive from self-generated mental events, such as verbal imagery and inner speech, where the self is usually the one who is observing or discussing rather than the one who is being observed or being discussed.

In schizophrenia, patients may perceive hallucinations under pressure of increased attention to environmental cues and events that relate to their social fears. Personality deficits and enduring social anxiety may be core problems that precede symptomatic illness. Premorbid feelings of inferiority can be found in at least a subgroup of patients with schizophrenia (Kendler & Hays 1982). Poor social adjustment and lack of social confidence are common characteristics in children and adolescents who later develop schizophrenia (Jones et al. 1995). If there is an additional biological predisposition to underconstrained perception, hallucinations and other psychotic symptoms may develop at times of insurmountable social stress and rising interpersonal anxiety. Despite their withdrawn state, patients with established schizophrenia are still sensitive to social factors, since psychotic episodes often follow an insult to the patient’s self esteem (Gabbard 1990) and develop in circumstances of high levels of critical comments and hostility (Brown et al. 1972). Patients with social anxiety and low self-esteem are likely to observe their social environment intensely for hints that allow them to make inferences about their social value and acceptance. In particular, patients may pay increased attention to how other people think and talk about them. In biologically predisposed subjects, this increased attention could result in underconstrained perceptions; and according to what the patient attends to, hallucinations would take the form of voices that discuss them or comment on them, thereby confirming social suspicions and allowing unconscious projections. The fact that schizophrenics among all possible hallucinations predominantly perceive hallucinatory voices, and that these voices are in the second or third person, is likely to be of value for understanding the mechanism of hallucinations in schizophrenia.

With regard to their content, clinical observations suggest that verbal hallucinations are context-dependent and have a predictable quality (Nayani & David 1996). They reflect the patient’s beliefs about his subordination, disparagement and marginalisation in social relationships (Linn 1977; Birchwood et al. 2000). Verbal hallucinations may represent parental authority (Nayani & David 1996) and reveal psychodynamic influences such as guilt, wish fulfilment or gratification of repressed impulses (Asaad & Shapiro 1986). In short, rather than being symptoms that randomly afflict the patient, verbal hallucinations appear to be intricately linked to the patient's psychological condition, reflecting his experiences, preoccupations and anxieties, both on a conscious and unconscious level. The role of attentional factors in the generation of hallucinations is further confirmed by the fact that attention-commanding properties in stimulation from the environment can suppress hallucinations (Margo et al. 1981). The role of attentional factors can also be demonstrated in bereavement. People without mental illness can hallucinate their dead relative’s voice or presence in the context of intense yearning and searching for the deceased. This may seem obvious, but only a perspective that accepts that perception is generally a subjective creation, whether or not it is currently constrained by sensory information, allows us to explain the content of these hallucinations with reference to normal attentional mechanisms.

3. Hearing impairment

Sensory impairment may contribute to random background activity in specific thalamic nuclei. Noise is, to some extent, inherent in the flow of sensory information. For instance, retinal dark discharge, which is the maintained discharge of retinal ganglion cells, constitutes the background noise from which the visual signal must be discriminated. The proportion of noise in sensory input would be expected to increase as a result of pathology of peripheral sensory organs. Hypothetically, thalamic relay cell excitability may be up-regulated in an attempt to enhance or recover lost information. Such compensatory up-regulation or indeed simply up-regulation by cholinergic mechanisms during arousal would amplify noise in specific thalamic nuclei, again predisposing to underconstrained thalamocortical activation.

3.1. Auditory hallucinations and acquired deafness

Acquired hearing impairment is associated with auditory hallucinations. This is particularly clear in the case of musical hallucinations in non-psychotic patients. Musical hallucinations, which include the hearing of instrumental and vocal music, are predominantly found in elderly female patients with progressive hearing loss (Berrios 1990; Pasquini & Cole 1997). Psychiatric illness or personality factors are thought to play a minimal role (Berrios 1990; Wengel et al. 1989). However, psychiatric illness, mostly depression, can contribute to the development of musical hallucinations in patients with hearing loss, and in this case musical hallucinations tend to respond to the appropriate psychiatric treatment (Wengel et al. 1989; Pasquini & Cole 1997). Male gender, acute onset and the absence of deafness or psychiatric illness are factors suggesting the presence of brain disease, usually affecting the non-dominant hemisphere (Berrios 1991).

Non-musical auditory hallucinations such as tinnitus, irregular sounds or voices can also occur in association with hearing impairment and in the absence of any psychiatric disturbance or organic condition (Hammeke et al. 1983). Additional, central disinhibiting factors often contribute to the development of auditory hallucinations in patients with ear disease (Gordon 1987), even in the case of musical hallucinations (Aizenberg et al. 1987) where the contribution of sensory impairment is particularly prominent. Central disinhibiting factors that have been implicated include cerebrovascular disease, organic changes related to aging and alcoholism as well as psychological factors such as paranoia, persistent anxiety and depression (reviewed in Gordon 1987). Although auditory hallucinations are less likely to occur without independent evidence of neurological or psychiatric disorder, the contribution made by deafness appears to be crucial since hallucinations often vanish after remission of ear disease (Gordon 1987).

It is often believed that auditory hallucinations in patients with hearing loss result from sensory deprivation leading to central disinhibition and release of past memories (reviewed in Asaad & Shapiro 1986; Hammeke et al. 1983). The release theory of hallucinations does not explain hallucinations unrelated to sensory deprivation and it remains elusive which mechanisms are to be disinhibited and how past memories are to be perceived. Alternatively, it is proposed that peripheral otopathic conditions, such as otosclerosis or chronic otitis media, disrupt sensory constraints normally imposed on the process of perception, while central disinhibiting factors may contribute to pathological activation of thalamocortical circuits by enhancing cortical excitability.

3.2. Sensory impairment in late paraphrenia

There is a well-established association between deafness and late paraphrenia, a form of schizophrenia occurring in late life that is characterised by prominent paranoid delusions and auditory hallucinations. Pure-tone audiometric assessment of elderly psychiatric patients revealed a greater degree of hearing loss among patients with paranoid psychosis than those with affective psychoses (Cooper et al. 1974). Hearing impairment among patients with paranoid psychosis was most commonly due to chronic middle ear disease and had usually begun well before the onset of paranoid illness (Cooper et al. 1974). Paranoid patients with deafness of early onset tended to have relatively intact premorbid personality, suggesting that in these patients deafness could have played a relatively specific role in causing psychosis (Cooper et al. 1976). Furthermore, chronic deafness was identified as one of several independent premorbid characteristics that successfully differentiated between paranoid and affective psychoses in a group of patients aged over 50 (Kay et al. 1976). Thus, chronic hearing loss may play an important role in the aetiology of paranoid-hallucinatory psychoses of later life (Kay et al. 1976).

Almeida et al. (1995) found that patients with late paraphrenia were four times more likely than matched elderly control subjects to have hearing impairment. Patients were also more likely than controls to exhibit neurological soft signs, but there was no difference between patients and controls in the frequency of a family history of schizophrenia (Almeida et al. 1995). Patients with late paraphrenia are also more likely to have visual impairment. Cooper and Porter (1976) found major ocular pathology (predominantly cataracts) in more than half of patients with late paraphrenia, significantly more than in elderly patients with depression. Pearlson et al. (1989) established that among elderly schizophrenic patients those with illness onset after the age of 45 had more auditory and visual sensory impairment than those with illness onset before age 45. Howard et al. (1994) confirmed the high prevalence of auditory and visual sensory impairment among patients with late paraphrenia and found that visual impairment was associated with the presence of visual hallucinations.

3.3. Hearing impairment and schizophrenia

Similarly, hearing impairment in childhood or early adulthood was found to be a risk factor for the later development of schizophrenia (O'Neal & Robins 1958; David et al. 1995). In a case-control study, Mason and Winton (1995) found an association between middle ear disease and schizophrenia. The odds ratio of middle ear disease in patients with schizophrenia was increased further when patients with ear disease occurring after the onset of schizophrenia were excluded or when those patients were excluded who had a family history of schizophrenia or a history of brain damage, which suggested that middle ear disease may be a predisposing factor for the development of schizophrenia (Mason & Winton 1995). Gordon (1996) suggested that inner ear hypersensitivity might explain the link between middle ear disease and development of schizophrenia. Hypersensitivity to noises (possibly an indication of compensatory up-regulation) is common in incipient ear disease and may be the basic symptom on which psychotic phenomena are later constructed (Gordon 1995).

Auditory sensory impairment may also be due to a disturbance affecting auditory nuclei or pathways in the brainstem. Evoked potentials to brief clicks occurring within 10 msec of auditory stimulus presentation reflect the conduction of sensory information through auditory brainstem pathways and nuclei. Abnormal brainstem auditory-evoked potentials characterised by reduced amplitudes and missing peaks were found particularly in schizophrenic patients with prominent negative symptoms (Hayashida et al. 1986; Igata et al. 1994). On the other hand, Lindstrom et al. (1987) reported that among schizophrenic in-patients abnormal auditory-evoked brainstem responses were associated with the presence of auditory hallucinations and suggested that interference with auditory brainstem pathways might be causally related to auditory hallucinations.

3.4. Progression of auditory hallucinations in hearing impairment

Gordon (1987) suggested that post-otitic middle ear deafness is an important cause of paranoid hallucinatory psychoses of later life. He described how hallucinations might start with tinnitus, gradually assuming more definite forms until complex verbal hallucinations and finally persecutory delusions arise (Gordon 1987). Marneros et al. (1997) reported a patient with otosclerosis whose hallucinations underwent a progression from tinnitus to musical hallucinations and finally unilateral auditory verbal hallucinations and other psychotic symptoms (symptoms disappeared completely after surgical treatment for otosclerosis).

Tinnitus and other unformed hallucinations that are perceived in a state of peripheral sensory loss may not necessarily lead to the development of complex verbal hallucinations and psychosis. The relative absence of psychological disturbances in patients with musical hallucinations due to hearing impairment may be important in this respect. Psychosis may develop only if theafflicted person fails to gain insight into the abnormal nature of these experiences. An anxious or paranoid person may attribute the source of a sudden unusual voice to the social environment more readily because such perception may already confirm fears and suspicions harboured about other people. Not surprisingly therefore, paranoid personality traits and social isolation are frequent premorbid characteristics of elderly patients who develop late paraphrenia. Under conditions of chronic fear or social isolation, the person may increasingly pay attention to utterances from the social environment. Lack of insight would ensure that the psychotic patient continues to focus attention on the presumed external source of voices, and in the focus of attention verbal underconstrained perceptions wouldhallucinations would continue to be generated.

4. Charles Bonnet syndrome

The Charles Bonnet syndrome refers to complex visual hallucinations that are usually recognised as unreal and develop in the absence of a disturbance of consciousness or major psychopathology (Gold & Rabins 1989; Schultz & Melzack 1991). The Charles Bonnet syndrome is commonly associated with impaired vision or blindness due to ocular pathology such as macular degeneration (Holroyd et al. 1992), retinal haemorrhage and catar