Psychological Design

SCHIZOPHRENIA: A COMPREHENSIVE META-HYPOTHESIS

Abstract

After generations of enquiry, and currently around 6150 papers being published annually (mostly focusing on the minutiae of the syndrome) there is a normative expectation that the pathogenesis of schizophrenia will never be unearthed. This in-depth review of all the prominent hypotheses of schizophrenia reveals that none can account for the heterogeneous presentations of the syndrome, especially one that ‘reaches beyond the categorical definitions to recognize the ‘constellation’ of schizophrenia features’ as the DSM-IV instructs. But by integrating and developing these hypotheses (including competing ones like the glutamate vs. dopamine hypothesis rivalry), the reams of data appear to congeal into a remarkably coherent meta-hypothesis. For the first time since the DSM-IV diagnostic criteria was published, a comprehensive and holistic explanation of all the symptoms and signs (as well as other findings known to the scientific community) is presented. The role that the striatal D1/D2 heteromers have in modulating saliency and mediating both bottom-up and top-down perception is expanded, bearing in mind that the switching of the striatal D1/D2 heteromers from a low affinity state (D2Low), to high (D2High) is already putatively the root of schizophrenia. Here links are found that associate top-down attention with presynaptic tonic dopamine synthesis and the function of D2High. Bottom-up attention, on the other hand, is
given to powerful phasic synthesis – the action of D2Low.

SCHIZOPHRENIA: A COMPREHENSIVE META-HYPOTHESIS 2

All attentional modes are essential to healthy perception, and as bottom-up attention subsides and top-down attention increases, eight very closely related syndromes can be identified, each with a distinct presentation and psychogenesis. These converge roughly into the
main schizophrenia subtypes. The meta-hypothesis that emerges from this review is highly triangulated and supported by robust published evidence.

By its narrowest definition, psychosis is a delusional and/or hallucinatory experience (American Psychiatric Association, 1994).Psychosis (thus defined) occurs in a number of conditions, but when psychotic symptoms are particularly bizarre and enduring, they characterize the most prominent symptoms of schizophrenia (Schneider, 1959). Next to schizophrenia, other psychoses simply aren’t as bizarre. The causes of schizophrenia have never been resolved and debate abounds from disciplines as diverse as linguistics and genetics about what mechanisms may be involved. A Medline review reveals 6146 articles since last year alone1. All heuristically use hypotheses, but these are as fragmented as their findings: they study single facets of the disorder and produce new hypotheses to explain their findings. Comprehensive hypotheses rare. These are discussed below but their explanations invariably fall far short of plausibility considering how bizarre, debilitating and heterogeneous schizophrenia is. This paper traverses several prominent hypotheses and integrates diverse streams of data to present a cogent meta-hypothesis to draw together all the diagnostic criteria listed in the DSM-IV to a single molecular dysfunction. This model first picks up on a number of perception-based hypotheses. These conceive perception as an input and behavior as an output. Psychotic symptoms are thus output errors, caused by the dysfunction either within perception, working memory, or any number of other areas of cognition. One class of hypotheses with an enduring legacy is that symptoms are caused by an inability to filter out (attenuate) perceptions of little or no importance resulting in a tendency to make irrelevant observations (Frith, 1979; Gray, Feldon, Rawlins, Hemsley, & Smith, 1991; Kapur, 2003). More recently, one of these hypotheses has gained particular attention because it links dopamine – a known active agent in psychosis, to perceptual saliency:
the front end of the perceptual process (Heinz & Schlagenhauf, 2010; Howes & Kapur, 2009; Kapur, 2003; Kapur, Mizrahi, & Li, 2005; Schwartz, Wiggins, Naudin, & Spitzer, 2005). In the aberrant saliency model (as this is called) a salient event is a perception marked by significance (something of importance relative to other percepts). Kapur posits that the importance or saliency of the event is tagged by the stimulation of the D2 receptors within the striatal dopamine neurons. The tagging process ensures that selective attention is directed to the percept that has been tagged as important. Salient events (ones tagged as important) should reflect the gamut of common reality as experienced by healthy people. The aberrant saliency hypothesis proposes that a functional oversupply of dopamine may harness attention ‘independently of cue and context’ (Kapur et al., 2005). As an example, Kapur refers to Bowers & Freeman (1966) who recall a patient describing his symptoms thus: ‘I developed a greater awareness… my senses were sharpened. I became fascinated by the little insignificant things around me.’ Aberrant saliency is presumed to cause disorganized thought, hallucinations and subsequent delusions. The pathogenesis is explained by a competing hypothesis, which is similar in many details: Fletcher & Frith’s Bayesian attenuation failure hypothesis (2009). In this model, normality is a state where familiar events are attenuated and go unnoticed. That is, people get so used to normality that they fail to notice it. However, when there is an attenuation failure, normal experiences appear uncanny and essential cognitive processes are aborted and left unresolved. Essentially Kapur, Fletcher & Frith are in agreement here.

While the prima facie evidence appears to bind dopamine, saliency and psychosis into a tight relationship, these hypotheses are bereft of details and need exploration. This is especially important because dopamine hypotheses are used to advocate specific treatments and account for serious medical conditions. But before we can discuss the implications of a dysfunctional system, a better understanding about the nature of the healthy system is needed, and for the purposes of this article, four research questions have been posed. Firstly, we need to ask exactly what saliency is. How much do we know about the neurological and perceptual processes involved in directing attention to important information? Secondly, if aberrant saliency is the basis for florid psychotic symptoms (hallucinations, delusions, etc) as both Fletcher & Frith (2009) and Howes and Kapur (2009); Kapur (2003); Kapur et al. (2005) assert, then what are the deficit symptoms – the distinct lack of responsiveness that was once thought to be at the core of the schizophrenic diagnosis (Bleuler,
1950; Parnas, 2011)?

Thirdly, must aberrant salience (the mis-tagging of important events) equate to over- stimulation – or even dopamine oversupply? Certainly, this is what Fletcher & Frith and Kapur both propose (Fletcher & Frith, 2009; Kapur, 2003; Kapur et al., 2005), yet about a third of schizophrenia cases don’t respond to dopamine antagonists (H. M. Jones, 2004), and how can this be explained? Lastly, is aberrant saliencies restricted to saliency as the concept is traditionally defined? That is, solely to stimulus-led perception, or does it also affect intention-led perception? In other words, does it only affect attention that is stimulated by bottom-up, unsought stimuli – or are there other events that may also trigger saliency responses – even when stimuli are anticipated?

The answers to these questions will help us not only to better understand schizophrenia, but will shed light on the processes of perception more generally.

Healthy salience

The concept of salience is based on the assumption that we selectively attend to a limited number of competing streams of information by filtering out or de-prioritizing irrelevant information (Milstein, Dalley, & Robbins, 2005). In this general framework, saliency is the measure of the influence that a singleton (a technical term for a potential percept, prior to it actually being singled out and declaratively perceived) has in gaining attention. At most, a single percept can dominate thought. It can, quite literally, ‘take possession of the mind’ (James, 1890, p. 404). At a minimum, the percept may remain latent; that is, beyond the threshold of declarative consciousness (Duckworth, Bargh, Garcia, & Chaiken, 2002; Hassin, Bargh, Engell, & McCulloch, 2009; Hassin, Bargh, & Zimerman, 2009). In between these two points, a singleton might make a partial incursion on working memory, occupying one or more ‘places’ of what is
often reckoned to be a limited capacity system. Most models allow four to seven items in active awareness (Cowan, 2005)

The factors of saliency

Saliency refers to properties of a singleton that demand selective attention; that is the directed awareness of the perceiver. In vivo studies of monkeys reveal that at least three factors will capture attention (Self citation, 2012b). Because all three are referred to as salience, a new taxonomy is proposed to maintain specificity:

High contrasts between the singleton and its context. Sudden sounds, flashing lights and bright colours etc, are highly prominent and demand reaction (Ljungberg, Apicella, & Schultz, 1992). As a subject learns about the presence of a prominent object, the initial burst of saliency immediately begins to wane – eventually to the point of extinction (Berns, Cohen, & Mintun, 1997; Mirenowicz & Schultz, 1996). It appears that prominence is automatically moderated by the glutamatergic and cholinergic neurotransmission systems (Carlsson, 1995; Corringer, Changeux, Bronner, Edelstein, & Smit, 2008) these then gate the secondary dopaminergic system (more about that bellow).
The kind of arousal that prominence evokes is highly automated, instinctive and is central to orientation reflexes (Ward, 2008). This form of salience must be non-cognitive because it generally survives states of hypoarousal such as vegetative comas (Laureys, 2005; Schiff et al., 2002) it will also activate both excitatory and inhibitory synapses of layer 2/3 of pyramidal neurons even in vitro – the result being a rhythmic oscillation in the gamma frequency – a frequency recognised as relating to sensory processing (Feldmeyer, 2010).

Significance
This form of salience is highly cognitive. It is the potential importance of the concept (schemata). There are three things that increase the significance of a singleton: a) anything that helps build meaning, b) anything that contributes to understanding, and c) anything that may needed to sustain existence (Self citation, 2010, 2011). In other words, we react to significant concepts, not only prominence.

Worth. Salience has a pre-cognitive affective or hedonic bias (Barrett & Bar, 2009). Monkeys respond to stimuli according to their desirability. The promise of a rewarding experience generates a much greater impact than the threat of a negative one (Mirenowicz & Schultz, 1996; Schultz, 1998). Worth may further be broken down into affective and hedonic subcategories. Affective worth is how a singleton makes you feel, whereas hedonic worth relates to an expectation of reward or punishment. Worth appears to be mediated by the amygdala, and remains either intact or slightly under engaged in schizophrenia (Anticevic et al., 2012; Becerril & Barch, 2010). This contrasts with affective psychoses like bipolar disorder, in which a large body of literature reports over-engagement of the amygdalae (S. B. Perlman et al., 2012).

A fourth saliency factor should also be added to the list. This factor was not identified using the same series of animal experiments, possibly because it is peculiar to healthy humans.

Opportunity
This is engagement – just for sensory variation, for stimulation, because we are active beings and need to do something but not necessarily for affective or hedonic gratification. It seems that no other animals have a go at things without good reason or choose to avoid a favorite food, just to try something different. Such behavior is very compelling for humans, and is therefore undeniably salient on some highly intellectual level, yet it doesn’t fit into the categories mentioned above. Nor does it behave the same way: opportunity is engagement with experience.. Opportunity is the salience of choice. The choice to try something off-putting, for no reason
other than the fact that it’s possible.

The modes of perception and attention.

The psychological canon allows for two modes of selective attention (see table 1.) There is top-down attention, which is the deliberate focus of declarative resources; and bottom-up attention,which is the attention brought to a subject by the interruption of top-down attention. As attention turns to attend to the interruption of an unexpectedly salient singleton, top-down intent changes.
Prior engagements are abandoned to focus on the rude intruder: ‘What is going on? Does it have relevance? How does it fit into what I know (my ontology)?’ If the transfer of focused attention to a new stimulus is automatic and unavoidable, then intent plays no part and is thus called bottom-up (Theeuwes, Atchley, & Kramer, 2000). This use of the term ‘bottom-up’ is thus a bit ambiguous, because it refers to the top-down attention that is given to bottom-up stimuli. To resolve this ambiguity the terms listed in column 2 of table 1 are used.

Raw attention
If there is such a thing, in its purest form, bottom-up attention must be a non-cognitive observation of unprocessed perceptual stimulus. Theoretically, in this raw state, nothing could be salient, because it would completely lack any kind of definition or ontological associations. Even so, automatic bottom-up attentional processes continue to monitor prominence, even during vegetative states, anesthesia, deep sleep, and in vitro (Feldmeyer, 2010; Laureys, 2005; Schiff et al., 2002) implying that this raw data feed bypasses cognition. This form of bottom up attention is referred to as raw attention.

Bottom-up attention
The cognitive attention given to anything that is not driven by intention, such as prominence and negative feedback is widely known as bottom-up attention. Bottom-up attention is a Bayesian process, it flags singletons that differ from expectations which would otherwise be ignored (for example, you crack an egg and it smells bad – you notice, but weren’t expecting it). In bottom-up attention, saliency is thus primarily a response to the unexpected, the undefined, misplaced and unanticipated, in other words mismatches between singleton-data and related schemata (Clark, 2012 – in press; Friston, 2003; Stephan, Penny, Daunizeau, Moran, & Friston, 2009). Functionally bottom-up attention is a response to difference and novelty, and this is how the term is normally used (Theeuwes et al.,2000). An example of a bottom-up Bayesian mismatch detection is when you see a red poppy in a green field or a tiger on your lawn, the former is inconsequential, the latter critical. In another body of literature, another feedback mechanism called error related negativity (ERN) is also proposed. This is a feedback loop specifically for negative affective or negative hedonic feedback – ERN occurs when an event is worse than expectations (Falkenstein, Hohnsbein, Hoormann, & Blanke, 1990; Holroyd & Coles, 2002). ERN is Bayesian too – but it isn’t driven by bottom-up attention because the negative event happens within the general range of expectations – ERN therefore fits in another category – active attention, a construct that fills the gap between top-down and bottom-up attention.
Bottom-up attention isn’t only given to external stimuli, it’s give to any unanticipated stimuli. Bodily (somatogenic) signals like hunger, tiredness, pain etc. would normally stimulate endogenous bottom-up stimuli because they’re unsolicited: we don’t usually ask ‘am I hungry, tired or in pain?’ because we will feel hungry or tired spontaneously. Initially these signals are received using raw attention, but come to attention with bottom-up attention. It must be noted that top-down attention can be directed to somatogenic stimuli, therefore while somatic and other bottom-up experience is normally bottom-up, it may occasionally be top-down.

Top-down attention
Top-down attention is the mode of attention where percepts are anticipated. For example, if I am hunting for truffles I ignore stones, but when I see something that matches my expectation of a truffle, it will elicit recognition salience: a ‘bingo!’ response. The top-down mode of attention has an exclusive domain of interest: the known. It is only concerned with recognizing significance saliency and only the correspondence between a singleton expectations (Coull, 2005). Unlike bottom-up attention, top-down attention is not Bayesian, because of its tight focus. Because top-down attention assesses only matches against intentions, top-down attention yields only positive feedback (you don’t have an interest in things that you weren’t looking for – these can only be noticed by other modes of attention.) This sets up a confirmation bias (Nickerson, 1998). Things that are actively sought trigger top-down recognition reactions easily, yet negative outcomes are referred to the bottom-up system, and will only be regarded if it is salient in its
own right (usually prominent) or significant in some way. There are two levels of top-down attention; focal and incidental. If we want to do something, we may focus on it intensely, in which case intention is highly deliberative. In such cases saliency is only used for positive feedback. On the other hand, the attention we give simple and well learned tasks is incidental. Here top-down attention that is incidental is reclassified as active attention.

Active attention
When a bottom-up singleton interrupts a deliberate (top-down) action, it comes as a distraction. But such encroachments are not regular (and may even be rare, depending on one’s lifestyle) because they depend on bottom-up stimuli, which may or may not be present. Despite this, we are not totally driven by intent (as other animals might be). In real terms, for humans, there is a large and undefined grey-area that separates the top-down and bottom-up modes of attention. If bottom-up attention is limited only to detecting prominence, and top-down attention covers only significance, then what of the rest? What of ERN? When things are noticed when they are worse than expectations but are generally within the range of unconsidered probability – such as when someone reaches into a fire and gets burned? Another whole range of experience is awareness on the edge of latent automaticity, which is nevertheless still noticed. When we do any routines we know well – such as drive, we do so automatically for the best part, yet we remain more or less
aware as we do so. There are many opportunities that we choose to engage in, just because they are there. Likewise, events are noticed but neither because we are looking for them, nor because they are prominent. Active attention is a term to describe the middle ground between bottom-up and top-down. It may be environment led, yet without the sharpness of novelty. Active attention is where top-down intention directs a ‘fuzzy’ search or where bottom-up attention reveals something that you had an interest in, but had not been looking for. Active attention is not driven by saliency, and isn’t a mode of attention in its own right, but complex combinations of bottom-up and top down processes, which are led by habits of engagement, by choice and simply by the availability of resources. And as such, active attention doesn’t sit comfortably with the assumed dynamics of the canonical model of attentional salience.
Attention is an expansive resource that doesn’t retreat and become passive when it’s not engaged in any clear intent or with salient environmental stimuli. If anything, the opposite: Active attention is the awareness that accompanies the continual activity of the mind and body, on the lookout for something to do, to experience and to think about. Indeed, studies have demonstrated that when environments are depleted of natural stimuli and opportunities for active attention, even healthy people start to hallucinate them (Grassian, 1983; Weckowicz, 1957) or fabricate them using whatever material is available – even human feces (Osmond, 1958). The behavior that takes place around active perception is largely automatic and involves stereotypical actions that are associated with whatever an object suggests – or affords the perceiver, thus objects of engagement are called affordances (Gibson, 1979; Withagen, de Poel, Araújo, & Pepping, 2012). An example is when you find yourself eating a cookie that you found when doing something altogether different. It wasn’t an intentional act, but neither was it unintentional, it was just automatic. Active-perception is very basic and instinctive because it links opportunity directly to action and in evolutionary terms, this is the primary role of perception. Simple creatures must act according to opportunities offered to them, but they don’t have the cognitive apparatus to intellectualize or otherwise make that action declarative (Bargh & Dijksterhuis, 2001). Active perception is evident in children too: ‘From birth, babies will actively engage with the perceptual environment. They will imitate expressions(Meltzoff & Moore, 1977), grasp a graspable object, and with age they will want to touch textures, run fingers down Brancusi’s sculptures, absorb themselves in colours and scents, feel the curvature and warmth of surfaces, immerse themselves in tastes, experiment with sounds, etc. They will want to squeeze bubble-wrap
and juice cartons, just to hear them pop. They light sparklers to see the sparks fly off in all directions; they burn incense to scent the air, and climb snowy mountains, just to feel the rush of zooming down again.’ (Self citation, 2012b) Systems that depend on attention

Perception is the most tangible link between the shared reality and personal experience. A lot depends on it and this brief section cannot even address all the facets of the psyche that are affected by attentional dysfunction, much less do them justice. Even so, it’s important to sketch out the interconnected systems of perception, automaticity, the sense of self and learning are needed because the most characteristic symptoms of schizophrenia can be explained as dysfunctions of these downstream functions all of which require attentional salience to function in memory (the ontology). This is constructed using a ‘dumb’ but efficient Hebbian
processes ‘cells that fire together wire together’ (Shatz, 1996). New knowledge isn’t learned this way, but associations are strengthened. This process creates the automatic ontology – processes that can be enacted without much consideration. New knowledge is processed through anti-Hebbian processes in which aberrance is detected and considered thoughtfully. The whole ontology is used to process all perception, but the ontology itself is opaque – it is perceived as something else – ‘the self’.

Perception

Even with robust attentional salience, perception is never perfect. Raw perceptual data is filtered and interpreted by our gross perceptual mechanisms. The sensory world is experienced through the available senses, so it is neither directly nor fully experienced. Human sight is the experience of reflections or luminosity of a narrow bandwidth (400-790 THz) of vibrations, 1-400 Thz is experienced as warmth (Gibson, 1979; Matlin, 1988), and the 17Hz-17kHz range as sound and pressure (Moore, 1997), but bats hear frequencies up to 115kHz (Heffner, 1983) and guppy fish can see light down to 300Thz (E. J. Smith et al., 2002). To sense these and other frequencies, humans have to use specialized equipment (such as radios or hyperspectral cameras), which translate frequencies into our native sensitivity range of visible light and audible sound. The next task ‘of the brain… can seem impossible: it must discover information about the likely causes of impinging signals without any form of direct access to their source’ (Clark, 2012 – in press, p. 6). Sensations must be iteratively interpreted and experienced through cortical processes. Information from the sensory organs travel to a series of perceptual cortices which process information of increasing complexity and abstraction. The later cortices (such as V4 and V5) are globally integrated to enable the processing of unknown, unpredictable and abstract stimuli. All the senses have lower-order processing areas and share the higher order ones -although even primary inputs are shared to some degree (Foxe et al., 2002).A bottom-up process is required to modulate raw perceptual data of the environment (including the body), but these stimuli can only be known and used in translation; that is through the top-down process of learning: decoding and recoding, creating and remodeling schemata. The heavy reliance on top-down processes doesn’t mean that what we experience isn’t real – but it does mean that experience is highly mediated and a partially inferred best guess (Sanders, 2004).

The implications of this are broad:

1. The fact that all perception is in some way translated into electrical and chemical impulses before it is reinterpreted as sensory perception should modify the way we understand hallucinations. In a profound way, all we know is, in some way hallucinated. I
don’t suggest that healthy experience is just as unreal as psychiatric hallucinations, but wish to point out that the qualifier of ‘hallucination’ is the departure from common and grounded sense of reality, and not the extension of experience into fantasy.
2. Perception will be distorted if the ontology (which informs perception) is dysfunctional.
3. The ontology is informed by personal experience, structured into schemata (including lexical schemas) and is fact-checked through a cultural lens. These form the basis of perception. This is reflected in the DSM-IV: ‘Ideas that may appear to be delusional in one culture (e.g., sorcery and witchcraft) may be commonly held in another. In some cultures, visual or auditory hallucinations with a religious content may be a normal part of religious experience (e.g., seeing the Virgin Mary or hearing God’s voice). In addition, the assessment of disorganized speech may be made difficult by linguistic variation in narrative styles across cultures that affects the logical form of verbal presentation.’ (American Psychiatric Association, 1994, p. 285).
4. The imperfect match between perception and experience means that all people have the capacity to be fooled by what they believe they perceive. This leads to one of the most commonly accepted hypotheses for hallucinatory experience – the illusion-hallucination
continuum (eg, Brébion et al., 2000; Frith & Done, 1988; Harvey, 1987; Johns et al., 2001). While direct studies of this phenomenon do demonstrate significant illusion/hallucination correlations, and therefore a possible role in psychosis, the effect size is still too small to explain psychotic symptoms (Aleman, Böcker, Hijman, de Haan, & Kahn, 2003; Woodward, Menon, & Whitman, 2007).

5. The higher the reliance on top-down and active attentional processes are (that is, the higher the top-down vs bottom-up sampling rate is), the more likely that perception will be distorted from reality, because sampling loses its fact-checking facility and negative
feedback tends to get ignored, creating a confirmation bias. This perspective is known in philosophy as the intellectualist position (A. D. Smith, 2007). Theorists who take an intellectualist stance on perception have an easy time understanding hallucinatory experience, because, for them the problem is resolved by asserting that top-down attention leads perception, therefore, a dysfunctional top-down attentional system will cause hallucinations. These models don’t acknowledge the possibility of direct perception. Intellectualist models only outline the higher order experience of knowing (a top-down function): they understand perception as a cognitive process that employs raw perception only as feedback. Current intellectualist models argue along the following lines: The brain is modeled as an organ that assigns meaning to raw perception. On one hand, raw perception is sensed, but sensory information remains meaningless until it is associated with schemata within ontology.

Clark (2012 – in press) and Fletcher and Frith (2009) emphasize that this follows an anti-Hebbian, Bayesian logic in order to maintain efficiency. In their models, singletons are only noticed when they are aberrant, at which point learning must take place. Grossberg (2009) is interested in the cytoarchitecture of perception. His model is Hebbian and distinctly not Bayesian, but is no less intellectualist. He proposes that the very laminar structure of the neocortical neurons match raw data to schemata by matching predictive representations of the world with raw bottom-up stimuli. The model Grossberg proposes is that neither top-down intentions nor raw bottom-up stimuli will trigger a neuronal excitatory reaction (recognition salience) on their own, except in psychosis (Grossberg, 2003). It is when two inputs: a tonic top-down intention and the other, a raw bottom-up singleton come together, that a saliency signal is transmitted horizontally (from one neuron column to another) through dendrites and vertically (to other brain regions) through axonal connections. In this model, intent quite literally works in a top-down fashion, because it stimulates the first layer of the neocortex. This tonic stimulation primes perception to receive appropriate input. Bottom-up perception works the opposite way around, from the sixth layer upwards. This tonically primes the neuron too, but the excitatory potential of the neuron is only achieved when both top-down and bottom up stimulation occurs. On their own, intellectualist approaches are limited. A distinction is not made between knowing and experience. Intellectualist approaches must face a very valid criticism; ‘I think is not I am, unless by thought I can equal the world’s concrete richness.’ (Merleau Ponty, 1943), This is not a critique of the reductive efficiency of Descartes, Bayes, Hemholtz or their inheritors (like Clark, Fletcher, Frith, Friston or Grossberg), but a critique of the lack of richness of experience that such approaches imply. As it happens, an automatic form of perception continues when top-down attention ceases, even in conditions such as a vegetative coma (Laureys, 2005; Schiff et al., 2002; Zeman, 2003), when sleep-walking (Plazzi, Vetrugno, Provini, & Montagna, 2005) or even in vitro (Feldmeyer, 2010). This means that raw perceptual processes are independent of any form of top-down attention, confirming Grossman’s assertion that attention and perception work in parallel. Habitual engagement (active attention) integrates the two. In this model, raw awareness provides a non-cognitive backdrop to experience: a richness and fullness instead of a perceptual vacuum. During active perception, raw sense data is not disposed of, just because it’s understood, nor is it disposed of merely because it is not. On the contrary, these conditions are when we may really start to enjoy experience. When we look at a painting and discover the subject, the richness in how the figures are presented doesn’t evaporate. The active-attention
model accepts the intellectualist models of perception, but also the concept of direct perception, thereby lowering the threshold for experience. The active attention model proposes that experience becomes declarative iteratively, exactly as perceptions are engaged with and acted on iteratively. At the most fundamental (precognitive) level, action is automated, not necessarily by
attention, but by habits that have formed direct associations with perceptual activations and actions: given familiar singletons, it isn’t necessary to recognize them before actions commence. We do not see a flat, hard surface and think; ‘hey – I can walk on that.’ We just walk (Bargh & Dijksterhuis, 2001; Gibson, 1979). The factors of salience will determine how much attention a singleton is given as a parallel process (see 1.1). In agreement with Clark, prominence (mismatches between expectations and experience) is a factor. But other factors (significance, worth and opportunity – see 1.1) also capture attention, and all are important. The mechanics of opportunity are especially interesting in the context of active attention because we do engage through action, but our awareness of such
engagement may be implicit (Hassin, Bargh, Engell, et al., 2009).
The associations that intellectualist models emphasize are plastic and changeable and thus are intrinsically vulnerable. Grossberg recognizes this and insists that a resonance between the raw data of reality and intentions is essential to create stability of meaning, to prevent it from being re-written with every new perception (Grossberg, 2003). Meaning cannot survive (long) when sensory cortical inter connectivity is lost, even if bottom up sensory perception continues. Dementia within the association cortices causes the singletons of perception to vanish, leaving behind nothing but a full, but unintelligible jumble of sensory indistinctness. In some cases, patients with specific dementia in the higher visual cortices cannot even distinguish between the categories of visual information – between colors, lines and shapes, much less pick out contents of a view, even though the raw visual information is still omnipresent (Sacks, 2010).

The sense-of-self

The sense-of-self is normatively considered irrelevant or only of peripheral interest to perception, but this viewpoint is difficult to defend. All the modes of attention serve a single perspective, and this is experienced as the locus of the sense-of-self; I am the center of my experience. There are many possible reasons for this, including (at a minimum):
1) Because all the organs of perception report to a single brain (although nominally to different
parts of the brain);
2) Because perception is served, and serves a single and personal epistemology;
3) Because the actions perception triggers all occur within a single physiology (Braund, in
preparation);
4) Because all of these functions unite a single sense of identity and ego (Shoemaker, 1968).
5) Because we have narratives of existence (R. A. Jones, 2010)
6) Because we act on self-knowledge and a sense of coherence emerges from these actions (Deci & Ryan, 1991 (1990));
7) We find anything that contributes to meaning, comprehensibility and manageability perceptually significant and salient (Antonovsky, 1987) The holism these interconnected and interdependent systems generates normally taken for granted, but it is vulnerable. The person (as philosophically conceived here) is an ego-centric composite of a body, perceptual system and epistemology, and for the best part, the sense of self doesn’t atomize when these constructs do – blindness, injury and ignorance are all undoubtedly risk-factors, but for the best part, the phenomenological sense of self is relatively robust in these domains. An experiment was conducted where the visual first person perspective was challenged, in this study the participants’ eyes were masked and all sight was channeled through monitor glasses with a wireless closed circuit video feed. The camera was put at a fixed angle in the corner of the room, creating the illusion that ‘my body is there – not here.’ Even in these circumstances, the sense of self was generally maintained; when participants wrote, the writing faced the camera, not the body, but the participants had a choice – they could also choose to engage their proprioception rather than visual modality, meaning that their writing faced their bodies (Mizumoto & Ishikawa, 2005). This suggests that the self-perspective is maintained by all the senses and can survive major disruptions. What is more vulnerable is the ability to reflect on the narratives and actions that we associate with the constructs of self, identity and self-agency. These narratives are sometimes oppositional: ‘I am x because I am not y’ (Festinger & Carlsmith, 1959). They are sometimes affirmative: ‘I am x because my actions are x-like’ (Bem, 1967). Both arguments rely on perception for the basic information to know what x and y are, and on narratives to contextualize x and y. Within these narratives, the sense of self is established relative to actions and thoughts, whether they are oppositional or reconciliatory.
I contend that both positions are valid, certainly in terms of contemporary neuroscience, because the more deliberate actions are, the more resistance: neural excitation and inhibition occur together, except during highly automated tasks, which are far more parsimonious, and don’t necessarily engage attention at all (Bargh & Dijksterhuis, 2001). Sometimes excitation and inhibition are in equal measure, in which case the balance suggests that no actions take place, but the greater the excess of excited neurons, the more spontaneous action occurs (Self citation, 2012a). It might be resistance to action (inhibition) that is recognized as self-agency, rather than actions themselves, whether successful resistance as in Festinger & Carlsmith’s model or unsuccessful as in Bem’s. If this is the case, active attention will be insufficient to maintain a sense-of-self.

Learning

The human search for comprehensibility and meaning appears to be endless. Every unexpected singleton is met with top-down questions about why, who, what and when – even when dreaming we are constantly trying to figure things out. This process involves linking singletons to the lexical and motor schemata that are most challenged by the new singleton (robin fits with birds, superhero  side-kicks; with rhymes bobbin and coffin; with names and other people called Robin like Robin Hood, with similar names such as Robert and so on.) This is a logical process which has been adopted to simulate intelligence in computing and has been adapted for very diverse models of learning (Baars & Franklin, 2003; Clark, 2012 – in press; Grossberg, Carpenter, & Ersoy, 2005; Johnson Abercrombie, 1960; McClelland, McNaughton, & O’Reilly, 1995; Rumelhart, McClelland, & UCSD PDP Research Group, 1986). If concepts defy knowing, they may still be placed into abstract schemata. Concepts like God, love and infinity are unknowable, but nevertheless have many associations across the global neural network. Abstract schemata are sometimes less grandiose but no less vague. If a singleton doesn’t initially appear to be of personal significance (ie, few significant associations), it may be bundled with other concerns for someone-else to deal with. Regardless of whether their nature is specific or abstract, concepts and their associations
become schemata and are useful for defining other singletons (Johnson Abercrombie, 1960). This is an intrinsically antiHebbian process, because it wrestles apart automatic associations and creates finer and finer distinctions within the epistemology. But there is also ‘dumb’ learning, which follows Shatz’ (1996) aphorism to describe Hebb’s rule of synaptic plasticity; ‘cells that
fire together, wire together.’ Where real learning creates associations, Hebbian learning reinforces them through long term potentation (ie, usage). The mechanisms are highly complex, but recent reviews link automatic functions to the strength of associations in the striatal association area (Ashby, Turner, & Horvitz, 2010). Under normal circumstances, Hebbian and antiHebbian systems work together. Whereas the anti-Hebbian system is primarilly used for initial learning (as described by Clark and others), the associations that are generated are strengthened by Hebbian processes, the oppositeis also true; disuse leads to synaptic loss andsometimes dementia. Schemata that have been committed to memory for long periods are known to lump together with similar concepts, if they can be recalled at all. This is demonstrated when people are given several stories and asked to relate them after only a short period and again after a long one. The phenomenon does not occur if subjects consider the stories regularly; instead, the stories develop details and a new character of their own (Johnson Abercrombie, 1960). This can be taken as preliminary evidence that the longer elements of the epistemology are left without focal attention, the more tacit they become. The more focus they receive, on the other hand, the more explicit (Gulick, 2006). Tacit knowledge is difficult to express because meanings are personal and hermetic, and the communication of knowledge requires explicit knowledge, not vague concepts. For example, someone who has forgotten their primary school science may say ‘light works because of electricity.’ The person is not wrong, but the information lacks specificity and isn’t very useful. Without regularly scanning knowledge for errors using a system similar to the one described by Clark or Friston (Clark, 2012 – in press; Friston, 2003), knowledge must become more tacit. Tacit awareness of well known schemata and routines appears to be an ideal level of
awareness to allow associated actions to take place automatically.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top