The Split-Brain Syndrome
Abstract and Keywords
The received view within psychology and philosophy is that the split‐brain (commissurotomy) procedure leads to a breakdown in the unity of consciousness. Disunity models of the split‐brain can be divided into two classes: two‐streams models, according to which patients have two streams of consciousness, and partial unity models, according to which patients have a merely partially unified consciousness. Both models are motivated by the cognitive and behavioural disunities that patients exhibit in certain laboratory conditions, but they struggle to account for the cognitive and behavioural unity that patients demonstrate in everyday life. Preferable to disunity models is a full unity ‘switch’ model, according to which consciousness in the split‐brain rapidly switches between hemispheres. It is argued that only the switch model can account for both the behavioural disunities that split‐brain patients exhibit under experimental conditions and the behavioural unities that they exhibit outside of such contexts.
Whereas discussions of the unity of consciousness in the late nineteenth century focused on various dissociative phenomena, discussions of the unity of consciousness since the late twentieth century have centred on the results of so-called ‘split-brain’ operations. First performed on humans by van Wagenen and Herren in the 1930s, the split-brain procedure involves severing the corpus callosum in order to prevent epileptic seizures spreading from one hemisphere to another. However, van Wagenen's and Herren's version of the operation was not particularly successful (Bogen 1995), and the procedure was abandoned until it was resurrected by Vogel and Bogen at Caltech in the late 1950s (Bogen & Vogel 1962). The revised form of the operation proved more effective, and in early 1970 Donald Wilson of Dartmouth Medical School commenced another series of split-brain operations (Wilson et al. 1977; Wilson et al. 1982). The last forty years have provided neurologists with a new range of weapons for controlling epilepsy, but the split-brain procedure remains a treatment option of last resort (Spencer et al. 1987).
Astonishingly, the split-brain operation has little impact on the patient's ability to cope with the demands of everyday life. Patients can drive, hold down jobs, and carry out routine day-to-day tasks. Bogen remarks on their ‘social ordinariness’, and early researchers were baffled by the apparent lack of cognitive impairment arising from the operation (Akelaitis 1943, 1944). The split-brain case against the unity of consciousness derives not from the everyday behaviour of split-brain patients but from the cognitive and behavioural disunities that they exhibit in carefully controlled studies.1 Most commentators take such studies to show that split-brain patients no longer enjoy a unified consciousness. In this chapter I argue that this assessment of the split-brain syndrome is open to challenge.
(p.190) 9.1 Setting the stage
Most split-brain studies have drawn on a pool of eleven patients, six from the Caltech series and the remainder from the Dartmouth series. The procedure performed on the original Caltech patients is known as a ‘commissurotomy’, and involves severing all of the corpus callosum, together with various other interhemispheric tracts, including the anterior commissure, the hippocampal commissure, and (in some cases) the massa intermedia of the thalamus. In another form of the procedure, referred to as a ‘callosotomy’, only the corpus callosum is sectioned. The differences between commissurotomy and callosotomy patients are not pronounced, and I will refer to them all as ‘split-brain patients’. Behaviourally speaking, the important division is between patients in whom the anterior portion of their corpus callosum has been sectioned and those in whom only the posterior portion of the corpus callosum has been sectioned. The former tend to exhibit the classic split-brain syndrome, while the latter show only minimal dissociations.
The split-brain syndrome is most clearly revealed under carefully controlled experimental conditions. In the typical split-brain experiment, information is presented to the patient in such a way that it is processed by only one hemisphere.2 In vision this is typically achieved by means of tachistoscopic presentation, in which the patient is required to maintain a central focus whilst information is flashed into one hemi-field long enough to be registered but not long enough to permit eye movements. For example, the word ‘key-ring’ might be presented to the patient such that ‘key’ falls within the patient's left visual field (LVF) and ‘ring’ falls within the patient's right visual field (RVF). The contralateral structure of the visual system ensures that stimuli projected to the LVF are processed in the right hemisphere and vice versa (see Figure 9.1). Other modalities can be studied in a similar fashion. Tactile perception in the split-brain is studied by presenting stimuli to only one of the patient's hands, or by presenting different stimuli to each hand. Typically, the patient will be unable to either name aloud objects correctly identified with the left hand, or select with their right hand an object palpitated by their left hand. With respect to olfactory information, the patient may be unable to match odours presented to one nostril with those presented to the other (Gordon & Sperry 1969). The situation with respect to audition is rather more complicated, and I will defer discussion of it until §9.5.
Source: Sperry ( 1974)
Although the precise nature of such disunities may vary from patient to patient—for example, patient L.B. can name LVF stimuli but cannot compare stimuli presented to different halves of his two visual fields, whereas patient N.G. can compare stimuli across her two visual hemi-fields but cannot name LVF stimuli (Johnson 1984a, 1984b)—this description captures the ‘classic’ split-brain syndrome and provides a useful anchor for discussion. What might it tell us about the structure of consciousness in the split-brain?
The first question to ask is whether both left-hemisphere (LH) and right-hemisphere (RH) representations are conscious. (In the terminology introduced in Chapter 5, this is the challenge of meeting ‘the positive moment’.) Some of the early split-brain commentators expressed doubts about an affirmative answer to this question. According to some authorities, only the patient's right hemisphere is conscious. The ‘non-speaking’ left hemisphere is unconscious, and the behaviours that it generates are produced by ‘zombie mechanisms’ (Eccles 1965; MacKay 1966; see §5.1 for discussion of zombie mechanisms).
The problems with this account of the split-brain are daunting. Although speech production is generally lateralized to the left hemisphere, certain split-brain patients (P.S., V.P., L.B., J.W.) have some capacity for both LH and RH speech production (Baynes & Gazzaniga 2000; Levy et al. 1971; Zaidel et al. 2003). More importantly, we should not make linguistic production—or, for that matter, linguistic comprehension—a precondition on the possession of consciousness. Such a principle would, implausibly, remove pre-linguistic children and aphasics from the realm of the conscious. One might argue that if a creature that is capable of verbal report cannot verbally report the contents of state P then one has reason to think that P is not a conscious state, but even this principle is contentious. Appealing to any such principle in this context is particularly problematic, for right-hemisphere guided behaviour in the split-brain is a plausible counter-example to the claim that the contents of consciousness must be verbally reportable.
Can the zombie model be defended by assimilating right-hemisphere guided behaviour to other instances of unconscious behavioural control? Perhaps the right hemisphere behaviour of split-brain patients is of a piece with the non-conscious motor control that has been found in normals (see §5.1). However, on close examination the parallel between right hemisphere behaviour and (p.193) unconscious motor control turns out to be rather weaker that it might have seemed at first. Normal subjects who demonstrate motor control outside of awareness remain aware of the stimulus; what is lost is merely awareness of the response to the stimulus. In other cases, such as blindsight, the subject is not conscious of the stimulus, but is conscious of his response to it. The split-brain case would be unique in that the subject would be unaware of both the stimulus and their response to it.
The fundamental problem with the zombie model, however, is that split-brain patients can produce complex goal-directed behaviour under right hemisphere control. Indeed, right hemisphere behaviour far surpasses in cognitive complexity the kinds of tasks that elsewhere occur outside of consciousness. Summarizing an experiment testing the right-hemisphere based abilities of split-brain patients (with left hemisphere language), Sperry et al. write, ‘The overall level of the right hemisphere's ability to identify test items and also the quality of the accompanying emotional and evaluative responses were of the same order approximately as those obtained from the right visual field and left hemisphere’ (Sperry et al. 1979: 163). We would have little hesitation in saying that a patient in a persistent vegetative state had emerged from the vegetative state if they were to perform actions that were as purposeful and environmentally sensitive as those that split-brain patients perform under right hemisphere control. In light of this, I conclude that both the left and right hemispheres in the split-brain can support consciousness. In other words, the ‘positive moment’ in the argument for disunity can be met.
What about the negative moment? Can it also be met? The dominant view within both neuropsychology and philosophy is that it can. Most commentators are united in the view that there are times at which the split-brain patient has phenomenally disunified conscious states. Let us call such views disunity accounts of the split-brain. Within the broad family of disunity accounts we need distinguish two-streams accounts from partial unity accounts. Proponents of the two-streams account hold that the conscious states that a split-brain patients has at any one time can be assigned to one of two non-overlapping sets, where the members of each set are mutually unified but no member of either set is phenomenally unified with any member of the other set. Although few theorists put their views in quite this way, the two-streams model is arguably the received account of the split-brain syndrome.3 In the words of one paper, ‘…the disconnected hemispheres in both animal and human subjects are separately conscious in parallel at a moderately high and approximately equal (p.194) value’ (Sperry et al. 1979: 153). By contrast, proponents of the partial unity account hold that consciousness in the split-brain has a branching structure: patients have simultaneous experiences (e 1, e 2, and e 3) such that e 1 and e 2 are each phenomenally unified with e 3 but not with each other (see §2.4). To the best of my knowledge the partial unity model of the split-brain was first explicitly presented by Lockwood (1989), although there are earlier hints of it in both the philosophical and neuropsychological literatures.4
Both the two-streams and partial unity models are at odds with the unity thesis, for each model holds that there are times during which patients have experiences that are not phenomenally unified with each other. However, one could attempt to reconcile these models with the spirit behind the unity thesis by rejecting the identification of subjects of experience with organisms. For example, one could follow neo-Lockeans in identifying subjects of experience with psychological networks or intentional systems. On such accounts, the truth of phenomenal disunity treatments of the split-brain need not spell the end of the unity thesis, for it could turn out that severing the corpus callosum not only splinters the patient's stream of consciousness but also creates multiple—and perhaps even overlapping—subjects of experience. Just what to say about the unity thesis should we identify subjects of experience not with organisms but with intentional systems raises many complex issues, not least of which is the problem of how to individuate such systems. However, I will leave this issue for others to pursue, for if the account of the split-brain that I offer in §9.5 is sound then the unity thesis can be saved without denying that subjects of experience can be counted by counting conscious organisms. But before we get to that, we must first examine what there is to be said on behalf of disunity accounts of the split-brain.
9.2 The case for disunity
Common to the two-streams and partial unity models is the claim that split-brain patients are phenomenally disunified—that the ‘negative moment’ can be sustained. So our first order of business is to examine the case for phenomenal disunity. I will consider three arguments for thinking that split-brain patients have phenomenally disunified conscious states.
(p.195) The argument from agentive disunity
Although split-brain patients generally display a high degree of agentive unity, there are contexts in which this unity is lost, or at least impaired to some degree. One manifestation of agentive disunity in the split-brain is inter-manual conflict, in which the patient's left hand becomes anarchic and interferes with what the patient is trying to do with his or her right hand. For example, the anarchic hand might interfere with the patient's attempts to button up his shirt. Anarchic hand behaviour is common in the weeks immediately after surgery, but typically subsides within a few months (Bogen 1998; Wilson et al. 1977; Fergusen et al. 1985).
A second form of agentive disunity is spatial decoupling. Although cognitively intact subjects find it difficult to simultaneously draw distinct spatial patterns (for example, a square and a circle) with each hand, split-brain patients have no such difficulties (Preilowski 1972; Zaidel & Sperry 1977). In one study, patient J.W. and three normal controls were required to draw with each hand pairs of lines that differed in orientation (Franz et al. 1996). As expected, control subjects were impaired on trials in which the two stimuli had divergent orientations (relative to when they had identical orientations), whereas J.W. showed improved performance on such trials.
It seems to me that neither inter-manual conflict nor spatial decoupling provide much support for thinking of split-brain patients as phenomenally disunified. The behaviour of the anarchic hand is best thought of as triggered by stimulus-driven intentions: seeing a button elicits a prepotent response to undo it; seeing a fork elicits a prepotent response to pick it up, and so on. The intention might be triggered by a conscious perception of the stimulus, but the subject herself need not be conscious of the intention. Indeed, we are frequently unaware of stimulus-driven intentions. (Consider what it is like to navigate a crowded street whilst engaged in conversation.) Spatial decoupling is likewise consistent with phenomenal unity. Again, the kind of motor independence seen here involves only subconscious mechanisms; we needn't posit separate conscious intentions or goals in each hemisphere. We might also note that although split-brain patients can produce spatially decoupled actions, they do not produce temporally decoupled actions (see Franz et al. 1996; Kennerley et al. 2002; Tuller & Kelso 1989). Split-brain patients may be less ‘agentively unified’ than cognitively intact individuals, but we should not think of each hemisphere as harbouring an autonomous agent. Each of the two hemispheres may be able to process perceptual information autonomously, but they are not able to select motor responses independently and simultaneously (Ivry et al. 1998; Pashler et al. 1994; see also Reuter-Lorenz 2003).
(p.196) Paradoxically, the most striking form of agentive disunity in the split-brain may involve inter-hemispheric cooperation rather than inter-hemispheric conflict. In a phenomenon known as ‘cross-cuing’, patients employ environmental cues to transfer information between hemispheres. For example, the patient may run a fingernail down the teeth of a comb in order to produce a characteristic noise by which the object can be identified (Zaidel et al. 2003: 365). Employing the perceptual space shared by both hemispheres, the right hemisphere may attempt to ‘tip-off’ the left hemisphere by visually fixating on a related object. Cross-cuing can also be mediated by movements of the head or tongue, subvocal speech, and various forms of imagery (Bogen 1998; Corballis 1995; Gazzaniga & Hillyard 1971). Like bridge partners trying to tip each other off about their hand or members of a jazz duo attempting to keep each other ‘in the pocket’, we could think of the two hemispheres as distinct agents intent on securing a common goal.
But as tempting as it might be, I think we have good reason not to model cross-cuing on intersubjective communication. For one thing, the cuing hemisphere does not take itself to be communicating with another agent, nor does the receiving hemisphere think of itself as the recipient of a communicative act. In a revealing comment, J.W. exclaimed, ‘Are you guys trying to make two people out of me?’ (MacKay & MacKay 1982: 691). If we were to accept the intersubjective conception of cross-cuing, we would have to think of J.W. as (self?) deluded in taking himself to be a single agent. That, I think, would be an unwelcome result. (Contrast the kind of cross-cuing carried out by J.W. with that which occurs between conjoined twins, which clearly does take place between different agents.) Although there is some evidence that split-brain patients may have distinct RH and LH self-conceptions (LeDoux et al. 1977; Sperry et al. 1979), there is no evidence that each of the patient's hemispheres thinks of itself—or, indeed, of its neighbour—as an autonomous locus of agency.
Unwelcome or not, we might be forced to model cross-cuing on inter-subjective communication if it were the only account going. However, it turns out that it's not. In fact, there are two alternatives to the ‘communicative’ conception of cross-cuing. On the one hand, we could model cross-cuing on the kind of information transfer that occurs between sub-personal homunculi in (say) decision-making or memory retrieval. On the other hand, we could model cross-cuing on self-directed agency; in other words, we could think of cross-cuing as a technique by means of which the patient attempts to manipulate his or her own mind. Just as one might try to jog one's memory by engaging in certain types of imagery or by talking to oneself, so too the split-brain patient might deliberately deploy various tricks in order to transfer information between (p.197) hemispheres. It is not clear to me which of these two models best fits cross-cuing, but each of them seems preferable to the communicative model.
Split-brain patients depart from the norms of unified agency in a number of ways, but none of these departures indicates that we should think of patients as ‘housing’ two distinct conscious agents. Instead, the evidence suggests that split-brain patients are single agents who are attempting to get by as best they can (Gillett 1986). Split-brain patients might enjoy significantly less agentive unity than you or I, but we should not conceive of them as ‘composites’ of two conscious agents.
The representational disunity argument
Another argument for phenomenal disunity appeals to a lack of representational integration between left hemisphere and right hemisphere representations. Consider a typical split-brain patient (‘S’) in a key-ring experiment.
(1) S has, simultaneously, an experience with the content 〈‘key’〉 and an experience with the content 〈‘ring’〉.
(2) Any subject with simultaneous experiences of 〈‘key’〉 and 〈‘ring’〉 that are phenomenally unified with each other will also have an experience with the content 〈‘key’ & ‘ring’〉.
(3) S does not have an experience with content 〈‘key’ & ‘ring’〉.
(4) S's experiences of ‘key’ and ‘ring’ are not phenomenally unified.
(1) appears to be secure. Since the two words were flashed to S simultaneously, it seems to follow that if S was indeed aware of both words then she was aware of them simultaneously. And, as I argued in §9.1, there is good reason to think that each hemisphere in a split-brain subject can support consciousness.
The truth of (2) seems to follow from what in Chapter 5 I called RIP:
Representational Integration Principle (RIP): For any pair of simultaneous experiences e 1 and e 2, if e 1 and e 2 are phenomenally unified then, ceteris paribus, their contents will be available for representational integration.
Note that the kind of integration in question here is rather undemanding. The argument does not even require that S be aware of the presented word as ‘key-ring’. All (2) requires is that S be aware of the words ‘key’ and ‘ring’ together in the form of a representation that has as its content 〈‘key’ & ‘ring’〉. But S doesn't even appear to possess this representation.
Is (3) secure? Is it possible that S could in fact have a conjunctive experience of the words ‘key’ and ‘ring’, but simply be unable to report this content because of (p.198) cognitive bottlenecks of the kind seen in the Sperling experiments (see §4.2)? Although I was at one point attracted to this proposal (Bayne & Chalmers 2003), I now regard it as unsatisfactory. Bottleneck models might be able to explain why S cannot report an experience of 〈‘key-ring’〉 or an experience of 〈‘key’ & ‘ring’〉, but they are ill-equipped to explain why she cannot use the contents of her representation of ‘key’ in ways in which she can use her representation of ‘ring’ and vice versa. It is not as if saying ‘key’ or picking out a ring with one's left hand is more cognitively demanding than saying ‘ring’ or selecting a key from amongst an array of objects. S would presumably have little difficulty forming a conscious representation of ‘key-ring’ were we to project it into either her LVF or her RVF. Furthermore, where there are cognitive bottlenecks, we are usually able to report that we are aware of more than we can directly report. Subjects in the Sperling experiments cannot report the contents of their experience of the entire matrix, but they can (and do) indicate that they had an experience of the matrix whose contents outstripped what they could report. But S fails to have even this indirect form of access to her experiential content. She doesn't, for example, say that she was aware of two words, only one of which she could identify. It seems reasonable to conclude that S manifests no evidence of having an experience of 〈‘key-ring’〉 because she has no such experience.
In sum, premises (1) (2) and (3) are individually plausible, and together they entail phenomenal disunity. The representational disunity argument looks to be secure.
The access disunity argument
A third argument for phenomenal disunity appeals to the relationship between access unity and phenomenal unity. Once again we can use the key-ring experiment as our ‘template’.
(1) S has, simultaneously, an experience with the content 〈‘key’〉 and an experience with the content 〈‘ring’〉.
(2) If simultaneous experiences of 〈‘key’〉 and 〈‘ring’〉 are phenomenally unified with each other then they will be access unified: their contents will be available to the same range of consuming systems.
(3) S's representations of ‘key’ and ‘ring’ are not access unified: although the contents of both states are available for high-level consumption, they are not available to the same consuming systems.
(4) So, S's experiences of ‘key’ and ‘ring’ are not phenomenally unified.
Given that the first premise of this argument is identical to that of the previous argument we need not comment any further on it here. But what about (2) and (3)?
(p.199) In arguing for (2) we might appeal to the Conjoint Accessibility Principle:
Conjoint Accessibility Principle (CAP): For any pair of simultaneous experiences e 1 and e 2, if e 1 and e 2 are phenomenally unified then, ceteris paribus, their contents will be available to the same consuming systems.
As I argued in Chapter 5, the force of the inference from access disunity to phenomenal disunity depends on the degree to which the states in question are access disunified: the less co-accessible the contents of the states, the more reason we will have for denying that they are phenomenally unified. So, that leads us to premise (3): are there any consuming systems to which the subject's experiences of 〈‘key’〉 and 〈‘ring’〉 might both be available? Given the difficulties in individuating consuming systems it would be premature to give a definitive answer to this question (see p. 213), but the evidence suggests that few consuming systems will have access to both contents. In short, the access disunity argument also seems to be secure.
It's time to recap. I have examined three arguments for the claim that split-brain patients are phenomenally disunified. Although none is decisive, the arguments from representational and access disunity are strong, and their combined force goes some way towards explaining—not to mention justifying—the dominance of disunity models. So let us temporarily proceed on the assumption that the typical split-brain patient is indeed phenomenally disunified. The question I turn to now is whether the split-brain patient has two separate streams of consciousness or a merely partially unified consciousness. As we will see, neither proposal is unproblematic.
9.3 Everyday integration
In the period immediately following surgery (right-handed) patients typically experience unilateral apraxia—an inability to execute with their left hand actions that are verbally described or named by the examiner, despite being able to imitate these actions when demonstrated. Long-term deficits following the split-brain procedure include impairments in making and executing decisions, problems with short-term memory, and frequent absent-mindedness.5 Nonetheless, the split-brain procedure has surprisingly little impact on the everyday lives of patients. They can cook, cycle, swim, and play the piano, and naïve observers are rarely aware that they suffer from cognitive impairments. It is far from clear how (p.200) disunity models might account for this ‘social ordinariness’. How is it possible that individuals with a disunified consciousness are able to exhibit such high degrees of cognitive and behavioural unity?
Although the challenge of everyday unity confronts both two-streams and partial unity theorists alike, it is most pronounced with respect to the two-streams account, and I will focus my discussion on two-streams responses to it. However, many of the points that apply to two-streams accounts of everyday unity also apply mutatis mutandis to partial unity accounts.
There are two treatments of the everyday unity of the split-brain patient to be found within the two-streams literature: a ‘contextualist’ treatment and a ‘duplication’ treatment. Both treatments take as their point of departure the fact that in everyday contexts the two hemispheres will have access to (roughly) the same environmental features, and hence their contents will ‘mirror’ each other. As Sperry notes, ‘the two retinal half-fields of the eyeball move as one, and eye movements are conjugate, so that when one hemisphere directs the gaze to a given target the other hemisphere is automatically locked in at all times on the same target’ (Sperry 1974: 7). Further, the restrictions that prevent cross-cuing in the laboratory do not apply outside of the laboratory, and thus the patient can ensure that information is duplicated between the hemispheres even when their environmental relations fail to ensure this.
From this common starting point the two accounts proceed in quite different directions. Contextualists hold that split-brain patients have two streams of consciousness only in certain experimental conditions, and that in everyday life they enjoy a single, fully unified consciousness (Marks 1981; Tye 2003). As such, contextualists argue that the inter-hemispheric mirroring of content is conducive to conscious unity. Duplicationists, by contrast, take the mirroring of content to result in (disunified) phenomenal duplicates—conscious states whose contents are fully identical. In other words, they attempt to account for everyday unity without restricting the scope of the two-streams account. Let us begin with contextualism.
Marks captures the contextualist position in the following question, ‘Why should neural processes unrelated by direct causal routes not be the physical basis for a single mental state?’ (Marks 1981: 23). The idea, I take it, is that split-brain patients have what we might call ‘scattered experiences’. Consider two neural states, N1 and N2, located in the left and right hemispheres respectively. Although each of these states could have realized a K-type experience on its own had the other state been inactive, the patient has only a single token of a K-type experience even when both N1 and N2 are active. Rather than N1 constituting one K-type experience and N2 constituting another K-type (p.201) experience, their ‘sum’ or ‘composite’ realizes a single K-type experience, despite the fact that there are no direct causal connections between them.
The plausibility of this proposal surely depends in no small measure on how experiences ought to be individuated. I have recommended that experiences ought to be individuated in tripartite terms—that is, by reference to times, subject, and phenomenal properties. This conception of experience does leave conceptual space for scattered experiences to inhabit, for it doesn't require that the physical basis of token experience have any kind of internal causal unity. But even if scattered experiences are possible, it is far from clear that they will be able to do the work that the contextualist requires of them. How could the patient's K-type experience produce integrated thought and action if its causal basis is distributed between two causally isolated hemispheres?
Things may be even worse for those attracted to a vehicular conception of experience, as I suspect most contextualists are. How could the mereological sum of N1 and N2 realize a single token experience given the lack of causal commerce between them (Schechter 2010)? N1 and N2 might individually realize a K-type experience, but given that N1's causal influence is restricted to the patient's left hemisphere and N2's causal influence is restricted to the patient's right hemisphere it is difficult to see how the sum or composite of these two states could realize a K-type experience. It looks as though N1 and N2 must realize the K-type functional role as individuals, and if that is right then it is hard to see how their composite—N1 & N2—could itself realize that functional role. The only real loci of causal powers here are N1 and N2. The ‘sum’ of these two events is not itself a neural event in any non-gerrymandered sense, and even if it were it seems not to possess psychologically relevant causal powers of its own.
Further shortcomings of the ‘scattered experiences’ proposal can be brought into view by comparing N1 and N2 with another pair of neural states, N3 and N4, where N3 realizes a K-type experience in one patient and N4 realizes a K-type experience in another patient. Contextualists need to explain why it is that the sum of N1 and N2 constitutes a scattered K-type experience but the sum of N3 and N4 does not. In response, contextualists might be tempted to point out that N3 and N4 occur in the context of different minds—different psychological economies—whereas N1 and N2 do not, but it is precisely this latter claim that stands in need of explanation. Indeed, in a sense the objection from everyday integration just is the challenge of explaining how activity in two (disconnected) hemispheres could give rise to a single mind (or the appearance thereof).
Contextualists don't only need to provide an account of how the appeal to scattered experiences might explain everyday unity in the split-brain, they also need to explain how the transition between everyday and experimental contexts could bring about changes in the structure of the patient's consciousness (p.202) (Nagel 1971). How could moving back and forth between experimental and everyday environments transform the structure of the patient's consciousness from one in which the patient's experiences are all contained within a single stream to one in which they are parcelled out between two streams? After all, it is plausible to suppose that phenomenal structure supervenes on neural structure, and the patient's neural structure appears not to be fundamentally altered by the transition between everyday and experimental contexts.
As I understand their position, Marks and Tye reject the assumption that phenomenal structure supervenes on neural structure. On their view, the transition between unity and disunity occurs merely as a result of changes in the contents of the respective hemispheres. The patient has a unified consciousness in everyday life because—to put it somewhat loosely—the conscious content of the one hemisphere is ‘mirrored’ by that of the other hemisphere, whereas the patient's consciousness is disunified in laboratory conditions because the contents of the one hemisphere differ from those of the other. But as we have noted, it is unclear why we should interpret this ‘mirroring’ in terms of scattered (or ‘disjunctive’) experiences whose neural bases are distributed between the hemispheres as opposed to duplicate experiences that are fully located in one or the other hemisphere.
There is, however, another way in which the contextualist might attempt to account for transitions in the structure of the patient's consciousness, a way that does not involve giving up on the thought that the neural activity in one hemisphere must be causally integrated with that in the other in order for the subject to have a unified consciousness. Drawing on Hurley's ‘vehicle externalist’ account of consciousness, the contextualist might hold that the unity of consciousness involves a ‘dynamic singularity in the field of causal flows that is centred on but not bounded by a biological organism’ (1998: 207). This kind of contextualist might suggest that the difference between everyday and laboratory environments resides in the fact that only in everyday contexts does the ‘causal flow’ that underpins consciousness include both hemispheres within its orbit, whereas it is unable to extend beyond the reach of a single hemisphere when the patient is (say) required to maintain central fixation.6
Although I myself am not much attracted to vehicle externalism, I do think that it ought to be taken seriously. Moreover, being able to ground a plausible account of the split-brain would itself speak in its favour. But in fact the vehicle externalist proposal just sketched is really more of a promissory note than anything else, and it may not be one that is easily cashed. The contextualist who is tempted to go externalist needs to explain why the ‘causal flow’ that (p.203) constitutes the patient's phenomenal field might be able to straddle both hemispheres in everyday contexts but not in experimental contexts.
As far as I can see, the only plausible move to be made here appeals to the thought that the cognitive demands that split-brain patients face in experimental contexts are more taxing than those that everyday life places on them, and that it is these demands that govern the structure of the relevant causal flows. Although there is something to this suggestion (as we will see in §9.5), it is highly implausible to suppose that high cognitive load could generally lead to phenomenal division. Think of shadowing tasks, in which one has to repeat the words presented to one's left ear while listening for target words in the auditory stream presented to one's right ear. Such tasks are taxing, but they don't cause one's auditory states to divide into two distinct streams of consciousness. A further difficulty with the ‘cognitive load’ proposal is that one can elicit representational and access disunities of the sort the two-streamer appeals to even under conditions of minimal cognitive load. After all, identifying the word ‘key-ring’ does not seem to be particularly onerous.
It is time for an interim summary. I have argued that contextualist versions of the two-streams model are not promising. Neither the ‘isolated realizers’ version of contextualism developed by Marks and Tye nor the vehicle externalist version of contextualism that we have just examined provide us with a plausible account of everyday unity in the split-brain.7
Does this mean that the two-streams model is sunk? No. Contextualism is in fact something of a minority view within the two-streams camps, and most two-streamers attempt to account for everyday unity in the split-brain by adopting what I call the duplicationist account. To return to our earlier example (see p. 200) rather than supposing that N1 and N2 form the basis of a single K-type experience (as the contextualist does), the duplicationist takes N1 to form the basis of one K-type experience and N2 to form the basis of another, numerically distinct, K-type experience.8
This duplicationist strategy avoids many of the objections that troubled the contextualist's treatment of the inter-hemispheric mirroring of content, but it has troubles of its own. The primary objection to it is that the very notion of duplicate experiences is of dubious coherence. As I pointed out in Chapter 3, the tripartite account of experiences leaves no conceptual space for duplicates, for by definition duplicates have the same content, are had by the same subject of experience, and occur simultaneously. (Remember that we are here identifying the subject with the split-brain patient.) So, in order to embrace the (p.204) duplication gambit the two-streams theorist must either reject the tripartite conception of experience, or argue that each stream is to be identified with a distinct subject of experience.
Of course, the duplicationist could reject the tripartite account of experience. She could, for example, individuate experiences in neural or functional terms. Such views might allow a person to have multiple experiences with the same phenomenal content as long as the various experiential tokens involve (supervene on; are grounded in) distinct neural events. Although the issues here are complex, I think that we have good reason to retain the tripartite account. Experiences are properties of subjects of experience—they are not properties of neuronal assemblies or sub-personal homunculi. Talk of ‘the left hemisphere doing this’ or ‘the right hemisphere knowing that’ is tempting—indeed, I myself have engaged in it on occasions—but it needs to be taken with a large pinch of salt. And the fact that experiences are states of subjects ought to be reflected in their identity conditions.
Should the two-streamer allow that the split-brain subject might be or ‘house’ two subjects of experience? I think the costs of going down this road are high. As we have already noted, the patients don't think of themselves as multiple. There is no sense in which the left hemisphere thinks of itself as one subject and the right hemisphere thinks of itself as another subject. Not only do patients themselves not think of themselves as multiple, experimentalists do not typically think of them as multiple either. Although researchers are sometimes inclined to think of themselves as dealing with two subjects, such contexts are very much the exception to the rule. For the most part, there is little doubt that ‘I’-thoughts associated with the left hemisphere refer to the same individual as ‘I’-thoughts associated with the right hemisphere, and if ‘I’-thoughts associated with each hemisphere are co-referential then we must be dealing with a single subject of experience. These considerations fall short of demonstrating that split-brain patients cannot enjoy duplicate experiences, but they do put that proposal under significant pressure.
There are two final issues to consider before I draw this section to a close. The first is this. At the start of this section I noted that contextualist and duplicationist accounts of everyday unity in the split-brain share a common assumption—namely, that in everyday environments the contents of the patient's two hemispheres will ‘mirror’ each other. I have proceeded by granting this assumption, but in fact there is good reason to challenge it. The two hemispheres might enjoy a partial mirroring of content in everyday environments, but given hemispheric specialization in perceptual processing it is very unlikely that their contents will fully mirror each other. Secondly, even when the two hemispheres do receive the same perceptual input, this may not lead to (p.205) behavioural integration, for the behavioural upshot of perceptual information is not a function of that information alone but depends on the broader psychological context in which it is processed. For example, stimuli that the left hemisphere regards as aversive might not be so regarded by the right hemisphere, or vice versa. Given that the two hemispheres have different memory stores and cognitive styles (Roser & Gazzaniga 2004; Schiffer et al. 1998), even complete mirroring of content might be expected to lead to behavioural disunity were both hemispheres conscious in parallel.
We have seen that there are two lines of response open to the two-streams theorist in accounting for everyday behavioural unity in the split-brain: a ‘contextualist’ response, according to which split-brain patients enjoy phenomenal disunity only in laboratory contexts, and a ‘duplicationist’ response, according to which everyday unity is accounted for by supposing that the subject's two streams are mirror images of each other. In my view, neither proposal succeeds. Let us turn now from the prospects of the two-streams account of the split-brain to those of the partial unity model.9
9.4 Partial unity
The two-streams model goes hand in glove with the assumption that the split-brain operation bisects a single cognitive-behavioural workspace into two such workspaces, each of which is sealed off from the other. According to this picture, information presented to each hemisphere can be integrated only with information available to that hemisphere, and each of the patient's consuming systems has access to the contents of one and only one workspace. This picture is problematized by the phenomenon of inter-hemispheric integration.10
First, some background. Particular sections of the corpus callosum are responsible for the transfer of distinct types of information. The anterior mid‐body transfers motor information, the posterior mid‐body transfers somatosensory information, the isthmus transfers auditory information, and the splenium transfers visual information.11 This specialization suggests that it should be possible for patients to exhibit domain-specific splitting, and indeed this turns out to be the case.
Gazzaniga and LeDoux describe a patient (D.H.) with a partial callosotomy, who appeared to be split for touch but not vision:
(p.206) Tactual information in the left hand and right hemisphere remained isolated from the right hand and the left hemisphere. Yet, when a visual stimulus, such as the picture of an apple, was lateralized to either hemisphere, either hand could manually retrieve the apple, unaided by visual exploration. (Gazzaniga & LeDoux 1978: 10; see also Gazzaniga & Freedman 1973)
Even patients with a complete callosotomy appear to be capable of selective inter-hemispheric integration. Gazzaniga et al. (1963) describe a patient who was unable to localize with the one hand brief touches that had been applied to the foot, leg, arm, hand, and trunk of the contralateral side of his body, but was able to both localize with either hand and verbally report touches that had been applied to any region of his head or face. Selective inter-hemispheric integration is not restricted to touch but can also be found within vision. Ambient (peripheral) information about relative motion and size can be transferred between hemispheres (Trevarthen 1970; Trevarthen & Sperry 1973), as can information about visual field location (Holtzman 1984). Apparent motion can also occur across the visual hemi-fields: stimuli that are presented to each hemi-field sequentially can generate a percept of a single stimulus moving from one hemi-field to the other.12
Finally, there is evidence for the integration of perceptually based categorical information. In one experiment, patients were required to determine which of three pictures shown in one hemi-field matched a picture shown in the other hemi-field (Cronin-Golomb 1986; see also Zaidel & Iacoboni 2003). For example, the patient might be presented with a picture of a fish in his LVF and pictures of a pig, spider, and duck in his RVF. (The duck matches the fish, for both animals are associated with water.) Each of the three patients could match inter-field stimuli to high levels of reliability, suggesting that high-level, categorical information could be transferred between hemispheres.
Inter-hemispheric integration suggests that the ‘split’ of consciousness seen in the split-brain might be partial rather than complete—that some patients might have a single, partially unified stream of consciousness rather than two streams of consciousness. To convert inter-hemispheric integration into an argument for partial unity we need only suppose that inter-hemispheric integration involves token experiences that ‘straddle’ the two hemispheres. Consider again D.H., who is split for touch but not vision. Suppose that D.H. has, at the one time, tactile experiences in both hands and visual experiences in both hemi-fields. We have some reason to think that although his visual experiences are unified with each (p.207) other this is not the case for his tactile experiences. But what is the relationship between D.H.'s visual experiences and his tactile experiences? Presumably each of his visual experiences is unified with both his left hand tactile experience and with his right hand tactile experience, even though those experiences are not unified with each other. But if that is right, then D.H has a partially unified consciousness, for (at a single time) he will have a pair of experiences that are unified with each other but not with a third experience (see §2.4). Although D.H. is unusual among split-brain patients in being split for touch but not for vision, we have seen that even the most split of split-brain patients appears to retain the ability to integrate certain types of information, and comparable arguments could be constructed for them also.
The argument from integration is provocative. Without doubt, it demonstrates that the partial unity model of the split-brain needs to be taken seriously—much more seriously than it is usually taken. But it is not, I think, wholly decisive. To provide a watertight argument from integration one needs to show both that the relevant behavioural responses are indeed a manifestation of inter-hemispheric integration and that the integration in question is conscious. Neither of these tasks will be at all trivial to accomplish.
There are (at least) two ways in which an apparently integrative response might actually result from the operation of a single hemisphere. On the one hand, individual hemispheres sometimes have a wider range of processing abilities than is often suspected. For example, there is evidence that the right hemisphere has some ability to monitor the full visual field even when the corpus callosum is severed (Mangun et al. 1994). Given this, one has to ensure that visual field judgements that appear to involve the integration of LVF and RVF content cannot be produced by the right hemisphere under its own steam.
Even when behaviour does involve both hemispheres it may not involve inter-hemispheric integration. In an elegant study, Kingstone and Gazzaniga (1995) showed just how easy it is for split-brain patients to produce integrative behaviour in the absence of true representational integration. In an initial experiment, their subject (J.W.) appeared to demonstrate inter-hemispheric integration. For example, when the word ‘ten’ was presented to his left hemisphere (RVF) and the word ‘clock’ was presented to his right hemisphere (LVF) J.W. drew a clock showing 10 o'clock. However, Kingstone and Gazzaniga showed that J.W. produced his ‘composite’ pictures not by integrating information between hemispheres, but by transferring control of his drawing hand from one hemisphere to another. Using a single hand, J.W. would first sketch a clock under the guidance of his right hemisphere, and then add clock-hands indicating 10 o'clock under the guidance of his left hemisphere. As Kingstone and (p.208) Gazzaniga put it, the only integration to be found here occurred ‘on the sheet of paper in the drawing itself’ (1995: 324; see also Miller & Kingstone 2005).
Moreover, even when inter-hemispheric integration occurs it may not involve a ‘phenomenal bridge’ between the patient's two hemispheres (Corballis 1994). The argument from integration turns on the assumption that bilateral integration involves experiential events that ‘straddle’ the patient's two hemispheres, but conscious states in one hemisphere might have an effect on those in the other hemisphere even in the absence of inter-hemispheric phenomenal unity. A plausible example of this involves the inter-hemispheric transfer of affect. In one oft-cited vignette, a patient was unable to verbally identify a picture of Hitler that had been presented in his left visual field, but his description of it was clearly influenced by negative affect—affect that had presumably been generated in the right hemisphere (Sperry et al. 1979). Rather than suppose that the patient had a single experience of negative affect that bridged his two hemispheres, perhaps a right hemisphere state of negative affect merely primed a similar state in his left hemisphere. In other words, it seems plausible to suppose that inter-hemispheric integration can take place outside of consciousness.
In fact, Naikar (1996) advances just this treatment of apparent motion across the vertical midline in patient L.B. Rather than supposing that L.B.'s experience of the first light (as located in one hemifield) was phenomenally unified with an experience of the second light (as located in the other hemifield), Naiker suggests instead that (unconscious) sub-cortical mechanisms registered a shift of spatial attention from one hemi-field to the other, giving rise to a judgement of apparent motion across the vertical midline.
Might all instances of inter-hemispheric integration be explained in this deflationary manner? Perhaps; it is difficult to tell from the experimental literature. In principle, there are two ways in which one could rule out priming treatments of inter-hemispheric integration. One way to block priming accounts of integration would be to show that the kind of integration in question could not have occurred unless the states involved were phenomenally unified with each other. For example, one could undermine the ‘priming’ account of apparent motion by showing that visual representations cannot produce a representation of apparent motion unless they are both conscious and phenomenally unified with each other. In practice however, this line of response will be difficult to establish, for we know so little about the kinds of information transfer that can occur outside of phenomenal integration. Another strategy one might employ in order to block priming interpretations would be to show that the representations produced by the relevant instance of inter-hemispheric integration are not only integrated with both right- and left-hemisphere conscious states, but that their (p.209) contents are also available to both right- and left-hemisphere ‘consuming systems’. Although perhaps more feasible than the previous strategy, this strategy is also likely to be difficult to implement. Not only would one need to show that the consuming systems in question are unique to each hemisphere (rather than shared between the hemispheres), but one would also need a principled way of assigning consuming systems to hemispheres (and, by implication, to streams of consciousness). Although we can be reasonably confident that some behavioural capacities—such as control of the contralateral hand for fine motor movements—involve consuming systems that are unique to each hemisphere, other forms of motor control can be passed back and forth between hemispheres, so we cannot individuate consuming systems in terms of coarse-grained behavioural capacities. In short, it will not be easy to rule out deflationary ‘priming’ interpretations of inter-hemispheric integration.
But the deepest challenges facing the partial unity model are not empirical but conceptual. In weighing the overall plausibility of the approach we need to consider both its empirical credentials and its intrinsic plausibility (or the lack thereof). In Chapter 2 I noted that the question of whether the model is even coherent is far from settled. Contrary to what has often been claimed, the problem with partial unity is not that we cannot imaginatively project ourselves into a partially unified phenomenal perspective. Rather, the problem with partial unity is that it seems to be inconceivable. It seems to be central to our notion of a phenomenal perspective that phenomenal unity cannot fragment in the way that partial unity would require. This worry is far from decisive—after all, when it comes to consciousness, inconceivability judgements are notoriously controversial—but it surely has some force. Perhaps it would be reasonable to waive such worries if the evidence in favour of the partial unity model were overwhelming, but that is far from being the case if the considerations presented above are sound.
Where does this leave us? Although §9.2 appeared to provide a solid case for disunity interpretations of the split-brain, we have seen that both the two-streams and the partial unity versions of the disunity approach face formidable challenges. In light of this, perhaps we ought to reconsider the assumption that split-brain patients do indeed have a disunified consciousness. Maybe the case for disunity developed in §9.2 was not quite as solid as it appeared to be.
9.5 The switch model
In the previous chapter I suggested that the cognitive and behavioural disunity seen in hypnosis is best accounted for by supposing that hypnotised subjects (p.210) have a single stream of consciousness that switches between two streams of processing. Drawing on ideas that were first advanced by Jerrre Levy (Levy 1977, 1990), I suggest that we can account for the cognitive and behavioural disunities seen in the split-brain in precisely the same way (Bayne 2008b). Rather than suppose that the patient's two hemispheres are conscious in parallel, we should think of consciousness in the split-brain as moving or switching from one hemisphere to another. Although both hemispheres can process information concurrently, they take turns supporting consciousness. In effect, the switch model paints patients as suffering from a kind of fluctuating perceptual extinction: when the left (right) hemisphere is activated stimuli in the LVF (RVF) are typically ignored in favour of stimuli in the RVF (LVF). The patient might be conscious of the word ‘key’ (due to right hemisphere activity), or she might be conscious of the word ‘ring’ (due to left hemisphere activity), but she will not be conscious of both ‘key' and ‘ring’ at once, even when the two words are simultaneously presented to her.
An important source of evidence for the switch model is to be found in an experiment conducted by Levy and colleagues involving chimeric stimuli—that is, stimuli that are created by conjoining two similar half-stimuli at the vertical midline (Levy et al. 1972). A chimeric face might be constructed from the left side of one person's face and the right side of another person's face. On some trials subjects were asked to point to a matching stimulus (either with their right or left hands), whilst on other trials subjects were required to name the stimulus. On almost all trials, subjects indicated only one match for each of the chimeric stimuli.
In their original report, Levy and co-authors suggested that the non-responding hemisphere had a rival percept that it failed to express, and it is this interpretation of the experiment that seems to have entered the literature (see e.g. Marks 1981). However, in subsequent work Levy argued for a very different account of the data:
With one half-stimulus joined at the mid-line to a different half-stimulus to make a ‘chimera,’ each hemisphere would receive equivalent, but different stimulus input, and if two perceptions were gained, they would be in conflict as evidence concerning the object of interest, and the motor responses guided by these percepts would be in conflict… [But] if…perception is the preparation to respond, then, except in special circumstances, there should be a single perception linked to a single response under conditions of competitive stimulus input…Our studies overwhelmingly confirmed the predictions of this conceptual model. For all patients examined, and for tasks including the perception of faces, nonsense shapes, pictures of common objects, patterns of Xs and squares, words, word meaning, phonetic images of rhyming pictures, and outline drawings to be matched to colors, patients gave one response on the vast majority of competitive trials. Further, the nonresponding hemisphere gave no evidence that it had any perception at all. Thus, if (p.211) the right hemisphere responded there was no indication, by words or facial expression, that the left hemisphere had any argument with the choice made, and, similarly, if the left hemisphere responded, no behavior on the part of the patient suggested a disagreement by the right hemisphere. (Levy 1990: 235; see also Levy 1977)
Levy's ‘switch’ account of this data strikes me as rather more elegant than her earlier ‘non-responding’ account. Rather than take the non-responding hemisphere to have a conscious percept that it was either unable or unwilling to express, it seems simpler to suppose that whatever percepts it might have had were unconscious.
Levy's experiments are not the only evidence of perceptual extinction in the split-brain. In a follow-up experiment, six split-brain subjects were given a mixture of dot and numeral counting exercises via the usual tachistoscopic method (Teng & Sperry 1974). Between one and five dots were flashed either in the LVF or RVF alone, or in the two visual fields simultaneously. All but one of the six split-brain subjects showed massive amounts of extinction, with patients reporting only those stimuli restricted to a single hemi-field. Gazzaniga and colleagues also found perceptual extinction in a study that examined the bilateral integration of tactile stimuli in a split-brain patient. The authors noted that ‘it was as if a strong shift of attention to one hemisphere had tended to extinguish perceptual awareness in the other’ (Gazzaniga et al. 1963: 211).
The dynamics of switching in the split-brain appears to depend on a number of factors, one of which is the response required of the subject. Some studies have noted that differences between motor responses can influence the lateralization of awareness in the split-brain. For example, Levy et al. (1972) reported that verbal responses tended to favour completion of the right half of the stimulus, while manual responses favoured completion of the left half of the stimulus. Trevarthen found that the lateralization of conscious perception could be switched merely by requiring subjects to respond with one hand rather than the other (Trevarthen 1974b: 195). In another series of studies, Levy and Trevarthen found that the lateralization of awareness could also be switched by modulating the cognitive content of the responses required by patients (see Figure. 9.2). For example, asking patients to match chimeric stimuli based on their visual appearance favoured the LVF (implicating the right hemisphere) whereas instructing them to match stimuli based on their function favoured the RVF (implicating the left hemisphere) (Levy & Trevarthen 1976; see also Sperry 1974 and Iacoboni et al. 1996). As Teng and Sperry put it, ‘the apparent distribution of attention between the two hemispheres is not static, but may change with the nature of the task…’ (Teng & Sperry 1973: 137). The response-dependence in evidence here is not unique to the split-brain—we have seen it previously in unilateral neglect (Chapter 5) and anosognosia (Chapter 7)—but it does seem to be remarkably pronounced in the split-brain syndrome.
Source: Based on a figure in Levy 1977
Further support for the switch model is provided by studies of auditory processing in the split-brain. Because information from each ear projects to both hemispheres, commissurotomy does not itself lead to the lateralization of (p.213) auditory information. However, lateralization can be achieved by means of a dichotic listening paradigm, in which patients are simultaneously presented with competitive stimuli to each ear. Under these conditions, ipsilateral processing is suppressed in favour of contralateral processing, with inter-hemispheric competition ensuring that information from only a single ear enters consciousness at any one point in time. Moreover, which of the two hemispheres dominates auditory processing can be modulated by task demands. In one study, Milner and colleagues found that patients presented with different digits to each ear complained that they could hear nothing in the left ear even though they had expected to hear numbers in both ears (Milner et al. 1968; see also Milner et al. 1990). Because patients were required to verbally report the numbers, the left hemisphere suppressed input from the ipsilateral ear in favour of input from the contralateral ear. However, right hemisphere activation could be elicited by changing the response required of patients. Subjects were presented with pairs of competing instructions, one to each ear, with each instruction naming an object to be retrieved by touch from amongst a group of nine objects hidden behind a screen. Subjects who were instructed to use their left hand to retrieve objects tended to follow the instruction given to the left ear, with partial to complete neglect of the instruction presented to the right ear. In other words, use of the left hand appears to have favoured right hemisphere processing, leading in turn to the suppression of auditory information entering through the right ear.
We can now see where the arguments for the disunity models of the split-brain go wrong. Consider again the key-ring experiment. The representational and access disunity arguments assume that the patient's two hemispheres must be simultaneously conscious because the stimuli are simultaneously projected to the patient's visual hemifields, and because each hemisphere can respond to the stimulus in its hemi-field as and when required. But both inferences are contentious. Perhaps the ability of patients to respond in this way is the result of consciousness switching rapidly and effortlessly between hemispheres in response to the demands of the patient's context. The hemisphere that is silent on any one trial may be so because it is unconscious rather than because it is unable (or unwilling) to ‘speak’.14
So much for how the switch model might account for the experimental data—how might it explain everyday unity in the split-brain? There are three lines of (p.214) thought we might appeal to here. Most straightforwardly, it is possible that split-brain patients generally get by on a single conscious hemisphere. Split-brain control is usually dominated by the left hemisphere, and experimenters often comment on the difficulties they confront in eliciting responses from the right hemisphere. Even when the right hemisphere does initiate a response, the left frequently takes over and finishes it, sometimes to the detriment of the patient's performance (Nebes & Sperry 1971; Sperry 1974; Zaidel & Sperry 1973).
Secondly, inter-hemispheric switches might be both smooth and rapid, generating the impression that the patient is conscious of more than she is—in much the way that our fluid interaction with the environment perhaps generates the impression that we are conscious of more than we are. The ability of some patients to control both arms from each hemisphere might also mask inter-hemisphere switching. A single right-handed action might be under the control of first one hemisphere and then the other (Kingstone & Gazzaniga 1995).
Finally, it may be that the lower cognitive demands of everyday life allow patients to deploy a kind of non-focal, low-level attention that can straddle both hemispheres. Trevarthen notes that in certain situations patients adopt a ‘particular condition of mental orientation’, in which awareness is bilaterally distributed. In this condition,
visual events were noticed by commissurotomy patients in an undivided bilateral and temporally unified space around the body and their strength, quality or motion, and general spatial configuration were related to this space much as they would be by normal subjects. (Trevarthen 1974b: 195)
Whereas experimental contexts generally require focused attention, everyday life may permit patients to enter into a state of non‐focal awareness of the kind that Trevarthen describes. Moreover, although focal consciousness is restricted to a single hemisphere at a time, it might be possible for non-focal awareness to be distributed across the patient's two hemispheres.
The foregoing goes some distance towards establishing the switch model as a viable account of the structure of consciousness in the split-brain. I turn now to the question of whether this prima facie plausibility can withstand critical scrutiny.
9.6 Objections and replies
Let us examine some objections to the switch model.
First objection: ‘We know that split-brain patients have independent attentional systems in each hemisphere. Given the intimate connections (p.215) between attention and consciousness, surely attentional disunity in the split-brain constitutes strong evidence for phenomenal disunity.’
Do split-brain patients have independent attentional systems? There is certainly some evidence for this conclusion. An early study found that the two hemispheres could carry out visual discrimination tasks independently and in parallel, suggesting that there was some degree of attentional division in the split-brain (Gazzaniga & Sperry 1966). This finding is supported by a number of other studies that indicate that there is some degree of hemispheric independence in visual attention within the split-brain (Arguin et al. 2000; Luck et al. 1989; Luck et al. 1994; Mangun et al. 1994).
However, these findings need to be balanced against a number of other findings that suggest that attention remains fundamentally unified in the split-brain. Based on an experiment requiring split-brain subjects to either synchronize the taps of their two index fingers or to alternate them as rapidly as possible, Kreuter et al. (1972) concluded that ‘a maximum effort by one hemisphere does withdraw capacity from the other, an effect which in the absence of the corpus callosum is presumably mediated by a “capacity distributing system” located in the brain stem’ (Kreuter et al. 1972: 460). Holtzman & Gazzaniga (1982) found that cognitive load in one hemisphere interfered with performance in the other hemisphere on a task requiring the subject to match the presented stimulus to a template, and a follow-up study involving lexical memory produced a similar result (Gazzaniga 1987).
Semantic priming studies also suggest that there is attentional integration in the split-brain. In one experiment, L.B. was presented with two words, one to each visual hemi-field, and instructed to categorize (living vs. non-living) the RVF word but ignore the LVF word (Lambert 1991). His performance on this task showed an inhibitory effect from the unattended word that closely resembled that seen in neurologically normal individuals. A subsequent study found normal levels of negative priming in three split-brain patients: LVF items inhibited responses to categorically related items in the RVF (Lambert 1993). Lambert concluded that there is a single system of selective attention in the split-brain, involving sub-cortical connections.
Indeed, even studies of visual attention are far from unanimous in suggesting that the division of the corpus callosum brings about a corresponding division in visual attention. In a spatial priming study, patients were required to produce speeded responses to targets in either the LVF or RVF (Holtzman et al. 1984). Targets were preceded by spatial cues that directed the subject's attention to either the LVF or RVF (focused attention) or both visual fields simultaneously (divided attention). These conditions were compared with neutral trials on (p.216) which the cue carried no information about the location of the target. The authors found that valid focused attention trials showed a response time advantage over divided attention and neutral trials, suggesting that some form of spatial attention remains unified in the split-brain.
Studies of sustained attention present a somewhat mixed picture. An early study concluded that complete commissurotomy patients had independent systems responsible for sustained attention with this state of divided attention being maintained for prolonged periods (Ellenberg & Sperry 1980). In contrast, a later vigilance study concluded that split-brain patients appear to have a single, and rather depleted, source of sustained attention (Dimond 1976). There were frequent occasions during the course of this study when one or other of the two hemispheres appeared to be completely unresponsive to external stimulation. Perhaps, the author suggested, these periods of unresponsiveness resulted from ‘the loss of attentional capacity from one hemisphere, without at the same time leading to an overall loss of attentional capacity’ (Dimond 1976: 354).
What should we make of all of this? I'm not sure that there is any tidy story to tell. Gazzaniga once claimed that ‘attention remains largely integrated in the split-brain patient’ (Gazzaniga 1987: 119). That claim may have been something of an over-statement but it is certainly true that many forms of attention appear to remain integrated in the split-brain. At the very least, split-brain patients do not have ‘two attentional systems’. Some forms of attention may be split in the split-brain but others appear to remain ‘intact’, either because attention straddles both hemispheres or because it alternates between hemispheres. In fact, rather than undermining the claim that consciousness remains unified in the split-brain the work on attention may actually provide some support for the view.
Second objection: ‘The switch model is anatomically implausible. How could consciousness move between hemispheres given that the main band of fibres connecting the cortical regions has been severed?’
I don't have a model of how the switch model is neurophysiologically implemented, but it is possible to deflate the objection without such a model. As we have just seen, some attentional systems remain unified in the split-brain, and it is possible that these systems play an important role in ‘shuttling’ consciousness between hemispheres as and when required. Secondly, we have independent evidence that consciousness can be unified in the absence of a corpus callosum, for individuals born without a corpus callosum (acallosals) show few of the signs of behavioural disunity that characterize the (p.217) split-brain syndrome.15 And, third, we know that the mechanisms responsible for modulating wakefulness remain unified in the split-brain. Split-brain patients do not exhibit unihemispheric sleep (Sperry 1974), unlike dolphins, grey whales, white whales, southern sea lions, northern fur seals, and certain bird species.16 Although the relationship between the mechanisms underlying wakefulness and those underlying consciousness is uncertain, it is possible that the sub-cortical systems implicated in modulating wakefulness and arousal play an important role in maintaining coherent relations between the hemispheres.
Third objection: ‘What about the data that appeared to motivate the partial unity account? How does the switch model account for bilateral integration in the split-brain?’
The first point to make is that the force of the evidence for conscious bilateral integration is somewhat uncertain. As I noted earlier (§9.4), it is possible that behaviour that appears to involve inter-hemispheric integration might actually be produced by a single hemisphere exerting control over (say) both the right and left hands. Furthermore, even where inter-hemispheric integration does occur, it might involve only inter-hemispheric priming rather than any kind of phenomenal bridge between the hemispheres.
But suppose that split-brain patients do occasionally have experiences that ‘straddle’ their two hemispheres: would that be at odds with the switch model? Not necessarily. The critical question is whether such experiences are also unified with experiences that are not themselves unified with each other. Consider D.H., who is split for touch but not vision. Suppose that, at a particular time, D.H. is presented with visual stimuli in both hemi-fields and tactile stimuli to both hands. Perhaps it is possible that D.H.'s visual experiences can be phenomenally unified with either the right hand stimulus or the left hand stimulus, but not both at once. More generally, it is possible that although consciousness can straddle the two hemispheres in different places at different times, it cannot fragment in the way demanded by partial unity. Of course, this is all very speculative. In order to evaluate this proposal we need to know more about what patients such as D.H. can be simultaneously conscious of—in particular, we would need to know whether D.H. can be simultaneously (p.218) conscious of tactile stimulation to both right and left hands. The issue of partial unity must be left open here. Although it could turn out that we need to reject the full-unity version of the switch model in favour of one that allows partial unity, we ought not embrace partial unity just yet.
Fourth objection: ‘The switch model holds that the contents of consciousness are sequentially informed by processing in each hemisphere. If this were so, however, then one would expect patients to report sudden changes in the contents of their consciousness. After all, subjects who experience binocular rivalry are aware of the alteration between rival percepts, so why are split-brain patients apparently unaware of the alteration between rival percepts if, as the switch model claims, they are subject to inter-hemispheric rivalry?’
I regard this as one of the most challenging objections to the switch model, and I am not sure that I have a completely satisfactory solution to it. But the following observations do, I think, blunt its force.
For one thing, there is extensive representational overlap between hemispheres (as two-streams theorists have often noted). Perhaps the patient's perspective of the world when her consciousness is informed by right hemisphere processing will not in general be significantly different from that which she enjoys when it is informed by left hemisphere processing. But of course there will be occasions in which the contents of the patient's experience will undergo radical changes as consciousness switches from one hemisphere to another. Should we not expect the patient to notice such changes when they occur? Perhaps not. The distinction between changes in the contents of consciousness and the conscious representation of those changes is no mere formal nicety but has some psychological robustness, as we noted in discussing change blindness. Indeed, it is not uncommon for patients to be unaware of quite radical changes to the contents of their own consciousness (see Chapter 7). For example, patients with achromatopsia (agnosia for colour) frequently fail to realize that they have lost their experience of colour, and are often puzzled by questions about the reliability of their colour vision (Cowey 2009). Unilateral neglect furnishes us with even more striking examples of the failure to keep track of ‘gaps’ in one's own consciousness. In a famous study, patients were asked to imagine themselves standing in Milan's Piazza del Duomo with their back to the cathedral (Bisiach & Luzzatti 1978). As one might predict, they failed to describe buildings on their left. When asked immediately afterwards to describe what they would see if looking at the cathedral from the opposite end of the square, the same patients named the previously neglected buildings but neglected those that they had just mentioned. At no point did the patients (p.219) attempt to integrate their successive reports, nor did they express any concern about the obvious inconsistency between them. In the same way that the ability to track perceptual continuity is impaired by the very damage that causes neglect, so too it is possible that the very operation which causes inter-hemispheric rivalry in the split-brain also prevents patients from becoming introspectively aware of that rivalry. Perhaps, as I noted in Chapter 7, it is possible that introspective access to an experience requires activating the very neural areas that generated it in the first place (Bisiach 1988). If that were the case, then in order to be aware of inter-hemispheric switches in consciousness the patient would be required to simultaneously activate both hemispheres—precisely what the switch account rules out.
Fifth objection: ‘Suppose that the contents of consciousness in the split-brain do switch between hemispheres in the way that you've been suggesting. Wouldn't it be more plausible to describe such a scenario as one in which the subject has two streams of consciousness which are sequentially active rather than a single stream of consciousness whose contents are drawn from each hemisphere in succession? Come to think of it, what exactly is the difference between claiming that consciousness in the split-brain switches between two streams of cognitive activity and claiming that split-brain patients have two streams of consciousness that operate sequentially?’
Let us begin with the question of what, if anything, distinguishes these two versions of the switch model. If we think of a stream of consciousness in purely phenomenological terms, then the difference between the one-stream and two-streams versions of the switch model reduces to the question of whether inter-hemispheric switching interrupts the continuity of consciousness. If the split-brain subject enjoys no interruption in the continuity of consciousness, then we ought to say that he or she has but a single stream of consciousness. Of course, it is difficult to know whether or not inter-hemispheric switching brings with it a phenomenal gap—a hiatus in experience. The patient reports no such gap, but this has little to no evidential weight, for if there were such a gap it would not be available for report.
If we individuate streams of consciousness in terms of the underlying mechanism(s) responsible for consciousness, then the difference between the one-stream and two-streams versions of the switch model turns on whether there is a single mechanism of consciousness in the split-brain or two. On this point there is a clear difference between the two versions of the switch model. The single-stream model holds that consciousness in the split-brain patient has a singular (presumably sub-cortical) ground, whereas the two-streams model holds that the left and right hemispheres have independent mechanisms of consciousness—that there is no common substrate of consciousness in the split-brain. On this conception of (p.220) streams of consciousness, it seems to me that the evidence clearly favours the one-stream model. There is but one substrate of consciousness in the split-brain, and it is grounded in an (undivided) sub-cortical network. Right and left hemisphere activation in the split-brain generates conscious content only because it is suitably incorporated into that system (see also §10.6).
I have argued that the near-consensus in favour of disunity models of the split-brain is far from secure. Such models come in two forms: two-streams models and partial unity models. Two-streams models face challenges on two fronts: not only do they struggle to account for the unity of the split-brain patient's everyday behaviour they may also be at odds with the fact that patients demonstrate some degree of inter-hemispheric integration within experimental contexts. Does this mean that we should think of the split-brain in terms of partial unity? I think not. Although partial unity models may provide a better account of the behavioural evidence than two-streams models, it is not at all clear that such models are coherent. Preferable to both two-streams and partial unity models is the switch model, which—I argued—provides the best account of both the behavioural disunities that split-brain patients exhibit under experimental conditions and the behavioural unities that they exhibit outside of such contexts.
Of course, any claim about ‘the’ structure of consciousness in the split-brain must be at best provisional. Despite the undoubted utility of the notion of a ‘split-brain syndrome’, the individual differences between split-brain patients are not insignificant and should not be overlooked (Kinsbourne 1974). Indeed, there may not even be a single account of ‘the’ structure of consciousness in any one split-brain patient. Perhaps there are patients who have a single unified consciousness in some contexts, two streams of consciousness in other contexts, and a partially unified consciousness in still other contexts. Leaving these speculations to one side, it is undoubtedly the case that the split-brain data are extremely complex, and it would be foolish to pretend that the evidence points unequivocally in favour of the switch model. However, that is not my claim. My claim, rather, is that the switch model ought to be regarded as a serious rival to the two-streams and partial unity models. And if the switch model is right, then the split-brain syndrome does not undermine the unity thesis.
Does this mean that the unity thesis is secure? I think so. In Chapter 6 we saw that there is no good evidence of phenomenal disunity within everyday forms (p.221) of conscious experience. In Chapter 7 I argued that although anosognosia, schizophrenia, and multiplicity involve the breakdown of certain forms of the unity of consciousness, none of these phenomena threaten the unity thesis. In Chapter 8 we saw that the appearances of disunity within hypnosis can be accounted for in terms of switches in consciousness. And in this chapter I have defended a parallel model of the structure of consciousness in the split-brain syndrome. The upshot of all this appears to be that there is indeed a deep and robust sense in which consciousness is unified, a sense that is captured by the unity thesis.
With the conclusion of this chapter I also conclude Part II of this book. In Part III I examine some of the potential implications of the unity of consciousness. In particular, I will explore ways in which the unity of consciousness might inform our understanding of consciousness, bodily experience, and the self. (p.222)
(1) For reviews of the split-brain syndrome see Baynes & Gazzaniga (2000); Beaumont (1981); Gazzaniga (1995, 2005); Seymour et al. (1994); Spencer et al. (1988); Wolford et al. (2004); Zaidel et al. (2003).
(4) For philosophy see Moor (1982); for neuropsychology see Sperry (1976, 1984); Trevarthen (1974a) and (1974b). It is also possible to read Nagel's ( 1971) account of the split-brain as prefiguring the partial unity model in certain ways, although the central thrust of Nagel's position is that consciousness in the split-brain has no determinate structure.
(6) I should add that I am considering here a contextualist appropriation of Hurley's vehicle externalism. I am not suggesting that this was Hurley's own position.
(9) The material in this section has benefited greatly from discussions with Lizzie Schechter.
(14) Note that the role assigned to contextual demands by the switch account differs from that assigned to them by contextualist versions of the two-streams account (see §9.3). According to the switch account, context plays a role in determining which of the two hemispheres monopolizes consciousness and suppresses the other hemisphere, whereas the two-streams account holds that context determines whether or not the patient's consciousness is unified or divided.
(15) On this point see Ettlinger et al. (1974); Ferriss & Dorsen (1975); Gott & Saul (1978); Lassonde et al. (1988); Lassonde et al. (1991); Reynolds & Jeeves (1977); Saul & Sperry (1968); Chiarello (1980); although see Forget et al. ( 2009), Jeeves (1979), and Lassonde et al. (1995) for evidence of cognitive impairment in acallosals.