NOTE: Please see “Unmasking Multiple Drafts.” (2006) Philosophical Psychology, Vol. 19, No. 4, pp. 477-494 for citation purposes. This version is NOT the final version. There are some slight differences.
Unmasking Multiple Drafts
Any serious theoretical model of consciousness should be able to account for methodologically sound findings in the neurosciences and psychology. The details of the empirical data need to be carefully assessed – the theoretician ignores those details at the peril of their account of consciousness. This paper presents a case in point. An important part of Dennett & Kinsbourne’s (1992) assault on the Cartesian Theatre [1] model of the mind is their argument that their own Multiple Drafts model of consciousness is scientifically the best of three candidates they consider to explain the well-studied phenomenon of metacontrast backward masking. Their strategy for persuading us is to offer an Orwellian model as an indiscernible rival to the Stalinesque model researchers adopt.[2] The supposed lack of a useful distinction, on their account, is what they rely on to set Multiple Drafts in relief and motivate us to infer that Multiple Drafts is thus the preferable explanation for the phenomenon of metacontrast. Their strategy fails.
In this paper I argue that the Stalinesque interpretation is the best explanation for metacontrast. Section 1 is devoted to presenting a slightly more robust description of metacontrast.[3] The Dennett & Kinsbourne description is inadequate, leaves many important points obscured, and I have yet to see a more detailed description in the philosophical literature than what they offer. Next, in Section 2, I contrast the Stalinesque model with both the Multiple Drafts and Orwellian models. Most of the data is anomalous under the Multiple Drafts reading Dennett & Kinsbourne offer us, and, upon reflection, the Orwellian story appears to add unnecessary complexities. Finally, in Section 3, I consider a more general ramification of the discussion, namely that neither an Orwellian nor a Stalinesque account of any given neural architecture is, by itself, sufficient to commit us to a Cartesian Theater.
1.1 Dennett & Kinsbourne’s Description
Dennett & Kinsbourne describe the phenomenon of metacontrast as follows:
If a stimulus is flashed briefly on a screen and then followed, after a brief interstimulus interval, by a second “masking” stimulus, subjects report seeing only the second stimulus. (And if you put yourself in the subject’s place you will see for yourself; you will be prepared to swear that there was only one flash.) The standard description of such phenomena is that the second stimulus somehow prevents conscious experience of the first stimulus (in other words, it somehow waylays the first stimulus on its way to consciousness). But people can nevertheless do much better than chance if required to guess whether there were two stimuli. This only shows once again that stimuli can have their effects on us without our being conscious of them. This standard line is, in effect, the Stalinesque model of metacontrast: The first stimulus never gets to play on the stage of consciousness; it has whatever effects it has entirely unconsciously. But we have just uncovered a second, Orwellian model of metacontrast: Subjects are indeed conscious of the first stimulus (which would “explain” their capacity to guess correctly) but their memory of this conscious experience is almost entirely obliterated by the second stimulus (which is why they deny having seen it, in spite of their tell-tale better-than-chance guesses). (1992, p. 193)
This description falls short of being a sufficiently complete presentation of the phenomenon. First, though they are correct in holding that the standard account is Stalinesque, they don’t explain why researchers think about the phenomenon this way. One deficit in their presentation is that they neglect to mention that the Stalinesque model is based on general assumptions about the fundamental features of neural architecture and, further, those assumptions rest on even more general evolutionary assumptions.
Researchers assume that both the neural architecture of the visual system and consciousness arose due to evolutionary pressures (Bachmann (2000) and Breitmeyer (1984)). Given the background paradigm of evolution, Bachmann (2000) builds a theory of the neural basis of consciousness according to which early development in the evolution of the visual system did not involve conscious perception, whereas later developments are correlated with conscious experience. Moreover, he argues that this evolutionary history (from unconscious processing to conscious perception) is retraced during any given process from stimulus onset to conscious awareness. Breitmeyer (1984) reasons that visual awareness is a later evolutionary development and thus we should be able to trace the neurology of the human visual system (considered analogous to visual systems found in cats and monkeys) using psycho-physical experiments as guides to pinpoint the timing and location of episodes of either conscious or unconscious processing.
The problem that Dennett & Kinsbourne (1992) pose for these assumptions rests on a denial that we can actually perform these traces of the timing of conscious and unconscious processing at such a micro-level.[4] But such absolute skepticism concerning neuroscience seems misplaced; surely, the issue is an empirical one (Block, 1992; Flannagan, 1992, pp. 79-85; Johnsen, 1997, pp. 69-73; and Van Gulick, 1992).
So it is not the case that researchers just assume that the explanation of metacontrast is Stalinesque. The very notion of the development of a representation over time in the neurological structure of the brain from unconscious processing (beginning with stimulus onset) to conscious perception and the desire to discover the actual neural correlates of these processes are what drive the research. Given their background assumptions and the evolutionary framework within which those assumptions were generated, it is unremarkable that researchers think the data support their hypotheses. It is not merely that “this only shows once again that stimuli can have their effects on us without our being conscious of them,” but rather that the data primarily indicates that we may be correct about our assumption that some unconscious processing can be experimentally isolated based on the timing of the representation. As it turns out, the data seems to reveal that unconscious processing can affect our behavior, and this becomes a further consideration in research design. Given the fuller background explanatory goals of the researchers, the issues and the claims are more complex than Dennett & Kinsbourne’s discussion allows.
A second point that is obscured concerns the claim that “people can nevertheless do much better than chance if required to guess whether there were two stimuli.” While this is true, again the issue is more complex. The fact that people can perform better than chance at guessing the presence of some “masked” stimulus is not believed to be part of the metacontrast “masking” phenomenon (concerning target visibility), but rather another phenomenon (target localization) a metacontrast model with sufficient scope should explain – motion, location and target detection are thought to be immune to metacontrast “masking” (Breitmeyer, 1984; Breitmeyer & Ogmen, 2000; and Ogmen et. al., 2003). The version of metacontrast that Dennett & Kinsbourne discuss seems to be an experiment designed to have the subject report the visibility of the surface properties of the disc (e.g., color and brightness), which are susceptible to metacontrast “masking.” Thus, claiming subjects saw only “one flash” is part of the metacontrast “masking” phenomenon, however, a better than chance identification that there were two stimuli presented is not, but instead another phenomenon needing an explanation; in fact, some metacontrast models cannot explain the latter phenomenon (Breitmeyer, 1984, p. 228). It seems, then, that Dennett & Kinsbourne are conflating these two different, though related, phenomena, without argument.[5]
A third complaint concerns Dennett & Kinsbourne’s basic description of metacontrast: “If a stimulus is flashed briefly on a screen and then followed, after a brief interstimulus interval, by a second “masking” stimulus, subjects report seeing only the second stimulus.” This is not a full description of metacontrast, but only a description of a very isolated, though quite startling, effect of metacontrast. I develop this criticism in the next subsection.
1.2 Basic Disc/Ring Metacontrast
Different models of metacontrast have different explanatory and predictive power (Breitmeyer, 1984; and Breitmeyer & Ogmen, 2000), which makes for interesting problems in choosing among them, but those problems are beyond the scope of this paper. For this discussion I shall present a necessarily abbreviated account of a model utilizing the theory of transient/sustained channel inhibition (Breitmeyer, 1984; Breitmeyer & Ogmen, 2000; and Ogmen et. al., 2003) because of its explanatory scope, predictive success and relevance to the Dennett & Kinsbourne description. First, we need to define a few technical terms.
The phenomenon of metacontrast is the reduced visibility of some stimulus (in this case a disc – T) in the presence of another stimulus presented later (in this case a ring – M1) at various SOA’s (Bachmann, 2000; Breitmeyer, 1984; Breitmeyer & Ogmen, 2000; and Ogmen et. al., 2003).[7] It is the reduced visibility of the disc as a function of various SOAs of the following ring, producing a U-shaped curve, that is of most theoretical interest to researchers, not merely the striking invisibility of the disc at optimal SOA (optimal SOA is when the disc is reported as not visible and would be the bottom of the U-shaped function – this is the only SOA that Dennett & Kinsbourne describe). The gradual reduction of visibility of the disc as the ring is presented at increasingly later SOA’s is counterintuitive. This gradual reduction of visibility and then gradual recovery of visibility after optimal SOA is thought to reveal something about visual processing and, by extension, the functional architecture of the brain.
The depth of suppression of the visibility of the first stimulus is also affected by the spatial separation between the outer contour of the first stimulus and the inner contour of the following stimulus. So, for example, when the spatial separation between the outer contour of T and the inner contour of M1 is optimal (M1 fits tightly around T), at the optimal SOA the subject will report that T is not visible. As the space separating T and M1 is increased the subject will report a corresponding increase in visibility of the disc. We will return to this important point in Section 2.2 below.
The transient/sustained channel model was based on the discovery of different latencies of retinal ganglion in cats. Some ganglion would respond more quickly than others. The thought was that these different latencies would correspond to two neural channels that would carry information through the brain continuing at different latencies, though not necessarily the exact same latencies as observed in the retina. If this were so, then there might be some inter-channel inhibition that would help explain and predict the metacontrast phenomenon. Specifically, the visual information in the transient channel would arrive at some cortical area (say, V1) before the sustained channel information.[8] The model predicts that surface property information about the disc in the slower sustained channel could be inhibited by motion and position information of the later presented ring if that information about the ring arrived first at some cortical area (say, maybe, V1) via the faster transient channel. Depending on the timing difference of the two stimuli at the retina, and the respective latencies of the transient and sustained channels, the transient information of the later presented ring would interfere with the sustained information of the earlier presented disc to differing degrees. If the neurons in some cortical area (yet to be determined) where these channels are thought to interact are firing at their peak so that the faster transient channel information about the ring coincides with the slower sustained channel information about the disc, then the sustained channel is thought to suffer the greatest inhibition.
These theoretical channels have their neurological correlates in the magnocellular (M) pathways (for the transient information) and the parvocellular (P) pathways (for the sustained information) (Bar, 2003; Gazzaniga et. al., 1998, pp. 126-150; Lamme & Roelfsema, 2000). The M-pathways are primarily associated with coarse-grained information, especially motion and position and continue on past V1 via a dorsal (parietal) pathway toward the pre-frontal cortex, whereas the P-pathways are primarily associated with fine-grained information such as surface properties and shape and continue past V1 via a different ventral pathway toward V4 and beyond (Bar, 2003; Gazzaniga et. al., 1998, pp. 49-52, 126-150; Lamme & Roelfsema, 2000). Thus, the explanation for the ability to tell that something was there (fast M-pathway – dorsal), yet not be visually aware of its surface properties (slow P-pathway – ventral) is explained by these different pathways (Crick & Koch, 2003). The transient information about the disc (motion and position) outruns the transient information of the mask, and continues uninhibited up through the dorsal pathway. Whereas the sustained information about the disc (in this discussion, the surface properties) is inhibited by the transient information about the mask (by degrees, depending on the timing of the stimulus onset) but the sustained information about the ring (surface properties) does continue on uninhibited via the ventral pathways to the temporal cortex.
1.3 More Complex T-M1-M2 Metacontrast
This second experimental design is built on the same transient/sustained channel assumptions above. A few of the questions this experiment addresses are 1) whether the sustained channel activity of another ring (M2) can interfere with the transient channel activity of the first ring (M1), 2) whether M2 can mask M1 as well, and further, 3) how these combinations affect the visibility of the disc (T)? All three stimuli are briefly presented (for, say, 10ms.) at constant luminance. The T-M1 SOA is held at an optimal SOA (of say, 60ms); i.e., when the subject will not report T as visible. The onset of M2 is then varied from 240ms prior to M1 up to 120ms after M1 (the M2-M1 SOAs vary between -240ms and +120ms). The results are as follows: When M2 is presented prior to M1, T’s reported visibility increases (a phenomenon known as ‘target recovery’) until M2 and T are synchronous, i.e., when M2 precedes M1 by 60ms. Given the assumption of transient/sustained channel inhibition, this is predicted. It is assumed that the slow sustained activity caused by M2 suppresses the fast transient activity caused by M1 as M2 is presented closer in time to the presentation of M1. Thus T is increasingly visible because the transient channel activity caused by M1 is increasingly inhibited and this has a decreasingly inhibitory effect on the sustained channel activity caused by T.
When M2 is presented after the presentation of T, the visibility of T decreases as M2 is presented temporally closer to the presentation of M1. When M2 is presented after M1, M2 acts as a metacontrast mask of M1 thus reducing M1’s visibility. However, the reduced visibility of M1 has no effect on the reduced visibility of T. The explanation is as follows: At the later SOA’s, the transient channel activity caused by M2 begins to inhibit the sustained channel activity caused by M1, producing the classic U-shaped metacontrast function of M1. However, the sustained channel activity caused by M2 is no longer suppressing the transient channel activity caused by M1. Thus, the transient channel activity caused by M1 is again suppressing the sustained channel activity caused by T, and as M2 is presented at even later SOA’s after optimal SOA relative to M1, M1 regains visibility as a function of the SOA of M2. These effects are predicted by the model and verified under experimental conditions.
This is a very simplified and brief presentation of relevant features of metacontrast, from the view of a transient/sustained channel model. The actual neural activity is far messier. There are feed forward and feed back loops within the dorsal and ventral pathways that presumably also have effects on the system (Bar, 2003; Di Lollo et. al., 2000; Hochstein & Ahissar, 2002; Lamme & Roelfsema, 2000). Further, even though researchers are trying to pinpoint the location and timing of the level of processing necessary for consciousness, there is no definite time or location identified yet, nor a definite location or time for the alleged interactions that are assumed responsible for the phenomenon of metacontrast. Even though we can check on individual neurons, consciousness is experienced in a neural network, not a petri-dish, and this is an important point of Dennett & Kinsbourne’s overall paper that is out of the scope of our present discussion.
2.1 Multiple Drafts Explanation of metacontrast
Here is Dennett & Kinsbourne’s explanation for the phenomenon of metacontrast within their Multiple Drafts model:
When a lot happens in a short period of time, the brain may make simplifying assumptions…In metacontrast, the first stimulus may be a disc and the second stimulus a ring that fits closely outside the space where the disc was displayed. The outer contour of a disc rapidly turns into the inner contour of a ring. The brain, initially informed just that something happened (something with a circular contour in a particular place), swiftly receives confirmation that there was indeed a ring, with an inner and outer contour. Without further supporting evidence that there was a disc, the brain arrives at the conservative conclusion that there was only a ring. Should we insist that the disc was experienced because if the ring hadn’t intervened the disc would have been reported? Our model of how the phenomenon is caused shows that there is no motivated way of settling such border disputes: Information about the disc was briefly in a functional position to contribute to a later report, but this state lapsed; there is no reason to insist that this state was inside the charmed circle of consciousness until it got overwritten, or contrarily, to insist that it never quite achieved this state. Nothing discernible to “inside” or “outside” observers could distinguish these possibilities (1992,p. 195).
Flanagan (1992, p. 84) suggests that this description of “lapsing” doesn’t seem to be much different than the pre- and post-conscious “bumpings” ascribed to the Stalinesque or the Orwellian accounts. This is a point worth considering. How is “lapsing” all that different from getting “bumped,” (either pre- or post-consciously)? Does the word choice of “lapsed” instead of “overwriting” or “waylaying” really make much of a difference? Doesn’t this description just amount to the same thing, except that it helps to avoid sounding either Stalinesque or Orwellian? But a mere linguistic device does not change the ontology. As I mentioned in Section 1.1, Dennett & Kinsbourne do not think we can discover the actual neurological correlates associated with the timing of the metacontrast phenomenon. But a failure to discover those correlates does not mean they in fact don’t exist, and the fact that we haven’t discovered them yet does not entail that we won’t. I will return to this tension in their account in Section 3 below.
Consider the first five sentences of the quote above. Dennett & Kinsbourne attempt to offer an explanation of how it is that subjects seem to be aware of more than one stimulus (our better than chance guesses) while the same subjects steadfastly deny seeing a disc. Given their commitment to avoiding any discussion of the timing of consciousness in any possible specific neurological architecture involved in metacontrast, since they believe this would ultimately commit them to a Cartesian Theater, they speak in more generalized terms. But, under scrutiny, this backfires. They begin their explanation with, “[w]hen a lot happens in a short time, the brain may make simplifying assumptions.” As we now know, based on our discussion in Section 1 above, the Dennett & Kinsbourne description of metacontrast only refers to optimal SOA of M1 to T (say, 60ms). How do they account for all of the other SOAs, from simultaneous presentation of the T-M1 sequence up to optimal SOA?
For example: if T is a black disc presented for 10ms at a T-M1 SOA of 20ms, the subject will report a gray disc, i.e., T is visible, though slightly faded. An SOA of 20ms is a lot shorter than 60ms. What sort of “simplifying assumptions” does the brain make here? Given Dennett & Kinsbourne’s description above, we should expect that “[w]ithout further supporting evidence that there was a disc, the brain [would arrive] at the conservative conclusion that there was only a ring.” Given this “conservative conclusion” that there was only a ring, and given the lack of evidence for a disc while the brain is receiving confirming evidence for a ring, how can Dennett and Kinsbourne explain the report of a slightly faded disc in this even briefer inter-stimulus interval than the one they discuss? The brain, according to their explanation, merely noted that “something happened (something with a circular contour in a particular place)” and then the brain very quickly received confirmation of this contour event with the ring. Since they are relying on an unspecified “brief interstimulus interval” for their discussion, and that interval, on their account, only results in the total lack of visibility of the disc (which we now know occurs at optimal SOA), I see no way to resolve the problem when considering the much briefer inter-stimulus intervals that researchers think are equally interesting.
Further, consider the brief example of ‘target recovery’ discussed above in Section 1.3: now we have a situation where there is an even greater amount that happens in a short period of time. Consider the following T-M1-M2 combinations:
1. If M2 is presented 60ms before T, and thus 120ms before M1, the subject will report a near-full recovery of the visibility of T in the sequence of M2-T-M1 where the SOA of 60ms between T and M1 is optimal such that the subject would not see T if M2 were not presented. Given Dennett & Kinsbourne’s discussion above, there is no way for them to explain this – on their account the subject should not report T as visible since it is merely the briefness of the interstimulus interval between T and M1 that causes the metacontrast suppression of T.
2. If the sequence presents M2 30ms after T (i.e., between T and M1), then the subject will report that all three stimuli (T-M2-M1) are visible (although T will be reported as less bright than before). Again, following Dennett and Kinsbourne’s explanation of how Multiple Drafts handles the phenomenon of metacontrast, we should not expect the subject to report T as visible at all – with the addition of M2 a whole lot more is happening in the brief interstimulus interval associated with T’s lack of visibility, yet the subject reports seeing T.
3. Finally, when M2 is presented 60ms after M1, neither T nor M1 are reported as visible but only M2 is reported as visible to the subject. Although Dennett and Kinsbourne’s account may explain why it is that M1 is not perceived relative to M2, how does T’s outer contour “turn into” M1’s inner contour when M1 is not reported as visible? T’s outer contour cannot “turn into” M1’s inner contour because M1’s inner contour is not visible to the subject, so transitivity of contours (T to M1 to M2) is out. Moreover, it just won’t do to insist that T, as well as M1, is also interpreted by the brain as M2, because the SOA between T and M2 here is 120ms and at that SOA the subject does report T as visible in the absence of any presentation of M1. Given Dennett and Kinsbourne’s description we cannot account for T’s continued invisibility to the subject – the phenomenon is just mysterious.
Based on Dennett and Kinsbourne’s explanation of how Multiple Drafts handles metacontrast, the Multiple Drafts model now seems unable to account for any of these findings in what is considered a standard masking experiment. Multiple Drafts may fit the data of Dennett and Kinsbourne’s limited description of metacontrast, but it has absolutely no fit with the data given a more complete (though simplified) description of metacontrast. The timing of the stimulus onset is what is thought to be the key to eliciting the metacontrast phenomenon. In their attempt to deny this, and thus to avoid any comment concerning the neurology involved in the phenomenon, they have eliminated their model from contention for being the best model to describe metacontrast. Moreover, this failure casts doubt on the adequacy of Multiple Drafts as a model of consciousness.
2.2 Orwell or Stalin?
Still, even though Multiple Drafts is not the model of choice for explaining metacontrast, Dennett and Kinsbourne have also introduced the so-called Orwellian model. Block (1992), Flanagan (1992, pp. 82-83), and Van Gulick (1992) each offer some general observations concerning how we could rely on neurological evidence to decide between the Orwellian and Stalinesque models. These versions have the common feature of isolating the relevant neural systems necessary for consciousness, determining if there was neural activity in those systems corresponding to the presentation of the first stimulus and then inferring whether the subject was conscious of that stimulus. So, if the relevant systems associated with consciousness show activity in the presence of the first stimulus, then we could claim the subject was conscious of the first stimulus. Though we are much further along in our understanding of the neurology of the brain than we were in 1992, we are still far short of understanding those systems in precise detail.[9] However, we can choose between the two models without understanding the neurology exactly by appealing to the notion of inference to the best explanation.
Lipton (1991) explains that we tend to choose the ‘loveliest’ theory, that is, the theory that would give us the most understanding, if it were true. “It is not that the truth of the theory is the best explanation of its explanatory or predictive success; it is simply that the theory provides the best explanations of the phenomena that the evidence describes” (Lipton, 1991, p. 184). Explanatory scope and predictive power are very strong indicators for theory choice (Breitmeyer (1984), Kuhn (1977), Lipton (1991)). Breitmeyer (1984), and Breitmeyer & Ogmen (2000) discuss explanatory scope and predictive power as important for choosing between any metacontrast models. For example, if a model both describes the metacontrast U-shaped function of SOA and explains why motion, position and target detection are immune to that function, especially if it can also predict said immunity (see Ogmen et. al. (2003)), then it should be preferred over a model that can only describe the metacontrast U-shaped function of SOA (even if that description is on equal footing with the former).
As a first pass at deciding between Orwell and Stalin, we now discuss another experiment. Breitmeyer, et. al. (2004) recently employed a series of metacontrast experiments concerned with reaction time to color priming. We should recall, from Section 1 above, that, along with motion and position, target detection is thought to be unaffected by metacontrast masking. These experiments are designed to reveal some behavior resulting from what is thought to be unconscious processing – motion, position and target detection are associated with early evolutionary developments in the visual system, assumed to be unconscious. (Bachmann (2000), Crick & Koch (2003, p. 120), and Gazzaniga et. al. (1998, p. 534) argue that it is in our best interest as creatures with further evolved visual systems for early evolutionary developments to remain unconscious.) Briefly, the setup is as follows: one of three colored discs (blue, green and white) is randomly presented followed by one of two colored rings (blue or green), also randomly presented, after a brief SOA in the standard metacontrast masking setup. The task is to identify the color of the ring as quickly and accurately as possible.
When the T-M1 sequence was presented at optimal SOA and optimal spatial proximity, the green disc acted as a positive prime for the subject when the subject was presented with a green ring by reducing the reaction time to the green ring, and acted as a negative prime by increasing the reaction time to the blue ring. And vice versa with the blue disc. Oddly enough, however, the white disc acted like the green disc by decreasing reaction times for the green ring and increasing the reaction times for the blue ring. Analysis of the white disc stimulus revealed that the green phosphor contributed 68% to produce the white disc, whereas the blue phosphor contributed only 12% (in the RGB monitor used for the experiments). Thus, the wavelength composition of the white stimulus was much closer to the wavelength composition of the green stimulus than that of the blue stimulus. The priming effects were associated with the wavelength composition of the stimuli.
Further, in follow-up experiments conducted by Breitmeyer, Ogmen, H. and Todd, S.J. (in press), when the T-M1 sequence was again presented at optimal SOAs, but at spatial separations such that the subjects reported seeing the priming discs (recall the discussion in section 1.2 above), although the green and blue discs primed in similar ways as they did when not visible to the subject, the white disc now tended to be equally confusable with both the blue and green discs – it acted as a neutral prime for both the green and the blue rings. Wavelength is associated with stimulus-dependent neural activity (of, say, V1), whereas color perception is associated with percept-dependent neural activity (of, say, V4) (Zeki, 1997, pp. 794-804). The Stalinesque account would say that the “masked” disc did not undergo the subsequent percept-dependent processing necessary to be consciously perceived. The Orwellian would have to say that the disc was consciously perceived, but that it was almost totally overwritten – though very selectively (only the color and other perceived surface properties, not the physical wavelength information). This involves at least one extra, very complicated, step in explanation. In a word: Stalin gets the job done by distinguishing levels of processing, whereas Orwell must tell a story that the disc was perceived at a single level of processing, but memory of that event was selectively overwritten.
Also, consider Damasio et. al. (1980): bilateral lesions on V4 result in total cerebral achromatopsia. Such subjects cannot even imagine color, though they remember having experienced color (Sacks, 1995, pp. 3-41). These cases present neurological evidence that some specific areas of the brain are necessary for color processing, though other surface properties such as brightness are still perceived. (See also Gazzaniga et. al., 1998, pp. 144-146) Combined with the metacontrast reaction time experiment we can see an evolution of microprocessing – from wavelength-dependant processing, to percept-dependant processing of perceived surface properties like color. If I’m right about this, then we should expect to find the inhibition of certain surface properties somewhere between V1 and V4. And if that can be accomplished, we may be able to figure more precise timings for different qualities of representations, and if those timings are faster than 100ms, in keeping with the body of literature on neurology, we have good reason to believe these early processes are unconscious. Again, the Stalinesque account would explain this by distinguishing levels of processing. With the Orwellian account we must declare that achromatopsics are conscious of color but forget the experience!
Still, this is not conclusive. Without the neurological evidence that could actually decide the case, we must also consider less certain criteria for selecting between the two. An additional consideration could be that the Orwellian “model” served as a foil to make us doubt that the Stalinesque model really had any useful distinctiveness and was staged to motivate us to prefer the Multiple Drafts model. Consider the description that Dennett & Kinsbourne provide us for the Orwellian model, given the distributed parallel processing in the brain: “Each of these distributed microtakings is an episode of conscious judgment (multiple minicinemas). But then why don’t we all have either a kaleidoscopic and jumbled ‘stream of consciousness’ or (if these distributed microtakings are not ‘co-conscious’) a case of ‘multiple selves’?…[H]ow can the manifest coherence, seriality, or unity of conscious experience be explained?” (1992, pp. 234-235) Whereas when they describe the Stalinesque model claiming that each “distributed microtaking is an episode of unconscious judgment, and consciousness of the taken element must be deferred to some later process,” their complaint is: “But how long must each scene wait, pending potential revision, before the curtain rises on it?” (Dennett & Kinsbourne, 1992, p. 234) It seems the latter of the two models has the fewer problems, on Dennett and Kinsbourne’s own account. On the Orwellian version, as they define it, we now have to consider how all of these microtakings can be conscious and yet we still have a kind of unity of conscious experience, whereas the Stalinesque version does not pose this particular problem.
Consider again the Stalinesque version: neurologically, we may be able to find at least necessary conditions among the multitude of neural processes for conscious awareness of any given property of some stimulus – such as the color priming experiment in conjunction with cerebral achromatopsia. In other words, with Stalin we are just dealing with the basic problem (as difficult as that is) of trying to find the neural correlates of consciousness. With Orwell, every neural event is a conscious experience and we now have to explain how this “kaleidoscope” of microtakings can square with our basic conscious experience, and how the brain knows which of these non-trivially many events to overwrite in order to preserve our experience of a unified stream of consciousness.
Moreover, when discussing the use of the clock in Libet’s (1985) discussion of “readiness potential” experiments, Dennett & Kinsbourne claim that “there is essentially continuous representation of the spot…in various different parts of the brain, starting at the retina and moving up through the visual system” (1992, p.198). So, given their description of Orwell, and their notion of a microtaking, the stimulus onset at the retina is a conscious microtaking.[10] That is just too hard to swallow. With Stalin, our basic conscious experience is not up for question, only when and where we have it, and we don’t have to think it happens before the standardly accepted 100-300ms from stimulus onset to consciousness (especially, at the retina). With Orwell, sorting through the now baroque morass to be able to design an experiment is just too unwieldy. The principle of parsimony would here favor Stalin.
But there is a further concern. As discussed in Section 1.1 above, the Stalinesque account is more completely understood as a background assumption of how the neurology of the brain developed embedded in the paradigm of evolution. If we really want to buy the Orwellian account, we would have to give an interesting evolutionary story explaining how a bat (or, by extension, even a fly) could be a flying zombie and still have every microtaking be a conscious microtaking! The Orwellian is in a dilemma. Either humans have evolved a visual system unique from all other species that allows us to retain certain information in consciousness while the rest are selectively erased (and those mechanisms for selective erasure now need to be searched for neurologically) or the bat and the fly’s neural microtakings are, by stipulation, conscious. With Stalin we only have to point out that some system does not have the necessary neurological correlates for consciousness (given we find them). Again we face the problem of parsimony. We have no further motivation to fix up the Orwellian model. The one who presents a theory accepts the burden of placing it comfortably within the broader scientific context. Based on these considerations, inference to the best explanation would surely favor a Stalinesque account. It is supported by our existing evolutionary stories, it provides a better understanding of a complex process without further complicating those processes, and it requires the least adjustment (none) to our existing scientific paradigms.
Section 3: Neurological Multiple Drafts
Now we get to the crux of the problem with Dennett & Kinsbourne’s account and explanation of metacontrast. They convincingly argue that there is a difference between the timing of the representation and the representation of time. (However, as noted throughout our discussion of metacontrast, the timing of the representation is crucial to obtaining the very phenomenon of metacontrast – it is a function of SOA.) Most of us can agree that the Cartesian Theater is not a very good model of neural architecture (though some have at times embraced some form of the Cartesian Theater, such as Vogeley et.al., (1999)). Yet Dennett and Kinsbourne seem unsure as to how we do not all agree that it follows from those two points that there is no way to decide the case of the when and where of consciousness at the neurological level for any given neurological process – that the Stalinesque/Orwellian distinction totally breaks down. Two important underlying assumptions for their conclusion are 1) that psychophysical experiments and greater neurological understanding will not reveal any necessary “moment” or place for conscious experience and 2) any commitment to either Orwell or Stalin is sufficient to commit one to a Cartesian Theater. Our discussion of metacontrast has cast doubt on the truth of point 1. Let us now consider point 2.
As Flanagan (1992, p. 81) points out, it is uncharitable to saddle anyone taking either an Orwellian or Stalinesque position with the mythical Cartesian Theater. Bachmann (2000), for example, denies that development of a visual representation necessary for conscious awareness is a commitment to a Cartesian Theater. Given Dennett & Kinsbourne’s discussion, I think they would agree. The difference enters when Bachmann makes the claim that these developments are potentially traceable. Dennett & Kinsbourne think that any line dividing conscious and unconscious processing would be arbitrary. However, as Johnsen (1997, p.73) notes, although such a line may be arbitrary, that doesn’t make it meaningless – as neurological sciences improve, the line may not be as arbitrary. (In fact, it seems that Dennett & Kinsbourne think such a line would be capricious.) But, given our discussion in section 2.2 about the recent color priming experiments and achromatopsia, it does not seem that a potential line drawn somewhere between V1 and V4 for color perception is meaningless or entirely arbitrary. There seem to be useful theoretical and empirical reasons for drawing such a line. However, I don’t think this commits us to a Cartesian Theater – even Dennett & Kinsbourne admit that the different properties are processed at different latencies. They claim their model opposes the Cartesian Theater, and their description of the various phenomena they discuss is self-consciously vague in order to avoid a commitment to either Orwell or Stalin. I think this fear is misplaced, since I do not think a commitment to either Orwell or Stalin necessitates a commitment to a Cartesian Theater. Let us now consider their discussion of the British Empire (Dennett & Kinsbourne, 1992, pp. 188-189, pp. 235-236.), which they use to help us understand that the representation of time is different from the timing of the representation.
The Battle of New Orleans takes place after the treaty ending the War of 1812 is signed. India finds out about the battle before it finds out about the treaty, yet no one is confused about the information – the postmark tells the story, is the representation of time. At some fuzzy time after both the treaty and the battle (say, a couple of weeks), the British Empire knows about both and about their timings. Yet there is no specific place and time one can discern that would precisely specify when and where the British Empire knew the combined story – or so the argument goes.
Consider: “The British Empire” is a governmental entity that has both a centralized location and geographic and temporal smear. If India does not get the news of either the treaty or the battle, it is still plausible to think that, in general, the British Empire, to some degree knows both. If neither India nor Australia receives the news it is still plausible to think that, in general, though to a slightly lesser degree than the previous example, the British Empire knows both. However, if Parliament and the Queen (or King) do not get news of either the treaty or the battle, then no matter how many colonies “know” the information, it seems just wrong to claim that the “British Empire” really knows both. Further, it is the function of the official governmental body based in England to put its stamp of approval on the treaty event. Until that happens, it is always an open question whether both events will be acknowledged as actually happening in the way the “rumors” have it.
Assume the information finally arrives. If the official rulers decide that the “treaty” was never properly signed (they believe a homeless worldly wanderer from the United States forged the signature of an official representative of the British Empire), then the treaty never happened and the Battle of New Orleans did not take place after the signing of the treaty. There is a necessary event that must take place for the British Empire to be said to know the information about the treaty – viz., the official government must declare that the event happened. This does nothing to undermine the argument that the time of representation is different than the time represented, nor does it undermine the argument concerning the spatial and temporal smear of the representation – there is no one place where it all comes together. It does, however, parallel the case of metacontrast and achromatopsia, since there is a place that certain information must pass through, and the timing of this event is traceable from the moment of sending the messages.
This is not a commitment to a Cartesian Theater, only to a level of processing necessary for the brain to resolve some particular property represented – color, and similarly for the British Empire to be aware of both the treaty and the battle as well as their timing. V4 is apparently necessary for color processing, while Parliament is apparently necessary for treaty processing. Notice, the Battle of New Orleans does not necessarily need the official stamp for that information to be known to the British Empire. Likewise, motion, position and detection do not need to pass through V4. Both accounts are Stalinesque, yet both accounts are compatible with spatial and temporal smear. The findings in neurology and metacontrast experiments support a Stalinesque model of neural processing, yet a commitment to a Stalinesque model does not entail a commitment to a Cartesian Theater.
Conclusion
I’ve shown that the Dennett and Kinsbourne description of metacontrast is inadequate and I have offered a more detailed account of metacontrast. I’ve also demonstrated that the Stalinesque model is preferable to both the Multiple Drafts model and the Orwellian model as the explanation for the phenomenon of metacontrast. Most of the data is anomalous under the Multiple Drafts reading Dennett and Kinsbourne offer us and, upon reflection, the Orwellian story appears to add unnecessary complexities. Finally I have argued that a Stalinesque model (and by extension an Orwellian model) of any given neural architecture is, by itself, not sufficient to commit us to a Cartesian Theater. The Stalinesque model of metacontrast allows us to predict, with scientific accuracy, which draft will be reported, and any adequate model of consciousness should account for this.
Notes:
[1] The Cartesian Theater model of the mind requires a discernible time and place in the brain where discriminations from various modalities must “come together” to be consciously experienced (Dennett & Kinsbourne (1992)).
[2] Both of these models are allegedly versions of the Cartesian Theater (I argue they are not). For this paper, on the Stalinesque account conscious experience of a visual stimulus is prevented by the later presentation of a second visual stimulus. On the Orwellian account subjects are conscious of the first stimulus, but the memory of that event is almost totally wiped out by the second stimulus (Dennett & Kinsbourne, 1992). On the Multiple Drafts model, the response to or the judgment of any input or processing by any neural configuration is called a micro-taking, and any “unified taking is broken up in cerebral space and real time…fragmented into many distributed moments of microtaking.” On the Stalinesque model, “each of these distributed microtakings is an episode of unconscious judgment, and the consciousness of the taken element must be deferred to some later process.” On the Orwellian model, “each of these distributed microtakings is an episode of conscious judgment (multiple minicinemas).” (Dennett & Kinsbourne, 1992, p. 234)
[3] I say slightly because I limit myself to simplified, though important, variations of the basic disc/ring combination that Dennett & Kinsbourne describe. This barely scratches the surface of the metacontrast literature.
[4] Dennett & Kinsbourne’s arguments for differentiating the time of representation from the representation of time are one thing. The assertion that we cannot, even in principle, find a meaningful place and time necessary for conscious perception in such micro-frameworks as neurology is quite another (Dennett & Kinsbourne, 1992, p. 240). We can agree on the former, without agreeing on the latter. And as Dennett & Kinsbourne note (1992, p. 235), some of us in fact do, though they seem mystified by that disagreement – they think their skepticism just follows from the said differentiation coupled with a rejection of the Cartesian Theatre. I think we can even agree that “the Stalinesque/Orwellian distinction must break down at some scale of temporal resolution” without agreeing that the scale in question is at every neurological level. E.g., we might think it makes sense to draw the line between conscious and unconscious processing through some general location in the brain rather than a particular neuron. The tension seems to be that they think 1) there are no meaningful theoretical motivations to choose Stalinesque over Orwellian models (I argue there are), and 2) either a Stalinisque or Orwellian description of any neurological process commits one to a Cartesian Theatre (I argue it doesn’t).
[5] Researchers may be wrong about this point. However, the Dennett & Kinsbourne discussion does not reveal that their claim may be controversial, and thus not representative of the standard view. Having said that, I set this particular point aside.
[6] SOA is different than “a brief interstimulus interval” separating the presentation of the stimuli and is another point of contention with Dennett & Kinsbourne’s description. Though this is an important technical point that also has bearing on the discussion, it is not necessary to develop the distinction for the purposes of this paper, so I set that discussion to one side.
[7] I will be using these sources as the basis of my discussion. Unless a particular point is relevant to only one of these sources, I will no longer continue citing them. Also, I will be using the terms ‘visible’, ‘visibility’, etc., as is often found in the psychophysical literature; that is, as referring to the subject’s report of the apparent brightness of a visual stimulus. I do this to be consistent with the psychological literature that I am relying on, and I am thus setting aside issues concerning whether ‘visible’ is the appropriate term when discussing a subject’s report of what he/she is aware of or sees, rather than what is actually present in the visual field. In the metacontrast experiments discussed here, a visual stimulus (e.g., T) is always presented at the same luminance, but the reported visibility of that stimulus varies with the SOA of a following stimulus (e.g., M1).
[8] There are reasons for thinking the interaction happens after the channels pass through the Lateral Geniculate Nucleus that have to do with binocular vision, but that discussion is outside the scope of this paper.
[9] See, for example, Dehaene et. al. (2004) and Kamitani & Tong (2005) for some recent fMRI studies. Although quite an advance from 1992, such studies still fall far short of providing the precise empirical evidence necessary for undermining the so-called Stalinesque / Orwellian Janus-faced problem. And this is just what one should expect given the Dennett and Kinsbourne characterization of the problem.
[10] Dennett &
Kinsbourne admit the possibility of this odd Orwellian commitment: “[C]onsider
the insanely radical Orwellian hypothesis that we are conscious of the
dizzying swim of imagery on our moving retinas in spite of what we subsequently
say – we just continually forget it.” (1992, p. 239).
References
Bachmann, T. (2000). Microgenetic Approach to the Conscious Mind.
John Benjamins Publishing Co., Amsterdam/Philadelphia.
Bar, M. (2003). A cortical mechanism for triggering top-down facilitation in visual
object recognition. Journal of Cognitive Neuroscience 15:4, pp. 600-609.
Block, N. (1992) Begging the question against phenomenal consciousness.
Behavioral and Brain Sciences, 15, 205-206.
Breitmeyer, B.G. (1984). Visual Masking: an integrative approach.
Clarendon Press: Oxford.
Breitmeyer, B.G., and Ogmen, H. (2000). Recent models and findings in visual
backward masking: a comparison, review, and update. Perception &
Psychophysics, 62,(8), 1572-1595.
Breitmeyer, B.G., Ro, T., and Singhal, N.S. (2004) Unconscious
color priming occurs at stimulus- not percept-dependent levels of
processing. Psychological Science. Vol. 15. No. 3. pp. 198-202
Breitmeyer, B.G., Ogmen, H. and Todd, S.J. (in press) Unconscious, Stimulus-Dependent
and Conscious, Percept-Dependent Priming with Chromatic Stimuli. Perception &
Psychophysics.
Crick, F. and Koch, C. (2003). A framework for consciousness. Nature neuroscience.
Volume 6 no 2. 119-126.
Damasio, A., Yamada, T., Damasio, H., Corbett, J., McKee, J. (1980) Central
achromatopsia: behavioral, anatomic, and physiologic aspects.
Neurology 30. 1064-1071.
Dehaene, S., Jobert, A., Naccache, L., Ciecie, P., Poline, J. B., Le Bihan, D., and
Cohen, L. (2004). Letter binding and invariant recognition of masked words: behavioral and neuroimaging evidence. Psychological Science. Vol. 15, No. 5. 307-313.
Dennett, D. (1991). Consciousness Explained. Little Brown and Company Limited.
when of consciousness in the brain. Behavioral and Brain Sciences, 15, 183-243.
Di Lollo, V., Enns, J., and Rensink, R. (2000). Competition for consciousness among
visual events: The psychophysics of reentrant visual processes. Journal of
Experimental Psychology: General. Vol. 129, No. 4, 481-507.
Flanagan, O. (1992). Consciousness Reconsidered. MIT Press: Cambridge, MA.
Gazzaniga, M.S., Ivry, R.B., and Mangun, G.R. (1998). Cognitive Neuroscience:
the biology of the mind. W.W.Norton & Company: New York.
Hochstein, S., and Ahissar, M. (2002). View from the top: Hierarchies and reverse
hierarchies in the visual system. Neuron Vol. 36. 791-804.
Johnsen, B. (1997). Dennett on qualia and consciousness: a critique.
Canadian Journal of Philosophy. Volume 27, Number 1. pp. 47-82.
Kamitani, Y., and Tong, F. (2005). Decoding the visual and subjective contents of the
human brain. Nature Neuroscience Vol. 8, No. 5. 679-685.
Kuhn, T. (1977). The essential tension: selected studies in scientific tradition and change.
Chicago: University of Chicago Press.
Lamme, V.A.F., and Roelsema, P.R. (2000). The distinct modes of vision offered by
feedforward and recurrent processing. Trends Neuroscience 23, 571-579.
Libet, B. (1985). Unconscious cerebral initiative and the role of conscious will in
voluntary action. The Behavioral and Brain Sciences 8, 529-566.
Lipton, P. (1991). Inference to the Best Explanation. Routledge, London.
Ogmen, H., Breitmeyer, B.G., and Melvin, R. (2003). The what and where of visual
masking. Vision Research. 43, 1337-1350.
Sacks, O. (1996). An Anthropologist on Mars. Vintage Books: New York.
Van Gulick, R. (1992) Time for more alternatives.
Behavioral and Brain Sciences, 15, 228-229.
human self model are implemented in the prefrontal cortex.
Consciousness and Cognition 8, 343-363.
Zeki, S. (1997). The color and motion systems as guides to conscious visual perception.
In K. S. Rockland, Kaas, J.H., and Peters, A. (eds.), Cerebral cortex (vol. 12);
Extrastriate cortex of primates (pp. 777-809). New York: Plenum.