Margolis (1998a,b) has offered the four-century persistence of Tycho's illusion among expert astronomers and historians of science as evidence that cognitive illusions cannot be treated solely as effects produced by clever experimenters working with naive subjects. It is therefore relevant that another cognitive illusion has persisted for the better part of a century among experts who compare visual neurophysiology with visual sensation. In this case, a powerful illusory correlation leads experts to conclude that neural onset transients enhance the perceived brightness of brief flashes. The illusion is partially dissipated by the provision of a visual display which contains no evidence of brightness enhancement even though it clearly contains neural onset transients. Despite exegesis of the basis of this cognitive illusion in prominent publication venues, and despite authoritative endorsement of the soundness of the exegesis, this powerful cognitive illusion continues to affect expert thinking. Moreover, because the dissipation is only partial in this particular case, some degree of cognitive isolation exists between the two ways of comprehending these data. Hence this psychobiological case not only provides another illustration of the power of certain habits of mind to dominate other such habits, but the partial character of the dissipation provides more generality for Margolis's position.
2. Margolis then reviews the properties of various astronomical systems, such as the well-known Ptolemaic and Copernican systems. In the course of the transition from one of these systems to the other, an intermediate system was proposed by Tycho Brahe some 400 years ago. The discussion of the implications of Tycho's system by Tycho himself and by everyone else was marked by a cognitive illusion which led the very best experts of his time, as well as modern historians of science, to believe that such a system was impossible in a world in which the heavenly bodies were carried by solid spheres.
3. Margolis provides a clever visual display in the form of a cutout whose parts can be moved relative to each other in a way that completely unmasks the illusion. Because the display produces an immediate and compelling change, Margolis concludes that the illusion depends on an entrenched habit of mind which is effectively challenged only when a person actually sees (via the cutout), rather than merely imagines, how the Tychonic system must move. And because this episode involves real people, some of them extraordinarily famous, along with their struggle to understand a scientific question which was important to them, Margolis correctly argues that this is not just a form of entertainment in which clever people trick innocents. Instead, it demonstrates how entrenched intuition can trump logic.
4. A similar failure occurred at the very dawn of cellular psychobiology when a cognitive illusion appeared which influenced the contributions of one of the great neurophysiologists of the time, Edgar Adrian, who shared the Nobel Prize in Physiology/Medicine in 1932. The Prize recognized the seminal contributions Adrian and his teacher, Keith Lucas, had made to the embryonic field of single cell electrophysiology. This work was summarized in Adrian (1928). Among their contributions were: (A) an application of electronic amplification via vacuum tubes to electrophysiology; (B) the discovery of action potentials (APs) propagating in single axons; (C) the promulgation of the all-or-none law to characterize these APs; and (D) the use of these discoveries to investigate the way single sensory axons encode stimuli.
5. The cognitive illusion came on the stage in complete form in a passage which concluded Adrian's 1928 book and was devoted to sensory coding. Two figures were used to illustrate the point. A link to the first is provided below as our Figure 1 (Adrian's Figure 30). It displays two sets of real data, which include (A) cellular recordings of the AP frequencies (in spikes per sec) of neural responses evoked in eel optic nerve fibers by light stimuli of varying intensity and (B) brightness matches (in relative units) made by human observers (studied by David Broca and Andre Sulzer around the turn of the century) to lights of comparably varying intensity. In both sets, data obtained at a given intensity are represented by a smooth curve and the luminance associated with each curve is indicated in meter candles. The time axis is described more sparely: Both sets of data are presented as functions of time on an axis which is demarcated in 0.1 sec intervals but otherwise unlabeled. (It is this inattention to detail which set the stage for Adrian's cognitive illusion, which depends on a conflation of two quite distinct temporal variables.)
Figure 1 (Adrian 1928, Figure 30).
6. Inspection will show that both data sets exhibit the same trends: At small values of time, low frequencies or dim brightness matches are found. Then both data sets peak in a middle range of time values and then decline at yet greater time values. Furthermore, increasing the intensity increases both AP frequencies and brightness matches. Both sets of peaks shift towards shorter time values as the intensity is increased. (In the behavioral literature, the peak overshoot is known as either brightness enhancement or the Broca-Sulzer effect.) One discrepancy between the two data sets is that the human brightness matches peak at time values that range from about 50 to 150 msec while the eel AP frequencies peak at time values that vary from about 300 msec to 500 msec. Eels being cold blooded, this retardation of their data relative to the human data was not particularly troubling, and the parallelism was taken by Adrian as evidence that "the general form" of physiology and behavior "is surprisingly alike." Thus, a plausible explanation had been advanced for an otherwise paradoxical phenomenon -- brief light flashes with less total energy may appear brighter than longer flashes with more total energy because of neural onset transients.
7. The complexities of the actual data were harmonized by producing a more schematic version, presented as our Figure 2 below (Adrian's Figure 31). This drawing is based on several inspired insights addressing certain basic issues this infant field needed to resolve immediately. The correctness of these inspirations varied; some turned out to be entirely right while at least one (i.e., the cognitive illusion) was entirely wrong.
Figure 2 (Adrian 1928, Figure 31).
8. The most correct inspiration was Adrian's realization that a continuous sensory stimulus had to generate a continuous and graded "excitatory process in receptor[s]." This sustained graded process then generated the discrete axonal APs. Confirmation of the essential correctness of this inspiration did not come for several decades, until the intracellular micropipette eventually provided direct recordings of what are now called receptor potentials (RPs). Adrian's prescience in this regard is sadly confirmed by the fact that RPs today are still not universally included in expositions of basic sensory physiology.
9. Also of lasting significance was Adrian's appreciation of the role of frequency coding in the representation of sensory quantity by APs. He wrote (p. 114):
"Provided that there is nothing to distract our attention the intensity of the sensation at any moment turns out to be proportional to the frequency of the impulses in the sensory nerve fibre."
10. In this, he was aided by the already established psychophysical understanding of the compressive nature of sensory quantity, evident in the long duration brightness matches shown in Figure 1. This compression has turned out to be far more complex than Adrian thought, with the result that this is still an active current area of research [see the reviews by Krueger (1989) and Laming (1997)]. It has turned out that there are quite a few interesting ways to "distract our attention." Furthermore, the frequency coding concept per se has turned out to need substantial refinement [see the reviews in Uttal (1973) and Wasserman (1992)]. Nevertheless, Adrian launched this field of research in the right direction. Although many complexities have subsequently developed, modern expositions which are comprehensive and complete generally make it clear that RP amplitude and AP frequency are intimately linked in roughly the way envisioned by Adrian.
11. Finally, the trace representing "Sensation" in Figure 2 expresses the cognitive illusion which resulted from Adrian's use of the same units, namely, seconds, of the same variable, namely, time, to characterize two completely different parameters: flash duration versus time after stimulus onset. This illusion is so compelling that authoritative endorsement of Adrian's explanation came from influential scholars; indeed, the proposition that the Broca-Sulzer effect is caused by neural onset transients became such a commonplace that exegesis was deemed unnecessary. For example, Le Grand (1968; p. 422) simply said that the "connection" of the two sets of data was "obvious."
12. The illusion is all the more remarkable because, on some cognitive level, Adrian was not unaware that there was a genuine difference between these two temporal parameters. For example, he said (p. 116-7) that he was comparing "the curves which show the rise and fall in the brightness of the sensation produced by a flash of light in man." with the "curves showing the rise and fall in the frequency of impulses in the eel's optic nerve when a light is thrown on the eye and allowed to remain there." The powerful effect of this illusion on his cognitions is aptly indicated by his immediately following statement (p. 117) that: "The curves produced by a flash of light are of the same form, but we have more complete data for the steady exposures."
13. The latter statement was only possible because Adrian had actually had the data in hand to prepare a display in which the neural responses to longer flash durations could be directly compared to those evoked by shorter flash durations. Had he done so -- that is, had he formally constructed an adequately detailed analysis of this question -- he would certainly have seen that longer flashes produce neural responses that are always greater than or equal to those evoked by shorter flashes. A fair correlate of brightness enhancement, of course, would require that some feature of the longer responses be reduced relative to a corresponding feature of the shorter flashes. To this day, to our knowledge, no such reversal has ever been reported.
14. The illusion is immediately dispelled when an appropriate visual display is created. As with Margolis's cutout, the viewer of this display can actually see the critical information, rather than just imagine it. Figure 3 contains the display and shows how Wasserman and Kong (1974) dispelled the illusion by presenting data derived from photoreceptor RPs. [A more conventional Cartesian plot of one feature of these same data was given later in Figure 8-11 of Wasserman (1978).] It will be appreciated (see Kong and Wasserman, 1978) that formally similar plots can be constructed using optic nerve AP frequencies (i.e., the same neural variable which had been displayed above in Figure 1).
Figure 3 (Wasserman & Kong 1974, reproduced with permission]
15. Each panel of Figure 3 contains superimposed RPs which were evoked by a given flash intensity. The various panels represent fourfold changes in intensity and the log of each panel's intensity is indicated within each panel with the lowest intensity set at zero. Each panel contains four traces which superimpose the RPs evoked by flashes of durations which also differ in fourfold steps, namely 10, 40, 160, and 640 msec. The traces are aligned so that they all start at the moment when the stimulus is turned on, with the useful result that the responses to longer durations can readily be distinguished in Figure 3 because they always evoke responses which obviously last longer. (A modest but tolerable degree of complexity is introduced by the fact that fluctuations caused by the quantal nature of the light stimulus are evident at lower intensities but severely attenuated optic nerve APs are evident at higher intensities.)
16. This display then clearly dissociates the two independent variables of this experiment, intensity and flash duration. Each waveform traces the magnitude of the physiological response evoked by a flash of light as a function of the time elapsed after the start of the light flash. Considered from that standpoint, all the plotted waveforms exhibit a rise and fall and some of them exhibit a sag and leveling off after an initial transient peak of the type observed by Adrian, with the peak-cum-sag occurring at higher intensities and longer durations.
17. The illusion is dispelled when these data are considered from the standpoint of flash duration. From that standpoint, the totality of each trace in Figure 3 simply represents a given intensity and duration. It is only the differences among traces within a given panel which represent the effect of flash duration. Considered from this detailed perspective, it is evident that longer flashes always produced waveforms which either equaled or exceeded those produced by shorter flashes. Other than some small quantal irregularities, none of the seven panels in Figure 3 display any exceptions to this generalization. Hence, no correlate of the Broca-Sulzer brightness enhancement effect can be found in any characteristic of the physiology, even though the neural transients in this display are as evident as they ever are. The presence of the Adrian illusion in the frameworks given in Figures 1 and 2 and the absence of this illusion in the framework of Figure 3 show that, as Harris (1988) points out, reference frames provide a fundamental aspect of cognitive coding.
18. Authoritative endorsement of this correction of Adrian's inspiration came when Uttal (1981), reproduced the illusion-dispelling displays which had been published by Wasserman and Kong in his major synthesis of this field. He gave a fair summary of the issue and concluded (p. 501) that "abundant reasons to question this association have been forthcoming in recent years."
19. The potency of this illusion, however, is so great that it continues to recur. Indeed, such recurrences are a regular phenomenon (e.g., Ward, 1991). A partial cause of recurrence may be the partial encapsulation (see below) of the Adrian illusion. Despite great familiarity with it, inspection of Figure 1 in isolation still tends to reinstate this cognitive illusion partially. Perhaps recurrences also take place because explaining this particular cognitive illusion requires an interdisciplinary analysis; this keeps it out of most secondary or tertiary bibliographic sources. Moreover, most primary sources do not offer a mechanism for providing a timely commentary (e.g., Wasserman, 1991) in the same venue as the original publication. It is most likely, however, that these recurrences are spontaneous, as they give no indication that they are being forwarded in an attempt to refute the view that the "obvious" association of brightness enhancement and neural transients is a cognitive illusion.
20. These recurrences hence particularly reinforce Margolis's view that it would be naive to hold that we are less vulnerable to such illusions than were our predecessors. The only protection is to devise procedures that mechanize our review of obvious connections. Such protection may sometimes come from fully formal mathematical procedures, but a very large part of the primate brain is devoted to analyzing visual information. Hence we wish very strongly to support Margolis' view that properly developed visual displays can dispel cognitive illusions that originate in partially developed sketches.
21. Finally, it is appropriate to comment on Margolis's view that cognitive illusions express a convergence of science, whether its subject be the stars or the brain, with perception. An appropriate comment, however, would note that perception involves a variety of phenomena besides blindsight. Indeed, both Pearson (1988) and Munafo (1988) have already offered other suggestions, still others will no doubt be forwarded. Such an abundance is possible because it has long been understood that perception involves many processes which are collectively very much like those involved in the formulation and testing of scientific hypotheses. Without attempting to summarize the extensive philosophical discussion on this subject (which began in antiquity), suffice it to say that even in the early days of modern psychology, Helmholtz (1866/1910/1925) provided a crisp statement of this view, saying (p. III/3): "The general rule determining the ideas of vision that are formed whenever an impression is made on the eye... is that such objects are always imagined as being present in the field of vision as would have to be there in order to produce the same impression on the nervous mechanism, the eyes being used under ordinary normal conditions." Thus, just as perceptions are ideas that occur in the perceiver's mind, so astronomical theories are ideas that occur in the astronomer's mind.
22. Both perceptions and scientific theories are therefore rational, in the sense that they are derived from a reasoning process, but they can both also be irrational, in the sense that they lead to a false conclusion. Helmholtz stated the former proposition thus (P. III/4):
"The psychic activities that lead us to infer that there in front of us at a certain place there is a certain object of a certain character, are generally not conscious activities, but unconscious ones. In their result they are equivalent to a conclusion."
He stated the latter proposition as follows (P. III/5):
"But, moreover, just because they are not free acts of conscious thought, these unconscious conclusions from analogy are irresistible, and the effect of them cannot be overcome by a better understanding of the real relations."
23. Hence, to take an example discussed by Margolis, the Mueller-Lyer illusion does not disappear simply because we can use a ruler to measure it. Rather, the mental constructions that underlie the perception of the Mueller-Lyer diagram are executed by neural structures that are, in Fodor's memorable phrase (Fodor, 1985), "encapsulated" with regard to other mental activities. Encapsulation here does not have its most common meaning of enclosure. Rather, it has become a technical term which succinctly characterizes the fact that the act of measuring the Mueller-Lyer arrows with a ruler produces rational information which cannot penetrate the neural processes that produce our immediate perceptions.
24. Likewise, as Munafo (1998) rightfully observes, science modifies its views in ways that are much like those used by the child to refine its perceptions: by benefiting from challenging experience. Illusions are thus a most fitting subject for discussion because the word "illusion" refers to the mocking and playful interactions of competing views of the world -- science as well as childhood being simply extended periods of play. Indeed, even though we hope we have not been ambiguous in what we have written here, we would not be surprised if more playful commentators were to take some pleasure in reversing the terms of our commentary, turning it into a textual counterpart of an ambiguous reversible figure.
We wish to express our appreciation for the helpful side discussions which we have had with Howard Margolis by e-mail while writing this commentary. Figure 3 is reproduced with permission from Wasserman and Kong (1974); Copyright 1974 American Association for the Advancement of Science.
Adrian, E.D. (1928) The Basis of Sensation. Christophers (London).
Fodor, J.A. (1985). Precis of The Modularity of Mind. Behavioral and Brain Sciences, 8, 1-42.
Harris, L. (1998) The Mars/Sun collision illusion: Motion is not visualizable in two different reference frames simultaneously. Commentary on Margolis. PSYCOLOQUY 9 (33) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/psyc.98.9.33.cognitive-illusion.2.harris
Helmholtz, H. von (1866/1910/1925) Concerning the perceptions in general. Section 26 of the translation of the third German edition of the "Treatise On Physiological Optics." J.P.C. Southall (Ed.). Dover Publications.
Kong, K.-L. (1978) Temporal summation in the receptor potential of the Limulus lateral eye: Comparison between retinula and eccentric cells. Sensory Processes, 2, 9-20.
Laming, D. (1997) The Measurement of Sensation. Oxford.
Krueger, L.E. (1989) Reconciling Fechner and Stevens: toward a unified psychophysical law. Behavioral and Brain Sciences, 12, 251-320.
Margolis, H. (1998a) Tycho's illusion and human cognition. Nature, 392, 857.
Margolis, H. (1998b) Tycho's illusion: How it lasted 400 years, and what that implies about human cognition. PSYCOLOQUY 9 (32) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/psyc.98.9.32.cognitive-illusion.1.margolis
Munafo, M. (1998) The Tycho Illusion: Two ages and two illusions. Commentary on Margolis. PSYCOLOQUY 9 (39) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/psyc.98.9.39.cognitive-illusion.5.munafo
Pearson, D. (1998) Imagery need not be blind to fail. Commentary on Margolis. PSYCOLOQUY 9 (34) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/psyc.98.9.34.cognitive-illusion.3.pearson
Uttal, W.R. (1973) The Psychobiology of Sensory Coding. Harper and Row.
Uttal, W.R. (1981) A Taxonomy of Visual Processes. Erlbaum.
Ward, L.M. (1991) Informational and neural adaptation curves are asynchronous. Perception and Psychophysics, 50, 117-128.
Wasserman, G.S. (1978) Color Vision: An Historical Introduction. Wiley.
Wasserman, G.S. (1991) Time and duration: A persistent illusion. Perception and Psychophysics, 50, 603-604.
Wasserman, G.S. (1992) Isomorphism, task dependence, and the multiple meaning theory of neural coding. Biological Signals, 1, 114-142. ftp://ftp.cis.ohio-state.edu/pub/neuroprose/wasserman.mult_mean.ps.Z
Wasserman, G.S. and Kong, K.-L. (1974) Illusory correlation of brightness enhancement and transients in the nervous system. Science, 184, 911-913.