David Pearson (1998) Imagery Need not be Blind to Fail. Psycoloquy: 9(34) Cognitive Illusion (3)

Volume: 9 (next, prev) Issue: 34 (next, prev) Article: 3 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 9(34): Imagery Need not be Blind to Fail

Commentary on Margolis on Cognitive-Illusion

David Pearson
Department of Psychology
University of Aberdeen
Aberdeen AB24 2UB, UK.



Margolis (1998) argues that the mental rotation essential for an illusory collision in a Tychonic system cannot be "seen," because if conscious imagery represented the rotation the illusion would be immediately dispelled. However, a number of research studies have demonstrated circumstances in which individuals can fail to discover quite simple insights on the basis of mental imagery alone. Rather than arising from a form of "imagery blindsight," the persistence of the illusion can be accounted for by existing cognitive theories of mental discovery.


blindsight, cognitive illusion, mental image, persuasion, psychology of science.
1. Margolis (1998) discusses the intriguing case of a perceived illusory collision in a Tychonic model of the solar system which has persisted for 400 years, despite the model's being carefully evaluated by the best experts in the field. Strikingly, however, the illusion can easily be dispelled by visually presenting a dynamic cutout of Tycho's original diagram. Margolis argues that perception of the illusory collision cannot be based on conscious mental imagery because "if conscious imagery showed us what I just described (with the sun carrying the orbits of Venus and Mercury but the larger orbits staying in place), it would be obvious that it was mistaken" (Margolis, paragraph 20).

2. There are a number of empirical studies, however, which have demonstrated that people can often fail to reach novel insights on the basis of conscious imagery alone, despite the fact that such insights become obvious when the same material is made available as a percept rather than a mental image. In an early demonstration of this Reed & Johnsen (1975) showed that participants were extremely poor at discovering hidden figures within imaged patterns, compared to when the same patterns were made visually available as drawings. More recently, Verstijnen (1997) has shown that the detection of hidden figures in imaged patterns is only at chance level, regardless of whether the participants are psychology or industrial design engineering students, but performance is significantly improved when participants are allowed to represent visually the patterns by sketching.

3. The persistence of the illusory collision in a Tychonic system, and the ease with which it is dispelled by viewing a dynamic model of the same system, highlights a discrepancy between trying to reach a conclusion based on a percept and trying to reach the same conclusion based on an equivalent mental image. Much of the literature on imagery over the last fifty years has tended to emphasise the similarities between imagery and perception, but more recently many of the differences between the two processes have become apparent. In a classic demonstration of this, Chambers & Reisberg (1985) used the duck/rabbit ambiguous figure, the simple outline that is initially perceived by the viewer as representing either a duck or a rabbit, but within a few seconds the alternative interpretation becomes readily apparent. After this, continued viewing of the figure results in spontaneous reversals of the percept between the two alternative interpretations similar to the alternations experienced while viewing a Necker cube.

4. Chambers & Reisberg presented the duck/rabbit figure to naive participants long enough for them to form a mental image of the figure, but but not long enough for the alternative interpretation to be deduced from the percept alone. Participants were then asked to try to find the alternative interpretation purely by consulting their mental image of the figure. Surprisingly, none were able to do this, despite being exposed to other ambiguous figures in a training session. This cannot be attributed to the image's failing to represent the duck/rabbit figure accurately, as when participants were allowed to draw their image the alternative interpretation once again became readily apparent. Hence, a discovery which is very easy to make using a percept is very difficult to make using a mental image, even when the image represents the same visual information as the percept.

5. Reisberg (1996; Reisberg & Logie, 1993) has accounted for this powerful effect in terms of a mental reference frame which can constrain the types of discoveries that can be made using a mental image alone. Reisberg argues that images, unlike percepts, cannot be inherently ambiguous, but are instead always interpreted within the context of a frame of reference which specifies how the image should be understood (e.g., which part is "front", which is "back"). Hence, when individuals form an image of the duck/rabbit figure this image is contained within a reference frame based on the original interpretation of the figure; i.e., it depicts either a duck or a rabbit, but not both. A successful reinterpretation of the image therefore requires a reversal of this reference frame, and this can be extremely difficult without some form of external support. Chambers & Reisberg (1992) found that the distortions of images of the duck/rabbit figure depended on the reference frame in which the figure was first interpreted: Participants who imaged the figure as representing a duck tended to emphasise the "beak" aspects of the figure and to de-emphasise the bump on the figure representing the rabbit's mouth.

6. One weakness of basing such a theory on the reversal of ambiguous figures is that it does not capture the transformation/manipulation aspect of much imagery use during reasoning. It was therefore not apparent whether similar constraints would apply to other forms of imagery use. In an attempt to address this issue, together with colleagues at Aberdeen University (Pearson, Logie & Green, 1996), I carried out a series of experiments using a guided image manipulation task (Finke et al., 1989). This required participants to mentally manipulate alphanumeric symbols in response to verbal instructions and then to make a novel interpretation of the resulting figure. For example, "Imagine a capital `J'. Imagine a capital `D'. Rotate the `D' 90 degrees to the left. Place the `D' directly on top of the `J'." Participants were then asked whether the resulting image reminded them of any object or symbol (in this case, an umbrella). In a series of such experiments we contrasted performance on this task using imagery alone with when participants were allowed to support their images visually by drawing. In all cases performance based on drawing-aided perception was significantly better than performance based on imagery alone. Many participants expressed great surprise at how "obvious" the answer was once they were allowed to draw their mental image on a sheet of paper. This is reminiscent of Margolis's account of Owen Gingerich's surprise on viewing the dynamic cutout of Tycho's model.

7. Considering the above results, we realised that even drawing could only provide a representation of the static aspects of the task, and that participants would still have to mentally carry out the dynamic aspects such as rotation. Recently we have explored this using a computer graphics package which allows participants to visually perceive their rotation and size-changes of the alphanumeric symbols (Pearson & Logie, in prep.) We have found that performance of the manipulation task using the graphics package, in which the dynamic aspects can be perceived, is significantly better than when using either imagery alone or aided only by drawing.

8. This result may shed some light on why the perception of a dynamic model of Tycho's system immediately removes the illusory collision, while perception of a static representation of the model clearly cannot. If a mental discovery depends on some form of dynamic transformation, static representations such as drawings may not provide sufficient support for a successful reinterpretation of an image's reference frame.

9. In terms of the Tychonic system, the image generated to represent the model is interpreted within an extremely strong frame of reference which is based on an incorrect understanding of the model. I agree with Margolis that the knowledge that creates the illusory collision can be considered largely unconscious, and consists of (a) prior experience of small objects tending to circle large objects, (b) possibly an unconscious influence of the Copernican view of the solar system, and (c), for the majority of people, prior knowledge that the illusory collision "exists," i.e., that it has been deduced by others. In the studies described above, participants do not consciously recognise the forces that continue to cause them to interpret `J' and `D' as separate letters rather than constituent parts of a recognisable object. Similarly, prior experience of ambiguous figures still does not make it any easier to reinterpret the duck/rabbit figure using imagery alone. However, unlike Margolis, I continue to feel that conscious imagery is very much at work here. Also, the persistence of the illusion, rather than being related to a phenomenon like blindsight, is more likely to be due to the way mental images function as internal representations generated from prior knowledge, in contrast to the types of external representations that we can derive from perception.

10. Of course, for every demonstration of how imagery can fail during reasoning there are numerous fascinating examples of cases where imagery can clearly lead to novel and inventive discoveries (e.g., Einstein's use of combinatory play; Kekule's insight into the structure of benzene, etc.). Since the original publication of the Chambers & Reisberg paper, there has been substantial debate about the circumstances under which imagery may or may not support mental discoveries. I would recommend the book "Stretching the Imagination" (Cornoldi et al., 1996) to those interested in pursuing these issues further.


Chambers, D., & Reisberg, D. (1985). Can mental images be ambiguous? Journal of Experimental Psychology: Human Perception and Performance, 11, 317-328.

Chambers, D., & Reisberg, D. (1992). What an image depicts depends on what an image means. Cognitive Psychology, 24, 145-174.

Cornoldi, C. Logie, R.H. Brandimonte, M.A. Kaufmann, G. and Reisberg D. (1996; Ed.), Stretching the imagination: Representation and transformation in mental imagery. New York: Oxford University Press.

Finke, R., Pinker, S., & Farah, M.J. (1989). Reinterpreting visual patterns in mental imagery. Cognitive Science, 13, 51-78.

Margolis, H. (1998) 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

Pearson, D.G. & Logie, R.H. (in preparation). Seeing is believing: The importance of perceiving movement during mental discovery.

Pearson, D.G., Logie, R.H., & Green, C. (1996). Mental manipulation, visual working memory, and executive processes. Psychologische Beitrage, 38, 324-342.

Reed, S.K., & Johnsen, J.A. (1975). Detection of parts in patterns and images. Memory & Cognition, 3, 569-575.

Reisberg, D. (1996). The nonambiguity of mental images. In C. Cornoldi, R.H. Logie, M.A. Brandimonte, G. Kaufmann, and D. Reisberg (Ed.), Stretching the imagination: Representation and transformation in mental imagery. New York: Oxford University Press.

Reisberg, D., & Logie, R.H. (1993). The ins and outs of visual working memory: Overcoming the limits on learning from imagery, In M. Intons-Peterson, B. Roskos-Ewoldsen, & R. Anderson (Eds.), Imagery, creativity, and discovery: A cognitive approach, pp. 39-76. Amsterdam: Elsevier.

Verstijnen, I.M. (1997). Sketches of Creative Discovery. Thesis Technische Universiteit Delft.

Volume: 9 (next, prev) Issue: 34 (next, prev) Article: 3 (next prev first) Alternate versions: ASCII Summary