Professional Documents
Culture Documents
The Poggendorff Illusion Explained by Natural Scene Geometry
The Poggendorff Illusion Explained by Natural Scene Geometry
The Poggendorff Illusion Explained by Natural Scene Geometry
scene geometry
Catherine Q. Howe, Zhiyong Yang, and Dale Purves*
Center for Cognitive Neuroscience and Department of Neurobiology, Duke University, Durham, NC 27708
effect is reduced but not completely abolished (2, 3, 5). of ⫾25 mm (Fig. 2A). By using this system, we collected ⬇100
Although a variety of theories have been proposed to account wide-field images extending 333° horizontally and 80° vertically
for the Poggendorff illusion (see Discussion), none of these at a resolution of 0.144°. Of these images, 25 were obtained in
explanations rationalizes the full range of behavior illustrated in fully natural environments in undeveloped terrain, and the rest
Fig. 1. Here, we test the hypothesis that the Poggendorff effect were obtained in outdoor or indoor settings that included human
and its altered behavior in different presentations are generated constructions (see also ref. 7). Only the midportion of the
by the accumulated experience of humans with the various wide-field images (⬇110° horizontally ⫻ 52° vertically) was used
sources of interrupted linear features in typical environments. By for analysis.
collecting a range image database with information about the 3D
structure of a large series of natural scenes, we assessed whether
these geometrical illusions are indeed explained by the proba- Freely available online through the PNAS open access option.
bility distribution of the relative locations of line segments across *To whom correspondence should be addressed. E-mail: purves@neuro.duke.edu.
a spatial interval. © 2005 by The National Academy of Sciences of the USA
Empirical Framework and Hypothesis. When the continuity of a percepts according to the accumulated experience in dealing
physical object is interrupted by occlusion, the location of its with the physical sources of retinal stimuli. If this hypothesis is
reappearance on the other side of the interruption is inevitably correct, the percepts elicited by the stimuli in Fig. 1 should
uncertain: there are many ways the object could have traversed reflect the past experience of human observers with the physical
the interval, and the interruption does not provide information sources of straight-line segments that are in the same orientation
about which of the possible ways actually generated the stimulus. but separated by a spatial interval. More specifically, given a line
Thus, the oblique line segments in the standard Poggendorff segment on one side of an implied occluder, the perceived
presentation in Fig. 1 could arise from a single object with an position of a second line segment in the same orientation but on
infinite variety of possible configurations or from two different the other side should be predicted by the relative probability of
objects altogether. In short, the stimulus reflects the fundamen- occurrence of the physical sources of line segments projected in
tal uncertainty of how the physical sources of projected line that orientation.
segments are actually continued across a spatial interval. The
hypothesis examined here is that the way humans produce useful Sampling the Database. Fig. 2B shows examples of the geometrical
behavioral responses in the face of this uncertainty is to generate templates we applied to the range images to sample the physical
occurrence of the physical sources of right oblique line segments be biased one way or another by an observer’s past experience.
at different vertical locations relative to the left segment should Accordingly, a right line segment that was actually collinear with
determine how observers perceive the relative positions of the the left segment would be seen as such. As shown in Fig. 3A,
two segments in the Poggendorff stimulus. Fig. 3A shows that however, the real-world sources of the right line segment, given
when the left oblique segment is oriented downward and to the the left oblique segment and the vertical occluder, are biased
right, as in Fig. 1 A, the probability distribution of the possible toward locations above the point of collinearity with the left
sources of the right segment is biased toward locations that segment. Assessed against this body of past experience, then, a
correspond to negative values of ⌬. In other words, the majority right line segment that is geometrically collinear with the left
of the physical sources that give rise to projections of right segment would appear shifted downward. Thus, the probability
oblique line segments in the same orientation as the left segment distribution derived from the fully natural scenes in the database
will, in the past experience of human observers, have projected can indeed account for the standard Poggendorff effect.
above the point at which an extension of the left oblique line Conversely, when the orientation of the interrupted line in the
intersects the right vertical line. This probability distribution stimulus is upward from the lower left, as in Fig. 1B, the probability
Howe et al. PNAS 兩 May 24, 2005 兩 vol. 102 兩 no. 21 兩 7709
Fig. 4. The probability distribution of the physical sources of the right line Fig. 5. Probability distributions of the physical sources of the right line
segment changes progressively as the orientation of the interrupted line (A) segment when the standard Poggendorff stimulus is decomposed or rotated.
or the width of the interruption (B) is altered. ␣ represents the angle of (A) The probability distribution obtained when only the acute components of
intersection between the interrupted oblique line and the vertical lines, and the stimulus are present, compared with the distribution when only the
w is the width of the interruption. w ⫽ 1° in (A) and ␣ ⫽ 45° in (B); both values obtuse components are considered. In both cases, the orientation of the
were measured in terms of visual angle. interrupted line was 45° and the width of the interruption 1°. (B) The prob-
ability distribution obtained when the standard Poggendorff stimulus is
rotated such that the interrupted line is horizontal (see Fig. 1G). The distri-
bution for the standard stimulus is replotted from Fig. 3A for comparison.
distribution of the sources of the right line segment is biased toward
locations below the point at which a continuation of the left line
intersects the right vertical line (i.e., toward values of ⌬ ⬎ 0) (Fig. acute angles in the standard stimulus are presented but little
3B). By the same reasoning, a right oblique segment that is actually affected when the presentation is restricted to the obtuse com-
collinear with the left segment will in this case be seen to be ponents (see Fig. 1F). As shown in Fig. 5A, when only the acute
somewhat above the point where an extension of the left segment elements are used as the templates for sampling the range images
intersects the right vertical line, as indeed it is.
(see Fig. 2D), the mode of the probability distribution (arrow) is
A similar explanation applies to the Poggendorff stimulus in
very near the point where ⌬ equals zero. Accordingly, the
which an oblique line is interrupted by two horizontal lines.
Poggendorff effect should be largely abolished. Conversely,
When the left oblique segment is oriented downward and to the
when only the obtuse components are used, the probability
right, as in Fig. 1C, the probability distribution of the sources of
distribution is similar to that of the standard Poggendorff
the right line segment is biased toward locations to the left of the
point at which ⌬ ⫽ 0 (Fig. 3C). Thus, the location of the right stimulus; thus, the illusion would be expected to retain its full
segment appears shifted to the right from the continuation of the magnitude.
left line, which is again consistent with what people see. Finally, we determined the probability distribution of the
sources of the right line segment when the overall orientation of
Explanation of Other Features of the Poggendorff Effect. Changing the Poggendorff stimulus is rotated so that the interrupted line
either the orientation of the interrupted line or the width of the is horizontal, as in Fig. 1G. Compared with the standard
interruption systematically also shifts the relevant probability presentation, the distribution when the stimulus is rotated 90°
distributions in a manner consistent with the altered magnitude has a mode closer to ⌬ ⫽ 0, as shown in Fig. 5B. As a result, the
of the perceptual effects in these presentations. The effects of Poggendorff effect should be reduced but not abolished when
varying the intersecting angle of the interrupted line (Fig. 1D) the presentation of the interrupted line is horizontal, as it is.
and of varying the width of the interruption (Fig. 1E) on the
corresponding probability distributions of the physical sources of Statistics Derived from Different Types of Scenes. Theories purport-
the right line segment are shown in Fig. 4 A and B, respectively. ing to explain one or more aspects of the Poggendorff illusion
As the orientation of the interrupted line becomes closer to the have often been based on intuitions about the ‘‘carpentered’’
orientation of the lines defining the interruption (vertical in world of human artifacts (9, 10). It was thus of interest to ask how
the example in Fig. 4A) or as the interruption becomes wider, the the probability distributions derived from fully natural scenes
mode of the distribution shifts progressively away from the point (presumably representative of the human visual environment
at which ⌬ ⫽ 0, in accord with the fact that the perceived shift during evolution) compare with distributions derived from en-
in the apparent location of the right line segment increases as the vironments in which human construction is prevalent. We there-
angle of the intersection becomes smaller or as the width of the fore computed the relevant probability distributions from the set
interruption increases (3). of scenes in the database that included some or many man-made
Another puzzle that can be explained in these terms is the fact objects. As shown in Fig. 6, the pertinent probability distribu-
that the Poggendorff effect is greatly diminished when only the tions obtained from the two types of scenes are generally similar.
Discussion
Despite numerous studies (1), the Poggendorff effect remains
one of the most controversial of the geometrical illusions. Here
we have shown that this effect and its complex behavior in
different presentations can be accounted for by the relationship
NEUROSCIENCE
Fig. 7. Physical basis for the bias in the probability distributions of the between the stimulus in the image plane and its possible sources
sources of Poggendorff stimuli. (A) In the standard stimulus, a right line in natural visual environments. From the analysis presented, it
segment above the point at which ⌬ ⫽ 0 (gray arrowhead) is, on average, should be apparent that the biased image–source relationships
closer to the center of the plane containing the physical source of the left seg- that account for the Poggendorff effect and its variants are
ment and the vertical lines than is an otherwise comparable right line segment rooted in the statistical properties of the physical world in which
below the point at which ⌬ ⫽ 0. The distances of the right line segments from human observers reside. Thus, this relationship has to be derived
the center of the plane (black dot within the dashed line that outlines the empirically by examining the natural environment and cannot be
plane) are indicated by the gray dotted lines; the difference in the lengths of
deduced from principles of projective geometry per se.
the two gray dotted lines is indicated by the black bar below the diagram. (B)
As the orientation of the interrupted line becomes more vertical (Left) or as
the width of the interruption increases (Right), the same shift in the vertical Previous Explanations of the Poggendorff Effect. The explanation
position of the right line segment means a larger difference in the distances most often offered in the past has been that the illusion derives
of the right line segments from the center of the plane. The differences in the from a misperception of the angles in the stimulus (11, 12). For
lengths of the gray dotted lines are again indicated by the black bars. instance, an overestimation of the acute angles in the standard
Howe et al. PNAS 兩 May 24, 2005 兩 vol. 102 兩 no. 21 兩 7711
stimulus in Fig. 1 A and an underestimation of the obtuse ones fundamentally uncertain significance of any given retinal pro-
would presumably affect the apparent orientation of the line jection for visually guided behavior. Because 3D spatial rela-
segments on either side of the interruption toward horizontal, tionships in the physical world are transformed into 2D rela-
thus causing them to appear noncollinear. This explanation, tionships in the image plane, the physical reality underlying any
however, has been disputed (1) on the grounds that it cannot given geometrical relationship in the retinal image is necessarily
rationalize the paradoxical abrogation or preservation of the ambiguous: a given configuration in the image plane could have
illusion elicited, respectively, by presentation of only the acute or been generated by many different physical sources. It seems
only the obtuse components of the stimulus (see Fig. 1F). Nor inevitable, therefore, that this problem be addressed by taking
does this explanation account for the additional features of the advantage of past experience with retinal images and their
Poggendorff effect shown in Figs. 1D, E, and G. physical sources. The behavioral interactions of humans and
Another explanation is based on the idea that the geometrical other animals with the physical sources of images in natural
information in the retinal projection might be inappropriately visual environments have presumably led to the instantiation of
‘‘interpreted’’ by obser vers. For example, the ‘‘depth- the relevant statistical relationships in visual system circuitry
processing’’ theory suggests that the oblique lines in the Poggen- over evolutionary time. Indeed, a growing body of evidence
dorff stimulus are interpreted as lines extending in depth, and supports this concept of vision (see, for example, refs. 15 and 16).
are therefore perceived to be noncollinear (10, 13, 14). This More specifically, statistical analyses of natural geometry perti-
category of explanation is also limited in that it considers only nent to the perception of surface shape and orientation (17),
some aspects of the projective geometry of the stimulus and its egocentric distance (18), spatial intervals (6), and object size (19)
possible sources and cannot explain the full range of the percepts have successfully explained complex perceptual phenomena in
elicited by the configurations in Fig. 1. each of these several categories of perceived geometry.
Biological Rationale. An important question not yet fully ad- We thank F. Long, S. Nundy, D. Schwartz, and J. Voyvodic for useful
dressed is why, from the perspective of biological success, comments and criticisms. This work was supported by the National
generating geometrical (and other visual) percepts on a statis- Institutes of Health, the Air Force Office of Scientific Research, and the
tical basis is useful and indeed necessary. The answer lies in the G. B. Geller Endowment.
1. Robinson, J. O. (1998) The Psychology of Visual Illusion (Dover, Mineola, NY). 10. Gregory, R. L. (1997) Eye and Brain (Princeton Univ. Press, Princeton, NJ).
2. Day, R. H. & Dickinson, R. G. (1976) Br. J. Psychol. 67, 537–552. 11. Blakemore, C., Carpenter, R. H. S. & Georgeson, M. A. (1970) Nature 228, 37–39.
3. Weintraub, D. J. & Krantz, D. H. (1971) Percept. Psychophys. 10, 257–264. 12. Burns, B. D. & Pritchard, R. (1971) J. Physiol. (London) 213, 599–616.
4. Day, R. H. (1973) Percept. Psychophys. 14, 590–596. 13. Gillam, B. (1971) Percept. Psychophys. 10, 211–216.
5. Leibowitz, H. & Toffey, S. (1966) Vision Res. 6, 101–103. 14. Gregory, R. L. (1963) Nature 199, 678–780.
6. Howe, C. Q. & Purves, D. (2002) Proc. Natl. Acad. Sci. USA 99, 13184–13188. 15. Purves, D. & Lotto, B. (2003) Why We See What We Do: An Empirical Theory
7. Huang, J., Lee, A. B. & Mumford, D. (2000) in Proceedings of IEEE Conference of Vision (Sinauer, Sunderland, MA).
on Computer Vision and Pattern Recognition (IEEE, Piscataway, NJ), Vol. 1, pp. 16. Rao, R. P. N., Olshausen, B. A. & Lewicki, M. S. (2002) Probabilistic Models
324–331. of the Brain: Perception and Neural Function (MIT Press, Cambridge, MA).
8. Howe, C. Q. & Purves. D. (2005) Proc. Natl. Acad. Sci. USA 102, 1228–1233. 17. Yang, Z. & Purves, D. (2003) Network: Comp. Neural Syst. 14, 371–390.
9. Gillam, B. (1998) in Perception and Cognition at Century’s End, ed. Hochberg, 18. Yang, Z. & Purves, D. (2003) Nat. Neurosci. 6, 632–640.
J. (Academic, New York), pp. 98–136. 19. Howe, C. Q. & Purves, D. (2004) J. Cognit. Neurosci. 16, 90–102.