Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Additional evidence that contour attributes are not essential cues for object recognition

Additional evidence that contour attributes are not essential cues for object recognition It is believed that certain contour attributes, specifically orientation, curvature and linear extent, provide essential cues for object (shape) recognition. The present experiment examined this hypothesis by comparing stimulus conditions that differentially provided such cues. A spaced array of dots was used to mark the outside boundary of namable objects, and subsets were chosen that contained either contiguous strings of dots or randomly positioned dots. These subsets were briefly and successively displayed using an MTDC information persistence paradigm. Across the major range of temporal separation of the subsets, it was found that contiguity of boundary dots did not provide more effective shape recognition cues. This is at odds with the concept that encoding and recognition of shapes is predicated on the encoding of contour attributes such as orientation, curvature and linear extent. attributes provide the essential cues for object recognition. Background "Stationary visual percepts, a tree, a stone, or a book, are That study displayed dots that were positioned on the as a rule extremely reticent as to the nature of the neural outer boundary of namable objects, varied the number of events which underlie their existence. We may hope to dots that were displayed in progressively larger samples, learn more about brain correlates if we turn to instances and manipulating the spatial positioning of dots within in which percept processes seem to be in a more active those samples. The display dots were chosen to provide state." Kohler [1] subsets that either: a) formed contiguous strings that approximated line segments, b) were at randomly selected Cognitive, computational and neural theories of object positions around the boundary of the shape, or c) were at recognition all share the concept that essential cues are evenly spaced positions around the boundary. For each provided by the orientation, curvature and linear extent of condition the number of dots to be displayed (as a per- the lines and edges that lie at the boundary of an object centage of the total number of dots in the perimeter) was and its component parts. We can describe these cues as increased until the participant identified the object. The "contour attributes," and describe the mechanisms for greatest percentage of dots was required for recognition registering and encoding these attributes as "contour fil- when the subsets formed contiguous strings, and the ters." smallest percentage was needed when the dots in the sub- sets were positioned with even spacing. The contiguous A previous study from this laboratory [2] offered evidence strings would have delivered the most information about and arguments against the proposition that contour contour attributes, yet this treatment condition provided Page 1 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 the least effective cues for eliciting recognition of the respectively. The horizontal and vertical dimensions of shapes. the full array were 7.74 × 7.74 arc°. There were two additional reasons that the previous The circuits of the display board provided for activation of results [2] were at odds with the concept that contour a given LED by specifying an x, y address position within attributes provide essential cues for recognition. First, it is the array, under the control of a microprocessor running widely believed that orientation-selective cells in primary with a clock speed of 24 Mhz. Rise and fall time for emis- visual cortex serve as the basic filters that register the con- sion was in the range of 100 ns. Room illumination was tour attributes [3,4]. Yet recognition of many shapes was from standard ceiling-mounted fluorescent fixtures that possible with only a small sampling of dots, and with the were fitted with opaque panels to block most of the light. space between adjacent dots being wider than the recep- This provided ambient illumination of 13.3 lux. Lumi- tive fields of orientation-selective cells. Second, even if nance of an emitting LED was set at 7 Cd/m . one proposed new principles for registering alignments among these dots, it is not obvious how one would know The experiment displayed 64 shapes, these being the same which dots to connect. as listed in Table 1 of reference [2], but with the elimina- tion of five shapes in order to provide an equal number of The present experiments used the minimal transient dis- shapes for each treatment condition. Each shape was rep- crete cue (MTDC) protocol [5-7] to examine whether dot resented by discrete boundary dots that formed a con- subsets that should activate contour filters (because of nected string of adjacent positions within the 64 × 64 spatial contiguity) provide better shape cues than do non- array. Boundary dots for one shape, the frog, are illus- contiguous subsets. This protocol uses brief and succes- trated in the left panel of Fig. 1. sive display of subsets that were chosen from the full inventory of boundary dots. Successful recognition of the As implemented here, the MTDC protocol was as follows. shape required integration of the information provided by For a given shape, only some dots from the full inventory each subset, and the level of performance reflected the were shown, these being designated as the display set. The degree to which that information was useful. The prior size of the display set was determined on the basis of test- research [6,7] found that millisecond-level separation of ing done with other subjects. These tests established the subsets produced significant declines in recognition. This number of evenly spaced dots needed to provide a 75% was the case even when the total display time for a given hit rate, i.e., successful recognition of a given shape by shape – and thus duration of cue persistence – was con- trolled. This means that the effectiveness of the shape cues, and their ability to be integrated, are reflected in the rate at which recognition declines when time differentials are inserted between the successive cues. For the present experiment, the subsets provided either contiguous sequences of four dots, or four dots chosen at random locations in the boundary. Each contiguous dot subset would provide information about the local con- tour attributes of the shape. Each random dot subset would not provide this information, and to the extent that the subset might activate contour filters, would deliver The left fu the shapes Figure 1 ll inventory of panel uses u dots th nfilled at mar dots ked the boundary for on to show the positions of e of the The left panel uses unfilled dots to show the positions of the inappropriate cues regarding alignments of the boundary. full inventory of dots that marked the boundary for one of Therefore, to the extent that contour attributes are essen- the shapes. This full inventory was never displayed. Rather, a tial cues for shape recognition, performance levels should sampling of dots, designated as the display set, was shown, be higher for contiguous than for random subsets. The using a sample size that was expected to yield recognition on experimental results did not support this prediction. approximately 75% of the trials. An example of a display set is shown by the filled dots in the right panel. For a given dis- Methods play, a random starting point was picked from among the list The apparatus for display of shapes (display board) was of addresses, shown by the arrow, and then every Nth dot the same as used in earlier experiments [2,5,6]. Briefly location was selected for inclusion in the display set (with the outlined here, it consisted of a 64 × 64 array of red LEDs. value of N being that which would yield the number of dots The participant sat at a distance of 3.5 m from the display needed in the display set). The size of display dots has been exaggerated for purposes of illustration. board, and at this distance, the diameter of each element and the center-to-center spacings were 4.95 and 7.42 arc', Page 2 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 75% of the participants, if all the dots were simultane- ously displayed. This might be regarded as providing a given subject with a 75% probability of identifying the shape, and if only for convenience, it may be described in this manner in what follows. To specify the display set for each shape (done independ- ently for each participant), the first address to be included in the set was chosen at random. From there, counting in th a clockwise direction, every N address was chosen, the value of N being that which would yield the 75% hit rate. An example of one possible display set is illustrated in the right panel of Fig. 1. Each display set was further broken into randomly chosen subsets containing four dots each (with one residual sub- set potentially having fewer than four). Spatial position- ing of dots within the subsets provided the experimental treatment designated as "proximity," and temporal sepa- ration of subsets provided the treatment designated as "temporal separation," also known as T3. There were two levels of the proximity condition, requir- ing either that the dots of the subset be contiguous, or that they be randomly selected from among the members of the display set. Note that contiguity is relative, in that the th position within the full display set consists of every N inventory of boundary dot positions. For the random con- dition, there was an additional restriction that each dot in the subset must lie at least three steps away from other positions within the display set. The panels of Fig. 2 illustrate four subsets of contiguous dots sampled from the display set that is shown in Fig. 1. A given display tain were displayed successively Figu ing four dots each (plus any residual), re 2 set was broken at random into subsets con- and these subsets A given display set was broken at random into subsets con- The left panels show the location of each subset, superim- taining four dots each (plus any residual), and these subsets posed on the full complement of boundary dot positions. were displayed successively. This figure provides examples of The right panels show how each subset would appear contiguous subsets. The left panels show the location of sub- upon display. Each subset would be shown in rapid suc- set dots within the full inventory of dots. The right panels cession using one of the T3 intervals described below. show how each subset would appear on the display board, with the time interval for display of a given subset being 0.4 The panels of Fig. 3 provide a corresponding illustration ms. of randomly positioned subsets that again are members of the display set shown in Fig. 1. From a comparison of the right column of panels in Figs. 2 and 3 one can see the essence of the experimental concept. The contiguous-dot T3 specified the time interval between offset of one subset subsets reflect the local contour attributes of the frog, and onset of the next. There were four levels of T3, these whereas the random-dot subsets do not. being 1, 3, 9 and 27 ms. These time parameters are illus- trated in Fig. 4. Dots were displayed successively. Each dot was displayed for 0.1 ms, this being designated as T1. T2 specified the Total display time for a given shape was determined by interval from onset of one address within a subset to the the number of dots in the display set multiplied by 0.1 ms, next address of the same subset. This was also fixed at 0.1 plus the T3 interval multiplied by the number of subsets ms, providing for zero delay between offset of one address minus one. For T3 = 1 ms, the minimum and maximum and onset of the next. With these T1 and T2 intervals, all display times were 4.3 and 62 ms respectively, and the addresses within a given subset were displayed in 0.4 ms. mean display time was 16.6 ms. For T3 = 3 ms, minimum Page 3 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 first dot (21,43) T1 (pulse width) = 0.1 ms second dot (12,31) x,y T2 (pulse spacing) = 0.1 ms etc...(x,y) T3 (subset spacing) = 1, 3, 9 or 27 ms 45,18 29,33 21,43 12,31 22,57 15,59 41,40 35,11 The time intervals for displa Figure 4 y of shapes are illustrated The time intervals for display of shapes are illustrated. Fig. 1a shows the duration of emission from any given LED, desig- nated as T1, this being 0.1 ms. Fig. 1b specifies that the onset- to-onset interval of successive pulses within a given subset, designated as T2, which was also 0.1 ms. In Fig. 1c the pulses are illustrated as a string of beads. The four addresses of each subset are displayed as a group, separated by the T3 interval. For display of a given shape, theT3 interval was 1, 3, 9 or 27 ms. then shapes were assigned at random from the ranked list to the eight treatment combinations. The net effect of the assignment was to provide each treatment level with a sampling of shapes that were approximately equal in dif- ficulty. Each participant saw a given shape only once, and the order for display of the shapes (and thus the treatment combinations) was random. Th were ran Figure 3 is figure provides domly selected examples of subsets in which the dots Eight USC undergraduates served as participants, each dis- This figure provides examples of subsets in which the dots playing normal or corrected to normal visual acuity. Each were randomly selected. Again, the left panels show the loca- was naïve to the goals of the experiment, and was paid for tion of the subset dots within the full inventory of dots, and his or her participation. the right panels show how the dots in each subset would appear. Note that contour attributes, e.g., orientation, curva- Results ture and length, can be seen in the contiguous subsets shown The response variable was binary, i.e., recognize or failure in Fig. 2, but are not present in the randomly chosen subsets to recognize. Participants were treated as random samples shown here. from the population of possible participants. The order of presentation of shapes was randomly specified for each participant, and the treatment combination shown for a and maximum display times (rounded up) were 10 and given shape and participant was selected at random. Thus 150 ms, with a mean of 40 ms. For T3 = 9 ms, minimum the appropriate statistical model is a Generalized Linear and maximum display times were 28 and 414 ms, with a Mixed Model [8] with random effects of Participant and mean of 126 ms. For T3 = 27 ms, minimum and maxi- Shape, and fixed effects of Proximity and T3 interval. Logit mum display times were 82 and 1206 ms, with a mean of values (log (proportion/1-proportion) were calculated, 320 ms. and treatment differences were compared using the stand- ard error of the difference (SED) for these values. Model The two levels of dot proximity and four levels of T3 pro- predictions and standard errors of the mean for each of vided eight treatment combinations. For each participant these predictions are shown in Table 1. the inventory of 64 shapes were ranked for difficulty level, i.e., the number of dots required for a 75% hit rate, and Page 4 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 Table 1: Generalized Linear Mix Model Values for Treatment Conditions Contiguous Subsets Random Subsets T3 (ms) Mean SEM Backtransformed Mean SEM Backtransformed 1 1.056 0.369 0.742 1.435 0.382 0.808 3 0.879 0.363 0.707 0.537 0.359 0.631 9 0.213 0.354 0.553 0.284 0.354 0.570 27 -0.442 0.358 0.391 -1.425 0.386 0.194 The Generalized Linear Mixed Model transforms the percent recognition of shapes into logit values, i.e., log (proportion/1-proportion). The mean logit values and the standard errors of these means are given for each T3 interval for the contiguous and random subset data. The logit values have also been backtransformed into model predictions of recognition rate. This statistical analysis found no significant difference (p mark the outer boundary of the shape, the number of dots = 0.59) in recognition rate for the proximity condition, being just sufficient for recognition of the shape if all of i.e., recognition of shapes was not different as a function them are shown with minimal delay. By choosing which of whether the subset dots were contiguous or were at ran- dots to sample, and introducing delays between succes- domly selected positions. sive samples that are chosen, one can assess the effective- ness of the shape cues being provided by the samples. There was a significant (p < 0.001) linear decline in recog- nition rate as a function of the temporal separation The present goal was to examine whether contiguous sub- between subsets, and the quadratic component was not sets of dots would be more effective at eliciting recogni- significant (p = 0.22). The model predictions were back- tion of shapes than would subsets having an equal transformed into values that reflect the percentage of number of dots that were randomly chosen from the full shapes that were recognized for each of the treatment con- inventory of dots. The contiguous subsets should provide ditions. These predictions are very near the arithmetic mean recognition percentages that are plotted in Fig. 5 for contiguous and random subsets at each of the T3 inter- vals. 80 Although the difference between contiguous and random treatment conditions was not significant, inspection of the means plotted in Fig. 5 suggest the possibility that the treatments were not comparable at T3 = 27 ms. To for- mally evaluate this, pairwise comparisons of means were calculated, properly adjusting for the number of compar- isons. There were no significant differences at the first three T3 intervals, but the difference at T3 = 27 ms was sig- nificant at p < .02. This differential could be a simple experimental artifact, in that a treatment will not always yield data that fits the overall trends. Discussion 0.0 0.3 0.6 0.9 1.2 1.5 A great many, perhaps a majority, of shape recognition Temporal Separation of Subsets ( log milliseconds) theories propose that contour attributes, i.e., orientation, curvature and linear extent, provide the elemental features that define the shape of an object. Selfridge [9] may have Mean percent rec tion set Figure 5 s is plotted against the time ognition for each of th interval separating each su e proximity condi-b- been the first to characterize the perceptual process in Mean percent recognition for each of the proximity condi- terms of an assemblage of filters, each having the ability tions is plotted against the time interval separating each sub- to register a distinctive contour attribute, but many others set. Contiguous subsets are shown with filled circles, and have followed this lead [see [10-14]]. random subsets are shown with open circles. The decline in recognition was significant across the tested time intervals, The minimal transient discrete cue (MTDC) protocol [5-7] but the proximity conditions did not produce differential lev- provides a means to evaluate the validity of this hypothe- els of shape recognition. sis. This method briefly displays a spaced array of dots that Page 5 of 9 (page number not for citation purposes) Percent Shape Recognition Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 a more effective stimulus for the filters that are presumed assemblage of these contour filters delivers the full com- to register the contour attributes. If shapes are specified on plement of contour attributes needed for recognition. the basis of their contour attributes, then the contiguous subsets should convey the best partial shape cues, and one However, previous results from this laboratory [2] raise would expect these subsets to be more effective for elicit- the question of whether shape analysis depends on activa- ing recognition. tion of orientation-selective cells. That study found that recognition was possible when the full complement of The overall result was that contiguous and randomly dots being shown was relatively sparse. Recognition was selected subsets contributed equally to shape recognition, well above chance when dot spans exceeded the length of even though the randomly selected subsets did not dis- orientation-selective receptive fields [23]. That outcome play cues that relate to the orientation, curvature and lin- suggests that each dot is acting as an independent marker ear extent of the boundary. This indicates that under the of boundary position, and that shape is defined by an present test conditions, contour attributes did provide unspecified – not yet known – relationship among the cues that are essential for shape perception. individual markers. Even when the orientation-selective cells are activated by an array of dots, the essential infor- For the present task conditions, one might speculate that mation might be the locations that have been specified information persistence allowed successive dots to accu- rather than the collinearity in the array. mulate, such that dots from the random subsets could eventually form contiguous strings that provided contour With respect to the present results, one might wonder attributes. There is persistence of brief visual stimuli, as whether the contiguous subsets were effective stimuli for reported by Sperling [15], Neisser [16], Haber and Stand- the orientation-selective cells. Perhaps the cells did not ing [17], and Eriksen and Collins [18,19], among others, respond to the very brief presentation of just four dots. and reviewed by Coltheart [20], Long [21], and Nisly and There are three reasons to suggest that the subsets deliv- Wasserman [22]. Whereas local contour information was ered adequate stimulation. not provided by a given random subset, one could argue that the contour-filtering process simply waited for a First, although the stimulus duration was very brief, the number of the subsets to be delivered, after which the con- flashes were easily visible, i.e., consciously perceived. It is tour attributes could be extracted from the aggregate pool generally accepted that conscious awareness of a visual of dots. stimulus requires processing by the primary visual cortex, thus the stimulus strength was adequate for activating its Recent work using the present experimental protocols, neurons. however, has found that millisecond and even submilli- second differentials in the display of dot subsets can pro- Second, the span of each contiguous subset was a suitable duce significant differences in shape recognition [6,7]. fit to the size of receptive fields. Sceniak et al. [23] exam- The result that is most critical to this discussion was pro- ined receptive field size of orientation-selective cells in V1 vided by the second experiment in each of the cited stud- of Macaque, and found the average space constant to be ies, wherein the total time (and thus duration of 60 arc', and the average length-summation tuning curve to persistence) for a given shape was held constant. Under be 49 arc'. The four-dot array of the contiguous subsets these conditions, it was found that varying the interval spanned 35 arc' for horizontal or vertical alignments, and between successive dots impaired recognition, with tem- 47 arc' for diagonal alignments. Therefore each of the con- poral separation of as little as half a millisecond being sig- tiguous subsets displayed an image size that would pro- nificant. Shape-relevant contour attributes are delivered vide four dots to the receptive fields. directly by the contiguous dot subsets, but they could be provided by random subsets only through aggregation. Third, there is direct electrophysiological evidence that an The prior studies demonstrate that the cues do not aggre- array of briefly flashed dots will stimulate the cortical gate without a recognition penalty. cells. Jones & Palmer [24] examined responsiveness of ori- entation-selective cells with successive stimulation of When neural substrates for shape perception are dis- local points across the receptive fields, the typical dura- cussed, most see the orientation-selective cells character- tion of each stimulus being 50 ms. They reported that the ized by Hubel & Wiesel [3,4] as providing the first step for responses that could be elicited by stimulating one loca- registering contour attributes. A given cell can be activated tion at a time was too weak to be of practical value in the by a contour, and because the firing rate is influenced by analysis of receptive field structure. However, simultane- the orientation, length, and (possibly) curvature of the ous activation of three sites within the receptive field contour, the response is thought to convey information yielded usable data. As indicated above, the contiguous about these attributes. It is further suggested that an subsets of the present experiment displayed four dots that Page 6 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 would register on a given receptive field, and this would put a premium on very tight temporal proximity within a provide a stronger stimulus than was found to be effective stimulus pattern. More advanced image-processing sys- by Jones & Palmer [24]. tems, such as primary visual cortex, might have similar requirements for simultaneity, but with a longer time con- The more general point is that the random subsets as well stant. This could explain the differential at T3 = 27 ms as as the contiguous subsets were seen by the subject and a contribution to the temporal-integration process by ori- delivered sufficient stimulation to elicit recognition. If entation-selective cells that could not be accomplished in one took the position that the contiguous arrays provided the retina. an insufficient stimulus for activating orientation-selec- tive cells, it would mean that recognition was accom- The finding that the contour attributes did not benefit rec- plished without any contribution from these cells. ognition under the present test conditions should not be taken as a blanket rejection of a useful role in the percep- It is possible, that the cues used for this experiment may tion of objects. The fact that we can detect edges with a be especially salient for activating a primitive shape contrast differential as small as 3% speaks to the benefit of encoding system. The pattern provided by the full com- these filters for registering the presence of a boundary. plement of dots is very similar to a silhouette, and recog- Doubtless this is useful for detecting an object that is nition is best when there is maximal simultaneity of the almost the same color or luminance as the background, or flashed dots. This is not unlike conditions that might face where it must be seen through haze. Contour filters may an early vertebrate – perhaps a fish – who detects simulta- make it possible to see the object's boundaries under a neous movement through small openings in a wall of sea- variety of degraded conditions, and there is ample evi- weed. The pattern that is seen could be a predator, or dence that alignment of lines and edges provides a basis might be prey, and successful recognition by the creature for object completion. It is possible, however, that this would have implications for survival. It is likely that these processing allows the position of discrete markers to be recognition skills evolved, and are present in a great many specified. Shape perception, per se, may then be based on present-day animals that have no cortex. metric relationships that have little or nothing to do with collinearity of the markers. Recent evidence from this laboratory [25], gathered and published after the present research was conducted, has It is unclear why so many insist that shape is defined by demonstrated that the retina contains a neural system that the orientation, curvature and linear extent of the con- is sensitive to millisecond-level simultaneity when the tours. We know that all manner of cues can contribute to subsets consist of dot pairs. This suggests that the present identification of objects, but have no trouble discarding task draws on primitive shape-encoding mechanisms that most of them as being ancillary. Fig. 6 illustrates the situ- detailed boundary detailed boundary boundary markers internal contours texture color S Figure 6 timuli that are available as shape cues are listed above each image Stimuli that are available as shape cues are listed above each image. The right image provides the number of dots in a display set that allows for 75% successful recognition of the rooster when the dots are displayed successively, each being shown for 0.1 ms, and with a T3 interval of zero. Page 7 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 ation. The left image shows a detailed colored sketch that Temporal separation between members of subset pairs; can be readily identified as a rooster. In fact the image is T3: Temporal separation between subset pairs; V1: Pri- devoid of various depth cues that would be present in the mary visual cortex; 2D: Two dimensional; SEM: Standard real object. Nonetheless, we accept that the 2D image has error of the mean. the shape of a rooster, so the depth cues must be ancillary to our concept of shape. Competing interests The author declares that he has no competing interests. The middle image has eliminated internal contours, tex- ture, and color, replacing all these cues with uniform Authors' contributions black. Yet this silhouette is readily identified as being in EG conceived of the study, designed the study, tested all the shape of a rooster. The internal parts, color and texture participants, and wrote the article. Technical assistance for must be at least somewhat ancillary, i.e., nonessential. programming and data analysis was provided by contract, as noted below. EG has read and approves of the final The right image has replaced the boundary edge with an manuscript. array of dots, and we can still see the stimulus as having the shape of a rooster. Contour attributes of the boundary Acknowledgements I wish to thank David Gorin for writing the custom applications used in this have been eliminated, but many will insist that they must research, and Dr. Leigh Callinan for statistical analysis of data. Ambient and be inferred in order to identify the shape. LED luminance values were measured by Drs. Ronald Henry and Andrew Jones. This research was supported, in part, by the Neuropsychology Foun- Previous research demonstrated that as few as 19 dots dation. allowed for recognition of the rooster by half of the sub- jects [2]. It was hypothesized that the individual dots serve References as markers of boundary positions, and the information 1. Kohler W: Dynamics in Psychology New York: Liveright Publishing; needed for encoding and storage of shapes might be based 1940:67-68. 2. Greene E: Recognition of objects that are displayed with on metric relationships among these markers. For the incomplete sets of discrete boundary dots. Percept Motor Skills image provided by a natural object, a great many more 2007, 104:1043-1059. 3. Hubel DH, Wiesel TN: Receptive fields of single neurons in the markers are activated by its contours. But here also, some cat's striate cortex. J Physiol 1959, 148:574-591. unspecified metric relationship among the markers may 4. Hubel DH, Wiesel TN: Receptive fields, binocular interactions provide the basis for recognition, rather than collinearity and functional architecture in the cat's visual cortex. J Physiol 1962, 160:106-154. of the markers. 5. Greene E: Simultaneity in the millisecond range as a require- ment for effective shape recognition. Behav Brain Funct 2006, 2:38. Conclusion 6. Greene E: Spatial and temporal proximity as factors in shape Contours provide a number of cues that might contribute recognition. Behav Brain Funct 2007, 3:27. to identifying a given shape. Investigators and theorists 7. Greene E: Information persistence in the integration of partial cues for object recognition. Percept Psychophys 2007, 69:772-784. have focused on a specific set of attributes that are pro- 8. Schall R: Estimation in generalized linear models with random vided by contours, in particular suggesting that orienta- effects. Biometrika 1991, 40:917-927. tion, curvature and linear extent serve to characterize and 9. Selfridge OG: Pattern recognition and learning. In Information theory Edited by: Cherry C. New York: Academic Press; specify the shape. This emphasis has been augmented by 1957:345-353. evidence that neurons in visual cortex respond more vig- 10. Sutherland NS: Outlines of a theory of visual pattern recogni- tion in animals and man. Proc R Soc Lond B Biol Sci 1968, orously at a particular orientation of the contour, with 171(24):95-103. response strength being a function of length, and in some 11. Hinton GE: A parallel computation that assigns canonical cases, curvature. The fact that these neurons also specify object-based frames of reference. In Proceedings of the Seventh International Joint Conference on Artificial Intelligence Los Altos, CA: location of a contour segment is given minimal attention. International Joint Conferences on Artificial Intelligence; It is conceivable that the locations that are registered by 1981:683-685. 12. Marr D: Vision: A Computational Investigation into the Human Represen- contour filters provide the information that is most essen- tation and Processing of Information New York: WH Freeman; tial for characterizing a given shape. 1982:51-79. 13. Quinlan PT: Differing approaches to two-dimensional shape recognition. Psychol Bull 1991, 109:224-241. Abbreviations 14. Palmer SE: Vision science: photons to phenomenology Cambridge, MA: MTDC: Minimal transient discrete cue; Arc': Minutes of MIT Press; 1999. visual angle; Cd/m : Candela per meter squared; arc°: 15. Sperling G: The information available in brief visual presenta- tions. Psychol Monogr 1960, 74:1-29. Degrees of visual angle; LED: Light emitting diode; Mhz: 16. Neisser U: Cognitive Psychology New York: Appleton-Century-Crofts; Megaherz; Log : Natural log; lux: Lumen per meter 17. Haber RN, Standing L: Direct measures of short-term visual squared; m: Meters; ms: Milliseconds; ns: Nanoseconds; storage. Quart J Exp Psychol 1969, 21:43-54. N: Number used to specify number of dots from address 18. Eriksen CW, Collins JF: Some temporal characteristics of visual list to be displayed; p: Probability; T1: Pulse width; T2: pattern perception. J Exp Psychol 1967, 74:476-484. Page 8 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 19. Eriksen CW, Collins JF: Sensory traces versus the psychological moment in the temporal organization of form. J Exp Psychol 1968, 77:376-380. 20. Coltheart M: Iconic memory and visible persistence. Percept Psychophys 1980, 27:183-228. 21. Long GM: Iconic memory: A review and critique of the study of short-term visual storage. Psychol Bull 1980, 88:785-820. 22. Nisly SJ, Wasserman GS: Intensity dependence of perceived duration: data, theories, and neural integration. Psychol Bull 1989, 106:483-496. 23. Sceniak MP, Hawken MJ, Shapley R: Visual spatial characteriza- tion of macaque V1 neurons. J Neurophys 2001, 85(5):1873-1887. 24. Jones JP, Palmer LA: The two-dimensional spatial structure of simple receptive fields in cat striate cortex. J Neurophys 1987, 58:1187-1211. 25. Greene E: Retinal encoding of ultrabrief shape recognition cues. PLoS ONE 2007, 2(9):e871. Publish with Bio Med Central and every scientist can read your work free of charge "BioMed Central will be the most significant development for disseminating the results of biomedical researc h in our lifetime." Sir Paul Nurse, Cancer Research UK Your research papers will be: available free of charge to the entire biomedical community peer reviewed and published immediately upon acceptance cited in PubMed and archived on PubMed Central yours — you keep the copyright BioMedcentral Submit your manuscript here: http://www.biomedcentral.com/info/publishing_adv.asp Page 9 of 9 (page number not for citation purposes) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Behavioral and Brain Functions Springer Journals

Additional evidence that contour attributes are not essential cues for object recognition

Greene, Ernest
Behavioral and Brain Functions , Volume 4 (1) – Jul 1, 2008
Download PDF
9 pages

Loading next page...
 
/lp/springer-journals/additional-evidence-that-contour-attributes-are-not-essential-cues-for-KAFAsASX5N

References (28)

Publisher
Springer Journals
Copyright
Copyright © 2008 by Greene; licensee BioMed Central Ltd.
Subject
Biomedicine; Neurosciences; Neurology; Behavioral Therapy; Psychiatry
eISSN
1744-9081
DOI
10.1186/1744-9081-4-26
pmid
18593469
Publisher site
See Article on Publisher Site

Abstract

It is believed that certain contour attributes, specifically orientation, curvature and linear extent, provide essential cues for object (shape) recognition. The present experiment examined this hypothesis by comparing stimulus conditions that differentially provided such cues. A spaced array of dots was used to mark the outside boundary of namable objects, and subsets were chosen that contained either contiguous strings of dots or randomly positioned dots. These subsets were briefly and successively displayed using an MTDC information persistence paradigm. Across the major range of temporal separation of the subsets, it was found that contiguity of boundary dots did not provide more effective shape recognition cues. This is at odds with the concept that encoding and recognition of shapes is predicated on the encoding of contour attributes such as orientation, curvature and linear extent. attributes provide the essential cues for object recognition. Background "Stationary visual percepts, a tree, a stone, or a book, are That study displayed dots that were positioned on the as a rule extremely reticent as to the nature of the neural outer boundary of namable objects, varied the number of events which underlie their existence. We may hope to dots that were displayed in progressively larger samples, learn more about brain correlates if we turn to instances and manipulating the spatial positioning of dots within in which percept processes seem to be in a more active those samples. The display dots were chosen to provide state." Kohler [1] subsets that either: a) formed contiguous strings that approximated line segments, b) were at randomly selected Cognitive, computational and neural theories of object positions around the boundary of the shape, or c) were at recognition all share the concept that essential cues are evenly spaced positions around the boundary. For each provided by the orientation, curvature and linear extent of condition the number of dots to be displayed (as a per- the lines and edges that lie at the boundary of an object centage of the total number of dots in the perimeter) was and its component parts. We can describe these cues as increased until the participant identified the object. The "contour attributes," and describe the mechanisms for greatest percentage of dots was required for recognition registering and encoding these attributes as "contour fil- when the subsets formed contiguous strings, and the ters." smallest percentage was needed when the dots in the sub- sets were positioned with even spacing. The contiguous A previous study from this laboratory [2] offered evidence strings would have delivered the most information about and arguments against the proposition that contour contour attributes, yet this treatment condition provided Page 1 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 the least effective cues for eliciting recognition of the respectively. The horizontal and vertical dimensions of shapes. the full array were 7.74 × 7.74 arc°. There were two additional reasons that the previous The circuits of the display board provided for activation of results [2] were at odds with the concept that contour a given LED by specifying an x, y address position within attributes provide essential cues for recognition. First, it is the array, under the control of a microprocessor running widely believed that orientation-selective cells in primary with a clock speed of 24 Mhz. Rise and fall time for emis- visual cortex serve as the basic filters that register the con- sion was in the range of 100 ns. Room illumination was tour attributes [3,4]. Yet recognition of many shapes was from standard ceiling-mounted fluorescent fixtures that possible with only a small sampling of dots, and with the were fitted with opaque panels to block most of the light. space between adjacent dots being wider than the recep- This provided ambient illumination of 13.3 lux. Lumi- tive fields of orientation-selective cells. Second, even if nance of an emitting LED was set at 7 Cd/m . one proposed new principles for registering alignments among these dots, it is not obvious how one would know The experiment displayed 64 shapes, these being the same which dots to connect. as listed in Table 1 of reference [2], but with the elimina- tion of five shapes in order to provide an equal number of The present experiments used the minimal transient dis- shapes for each treatment condition. Each shape was rep- crete cue (MTDC) protocol [5-7] to examine whether dot resented by discrete boundary dots that formed a con- subsets that should activate contour filters (because of nected string of adjacent positions within the 64 × 64 spatial contiguity) provide better shape cues than do non- array. Boundary dots for one shape, the frog, are illus- contiguous subsets. This protocol uses brief and succes- trated in the left panel of Fig. 1. sive display of subsets that were chosen from the full inventory of boundary dots. Successful recognition of the As implemented here, the MTDC protocol was as follows. shape required integration of the information provided by For a given shape, only some dots from the full inventory each subset, and the level of performance reflected the were shown, these being designated as the display set. The degree to which that information was useful. The prior size of the display set was determined on the basis of test- research [6,7] found that millisecond-level separation of ing done with other subjects. These tests established the subsets produced significant declines in recognition. This number of evenly spaced dots needed to provide a 75% was the case even when the total display time for a given hit rate, i.e., successful recognition of a given shape by shape – and thus duration of cue persistence – was con- trolled. This means that the effectiveness of the shape cues, and their ability to be integrated, are reflected in the rate at which recognition declines when time differentials are inserted between the successive cues. For the present experiment, the subsets provided either contiguous sequences of four dots, or four dots chosen at random locations in the boundary. Each contiguous dot subset would provide information about the local con- tour attributes of the shape. Each random dot subset would not provide this information, and to the extent that the subset might activate contour filters, would deliver The left fu the shapes Figure 1 ll inventory of panel uses u dots th nfilled at mar dots ked the boundary for on to show the positions of e of the The left panel uses unfilled dots to show the positions of the inappropriate cues regarding alignments of the boundary. full inventory of dots that marked the boundary for one of Therefore, to the extent that contour attributes are essen- the shapes. This full inventory was never displayed. Rather, a tial cues for shape recognition, performance levels should sampling of dots, designated as the display set, was shown, be higher for contiguous than for random subsets. The using a sample size that was expected to yield recognition on experimental results did not support this prediction. approximately 75% of the trials. An example of a display set is shown by the filled dots in the right panel. For a given dis- Methods play, a random starting point was picked from among the list The apparatus for display of shapes (display board) was of addresses, shown by the arrow, and then every Nth dot the same as used in earlier experiments [2,5,6]. Briefly location was selected for inclusion in the display set (with the outlined here, it consisted of a 64 × 64 array of red LEDs. value of N being that which would yield the number of dots The participant sat at a distance of 3.5 m from the display needed in the display set). The size of display dots has been exaggerated for purposes of illustration. board, and at this distance, the diameter of each element and the center-to-center spacings were 4.95 and 7.42 arc', Page 2 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 75% of the participants, if all the dots were simultane- ously displayed. This might be regarded as providing a given subject with a 75% probability of identifying the shape, and if only for convenience, it may be described in this manner in what follows. To specify the display set for each shape (done independ- ently for each participant), the first address to be included in the set was chosen at random. From there, counting in th a clockwise direction, every N address was chosen, the value of N being that which would yield the 75% hit rate. An example of one possible display set is illustrated in the right panel of Fig. 1. Each display set was further broken into randomly chosen subsets containing four dots each (with one residual sub- set potentially having fewer than four). Spatial position- ing of dots within the subsets provided the experimental treatment designated as "proximity," and temporal sepa- ration of subsets provided the treatment designated as "temporal separation," also known as T3. There were two levels of the proximity condition, requir- ing either that the dots of the subset be contiguous, or that they be randomly selected from among the members of the display set. Note that contiguity is relative, in that the th position within the full display set consists of every N inventory of boundary dot positions. For the random con- dition, there was an additional restriction that each dot in the subset must lie at least three steps away from other positions within the display set. The panels of Fig. 2 illustrate four subsets of contiguous dots sampled from the display set that is shown in Fig. 1. A given display tain were displayed successively Figu ing four dots each (plus any residual), re 2 set was broken at random into subsets con- and these subsets A given display set was broken at random into subsets con- The left panels show the location of each subset, superim- taining four dots each (plus any residual), and these subsets posed on the full complement of boundary dot positions. were displayed successively. This figure provides examples of The right panels show how each subset would appear contiguous subsets. The left panels show the location of sub- upon display. Each subset would be shown in rapid suc- set dots within the full inventory of dots. The right panels cession using one of the T3 intervals described below. show how each subset would appear on the display board, with the time interval for display of a given subset being 0.4 The panels of Fig. 3 provide a corresponding illustration ms. of randomly positioned subsets that again are members of the display set shown in Fig. 1. From a comparison of the right column of panels in Figs. 2 and 3 one can see the essence of the experimental concept. The contiguous-dot T3 specified the time interval between offset of one subset subsets reflect the local contour attributes of the frog, and onset of the next. There were four levels of T3, these whereas the random-dot subsets do not. being 1, 3, 9 and 27 ms. These time parameters are illus- trated in Fig. 4. Dots were displayed successively. Each dot was displayed for 0.1 ms, this being designated as T1. T2 specified the Total display time for a given shape was determined by interval from onset of one address within a subset to the the number of dots in the display set multiplied by 0.1 ms, next address of the same subset. This was also fixed at 0.1 plus the T3 interval multiplied by the number of subsets ms, providing for zero delay between offset of one address minus one. For T3 = 1 ms, the minimum and maximum and onset of the next. With these T1 and T2 intervals, all display times were 4.3 and 62 ms respectively, and the addresses within a given subset were displayed in 0.4 ms. mean display time was 16.6 ms. For T3 = 3 ms, minimum Page 3 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 first dot (21,43) T1 (pulse width) = 0.1 ms second dot (12,31) x,y T2 (pulse spacing) = 0.1 ms etc...(x,y) T3 (subset spacing) = 1, 3, 9 or 27 ms 45,18 29,33 21,43 12,31 22,57 15,59 41,40 35,11 The time intervals for displa Figure 4 y of shapes are illustrated The time intervals for display of shapes are illustrated. Fig. 1a shows the duration of emission from any given LED, desig- nated as T1, this being 0.1 ms. Fig. 1b specifies that the onset- to-onset interval of successive pulses within a given subset, designated as T2, which was also 0.1 ms. In Fig. 1c the pulses are illustrated as a string of beads. The four addresses of each subset are displayed as a group, separated by the T3 interval. For display of a given shape, theT3 interval was 1, 3, 9 or 27 ms. then shapes were assigned at random from the ranked list to the eight treatment combinations. The net effect of the assignment was to provide each treatment level with a sampling of shapes that were approximately equal in dif- ficulty. Each participant saw a given shape only once, and the order for display of the shapes (and thus the treatment combinations) was random. Th were ran Figure 3 is figure provides domly selected examples of subsets in which the dots Eight USC undergraduates served as participants, each dis- This figure provides examples of subsets in which the dots playing normal or corrected to normal visual acuity. Each were randomly selected. Again, the left panels show the loca- was naïve to the goals of the experiment, and was paid for tion of the subset dots within the full inventory of dots, and his or her participation. the right panels show how the dots in each subset would appear. Note that contour attributes, e.g., orientation, curva- Results ture and length, can be seen in the contiguous subsets shown The response variable was binary, i.e., recognize or failure in Fig. 2, but are not present in the randomly chosen subsets to recognize. Participants were treated as random samples shown here. from the population of possible participants. The order of presentation of shapes was randomly specified for each participant, and the treatment combination shown for a and maximum display times (rounded up) were 10 and given shape and participant was selected at random. Thus 150 ms, with a mean of 40 ms. For T3 = 9 ms, minimum the appropriate statistical model is a Generalized Linear and maximum display times were 28 and 414 ms, with a Mixed Model [8] with random effects of Participant and mean of 126 ms. For T3 = 27 ms, minimum and maxi- Shape, and fixed effects of Proximity and T3 interval. Logit mum display times were 82 and 1206 ms, with a mean of values (log (proportion/1-proportion) were calculated, 320 ms. and treatment differences were compared using the stand- ard error of the difference (SED) for these values. Model The two levels of dot proximity and four levels of T3 pro- predictions and standard errors of the mean for each of vided eight treatment combinations. For each participant these predictions are shown in Table 1. the inventory of 64 shapes were ranked for difficulty level, i.e., the number of dots required for a 75% hit rate, and Page 4 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 Table 1: Generalized Linear Mix Model Values for Treatment Conditions Contiguous Subsets Random Subsets T3 (ms) Mean SEM Backtransformed Mean SEM Backtransformed 1 1.056 0.369 0.742 1.435 0.382 0.808 3 0.879 0.363 0.707 0.537 0.359 0.631 9 0.213 0.354 0.553 0.284 0.354 0.570 27 -0.442 0.358 0.391 -1.425 0.386 0.194 The Generalized Linear Mixed Model transforms the percent recognition of shapes into logit values, i.e., log (proportion/1-proportion). The mean logit values and the standard errors of these means are given for each T3 interval for the contiguous and random subset data. The logit values have also been backtransformed into model predictions of recognition rate. This statistical analysis found no significant difference (p mark the outer boundary of the shape, the number of dots = 0.59) in recognition rate for the proximity condition, being just sufficient for recognition of the shape if all of i.e., recognition of shapes was not different as a function them are shown with minimal delay. By choosing which of whether the subset dots were contiguous or were at ran- dots to sample, and introducing delays between succes- domly selected positions. sive samples that are chosen, one can assess the effective- ness of the shape cues being provided by the samples. There was a significant (p < 0.001) linear decline in recog- nition rate as a function of the temporal separation The present goal was to examine whether contiguous sub- between subsets, and the quadratic component was not sets of dots would be more effective at eliciting recogni- significant (p = 0.22). The model predictions were back- tion of shapes than would subsets having an equal transformed into values that reflect the percentage of number of dots that were randomly chosen from the full shapes that were recognized for each of the treatment con- inventory of dots. The contiguous subsets should provide ditions. These predictions are very near the arithmetic mean recognition percentages that are plotted in Fig. 5 for contiguous and random subsets at each of the T3 inter- vals. 80 Although the difference between contiguous and random treatment conditions was not significant, inspection of the means plotted in Fig. 5 suggest the possibility that the treatments were not comparable at T3 = 27 ms. To for- mally evaluate this, pairwise comparisons of means were calculated, properly adjusting for the number of compar- isons. There were no significant differences at the first three T3 intervals, but the difference at T3 = 27 ms was sig- nificant at p < .02. This differential could be a simple experimental artifact, in that a treatment will not always yield data that fits the overall trends. Discussion 0.0 0.3 0.6 0.9 1.2 1.5 A great many, perhaps a majority, of shape recognition Temporal Separation of Subsets ( log milliseconds) theories propose that contour attributes, i.e., orientation, curvature and linear extent, provide the elemental features that define the shape of an object. Selfridge [9] may have Mean percent rec tion set Figure 5 s is plotted against the time ognition for each of th interval separating each su e proximity condi-b- been the first to characterize the perceptual process in Mean percent recognition for each of the proximity condi- terms of an assemblage of filters, each having the ability tions is plotted against the time interval separating each sub- to register a distinctive contour attribute, but many others set. Contiguous subsets are shown with filled circles, and have followed this lead [see [10-14]]. random subsets are shown with open circles. The decline in recognition was significant across the tested time intervals, The minimal transient discrete cue (MTDC) protocol [5-7] but the proximity conditions did not produce differential lev- provides a means to evaluate the validity of this hypothe- els of shape recognition. sis. This method briefly displays a spaced array of dots that Page 5 of 9 (page number not for citation purposes) Percent Shape Recognition Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 a more effective stimulus for the filters that are presumed assemblage of these contour filters delivers the full com- to register the contour attributes. If shapes are specified on plement of contour attributes needed for recognition. the basis of their contour attributes, then the contiguous subsets should convey the best partial shape cues, and one However, previous results from this laboratory [2] raise would expect these subsets to be more effective for elicit- the question of whether shape analysis depends on activa- ing recognition. tion of orientation-selective cells. That study found that recognition was possible when the full complement of The overall result was that contiguous and randomly dots being shown was relatively sparse. Recognition was selected subsets contributed equally to shape recognition, well above chance when dot spans exceeded the length of even though the randomly selected subsets did not dis- orientation-selective receptive fields [23]. That outcome play cues that relate to the orientation, curvature and lin- suggests that each dot is acting as an independent marker ear extent of the boundary. This indicates that under the of boundary position, and that shape is defined by an present test conditions, contour attributes did provide unspecified – not yet known – relationship among the cues that are essential for shape perception. individual markers. Even when the orientation-selective cells are activated by an array of dots, the essential infor- For the present task conditions, one might speculate that mation might be the locations that have been specified information persistence allowed successive dots to accu- rather than the collinearity in the array. mulate, such that dots from the random subsets could eventually form contiguous strings that provided contour With respect to the present results, one might wonder attributes. There is persistence of brief visual stimuli, as whether the contiguous subsets were effective stimuli for reported by Sperling [15], Neisser [16], Haber and Stand- the orientation-selective cells. Perhaps the cells did not ing [17], and Eriksen and Collins [18,19], among others, respond to the very brief presentation of just four dots. and reviewed by Coltheart [20], Long [21], and Nisly and There are three reasons to suggest that the subsets deliv- Wasserman [22]. Whereas local contour information was ered adequate stimulation. not provided by a given random subset, one could argue that the contour-filtering process simply waited for a First, although the stimulus duration was very brief, the number of the subsets to be delivered, after which the con- flashes were easily visible, i.e., consciously perceived. It is tour attributes could be extracted from the aggregate pool generally accepted that conscious awareness of a visual of dots. stimulus requires processing by the primary visual cortex, thus the stimulus strength was adequate for activating its Recent work using the present experimental protocols, neurons. however, has found that millisecond and even submilli- second differentials in the display of dot subsets can pro- Second, the span of each contiguous subset was a suitable duce significant differences in shape recognition [6,7]. fit to the size of receptive fields. Sceniak et al. [23] exam- The result that is most critical to this discussion was pro- ined receptive field size of orientation-selective cells in V1 vided by the second experiment in each of the cited stud- of Macaque, and found the average space constant to be ies, wherein the total time (and thus duration of 60 arc', and the average length-summation tuning curve to persistence) for a given shape was held constant. Under be 49 arc'. The four-dot array of the contiguous subsets these conditions, it was found that varying the interval spanned 35 arc' for horizontal or vertical alignments, and between successive dots impaired recognition, with tem- 47 arc' for diagonal alignments. Therefore each of the con- poral separation of as little as half a millisecond being sig- tiguous subsets displayed an image size that would pro- nificant. Shape-relevant contour attributes are delivered vide four dots to the receptive fields. directly by the contiguous dot subsets, but they could be provided by random subsets only through aggregation. Third, there is direct electrophysiological evidence that an The prior studies demonstrate that the cues do not aggre- array of briefly flashed dots will stimulate the cortical gate without a recognition penalty. cells. Jones & Palmer [24] examined responsiveness of ori- entation-selective cells with successive stimulation of When neural substrates for shape perception are dis- local points across the receptive fields, the typical dura- cussed, most see the orientation-selective cells character- tion of each stimulus being 50 ms. They reported that the ized by Hubel & Wiesel [3,4] as providing the first step for responses that could be elicited by stimulating one loca- registering contour attributes. A given cell can be activated tion at a time was too weak to be of practical value in the by a contour, and because the firing rate is influenced by analysis of receptive field structure. However, simultane- the orientation, length, and (possibly) curvature of the ous activation of three sites within the receptive field contour, the response is thought to convey information yielded usable data. As indicated above, the contiguous about these attributes. It is further suggested that an subsets of the present experiment displayed four dots that Page 6 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 would register on a given receptive field, and this would put a premium on very tight temporal proximity within a provide a stronger stimulus than was found to be effective stimulus pattern. More advanced image-processing sys- by Jones & Palmer [24]. tems, such as primary visual cortex, might have similar requirements for simultaneity, but with a longer time con- The more general point is that the random subsets as well stant. This could explain the differential at T3 = 27 ms as as the contiguous subsets were seen by the subject and a contribution to the temporal-integration process by ori- delivered sufficient stimulation to elicit recognition. If entation-selective cells that could not be accomplished in one took the position that the contiguous arrays provided the retina. an insufficient stimulus for activating orientation-selec- tive cells, it would mean that recognition was accom- The finding that the contour attributes did not benefit rec- plished without any contribution from these cells. ognition under the present test conditions should not be taken as a blanket rejection of a useful role in the percep- It is possible, that the cues used for this experiment may tion of objects. The fact that we can detect edges with a be especially salient for activating a primitive shape contrast differential as small as 3% speaks to the benefit of encoding system. The pattern provided by the full com- these filters for registering the presence of a boundary. plement of dots is very similar to a silhouette, and recog- Doubtless this is useful for detecting an object that is nition is best when there is maximal simultaneity of the almost the same color or luminance as the background, or flashed dots. This is not unlike conditions that might face where it must be seen through haze. Contour filters may an early vertebrate – perhaps a fish – who detects simulta- make it possible to see the object's boundaries under a neous movement through small openings in a wall of sea- variety of degraded conditions, and there is ample evi- weed. The pattern that is seen could be a predator, or dence that alignment of lines and edges provides a basis might be prey, and successful recognition by the creature for object completion. It is possible, however, that this would have implications for survival. It is likely that these processing allows the position of discrete markers to be recognition skills evolved, and are present in a great many specified. Shape perception, per se, may then be based on present-day animals that have no cortex. metric relationships that have little or nothing to do with collinearity of the markers. Recent evidence from this laboratory [25], gathered and published after the present research was conducted, has It is unclear why so many insist that shape is defined by demonstrated that the retina contains a neural system that the orientation, curvature and linear extent of the con- is sensitive to millisecond-level simultaneity when the tours. We know that all manner of cues can contribute to subsets consist of dot pairs. This suggests that the present identification of objects, but have no trouble discarding task draws on primitive shape-encoding mechanisms that most of them as being ancillary. Fig. 6 illustrates the situ- detailed boundary detailed boundary boundary markers internal contours texture color S Figure 6 timuli that are available as shape cues are listed above each image Stimuli that are available as shape cues are listed above each image. The right image provides the number of dots in a display set that allows for 75% successful recognition of the rooster when the dots are displayed successively, each being shown for 0.1 ms, and with a T3 interval of zero. Page 7 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 ation. The left image shows a detailed colored sketch that Temporal separation between members of subset pairs; can be readily identified as a rooster. In fact the image is T3: Temporal separation between subset pairs; V1: Pri- devoid of various depth cues that would be present in the mary visual cortex; 2D: Two dimensional; SEM: Standard real object. Nonetheless, we accept that the 2D image has error of the mean. the shape of a rooster, so the depth cues must be ancillary to our concept of shape. Competing interests The author declares that he has no competing interests. The middle image has eliminated internal contours, tex- ture, and color, replacing all these cues with uniform Authors' contributions black. Yet this silhouette is readily identified as being in EG conceived of the study, designed the study, tested all the shape of a rooster. The internal parts, color and texture participants, and wrote the article. Technical assistance for must be at least somewhat ancillary, i.e., nonessential. programming and data analysis was provided by contract, as noted below. EG has read and approves of the final The right image has replaced the boundary edge with an manuscript. array of dots, and we can still see the stimulus as having the shape of a rooster. Contour attributes of the boundary Acknowledgements I wish to thank David Gorin for writing the custom applications used in this have been eliminated, but many will insist that they must research, and Dr. Leigh Callinan for statistical analysis of data. Ambient and be inferred in order to identify the shape. LED luminance values were measured by Drs. Ronald Henry and Andrew Jones. This research was supported, in part, by the Neuropsychology Foun- Previous research demonstrated that as few as 19 dots dation. allowed for recognition of the rooster by half of the sub- jects [2]. It was hypothesized that the individual dots serve References as markers of boundary positions, and the information 1. Kohler W: Dynamics in Psychology New York: Liveright Publishing; needed for encoding and storage of shapes might be based 1940:67-68. 2. Greene E: Recognition of objects that are displayed with on metric relationships among these markers. For the incomplete sets of discrete boundary dots. Percept Motor Skills image provided by a natural object, a great many more 2007, 104:1043-1059. 3. Hubel DH, Wiesel TN: Receptive fields of single neurons in the markers are activated by its contours. But here also, some cat's striate cortex. J Physiol 1959, 148:574-591. unspecified metric relationship among the markers may 4. Hubel DH, Wiesel TN: Receptive fields, binocular interactions provide the basis for recognition, rather than collinearity and functional architecture in the cat's visual cortex. J Physiol 1962, 160:106-154. of the markers. 5. Greene E: Simultaneity in the millisecond range as a require- ment for effective shape recognition. Behav Brain Funct 2006, 2:38. Conclusion 6. Greene E: Spatial and temporal proximity as factors in shape Contours provide a number of cues that might contribute recognition. Behav Brain Funct 2007, 3:27. to identifying a given shape. Investigators and theorists 7. Greene E: Information persistence in the integration of partial cues for object recognition. Percept Psychophys 2007, 69:772-784. have focused on a specific set of attributes that are pro- 8. Schall R: Estimation in generalized linear models with random vided by contours, in particular suggesting that orienta- effects. Biometrika 1991, 40:917-927. tion, curvature and linear extent serve to characterize and 9. Selfridge OG: Pattern recognition and learning. In Information theory Edited by: Cherry C. New York: Academic Press; specify the shape. This emphasis has been augmented by 1957:345-353. evidence that neurons in visual cortex respond more vig- 10. Sutherland NS: Outlines of a theory of visual pattern recogni- tion in animals and man. Proc R Soc Lond B Biol Sci 1968, orously at a particular orientation of the contour, with 171(24):95-103. response strength being a function of length, and in some 11. Hinton GE: A parallel computation that assigns canonical cases, curvature. The fact that these neurons also specify object-based frames of reference. In Proceedings of the Seventh International Joint Conference on Artificial Intelligence Los Altos, CA: location of a contour segment is given minimal attention. International Joint Conferences on Artificial Intelligence; It is conceivable that the locations that are registered by 1981:683-685. 12. Marr D: Vision: A Computational Investigation into the Human Represen- contour filters provide the information that is most essen- tation and Processing of Information New York: WH Freeman; tial for characterizing a given shape. 1982:51-79. 13. Quinlan PT: Differing approaches to two-dimensional shape recognition. Psychol Bull 1991, 109:224-241. Abbreviations 14. Palmer SE: Vision science: photons to phenomenology Cambridge, MA: MTDC: Minimal transient discrete cue; Arc': Minutes of MIT Press; 1999. visual angle; Cd/m : Candela per meter squared; arc°: 15. Sperling G: The information available in brief visual presenta- tions. Psychol Monogr 1960, 74:1-29. Degrees of visual angle; LED: Light emitting diode; Mhz: 16. Neisser U: Cognitive Psychology New York: Appleton-Century-Crofts; Megaherz; Log : Natural log; lux: Lumen per meter 17. Haber RN, Standing L: Direct measures of short-term visual squared; m: Meters; ms: Milliseconds; ns: Nanoseconds; storage. Quart J Exp Psychol 1969, 21:43-54. N: Number used to specify number of dots from address 18. Eriksen CW, Collins JF: Some temporal characteristics of visual list to be displayed; p: Probability; T1: Pulse width; T2: pattern perception. J Exp Psychol 1967, 74:476-484. Page 8 of 9 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:26 http://www.behavioralandbrainfunctions.com/content/4/1/26 19. Eriksen CW, Collins JF: Sensory traces versus the psychological moment in the temporal organization of form. J Exp Psychol 1968, 77:376-380. 20. Coltheart M: Iconic memory and visible persistence. Percept Psychophys 1980, 27:183-228. 21. Long GM: Iconic memory: A review and critique of the study of short-term visual storage. Psychol Bull 1980, 88:785-820. 22. Nisly SJ, Wasserman GS: Intensity dependence of perceived duration: data, theories, and neural integration. Psychol Bull 1989, 106:483-496. 23. Sceniak MP, Hawken MJ, Shapley R: Visual spatial characteriza- tion of macaque V1 neurons. J Neurophys 2001, 85(5):1873-1887. 24. Jones JP, Palmer LA: The two-dimensional spatial structure of simple receptive fields in cat striate cortex. J Neurophys 1987, 58:1187-1211. 25. Greene E: Retinal encoding of ultrabrief shape recognition cues. PLoS ONE 2007, 2(9):e871. Publish with Bio Med Central and every scientist can read your work free of charge "BioMed Central will be the most significant development for disseminating the results of biomedical researc h in our lifetime." Sir Paul Nurse, Cancer Research UK Your research papers will be: available free of charge to the entire biomedical community peer reviewed and published immediately upon acceptance cited in PubMed and archived on PubMed Central yours — you keep the copyright BioMedcentral Submit your manuscript here: http://www.biomedcentral.com/info/publishing_adv.asp Page 9 of 9 (page number not for citation purposes)

Journal

Behavioral and Brain Functions – Springer Journals

Published: Jul 1, 2008

There are no references for this article.