Perceptual and Motor skill^, 1975,41, 167-170. @Perceptual and Motor Skills 1975
ICONIC MEMORY OF DOT PATTERNS: PRELIMINARY REPORT1 LEE S. COHENE
Surnmry.-Pairs stimulus pair forms a variables of stimulus field were all shown srudents.
of complementary dot-pattern stimuli whereby each bigram were used in an iconic memory paradigm. The duration, interstimulus interval, dot density, and letter to affect recognition performance of 15 naive college
A relatively new and interesting technique for studying visual pattern perception involves the use of pairs of complementary dot-pattern stimuli. Dots are arranged in a configuration so as to form a recognizable pattern of letters or numbers (usually one co three elements in length). These dots are then equally and randomly distributed over two stimulus halves (S1 and S2) with addicional dots similarly distribuced for camouflage. Since little information concerning the nature of the composite form is available from either stimulus half alone, this method affords a unique opportunity to study the processing of visual pacterns while avoiding many of the complicating interactions that often arise with the use of strictly alpha-numeric materials. These stimuli have been used to study such visual phenomena as integration or summation (Eriksen & Collins, 1967, 1968; Hastings, 1972), eidetic imagery (Stromeyer & Psotka, 1970), masking (Cohene & Bechtoldt, 1974a, 1974b), and processing time (Cohene & Bechtoldt, in press). It appears that these dot-pattern stimuli provide a suitable method for studying iconic memory or short-term visual storage, simply by manipulating the interstimulus interval between S1 and S2 and, at the same time, this method is not subject to many criticisms (Dick, 1971; Holding, 1975) that have been leveled at Sperling's (1960) partial report technique. The present experiment represents a preliminary attempt to analyze the role of stimulus duration, interstimulus interval, dot density, and letter field in an iconic memory paradigm using dot-pattern bigrams. METHOD The subjects were 15 students at the University of Iowa who had 20-20 vision. The dot-pattern bigrarns were the same 10 stimuli that we have been using in previous experiments (Cohene & Bechtoldt, 1974b. in press). A projection apparatus was used to transmit these stimuli through a screen to the subject. 'This research was reported at the annual meeting of the Midwestern Psychological Association in Chicago, May, 1975. It was supported, in part, by a Grant-in-Aid of Research from Sigma Xi, The Scientific Research Sociery of North America and by the Graduate College of the University of Iowa. The ex~erimentwas conducted while the author held a Hiaher Education ~ e l l o w s h ifrom ~ the province of Quebec, Canada. 1 wish to express myWslocere appreciation to Harold Bechtoldt for his thoughtful comments. Requests for reprints may be sent to Lee S. Cohene, Department of Psychology, University of Toronto, Toronto M5S 1A1, Ontario, Canada.
168
L. S. COHENE
Each test trial consisted of the presentation of a 1.5-sec. fixation point followed immediately by S1 and S2 with some interstimulus interval between the stimulus halves. A forced-choice recognition procedure was employed, based upon a set of the ten possible letters, which was made available to the subject. The experimental design was a four-factor mixed arrangement. T h e effect of interstimulus interval was examined at the three different levels of 0, 20, and 40 msec., with five subjects assigned to each level. The remaining three variables were manipulated 50, and orthogonally within subjects. There were three levels of stimulus duration-25, 75 msec.-with the duration of S 1 and S2 always being equal. A 25-msec. condition with simultaneous onset and simultaneous offset of the dot patterns was also included, a t the end of practice, to obtain baseline data for comparison purposes. T o examine dot density, the composite consisted of either 75 dots o r 150 dots with the dots i n the more dense patterns being one-half in area to maintain equal brightness. Finally, responses for the bigcam stimuli were recorded according to letter position, i.e., left and right fields. The 10 stimuli were randomly presented rwice in each condition. A practice session in which the same stimuli were presented with simultaneous onset 2nd offset of S1 and S2 immediately preceded the test trials.
RESULTS AND DISCUSSION The CL-leveladopted for this experiment was p < .01. An analysis of variance of the factorial design of interstimulus interval by duration by dot density by letter field yielded significant main effects for the variables of inter= 16.00, p < .Ol), stimulus duration (P2,24 = 132.33, stimulus interval (F2,12 p < .01), and letter field (Fl,lo= 10.53, p < .01). There were also several interactions. The only significant interaction involving letter position was that of dot density by field (F1,12 = 13.34, p < .01). The mean recognition scores for the patterns containing 75 dots were 37.0 and 19.6 for the left and right positions, respectively, and for the patterns containing 150 dots, the left- and right-letter means were 28.8 and 22.2 respectively (maximum score equals 90). This interaction therefore arose as a result of the difference in the mean differences, both of which favored the left-letter position. The left-field superiority effect has been observed by us in previous experiments (Cohene & Bechtoldt, 1974b, in press) and presumably is a function of a set process and/or reading habits. Of major interest were the significant interactions t h a t were observed for interstimulus interval by duration (F4,24= 132.22, p < .01), duration by dot density (F1,34 = 12.61, p < .01), and interstimulus interval by duration by dot density (F4,PQ= 4.50, p < .01). These interactions are represented graphically in Fig. 1. The interaction of duration by interstimulus interval may be due largely to a convergence on a low performance level (approaching a "floor" effect) at the 75-rnsec. duration with a 40-msec. interstimulus interval. There was also the interaction of duration by dot density whereby the superiority in recognition of the 150-dot patterns over the 75-dot patterns at 25 msec. was reversed at 50
ICONlC MEMORY OF DOT PATTERNS
A......,A
40
150
&...
40
75
...,.
\
FIG. 1. Mean letter recognition ,as a
function of stimulus duration, Interinterval (ISI). . . . and dot densiw ( N ) . The concurrent conditions are represented as 25(c). The maximum possible value for each data point is 20; expected score due to guessing is 2.1. scimulus
I
50 Duration L msec.)
I
75
and 75 msec. Better recognition with the 150-dot stimuli has been observed before by us (Cohene & Bechtoldt, in press) and is obvious in the 25-msec. concurrent conditions of the present study; the interaction with duration may be attributed to a reduction in contour information that is available to a subject under the less ideal conditions of longer durations. The dot-density-by-duration crossover pattern was produced in both the 0- and 20-msec. interstimulus interval conditions. However, in the 40-msec. interstimulus interval condition the 75-dot stimuli resulted in better recognition than the 150-dot stimuli at the three stimulus durations of 25, 50, and 75 msec. This exception at the 40-msec. interstimulus interval created the interaction of duration by interstimulus interval by dot density. It therefore appears that the small amount of iconic memory of the dotpattern bigrams which existed at the short stimulus durations with zero interstimulus interval was greatly impaired by relatively small increases in stimulus duration and interstimulus interval. The decrease in performance with increases in interstimulus interval may be attributed to a decaying trace or icon. The effect of poorer performance with an increase in duration has been observed once before (Pollack, 1973). The reduced performance could possibly be due, in part, to luminance summation which can lead to reduced contrast (e.g., Eriksen & Hoffman, 1963), provided the energy of the traces of the stimulus halves was increasing over the 25- to 75-msec. range. Nevertheless, because the luminance of S1 and S2 was very near to that of the adapting field and because the luminance of S1 and S2 together was not very different from either stimulus half alone, the amount of masking by luminance summation would appear to have been negligible. In conclusion, the variety of interactions that were seen in this study points
170
L. S. COHENE
to the advantage of using dot-pattern stimuli with the iconic memory paradigm and to the need for further investigation of this and other visual phenomena such as eidetic imagery and masking with similarly conuolled patterns. REFERENCES
COHENE,L. S., & BECHTOLDT,H. P. Some temporal factors in visual pattern recognition 11. Proceedings o f the Iowa Academy o f Science, 1974, 81, 111-115. ( a ) COHENE,L. S., & BECHTOLDT,H. P. Visual recognition as a function of stimulus offset s, 15, 221-226. (b) asynchrony and duration. Perception and ~ s ~ c h o p h y s i c1974, COHENE,L. S., & BECHTOLDT,H. P. Visual recognition of dot pattern bigrams: replication and extension. American Journal of Psychology, in press. DICK,A. 0 . On the problem of selection in short-term visual (iconic) memory. Canadian Journal o f Psychology, 1971, 25, 250-263. ERIKSEN, C. W . , & COLLINS,J. F. Some temporal characteristics of visual pattern perception. Journal o f Experimental Psychology, 1967, 74, 476-484. ERIKSEN, C. W., 8r COLLINS,J. F. Sensory traces versus the psychological moment in the temporal organization of form. ]ourma1 of Experimental Psychology, 1968, 77, 376-382. ERIKSEN, C. W., & HOFFMAN, M. Form recognition at brief durations as a function of ada ring field and interval between stimulations. Journal of Experimental Psychofogy, 1963, 66, 485-499. HASTINGS, M. W. Temporal summation of visual form information as a function of interstimulus-interval, retinal location, and stimulus orientation. (Doctoral dissertation, Univer. of Maine, 1972) Dissertation Abstracts I n t e r n a t i o ~ l , 1973, 33, 6108B-6109B. (Univer. Microfilms No. 73-13, 068) HOLDING, D. H. Sensory storage reconsidered. Memory and Cognition, 1975, 3, 31-41. POLLACK, 1. Interaction effects in successive visual displays: an extension of the EriksenCollins paradigm. Perception and Psychophysics, 1973, 13, 367-373. SPERLING,G. T h e information available in brief visual presentations. Psychological Mono~raphs,1960, 74, No. 11 ( W h o l e No. 498). STROMEYER, C. G., & PSOTKA, J. The detailed texture of eidetic images. Nature, 1970,
225, 346-349.
Accepted May 7,1975.
ICONIC MEMORY OF DOT PATTERNS: PRELIMINARY REPORT1 LEE S. COHENE
Surnmry.-Pairs stimulus pair forms a variables of stimulus field were all shown srudents.
of complementary dot-pattern stimuli whereby each bigram were used in an iconic memory paradigm. The duration, interstimulus interval, dot density, and letter to affect recognition performance of 15 naive college
A relatively new and interesting technique for studying visual pattern perception involves the use of pairs of complementary dot-pattern stimuli. Dots are arranged in a configuration so as to form a recognizable pattern of letters or numbers (usually one co three elements in length). These dots are then equally and randomly distributed over two stimulus halves (S1 and S2) with addicional dots similarly distribuced for camouflage. Since little information concerning the nature of the composite form is available from either stimulus half alone, this method affords a unique opportunity to study the processing of visual pacterns while avoiding many of the complicating interactions that often arise with the use of strictly alpha-numeric materials. These stimuli have been used to study such visual phenomena as integration or summation (Eriksen & Collins, 1967, 1968; Hastings, 1972), eidetic imagery (Stromeyer & Psotka, 1970), masking (Cohene & Bechtoldt, 1974a, 1974b), and processing time (Cohene & Bechtoldt, in press). It appears that these dot-pattern stimuli provide a suitable method for studying iconic memory or short-term visual storage, simply by manipulating the interstimulus interval between S1 and S2 and, at the same time, this method is not subject to many criticisms (Dick, 1971; Holding, 1975) that have been leveled at Sperling's (1960) partial report technique. The present experiment represents a preliminary attempt to analyze the role of stimulus duration, interstimulus interval, dot density, and letter field in an iconic memory paradigm using dot-pattern bigrams. METHOD The subjects were 15 students at the University of Iowa who had 20-20 vision. The dot-pattern bigrarns were the same 10 stimuli that we have been using in previous experiments (Cohene & Bechtoldt, 1974b. in press). A projection apparatus was used to transmit these stimuli through a screen to the subject. 'This research was reported at the annual meeting of the Midwestern Psychological Association in Chicago, May, 1975. It was supported, in part, by a Grant-in-Aid of Research from Sigma Xi, The Scientific Research Sociery of North America and by the Graduate College of the University of Iowa. The ex~erimentwas conducted while the author held a Hiaher Education ~ e l l o w s h ifrom ~ the province of Quebec, Canada. 1 wish to express myWslocere appreciation to Harold Bechtoldt for his thoughtful comments. Requests for reprints may be sent to Lee S. Cohene, Department of Psychology, University of Toronto, Toronto M5S 1A1, Ontario, Canada.
168
L. S. COHENE
Each test trial consisted of the presentation of a 1.5-sec. fixation point followed immediately by S1 and S2 with some interstimulus interval between the stimulus halves. A forced-choice recognition procedure was employed, based upon a set of the ten possible letters, which was made available to the subject. The experimental design was a four-factor mixed arrangement. T h e effect of interstimulus interval was examined at the three different levels of 0, 20, and 40 msec., with five subjects assigned to each level. The remaining three variables were manipulated 50, and orthogonally within subjects. There were three levels of stimulus duration-25, 75 msec.-with the duration of S 1 and S2 always being equal. A 25-msec. condition with simultaneous onset and simultaneous offset of the dot patterns was also included, a t the end of practice, to obtain baseline data for comparison purposes. T o examine dot density, the composite consisted of either 75 dots o r 150 dots with the dots i n the more dense patterns being one-half in area to maintain equal brightness. Finally, responses for the bigcam stimuli were recorded according to letter position, i.e., left and right fields. The 10 stimuli were randomly presented rwice in each condition. A practice session in which the same stimuli were presented with simultaneous onset 2nd offset of S1 and S2 immediately preceded the test trials.
RESULTS AND DISCUSSION The CL-leveladopted for this experiment was p < .01. An analysis of variance of the factorial design of interstimulus interval by duration by dot density by letter field yielded significant main effects for the variables of inter= 16.00, p < .Ol), stimulus duration (P2,24 = 132.33, stimulus interval (F2,12 p < .01), and letter field (Fl,lo= 10.53, p < .01). There were also several interactions. The only significant interaction involving letter position was that of dot density by field (F1,12 = 13.34, p < .01). The mean recognition scores for the patterns containing 75 dots were 37.0 and 19.6 for the left and right positions, respectively, and for the patterns containing 150 dots, the left- and right-letter means were 28.8 and 22.2 respectively (maximum score equals 90). This interaction therefore arose as a result of the difference in the mean differences, both of which favored the left-letter position. The left-field superiority effect has been observed by us in previous experiments (Cohene & Bechtoldt, 1974b, in press) and presumably is a function of a set process and/or reading habits. Of major interest were the significant interactions t h a t were observed for interstimulus interval by duration (F4,24= 132.22, p < .01), duration by dot density (F1,34 = 12.61, p < .01), and interstimulus interval by duration by dot density (F4,PQ= 4.50, p < .01). These interactions are represented graphically in Fig. 1. The interaction of duration by interstimulus interval may be due largely to a convergence on a low performance level (approaching a "floor" effect) at the 75-rnsec. duration with a 40-msec. interstimulus interval. There was also the interaction of duration by dot density whereby the superiority in recognition of the 150-dot patterns over the 75-dot patterns at 25 msec. was reversed at 50
ICONlC MEMORY OF DOT PATTERNS
A......,A
40
150
&...
40
75
...,.
\
FIG. 1. Mean letter recognition ,as a
function of stimulus duration, Interinterval (ISI). . . . and dot densiw ( N ) . The concurrent conditions are represented as 25(c). The maximum possible value for each data point is 20; expected score due to guessing is 2.1. scimulus
I
50 Duration L msec.)
I
75
and 75 msec. Better recognition with the 150-dot stimuli has been observed before by us (Cohene & Bechtoldt, in press) and is obvious in the 25-msec. concurrent conditions of the present study; the interaction with duration may be attributed to a reduction in contour information that is available to a subject under the less ideal conditions of longer durations. The dot-density-by-duration crossover pattern was produced in both the 0- and 20-msec. interstimulus interval conditions. However, in the 40-msec. interstimulus interval condition the 75-dot stimuli resulted in better recognition than the 150-dot stimuli at the three stimulus durations of 25, 50, and 75 msec. This exception at the 40-msec. interstimulus interval created the interaction of duration by interstimulus interval by dot density. It therefore appears that the small amount of iconic memory of the dotpattern bigrams which existed at the short stimulus durations with zero interstimulus interval was greatly impaired by relatively small increases in stimulus duration and interstimulus interval. The decrease in performance with increases in interstimulus interval may be attributed to a decaying trace or icon. The effect of poorer performance with an increase in duration has been observed once before (Pollack, 1973). The reduced performance could possibly be due, in part, to luminance summation which can lead to reduced contrast (e.g., Eriksen & Hoffman, 1963), provided the energy of the traces of the stimulus halves was increasing over the 25- to 75-msec. range. Nevertheless, because the luminance of S1 and S2 was very near to that of the adapting field and because the luminance of S1 and S2 together was not very different from either stimulus half alone, the amount of masking by luminance summation would appear to have been negligible. In conclusion, the variety of interactions that were seen in this study points
170
L. S. COHENE
to the advantage of using dot-pattern stimuli with the iconic memory paradigm and to the need for further investigation of this and other visual phenomena such as eidetic imagery and masking with similarly conuolled patterns. REFERENCES
COHENE,L. S., & BECHTOLDT,H. P. Some temporal factors in visual pattern recognition 11. Proceedings o f the Iowa Academy o f Science, 1974, 81, 111-115. ( a ) COHENE,L. S., & BECHTOLDT,H. P. Visual recognition as a function of stimulus offset s, 15, 221-226. (b) asynchrony and duration. Perception and ~ s ~ c h o p h y s i c1974, COHENE,L. S., & BECHTOLDT,H. P. Visual recognition of dot pattern bigrams: replication and extension. American Journal of Psychology, in press. DICK,A. 0 . On the problem of selection in short-term visual (iconic) memory. Canadian Journal o f Psychology, 1971, 25, 250-263. ERIKSEN, C. W . , & COLLINS,J. F. Some temporal characteristics of visual pattern perception. Journal o f Experimental Psychology, 1967, 74, 476-484. ERIKSEN, C. W., 8r COLLINS,J. F. Sensory traces versus the psychological moment in the temporal organization of form. ]ourma1 of Experimental Psychology, 1968, 77, 376-382. ERIKSEN, C. W., & HOFFMAN, M. Form recognition at brief durations as a function of ada ring field and interval between stimulations. Journal of Experimental Psychofogy, 1963, 66, 485-499. HASTINGS, M. W. Temporal summation of visual form information as a function of interstimulus-interval, retinal location, and stimulus orientation. (Doctoral dissertation, Univer. of Maine, 1972) Dissertation Abstracts I n t e r n a t i o ~ l , 1973, 33, 6108B-6109B. (Univer. Microfilms No. 73-13, 068) HOLDING, D. H. Sensory storage reconsidered. Memory and Cognition, 1975, 3, 31-41. POLLACK, 1. Interaction effects in successive visual displays: an extension of the EriksenCollins paradigm. Perception and Psychophysics, 1973, 13, 367-373. SPERLING,G. T h e information available in brief visual presentations. Psychological Mono~raphs,1960, 74, No. 11 ( W h o l e No. 498). STROMEYER, C. G., & PSOTKA, J. The detailed texture of eidetic images. Nature, 1970,
225, 346-349.
Accepted May 7,1975.
