Sensory memory is the initial structure in the Atkinson and Shiffrin’s modal model of memory. This memory structure temporarily stores information from sensory stimulation for processing and transferring to short-term memory. Within fractions of seconds, this memory will degenerate if remain unattended (Shiffrin & Atkinson, 1969). There are two kinds of sensory memory which is iconic memory and echoic memory. Iconic memory is crucial as it stabilizes vision despite the presence of saccadic movements which disturbs visual adaptation to stimulus (Ciccarelli & White, 2012). Hence, any delay in between the sensory stimulation and the recollection will cause an effect to the iconic memory. The question is; how much are we able to recall from our iconic memory after a period of delay?
In 1967, Neisser coined the term iconic memory and assumed that all visual information are initially held there before being transported to longer lasting memory upon cue (Gegenfurtner & Sperling, 1993). Iconic memory involves vision persistence where trails of light retained for fragments of second in memory (Goldstein, 2011). Haber (1983) described iconic memory as the availability to perceive the information briefly after terminating visual stimuli. The persistence and decay of the visual information in iconic memory is first presented in Sperling’s partial report (Sperling, 1960).
In partial report paradigm, there is a brief delay prior to the selection cue in reporting the stimuli. The brief delay or known as the interstimulus interval (ISI) which is a time period in between two stimuli (Reed, 2013). ISI is the interval between the end of the visual stimulus and the onset of the cue tone before reporting the displayed information in the partial report technique (Chow, 1985).
The effect of ISI on iconic memory in partial report can be explained by Sperling’s theory of iconic memory. Sperling stated that observers have the ability to temporarily store a large capacity of visual information, however as the information decayed rapidly due to the delay, the subjects were only able to report an average of 4.5 items (Sperling, 1960). Hence, as the interval prolonged, the effectiveness of iconic memory decreases despite the large capacity in storage. Besides that, another theory that can explain the cue delay in partial report is the Bundesen’s theory of visual attention. This quantitative model assumed that the sum of perceptual processing resources which determines the rate of processing is inadequate for the information displayed. Processing resources are used to filter information as distractor and target using selective attention. After the allocation of the processing resources, the information races for encoding in the limited capacity visual short-term memory (Petersen and Andersen, 2012). As the encoding process is time consuming, the information decays as the selection cue delay prolongs thus affecting the items recollection.
The partial report paradigm was pioneered by Sperling’s (1960) dissertation in Harvard University. Due to the subjects’ immediate-memory capacity in whole report, he devised a technique called partial report which was to only report a particular array of items arrangement according to different cue tones for each rows projected after the stimulus was displayed. Three to four items per row were used due to limited perceptual span. Higher tone was for upper row, medium tone for middle row and lower tone for lower row. In his study, he aimed to investigate on information decay by measuring the accuracy of the report. This study was conducted on 5 trained subjects who were scheduled for 3 sessions weekly for a total of 12 sessions. Before the subjects report the information, he delayed the instructional tone for 0.0s, 0.15s, 0.30s and 1s after the stimulus display. The results indicated steep decline in accuracy of report as the delay was longer. 91% of the subjects were able to accurately report the stimulus but as the delay increased to 1s, the accuracy declined to 69%. He concluded most visual information were registered in the sensory memory but decayed rapidly within fractions of a second (Sperling, 1960). Through his research, he was able to show that in testing immediate-memory, not just knowing the limit of the perceptual span actually is but the ability to recall the items seen needed to be measured. With his findings that perceptual span is actually larger than the ability to recall them, his pioneering study pointed to the discovery of sensory storage.
Another study conducted by Merikle (1980) in University of Waterloo, Canada supports the findings by Sperling (1960) in terms of cue delay effect in partial report. This experiment aimed to test on partial report superiority as the cue delay increased for both categories (letters or numbers) and spatial cues (top or bottom). The partial report and whole report were tested on 18 subjects where they were shown 90 sets of stimulus displays from both types of stimulus and the cue were delayed for -250ms, 0ms and +250ms and. The results indicated that spatial cues are more effective than category cues. Both types of partial report condition performed better than whole report, but the decline in performance was greater than whole report when cue were more delayed (Merikle, 1980). This study illustrated that the advantage of partial report is compromised as the cue delay increased. Perhaps, the advantage is due to instruction to report according to rows enhances the accuracy as it involves spatial processing of the stimulus compared to whole report.
On the contrary, Darwin, Turvey and Crowder (1972) findings did not show any significant change in partial report over time. This study was conducted on 12 Yale University undergraduate. In contrast to previous studies by Sperling (1960) and Merikle (1980), auditory stimulus was presented and the indicating cue was in the form of slide projection. A set of 20 stimuli with combinations of monosyllable letters and numbers were given over headphones and the ISIs was 0s, 1s, 2s and 4s. They were asked to report the item and the location as indicated in the stimuli. The findings showed that there are no significant differences of performance between each interval (Darwin et al., 1972). One might argue that perhaps this is due to the echoic memory and not iconic memory. However, Sperling’s (1960) assumption on sensory storage decay across time in partial report should apply to auditory memory. Thus, the result should indicate significant rapid decrease of accuracy in partial report. The insignificant finding could be due to the complex instruction that also tests on their spatial processing apart from recalling the items.
Furthermore, Kuhbandner, Spitzer and Pekrun (2011) investigated on the consequences of emotion-inducing stimuli on the iconic memory decay. 45 with 32 females and 13 male students from University of Munich took part in this study. Emotion-inducing stimuli which were threatening, neutral and positive stimuli consisted of objects and animal pictures. The indicator was shown in a form of arrow after the ISIs of 14ms, 71ms, 229ms, 500ms and 1000ms. 8 trials were conducted for each stimuli conditions and delays. The subjects reported 4 chosen items out of 12 verbally. As expected, the iconic memory degenerated rapidly as the delay time increased however, for threatening stimuli; the results indicated slow decay of information (Kuhbandner et al., 2011). From this study, we can observe that emotion encourages persistence in iconic memory regardless of the delay period. Perhaps, due to the higher number of females in this study affected the results as emotional perception are different across genders. This study suggested that emotional processing occurred faster than selective attention in iconic memory processing.
As we can see from previous studies discussed, various adaptations of Sperling’s (1960) partial report paradigm ware used. However, most studies on partial report that supported Sperling’s findings were conducted at least thirty years ago. A current study on the interstimulus delay effect using neutral stimulus on iconic memory needs to be conducted in order to observe whether Sperling’s assumption are still applicable to this date. Past researches have conducted partial report experiments using traditional tachistoscopes (Sperling, 1960; Merikle, 1980) however for this present study, CogLab 2.0 computer application were used for better and more comprehensive material in data collection. For this research we aimed to observe the effect interstimulus intervals on the iconic memory. Iconic memory was measured by the percentage of the recalled items. We hypothesized that the increase in interstimulus interval decreases the percentage of recalled items.
In testing out the hypothesis, we conducted an experiment with repeated measure design. Data was gathered through convenient sampling. For this study, the independent variable was the interstimulus interval (ISI) whereas the dependent variable was percentage of the recalled items.
Participants in this study consist of 28 (10 males and 18 females) undergraduate Sunway University psychology students who are enrolled in Cognition and Perception course with the age range of 18 to 35 years old. As a part of the coursework, students were rewarded with 1% credit for participating. In selection of participants, the inclusion criterion was those who are wearing visual and hearing aids whereas the exclusion criterion was non-psychology students.
In conducting this experiment, we have used Wadsworth CogLab online laboratory 2.0 (Goldstein, 2011) computer program. Specifically, under the sensory memory section, we have selected partial report experiment.
Interstimulus interval. The interstimulus interval was between the end of the stimulus matrix display and cue tone onset. Each interval was varied at 20milisecond (ms), 100ms, 300ms, and 1000ms.
Percentage of recalled items. The numbers of correct letters reported for each delay trials are recorded and totaled regardless of the order of the letters. The percentages of the scores are then calculated according to each interstimulus interval conditions.
Ahead of the tutorial class, every student was informed to bring their own headphones to reduce distractions during the experiment. All students who were present on the day of the experiment gathered in the computer lab with one computer for each student. This experiment was only conducted in one session with all participants at once with estimated time of completion at about 20 minutes. All participants were assigned to complete every condition of the interstimulus intervals. As the class started under the supervision of a lecturer, the students were allowed to read the instructions and start the experiment at their own pace. They were required to fill in their student ID and put on their headphones before beginning the experiment. As they started, they were exposed to 3 different cue tones to familiarize them with indicator tone on which row to report. Low-pitched tone was for bottom row, medium-pitched sound for middle row and lastly high-pitched tone for the upper row. For a total of 60 trials, each trial started as they pressed the space bar and they had to fixate their vision on a small asterisk at the screen center. After one second and a half, 3?3 matrix of alphabets appeared on the screen for a duration of 150ms. The interstimulus interval started after the matrix flashed and at the offset of the tone. A tone was played as the indicator and the letters of the indicated row were typed. These procedures were repeated for each trial and the students were asked to keep their eyes fixated on the center throughout the experiment. The results were immediately displayed for each participant on their performance and submitted to e-Learn for pooled data collection.
To test the decrease of percentage of recalled items as the interstimulus interval increases, the results were analyzed using repeated measure ANOVA due to the nature of the experiment where all conditions were tested on within the same subjects. Interstimulus interval was classified as categorical variable with four conditions (20ms, 100ms, 300ms, and 1000ms) whereas the dependent variable, the percentage recalled was a continuous variable. The results illustrated significant decrease from the 20ms to 100ms interval on the percentage recalled, Wilk’s Lambda = .011, F (3, 25) = 4.56, p = .011. To observe the informational decay pattern, further test of pairwise comparison demonstrated that there was there was a significant decline between the 100ms interval and 1000ms interval (mean difference= 8.730; p = .02). Furthermore, between 300ms interval and the 1000ms, there was a significant decrease (mean difference = 8.10; p = .009). However, no significant decrease was found from 20ms interval to 100ms interval (mean difference = 1.99; p = 1.00) to 300ms interval (mean difference = 1.35; p = 1.00), and to 1000ms interval (mean difference = 6.75; p = .136). In addition, there was no significant decrease found between the 100ms interval and 300ms interval (mean difference = .635; p = 1.00) but. Generally, since there are greater decrease from 100ms and 300ms interval to 1000ms interval, there were overall decrease across the increasing interstimulus intervals. The mean scores for each interstimulus interval are shown in table 1.
Mean and standard deviation of percentage of recalled items after interstimulus delay
Interstimulus interval (ms)
Percentage of recalled items
This research was carried out to examine the effect of interstimulus delay on the effectiveness of iconic memory. Our hypothesis was testing on whether longer interstimulus delay causes lower performance in the percentage of the test items recalled. The statistical analysis on our data showed that there was indeed a significant decrease in percentage of recalled items across the increasing length of interstimulus interval thus supporting our hypothesis. Our result was found to be consistent with the findings from previous studies (Sperling, 1960; Merikle, 1980) which showed that the information decayed as the time delay increased. On the contrary, our outcome did not support the findings from the studies conducted by Darwin et al. (1972) which indicated no significant difference between the interstimulus intervals and study by Kuhbandner et al. (2011) which indicated persistence of iconic memory despite 1000s of delay.
A possible reasoning for this finding is the iconic memory decayed over time due to the use of neutral stimulus items where 12 random letters were briefly displayed. These random letters were non-relatable to one another to make sense of the information. According to the Atkinson and Shiffrin’s (1969) modal model of memory, the sensory memory acted as the pathway to the short-term memory and the information decayed rapidly if remained unrehearsed before further processing of the information. Thus, as the stimulus displayed provided no significance for participants to further process in short-term store and stimulates the information decay. This justification can be supported by the finding by Kuhbandner et al., (2011) as the usage of the emotion-inducing stimulus, particularly threatening stimulus caused longer visual persistence due to the human survival instinct. Hence, the type of visual stimulus used explains why iconic memory is short-lived.
Besides that, another possible rationalization on decreasing percentage of recalled items over time was due to blinking. The blinking action momentarily disrupts our vision as we are receiving the visual stimuli. Thomas and Irwin (2006) claimed that blinking restrained cognitive processing from their findings in conducting partial report experiment. Their findings showed that more errors were found under blinking condition. As blinking hinders cognitive processing, it is aligned to the Bundesen’s theory of visual attention where limited processing resources is available hence it requires more time to process more information (Petersen and Andersen, 2012). Blinking puts further setback in the information processing thus leading to the decay of unattended information.
Strengths and Limitations
As we conducted this experiment in one session where all participants were tested in one sitting, all participants experienced similar external conditions including lighting and temperature that may affect the attention. The similar extraneous conditions contributed to the strength of this study. Besides that, a relatively large number of samples for experimental design research also helped to strengthen this study.
For limitations, the convenience sampling method of only conducting the research on a class of psychology students is not representative of the population in Sunway University. Thus, it is difficult to generalize our findings as psychology students are more familiar with the CogLab experiments and the theoretical assumption of partial report. Furthermore, this test was conducted early in the morning as soon as the class begins. The students were not in full-alert state during that hour as they rushed to get to class hence may affect their cognitive processing.
Future studies and implications
To improve the present study, one of the ways that future researcher can apply is to conduct the experiment on subject from different courses in this university. To remove any biasness in sampling, psychology students should be an excluded as they have basic knowledge on cognitive processing. Besides that, another way to improve this study is by using other types of visual stimulus such as combination of letters and numbers per row to see whether it has an effect on their iconic memory.
The current finding suggests that there should be very short intervals in between visual aid presentation especially in videos and movies for visual persistence. Besides that, the finding implies that rehearsals from longer duration of visual display help in retaining the iconic memory.
In conclusion, this study focuses on iconic memory where we assumed that the percentage of recalled items decreases as there is increase in interstimulus interval. The statistical data analysis indicated that there is a significant decrease in percentage of recalled items as the interval period prolonged. The iconic memory theory (Sperling, 1960) and theory of visual attention (Petersen & Andersen, 2012) explained on how the delay affects the iconic memory performance. Iconic memory performance decreases over time due to the decay of information. The finding from this study has implication on filming industries to edit their video materials to reduce the iconic memory effect as cut-scene changes.