Testing the Existence of Iconic Memory Based on the Multi-Store Memory Model

1. Introduction

Memory was defined as a cognitive process that enables us to code, store and retrieve information from the past. Scientists suggested a lot of models that aimed to describe and explain the mechanism of memory and one of the most famous ones is the multi-store memory model proposed by Atkinson and Shiffrin in 1968. The model assumed that memory consists of three main sub­components that are characterized by different capacities and duration, i.e. how many units of information can be held for a certain amount of time by a particular component. These components are long-term, short-term, and sensory memory store which is divided into echoic, haptic, and iconic memory that is the research subject of this paper. Iconic memory encodes for visual inputs and its capacity is limited only by the perception so people can potentially remember unlimited visual information. However, the duration of iconic memory is very short since it has a span of less than one second. Paying attention is the only way to transfer information to the short-term memory store, where its time of decay is longer. However, attention is limited, hence it is impossible to pay attention to everything we see. George Sperling wrote once a statement that perfectly summarizes all that we know about iconic memory: “More is seen than be remembered”¹.

Sperling conducted a substantial investigation in an iconic memory store that purposed to test quantitatively its existence and consisted of 7 experiments. However, only 2 of them are relevant to this paper; the whole-report (control) experiment and the partial-report experiment which involved attention. 5 students participated. Repeated measures design was used and the condition applied (presence of attention) was the independent variable. The participants were presented with a 50 ms flash of a grid that was either alphanumeric or consisted only of letters. In the whole-report condition they had to recall as many characters as possible whereas in the partial-report condition, approximately 50 ms after the visual presentation, the participants heard a 500ms tone that indicated which row from the grid they had to recall. The dependent variable was the percentage of correctly remembered characters from the grid. The study was time-consuming and involved a large number of trials where grids had various configurations. The results showed that participants in the whole­report (non-attention) condition were able to recall a significantly lower percentage of characters from the grid than in the partial-report (attention) condition where the attendees accessed almost the whole grid. As a result, it was concluded that iconic memory does exist and its mechanism is consistent with the assumptions of the multi-store memory model, that is attention is responsible for the transfer of information to the short-term memory store.

In the study, I want to investigate the link between attention and the amount of information that can be held. This paper aims to test quantitatively the existence of iconic memory. The subject is relevant as it helps to understand how memory and its components work. According to the investigated model, iconic memory is a part of the first component of memory which information reaches. Moreover, it depends on sight which processes 83%² of the stimulus so it is justified to extend the knowledge about this subject. The theoretical prediction is that the amount of correctly recalled information will depend on paying attention. Attention will be operationalized as a partial­report condition, where participants had to remember 4 elements from the grid or whole-report condition where they were instructed to recall all, 12 elements from the grid. The amount of correctly recalled information is shown as the percentage of correctly recalled characters from a grid.

Therefore:

- H0: There will be no significant difference between the percentage of alphanumeric elements correctly remembered by the participants in the whole-report condition and the partial-report condition.
- Hi: The participants in the partial-report (attention focused) condition will correctly remember a significantly greater percentage of alphanumeric elements than in the whole­report condition.

The independent variable is the condition applied whereas the dependent variable is the percentage of correctly recalled characters from a grid.

2. Exploration

2.1: Design and method

The study was a laboratory experiment and involved the repeated measures design, which was used to exclude the problem of individual differences that would certainly occur if different subjects were used in each condition. The design is vulnerable to an order effect so the procedure included counter-balancing. The experiment was conducted online because of the sanitary restrictions related to the coronavirus pandemic.

2.2: Participants.

The target population was young Poles. Young people were chosen because Sperling also used people from this age group, possibly because then memory is usually most efficient. The sample consisted of 20 Poles (i2 males and 8 females) between the age of i6 and 30 who voluntarily agreed to take part in the experiment. The majority of them (i6) attended bilingual high school in XXXX. The remaining participants responded to the online announcement. Voluntary sampling was used because of its simplicity and low-time consumption.

2.3: Materials

1. Informed consent form³.
2. Briefing note⁴.
3. Debriefing note⁵.
4. A piece of paper and a pen to write down the answers.
5. Laptop and access to the Internet. It was important to minimize the variation of screen size to provide a better reliability of the results. A pre-experiment was done and it turned out that participants often were not able to remember anything from the alphanumeric grid when they performed the task on a small screen. Larger laptop screens resolved this issue.
6. Alphanumeric grid. Every alphanumeric grid consisted of 3 rows and 4 columns which gives 12 characters in total⁶. Such grids were used to improve reliability because they were among those applied in the original study.

2.4 Procedure

Participants got familiar with standardized instructions. They could sign the informed consent and they were informed about the right to withdraw at any time. The attendees were told that they were going to see a 50ms presentation of an alphanumeric grid. In the whole-report condition, they were instructed to remember as many alphanumeric characters as possible. In the partial-report condition, the participants heard a 500ms tone 50ms⁷ after the visual presentation that indicated which row of the grid they had to recall. The high, middle, and low tones corresponded to the top, middle, and bottom rows respectively. The tones were not similar to each other and before the task participant got acquitted with the tones and the rows that corresponded to them, to eliminate any misunderstandings. In both conditions, the participants were told to guess in case of uncertainty and asked to write down their responses in any order they wanted and to put them into a suitable field in the online form which the participants were obliged to send us. In each condition, the participants had one practice trial before the official task to get familiar with the procedure.

After completion, the participants were debriefed and they could also ask questions by submitting them in an online form. The participants could see their results as well as the final report by sending an e-mail to the address that was provided in the contact information section.

2.5: Control variables

1. The participants were told to be rested and to perform the task only once using a laptop, under quiet conditions with no distractions to eliminate potential confounding variables.
2. There was one set of participants. Ten of them were randomly chosen to start with the whole­report condition and 10 started with the partial-report condition to reduce the order effect.
3. The configuration and size of the grids were the same for every participant and every tone lasted 500ms to provide standardized conditions.
4. The participants have never studied Psychology and had sufficient English language skills. This was to avoid demand characteristics bias and comprehension difficulties.

3. Analysis

Since the variables were at least ordinal, the results between conditions can be compared using all measures of central tendency. All measures were used to show data transparently and reliably. Regarding the measures of dispersion, the standard deviation was calculated.

Note: In the partial-report condition one row does not report the whole stimulus, but only a part of it. That is why Sperling decided to multiply the number of correctly recalled characters from a row times the number of rows in the grid to make the data comparable. Consequently, in this paper data from the partial-report condition was firstly multiplied by three⁸ and then compared with data from the whole-report conditions and converted into percentages, as Sperling did.

Abbildung in dieser Leseprobe nicht enthalten

Table 1: Table showing the measures of central tendency and dispersion concerning whole-report and partial-report conditions.

Abbildung in dieser Leseprobe nicht enthalten

Graph 1: Graph showing the mean percentage of correctly recalled characters from the alphanumeric grid in whole-report and partial-report conditions.

The mean shows that participants recalled correctly a significantly higher percentage of words in the partial-report condition. However, it was also the condition with a bigger variance in scores since the standard deviation was higher. In the whole-report condition, the median was 16.67% whereas in the partial-report it was equal to 50.00%. The most frequent scores were 8.33% and 50.00% respectively and this highlights the difference between conditions and confirms the previous remarks. The ranges of scores were from 0.00% to 58.33% and from 0.00% to 100.00% respectively, making the data a bit unclear when it comes to this measure.

A related t-test was used to assess the significance of the results since it is suitable for repeated measures design and the distribution does not deviate severely from normality. Furthermore, there are no obvious outliers. The test showed that the t- value is 5.81 which exceeded the critical value of 1.73 for p < 0.05 so the results are significant⁹. The null hypothesis is rejected and the research hypothesis is accepted. Therefore, it seems that attention causes participants to recall correctly a significantly higher percentage of characters than without attention, showing that iconic memory exists.

[...]

¹ Sperling., 1960.

² Rosenblum L.D., Stolovitch, H. and Keeps, E., 2006.

³ See appendix 3.

⁴ See appendix 1 and 2.

⁵ See appendix 5.

⁶ See appendix 6.

⁷ The 50ms and 500ms intervals were obtained with the help of a special software called FFmpeg.

⁸ Because each grid had 3 rows.

⁹ p = 0.00001.