2014 Volume 56 Issue 6 Pages 453-460
Objectives: The aim of this study was to investigate whether the combination of extension of the encoding time and repetition of a test trial would improve the visual recognition memory performance in older adults. Methods: We evaluated visual memory performance in young and older adults on a Yes-No recognition memory test under four different conditions. The conditions consisted of combinations of encoding times of two and four seconds (E2 and E4) and first and second retrieval practice test trials (T1 and T2): E2T1, E2T2, E4T1 and E4T2. Performance was evaluated by measuring hit rates, false alarm rates, discrimination ability and response bias. Results: Older adults showed better improvement of hit rate and discrimination ability under the E4T2 conditions whereas young adults showed better memory performance under the E2T2 conditions. Conclusions: A longer encoding time and repetition of the test was effective in improving the visual memory performance in terms of the hit rates and discrimination ability of older adults. The results suggest that this strategy should be useful in providing a suitable work environment for older workers.
(J Occup Health 2014; 56: 453–460)
Aging of the population has increased due to the decline in fertility rate and increase in life expectancy in Japan. In 2006, the labor force participation rate of older workers aged 60 and older was reached to 30.2%1). Memory functions such as storing and retrieving recent information from memory correctly are very important in our daily lives and work. They allow or guide us to make a decisions or form responses in a proper way in ever-changing environments. A decline in visual memory, especially in the manufacturing industry, would be very critical to workers, as the workplace requires them to process lots of visual information with considerable speed. Additionally, it is suggested that the memory performance in older people is lower compared with that of young people when faced with new and/or complex information2).
One of the methods of evaluating the memory performance is the recognition memory test3). In this test, a participant is asked to remember a set of items for a certain period of time; this is referred to as the encoding phase. Later, the same items as those presented in the encoding phase (old items) and a new set of items are shown to the participant in a random order. The participants are asked to indicate if each item is an old (yes) or a new (no) item by retrieving the information from their stored memories (test phase). The researcher then calculates the hit rate (yes response to the old item) and false alarm rate (yes response to the new item) and also estimates the individual's discrimination ability (ability to discriminate between old and new items) and response bias scores (a decision criterion to apply when responding to the items)3). Researchers have used this method to evaluate the memory performance of different groups of participants under various conditions.
As we age, our visual memory declines; it is reported that this decline is more apparent than that in verbal memory4). Errors in visual memory may occur when the original items in the encoding phase have semantic or perceptual similarities to the new items in the test phase5). Compared with young adults, older adults make more errors when a new item in the test phase is categorically related to an item presented in the encoding phase6). Other researchers have reported that the declines in encoding speed in older adults lead to lower performance on visual memory tests than in young adults7). In daily life, we receive an enormous amount of scenic information from the real world. This information is highly complex, as each scene includes information about its theme, background and the relationships between objects or groups of objects. Older adults may experience difficulty remembering complex information such as real-world scenes during the encoding phase because they have slower processing speeds and difficulty in binding features from information in the environment8). Moreover, previous studies have found that older adults could encode the details of pictures but were less effective than young adults in retrieving them in the test phase, resulting in lower memory performance9). Lower visual memory performance is likely to increase errors and decrease the work efficiency of older adults, thereby reducing their quality of life. Thus, it is imperative to find solutions to enhance visual memory performance in older adults.
There are many ways to enhance visual memory performance during the encoding and test phases. Previous studies have found that increasing the time for presenting items in the encoding phase is equivalent to extending the duration of the encoding phase; this strategy led to an increase in the memory performance of subjects for scenes10). These studies, however, were focused on young adults or failed to indicate the ages of their participants. Prior studies have shown that increasing the number of test trials (or repeating the test trials) during the test phase can improve long-term memory retention. This phenomenon is known as the testing effect11, 12). Most of the studies related to repeated-test effects have also focused on young adults; few have examined them in older adults.
It is reported that older adults exhibit difficulty in remembering complex scenes during the encoding phase8) and have some trouble retrieving details of information during the retrieval phase13). In such a case, it may be possible that the combination of extending the duration of the encoding phase and increasing the number of test trials may be a better strategy for improving the memory performance of older adults than applying only one of them. Therefore, we hypothesized that the combination—but not application in an alternating manner—of the two strategies could improve visual memory performance in older adults. To confirm this hypothesis, we measured visual memory performance by using a yes-no recognition test in young and older adults. To increase the difficulty level for older adults, we used old and new pictures of natural scenes with high conceptual similarity (themes) in our experiment, expecting that the older adults would perform worse than the young adults. If older adults showed improvement in visual memory performance by combining the two strategies (extending the encoding time and repeating the test trial), the result could have useful implications for older adults both in the workplace and in daily activities.
The participants were 13 young men (mean age=22.1 years, SD=1.2, range=20–24 years) and 15 older men (mean age=67.3 years, SD=4.1, range 62–76 years). The young group consisted of undergraduate students from the University of Occupational and Environmental Health, Japan. The older group lived in the same area as the young group. All of the participants were native speakers of Japanese and voluntarily enrolled in this study. All of them reported their health status as good and denied any history of neurological impairment. The Medical Ethics Committee at the University of Occupational and Environmental Health, Japan, approved the study. Participants signed informed consent forms before the study began.
Stimuli and visual memory testWe employed a yes-no recognition test to evaluate the performance of visual memory. Pictures of colorful natural scenes were used as stimuli in this study. Figure 1 shows the details of the visual memory test used in this study. The experiment consisted of an encoding phase and a test phase. In the encoding phase, participants were asked to remember seven pictures of natural scenes that were sequentially presented on a screen against a gray background with a 0.5-sec interstimulus interval between picture presentations. After a 3-sec maintenance period, the 12 pictures were sequentially presented in the test phase. Participants were required to answer whether the pictures displayed were old or new pictures by pressing the “Yes” or “No” button, respectively, on the keyboard as soon as possible. Immediately after the participant pressed the Yes or No button, the next picture was presented on the screen. The maximum time allowed to answer each picture was 3 sec.
A diagram illustrating the details of the yes-no recognition test for measuring visual memory performance in young and older participants. The visual memory test began with the encoding phase and was followed by the test phase. In the encoding phase, the participant memorized seven pictures that were presented one at a time for 2 sec (E2) or 4 sec (E4) with an interstimulus interval (ISI) of 0.5 sec. After a maintenance period of 3 sec, the participants gave responses indicating whether the 12 pictures displayed one at a time were old or new pictures. They responded as quickly as possible by pressing the Yes or No buttons on the keyboard. The 12 pictures in the test phase were divided into two tests: Test 1 (T1) and Test 2 (T2). Each test consisted of six pictures: Three old pictures from the encoding phase and 3 new pictures from the test.
The first and the seventh pictures in the encoding phase were not used as old pictures to prevent the primacy and recency effects. Nineteen of the pictures in each phase were selected on the basis of conceptual similarity, such as beaches, forests, seacoasts and mountains, so that the new pictures (lures) were highly similar to the old pictures (see Fig. 2).
An example of a picture presented to participants in the encoding phase (left) and a new picture (lure) presented in the test phase (right). The pictures in the encoding and test phases were selected on the basis of conceptual similarity (same category). A lure picture refers to a new picture that is highly similar to an old picture presented in the encoding phase.
To evaluate the combination of extending the encoding time and repeating the test trial effect on memory performance, we employed two different encoding times combined with two different testing conditions. We used 2 sec for short encoding (E2) and 4 sec for long encoding (E4) as the time durations for the picture presentations in the encoding phase and set the test phase into two parts, the first test (T1) and the second test (T2). Each part of the test phase included six pictures (three old pictures and three new pictures). The E2 and E4 conditions were randomly presented in sixteen sets of the test. Participants were not told that there were variations in the encoding times and repetition of the test trial in this study.
ProcedureParticipants sat on an adjustable chair at a distance of 50–60 cm from the display. They performed a practice session to familiarize them with the equipment and procedures until they fully understood the recognition memory test. The participants were provided 20 minutes of rest after the practice session, and were then required to perform the 16 sets of the visual memory test. Participants decided whether the 192 pictures in the 16 sets of tests were old or new. The entire test took approximately 20 minutes for each participant. During the visual memory test, the Yes-No answers were recorded. Based on the encoding time conditions and the test conditions, the results were classified into four conditions: E2T1, E2T2, E4T1, and E4T2. Each condition included 48 trials (24 trials for old pictures and 24 trials for new pictures).
Visual memory performance analysisSignal detection theory was used as the basis for analyzing the accuracy of the participants on the yes-no recognition test. This theory may be applied when two possible stimulus types must be discriminated. In the yes-no recognition test, participants had to discriminate between a signal (old picture presented) and the absence of a signal (new picture presented). Based on this model, there were four possible responses: 1) hit, meaning a yes response was given for an old picture (correct response); 2) false alarm, meaning a yes response given for a new picture (incorrect response); 3) correct rejection, meaning a no response was given for a new picture (correct response); and 4) miss, meaning a no response was given for an old picture (incorrect response). When the hit rate is 1 or the false alarm rate is 0, these measures are undefined. Therefore, we corrected the hit and false alarm rates by using the following formulas provided by Snodgrass and Corwin3). The hit rate (H) and false alarm rates (FA) were calculated by the following formulas, respectively: H = (#Hit + 0.5)/(#OLD + 1), FA = (#FA + 0.5)/(#NEW + 1). In these formulas, #Hit represents the actual number of hits, #FA represents the actual number of false alarms, #OLD represents the number of old pictures in the test trial, and #NEW represents the number of new pictures in the test trial. Two more parameters were calculated from H and FA3). The first parameter was discrimination ability (dL), reflecting the participant's ability to discriminate between old and new pictures. The dL was calculated using the following formula: dL = ln [H*(1–FA)/(1–H)*FA], where H represents the Hit rate and FA represents the false alarm rate3). A higher dL score indicates greater ability to discriminate old and new items. The second parameter was response bias. The meaning of the term response bias in this study or in the field of signal detection theory is different from the meaning used in the field of epidemiology or occupational health, such as in the case of the healthy worker effect. It is an index reflecting the decision criteria determined by the participant's preference or tendency to answer “Yes” or “No”. Response bias (CL) was calculated by the formula CL = 0.5 *ln [(1–FA)*(1–H)/H*FA]. A positive values for CL is called a conservative CL (tending to respond “No”), a negative value is called a liberal CL (tending to respond “Yes”), and 0 is called a neutral CL3).
The comparisons of H, FA, dL and CL between conditions (E2T1, E2T2, E4T1 and E4T2) were analyzed by a three-way repeated-measures analysis of variance (ANOVA) with encoding time (E2 vs. E4) and repeating the test (T1 vs. T2) serving as the within-subjects factors and age group (young adults vs. older adults) serving as the between-subjects factor. Post hoc pair-wise comparisons were conducted using the Bonferroni correction. All statistical analyses were performed using IBM SPSS Statistics for Windows version 16.0. All data are expressed as means ± SDs.
Figure 3 (A) shows the H data of young and older adults under the four different conditions. We found a significant main effect of age (F (1, 26) = 10.86, p = 0.003, ηp2 = 0.30), indicating a lower H in older adults than young adults. There were main effect of encoding time (F (1, 26) = 8.61, p = 0.007, ηp2 = 0.25) and repeating the test (F (1, 26) = 31.73, p<0.001, ηp2 = 0.55) on H. There was no interaction effect between any combinations of two factors or among three factors. The results of the post hoc analysis showed that H in older adults was significantly lower than that of young adults under all conditions except E4T2. In young adults, the post hoc comparisons revealed that the H for E2T2 (0.90 ± 0.06) was significantly larger than that for E2T1 (0.85 ± 0.07). In older adults, the H for E2T2 (0.83 ± 0.08) was significantly higher than that for E2T1 (0.75 ± 0.11), and the H for E4T2 (0.90 ± 0.06) was significantly higher than that for E4T1 (0.79 ± 0.09). The H for E4T2 (0.90 ± 0.06) was significantly higher than that for E2T2 (0.83 ± 0.08) in older adults.
Hit rates (H) (A) and false alarm ratea (FA) (B) of the young and older adults under the 4 conditions (E2T1, E2T2, E4T1 and E4T2). The values represent means ± SDs of H and FA. Significant differences between conditions are indicated by an asterisk (*p<0.05), and significant differences between young and older adults under the same condition are indicated by a dagger (†p<0.05); significance of differences was determined using post hoc tests with Bonferroni correction.
Figure 3 (B) shows the FA data of young and older adults under the four different conditions. There was a significant main effect of age (F (1, 26)=31.95, p<0.001, ηp2=0.55) on FA, indicating a higher FA in older adults than young adults. There were interaction effects between repeating the test and age (F (1, 26)=5.75, p=0.024, ηp2=0.18) and encoding time and repeating test (F (1, 26)=17.02, p<0.001, ηp2=0.40) on FA. There was an interaction effect among three factors (F (1, 26)=18.85, p<0.001, ηp2=0.42) on FA. The post hoc analysis results indicated that the FAs of the older adults were higher than those of young adults under all conditions. The FAs in young adults were not significantly different between conditions. In older adults, the FAs for E2T2 (0.25 ± 0.08) and E4T1 (0.22 ± 0.07) but not E4T2 (0.19 ± 0.08) were significantly higher than that for E2T1 (0.16 ± 0.09). The FA for E4T2 (0.19 ± 0.08) was significantly lower than that for E2T2 (0.25 ± 0.08).
Discrimination ability (dL)The dL in the young and older groups under the four different conditions are shown in Fig. 4. There were main effects of age (F (1, 26)=36.82, p<0.001, ηp2=0.59) and repeating the test (F (1, 26)=12.96, p=0.001, ηp2=0.33) on dL. We found an interaction effect between encoding time and repeating the test (F (1, 26)=6.75, p=0.015, ηp2=0.21) on dL. There was an interaction effect among three factors (F (1, 26)=14.39, p=0.001, ηp2=0.36) on dL. The results of the post hoc analyses indicated that older adults showed a lower dL than young adults under all conditions. In young adults, the dL for E2T2 (5.34 ± 1.51) was significantly higher than that for E2T1 (4.52 ± 1.34). There were no significant differences between other combinations of conditions. In the older group, the dL for E4T2 (3.82 ± 0.91) was significantly higher than that for E2T2 (2.80 ± 0.77) and E4T1 (2.71 ± 0.74). Although it was not significant, the dL for E4T2 tended to be higher than that for E2T1 (3.05 ± 0.86) in older adults.
Discrimination ability (dL) of the young and older adults under the four conditions (E2T1, E2T2, E4T1 and E4T2). The values represent means ± SDs of dL. Significant differences between conditions are indicated by an asterisk (*p<0.05), and significant differences between young and older adults under the same condition are indicated by a dagger (†p<0.05); significance of differences was determined using post hoc tests with Bonferroni correction.
Figure 5 shows the CL in the young and older groups under four different conditions. There were significant main effects of encoding time (F (1, 26)=23.18; p<0.001, ηp2=0.47) and repeating the test (F (1, 26)=28.89; p<0.001, ηp2=0.53) on CL. There was no main effect of age on CL. There were no interaction effects of any combination of the three conditions on CL.
The response bias (CL) of the young and older adults under the four conditions (E2T1, E2T2, E4T1 and E4T2). The values represent means ± SDs of CL. Significant differences between conditions are indicated by an asterisk (*p<0.05); significance of differences was determined using post hoc tests with Bonferroni correction.
In the young group, the CL values under all conditions except E4T2 were conservative. If the CL value is more than zero, it means that the subject shows a tendency to respond “No” (conservative type). In the older group, only the CL for E2T1 was the conservative type. CL values for E2T2 and E4T2 in the older group showed liberal bias. If the CL value is less than zero, it means the subject shows the tendency to respond “Yes” (liberal type). The CL value for E4T1 in the older group was nearly zero (neutral or no bias).
Older adults demonstrated better performance in H and dL under the E4T2 conditions. Regarding the FA, we only found that the FA for the E2T2 conditions was worse than that of the other conditions, excluding E4T1, in the older adults. Therefore, we can conclude that the combination of extending the encoding time and repeating the test trial improved the visual recognition memory performance of the older adults in H and dL, but not in FA. Independent use of an extended encoding time or repeated test trials seems an inferior tactics for older adults. On the other hand, in young adults, the data for H, FA and dL indicated that the E2T2 conditions seemed to provide better visual memory performance, whereas the E4T2 conditions did not. For Hs, the combination of extending the encoding time and repeating the test trials seemed to be successful in improving the memory performance of the older adults to the same extent as the young adults. Previous studies have found that recognition tasks for testing memory performance for pictures could be increased when longer times were allowed for encoding information14, 15).
In the young adults, there was a significant difference between the dL values for E2T1 and E2T2, and there was a tendency for the dL of E4T2 to be larger than that of E4T1. In older adults, a similar tendency was observed, with the exception of E2T1 and E2T2. It seems that repeating the test trial but not extending the encoding time had an enhancing effect on the dL in both young and older adults, except under the E2 conditions in older adults. The reason why the older adults did not show a higher dL under the E2T2 conditions than under the E2T1 conditions might be that the encoding time of 2 sec was not long enough for them to encode the similar scenes in this experiment. Previous research has shown that performance on the recognition test and the accuracy of the answers to questions about the detail of objects in natural scenes were increased by lengthening the encoding time from 1 to 20 sec10). Brady and colleagues examined the ability to detect the changes in pictures of real-world objects by using three different encoding durations (1.2 sec, 6 sec and 18 sec) and reported that the ability to detect changes was higher for longer durations than shorter durations16). As older adults generally have difficulty processing the details of pictures during the encoding phase, shorter encoding times (such as the 2 sec used in our study) might not be sufficient time to detect the details of pictures. This might have caused confusion for older adults when they discriminated between old and new pictures within the same category. Further investigations are needed to determine whether a longer encoding time (more than 4 sec, such as 6 sec or longer) improves the dL and the other memory functions in older adults.
Our study suggests that repeating the test trial enhance the dL. Although the mechanism of this phenomenon is unclear, it may be explained by the retrieval effort hypothesis17). This hypothesis states that successful retrieval of information requiring more effort is more likely to enhance memory than successful retrieval requiring less effort. In our study, we did not inform participants that there were two sets of tests or that there was quite a time lag between the beginning of the first and the second tests (up to 18 sec) during the retrieval test phase. The rate of forgetting the old pictures in the encoding phase would be expected to increase as time passed18). Thus, participants might need to make a greater effort to retrieve information stored in their memories during the second test than during the first test. Based on the retrieval effortful hypothesis, greater effort leads to better performance on memory retrieval tasks.
Response bias reflects the decision criteria that an individual uses when they are faced with uncertainty in a recognition memory test. When people cannot make a decision based on a perfect memory, they may make the decision based on partial guessing. People show this tendency when they make decisions either to respond “No” (conservative type) or respond “Yes” (liberal type). In general, when the dL is high, a conservative CL makes the FA lower, with a slight reduction in the H. On the other hand, a liberal CL makes the H higher, with some increment in the FA when the dL is lower. In our study, the CL of young adults tended to be conservative under most of the conditions except for E4T2. As the dL values in young adults are higher, a conservative CL in young adults resulted in success in obtaining a higher H and lower FA in our study. The CL values of older adults were mostly liberal, except under the E2T1 conditions. Previous studies have reported that older adults prefer to respond to the gist of the presentation rather than the details of the information in recognition tests of new and old but similar items. This results in a tendency to use a liberal response criterion more often9, 19). It may be suggested that as the dL values in older adults are lower than those of young adults, the liberal CL in older adults may be partly responsible for their higher H and slightly higher FA in our study. It is not clear why the CL for E4T2 in young adults was liberal and that for E2T1 in older adults was conservative. Further investigations will be needed to clarify this.
To assist older workers with impaired memory performance, the workplace environment may need to make some adjustments to accommodate them. Because they have a slower information-processing speed, older workers may encounter difficulties in learning new information or work processes with high-speed tasks. Therefore, training programs and/or actual work environments for older workers can benefit from providing adequate time to encode new information combined with repetition and practicing new skills. This could also help to reduce errors due to memory problems.
Our study revealed that extending the encoding time and repetition of the test can contribute to improvement of the visual memory performances in terms of hit rate and discrimination ability, excluding the false alarm rate in older adults. This suggests that giving enough time for the memory encoding phase and repeated chances to retrieve the information from memory would help to improve the visual recognition memory performance in older adults in daily life, including employment.
