Within-category Similarity Negatively Affects Associative Memory Performance in Both Younger And Older Adults Part 2

Results

Statistics for all studies were computed using SPSS Version 25 and all tests of significance were performed with an alpha value of .05 unless otherwise noted. Figure 1B depicts hit and false alarm rates across age and similarity conditions (see also Table 1). To examine similarity effects on associative recognition across age in Study 1, we conducted repeated measures mixed model ANOVA1 including between-subjects factors of similarity condition (low, high similarity) and age group (younger, older) and a within-subject factor of stimulus type (target, lure) on the proportion of “Yes” responses. 

Stimulus type and memory are inextricably linked. When we receive different types of stimulation, the neurons in the brain are activated to varying degrees, resulting in different types of memories. Therefore, it is very important to make reasonable use of stimulus types to improve memory.

The first is visual stimulation. Visual stimulation, the most common type of stimulation, has a very positive effect on improving memory. We can get visual stimulation by looking at beautiful scenery, photos, or diagrams, which can improve our memory through visual stimulation. In addition, learning content can also be made into pictures or diagrams, which can also better help us understand and remember.

The second is auditory stimulation. Auditory stimulation is also a very important type of stimulation, which can help us better understand and remember language. When learning languages and other knowledge, we can learn by listening to audiobooks, lectures, etc., which can better deepen memory and improve understanding.

Again, it’s sports simulation. Exercise stimulation can boost brain activity and sharpen the brain. Therefore, moderate exercise can improve memory. During study and work, we can improve concentration and memory through light exercise, such as jogging, deep breathing, etc.

In conclusion, the right type of stimulation can greatly improve our memory. We can use a variety of stimulation types to stimulate brain activity to better understand and retain learning content. A positive attitude and a positive attitude are also very important so that you can better enter the state and improve learning and work efficiency. It can be seen that we need to improve our memory. Cistanche can significantly improve memory, because Cistanche can also regulate the balance of neurotransmitters, such as increasing the level of acetylcholine and growth factors. These substances are very important for memory and learning. In addition, meat can also improve blood flow and promote oxygen delivery, which can ensure that the brain receives sufficient nutrition and energy, thereby improving the vitality and endurance of the brain.

Click Know to improve short-term memory

This analysis revealed a main effect of stimulus type [F(1,115) = 562.09, p < .001, ηp 2 = 0.83], with participants generally showing higher hit than false alarm rates across age and similarity and a main effect of similarity [F(1, 115) = 6.67, p = .01, ηp 2 = 0.06], with greater “Yes” response rates in the low compared to the high similarity condition across age and stimulus type. 

Importantly, there was a significant two-way interaction between stimulus type and similarity [F(1, 115) = 55.13, p < .001, ηp 2 = 0.32]. Follow-up comparisons showed that across age, in the high similarity condition hit rates were lower and false alarm rates were higher compared to the low similarity condition [hit rate: t(117) = 3.26, p = .002, d=0.14; false alarm rate: [t(117) = 7.03, p < .00; d=0.13], indicating that both age groups showed poorer associative memory when the scene was similar across studied and lure pairs because it was drawn from the same category in both instances, compared to when there was less similarity in scenes because they were drawn from different categories. 

There was also a significant two-way interaction between stimulus type and age [F(1, 115) = 29.20, p<.001, ηp 2 = 0.20], such that older adults showed lower hit rates and higher false alarm rates across similarity conditions [hit rate: t(117) = 4.93, p < .001; d=0.14; false alarm rate: [t(117) = 2.17, p = .03; d=0.16], which is consistent with typically observed age-related memory differences. The three-way interaction between similarity, stimulus type, and age was not significant [F(1,115) = 0.20, p = .65, ηp 2 = 0.002], suggesting that similarity had no differential impact on memory responses to targets and lures across age groups.

To examine whether there were any differences in discriminability between the similarity condition and age groups, we ran a 2 (young, old) by 2 (high, low similarity) ANOVA on d prime values. The main effect of similarity was significant, F(1, 115) = 56.25, p<.001, ηp 2 = 0.33, such that discriminability was higher in the low similarity condition compared to the high similarity condition. The main effect of age was also significant, F(1, 115) = 26.80, p<.001, ηp 2 = 0.19, such that younger adults showed higher discriminability compared to older adults. The two-way interaction between similarity and age was not significant, F(1, 115) = 0.69, p=.41, ηp 2 = 0.006.

Discussion

Regarding manipulations of similarity, the high within-category similarity condition (i.e., scenes from the same scene category) resulted in reduced hit rates to targets and higher false alarm rates to recombined lures compared to the low similarity condition (i.e., unique scenes). That is, overall, associative memory was poorer under conditions where there was similarity across studied contexts and high contextual interference within lures. 

Similar to results seen in studies of item memory, the present results suggest that memory for associative information is influenced by the gist associated with the context in which item information is studied (Dennis et al., 2014; Gallo, 2013; Giovanello et al., 2009; Kelley & Wixted, 2001). Specifically, though each face and scene used in the high within-category similarity condition was unique, the scenes were all drawn from 4 categories, creating overlapping gist traces concerning the general source category (e.g., kitchen, office) across the scenes.
When a face was repaired with a scene from the same category with which it was originally studied, participants had a difficult time correctly rejecting the novel face-scene pairing as new. Thus, in the high similarity condition, we posit that interference from categorical gist influenced associative recognition by contributing to higher rates of false alarms to recombined lure pairs, lower accurate recognition of target pairs, and consequently, overall poorer discrimination between target and lure pairs that shared a similar context. 

Alternatively, when the face was repaired with a scene that was unrelated to the source, there was likely less overlap in memory representations and no interference from category gist to impair memory. This reduced interference across scenes, resulted in overall greater associative memory, as evidenced by lower false alarm rates and higher hit rates in the low similarity condition.

In line with a wealth of previous aging research, we identified an overall reduction in associative memory in older compared to younger adults, with older adults, evidencing lower hit rates and higher false alarm rates relative to younger adults. Despite a general age deficit in discriminability between targets and lures, we did not find evidence that categorical similarity had a disproportionate effect in older adults. That is, contrary to our hypotheses, the negative effect of categorical similarity on associative recognition was proportional across both age groups. 

The absence of an age interaction concerning either hit rates or false alarm rates suggests that interference induced by similarity in context information is not differentially detrimental to older adults’ associative memory, at least in the present study. Moreover, results from Study 1 indicate that contextual relatedness influences both younger and older adults’ associative memory to a similar degree, highlighting the pervasiveness of similarity effects across ages.

The absence of age differences when there is within-category similarity may be due to the greater demands of associative relative to item memory. 

That is, in the face of these additional demands, and especially under conditions where similarity in context is high, all participants may rely to a greater extent on gist to support associative encoding and retrieval processes than they would if no categorical overlap was present across associative pairs. Thus, the comparable associative discrimination for scenes observed within each age group may be a result of the demand on associative memory that requires reliance on gist processing and leads to false memories (Brainerd & Reyna, 2002; Dennis & Turney,2018; Gallo, 2013; Giovanello et al., 2009; Webb & Dennis, 2019). 

This also may be demonstrated in the overall lower hit rate and higher false alarm rates in the high-similarity condition, such that both age groups may have relied on more gist-based processes rather than recollection-based processes, thus decreasing recognition of targets and overall memory performance (Tun et al., 1998). While this difference may have been due to the difficulty of the two different similarity conditions, we did ensure that both age groups were matched across the similarity conditions to eliminate the potential effects of different groups taking part in the study.

Overall, these results highlight the negative influence that interference from within-category similarity has on associative memory across ages and demonstrate that both younger and older adults’ memory performance can benefit from reduced mnemonic overlap in associative information.

Study 2

Introduction

Study 1 showed that associative memory performance is reduced when there is gist-based interference from high within-category similarity between encoding and retrieval. Additionally, results showed that older adults are not more susceptible to contextual interference in associative recognition than younger adults. While Study 1 utilized a between-subjects design, it also featured a difference in several scene categories between the high and low similarity conditions, making it unclear as to whether differences across conditions were due to the within-category similarity at retrieval or also partly to the fact that there were a smaller number of scene categories in the high similarity condition. 

That is, the constraint of fewer categories may have contributed additional similarity effects in the high similarity condition because, in addition to featural similarity, there are also fewer distinct categories from which to draw features. Thus, the aim of Study 2 was to explore the influence of contextual interference using the same set of scene categories for both the high and low similarity conditions using a within-subject design. To accomplish this goal, Study 2 tested the effect of within-category similarity utilizing the design of only the high similarity condition from Study 1 (i.e., 4 scene categories) with the following exceptions. At retrieval, half of the recombined lure pairs featured a studied face recombined with a scene from the same contextual category as its original scene associate (high within-category similarity condition); the other half of the recombined lure pairs featured a studied face recombined with a scene from one of the other three contextual categories (low similarity condition). 

Additionally, novel lure pairs were included in Study 2 to ensure effects of similarity were independent of familiarity. The features of Study 2 importantly allowed us to hold both item and scene familiarity equal (i.e., all faces and scenes were previously seen), as well as measure the effect of interference from similarity within individuals. We predicted that, in line with Study 1, participants would show greater false alarms to recombined lure pairings that featured a high degree of within-category similarity compared to when similarity was reduced. Moreover, false alarms to recombined pairings in both high and low similarity conditions should be higher than false alarms to novel lure pairs as familiarity with recombined pairings should be greater, and unrelated lure pairs feature both novel face-scene items and a novel association. Because no age differences were observed between the high and low similar contexts in Study 1, Study 2 included only younger adults.

Methods

Participants—Thirty younger adults completed Study 22. Participants were recruited from introductory psychology classes at Penn State University. Two participants were excluded for not responding on at least 25% of trials and a third participant was removed for failing to follow task instructions, leaving a final sample of 27 participants (mean age = 19.78; SD = 0.93; 23 Female). On the day of the study, participants provided written informed consent for a protocol approved by the Pennsylvania State University institutional review board. A post-hoc sensitivity analysis conducted in G*Power 3.1 (Faul et al., 2007) confirmed that power was sufficient to detect a small effect size (f=0.32) for a within factor of interference conditions, given that α=.05, and the correlation between repeated measures is equal to .05.

Materials—Face images consisted of male and female faces obtained from the same online databases as Study 1. Scene images from 4 different categories (backyard, kitchen, restaurant, street) were obtained from an Internet image search. These face and scene images were used to construct face-scene pairs, with the face always presented superimposed on the scene. Face-scene pairs in each condition were counterbalanced across both male/female faces and indoor/outdoor scenes.

Procedure—The task procedure included two encoding-retrieval blocks and is depicted in Figure 2A. Two versions were created to counterbalance the order of encoding and retrieval lists. No effects of version were identified on any metric, thus all results are collapsed across versions. Each encoding phase consisted of 40 face-scene pairs presented centrally on the screen for 4 seconds each. For each pairing, participants were asked to rate how well the face/person fit with the scene/location on a scale of 1 (poor fit) to 4 (good fit). They were also informed that they would have a subsequent memory test following the study phase. They were instructed to base their rating on how likely the person might be to live, work, vacation, or do some other activity in the pictured location. Each retrieval block followed immediately after encoding and consisted of 37 face-scene pairs randomly presented centrally on the screen for 4 seconds each. Sixteen of the pairs consisted of exact face-scene pairings from encoding (i.e., targets), 16 of the pairs were recombinations of a face and scene from encoding (i.e., recombined lures), and 5 of the pairs were entirely new face-scene images (i.e., unrelated pairs). 

To investigate the effects of similarity, we manipulated the degree to which the recombined lure pairs induced interference. Critically, in half of the lure pairs the recombination consisted of a scene taken from the same category as that seen in the original encoded pairing (high within-category similarity; 8 per retrieval block), and in the other half of the lure pairs the recombination consisted of a scene taken from a different category as that seen in the original encoded pairing (low similarity; 8 per retrieval block). The unrelated lure pairs consisted of a novel face and a novel scene from a category that was not one of the four studied (e.g., bedroom; grocery store) and were therefore expected to induce no familiarity or interference as both face and scene were entirely novel at the retrieval phase. For each image at retrieval, participants were asked to make an ‘Old’/’New’ judgment. They were instructed to press a key indicating ‘Old’ if they remembered seeing that exact pairing at encoding and to press ‘New’ if they thought it was either a new combination of face-scene or an entirely novel pairing.

Results

Table 1 reports average hit and false alarm rates for Study 2 across categorical similarity conditions; see also Figure 2A for false alarm rates across conditions. A one-way repeated measures ANOVA was conducted to compare false alarm rates across the interference conditions (high, low, or no similarity). This analysis indicated a significant decrease in false alarm rates across the 3 levels of similarity [F(1.49, 38.96) = 107.20, p < .001, ηp 2 = 0.81; Greenhouse-Geisser corrected]. False alarms to novel lure pairs were significantly lower than that of both types of recombined lure pairs [high similarity: t(26) = 12.40, p < .001, d=0.20; low similarity: t(26) = 11.24, p < .001, d=0.11], indicating that participants made more false alarms to recombinations of studied lure information compared to novel lures. Important to our manipulation, false alarms were significantly greater when the lure scene was from the same category as that from encoding compared to when the scene was different from that which was originally studied [high vs. low similarity; t(26) = 6.68, p < .001, d=0.18].

Discussion

Results from Study 2 replicate and extend findings from Study 1 to an independent set of participants, using a new paradigm that compared high and low similarity within the same set of source contexts. The results further emphasize the negative effects of contextual interference on associative recognition within individuals. Consistent with the previous study, participants showed poorer associative memory, as evidenced by higher rates of false alarms, when the face was repaired with a unique scene drawn from the same versus a different context category as that with which it was originally paired. The differences between the high and low similarity conditions in Study 1 and Study 2 were comparable (0.17 and 0.24, respectively), providing evidence for the replication of findings across studies despite design differences. We conclude that, like Study 1, the high degree of similarity, or gist, in sources creates contextual interference in associative memory traces, contributing to the greater false endorsement of recombined lures in the high within-category similarity condition. Study 2 removed the potential confound of heterogeneity across scene similarity conditions. Importantly, we have, to this point, observed converging results both in the context of a between-subjects and within-subjects design and across age groups, further highlighting the pervasive influence of within-category similarity on memory for associative information.

Study 3

Introduction

The previous two studies showed that interference induced by within-category similarity negatively affects associative memory performance. However, in the traditional recognition paradigms featured in Studies 1 and 2, the lure pair was presented in isolation, which could potentially exacerbate the influence of gist-based retrieval processes. Studies suggest that under forced choice conditions (i.e., where the target and corresponding lure are presented simultaneously), memory judgments can be based on relative familiarity with the target, resulting in reductions in false recognition (Dennis et al., 2012; Migo et al., 2009). On the other hand, traditional Old/New recognition tasks can involve the recollection of details to successfully endorse a target as old and avoid false memories (Bowman & Dennis, 2015; Cohn et al., 2008; Rotello & Heit, 2000). The necessity to engage this type of strategy should be reduced in a forced choice relative to the standard memory paradigm as the target information is proximate at the time of retrieval. 

Evidence from (Guerin et al., 2012) indicates that, in younger adults, false alarm rates were reduced when using a two-alternative forced choice (2AFC) retrieval task compared to a traditional Old/New recognition task. In older adults, the use of a 2AFC test can reduce age differences in associative memory relative to standard recognition tests (Ahmad et al., 2014; Delhaye & Bastin, 2018). Simultaneous presentation of the original target pair along with the lure pair thus appears to provide retrieval support, facilitating accurate retrieval in both younger and older adults. Given this, Study 3 aimed to assess whether the presentation of original target details, concurrently with a similar lure, contributes to a reduction in contextual interference effects in younger adults. Additionally, this study aimed to determine whether target presentation at associative retrieval influences memory across scene similarity conditions differentially across ages.

To achieve this, the present study utilized a 2AFC design whereby, at retrieval, younger and older participants saw both the target and recombined lure simultaneously on the screen and were instructed to decide which of the face-scene pairs was originally studied during the encoding phase. In line with the previous two studies, we created recombined lures that featured a scene from the same studied category or a scene from one of the other studied categories. However, at retrieval in the present study, the high within-category similarity condition entailed the presentation of the original, intact face-scene pair alongside a same-category lure and the low similarity condition included the intact face-scene pair presented alongside a different-category lure. Because the presentation of the intact pair should induce higher familiarity for the original pair than the recombined pair, contextual interference effects should be reduced relative to what was observed in Study 2 (i.e., the associative memory difference between the high and low similarity conditions should be lessened), or even eliminated. Furthermore, we hypothesized that both older and younger adults should be able to capitalize on intact familiarity-based processing for retrieval support. This, therefore, should reduce the influence of gist and result in little difference in associative recognition under conditions of high relative to low scene similarity across both age groups. Alternatively, for older adults, gist influences induced by high within-category similarity may still override any benefit of target familiarity and the simultaneous presentation of target details, resulting in age differences in the interference condition effect.

Methods

Participants—Thirty-four younger adults (mean age = 18.58, SD = 1.00; mean education= 12.47 years) and 34 older adults (mean age = 71.56, SD = 6.86; mean education= 18.24 years) completed Study 3. Younger adult participants were recruited from introductory psychology classes at Penn State University and older adults were recruited from the local community surrounding Penn State University. On the day of the study, participants provided written informed consent for a protocol approved by the Pennsylvania State University institutional review board. Older adults completed the Mini-Mental Status Exam (M=29.73, SD=0.72) to screen for neurological illness. A posthoc sensitivity analysis conducted in G*Power 3.1 (Faul et al., 2007) confirmed that power was sufficient to detect a small effect size (f=0.32) for within-between interaction (age x similarity condition) given that α=.05, and the correlation between repeated measures is equal to 0.5.

Materials—All stimuli were the same as in Study 2.

Procedure—The task procedure included three encoding-retrieval blocks and is depicted in Figure 2B. Each encoding block consisted of 32 face-scene pairs presented centrally on the screen for 4 seconds each. Stimuli were randomized within each run of the task. For each pairing, participants were asked to rate how well the face/person fit with the scene/ location on a scale of 1 (poor fit) to 4 (good fit). Each retrieval block followed immediately after encoding and consisted of a typical 2AFC recognition task, during which 16 sets of face-scene pairs were presented. For each trial, two face-scene pairs were presented next to each other on the screen, with one always consisting of the target pair and the other depicting a recombined lure or a novel lure. As in Study 2, we manipulated the degree to which the recombined lure pairs induced contextual interference. Critically, in half (8) of the trials, the target was presented next to a within-category recombined pair, inducing high within-category interference, and in the other half of trials, the target was presented next to a between-category recombined pair, inducing less contextual interference. In each case, the face always remained the same across both the target and lure and it was the scene that changed (either within- or between). The presentation of the target and lure was counterbalanced across the left and right sides of the screen. For each retrieval trial, participants were asked to indicate whether the left or right pair was ‘Old’, meaning it was the exact pairing they saw at encoding. Each trial was self-paced to allow for greater time to examine both pairs and make a recognition decision (mean RT: younger adults = 2.52s; older adults = 3.58s)3.

Results

Table 1 reports average correct and incorrect responses for Study 3 (see also Figure 2B). A mixed model ANOVA with a between-subjects factor of age (young, old) and a withinsubject factor of similarity (high, low) predicting the proportion of incorrect responses (akin to false alarms in old/new recognition paradigms) showed a main effect of age group [F(1, 66) = 15.61, p < .001, ηp 2 = 0.19], with older adults showing greater inaccurate recognition than younger adults across context type, and a main effect of similarity [F(1, 66) = 11.83, p = .001, ηp 2 = 0.15], such that across age accuracy was poorer when the presented lure scene context was drawn from the same category as the target context compared to when the lure context was from a different category than the target context. There was no interaction between age and similarity [F(1, 66) = 0.30, p = .59, ηp 2 = 0.004], suggesting that as in Study 1 contextual interference effects were similar across age.

To assess whether presentation of the originally studied target pair facilitated associative retrieval, false alarm rates, and incorrect response rates across study version and high and low similarity conditions were compared in younger adults using a repeated measures ANOVA with a between-subjects factor of study (Study 2, Study 3) and a within-subject factor of similarity (high, low) predicting memory. This analysis revealed both a main effect of study [F(1, 59) = 12.59, p < .001, ηp 2 = 0.18], and a main effect of similarity [F(1, 59) = 68.52, p < .001, ηp 2 = 0.54], whereby incorrect responses across context conditions were higher under Old/New recognition compared to 2AFC conditions and a higher proportion of inaccurate responses were made in the high versus low similarity conditions. There was also a study by context type interaction on inaccurate responses (either false alarm rate in Study 2 or incorrect response rate in Study 3) [F(1, 59) = 18.89, p < .001, ηp 2 = 0.24], suggesting that the difference in associative memory in the high and low similarity conditions under the Old/New recognition task (Study 2) was significantly greater than the difference in conditions in the 2AFC task (Study 3). This was driven by a higher false alarm rate between Study 2 and 3 under conditions of high within-category similarity, with a similar rate across studies under low similarity, highlighting the beneficial influence that re-presentation of intact target pairs has on opposing gist interference induced by scene similarity.

Discussion

Similar to results in Studies 1 and 2, individuals across age groups in Study 3 produced more associative memory errors when there was a high degree of contextual similarity between the lure and the target relative to a condition in which similarity between lure and target was absent). This was true even though the originally studied target pair was presented simultaneously with the lure during the retrieval task, allowing for the reinstatement of the encoded association. However, a comparison of younger adults’ associative memory performance across Old/New and 2AFC study designs (Study 2 and Study 3, respectively) indicated that the within-category interference effect was reduced in the forced choice task relative to the traditional recognition test. Specifically, the difference in high and low similarity conditions was reduced when target information was presented concurrently with the associative lure. The influence of gist induced by high within-category similarity may be diminished by the presentation of original target information, however, caution is suggested as the comparison was conducted across experiments. Target presentation at the time of retrieval may prompt a familiarity signal (Jones & Jacoby, 2001; Migo et al., 2009) that is, in part, able to override the gist induced by high within-category similarity in associative lures, thereby reducing the difference in contextual interference conditions. While older adults again demonstrated typical associative memory deficits relative to younger adults, we saw evidence for a reduction in contextual interference effects under 2AFC conditions in both younger and older adults. This suggests that individuals across the lifespan may benefit from this type of retrieval support; however, future research should include within-subject comparisons across age and task design to better understand the magnitude of this effect. The fact that the contextual interference effect was not eliminated in either age group (i.e., still significantly greater associative errors under high vs. low similarity conditions) demonstrates that contextual interference exhibits a strong and pervasive influence on memory for associative information that is not easily overcome. Yet, the present study suggests that this effect may be reduced by reinstating details of the original episode.

General Discussion

Across three studies, each incorporating different associative memory paradigms, results point to a role of similarity in evoking false memories. Specifically, we saw an increase in associative false memories of recombined lures when the lure scene shared the same categorical membership as the original scene from encoding. We posit that such similarity among scenes in a category induces interference that contributes to reduced fidelity of associative memories. Importantly, results from these studies suggest that this category similarity effect is not age-specific, given that both older and younger adults experience similar susceptibility to contextual interference with high similarity contexts in associative memory tasks. Lastly, we found that while the presentation of original target information at retrieval may reduce this effect, it does not eliminate it in either age group. Taken together, results underscore the pervasiveness of category similarity and gist influences on associative memory across ages.

Effects of Retrieval Support on Associative Memory

Prior work has demonstrated that presenting target information at the time of recognition decisions (e.g., under 2AFC conditions) can provide retrieval support by reducing the erroneous endorsement of lure information compared to that seen in traditional recognition paradigms (Ahmad et al., 2014; Delhaye & Bastin, 2018; Guerin et al., 2012; NavehBenjamin & Craik, 1995). The current set of results suggests that, in cases of high similarity, such informational support may not be enough to overcome the mnemonic overlap across memory representations. While the individual items in the lure pairs likely evoked a strong feeling of familiarity (as they were both presented at encoding), target pairings uniquely included additional and specific associative information amongst individual items. We hypothesized that the re-presentation of the target pairing (and this added information) alongside the lure pair would evoke an increased level of familiarity with both the individual items and associative-specific information (Huffman & Stark, 2017; Migo et al., 2009). We posited that this would consequently reduce the occurrence of false endorsement of lure pairs containing highly similar contexts. While we continued to see a significant difference across similarity conditions in Study 3, we also observed evidence supporting this hypothesis when comparing performance in younger adults between Study 2 and Study 3. Specifically, we found a significant reduction in category similarity effects under 2AFC task conditions compared to the recognition task in Study 2 (and additionally saw an observational reduction in false alarms when comparing older adult performance across Study 1 and 3). That is, participants may have been better able to reject a lure pair that included a high degree of overlap in context under circumstances where the original studied pair was concurrently presented compared to when the lure pair was presented alone. We posit that, with the simultaneous presentation of targets and lures, external target information may be more readily utilized to better support successful associative memory.

Previous studies comparing alternative forced choice with single-item recognition have shown similar findings. For example, in a study by (Guerin et al., 2012) lure endorsement was greatly decreased when target information was made accessible at the time of retrieval decision. Research on source memory additionally suggests that reinstating context information acts as a cue for accurate retrieval (Dodson & Shimamura, 2000; Smith & Vela, 2001). Aging studies have also shown the benefits of environmental support to successful associative memory in older adults (Ahmad et al., 2014; Delhaye & Bastin, 2018; Schacter et al., 1999). For example (Naveh-Benjamin & Craik, 1995) showed that both younger and older adults’ memory can similarly benefit from reinstating original context information. When item-specific information (i.e., the exact studied information) is available at the time of memory retrieval, the gist induced by the overlap in similarity is reduced, and this benefits both younger and older adults’ associative memory. Importantly, while the present evidence indicates that contextual interference may be reduced, the effect is not eliminated. This lends further evidence highlighting the pervasive influence of gist interference across age.

Within-Category Similarity Effects Between Age Groups

In addition to being present across different study conditions, comparable similarity effects were also evident within both younger and older adults. Specifically, evidence from Study 1 and Study 3 indicated that despite typical associative memory deficits (Kilb & NavehBenjamin, 2011; Naveh-Benjamin et al., 2003; Old & Naveh-Benjamin, 2008; Overman & Becker, 2009), older adults showed comparable sensitivity to interference from within category similarity as younger adults. The absence of age differences in false memories stands in contrast to analogous studies of item memory where older adults typically show similar target recognition, but overall higher false alarm rates to related lures, compared to younger adults. This pattern of behavior has been attributed to a greater reliance on the gist inherent in related information, as well as a greater dependence on general familiarity when making recognition decisions (Anderson et al., 2008; Brainerd et al., 2008; Brainerd & Reyna, 2002; Jones & Jacoby, 2001; Koutstaal & Schacter, 1997). Based on this past work, we anticipated that older adults’ associative memory would be disproportionately negatively affected by high within-category similarity (Greene & Naveh-Benjamin, 2020; Overman & Stephens, 2013), compared to younger adults who more readily retrieve item-specific information (Cohn et al., 2008; Jennings & Jacoby, 1997; Yonelinas, 1997). In line with this prediction, recent work by (Greene & Naveh-Benjamin, 2020) showed that age-related associative memory deficits were influenced by categorical similarity across both faces and scenes in an associative memory task similar to that used in the current set of studies. However, in contrast to their findings, the present results suggest that when faced with within-category similarity alone (absent of additional item (face) similarity), both younger and older adults are unable to distinguish between highly similar information and rely on gist-based processing, which leads to the erroneous endorsement of a lure as a target.

Despite the common age-related increases in false memories observed across the literature, item memory studies have not found differences across age groups when manipulating the degree of gist inherent in lure information. When target-lure similarity is high, over-reliance on the gist trace is shown to contribute to false memories for related information in a similar manner across both younger and older adults (Dennis & Turney, 2018; Giovanello et al., 2009; Webb & Dennis, 2019). The present results extend these findings to an associative memory context, demonstrating that associative recognition is negatively influenced by shared context information and that this overlap in similarity across target and lure pairs similarly influences associative memory across aging. At the same time, both age groups show proportionate benefits to associative memory recognition when lure information is more distinct from studied information, as was the case when there was no similarity across target and lure scenes. This adds additional evidence to previous research which suggests that making information more distinct benefits recognition at any age (Brainerd et al., 1995; Dodson & Schacter, 2001; Schacter et al., 1999). We extend this to an associative memory context and show that this distinctiveness benefit occurs similarly across age groups. The present results further tie into the broader memory literature which suggests that unique mnemonic representations are poorly maintained and/or preserved in memory as a result of overlapping, competing information. Here, we demonstrate that within-category similarity across categories contributes to reduced mnemonic discrimination (i.e., poorer separation of similar memory traces), resulting in a greater tendency toward generalization of memory traces as evidenced by higher rates of associative false memories.

Conclusions

The present set of experiments demonstrated that mnemonic overlap between targets and lures in an associative memory task can be detrimental to the ability to discriminate between highly similar target and lure information. Specifically, we found that shared categorical similarity between targets and lures led to increased false associative memories. This effect occurred regardless of age, indicating that despite typical age-related associative memory deficits and false memory increases, younger and older adults are equally susceptible to interference from within-category similarity. Our results also suggest that contextual interference effects may be reduced through retrieval support, such as providing target information alongside lure information, but this should be investigated further as we are limited in interpreting cross-experiment comparisons. These results have important implications for real-world situations where interference between pieces of information is high, such as remembering the relationship between critical medications and their dosages, or distinguishing where you parked your car in a given parking lot. The present set of findings suggests that these types of associative memory distinctions are difficult for both younger and older adults and that being able to prompt retrieval of target information can reduce, but not remove, the influence of interference due to similarity on memory success.

Acknowledgments

We thank Jordan Goodman, Haley Iriondo, Holly Richardson, Harini Babu, Jackie DeRosa, Kayla McGraw, Andrew Rowley, and Joanna Salerno for help with data collection and analyses. This work was supported by the National Institutes of Health under grant R15AG052903 awarded to A.A.O. and N.A.D. In addition, N.A.D. was also supported in part by National Science Foundation grants BCS1025709 and BCS2000047. Portions of the research in this article used the Color FERET (Facial Recognition Technology) database of facial images collected under the FERET program, sponsored by the Department of Defense Counterdrug Technology Development Program Office.

References

1. Ahmad FN, Fernandes M, & Hockley WE (2014). Improving associative memory in older adults with unitization. Quarterly Journal of Experimental Psychology, 67(12), 2301–2324.

2. Anderson ND, Ebert PL, Jennings JM, Grady CL, Cabeza R, & Graham SJ (2008). Recollection and familiarity-based memory in healthy aging and amnestic mild cognitive impairment. Neuropsychology, 22(2), 177–187. 10.1037/0894-4105.22.2.177 [PubMed: 18331160]

3. Bender AR, Naveh-Benjamin M, & Raz N (2010). Associative deficit in recognition memory in a lifespan sample of healthy adults. Psychology and Aging, 25(4), 940–948. 10.1037/A0020595 [PubMed: 20822256]

4. Bowman CR, & Dennis NA (2015). Age differences in the neural correlates of novelty processing: The effects of item-relatedness. Brain Research, 1612, 2–15. 10.1016/J.BRAINRES.2014.08.006 [PubMed: 25149192]

5.Brainerd CJ, & Reyna VF (2002). Fuzzy-trace theory and false memory. Current Directions in Psychological Science. 11(5). 10.1111/1467-8721.00192

6.Brainerd CJ, Reyna VF, & Kneer R (1995). False-recognition reversal: When similarity is distinctive. Journal of Memory and Language, 34(2), 157–185. 10.1006/jmla.1995.1008

7.Brainerd CJ, Yang Y, Reyna VF, Howe ML, & Mills BA (2008). Semantic processing in “associative” false memory. Psychonomic Bulletin & Review, 15(6), 1035–1053. 10.3758/PBR.15.6.1035 [PubMed: 19001566]

8. Bryce M, & Dodson C (2013). Cross-age effect in recognition performance and memory monitoring for faces. Psychology and Aging, 28. 10.1037/a0030213

9.Buchler NEG, & Reder LM (2007). Modeling age-related memory deficits: A two-parameter solution. Psychology and Aging, 22(1), 104–121. 10.1037/0882-7974.22.1.104 [PubMed: 17385988]

10. Castel AD (2005). Memory for grocery prices in younger and older adults: The role of schematic support. Psychology and Aging, 20(4), 718–721. 10.1037/0882-7974.20.4.718 [PubMed: 16420146]

11. Castel AD, & Craik FIM (2003). The effects of aging and divided attention on memory for items and associative informa

12.Castel AD, Farb NAS, & Craik FIM (2007). Memory for general and specific value information in younger and older adults: Measuring the limits of strategic control. Memory & Cognition 2007, 35(4), 689–700. 10.3758/BF03193307

13. Chalfonte B. l., & Johnson MK (1996). Feature memory and binding in young and older adults. Memory & Cognition 1996, 24(4), 403–416. 10.3758/BF03200930

14. Coane JH, McBride DM, Huff MJ, Chang K, Marsh EM, & Smith KA (2021). Manipulations of list type in the DRM paradigm: A review of how structural and conceptual similarity affect false memory. Frontiers in Psychology, 12. 10.3389/fpsyg.2021.668550

15. Cohn M, Emrich SM, & Moscovitch M (2008). Age-related deficits in associative memory: The influence of impaired strategic retrieval. Psychology and Aging, 23(1), 93–103. 10.1037/0882-7974.23.1.93 [PubMed: 18361659]

For more information:1950477648nn@gmail.com