Comparison versus reminding

Participants accurately connected proverb pairs that shared a deep meaning. Participants were reminded more often of prior proverbs that shared the same meaning than prior unrelated instances. Participants in the comparison condition created generalizations more frequently when the proverbs shared the same meaning. Participants understood the deep meanings of the proverbs and often relied upon them to make connections between superficially different proverbs.

Participants in the reminding condition remembered more proverbs than those in the comparison condition. Prompting learners to think back through the study list and retrieve prior related proverbs produced better memory than prompting learners to compare two proverbs. Further, participants in the reminding condition were more efficient as they recalled more items, but spent less overall time studying the proverbs. The mnemonic benefits in the reminding condition are particularly prominent for the first items in related pairs. The mnemonic advantage for related P1s over unrelated P1s replicates research using different materials and contexts and has been labeled the “reminding effect” (Tullis, Benjamin, et al., 2014). The specificity of the benefits to P1 (and not P2) likely arose because the effortful retrieval of P1 during P2 enhanced later retrieval of P1.

An alternative explanation of the benefits seen in the reminding condition is that participants may have encoded the proverbs more deeply initially and the final test reflects differential encoding practices. The results of Experiment 1 suggest that reminding instructions may contribute to the mnemonic differences between conditions, but also shows that the reminding condition entails more than just deeper initial encoding. Participants in Experiment 1 remembered more unrelated proverbs than participants in the comparison condition, which supports the idea that reminding instructions improved overall encoding. This is particularly interesting because comparison participants spent numerically more time studying the proverbs than the reminding participants. However, the specificity of mnemonic benefits to related P1s shows that reminding instructions do not prompt learners to more deeply encode all the stimuli as compared to the comparison condition. Remindings specifically benefit memory for the first instances in related pairs, which suggests that the large mnemonic consequences of remindings arise from retrieval of the first presentation during encoding of the later related episode.

We find no evidence that comparison promotes forgetting of individual instances. In fact, we find some evidence that comparing related items promotes memory for those items more than comparing unrelated items. Participants in the comparison condition mostly created generalizations across related pairs and more items from related pairs were remembered than from unrelated pairs. Comparing across similar instances seemed to enhance memory for the individual items in a comparison more than across dissimilar instances. This suggests that participants are encoding the individual instances in related pairs, even though they create a generalization. However, the mnemonic benefits of comparison are much smaller than those of reminding.

Experiment 2 replicated and extended Experiment 1. During encoding, learners in the reminding group were able to successfully connect two superficially different proverbs that shared the same deep meaning, as indicated by the high rate with which P2 reminded learners of P1. Further, reminding led to better memory performance than comparison, even after a one-week retention interval. Participants in the reminding condition were better able to discriminate between studied and unstudied proverbs than participants in the comparison condition. Comparison led to lower hit rates of studied proverbs and higher false alarms to all unstudied proverbs. Participants in the reminding condition were also more efficient at their memory processes; they spent less time but had better recognition performance than those in the comparison condition.

If comparison created abstracted knowledge about the principles put forth by proverb pairs, participants in the comparison group would have shown selectively higher false alarms to unstudied but related proverbs than the reminding group. Some evidence supports this hypothesis. A 2 (condition) × 2 (type of false alarm) ANOVA on proportion of false alarms revealed a significant interaction (F(1, 67) = 4.42, p = 0.04, η_p ² = 0.06) and main effects of condition (F(1, 67) = 5.50, p = 0.02, η_p ² = 0.07) and type of false alarm (F(1, 67) = 32.46, p < 0.01, η_p ² = 0.33). The interaction indicates that comparison produced selectively more false alarms to unstudied, but related proverbs than did reminding. Comparison, then, may promote memory for generalizations of meaning across proverbs at the expense of memory for individual instances. However, we must be cautious with the interpretation of these data because scaling issues (i.e. floor effects) may be driving this interaction.

The results of Experiments 1 and 2 show that reminding supports better memory performance than comparison. In the reminding group, learners had to actively think back and retrieve prior related episodes. Remindings prompted learners to practice retrieval of a prior episode at a time when the prior episode could be recalled, re-exposed learners to the earlier episode, and supported long-term retention of that information. Comparison may have emphasized generalization at the expense of memory for the specific episodes. Results seem to suggest that participants in the comparison condition focused on abstracting a generalization from the two proverbs without encoding each proverb deeply.

Participants in the reminding condition more appropriately generalized and transferred their experiences to a new situation than participants in the comparison group. Comparison did not increase the hit rate to proverbs related to studied items, but almost doubled the false alarm rate to items that were not related to studied items. Consequently, comparison resulted in worse discrimination between proverbs that were related to studied items and proverbs that were unrelated to studied items. Comparison also resulted in a more liberal criterion for participants’ endorsement. Comparison did not support participants’ encoding of the studied principles or examples and seems to have prompted learners to overgeneralize the abstractions they created during the encoding phase. Reminding led to better transfer of studied principles to new situations and contexts than comparison. Participants in the reminding performed better on the transfer task than participants in the comparison condition. This is especially surprising because the study and test conditions were more similar for the comparison group than the reminding group: participants in the comparison group typed in themes during study and identified old proverb themes during the final test, while participants in the reminding condition typed in specific proverbs during study but identified old proverb themes during the final test. It is compelling that participants in the comparison group studied more similarly to how they were tested, but performed worse on the final task than participants in the reminding group.

Learners could have succeeded in the transfer task by remembering the general principles they abstracted from individual instances or by remembering and generalizing an individual instance. Research suggests that learners can succeed in new situations by applying generalized abstract knowledge (Gick & Holyoak, 1983) or by applying a specific prior instance (Ross, 1984). Therefore, remembering distinct prior episodes may be advantageous when trying to transfer knowledge to a new situation, as long as those prior episodes are not too tightly connected to the prior context. Reminding through structural similarities may promote both memory for the individual instances and the general principles that connect different episodes. Participants must abstract the deep meaning of the proverbs to create connections between them and the retrieval of the earlier episode during later presentations likely supports memory for the instances and the generalization. These results suggest that, relative to reminding, comparison may only support immediate generalization and not long-term memory or transfer performance. The generalization that participants create during comparison (and the individual instances compared) may be forgotten more quickly than during reminding so that participants perform poorly on later memory and transfer tasks.

Across three experiments, we compared the mnemonic and transfer consequences of comparison with those of reminding. In the first two experiments, reminding led to better memory for the individual instances than comparison. However, Experiments 1 and 2 suffer from the limitation that learners in the reminding group may have been using similar mental processes involving memory during both encoding and test. Because the mental processes during study and test may be more similar for the reminding group than comparison group, it may not be surprising that encoding through reminding is more beneficial for memory than encoding through comparison. In the last experiment, though, reminding showed cognitive benefits beyond just pure memory. Experiment 3 showed that reminding led to better generalization and transfer of that knowledge to a new situation. Learners who were in the reminding group more appropriately recognized proverbs that shared the same meaning as ones they had studied. Reminding not only supported memory, but also fostered appropriate transfer to new situations.

Comparison has been described as one of the most effective techniques to create generalized, transferrable knowledge (Gentner & Markman, 1997; Hummel & Holyoak, 1997). Yet, here we show that remindings led to better transfer than comparison. Our reminding task employs several well-documented cognitive principles to result in the mnemonic and transfer benefits showcased across the three experiments. First, we prompted learners to actively think back to proverbs that shared similar structural meanings with the current proverb, rather than earlier proverbs with similar superficial features. This likely encouraged learners to encode each proverb according to its deep meaning so that they could make connections between the current instance and future structurally similar instances. Encoding the proverbs in structural ways makes learners more likely to be able to later retrieve the example when they encounter a later analogous item (Loewenstein, 2010). Further, telling learners that prior instances are related to current items dramatically increases how much learners use prior examples to solve current ones (Gick & Holyoak, 1983). By giving learners a “hint” to look back through the study list, we likely increased the number of connections made between superficially different but deeply related proverbs. As learners look back through the study list, they bring together physically and temporally disparate events, the events become mentally contiguous, and learners can generalize across them.

Second, the reminding procedure forced learners to effortfully think back to and retrieve similar instances. The effortful retrieval of earlier episodes largely benefits memory for the retrieved episode (Roediger & Karpicke, 2006). The mnemonic benefits, especially for related P1s in the reminding condition, are likely due to practice retrieval. The presentation of a related P2 allows learners to practice retrieving P1 before P1 is completely forgotten. Successful practice retrieval of P1 ensures that memory for that prior item will persist over longer periods of time (Roediger & Karpicke, 2006). Practice retrieving has even been shown to benefit transfer of knowledge to new situations, tests, and contexts (Butler, 2010; Jacoby, Wahlheim, & Coane, 2010; Rohrer et al., 2010). Remindings, then, likely engender practice retrieval and, therefore, support learners’ memories of the exact studied items and learners’ generalization of studied information to new situations. Further, the separation of instances required for remindings likely takes advantage of the spacing effect, whereby repetitions presented spaced apart in time are remembered better than repetitions presented massed together (Benjamin & Tullis, 2010; Hintzman, 1974).

Third, our reminding procedure involved interleaving exemplars from different pairs. In the reminding condition, the structurally related pairs were separated by an intervening item from a structurally different pair. Interleaving exemplars from different categories can lead to enhanced memory for the items and superior category learning (Kornell & Bjork, 2008; Rawson, Thomas, & Jacoby, 2015). The benefits of interleaving exemplars from different categories is thought to arise because learners can better discriminate among categories when they are intermixed (Birnbaum, Kornell, Bjork, & Bjork, 2013; Kang & Pashler, 2012) to the degree that the categories are similar to each other (Carvalho & Goldstone, 2014). Reminding intermixes stimuli so that learners can compare and contrast between different pairs of proverbs, while the comparison process reduces the likelihood that learners compare across pairs. Interleaving, then, may cause the generalizations that arise through the reminding process to be appropriately broad.

Making physically separate events mentally contiguous through controlled memory search may be important to form complex memory traces involving both events (Jacoby, 1974). Hintzman (2011) argues that these complex memory traces incorporate the earlier stimuli into the representations of the later episode through recursive reminding and this incorporation results in the mnemonic benefits seen for related items. However, making related proverbs contiguous cannot be the fundamental cause of the benefits of reminding. Related proverbs in the comparison condition were both mentally and physically contiguous, yet were not remembered as well or transferred as appropriately. It seems likely, then, that the benefits of the practice retrieval that enables events to become mentally contiguous underlies the significant mnemonic benefits apparent in remindings.

Remindings may be most beneficial when learners can extract the deep structure of episodes and connect across superficially different instances. Comparison may be more beneficial in situations where learners need more explicit prompting to connect across very different individual instances. Further, comparison may be more beneficial with complex stimuli. In most applications of comparison during learning, the instances are presented simultaneously, so the learner does not need to occupy working memory with the retrieval of a prior episode during the presentation of a second. However, in remindings, learners need to retrieve the prior instance into working memory during presentation of the second item. Remindings, then, may require more working memory than comparison. As the to-be-learned information becomes more complex, space in working memory is likely at a premium and remindings may require more working memory than is available.

Many factors varied across the reminding and comparison conditions. Alternative experiments could have more rigidly controlled the similarities between the two conditions to isolate the factors causing differences between the conditions; however, controlling for those differences obscures the naturalistic integrity of the two conditions for real-world applications and implementation. For example, related proverbs in the reminding condition were separated by intervening unrelated proverbs while related proverbs in the comparison were simultaneously presented. Attempting to control for the simultaneous versus sequential presentation of exemplars would weaken the external validity of our procedures. The conditions we selected help answer the real-world question: if you have a limited amount of time to study related material, should you capitalize on the benefits of reminding or comparison? These results suggest that the benefits of reminding can outweigh those of comparison, at least for these stimuli.

How well the mnemonic and generalization benefits of reminding apply to different category structures and more complex materials, like complicated physics problems, remains unknown. Learners, especially novices, do retrieve and use prior problems when attempting to solve novel problems (Pirolli & Anderson, 1985; Reed, Dempster, & Ettinger, 1985). Learners need to recognize the deep structure of the problem before they can connect current problems with structurally similar prior problems. Without understanding structural similarities, learners often connect superficially similar problems that may not share deep structure (Ross, 1984). Further, the kind of information that is recalled from the prior problem during the presentation of the second likely changes what information is ultimately remembered and generalized. If learners primarily recall superficial features, without accessing deep structure, only memory for those superficial features may be enhanced. Ross (1987), however, suggests that remindings across problems can support access to the formula used to solve the prior problem. If an appropriate reminding supports retrieval of the formula, memory for that formula would likely be strongly enhanced.

The relationship between related exemplars and among the different related pairs may also moderate the advantages of remindings over comparison. Category learning research has examined how with-in and between-category similarity affects the advantages of comparison. Learners benefit from actively contrasting between confusable categories, but do not benefit much from contrasting distinct categories (Carvalho & Goldstone, 2014; Kornell & Bjork, 2008). For example, presenting two exemplars from confusable categories simultaneously can support memory for those exemplars and classification of later novel category members because it enables learners to discriminate between categories (Wahlheim, Dunlosky, & Jacoby, 2011). Here, however, we may find no mnemonic advantage of presenting exemplars from different categories together on the screen, as compared to presenting exemplars from the same category together or to presenting exemplars sequentially. The structure of our categories (each pair of related proverbs) may be low in with-in category similarity but high in between-category discriminability, such that comparison does not benefit category learning. Further, the structure of categories may determine the relative effectiveness of comparison and reminding. The benefits of remindings are likely limited by the probability of being reminded across instances. If categories have low within-category similarity, remindings may be very unlikely to occur and comparison may result in better learning and transfer than remindings.

Reminding may prove to be a desirable difficulty (Schmidt & Bjork, 1992), as compared to comparison. Comparison may allow for learners to abstract general principles from individual instances without much mental effort. With reminding, learners have to effortfully retrieve a prior related instance during the presentation of a temporally distant instance. This effortful retrieval supports memory for the first instance in a related pair (Roediger & Karpicke, 2006; Tullis, Benjamin, et al., 2014). Further, retrieving the first instance during the presentation of the second allows the learner to compare across the instances and create generalized knowledge across the episodes. Remindings seem to both enhance memory for the first presentation and generate abstracted knowledge across presentations and thereby influence both what is learned and how well it is learned.