Adaptive Memory: Story Processing Improves Recall Performance

Abstract

Across four experiments, participants were asked to process unrelated common nouns after listening to pleasantness processing instructions, survival processing instructions, or story processing instructions. In Experiment 1, participants were provided with a list of words and were asked to list various ways that the words were pleasant or unpleasant, list ways that the words could help them survive in an ancestral grasslands scenario, or use the words in the list to create a short story. In Experiment 2, participants were asked to perform similar tasks, but rather than being provided with a word list at the start of the experiment, words were presented one at a time at a fixed rate. In Experiment 3, participants were asked to rate words as they were presented one at a time regarding their perceived pleasantness, their relevance to an ancestral survival scenario, or the ease with which they could be included in a short story. In Experiment 4, a combined survival-story condition was included along with survival, story, and pleasantness processing instructions to determine whether combining story and survival processing produces an additive effect on recall performance. Across all four experiments, after a brief delay, the story processing condition produced either enhanced or comparable incidental recall performance to the survival processing scenario, and both of those conditions produced enhanced recall performance relative to the pleasantness processing scenario. These results suggest that processing information in the context of a story, like survival processing, may represent one of the best deep processing tasks identified to date.

The 2004 Indian Ocean earthquake, with an epicenter near Sumatra, triggered a series of deadly tsunamis that devastated coastal communities and killed more than 200,000 people. However, although other surrounding areas suffered severe death tolls (e.g., more than 100,000 casualties in Indonesia alone) the people living on the island of Simeulue, located less than 25 miles from the epicenter of the quake, largely avoided this gruesome fate. The fact that only seven people died on an island with a population of around 78,000 has not been attributed to an advanced warning system that quickly and efficiently informed residents of the impending danger, but instead to a more primitive method of information conveyance. In fact, experts suggest that a warning emitted from a technologically sophisticated system would have been too slow to allow residents of the island to avoid catastrophe. Instead, emergency planning steps regarding tsunami survival strategies had been imparted to residents by embedding them in lullabies and stories that they were familiarized with as children and retold as adults. When the ocean quickly receded in the aftermath of the earthquake, such stories may have provided both a warning and a plan of action to help guide residents to safety on higher ground (Rahman et al., 2018).

Communicating fitness-relevant information through stories, as in the example above, has long been known to represent one of the oldest and most universal means of transmitting information in human societies (Fisher, 1984). Narrative communication has been found to facilitate the transmission of cultural values (Imada & Yussen, 2012) and can even be used to help convey complicated scientific concepts to lay audiences (Dahlstrom, 2014). The purpose of the present study was to evaluate the degree to which creating a story using a list of unrelated nouns influenced participants’ subsequent free recall performance for the words. More specifically, the goal was to compare such ‘story processing’ with two other well-known deep processing tasks, namely, pleasantness and survival processing. A set of incidental learning instructions, like those employed in traditional levels of processing experiments (e.g., Craik & Tulving, 1975), were used to compare the effectiveness of these various forms of processing.

Over the years, memory researchers have analyzed the effectiveness of using stories to improve memory performance. For example, in one study designed to determine the effectiveness of creating stories on improving subsequent recall performance, Bower and Clark (1969) had participants study 12 lists each comprised of 10 unrelated nouns in one of two conditions. In one condition, the instructions indicated that participants should study and learn the words in each list to the best of their ability. In the other condition, participants were asked to use the words in each list to create a meaningful story by incorporating the words into a story in the order in which they were presented. Participants serial recall for the lists was tested both immediately following the presentation of each list and then again once all 12 study / test trials were completed. Although Bower and Clark observed ceiling effects for the recall tests immediately following the presentation of each list, as participants rarely failed to remember presented words, a marked difference in performance was observed on the final recall test covering all the presented information. On that final test, Bower and Clark reported the median percentages of words recalled as a function of serial position. Overall, the results indicated that the average of the median scores were 13% in the control condition and 93% in the condition in which participants were asked to create a story. Bower and Clark attributed this large observed difference in performance as a function of encoding condition to the effects of thematic organization. More specifically, they suggested that participants in the story condition were able to more effectively reconstruct the memory of the target words by focusing on the theme of each story that they created. They also noted that intrusions were very rare in the story condition, indicating an enhanced ability to discriminate between presented words and other participant generated words used in the story creation process.

In another study designed to analyze the effectiveness of storytelling on memory in a more applied setting, Oaks (1995) had participants learn information during a classroom session that explained a method of including media in instructional materials for classrooms. In one condition, participants listened to a traditional lecture whereas in another condition, the material was presented in the context of a story. Participants were subsequently tested over this information immediately after encoding the material, two weeks post instruction and again after a 5-week delay. Oaks compared participants’ memory as a function of the encoding condition to which they had been assigned for each of the three delay intervals and observed statistically significant differences across the groups. More specifically, participants in the story condition remembered more information than participants in the traditional lecture condition regardless of the amount of time that had passed since encoding. Oaks suggested that storytelling comprises a “natural” mnemonic technique, similar in effectiveness to other mnemonic techniques, like the keyword or pegword methods, in that they facilitate attention to, and the organization of the material participants are attempting to learn. In other words, creating a story instantiates a hierarchical semantic structure in which being reminded of a story's theme primes the initial sentence of the story and recall of individual sentences in the story, in turn, cue the specific to-be-recalled information. Thus, overall, Oak's research provides an applied example of how storytelling can be used to increase memory performance relative to more traditional memory improvement techniques.

Whereas Oak's (1995) study focused on the efficacy of story creation in aiding college students in recalling material presented during a lecture, such mnemonic enhancements as a result of storytelling have also been observed in older adults. Hill et al. (1991) had older adults study a 26-item list of unrelated concrete nouns after using the nouns to create a story, using the method of loci to encode the nouns, or in a control condition in which participants were not provided any training in the use of mnemonic techniques. The older adults’ memory was tested immediately after encoding the words, after a one-hour delay, and again after a three-day delay. Hill et al. observed statistically significant differences in free recall performance as a function of instructional condition across all three delay intervals. Although no differences in performance were found between the storytelling condition and the method of loci condition, both groups that received memory enhancement training outperformed those in the control condition regardless of the delay interval. Adding evidence to Oak's claim that storytelling represents not only an especially effective memory technique but, in function, a natural mnemonic device, Hill et al. noted that, during a pretest prior to any training in mnemonic techniques, no participants spontaneously relied on the method of loci although several participants reported using a story creation method to help them remember the list. Thus, even absent any form of memory improvement training, at least some participants spontaneously relied on storytelling to improve their recall performance for studied material.

As noted above, in at least several reported studies, when participants create stories during an initial encoding phase, their subsequent memory for the studied material markedly improves, relative to other well-known processing techniques. Despite this, the storytelling technique has not often been discussed in the context of the levels of processing literature (Craik & Lockhart, 1972). According to the levels of processing framework, some encoding strategies elicit deeper semantic-based encoding whereas other, more shallow encoding strategies elicit structural processing. When participants engage in deep processing, their memory for studied material is likely to be enhanced relative to other participants who rely on more shallow processing strategies. For example, a participant asked to consider whether a list item can be used to complete a provided sentence will be more likely to remember that word on a later test than another participant who is asked to count the number of times the letter ‘a’ appears in the word. Whereas tasks like evaluating words based on their perceived pleasantness have long been known to improve memory relative to other more shallow tasks, Nairne and colleagues introduced the concept of processing information for its survival value (Nairne et al., 2007). Their functionalist approach is based on the premise that our memory systems may have been “tuned” to remember survival-related content thereby enhancing the effectiveness of the survival processing strategy.

In a 2007 paper, Nairne et al. had participants briefly rate unrelated common nouns based on their pleasantness, the ease with which they could be used to assist with a moving task, or for their value in helping to survive in a grasslands environment, specifically with respect to finding food and water and protecting oneself from predators. Across four experiments, the survival processing instructions led to better memory than other well-known deep processing tasks, including relating studied information to oneself or the self-reference effect. The survival processing advantage was observed on both free recall and recognition tests and when using both between- and within-subjects designs. Thus, overall, the results supported the contention that our memory systems may have evolved, at least in part, to remember survival-related content and that focusing on the survival value of to-be-remembered information can improve memory for such stimuli relative to other well-known deep processing tasks.

In another study designed to compare the effectiveness of survival processing to other well-known deep processing tasks, Nairne and colleagues (2008) had participants study a list of 30 unrelated common nouns in one of six conditions. During the experiment participants were asked to rate each of the studied words, one at a time, on their survival value, pleasantness, ease with which the word evokes a mental image, ease with which a word evokes a personal experience (self-reference), a generation condition in which participants were asked to mentally switch two letters to complete a word and then rate it for pleasantness and finally, an intentional learning condition in which participants were informed that they would be asked to study and remember each of the words on the list. Following this encoding stage, participants completed a free recall test and the results indicated that the participants in the survival condition remembered more words than participants in the other five conditions. Thus, the results supported the idea that survival processing is unique and represents one of, if not the most effective, deep processing task described to date. In an attempt to demonstrate that the superior recall performance in the survival condition was not a result of the thematic content inherent to the survival processing task, Nairne and colleagues had participants complete a within-subjects design in which they either rated words on their survival value or on their ability to enhance the enjoyability of a vacation (a non-survival-related processing task that nonetheless was designed to include thematic content). In this condition, like those described above, participants recalled more words in the survival condition than in the vacation condition.

Since the initial demonstration of the effectives of survival processing in improving memory performance, a number of studies have examined the specific circumstances in which survival processing is more or less effective as an encoding strategy. For example, Otgaar et al. (2010) showed that the survival processing advantage not only exists for words but can also be found using pictures. Similarly, Weinstein et al. (2008) compared participants’ ability to remember unrelated common nouns after listening to Nairne et al.'s survival processing task or a task that used almost identical wording but placed participants in a modern city environment (a city survival task) or a task in which they rated the relevance of words to a moving scenario. If the explanation for the survival processing advantage relied primarily on participants schematic processing, one might expect both the grasslands survival scenario and the city survival scenario to lead to similar recall levels. However, consistent with the evolutionary account, the grasslands survival scenario led to better recall than the other two scenarios.

Although the city survival task was almost identical in wording to the grasslands survival task, it is possible that the city scenario elicited lower levels of novelty for participants or induced lower levels of arousal during the encoding phase of the study. To compare the grasslands scenario with a comparable scenario in terms of arousal level and the novelty of the stimuli, Kang et al. (2008) had participants rate items for either their survival value or their relevance to conducting a successful bank heist. In their study, like those mentioned above, survival processing still led to both better recall and recognition performance than other processing conditions. Overall, the survival processing advantage has been found to be quite robust, leading to better retention than other processing scenarios whether participants are asked to imagine themselves or others in survival situations (Kang et al., 2008), whether the participants involved in a study are older or younger (Otgaar & Smeets, 2010), and even when more specific scenarios involving grasslands survival activities like hunting and gathering tasks are compared to other comparable non-survival scenarios (Nairne et al., 2009).

Given such findings, one might suspect that it would be relatively simple to identify factors that may have been relevant to survival in the preagricultural past and design fitness relevant scenarios other than the grasslands survival scenario that are equally effective in improving retention across a variety of conditions. Sandry et al. (2013) ran three experiments designed to compare the grasslands survival scenario with other fitness-relevant scenarios. They predicted that if the survival processing advantage could be replicated using other similar scenarios, then fitness-relevance rather than other factors like arousal level, novelty, or the facilitation of item-specific and relational processing contribute to the oft observed survival processing advantage. In their study, Sandry et al. (2013) created scenarios related to fear and phobia mechanisms, mate detection mechanisms, incest avoidance mechanisms, cheater detection mechanisms, jealousy mechanisms, and status mechanisms. Across three experiments, the survival scenario led to better recall performance than all the other encoding conditions included in the study. Thus, at least using the specific wording employed in their fitness-relevant scenarios, only survival processing led to improved performance relative to the other conditions. Notably, although creating stories to facilitate the retention of information is potentially both fitness-relevant and has been previously demonstrated to be as effective as other non-naturalistic mnemonic techniques, story processing scenarios were not included amongst the alternative fitness relative scenarios in Sandry et al.'s (2013) study. The purpose of the present study was to fill this gap in the literature by comparing a story creation condition with survival processing and another deep processing control condition (pleasantness processing) across four experiments.

Although, to our knowledge, no study has compared recall after encoding words in a story creation condition with survival processing, at least one study has examined memory for stories after focusing on either the survival-related aspects of the story or, alternatively, asking participants to recall the details of the story. In a series of experiments, Seamon et al. (2012) demonstrated that when participants were asked to read a survival-related story with an emphasis on comprehending the presented material, participants’ recall performance for facts contained in the story was not enhanced following survival processing. More specifically, asking participants to focus on understanding the story led to similar levels of recall performance to asking them to focus on the survival value of the content of the story. This result suggests that encoding information embedded in stories, as opposed to other presentation methods, might improve memory, even in the absence of overt survival-related processing. Given that our ancestors likely used stories as a primary means of communicating important survival-related information, it suggests that the narrative structure of a story might be sufficient to confer mnemonic benefits to information embedded within it.

Survival processing is often considered to be one of the best deep processing tasks identified to date. The proximal explanation for the survival processing advantage likely involves enhanced item-specific processing rather than enhanced relational processing. For example, Burns et al. (2013) analyzed participants’ cumulative recall curves following both survival processing and after rating words’ effectiveness in assisting with a moving scenario. The pattern of results suggested that the survival scenario prompts more item-specific processing, but not more relational processing. Ultimately, however, the survival processing advantage has been hypothesized to result from the importance of remembering survival-related information throughout the course of humans’ evolutionary history. We predict that similar proximate and ultimate causes facilitate the retrieval of information following story processing. Unlike survival processing, story processing seems to occur naturally and likely has more immediate applications outside of the study of human memory, for example use in educational settings. To facilitate such applied applications, a better understanding of the theoretical explanations for any improvements in memory following story processing is required to create stories and exercises designed to optimize subsequent memory performance. The following experiments represent a first step towards gaining a better theoretical understanding of how the use of stories can improve memory. To achieve this goal, we test a straightforward prediction: If the same underlying processes that confer the survival processing advantage also result in enhanced memory following story processing, then both story processing and survival processing should lead to similar levels of recall when compared directly with one another and both tasks should lead to better recall than other well-known deep processing tasks. Although this prediction is straightforward, one challenge to testing it is that the methodologies used to examine the survival processing advantage and the story processing advantage have traditionally been quite different. In the experiments described below, we attempt to overcome this challenge and provide a direct comparison of recall performance following survival processing, story processing, and a well-known deep processing control condition, namely, pleasantness processing.

Experiment 1

Experiment 1 was designed with two primary goals in mind. First, we sought to replicate the mnemonic benefits of story processing observed by Bower and Clark (1969) using a more traditional levels of processing paradigm (e.g., Nairne et al., 2008). Second, we wanted to compare the number of common nouns recalled following encoding with a story processing scenario with the number recalled following both survival processing and pleasantness processing. Given the previous mnemonic benefit observed following story processing and the oft replicated survival processing advantage, we predicted superior recall performance in those two conditions relative to the pleasantness processing condition.

Method

Participants

An a priori power analysis was conducted using G*Power version 3.1.9.7 (Faul et al., 2007) to estimate the number of participants required to test our primary hypotheses using a one-way analysis of variance. A meta-analysis conducted by Scofield et al. (2018) reported medium to large survival processing effects for between-subjects designs like those used in the current study. For a one-way ANOVA with three groups, a medium to large effect size of .325 (Cohen, 1988), a significance criterion of α = .05, and power = .80, the recommended sample size was 32 participants per condition. In Experiment 1, 88 undergraduate psychology students attending the University of Mississippi participated in return for partial course credit.

Design and Materials

A between-subjects design was employed in which participants were randomly assigned to one of three instructional conditions. Twenty-seven participants were asked to create a story using a list of 20 unrelated common nouns selected from Van Overschelde et al.'s (2004) list of category norms. The selected words included one of the top three responses from 20 unique categories, for example, the word “diamond” selected from the category “a precious stone,” ensuring that common unrelated nouns were employed. None of the items chosen had any overtly pleasant or survival-related characteristics. For example, the first several items presented to participants in Experiment 1 were, “diamond,” “carrot,” “chair,” “dollar,” “house,” “magazine,” and “sandal.” Thirty-one participants listed survival-related attributes for each of the same words whereas another 30 participants listed pleasant or unpleasant attributes for each word.

Procedure

Participants arrived at the laboratory in small groups ranging from one to three people and were seated in separate cubicles. Other than the participants, the experimenter was the only other person in the room. After signing an informed consent form, the participants were handed a participant response sheet and listened to the experimenter read a set of instructions printed at the top of the form. In all three conditions, the instructions for completing the task were printed at the top of one side of a sheet of paper with 20 common nouns printed in a column on the left side of the page underneath the instructions. Participants were told that, during the experiment, they were to use the words printed beneath the instructions to complete the encoding task to which they had been assigned.

In the survival condition, instructions similar to those employed by Nairne et al. (2008) were displayed at the top of the page. Participants were asked to read the instructions silently to themselves as the experimenter read them aloud. The instructions read as follows, “In this task, we would like you to imagine that you are stranded in the grasslands of a foreign land without any basic survival materials. Over the next few months, you’ll need to find steady supplies of food and water and protect yourself from predators (Nairne et al., 2008, p. 177). For each of the words listed below, please describe how the word might help you survive in this situation.”

Participants in the pleasantness condition proceeded in a similar manner, with the exception that they were asked to list pleasant or unpleasant attributes to the right of each of the 20 common nouns printed in a column on the left side of the page. More specifically, those instructions read, “In this task, we are going to show you a list of words, and we would like you to list several pleasant or unpleasant attributes of each word in the space below.” Finally, in the story processing condition, participants were asked to read the following instructions to themselves as the experimenter read them aloud, “Many people find stories to be interesting and entertaining. Please write a short story in the space provided below and be sure to use each of the listed words in your story.”

Participants were provided with 10 min to complete the encoding task to which they had been assigned. If a participant indicated that they were finished prior to the end of the 10-min encoding period, the experimenter encouraged them to continue working on the task until the time allotted for the encoding period had expired. Following the encoding phase of the experiment, participants were asked to flip the participant response sheet over. The top half of the back side of the response sheet included 9 two-digit multiplication problems (e.g., 96 × 58). Participants were asked to complete as many of these problems as they could during a 60 s distractor period. The bottom half of the back side of each participant response sheet was blank. After completing as many multiplication problems as they could in the time allotted, participants were asked to write down as many list items as they could remember in the blank space underneath the math problems. This incidental free recall task lasted for 5 min, and if a participant indicated that they were no longer capable of remembering any more items, the experimenter asked them to continue working on the task until the time was up. After the free recall task, all the participants were debriefed and thanked for their participation.

Results and Discussion

The primary dependent variable of interest was the proportion of words that participants recalled as a function of the instructional condition to which they had been assigned. These values are displayed in Figure 1. A one-way analysis of variance revealed statistically significant differences in recall scores as a function of instructional condition, F (2, 85) = 23.47, MSe = 7.23, p < .001, η_p² = .36. Planned comparisons indicated that participants recalled more words in the story processing condition (M = .86) than in survival condition (M = .74), F (1, 56) = 10.80, MSe = 7.10, p < .01, η_p² = .16. In addition, participants recalled more words in the story condition than in the pleasantness condition (M = .61), F (1, 55) = 43.45, MSe = 7.78, p < .001, η_p² = .44. Finally, participants remembered more words in the survival condition than in the pleasantness condition, F (1, 59) = 14.73, MSe = 6.85, p < .001, η_p² = .20, thereby replicating the oft observed survival processing advantage. Overall, participants in the story processing condition had the highest levels of recall, participants in the survival processing condition recalled an intermediate number of words, whereas participants in the pleasantness processing condition recalled the fewest number of words.

Figure 1.

Proportion of words correctly recalled as a function of instructional condition in Experiment 1. Error bars represent 95% confidence intervals.

Another dependent variable of interest was the number of intrusions, or unstudied words, that participants wrote down during the free recall stage of the experiment. Intrusions were infrequent (M = .20) across all three instructional conditions. A one-way analysis of variance indicated that there were no statistically significant differences observed in the number of intrusions as a function of instructional condition, F (2, 85) = 1.20, MSe = 0.23, p > .05, η_p² = .03. Overall, encoding the words using one set of processing tasks or another did not appear to influence the number of unstudied words included in participants’ recall output.

Our results replicated and extended those reported by Bower and Clark (1969). More specifically, asking participants to create a story using a list of common nouns produced relatively high levels of recall on a subsequent memory test. Further, such recall levels were not only higher than another well-known deep processing task, in this instance, pleasantness, but also, were higher than survival processing, which is often considered to be one of the best deep processing tasks described to date. Similar to considering information in the context of one's survival, there is reason to predict that people may be “tuned” to remember information used in the creation of stories and that such adaptive processing may provide a strong mnemonic benefit.

Despite these observations, there were some major methodological differences between the design employed in Experiment 1, and those typically used in survival processing studies (e.g., Nairne et al., 2008). First, in the current experiment, the studied words were presented all at once rather than one at a time. Although such a presentation method allowed for a partial replication and extension of Bower and Clark's (1969) observation of a story processing effect, it did not guarantee an equal amount of processing time for each studied item. Second, this presentation format also allowed participants to process the words in the order in which they saw fit. An examination of the stories that participants created indicated that the studied words in each story were not typically inserted in the order in which they were presented in the column on the left side of each participant's response form. However, in the survival and pleasantness conditions, participants tended to respond to those items in the order in which they were presented on the page. It is possible that reordering the words for the purpose of including them in a story of the participant's own creation conferred some additional mnemonic benefit that was not typically utilized by participants in the other conditions. Furthermore, participants in the story condition rewrote each word while including it in their story, whereas participants in the pleasantness and survival conditions did not. Finally, most studies that have examined the survival processing advantage tend to include more than 20 to-be-remembered items. It is possible that the mnemonic advantage of creating a story observed in Experiment 1 is one of diminishing returns, in which increasing the number of stimuli to be remembered reduces the utility of the advantage. In Experiment 2, we sought to evaluate the mnemonic benefit of narrative processing relative to both survival and pleasantness processing when 30 unrelated common nouns were presented one at a time, thereby equating both the amount of time each word was processed across conditions and standardizing the order in which each word was considered. Thus, the design employed in Experiment 2 eliminated the potential for production effects to influence participants’ recall scores and standardized both study time per item and response order across conditions.

Experiment 2

Method

Participants

Ninety-one undergraduate psychology students from the University of Mississippi participated in the experiment in return for partial course credit.

Design and Materials

As in Experiment 1, a between-subjects design was employed with 30 participants included in the survival and story conditions and 31 participants in the pleasantness condition. The design and materials employed in Experiment 2 were similar to those used in Experiment 1, albeit with several notable exceptions. First, in the encoding phase of the experiment, participants were asked to process 30 unrelated common nouns selected from Van Overschelde et al.'s (2004) list of category norms as opposed to the 20 stimulus items used in the first experiment. Second, the words were presented one at a time for 10 s each on a computer monitor rather than all at once in a column on a sheet of paper. These methodological changes required participants in all three conditions to write each presented word as they recorded their responses and allowed for control over both the amount of time and the order in which each word was processed (Table 1).

Table 1.

Mean Ratings and Response Times (With Standard Errors of the Means) in Experiment 3.

	Rating		Response Time (ms)
Condition	Average	SEM	Average	SEM
Story	3.70	0.11	2107.2	0.09
Survival	2.97	0.07	2372.8	0.07
Pleasantness	3.21	0.09	2190.6	0.04

Procedure

Participants arrived at the lab in small groups ranging from one to three and were seated in separate cubicles containing personal computers. After signing an informed consent form, the participants were handed a participant response form which was blank on one side other than 30 evenly spaced horizontal lines, extending from the left side of the page to the right, on which participants would be asked to write their responses. At this point, participants were asked to press a key on the keyboard to initiate the experiment which resulted in one of three sets of instructions appearing on the screen. The instructions were similar to those used in Experiment 1 and included story, survival, and pleasantness processing instructions. In all three conditions, participants were asked to read the instructions presented on the computer monitor silently to themselves while the experimenter read them aloud. Participants were informed that words would be presented on the computer monitor, one at a time, for a brief interval. In the survival and pleasantness conditions, they were instructed to write the first presented word down on the first (top) line and then, when the next word was presented, to use the next available empty line on the page to continue this process until all 30 common nouns had been presented and processed. After writing each word, and prior to the presentation of the next word, they were asked to either list ways that the items could help them survive in a grasslands scenario, or list pleasant and unpleasant attributes of each item. In the story condition, they were asked to create a story and to write a new line of the story using each presented word in the space provided.

More specifically, in the survival condition, while each word was displayed on the screen, participants were asked to write the word on the first line of the participant response sheet and then list the ways that the item could help them survive in a grasslands scenario in the space immediately adjacent to the word. Participants in the pleasantness condition proceeded in a similar manner, with the exception that they were asked to write down each presented word, followed by listing pleasant and unpleasant attributes of the word in the space provided. Finally, in the story condition, participants went through an identical process except that they were asked to write one line of a story that they were asked to create, being sure to include each presented word in each line of their story. Thus, across all three conditions, despite the differences in encoding instructions, participants were asked to write down each presented word one time while engaging in the processing task.

After all 30 words had been presented and the participant responses had been recorded, participants were asked to flip the participant response sheet over. The top half of the back side of the response sheet included 18 three-digit addition and subtraction problems (e.g., 167–125). Participants were asked to work on completing as many of these problems as possible during a 60 s distractor period. The bottom half of the back side of each participant response sheet was blank. After completing as many problems as they could in the time allotted, participants were asked to write down as many list items as they could remember in the blank space underneath the math problems. This incidental free recall task lasted for 5 min, and if a participant mentioned that they were no longer capable of remembering any more items, the experimenter asked them to continue working on the task until the time was up. After the recall task, all the participants were debriefed and thanked for their participation.

Results and Discussion

Free Recall

The primary dependent variable of interest was the proportion of words that participants recalled as a function of the instructional condition to which they had been assigned. These values are displayed in Figure 2. A one-way analysis of variance revealed statistically significant differences in recall scores as a function of instructional condition, F (2, 88) = 9.76, MSe = 13.82, p < .001, η_p² = .18. Planned comparisons indicated that participants recalled more words in the survival condition (M = .62) than in the pleasantness condition (M = .52), F (1, 59) = 10.60, MSe = 13.40, p < .01, η_p² = .15. In addition, participants recalled more words in the story condition (M = .65) than in the pleasantness condition, F (1, 59) = 19.78, MSe = 12.45, p < .001, η_p² = .25. However, despite a slight numerical advantage for the proportion of words recalled by participants in the story condition when compared with the survival condition, this difference was not statistically significant, F (1, 58) = 0.90, MSe = 15.62, p > .05, η_p² = .02. Thus, overall, participants recalled more words in the survival and story conditions than in the pleasantness condition and there was not a statistically significant difference in the number of words recalled in the story and survival conditions.

Figure 2.

Proportion of words correctly recalled as a function of instructional condition in Experiment 2. Error bars represent 95% confidence intervals.

Intrusions

We also examined the number of intrusions, or words not presented on the list, that participants wrote down during the free recall phase of the experiment. Intrusions were infrequent (M = .42 words per participant across conditions). A one-way analysis of variance indicated that there were not statistically significant differences in the number of intrusions as a function of instructional condition, F (2, 88) = 0.03, MSe = 0.02, p > .05, η_p² = .001.

Experiment 2 provided more evidence in favor of the idea that story processing may represent an adaptive processing technique equal in its effectiveness to survival processing. With an increased list length (30 items) and both item presentation order and processing time held constant across conditions, participants remembered similar numbers of items following both survival and story processing. However, participants in the story processing condition did not outperform their peers in the survival condition. It is possible that restricting the order in which each word could be included in the story reduced the coherence or optimal structure of the story thereby limiting the extent to which relational processing could occur amongst the to-be-remembered items. Furthermore, despite the success of Experiments 1 and 2 in demonstrating the potential mnemonic benefits inherent in story creation, the methodology employed in both experiments differed markedly from a number of recent studies examining the effectiveness of survival processing relative to other well-known deep processing tasks. In many such studies, participants listen to a processing scenario and then provide a numerical rating to each individual stimulus item with respect to its relevance to the processing task employed. Although story creation does not seem, to us, easily amenable to such a research design, in Experiment 3, we attempted to examine the effectiveness of story processing relative to survival and pleasantness processing using such methods.

Experiment 3

Method

Participants

Ninety undergraduate psychology students from the University of Mississippi participated in the experiment in return for partial course credit.

Design and Materials

A between-subjects design was employed, with 30 participants included in each of the three instructional conditions. In the initial phase of the experiment, participants were asked to rate 29 unrelated common nouns selected from Van Overschelde et al.'s (2004) list of category norms. We had originally planned on using a 30-item word list, but due to a programming error, the last word on the list was not presented to any of the participants in the experiment. A third of the participants rated words with respect to their relevance in a survival situation, another 30 participants rated the pleasantness of the words, and a final 30 participants rated the words in terms of the ease with which they could be included in a story. The instructions for both the survival and pleasantness tasks were identical to those used by Nairne et al. (2008). More specifically, an experimenter read the following instructions to participants in the pleasantness condition:

In this task, we are going to show you a list of words, and we would like you to rate the pleasantness of each word. Some of the words may be pleasant and others may not, it's up to you to decide. (Nairne et al., 2008, p. 177)

In addition, the survival instructions were as follows:

In this task, we would like you to imagine that you are stranded in the grasslands of a foreign land, without any basic survival materials. Over the next few months, you’ll need to find steady supplies of food and water and protect yourself from predators. We are going to show you a list of words, and we would like you to rate how relevant each of these words would be for you in this survival situation. (Nairne et al., 2008, p. 177)

Finally, the instructions for the story processing condition were as follows:

In this task, we would like you to imagine that you are a well-known author working on a deadline. We are going to show you a list of words and we would like you to use the words in the list to create an interesting and entertaining short story. Some of the words may be easy to include in your story whereas others may be more difficult. We would like you to rate how easy or difficult it is to include each word in your story.

All stimuli were presented and all responses were recorded using personal computers.

Procedure

Participants arrived at the lab in small groups ranging from one to three participants and were seated in separate cubicles containing personal computers. Other than the participants, the experimenter was the only other person in the room. After signing an informed consent form, the participants listened to the experimenter read a set of instructions that informed them that they would be asked to rate a list of words on a scale ranging from 1 (totally unpleasant / irrelevant / very difficult) to 5 (extremely pleasant / relevant / extremely easy). The participants were asked to rate the words with respect to the pleasantness of each word, the word's relevance to a survival situation, and the ease with which the word could be included in a short story, depending on the encoding condition to which they had been randomly assigned. They were told that during each trial, a word would appear on the computer screen with a rating scale beneath it. They were asked to type a numerical response ranging from 1 to 5, at which point the word and their response would remain on the screen until the presentation of the next word. Participants were encouraged to attempt to use the entire rating scale while rating the words.

The participants completed five practice trials prior to rating the experimental stimuli to get them acquainted with both the procedure and the timing of the rating task. During each trial, a word was presented for 5 s. At the conclusion of each 5 s presentation, a 0.5 s interstimulus interval preceded the presentation of the next word. After each participant rated the five practice words, the experimenter asked them if they had any further questions about the rating task. After answering any remaining questions, the experimenter asked participants to click on box that said “Begin the Experiment” to start the experimental trials. At that point, the process of word presentation, coupled with each participant providing a numerical rating, continued until each participant had rated all 29 stimulus items.

After rating the last word on the list, participants were asked to complete simple single digit addition and subtraction problems for 60 s. Following this distractor task, the incidental free recall portion of the experiment commenced. Participants were asked to type as many of the previously rated items as possible. Each item that participants typed remained on the screen for the duration of the recall task. The recall task lasted for 5 min, and if a participant mentioned that they were no longer capable of remembering any more items, the experimenter asked them to continue working on the task until the time was up. After the recall task, all the participants were debriefed and thanked for their participation.

Results

Free Recall

In the present experiment, the primary dependent measure of interest was the proportion of words that participants recalled as a function of instructional condition, which are displayed in Figure 3. A one-way analysis of variance revealed statistically significant differences in recall scores as a function of instructional condition, F (2, 87) = 21.84, MSe = 6.39, p < .001, η_p² = .33. Further analyses indicated that participants remembered more words in the survival condition (M = .47) than in the pleasantness condition (M = .34), F (1, 58) = 32.38, MSe = 6.81, p < .001, η_p² = .36, and that participants remembered more words in the story condition (M = .46) than in the pleasantness condition, F (1, 58) = 45.28, MSe = 4.37, p < .001, η_p² = .44. However, there was not a statistically significant difference in the proportion of words recalled in the survival condition and the story condition, F (1, 58) = 0.07, MSe = 8.00, p > .05, η_p² = .001. Thus, overall, participants remembered more words in the survival condition and the story condition than in the pleasantness condition whereas there was not a statistically significant difference observed between the story processing condition and the survival processing condition.

Figure 3.

Proportion of words correctly recalled as a function of instructional condition in Experiment 3. Error bars represent 95% confidence intervals.

Intrusions

Another dependent variable of interest was the number of words that participants recalled that were not presented during the word rating phase of the experiment. Such words could include practice items, other related words, or any word other than the 29 items presented during the rating phase of the study. Although such intrusions were rare, a one-way analysis of variance revealed statistically significant differences in the number of intrusions as a function of instructional condition, F (2, 87) = 3.79, MSe = 0.58, p < .05, η_p² = .08. Subsequent analyses indicated that participants remembered more nonpresented words in the survival condition than in the pleasantness condition, F (1, 58) = 5.56, MSe = 0.51, p < .05, η_p² = .09, and that participants remembered more nonpresented words in the story condition than in the pleasantness condition, F (1, 58) = 9.36, MSe = 0.40, p < .01, η_p² = .14. However, the difference between the number of nonpresented words recalled in the survival condition and the story condition was not statistically significant, F (1, 58) = 0.08, MSe = 0.84, p > .05, η_p² = .001. Thus, overall, participants had a higher rate of intrusions in the survival condition and the story condition than in the pleasantness condition.

Ratings

In addition to the proportion of studied words and intrusions that participants recalled, we also analyzed the ratings that participants provided as a function of instructional condition (on a scale from 1 to 5). A one-way analysis of variance revealed statistically significant differences among the ratings that participants provided as a function of instructional condition, F (2, 87) = 16.91, MSe = 0.24, p < .001, η_p² = .28. Subsequent analyses indicated that participants’ ratings were higher in the pleasantness condition than in the survival condition, F (1, 58) = 4.77, MSe = 0.19, p < .05, η_p² = .08, and that participants’ ratings were higher in the story condition than in the pleasantness condition, F (1, 58) = 11.87, MSe = 0.29, p < .01, η_p² = .17. Finally, participants’ ratings were higher in the story condition than in the survival condition, F (1, 58) = 32.16, MSe = 0.25, p < .001, η_p² = .36. Thus, overall, participants’ ratings were highest in the story condition, intermediate in the pleasantness condition and lowest in the survival condition. Although the ratings differed as a function of condition, it did not appear that higher (or lower) ratings were predictive of subsequent recall performance. However, it is possible that because participants provided higher ratings in the story condition than in the survival condition, a congruency effect could have enhanced participants’ memories for the items in the story condition relative to the survival condition. To examine this possibility, we used words, rather than participants, as the basis of the analysis and obtained a similar result (e.g., Nairne et al., 2008). The difference in the proportion of times participants recalled each word in the survival condition and the story condition was not statistically significant, despite the observed differences in ratings, F (1, 56) = .14, MSe = 28.41, p > .05, η_p² = .002.

Reaction Times

A final dependent variable of interest was the amount of time that it took participants to provide each rating as a function of the instructional condition to which they were assigned. A one-way analysis of variance revealed statistically significant differences among the amounts of time that it took participants to provide ratings as a function of instructional condition, F (2, 87) = 3.76, MSe = 0.15, p < .05, η_p² = .08. Subsequent analyses indicated that participants took longer to provide ratings in the survival condition than in the pleasantness condition, F (1, 58) = 5.82, MSe = 0.08, p < .05, η_p² = .09, and that the difference in participants’ response times in the story condition and the pleasantness condition was not statistically significant, F (1, 58) = 0.67, MSe = 0.16, p > .05, η_p² = .01. Finally, participants took longer to respond in the survival condition than in the story condition, F (1, 58) = 5.29, MSe = 0.20, p < .05, η_p² = .08. Thus, overall, participants’ reaction times were higher in the survival condition than in both the story condition and the pleasantness condition. As with the ratings, although reaction times differed as a function of condition, they were not predictive of subsequent recall performance.

The results obtained in Experiment 3 again support the conclusion that asking participants to consider list items in the context of a story can lead to recall levels similar to asking them to consider items in terms of their survival value. However, one limitation associated with the experiment was the absence of a means by which to verify that participants created stories as instructed. Given the relatively short processing interval (5 s per word) and the requirement of entering a number as opposed to any written output, it seems unlikely that all participants in the story condition created a story, as was evident based on participants’ output in Experiments 1 and 2. Perhaps a better worded story rating task or other methodological changes could improve the effectiveness of the task relative to other well-known deep processing tasks including survival processing.

Experiment 4

Taken together, the results from the experiments reported above suggest that encoding information in the context of a story leads to similar levels of recall to survival processing and higher levels of recall than pleasantness processing. We hypothesize that similar underlying cognitive mechanisms are responsible for the increase in recall performance following both survival and story processing. One way to test this prediction is to compare participants’ recall following both survival and story processing with recall in a condition in which participants are asked to encode words while creating a story related to survival. An observation of similar recall levels across all three conditions would support the conclusion of similar underlying cognitive mechanisms contributing to the observed effects whereas an observation of an additive effect of both story and survival processing (i.e., better recall performance in the combined survival-story condition than in either the survival or story processing conditions) would suggest that different processes underly the advantage conferred by story and survival processing. Thus, in Experiment 4, we sought to replicate and extend the results obtained in Experiment 2 by adding an additional condition in which participants encoded words while creating a story related to survival.

Method

Participants, Design, Materials, and Procedure

One hundred and twenty undergraduate psychology students attending the University of Mississippi participated in Experiment 4 in return for partial course credit. The design, materials, and procedure employed in Experiment 4 were identical to those used in Experiment 2, with one exception. In addition to asking participants to encode words in one of three between-subjects conditions (survival, story, and pleasantness processing) a fourth condition was included in which participants were asked to create a story related to a survival scenario. The instructions provided to participants in that condition were as follows:

In this task, we would like you to imagine that you are stranded in the grasslands of a foreign land, without any basic survival materials. Over the next few months, you'll need to find steady supplies of food and water and protect yourself from predators.

In a moment, we are going to show you a list of words. We would like you to write a short story using each of the words on the list about how you would survive in this situation.

In every respect other than the inclusion of this new combined survival-story condition, the procedure used in Experiment 4 was identical to that employed in Experiment 2. Thirty participants were randomly assigned to one of the four between-subjects conditions. As in Experiment 2, after completing the encoding task to which they were assigned, participants engaged in a brief math distractor task and then completed an incidental free recall test for the list items lasting 5 min. Following this, participants were debriefed and thanked for their participation.

Results and Discussion

Free Recall

An examination of the stories that participants generated in the combined survival-story condition indicated that all 30 participants in that condition followed the instructions and included survival elements in their stories. Most participants did this at the outset of the encoding period, for example, one participant wrote the following as the first sentence in their story after being presented with the word “diamond,” “I woke up stranded on a desert island surrounded by diamonds.” Notably, only one participant in the story condition spontaneously included survival elements in their story. Thus, although it is not uncommon for stories to include survival-related elements, participants did not tend to include such content unless prompted to do so.

The primary dependent variable of interest was the proportion of words that participants recalled as a function of the instructional condition to which they had been randomly assigned. These values are displayed in Figure 4. A one-way analysis of variance revealed statistically significant differences in recall scores as a function of instructional condition, F (3, 116) = 12.44, MSe = 10.65, p < .001, η_p² = .24. Planned comparisons indicated that participants recalled more words in the survival condition (M = .64) than in the pleasantness condition (M = .50), F (1, 58) = 28.36, MSe = 9.18, p < .001, η_p² = .33. In addition, participants recalled more words in the story condition (M = .65) than in the pleasantness condition, F (1, 58) = 25.87, MSe = 10.72, p < .001, η_p² = .31. Participants also recalled more words in the combined survival-story condition (M = .64) than in the pleasantness condition, F (1, 58) = 29.14, MSe = 8.79, p < .001, η_p² = .33. As in Experiment 2, the difference between the number of words recalled in the survival condition and the story condition was not statistically significant, F (1, 58) = 0.02, MSe = 12.50, p > .05, η_p² = .00. Notably, the difference between the number of words recalled in the combined survival-story condition and the survival condition was not statistically significant, F (1, 58) = 0.002, MSe = 10.57, p > .05, η_p² = .00. Furthermore, the difference in the number of words recalled in the combined survival-story condition and the story condition was not statistically significant, F (1, 58) = 0.03, MSe = 12.11, p > .05, η_p² = .001. Thus, no additive effects on recall performance were observed in the combined survival-story condition, consistent with the hypothesis that similar underlying cognitive mechanisms contribute to the recall advantage observed across the combined story-survival condition, and the story and survival conditions.

Figure 4.

Proportion of words correctly recalled as a function of instructional condition in Experiment 4. Error bars represent 95% confidence intervals.

Intrusions

We also examined the number of intrusions, or words not presented on the list, that participants wrote down during the free recall phase of the experiment. Intrusions were infrequent (M = .17 words per participant across conditions). A one-way analysis of variance indicated that there were not statistically significant differences in the number of intrusions as a function of instructional condition, F (3, 116) = 0.11, MSe = 0.21, p > .05, η_p² = .003. Overall, Experiment 4 provided more evidence in favor of the idea that story processing may represent an adaptive processing technique equal in its effectiveness to survival processing. In addition, we did not observe an additive effect of survival and story processing in a condition in which both sets of processing instructions were combined. As noted above, this result is consistent with the hypothesis that similar cognitive processes are responsible for the increase in recall performance following both survival and story processing relative to pleasantness processing, a well-known deep processing task. One question left unanswered is the extent to which item-specific and relational processing contribute to the story processing advantage. Burns et al. (2013) have demonstrated that the survival processing advantage likely relies more on enhanced item-specific than relational processing. We recommend that future research on this topic examine participants’ cumulative recall curves following both story processing and survival processing in addition to using retention measures more sensitive to relational processing, like reconstruction tasks, to better understand the relative contribution of item-specific and relational processing to the story processing advantage.

General Discussion

Across four experiments we compared the relative effectiveness of story processing to both pleasantness processing and survival processing. In Experiment 1, when participants were presented with a list of words and were asked to use those words to create a story, their subsequent recall performance for those words was better than in both the survival processing condition and the pleasantness processing condition. In Experiment 2, when participants were asked to perform similar tasks with both the item presentation order and the item presentation rate controlled across conditions, the difference in participants’ recall performance in the survival and story processing conditions was not statistically significant and both were better than recall performance in the pleasantness processing condition. In Experiment 3, when participants were asked to rate presented words one-at-a time as a function of their pleasantness, the ease with which they could be included in a story, or their survival relevance, again, both survival and story processing led to enhanced recall performance relative to that of participants assigned to the pleasantness processing condition. Finally, in Experiment 4, when a combined survival-story processing condition was compared with survival, story, and pleasantness processing, all three conditions led to enhanced recall performance relative to pleasantness processing. Overall, the results of these experiments suggest that story processing may represent a natural mnemonic technique equal in its effectiveness to survival processing and that our memory systems may be “tuned” to remember information embedded in stories.

It is possible that story processing elicits both item-specific processing, as participants consider the meaning of each word while creating the story and relational processing, as they consider how the item can be included in the overall framework of the story. Forcing participants to include words in a story in a predetermined order might inhibit their ability to optimally structure the story, ultimately reducing the extent to which its individual components are subsequently remembered. Similarly, reducing the amount of time that participants are allotted to create a story and asking them to consider the ease with which an item can be included in a story may also reduce the extent to which story creation improves recall performance relative to survival processing. Both manipulations (controlling item presentation order and item presentation rate), although common in the levels of processing literature, seem counter to the way a participant might develop a story naturalistically to attempt to better remember information.

As noted above, whereas the proximate cause of the survival processing advantage is likely enhanced item-specific processing (e.g., Burns et al., 2011, Burns et al., 2013), the ultimate cause is believed to be evolutionary pressure to remember survival-related information (e.g., Nairne et al., 2007). Although the typical research methods used to test the effectiveness of both the survival and storytelling advantages are typically quite different, both such proximate and ultimate factors appear likely to contribute to mnemonic advantages following both types of processing. The comparable recall performance observed in the present experiments following both story processing and survival processing is consistent with the hypothesis that the underlying cognitive mechanisms driving such outcomes are similar in both cases. The lack of an additive effect on recall when both survival and story processing are combined is also consistent with this view. In addition, when a survival processing advantage is found outside the context of the traditional grasslands scenario, it is typically in cases where a story-like context is elicited, without necessitating a realistic or plausible scenario. For example, Soderstrom and McCabe (2011) have observed similar enhanced memory performance using a zombie processing scenario and Kostic et al. (2012) observed enhanced memory performance using a scenario involving being lost in outer space. Furthermore, Kroneisen and Makerud (2017) demonstrated that, at least for words that are high in imageability, survival processing leads to memory performance similar to well-known mnemonic techniques like the method of loci. Given the similarities between recall performance following survival and story processing observed in the present study, one might expect similar results when comparing recall following story processing to other well-known mnemonic techniques.

However, unlike the method of loci or survival processing, storytelling, as a natural mnemonic technique, seems to have more applicability in applied settings, such as through enhancing the memorability of educational content. Unlike other mnemonic strategies, storytelling often occurs naturally, without instruction, and is commonly used by educators to enhance learning in the classroom (Hill et al., 1991). The present results suggest that storytelling may well represent one of the best deep processing tasks identified to date, perhaps even better than survival processing in more applied settings, and as such, a better understanding of the cognitive mechanisms underlying the story processing advantage could allow for the creation of better uses in applied settings. Such research could examine the relative contributions of item-specific and relational processing to the mnemonic enhancements induced through storytelling using techniques like the observation of cumulative recall curves (e.g., Burns et al., 2013) or by examining other forms of retention such as recognition performance or memory for serial order. In addition, it is possible that creating stories may improve memory performance over both short and longer retention intervals and that it may be especially well suited to improving memory for “real world” information like complicated lecture material or even survival-related content, in addition to more basic information like lists of words provided to participants while completing a memory experiment. It may also be fruitful to explore how such storytelling strategies can be effectively communicated to educators and students. Whether such predictions are ultimately borne out through further research, we believe that an increased focus of attention on story creation as a natural mnemonic may prove fruitful in identifying and developing mnemonic techniques that can be employed to improve retention for various forms of information under a wide variety of experimental and naturalistic conditions.