The study has a basic design with a single independent variable with two levels (information type: expected information condition, unexpected information condition). The dependent variables are memory accuracy and false memory in recall and recognition tests.

Eight participants were excluded because they stated that they did not hear the first sentence of the passage from the voice recording that contained critical information about the main character’s age. The inclusion criteria were being healthy young adult university students, the exclusion criterion was not being a native speaker of Turkish. Participants were randomly assigned to the conditions in which the passages included either information that is expected (expected information condition) or unexpected (unexpected information condition) of the main character’s age. A total of 88 participants were recruited for the experiment. An equal number of participants were assigned to each condition, consisting of 33 males and 55 females within the age range of 18 and 28. In the expected information condition, there were 27 females and 17 males. In the unexpected information condition, there were 28 females and 16 males¹1

The sample size was determined based on similar prior research (e.g., Araya et al., 2003; Kleider et al., 2008).

The experimental procedure involved listening to a passage that either contained expected or unexpected information based on age stereotypes, responding to the misleading question about the passage, completing a non-verbal filler task, completing the recall and the recognition test for the information in the passage, responding to the manipulation check question about the passage, and finally filling in Kogan’s Attitude Toward Older People Scale (KAOP), Positive and Negative Ageism Scale (PNAS), and the demographic form. The subsequent section provides detailed information about the materials.

The voice recordings and the texts for the passages, the misleading question, and the questions for the recognition test are presented in Gülseren and İkier (2025)²2

All materials, instructions, and their English versions are presented in the OSF repository in Gülseren and İkier (2025).

The study was approved by Bahçeşehir University Ethics Review Board. The study was conducted by having an online face-to-face meeting with each participant on the Zoom platform. A flowchart for the procedure is provided in Figure 1.

All participants provided informed consent. First, the voice recording for the passage was sent to the participants and the participants downloaded the recording to their computer to prevent problems related to the internet connection while listening to the passage. The participants were asked not to start the recording themselves. The experimenter started and ended the recording by accessing the participants’ screen for the duration of the recording.

All participants used headphones to eliminate external noise. First, participants were asked to put their headphones on. Then, the sound quality and level were checked by the experimenter with a short voice control recording that included only two words (i.e., “voice control”). The sound level was optimized to minimize information loss. Then the participants were asked to carefully listen to the passage without a break. They were told that once the voice recording starts, it cannot be paused or re-played. They were asked to remember the sentences in the passage for a subsequent test. Then, the recording was started and stopped at the end by the experimenter. Participants were asked to delete the voice recordings permanently from their computers after the recording was stopped.

Next, participants were given the misleading question and responded to the question as “yes” or “no”. Then, they were engaged in the non-verbal filler task in which they were asked to respond as accurately as possible. Then, participants completed the recall test and the recognition test. Afterward, participants responded to the manipulation check question. Finally, they filled out the KAOP, the PNAS, and the demographic form. All participants were debriefed.

The results section starts with the descriptive statistics for the sample and a check for control variables. First, descriptive statistics for years of education and sex distribution were provided. Then, responses for the manipulation check question were compared with an independent measures t-test, to ensure control. Thereafter, KAOP and PNAS were analyzed and compared across conditions with an independent measures t-test. KAOP total score and PNAS positive and negative scores were correlated with memory measures (lenient recall, strict recall, verbatim recall, and recognition) by using Pearson correlation.

Then, the section continues with the tests of the main hypotheses of the study. Recall and recognition performance was assessed and compared across the conditions with independent measures t-test. A chi-square test was conducted to compare the two conditions for the generation of misinformation in the recall and the recognition tests.

Errors on the memory tests (omission, commission, and distortion) were assessed and compared across the conditions by using independent measures t-test. Lastly, the two conditions were compared in terms of the expected, unexpected, and unrelated commission errors by using independent measures t-test.

One participant’s datum was removed from the expected information condition because his z-score was above three in the recall test. The final sample that we conducted the analyses on consisted of 87 participants.

Responses to the manipulation check questions showed that the characteristics and behaviors of the main character in the passage were more expected of his age in the expected information condition, M = 3.86, SE = 0.12 than in the unexpected information condition, M = 2.20, SE = 0.15, t(85) = 8.61, p < .001, d = 1.86.

Participants in the two conditions did not differ in their attitudes toward older adults. For KAOP, M = 130.35, SE = 2.51 for the expected information condition, and M = 124.30, SE = 2.68 for the unexpected information condition, t(85) = 1.65, p = .103, d = 0.35. For, PNAS positive subscale M = 42.28, SE = 0.81 for the expected information condition, M = 42.05, SE = 1.19 for the unexpected information condition, t(85) = 0.16, p = .872, d = 0.03. For PNAS negative subscale, M = 41.07, SE = 0.84 for the expected information condition, and M = 39.41, SE = 0.97 for the unexpected information condition, t(85) = 1.30, p = .20, d = 0.28.

Based on the scores for KAOP and PNAS, our participants in general had relatively positive attitudes toward older adults. Exploratory analyses showed that the attitude scores did not correlate with any of the memory measurements, all p’s > .05.

Recall scores were calculated by three different scoring criteria: lenient criterion, strict criterion, and verbatim criterion. Two independent raters scored the recalled information based on these criteria. Passages were broken down into idea units, which is a widely used procedure in prose memory research (e.g., Bransford & Johnson, 1972). The resulting 58 idea units were designated based on semantic units such as subjects, adjectives, propositions, and verbs which corresponded to either single words or phrases (Bransford & Johnson, 1972). The idea units are presented in Gülseren and İkier (2025).

For the lenient criterion scoring, a flexible approach was taken in which synonyms and analogies of the idea units that conveyed the same meaning, different conjugations of the correct verbs, and changes in the location of the idea units were counted as correct. For the strict criterion scoring, only the exact same idea units were counted as correct, but different conjugations of the correct verbs and changes in the location of the idea units were also accepted. For the verbatim criterion scoring, the idea units were not used, and only the exact same sentences were counted as correct.

The interrater reliability for the independent raters was calculated for the lenient and strict criterion recall scores by using Pearson correlation. Results showed a strong significant positive correlation between the scores of the raters for the lenient criterion, r(86) = .99, p < .001, and for the strict criterion, r(86) = .99, p < .001. Minor differences in scoring were discussed among the raters and a compromise was reached for the final scoring. For the verbatim criterion, the scoring was first completed by the first rater and then checked by the second rater to ensure that only the same sentences were counted as correct.

Percentages for the recall and the recognition test scores are presented in Figure 2.

Recall scores for the lenient criterion did not reveal a difference between the expected information (M = 33.71, SE = 1.52) and the unexpected information (M = 35.58, SE = 1.76) conditions, t(85) = 0.80, p = .43, d = 0.17. Similarly, there was no difference between the conditions for the strict scoring criterion, M = 14.91, SE = 0.99 for the expected information, and M = 16.06, SE = 1.21 for the unexpected information condition, t(85) = 0.73, p = .47, d = 0.16. The verbatim criterion for recall also revealed no difference between the conditions, M = 3.78, SE = 0.98 for the expected information and M = 5.68, SE = 1.03 for the unexpected information condition, t(85) = 1.34, p = .19, d = 0.29.

Recognition scores for the unexpected information condition (M = 63.35, SE = 2.65) were significantly higher than the expected information condition (M = 55.81, SE = 2.54), t(85) = 2.05, p = .04, d = 0.44.

In summary, the memory performance differences across the expected and unexpected information conditions were solely revealed in the recognition test.

Pearson chi-square analysis was conducted to determine whether there were differences in the false recall and recognition of the misinformation. For the recall test, intrusions with the misinformation were minimal and similar in the two conditions: Three participants in the expected information condition (n = 43) and two participants in the unexpected information condition (n = 44) generated the misinformation in the recall test, χ²(1, N = 87) = 0.19, p = .66, Cramer’s V = 0.05.

For the recognition test, however, 24 participants in the expected information condition and 14 participants in the unexpected information condition falsely recognized the misinformation as accurate.³3

There were two options that included the misinformation. Option A had an incorrect detail in the sentence along with the misinformation, whereas Option B was the correct sentence except for the misinformation. We report the percentage of participants who chose each of these options. Option D was the correct option. Expected information condition: Option A, 18.6%; Option B, 37.2%; Option C, 20.9%; Option D, 23.3%. Unexpected information condition: Option A, 11.4%; Option B, 20.5%; Option C, 13.6%; Option D, 54.5%.

Further exploratory analyses were carried out to explore the type of errors in the recall test. Out of the 58 idea units in the passage, 10 critical idea units were selected that contained information about the characteristics (e.g., energetic, sporty), habits (e.g., going for a run every morning), actions (e.g., running after and catching the snatcher), and social life (e.g., having a crowded group of friends) of the main character. Depending on the condition, the idea units constituted either expected or unexpected information based on the age of the main character.

The number of omission, distortion, and commission errors were explored. An omission error refers to the absence of a critical idea unit. An omission error was recorded if the idea unit was absent in the scoring with the lenient criterion. A distortion error refers to a change in the content of the idea unit. A distortion error was recorded if the change in the idea unit was accepted as correct with the lenient criterion, but not with the strict criterion. Finally, a commission error refers to the addition of a new item to the passage. A commission error was recorded if an added item within the whole passage was not counted as correct even with the lenient scoring criterion. The commission errors were further divided and analyzed into three categories: 1) expected commission errors, 2) unexpected commission errors, and 3) unrelated commission errors. An expected commission error was recorded if the committed information was expected of the main character’s age based on stereotypes. An unexpected commission error was recorded if the committed information was unexpected of the main character’s age based on stereotypes. Unrelated commission errors were unrelated to the age of the main character.

Two independent raters scored errors. Pearson correlations showed strong correlations between the scores of the raters for the omission, r(85) = .96, p < .001, distortion, r(85) = .95, p < .001, and commission, r(85) = .91, p < .001 errors. Specifically, interrater reliability for the expected, r(85) = .89, p < .001, the unexpected, r(85) = .86, p < .001, and the unrelated, r(85) = .95, p < .001 commission errors was high. Minor differences in scoring were discussed among the raters and a compromise was reached for the final scoring.

Analysis for the omission errors did not reveal any difference between expected information (M = 5.00, SE = 0.25) and unexpected information (M = 4.77, SE = 0.28) conditions, t(85) = 0.60, p = .55, d = 0.13. There was also no difference across the conditions for the distortion errors. Specifically, M = 2.02, SE = 0.13 for the expected information, and M = 2.18, SE = 0.15 for the unexpected information condition, t(85) = 0.81, p = .42, d = 0.17.

There was a significant difference across the conditions for the overall commission errors. There were more commission errors in the expected information condition (M = 3.40, SE = 0.25) than in the unexpected information condition (M = 2.55, SE = 0.25), t(85) = 2.41, p = .02, d = 0.52. The results of the error analyses are presented in Figure 3.

A more detailed analysis of the commission errors showed that expected commission errors were more frequent in the expected information condition (M = 1.30, SE = 0.14) than in the unexpected information condition (M = 0.11, SE = 0.05) condition, t(85) = 7.93, p < .001, d = 1.70. On the other hand, unexpected commission errors were non-existent in the expected information condition and more frequent in the unexpected information condition (M = 0.82, SE = 0.15), t(85) = 5.52, p < .001, d = 0.17. The frequency of the unrelated commission errors did not differ across the expected information (M = 2.09, SE = 0.22) and unexpected information (M = 1.61, SE = 0.19) conditions, t(85) = 1.67, p = .10, d = 0.36. Thus, the commission errors were likely to be consistent with the information that was presented in the passage. The results of the analyses for the types of commission errors are presented in Figure 4.

Our findings imply that the type of memory test (i.e., recall versus recognition) may be a critical factor related to how expected and unexpected information is processed.

Our recognition memory results are consistent with the notion that information that is expected is more likely to be retrieved as a gist (Araya et al., 2003; Brainerd & Reyna, 2010). Recognition relies on familiarity more than recall (Kintsch, 1970), and this may be a reason that the differences across conditions were only revealed in the recognition test. Given that only the minor details of the correct sentences were changed in the incorrect choices of the recognition test, a sense of familiarity was probably preserved, increasing memory errors. Considering that the misinformation in the study was not stereotypical, yet that it was more likely to be falsely recognized in the expected information condition, the misinformation was possibly represented as a gist in memory that is open to distortion in the context of stereotypically expected information.

A useful measure to support this explanation could have been collecting “remember”/“know” judgments. While a remember judgment indicates conscious recollection of information, a “know” judgment indicates just a sense of familiarity in the absence of conscious recollection (Tulving, 1985). Previous investigations showed more know judgments for schematic information than for non-schematic information (e.g., Neuschatz et al., 2002). If these measurements were collected, they could be useful in explaining the underlying process for the recognition results and in explaining the discrepant results for the recall test. Presumably, there would be more know judgments in the expected information condition than in the unexpected information condition. Know judgments would also be presumably higher for the recognition test than for the recall test (Wixted & Squire, 2010). This limitation of the present study can be overcome in future studies, by including these measurements.

The absence of differences across conditions in the recall test can also be explained by considering the nature of the two memory tests. Recall tests are usually more difficult than recognition tests. Recall tests involve two stages, but recognition tests involve only a single stage (Anderson & Bower, 1972; Kintsch, 1970). In recognition tests, the old information needs to be identified as being presented. On the other hand, in recall tests, the information needs to be retrieved first and then selected as the correct information among the retrieved items. Failure of either of these stages may result in reduced recall. In the present study, the difficulty of the recall test may have also limited the amount of information that is retrieved, making the scores in the two conditions more similar to each other. The performance of the participants on the recall test was 34% and 36% for the expected and unexpected information conditions respectively, even when the scoring criterion was lenient. In the recognition test, the retrieval performance was higher. Specifically, it was 63% in the unexpected information and 56% in the expected information conditions. Thus, both familiarity-based retrieval and retrieval difficulties may have contributed to the observation of the current findings.

Due to the null results that we obtained in the recall test, we further analyzed the types of errors that were made in the recall of the passages. The results showed that commission errors were higher in the expected information condition than in the unexpected information condition. Further analyses showed that commission errors were likely to be consistent with the information in the passage. Thus, stereotypically expected information may have been likely to receive attentive processing.

In general, our results indicate that an event description that involves stereotypical information may generally result in worse recognition and more false retrieval. Implications of these findings can be considered in eyewitness testimony research as well because they indicate that an eyewitness may be less accurate about the details of an event when the event details involve stereotypically expected information.

We also would like to consider some procedural details that can affect memory performance for stereotypically expected or unexpected information and make some suggestions for future research. Kleider and colleagues (2008) showed that information that is unexpected based on stereotypes was retrieved better in an immediate test, whereas expected information was retrieved better in the delayed test. Their immediate test was right after a 20-minute distractor task, while the delayed test was two days later. In our study, testing was done immediately after a 4-minute filler task, and our results were similar to Kleider et al. (2008) results for the immediate test. In future research, the time course of the retrieved information that is expected or unexpected based on age stereotypes can be investigated to gain insight into its persistence and recovery.

In the present study, we focused on age-related physical performance and social behavior stereotypes. Prior research shows that older adults internalize both physical and cognitive aging stereotypes. Future research can investigate the processing of age-related expected and unexpected information about cognitive performance by younger adults. Extension to aging stereotypes of cognition would provide information about the generalizability of the current findings.

There can also be some individual difference variables that can affect memory processes for stereotypically expected or unexpected information. One individual difference variable may be the prejudiced attitudes toward older adults. Swift et al. (2018) claimed that aging stereotypes can be both positive and negative. Negative stereotypes of aging are more dominant than positive ones (e.g., Kite et al., 2005; Kite & Johnson, 1988) and this escalates ageism (Greenberg et al., 2004). The two groups that were tested in the present study did not differ in their attitudes toward older individuals. Exploratory analyses also showed that the attitude measures did not correlate with memory measurements. Thus, individual differences in attitudes cannot be an explanation for the group differences that we found in recognition. Attitudes of our participants toward older adults were also quite positive. It should be noted though our attitude measures were taken explicitly. Nosek and colleagues (2007) collected online data about implicit stereotyping over six years. This large-scale data revealed a striking finding: The strongest implicit prejudice effect was observed for aging attitudes, where 80% of participants had difficulty in pairing old people with good items and young people with bad items. Moreover, these implicit aging attitudes were found to be cross-culturally consistent. Other studies also confirmed the automaticity of implicit negative age stereotypes (e.g., Krings, 2004). Our results also imply more automatic, i.e., less effortful and more error-prone processing of information that is expected based on age stereotypes. However, a systematic investigation of the relationship between implicit aging attitudes and memory measures can provide more direct evidence of the effects of individual differences in attitudes toward older adults.

Individuals may differ in the way that they handle stereotypical information as well. Need for cognition is defined as an individual’s tendency to engage in effortful cognitive activity and individuals who are high in need for cognition semantically process information in a deep and elaborative manner (Cacioppo et al., 1996). According to the fuzzy-trace theory, semantic processing leads to meaning-based gist traces, and the activation of semantic connections can lead to false memory (Brainerd & Reyna, 2002). Previous studies showed increased false memory among individuals who have high needs for cognition both in recognition (e.g., Graham, 2007) and in recall (e.g., Leding, 2011). Future research can investigate whether the findings can be generalized to both stereotypically expected and unexpected information.

Additionally, the need for cognitive closure as a motivational desire to resolve ambiguous, uncertain information that causes discomfort (Kruglanski & Webster, 1996) may also be an individual difference variable that might affect how an individual deals with unexpected information. The need for cognitive closure results in a search for stereotypically consistent information and the avoidance of inconsistent information in order to prevent any conflict with pre-existing schemas (e.g., Dijksterhuis et al., 1996). Further research can be carried out to determine the systematic effects of the need for cognitive closure in the retrieval and false retrieval of stereotypical information.

The present study contributed to the literature on how aging stereotypes affect memory processes. The results show that stereotypically expected information is retrieved less accurately. Perpetuation of stereotypical ageist expectations hinders older adults’ performance and well-being (for a review see Chasteen & Cary, 2015). Therefore, the ultimate aim of research should be to reduce stereotypical processing. There are some promising studies pointing to factors that can decrease stereotyping such as motivation to generate accurate social impressions, taking the perspective of an out-group member, and being exposed to counter-stereotypical information (for a review, see Plous, 2003). It is important to understand the cognitive processes for the information that is expected based on stereotypes so that the findings in this line of research can eventually be useful in reducing prejudice and discrimination.