Automatic Activation of Age Stereotypes
Is Attention to Category Information Sufficient for Stereotype Priming?
Abstract
Abstract. The current study investigated category-based activation of stereotypes when processing of the category primes is mandatory. In three high-powered pre-registered experiments (total n = 211), we compared responses to age-stereotypic traits (e.g., lonely) after presenting matching versus mismatching category primes (old vs. young faces) of which the age information had to be remembered. Experiments varied in stimulus-onset asynchronies (SOA; 250 ms vs. 500 ms) and in the inclusion of neutral conditions of prime and target factors. Consistently across all experiments, no facilitation of matching category primes was observed, indicating that category information alone does not facilitate processing of matching stereotypes even if it is attended. The theoretical and practical implications for activation and representation of stereotypes are discussed.
According to prevailing textbook accounts of stereotyping, stereotypes are represented as mental associations between categories and stereotypic attributes, and become activated automatically upon encountering a member of a social category (Fiske, 1998; Schneider, 2004). For instance, when encountering an old person, one may perceive this old person to be slow because the category of old automatically increases the accessibility of stereotypical attributes that are characteristic for this category (such as being slow, forgetful, wise, or generous). In support of the spreading of activation account, Banaji and Hardin (1996; Blair & Banaji, 1996) reported the finding that participants were faster to categorize names (e.g., Carol, Adam) or pronouns (e.g., she, he) as male or female if they were preceded by gender congruent stereotypical primes (e.g., nurse, doctor). Although the facilitation effects were seen as evidence for automatic spreading of activation between stereotypes and categories, the effects can also be explained by response priming: Responses to target names were facilitated (delayed) by gender congruent (incongruent) primes simply because the prime triggers the same (a different) response as the target in a gender categorization task (Wentura & Degner, 2010). Supporting this explanation, recent studies using tasks that prevented response priming and were only sensitive to semantic priming effects (e.g., lexical decision task, henceforth LDT; Casper et al., 2011, Experiment 3) failed to reveal priming effects for stereotypic targets. In a high-powered pre-registered replication study of stereotype activation (Banaji & Hardin, 1996), Müller and Rothermund (2014) employed a semantic categorization task in which participants were asked to categorize gender-related (e.g., Carol, Adam) and gender-unrelated targets (e.g., Hamburg, Berlin) as a name of a person or a town after the presentation of gender-stereotypical primes (e.g., ballet, computer), which eliminates effects of response priming, since the response tendency that is triggered by the prime is unrelated to the response required for the target. No facilitation effects were found for matching prime–target pairs. Corroborating these findings, Tsamadi et al. (2020) and White et al. (2018) found stereotype-based priming effects only in tasks that are sensitive to response competition effects (i.e., response priming paradigms), but not in tasks that measure only semantic priming (i.e., semantic priming paradigms). A recent meta-analysis revealed that stereotype priming effects in paradigms that controlled for response priming are negligibly small (Kidder et al., 2018). These accumulating null findings challenge the assumption of a simple category-based activation of stereotypes.
More recent research revealed that an activation of stereotypes can be obtained only if category information is combined with matching context information (Casper et al., 2010, 2011; Huang & Rothermund, 2022; Wigboldus et al., 2003). For instance, the attribute “slow” was processed faster after primes in which information regarding the category “old” was combined with a matching context (e.g., “is crossing the street”), whereas no facilitation was observed for old primes that were combined with a context that is unrelated to the specific stereotypic attribute (e.g., “is watering the flowers”).
A direct comparison of those studies that reported evidence for such a context-dependent activation of stereotypes with studies that tested (and failed to find) priming effects of simple category primes reveals that in the former experiments, prime processing was enforced by including an additional memory task which required attention to the primes. In those studies that obtained null findings for simple category primes, on the other hand, no such measures were taken to ensure the processing of the prime information. Attention to primes, however, has been shown to be crucial for the emergence of priming effects. In a series of studies, Spruyt et al. (2007, 2009, 2011) manipulated selective attention for the semantic dimension of the stimulus information and found that semantic priming effects emerged only if participants were encouraged to attend to the semantic dimension of the stimuli that defined the relatedness of primes and targets (e.g., gender, valence). Specifically for stereotype priming, since individuals typically belong to multiple social categories, priming effects may only be obtained for the most salient category (Macrae et al., 1995). Accordingly, one would notice the age cues of an old man who sits in a university class, which might elicit stereotypic attributes about old people such as being slow, whereas encountering the same old man in a church meeting which is usually dominated by women may alternatively activate gender-related stereotypic traits such as being physically strong. Thus, if a priming procedure does not emphasize the category dimension on which attention should be focused, it is possible that the category that is related to the relevant stereotype dimension in the targets becomes ignored. Stereotype priming effects might thus be concealed or masked by not controlling for the attention to the relevant category information in the prime.
Against this backdrop, the main aim of the current study is thus to test whether category-associated stereotypical attributes can be activated in a conditionally automatic fashion by simple category information if the social category information contained in the primes is attended and processed. To enforce such a processing of the relevant category information in the prime without extending the stimulus-onset asynchrony between the prime and probe stimuli and thus inviting a more strategic processing of the primes, we employed an additional memory task which was inserted after the lexical decision task to check the attention paid to primes (Casper et al., 2010, 2011; Huang & Rothermund, 2022). Moreover, the additional memory task required that the age information in the primes was remembered and thus narrowed participants’ attention to the relevant category information that was assessed with the targets and prevented participants from focusing on other possible categories contained in the primes (e.g., gender). The LDT still guarantees that priming effects that are obtained with this procedure can be considered to be (conditionally) automatic in the sense of being unintentional, since the LDT does not require any processing of age information, nor does it profit from this information, which does not translate into one of the required responses (word vs. nonword). Due to the short SOA, priming effects can also be considered automatic in the sense of being efficient, since they do not depend on controlled/strategic processing resources, which require more processing time (Neely, 1989). However, since primes were presented above threshold and processing of the age information of the primes was required by the task, the resulting priming effects should not be considered being automatic in the sense of reflecting the result of unconscious or goal-independent processes (Bargh, 1994; Moors & De Houwer, 2006).
Overview of the Current Study
In a series of three experiments, a 2 (category prime: young vs. old) × 2 (target: stereotypically young vs. old) design was adopted to examine category-based stereotype priming effects which were computed as the difference in RTs for age stereotypical traits (e.g., slow) between matching (e.g., old) and mismatching (e.g., young) category primes.
In Experiment 2, we additionally included age-unrelated primes as a neutral condition for the prime factor. By comparing the difference in responses to age stereotypical targets between age-related and age-unrelated category primes, we examined the possibility of general activation effects of stereotypic traits driven by both matching (i.e., spread activation effects) and mismatching category primes (i.e., paradoxical activation effects). In this study, we also included age-unrelated targets as a neutral condition for the target factor.
In Experiment 3, we introduced a shorter SOA of 250 ms (compared with an SOA of 500 ms in Experiments 1 and 2) to exclude potential strategic processing of counteracting stereotype activation and also to capture fast-acting stereotype priming effects which may be seen only shortly after the onset of primes.
Experiment 1
Young or old pictures were presented as category primes after which young or old stereotypical traits were presented as targets in a lexical decision task. Similar to the study by Casper et al. (2011, Experiment 1), we inserted the additional memory task randomly after a quarter of the lexical decision trials, which has already been shown to be a valid manipulation to guarantee attention to category information in the category primes.
Method
Sample
We conducted a power analysis using G*Power 3 (Faul et al., 2007) for a repeated-measures ANOVA, which yielded a sample size of n = 65 to reach a power of 1 − β = .8 to detect a medium-sized effect1 (f = .25). For the power analysis, the alpha was set to .05, and calculations were based on the conservative assumption that response times for matching and mismatching combinations of category prime and stereotype targets were uncorrelated. Data were collected from 66 participants (49 male, M = 24.52 years, SD = 3.53, age range 18–30 years) who speak German as native language. The sample was recruited via Prolific, and participants were paid £1.25 for their participation. Three participants were excluded from the analyses due to a high amount of erroneous responses in the LDT or in the additional task (mean error rate > 40%). This study was pre-registered on OSF (https://osf.io/5nydf).
Materials
Based on the study by Casper et al. (2011), we selected 16 attributes representing stereotypes of young and old people (eight for the category young, eight for the category old; half of each set were positive and the other half were negative) as target words for the lexical decision task. The stereotypicality of the selected attributes was rated on a Likert scale ranging from 1 (= typical for young people) to 7 (= typical for old people) in another sample, N = 40. Stereotypicality ratings for stereotypically young (M = 2.17) and old traits (M = 5.56) differed as expected, t(39) = 22.06, p < .001, d = 3.49. Based on the attribute targets, the same number of nonword targets (e.g., sturk) was created by replacing letters of the word targets (e.g., stark). All the nonword targets are pronounceable and with similar length of the word targets. A complete list of trait words is shown in the Appendix (Table A1).
Category primes consisting of eight pictures of young faces and eight pictures of old faces (half male and half female for each age category) were taken from an established database for which age categorizations have been validated (Minear & Park, 2004). Similar to the study by Casper et al.'s (2011) in which significant priming effects of the category primes were found in combination with relevant contexts, we slightly blurred the faces but made sure that the age information was still clearly visible to avoid unwanted activation of stereotypic traits due to individual facial features. In order to test whether the facial primes could be easily categorized by age under speeded categorization conditions, we conducted an additional validation study (N = 86) in which pictures of young, middle-aged, and older adults were presented for 500 ms, and had to be categorized according to their age. The overall accuracy rate of age categorization for all faces that were chosen for our experiments was 85.42%.
Procedure
The formal experiment included two blocks of 64 trials of a lexical decision task. Within each block, each of the 16 stereotypical targets was assigned once to one of the two possible types of category primes, resulting in 32 trials. Another 32 trials were created by combining the same set of primes that were presented in the word trials with randomly selected nonwords. All trials within a block were presented in a random sequence. The presentation of stimuli and recording of responses were controlled by using the PsychoPy program.
The formal experiment started after eight practice trials using category primes and targets (e.g., powerful) that were not presented in the formal experiment. Each trial had the following temporal sequence of events (see Figure 1): First, a fixation cross was presented for 750 ms. Afterward, a face of a young or old person was presented for 500 ms. After the prime presentation, the prime was replaced immediately by the target (a word or a nonword; SOA = 500 ms) and participants had to classify the target as a word or nonword by pressing one of two keys on the keyboard (“D” = nonword or “L” = word). The target remained on the screen until a response was registered. After that, a blank screen appeared for 250 ms.
For a randomly selected quarter of all trials, an additional memory task was inserted into the trial procedure before the next trial started. In these trials, a category label (“young” or “old”) was shown as a probe and participants had to indicate whether the probe indicated the same or a different category information than the prime that had been presented before; responses were given by pressing one of two keys on the response pad (the same keys were used as for the lexical decision task: “D” = different, “L” = same). The probe remained on the screen until a response was registered. For both the lexical decision task and the additional task, an error or a “too slow” prompt would appear after wrong responses or responses longer than 1,500 ms.
Results
As is common practice for the lexical decision task, only responses to target words were included in all our analyses. Erroneous responses (2.6%) and outlier values that were more than three interquartile ranges above the 75th percentile of each subject’s individual reaction time distribution (“far out values” according to Tukey, 1977; 1.2%) or faster than 300 ms (0.02%) were excluded from the analysis. The average accuracy rate of responding for the additional memory task was 90.27%, indicating successful processing of the category information.
Since our data were hierarchically structured, we decided to use multilevel analysis to test the priming effects for stereotypic traits produced by category primes.2 The multilevel analysis was conducted on the basis of individual trials as nested within subjects in which prime (young = −1 vs. old = 1) and target (young = −1 vs. old = 1) were included as predictors on the level of trials. The multilevel analysis allowing for random intercepts yielded no main effects of prime or target, β = 1.28, t(3,810) = .52, p = .61, β = 1.33, t(3,810) = .54, p = .59. We found a marginally significant interaction between prime and target, β = 4.33, t(3,810) = 1.75, p = .08. However, the pattern is contrary to expectations with old (young) stereotypic targets being (nonsignificantly) facilitated by young (old) primes. See Table 1 for means and standard errors across conditions. Though the responses exceeding 1500 ms that were included in our final analyses are very rare, we still examined whether excluding these responses would reveal a different pattern of results. Thus, we conducted the multi-level analysis using only responses below 1500 ms. Similarly, we found no prime x target interaction effect, β = 2.86, t(3791) = 1.23, p = .22.
Discussion
Replicating previous studies on category priming (Casper et al., 2011; Kidder et al., 2018; Müller & Rothermund, 2014), we found no facilitation effects of age information on lexical decision times for age-related traits after category primes containing category information that matches the stereotypic trait that was presented as a target. Given that our null finding was obtained under conditions of focused attention to age information in the primes, this finding can be seen as further evidence against a simple category-based activation of age stereotypes, which rules out an explanation of the null finding in terms of a lack of feature-specific attention (Spruyt et al., 2007, 2009, 2011).
Experiment 2
Stereotype priming effects in Experiment 1 were computed as the difference in RTs between matching and mismatching category primes. Research from the semantic priming literature, however, may be taken to suggest that even a mismatching category can sometimes give rise to facilitation effects, similar to priming or association effects for antonyms and opposites (e.g., “black – white,” “winter” – “summer”; Lucas, 2000; Perea & Rosa, 2002). According to this idea, an old prime may not only facilitate attributes that are stereotypically old (e.g., “frail”) but may also facilitate the processing of opposite attributes that are part of the stereotypes of young people (e.g., “healthy”) if there is a strong association between old and not being healthy (see also De Houwer et al., 2015). If these paradoxical activation effects of mismatching category primes indeed influence responding in the LDT, the lack of a difference in RTs between matching and mismatching category primes that was found in Experiment 1 – and in similar studies that were reported in the literature (e.g., Casper et al., 2011; Kidder et al., 2018; Müller & Rothermund, 2014) – may not indicate a lack of stereotype activation effects but may instead reflect a general activation of stereotypes for matching and mismatching stereotypes alike.
In addition to examine the stereotype priming effects by comparing matching and mismatching category prime conditions, we also tested whether both matching and mismatching category primes elicit priming effects for stereotypical targets in Experiment 2. To achieve the goal, we introduced neutral conditions for the prime (an image of white noise) and for the target factor (age-unrelated traits; e.g., “cold”) and we tested whether such a general activation of matching and mismatching category stereotypes produces facilitation effects for the age-related category primes (both matching and mismatching) compared with the neutral prime condition. For age-unrelated targets, however, no such facilitation effects should occur.
Another change in Experiment 2 was that we inserted the category probe in each trial after the lexical decision response to see whether priming effects of category primes were magnified by more emphasis on the additional memory task assuring attention to the age information contained in the primes.
Method
Sample
We conducted a power analysis for a repeated-measures ANOVA with a small- to medium-sized effect (f = .2)3 and a power of 1 − β = .8 (alpha was set to .05), which yielded a sample size of n = 84. Again, a priori power calculations were based on the conservative assumption that our repeated measures were uncorrelated. Data were collected from 83 participants (47 male, M = 24.07 years, SD = 4.44, age range 18–31 years) who speak German as native language. The sample was recruited via Prolific, and participants were paid £2.50 for their participation. Five participants were excluded from the analyses due to high amount of erroneous responses in the LDT or in the additional task (mean error rate > 40%). This study was pre-registered on OSF (https://osf.io/9vtgr).
Materials and Procedure
Materials and the procedural details were identical with Experiment 1, except for the following changes: Eight age-unrelated attributes (half positive and half negative) were additionally selected as target words for the lexical decision task. Stereotypicality ratings for the age-neutral targets (M = 4.30) were higher than for the stereotypically young (M = 2.17; t[39] = 18.19, p < .001, d = 2.88) and lower than for the stereotypically old (M = 5.56; t[39] = −12.33, p < .001, d = 1.97) targets. The neutral and age-related target words were matched on length and frequency. An image of white noise was presented as a neutral category prime.4 To assure attention to the age information contained in the primes, a category probe (“young,” “old,” or “white noise”) was now inserted in each trial after the lexical decision response. In total, the experiment included a practice block of 18 trials and two experiment blocks of 144 trials of a lexical decision task. The primes and targets used in practice and experiment blocks were different stimuli. Within each experimental block, each of the 16 stereotypical plus eight age-unrelated targets was assigned once to one of the three possible types of category primes, resulting in 72 trials. Another 72 trials were created by combining the same set of primes that were presented in the word trials with randomly selected nonwords.
Results
Erroneous responses in the lexical decision task (6.1%) and outlier values that were more than three interquartile ranges above the 75th percentile of each subject’s individual reaction time distribution (“far out values” according to Tukey, 1977; 1.2%) or faster than 300 ms (0.1%) were excluded from the analysis.
The average accuracy rate of responding for the additional memory task was 91.12%, indicating successful processing of the category information.
A multilevel analysis with prime (young vs. old vs. neutral) and target (young vs. old vs. age-unrelated) as predictors on the level of trials yielded no main effect of prime, β = −.14, t(10,281) = −.07, p = .95, or target, β = −1.96, t(10,281) = −.97, p = .33. Similar to Experiment 1, we also found no interaction between prime and target, β = −2.82, t(10,281) = −1.14, p = .26, indicating no evidence for category priming effects. See Table 2 for means and standard errors across conditions.
After grouping the old and young primes and targets together into the age-related prime and target conditions, we conducted a 2 (prime: age-related vs. age-unrelated) × 2 (target: age-related vs. age-unrelated) multilevel analysis. The analysis revealed neither a main effect of the prime factor, β = −1.60, t(10,281) = −.86, p = .39, nor of the target factor, β = −.09, t(10,281) = .05, p = .96, nor of the prime × target interaction, β = −.83, t(10,281) = −.45, p = .65, indicating no evidence for general activation effects of age-related category primes. The multi-level analysis using only responses below 1500 ms also revealed no prime x target interaction effect, β = −3.14, t(10191) = −1.40, p = .16.
Discussion
The results of the second experiment replicate the findings of the first experiment in that there was no facilitation effect after matching category primes compared to nonmatching category primes. In addition, going beyond the findings from the first experiment, comparisons with the neutral prime (i.e., white noise picture) and neutral target (i.e., age-unrelated attributes) conditions revealed that there was no general facilitation effect in matching and mismatching prime conditions. The lack of finding a difference between matching and mismatching category primes in these and other previous experiments thus cannot be attributed to a neutralization of a facilitation effect in the matching category condition by a parallel activation of stereotypes by antagonistic categories.
Experiment 3
In the previous experiments, the temporal interval between the onset of the category prime and the onset of the stereotypical target was set to 500 ms. We chose this interval to guarantee optimal attention to the primes; furthermore, robust priming effects were found with this SOA for semantically associated picture–word pairs (e.g., sugar – sweet) in the lexical decision task in a previous study (Casper et al., 2011, Experiment 3). A recent meta-analysis by Kidder et al. (2018), however, reported evidence for more reliable stereotype priming effects in studies that used shorter SOAs (<350 ms), indicating that an associative activation effect of stereotypical targets might only be captured shortly after the onset of the category primes. Thus, the first two experiments that were reported above might not have been able to detect possible stereotype priming effects if these effects had a very fast decay function.
To capture priming effects that may appear only very early after the onset of primes, we introduced a short SOA of 250 ms in the third experiment. Furthermore, using a shorter SOA also allowed us to rule out the influence of controlled processes, which might have counteracted stereotype activation effects in Experiments 1 and 2. Although such an influence is unlikely, we still cannot rule out that participants may want to counteract stereotype activation to appear as being unbiased (Blair & Banaji, 1996; Blair et al., 2001; Dasgupta & Greenwald, 2001).
We adopted the same design as in the first experiment that focused only on the difference between matching and mismatching prime–target conditions, which guarantees a maximum number of trials entering into the matching and mismatching conditions. Two minor changes were made in Experiment 3. We presented unblurred pictures of old and young faces as primes to ensure successful processing of the age information even for the very briefly presented category primes. To increase the reliability of stereotype priming effects, we added an additional experimental block to increase the number of LDT trials that were entered into the analyses, resulting in three experimental blocks in total.
Method
Sample
We conducted a power analysis using G*Power 3 (Faul et al., 2007) for a one-tailed t-test5 comparing the matching and mismatching prime–target conditions (matched pairs), which yielded a sample size of n = 71 to reach a power of 1 − β = .8 to detect a small- to medium-sized effect (d = .3). For the power analysis, the alpha was set to .05. Data were collected from 72 participants (20 male, mean age = 21.44 years, SD = 3.99, age range 18–30 years) who speak German as native language. The sample was recruited via Prolific, and participants were paid £1.25 for their participation. Two participants were excluded from the analyses due to a high amount of erroneous responses in the LDT task (mean error rate > 40%). This study was pre-registered on OSF (https://osf.io/7cdf2).
Materials and Procedure
Materials and the procedural details were identical to Experiment 1, except for the following changes: Unblurred faces of young or old persons were presented as prime pictures for a duration of 250 ms. In total, the experiment included three blocks of 64 trials of a lexical decision task. Within each block, each of the 16 targets was assigned once to one of the two possible types of category primes, resulting in 32 trials. Another 32 trials were created by combining the same set of primes that were presented in the word trials with randomly selected nonwords. Similar to Experiment 1, the additional memory task was implemented after a random quarter of the lexical decision trials.
Results
Erroneous responses (3.8%) and outlier values that were more than three interquartile ranges above the 75th percentile of each subject’s individual reaction time distribution (“far out values” according to Tukey, 1977; 1.1%) or faster than 300 ms (0.04%) were excluded from the analysis. The average accuracy rate of responding correctly in the additional memory task was 88.16%, indicating successful processing of the category information in the primes.
A multilevel analysis with prime (young = −1 vs. old = 1) and target (young = −1 vs. old = 1) as predictors, allowing for random intercepts, yielded a significant main effect of prime, β = 4.96, t(6,316) = 2.80, p = .005, suggesting faster responses after young prime than after old primes. We found no main effect of target, β = 1.03, t(6,316) = .58, p = .56. Importantly, there was no interaction effect between prime and target, β = −1.10, t(6,316) = −.62, p = .53, suggesting no statistically detectable evidence of stereotype priming effects.6 The multi-level analysis using only responses below 1500 ms also revealed no prime × target interaction effect, β = −1.11, t(6308) = −.64, p = .52. See Table 3 for means and standard errors across conditions.
Discussion
Consistent with the first two experiments, we found no priming effects of stereotypical traits produced by the category primes even when the targets were presented very shortly (250 ms) after the onset of the primes. Thus, the absence of priming effects by category primes in the first two experiments cannot be attributed to the counteracting of stereotype activation or a fast decay of stereotype priming effects that went undetected due to the longer SOA. The null findings of stereotype priming effects across the three experiments with short as well as long SOAs demonstrate that regardless of the temporal interval between the onset of the category prime and the onset of the stereotypical traits, the processing of young and old category primes is unable to produce statistically detectable activation effects of stereotypical traits.
Further Analyses
To increase the power of our analyses, we used the data from the conditions that were common across our three experiments and conducted a three-level multilevel analysis (i.e., individual trials nested within subjects from three experiments). The joint analysis revealed no significant main effect of prime, β = 2.25, t(20,561) = .92, p = .36, or target, β = −3.15, t(20,561) = −1.10, p = .27. Similar to the results of the individual experiments, we found no interaction between prime and target, β = .23, t(20,561) = .16, p = .88, indicating no category priming effects across the three experiments.
To examine whether the stereotype priming effects were diminished by the repetition of primes throughout the experiment, we introduced block as an additional Level 1 factor (i.e., two blocks in Experiments 1 and 2, three blocks in Experiment 3) and conducted a multilevel analysis for each experiment. We found no interaction of the priming effect with the block factor (all ps > .30), speaking against the possibility that category priming effects were present during the beginning of the experiments but then washed out due to practice and excessive repetitions.
According to Verhaeghen et al. (2011), the emergence of stereotype priming effects is dependent on the objective associative strength between prime–target pairs, indicating that stronger prime–target associations may lead to stronger priming effects. We tested this idea in our study by introducing stereotype extremity (stereotypically extreme vs. less extreme targets) as an additional factor into the multilevel analysis for each experiment. Consistently across all experiments, we found no three-way interaction between prime, target, and stereotype extremity (all ps > .12), indicating no difference in the strength of priming effects between high and low stereotypical traits. Accordingly, stereotype priming effects were absent even for highly stereotypical target.
General Discussion
Across three experiments, we found no evidence for category priming effects although processing of the age information in the primes was mandatory. To further investigate whether the absence of category priming effects observed in our experiments and in previous studies (Casper et al., 2011; Müller & Rothermund, 2014) might be due to possible facilitation effects of mismatching category primes, we additionally included neutral baseline conditions for the primes and targets in the second experiment. Similar to the first experiment, the findings of Experiment 2 did not support the hypothesis of a category-based activation of stereotypes, neither for matching nor for mismatching category primes. In the third experiment, we introduced a short SOA of 250 ms to rule out possible strategic processing, which may counteract the stereotype activation, and to detect activation effects with a very fast decay. None of these measures revealed evidence for automatic stereotype activation.
The absence of a detectable stereotype priming effect for mere category primes in our studies is in line with the findings of a larger literature that tested stereotype priming with the lexical decision task (Kidder et al., 2018). The present study went beyond these previous experiments in that (a) we added an additional memory task that rendered processing of the category information in the primes obligatory to ensure a sufficient amount of feature-specific attention allocation to the relevant category information (Spruyt et al., 2007, 2009, 2011), (b) we included neutral baseline conditions to control for possible facilitation effects in the mismatching condition, and (c) we adopted short and long SOAs to rule out potential strategic efforts to counteract stereotype activation and to capture priming effects for stereotypical traits which may appear at different periods after the onset of category primes. Still, our findings consistently replicate the null findings that were also reported in most previous studies.
In apparent contrast with the accumulated null findings of simple category-based stereotype activation effects, some studies using the LDT did find stereotype prime effects (e.g., Kawakami et al., 2002; Verhaeghen et al., 2011). By comparing these studies with our study, we found one important methodological difference regarding the assessment of stereotype priming effects. In our studies as well as in other studies that failed to find evidence for category-based stereotype activation effects (e.g., Casper et al., 2011, Experiment 3; Müller & Rothermund, 2014), priming effects were calculated as the response time difference for matching versus mismatching combinations of stereotypic primes and targets (e.g., old-experienced vs. young-experienced). Those studies that reported evidence for a category-based activation of stereotypes, however, computed stereotype activation effects as response time difference between stereotype-related targets and stereotype-unrelated targets only within the matching category primes (e.g., old-experienced vs. old-active). Due to the lack of a mismatching category prime condition (e.g., young), these studies do not allow for a computation of priming effects properly. The alleged activation effects just reflect simple main effects of the target factor and thus cannot be interpreted as evidence for automatic stereotype activation (e.g., the same reaction time difference might also obtain for mismatching category primes). Given that stereotype activation refers to the activation effects of stereotypic traits that are exclusively driven by the matching category primes rather than mismatching category prime, the facilitation effects for stereotype-related targets that were reported by these studies may not be reliable evidence for stereotype activation.
The absence of detectable stereotype priming effect driven by simple category information has important implications for the activation and representation of stereotypes. The process that underlies stereotype activation was initially explained by a semantic network model in which stereotypes are mentally represented as associations between social category and related stereotypical attributes, and the mere processing of category information should increase the accessibility of associated stereotypical traits via a process of spreading of activation (e.g., Collins & Loftus, 1975). More recently, the parallel distributed processing model of memory (PDP; Rumelhart et al., 1986) has become a standard account to explain facilitation effects between concepts and the resulting priming effects. According to such an account, primes and targets are represented by a large number of simple, interconnected units, representing basic features or attributes. Priming effects are then explained by the amount of overlap or similarity between the mental representation of primes and targets, with associated primes and targets sharing multiple features, which allows the system to easily switch from the prime to the target representation. However, both the simple category-centered network model of stereotypes and the PDP model are challenged by null findings of simple category priming studies (Casper et al., 2011; Kidder et al., 2018; Müller & Rothermund, 2014; Tsamadi et al., 2020; White et al., 2018).
Going beyond these associationistic or connectionistic frameworks, recent research supports the notion that stereotype activation is context-dependent, suggesting that the stereotypes are mentally represented as context-specific schemas, each of which comprises and connects a combination of a social category and a specific situational context with a specific trait information that is relevant for the respective situation (Casper et al., 2010, 2011; Huang & Rothermund, 2022; Wigboldus et al., 2003). Corresponding to this context-dependent account, one recommendation for future stereotype priming research the current studies could provide is that not only category but also contextual information should be taken into account to activate stereotypes.
However, a previous study investigating stereotype activation with combinations of category and context primes still reported only small stereotype activation effects in the lexical decision task, ranging from 6 ms to 8 ms (Huang & Rothermund, 2022). These findings suggest that stereotypes might be represented in a more propositional format rather than in an associative fashion (De Houwer et al., 2020; Kurdi & Dunham, 2020), which specifies the exact meaning of how people belonging to a certain group are assumed to be or to behave in a certain situation (e.g., “old people walk slowly”). This propositional account of stereotypes merits further scrutiny; a promising avenue for future research is thus to focus on specific context-dependent age stereotypes and to use paradigms that are specifically designed to assess processes of an automatic activation of beliefs (Propositional Evaluation Paradigm, Müller & Rothermund, 2019).
Limitations and Outlook
Despite the consistent lack of significant priming effects in our study, we should be aware of the difficulty in ultimately rejecting a hypothesis (or accepting a null hypothesis), and these results must be interpreted with caution due to the limitations in terms of specific content and methodologies.
Consistently across three experiments, our results strongly favor the null hypothesis of no activation effects of social stereotypes for simple category primes. A sensitivity analysis for our study revealed that the joint analysis had sufficient power (1 − β = .8) to detect effects of d ≥ .02, meaning that our study is able to rule out the existence of even a negligibly small stereotype priming effect with a high probability. But still we cannot completely exclude the existence of an extremely small stereotype activation effect by mere category information.
Importantly, the present study did not test priming effects for social categories in general but instead focused only on age stereotypes. More research with other stereotypes is needed to yield a more comprehensive picture of the existence or lack of category-based stereotype activation effects.
Finally, although the lexical decision task (LDT) has been developed as a measure of associations in semantic memory (Neely, 1977, 1991; Wentura & Degner, 2010; Wentura & Rothermund, 2014) and has been widely used in stereotype priming research (Casper et al., 2010, 2011; Chasteen et al., 2002; Müller & Rothermund, 2014; Rothermund et al., 1995; see Kidder et al., 2018, for a review), our findings still require further exploration in future research using alternative task (e.g., naming task, semantic decision task).
Electronic Supplementary Materials
The electronic supplementary material is available with the online version of the article at https://doi.org/10.1027/1864-9335/a000503
References
(1996). Automatic stereotyping. Psychological Science, 7(3), 136–141. 10.1111/j.1467-9280.1996.tb00346.x
(1994). The four horsemen of automaticity: Awareness, intention, efficiency, and control in social cognition. In R. S. Wyer, JrT. K. Srull (Eds.), Handbook of social cognition: Basic processes; applications (pp. 1–40). Lawrence Erlbaum Associates, Inc.
(1996). Automatic and controlled processes in stereotype priming. Journal of Personality and Social Psychology, 70(6), 1142–1163. 10.1037/0022-3514.70.6.1142
(2001). Imagining stereotypes away: The moderation of implicit stereotypes through mental imagery. Journal of Personality & Social Psychology, 81(5), 828–841. 10.1037/0022-3514.81.5.828
(2010). Automatic stereotype activation is context dependent. Social Psychology, 41(3), 131–136. 10.1027/1864-9335/a000019
(2011). The activation of specific facets of age stereotypes depends on individuating information. Social Cognition, 29(4), 393–414. 10.1521/soco.2011.29.4.393
(2002). The activation of aging stereotypes in younger and older adults. The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences, 57(5), 540–547. 10.1093/geronb/57.6.P540
(1975). A spreading-activation theory of semantic processing. Psychological Review, 82(6), 407–428. 10.1037/0033-295X.82.6.407
(2001). On the malleability of automatic attitudes: Combating automatic prejudice with images of admired and disliked individuals. Journal of Personality and Social Psychology, 81(5), 800–814. 10.1037/0022-3514.81.5.800
(2015). The relational responding task: Toward a new implicit measure of beliefs. Frontiers in Psychology, 6, Article 319. 10.3389/fpsyg.2015.00319
(2020). Attitudes beyond associations: On the role of propositional representations in stimulus evaluation. In B. Gawronski (Ed.), Advances in experimental social psychology (pp. 127–183). Elsevier Academic Press. 10.1016/bs.aesp.2019.09.004
(2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39(2), 175–191. 10.3758/BF03193146
(1998). Stereotyping, prejudice, and discrimination. In D. T. GilbertS. T. FiskeG. Lindzey (Eds.), The handbook of social psychology (4th ed., Vol. II, pp. 357–411). McGraw-Hill.
(2022). Automatic stereotype activation: Further evidence against simple category-based activation. https://osf.io/c3v8x.
(2022). Implicit and explicit age stereotypes assessed in the same contexts are still independent. Experimental Aging Research. (In Press). 10.1080/0361073X.2022.2039507
(2002). Automatic stereotyping: Category, trait, and behavioral activations. Personality and Social Psychology Bulletin, 28(1), 3–15. 10.1177/0146167202281001
(2018). Sequential stereotype priming: A meta-analysis. Personality and Social Psychology Review, 22(3), 199–227. 10.1177/1088868317723532
(2020). Propositional accounts of implicit evaluation: Taking stock and looking ahead. Social Cognition, 38(Suppl), 42–67. 10.1521/soco.2020.38.supp.s42
(2000). Semantic priming without association: A meta-analytic review. Psychonomic Bulletin & Review, 7(4), 618–630. 10.3758/BF03212999
(1995). The dissection of selection in person perception: Inhibitory processes in social stereotyping. Journal of Personality and Social Psychology, 69(3), 397–407. 10.1037//0022-3514.69.3.397
(2004). A lifespan database of adult facial stimuli. Behavior Research Methods, Instruments & Computers, 36(4), 630–633. 10.3758/BF03206543
(2006). Automaticity: A theoretical and conceptual analysis. Psychological Bulletin, 132(2), 297–326. 10.1037/0033-2909.132.2.297
(2014). What does it take to activate stereotypes? Simple primes don't seem enough: A replication of stereotype activation (Banaji & Hardin, 1996; Blair & Banaji, 1996). Social Psychology, 45(3), 187–193. 10.1027/1864-9335/a000183
(2019). The Propositional Evaluation Paradigm (PEP): Indirect assessment of personal beliefs and attitudes. Frontiers in Psychology, 10, Article 2385–2398. 10.3389/fpsyg.2019.02385
(1977). Semantic priming and retrieval from lexical memory: Roles of inhibition less spreading activation and limited-capacity attention. Journal of Experimental Psychology: General, 106(3), 226–254. 10.1037/0096-3445.106.3.226
(1991). Semantic priming effects in visual word recognition: A selective review of current findings and theories. In D. BesnerG. W. Humphreys (Eds.), Basic processes in reading: Visual word recognition (pp. 264–336). Lawrence Erlbaum.
(1989). Semantic priming in the lexical decision task: Roles of prospective prime-generated expectancies and retrospective semantic matching. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15(6), 1003–1019. 10.1037/0278-7393.15.6.1003
(2002). The effects of associative and semantic priming in the lexical decision task. Psychological Research, 66(3), 180–194. 10.1007/s00426-002-0086-5
(1995). Selbstwertschützende Verschiebungen in der Semantik des Begriffs "alt" im höheren Erwachsenenalter (Protecting self-esteem by shifting the semantics of the concept "old" in old age). Sprache und Kognition, 14(2), 52–63. 10.1007/978-3-531-91900-3_9
(1986). A general framework for parallel distributed processing. In D. E. RumelhartL. McClelland (Eds.), Parallel distributed processing: Explorations in the microstructure of cognition (Vol. 1, pp. 45–76). MIT Press.
(2004). The psychology of stereotyping. Guilford.
(2011). Unconscious semantic activation depends on feature-specific attention allocation. Cognition, 122(1), 91–95. 10.1016/j.cognition.2011.08.017
(2007). Affective priming of nonaffective semantic categorization responses. Experimental Psychology, 54(1), 44–53. 10.1027/1618-3169.54.1.44
(2009). Modulation of automatic semantic priming by feature-specific attention allocation. Journal of Memory and Language, 61(1), 37–54. 10.1016/j.jml.2009.03.004
(2020). Stereotype-based priming without stereotype activation: A tale of two priming tasks. Quarterly Journal of Experimental Psychology, 73(11), 1939–1948. 10.1177/1747021820925396
(1977). Exploratory data analysis. Addison-Wesley.
(2011). Prime and prejudice: Co-Occurrence in the culture as a source of automatic stereotype priming: Prime and prejudice. British Journal of Social Psychology, 50(3), 501–518. 10.1348/014466610X524254
(2014). Priming is not priming is not priming. In D. C. Molden (Ed.), Understanding priming effects in social psychology (pp. 49–69). The Guilford Press.
(2018). Taking priming to task: Variations in stereotype priming effects across participant task. Social Psychology, 49(1), 29–46. 10.1027/1864-9335/a000326
(2003). When stereotypes get in the way: Stereotypes obstruct stereotype-inconsistent trait inferences. Journal of Personality and Social Psychology, 84(3), 470–484. 10.1037/0022-3514.84.3.470
Appendix
1We aimed at detecting medium sized effects based on effect sizes that were reported in previous priming studies in which attention to primes was obligatory (e.g., the smallest effect size for the priming effects reported in Spruyt et al., 2011, is also medium-sized, f = .30). Furthermore, to warrant practical relevance for contexts of stereotype activation and application, priming effects should at least be medium sized.
2As suggested by reviewers, we only presented the multilevel results for all our experiments. Since we pre-registered to use repeated measures ANOVAs to test our hypothesis, we also conducted repeated measures ANOVAs, which revealed highly similar results with the multilevel results (see the Electronic Supplementary Material, ESM 1).
3Since we were unable to detect the medium sized effect as we aimed for in Experiment 1, we slightly reduced the effect size we aimed for in Experiment 2 to a small to medium sized effect (f = .2).
4We used images of white noise as neutral category primes rather than photos of middle-aged persons because pilot testing revealed that presenting middle-aged primes led to extremely long response times to targets compared with young and old primes, possibly due to the fact that it is more difficult to identify a face as being middle-aged.
5We used one-tailed t-tests rather than an F test to compute the required sample size because the t-test is equivalent with an F test with df=1 in the numerator, but allows us to test our directional hypothesis of facilitation effects for matching category primes, which is in line with the entire literature on automatic stereotype activation (e.g., Blair & Banaji, 1996; Kidder et al., 2018).
6Since the power analysis was conducted for a t-test, we also conducted a one-tailed t-test to compare the matching and mismatching prime-target conditions in an analysis of aggregated data, and we found no difference, t(69) = −.74, p = .46.
