Implicit and Explicit Age Stereotypes Assessed in the Same Contexts are Still Independent

ABSTRACT

Objective: In a series of three studies (N = 187), we investigated the correlation between implicit and explicit age stereotypes, both of which were assessed in a context-dependent way. Methods: To assess implicit age stereotypes, we presented combinations of age category and specific context information as primes in a lexical decision task (LDT) with age stereotypic attributes as targets (e.g., “An old person is passing the crosswalk.” – “slow”). To assess explicit age stereotypes, stereotypic traits were rated for their fit with person descriptions containing the same category and context information as the implicit measure. Results: Category effects for the priming and rating tasks emerged within relevant contexts, however, we found no correlations between these two indicators, despite the fact that the same contexts were provided for explicit and implicit assessment. Conclusion: These findings indicate that implicit and explicit age stereotypes reflect independent belief systems that are activated under different operating conditions (automatic activation vs. controlled reasoning).

Data Availability Statement

The experiment in this article earned Open Materials and Open Data badges for transparent practices. Pre-registration, materials and data for the experiment are available at https://osf.io/beq6x/.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Ethics Approval Statement

Our study was approved by the Ethical Committee of the Faculty of Social and Behavioral Science, Friedrich-Schiller-Universität Jena (number FSV 19/37).

Notes

1. The three experiments address the same overarching research question, they share the same design and structure, and they use the same materials and tasks, which is why we decided to report them together in a single method and results section. Differences between the three experiments are marginal and refer to additional exploratory research questions (e.g., in Exp. 1A, we added background photos to further highlight the category information in the prime, which were dropped in later experiments; in Exp. 1C, we used pictorial stimuli as probes in the additional memory task in order to test whether this switch in the modality from verbal to pictorial materials induces a deeper semantic processing of the category and context information in the primes; in Exp. 1A we also included neutral primes in order to separate effects of matching and mismatching category information). None of these exploratory analyses yielded any meaningful or substantial results, which is also supported by the non-significant interactions of the experiment factor with any of the effects as reported in the joint analyses in the results section, which justifies our treating these three studies as independent conceptual replications of the same overarching research question.

2. A meta-analysis by Cameron, Brown-Iannuzzi, and Payne (2012) reported small to medium-sized correlations between sequential priming tasks and explicit attitude measures (average r = .20). The studies that were included in the meta-analysis did not investigate correlations for context-specific measures of attitudes. We thus decided to search for a medium-sized effect in our study because correlations should become stronger when contexts are provided to specify the exact meaning of the (implicit and explicit) stereotypes, guaranteeing a higher degree of conceptual overlap. Sample size was calculated in a way that allowed us to detect a medium sized effect with sufficient power in each experiment.

3. Prolific is an online portal that offers the option to recruit participants from a large database for online studies. Participant payment is organized via the portal. Prolific guarantees high quality data by regularly screening and eliminating participants from the database who fail to follow instructions. In addition, the portal allows for a flexible recruitment of specific samples (e.g., stratified by age, gender, or other socio-demographic criteria).

4. In Exp. 1A, we additionally included neutral category primes (i.e., “a middle-aged person”) as a third type of category, for exploratory reasons. Since this condition was not part of the other studies, we will not report results for this condition.

5. In order to test whether the age information was still clearly visible for the blurred pictures, we conducted a validation study (N = 78, 41% female, M_age = 24.19 years, SD = 4.45) in which blurred pictures of young, middle-aged, and older adults were presented for 500 ms (i.e., with an even shorter duration than in Exp. 1A), and had to be categorized according to their age. The overall accuracy rate of age categorization for the young and old faces that were chosen for Exp. 1A was above 92%, attesting to the ease with which age information can still be extracted from the blurred faces.

6. We report the results of one-tailed tests due to the directional hypotheses that we tested, predicting 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), and which also corresponds to the findings that were reported in the studies on which the current study was built (Casper et al., 2011). As a side note, we confirm that results of two-tailed tests for the respective priming effects yield the same conclusions as the one-tailed tests.

7. The statistical tests of normality together with the Q-Q plots indicate that the distributions of category priming effects in the three experiments met the assumption of normality. The only exception was the priming effect for the unrelated context condition in Exp. 1C, for which a significant deviation from normality was indicated by the Shapiro-Wilks test (p = .044), although the Q-Q plot for this variable did not reveal strong deviations from normality. Since this priming effect yielded a clearly non-significant result, none of the significant findings of our study violates the assumptions regarding the distribution of values.

Additional information

Funding

This work was supported by the Volkswagen Stiftung under Grant Az. 93 272.