2.4.1Overview
In Study 1, it was observed that implementation intentions are a useful strategy to improve one’s memory of inconsistent information that was specified in the if-part of the implementation intention. Thereby, less pieces of inconsistent information embedded in more pieces of consistent information were presented resulting in slightly different memory performance for the different kinds of information when comparing recall and recognition. However, in real-world situations, more than one piece of inconsistent information may occur. Thus, it is interesting to investigate whether implementation intentions not only support processing of one kind of inconsistent information, but also several pieces of inconsistent information. Comparing these two memory measures would directly help identify the processes in which implementation intentions are involved.
Nevertheless, before the effects of implementation intentions on overall processing of stereotype-inconsistent information can be investigated, a design has to be established that investigates the impact of cognitive load on the processing of several pieces of stereotype-inconsistent and stereotype-consistent information. Therefore, a paradigm was created in line with studies of Srull (1981), Wyer, Bodenhausen, and Srull (1984), and Sherman and Frost (2000). Participants first received some stereotype-consistent knowledge about the target person to activate the stereotype of the elderly. Then, participants were presented equal amounts of stereotype-inconsistent, stereotype-consistent, and stereotype-irrelevant information with an impression formation instruction. While reading the behavioral information about the target person, one-half of the participants was asked to engage in a cognitive load task of the central executive. After a short-interval delay, incidental free recall and recognition tests had to be answered. Following the memory tests, a judgment of the target person was collected to investigate the impact of memory on stereotypicality ratings.
Thus, following previous research (Dijksterhuis & van Knippenberg, 1995; Garcia-Marques et al., 2002; Macrae et al., 1993; Macrae et al., 1999; Sherman & Frost, 2000; Sherman et al., 1998; Sherman, Stroessner, Conrey, & Azam, 2005; Stangor & McMillan, 1992), the subsequent hypotheses were derived:
(1) In the no load condition, stereotype-inconsistent information will be remembered better than stereotype-consistent information in both the free recall and the recognition task.
(2) Under cognitive load, overall memory performance will be reduced. Furthermore, consistent information will be remembered better than inconsistent information on the free recall as well as the recognition task.
(3) Memory performance on free recall will predict memory performance on recognition. A simplistic analysis would suggest that the information retrieved for the free recall should serve as a cue for classifying the presented information as old or new. Thus, recognition memory should depend on recall memory as well as on encoding strategies.
(4) Participants under cognitive load will judge the target person as more stereotypic than participants not under cognitive load. If inconsistent information is neglected, impression ratings of the target person should be more stereotypical under cognitive load than if cognitive capacity is left to thoroughly process the inconsistent information.
2.4.2Method Participants
Forty-four students (22 female), from 19 to 32 (M = 23, SD = 2.93) years old from the Universität Konstanz participated in the experiment. Participants were randomly assigned to one of two experimental conditions: no load control and central executive load. Sessions lasted approximately 45 minutes and participants received 3 Euros for their participation.
Design
The study followed a 2 between (Cognitive Load: no load vs. cognitive load) ¯ 2 within (Behavior Stereotypicality: inconsistent vs. consistent) mixed-model design. Dependent variables were a free recall and a recognition test that tested memory for stereotype-consistent and stereotype-inconsistent information, which was varied within participants. Finally, the stereotypicality of the target was assessed.
Procedure
Participants arrived at the laboratory individually and were greeted by a female experimenter. They were seated in front of a personal computer9 and told that the experiment investigates impression formation based on behaviors. Then, participants were given some initial information about the target person to help establish an impression of the individual. As research has shown, stereotypes have to be activated prior to encoding for them to have an impact on memory and influence interpretation (Cohen, 1981; Rothbart et al., 1979; Srull &
Wyer, 1980). Specifically, the experimenter explained that different behaviors of an old man,
9 AMD K 6-2, 300 Mhz, operating system: FREE DOS
“Joseph Müller, who is 86 years old, and lives alone in an apartment in the suburbs on his own”
would appear on the computer screen one by one. Participants should quickly read through the behaviors and form an impression of Joseph Müller (see Appendix F, p. 172).
In the central executive load condition, participants were additionally asked to simultaneously generate random numbers synchronized to a metronome adjusted to a frequency of one beat every 1.5 seconds, and silently read over the behaviors of the target person (instructions were the same as in Studies 1, 2a, and 2b, see Appendix C, p. 129). The experimenter demonstrated this task and let the participant practice before starting the program.
The control group participants did not get this instruction nor practice time. All participants were further told that after reading the behaviors of Mr. Müller, other tasks would be presented on the computer screen and they should complete them one by one. After the instructions, the experimenter left the room until the experiment had concluded. Participants started the presentation of the behaviors on their own by pressing the space bar (computer presentation see Appendix F, p. 178-182).
After the presentation of the behaviors, several items taken from the Multiple Affect Adjective Checklist (Zuckerman, Lubin, Vogel, & Valerius, 1964) were administered to each participant. In order to control for mood effects after priming the stereotype of the elderly, participants were to answer eight bipolar emotion and arousal items on seven point scales (see Appendix F, p. 179). Subsequent to completing the questionnaire, participants worked on a filler task for approximately five minutes and then were asked to write down on a sheet of paper as many of Joseph Müller’s behaviors as they could remember (see Appendix F, p. 173). They had a time limit of ten minutes, indicated by a timer that counted down in minutes and seconds from ten to zero. If they finished earlier, they could stop by pressing the z-key so that the timer went to zero.
The thirty behaviors were presented again, one by one, together with thirty randomized filler behaviors; participants were asked to answer each question by pressing the key (‘;’) labeled for “yes, I saw this behavior performed by Joseph Müller” or another key (‘x’) labeled for “no, I did not see this behavior performed by Joseph Müller.” Next, participants were asked about their impression of Joseph Müller on eleven critical (stereotype-consistent) and ten filler traits (irrelevant to the stereotype of the elderly) on bipolar adjective scales. Finally, participants completed a paper-and-pencil questionnaire asking for demographic information and checking the cognitive load manipulation using a series of questions (see Appendix F, p. 174-177).
Participants were then paid and fully debriefed.
Pretested material
Behaviors. The ten pieces of stereotype-inconsistent (in opposition to the stereotype of an old man), ten pieces of stereotype-consistent (associated with the stereotype of an old man), and ten pieces of neutral information (irrelevant to the stereotype of an old man) used in the present study were selected from several extensive pretests.
First, 20 university students were asked to freely report typical behaviors performed by old men consistent) and typical behaviors performed by young men (stereotype-inconsistent). Based on this pretest, 60 counter-stereotypical items and 81 stereotypical items were selected. Additionally, 31 neutral behaviors were collected by a student researcher and the author herself. After integrating similar answers, the 20 most frequently named inconsistent behaviors, the 20 most frequent named consistent behaviors, and 21 neutral behaviors were then rated by an additional 20 university students on analogue scales for typicality ranging from not stereotypical at all to extremely stereotypical and valence ranging from negative to positive.
The ten behaviors that were described as most typical for old men and the ten behaviors that were indicated as least typical for old men (valence was also controlled for) were selected for the experiment. Examples of the stereotype-inconsistent behaviors were: “He spends a lot of time playing his PlayStation.” (Er verbringt viel Zeit mit seiner Play Station.) and “He always has a new girlfriend.” (Er wechselt sehr schnell seine Partnerinnen.). Examples of the stereotype-consistent behaviors were: “He takes his pills every morning and every evening.” (Er nimmt seine Medikamente jeden Morgen und jeden Abend.) and “He enjoys talking about the past.” (Er erzählt sehr gerne aus der Vergangenheit.). Examples of the stereotype-irrelevant behaviors were: “He has two rolls and an egg for breakfast.” (Er isst zum Frühstück zwei Brötchen und ein Ei.) and “He is happy when he sees a rainbow.” (Er freut sich, wenn er einen Regenbogen sieht. see Appendix F, p. 170).
The length of the sentences was also controlled for. Presentation times of four seconds per behavioral description were also pre-tested by recording reading latencies of 28 university students both with and without cognitive load. Mean reading latencies for all participants with or without cognitive load and across information types (inconsistent, consistent, and neutral) resulted in less than three seconds per description. To ensure that everybody had enough time to read the sentence, it was decided to present each behavior for four seconds. The first and last sentences were always neutral behaviors (not experimental ones), to allow for the controlling of primacy and recency effects.
Stereotypicality rating. Participants were asked to mark the position that best fit their opinion about Joseph Müller on a scale with seven possible responses. Stereotypic traits were conducted on the basis of former pretests conducted by students at the Universität Konstanz that investigated typical characteristics of the elderly. Critical items and their respective opposites
were: “sick-healthy” – krank-gesund, “old-fashioned-modern – altmodisch-modisch, “crabby-friendly” – mürrisch-freundlich, “stubborn-compliant” – stur-nachgiebig, “old-young” – alt-jung,
“static-dynamic” – statisch-dynamisch, “slow-fast” – langsam-schnell, “not spontaneous-spontaneous” – unspontan-spontan, “unsociable-sociable” kontaktarm-kontaktfreudig, “tradidional-modern” – konservativ-modern, “dependent-independent” – unabhängig-abhängig. Ten neutral attributes served as distractors (e.g. “likable-dislikable” – sympathisch-unsympathisch, “good-bad” – gut-schlecht, “boring-funny” – langweilig-lustig, “honest-intriguing” – ehrlich-intrigant). The attributes were presented in randomized order (see Appendix F, p. 171-172).
Manipulation checks. In the final questionnaire, participants assessed the difficulty of the impression formation task and the memory tests. Participants marked their answers on a 70 millimeter (2.76 inch) analogue scale labeled not at all on the left-hand side and very on the right-hand side. Questions regarding the manipulation of cognitive load were: “How difficult was it for you to form an impression of Mr. Müller while you were reading the description?” (Wie schwer fiel es Ihnen, sich beim Lesen der Beschreibung einen Eindruck von Herrn Müller zu machen?), “How much effort did you have to exert to form an impression of Mr. Müller?” (Wie sehr mussten Sie sich anstrengen, um sich einen Eindruck von Herrn Müller zu bilden?). Further, the following questions were asked: “How difficult was it for you to recall the behaviors of Mr. Müller?” (Wie schwer fiel es Ihnen, die Verhaltensweisen von Herrn Joseph Müller aus dem Gedächtnis wieder zu geben?), “How difficult was it for you to categorize the behaviors of Mr. Müller as old or new?” (Wie schwer fiel es Ihnen, die Verhaltensweisen von Herrn Joseph Müller als alt oder neu zu klassifizieren? see Appendix F, p. 174-177).
2.4.3Results Outliers
Four participants scored more than two standard deviations from the mean on important dependent variables; therefore, they were excluded from the analyses.
Manipulation checks
Given that the correlational relationship between the two impression formation questions (“How difficult was it for you to form an impression of Mr. Müller while you were reading the description?”, “How much effort did you have to exert to form an impression of Mr. Müller?”) was very high (r = .80, p < .001, Cronbach’s alpha = .89), an index of impression formation difficulties was computed.
A one-factorial (Cognitive Load: no load vs. cognitive load) ANOVA with this index as the dependent variable did not reveal a significant main effect of Cognitive Load, F(1, 38) =
2.68, p = .11, η² = .07). Nevertheless, means indicated that although participants in the cognitive load condition (M = 4.25, SD = 2.08) did not report more difficulties on the impression formation task compared to participants in the no load condition (M = 3.32, SD = 1.50), the expected tendency that cognitive load participants had more problems on the impression formation than control participants was found.
Participants were asked how difficult it was for them to recall or to categorize the behaviors of Mr. Müller. While the correlation between these two questions was not significant (r = .19, Cronbach’s alpha = .32), they were analyzed separately. Participants answers were analyzed by two one-factorial (Cognitive Load: no load vs. cognitive load) ANOVAs that revealed a significant main effect of Cognitive Load on the question “How difficult was it for you to recall the behaviors of Mr. Müller?”, F(1, 38) = 24.72, p < .001, η² = .39; no load: M = 3.02, SD = 1.33; cognitive load: M = 5.15, SD = 1.37), but not on the question “How difficult was it for you to categorize the behaviors of Mr. Müller as old or new?” (F < 1; no load: M = 3.23, SD = 1.83; cognitive load: M = 3.59, SD = 2.10). These results show that it was more difficult for participants in the cognitive load condition than for participants in the no load condition to recall the behaviors, but not when they recognized the information.
Furthermore, a t-test for the dependent variables showed that the free recall task (M = 4.03, SD = 1.71) was more difficult to answer than the recognition task (M = 3.40, SD = 1.95), t(39) = 1.70, p < .05.
The Multiple Affect Adjective Checklist (Zuckerman et al., 1964) did not show any effects on the dependent variables, indicating that priming the stereotype of the elderly did not generate any mood effects. Thus, it was not included in the analyses.
Free recall
The recall of behaviors was scored according to a standardized criterion. Written answers were scored as correct if they conveyed the same idea as the presented behaviors, regardless of the wording. However, statements that simply referred to the trait implied by a behavior, without conveying the action itself, were not scored as correct (cf. Wyer et al., 1984).
For the analysis, mean proportions of behaviors recalled by each subject were computed separately for stereotype-inconsistent and for stereotype-consistent behaviors.
Correctly recalled inconsistent and consistent items were analyzed in a 2 between (Cognitive Load: no load vs. cognitive load) ¯ 2 within (Behavior Stereotypicality: inconsistent vs. consistent) ANOVA. This analysis revealed a significant main effect of Cognitive Load, F(1, 37) = 21.66, p < .001, η² = .37, indicating that recall performance was higher for no load participants (M = .53, SD = .18) than cognitive load participants (M = .29, SD = .12).
Furthermore, a significant main effect of Behavior Stereotypicality was observed, F(1, 37) = 14.54, p = .001, η² = .28. In general, participants remembered the inconsistent information (M = .49, SD = .22) better than the consistent information (M = .35, SD = .23). This pattern of results was evident in both the no load (inconsistent: M = .61, SD = .22; consistent: M = .45, SD = .20; t(20) = 3.69, p < .01) and load conditions (inconsistent: M = .35, SD = .11; consistent:
M = .23, SD = .22, t(17) = 1.94, p < .05 (one-sided)). Nevertheless, the predicted Cognitive Load ¯ Behavior Stereotypicality interaction did not reach significance, F < 1 (see Figure 5).
0
Figure 5. Free recall performance of stereotype-inconsistent and stereotype-consistent information as a function of Cognitive Load (* p < .05, ** p < .01).
Recognition
On the recognition test, some participants showed almost perfect memory discrimination, therefore, A’, a nonparametric measure of recognition sensitivity (Grier, 1971), was computed. A’ reflects the degree to which participants were able to correctly discriminate presented behaviors from new behaviors, while controlling for guessing strategies and response biases. To compute each participant’s separate measures of recognition accuracy for stereotype-consistent and stereotype-instereotype-consistent behaviors, the proportions of hits (correct identification of presented behaviors) and false alarms (failure to reject new behaviors) were used (see Grier, 1971, for the exact formula).
*
**
The A’ recognition accuracy measures for inconsistent and consistent behaviors were analyzed in a 2 between (Cognitive Load: no load vs. cognitive load) ¯ 2 within (Behavior Stereotypicality: inconsistent vs. consistent) ANOVA. Thereby, a main effect of Cognitive Load was observed, which demonstrated that recognition accuracy was higher in the no load (M = .96, SD = .03) than in the cognitive load condition (M = .80, SD = .13), F(1, 38) = 50.58, p < .001, η² = .57.
Neither the main effect of Behavior Stereotypicality, F(1, 38) = 2.09, p = .16, η² = .05, nor the Cognitive Load ¯ Behavior Stereotypicality interaction, F < 1 reached significance (see Figure 6).
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
No Load Cognitive Load
Cognitive Load Condition
A' Proportion of Recognition Performance
Inconsistent Consistent
Figure 6. A´ recognition performance of stereotype-inconsistent and stereotype-consistent information as a function of Cognitive Load.
Correlational analysis
Moreover, to investigate if free recall performance predicts recognition performance, a correlational analysis was conducted for stereotype-inconsistent and stereotype-consistent information separately. The analysis for inconsistent information showed that as recall performance (M = .49, SD = .22) increased, recognition performance (M = .92, SD = .07) also increased, r = .47, p < .01. Similarly, recall performance for consistent information (M = .35, SD
= .23) predicted recognition performance for consistent information (M = .88, SD = .12), r = .57, p < .001.
Stereotypicality of the target person
In order to investigate differences in the judgment of the target person, critical items were partially reversed, such that high scores are indicative of high stereotypicality. A mean index was then computed (Cronbach’s alpha = .76). A one-factorial ANOVA with Cognitive Load (between: no load vs. central executive load) as the independent variable revealed no difference between control participants (M = 2.81, SD = 0.73) and central executive participants (M = 2.68, SD = 0.86), F < 1. Further, a linear regression analysis was computed to investigate whether memory performance predicts judgment of the target person without considering cognitive load, given that no effect of load was found.
Therefore, an index of stereotypical memory was separately computed for free recall and recognition performances to evaluate the prediction of this difference on judgment using the following formula (see Macrae et al., 1993): (inconsistent behaviors – consistent behaviors)/(inconsistent behaviors + consistent behaviors). A significant relationship was found for free recall performance (M = .22, SD = .42) on judgment ratings (M = 2.70, SD = 0.74), r = -.34, p < .05, indicating a negative linear relationship between stereotypical memory and judgment. There was also a marginally significant negative linear trend for recognition performance (M = .02, SD = .07) on judgment ratings (M = 2.75, SD = 0.79), r = -.28, p < .05, indicating that the increase in judgment ratings was explained by a decrease in the recognition performance of inconsistent information.
2.4.4Discussion
The present study was designed to investigate cognitive load effects on memory for equally distributed pieces of stereotype-inconsistent and stereotype-consistent information, comparing recognition and free recall tests. Therefore, participants were presented ten pieces of both inconsistent and consistent information regarding common behaviors of an old man. One-half of the participants was interrupted by the central executive load, while the other One-half could read the behaviors undisturbed. Unforeseen free recall and recognition tests were conducted, as well as a judgment rating of the assessed stereotypicality of the target person.
As predicted, memory performance decreased on both memory tests when cognitive load was induced indicating a successful manipulation of cognitive load. Furthermore, the pattern of results revealed a marginally significant trend consistent with the prediction that participants in the load condition reported more problems with impression formation than participants not under cognitive load. Thus, results of the present study replicated the finding that cognitive load of the central executive interrupts the encoding and storing of behavioral
information (Macrae et al., 1999; Sherman & Frost, 2000; see also Studies 1, 2a, and 2b of this dissertation).
Memory performance on the free recall task was expected to influence memory performance of recognition. Information that was retrieved from long-term memory in the free recall task should produce higher familiarity and thus should help one categorize the presented information as old or new (see Gillund & Shiffrin, 1984). Correlational analyses supported this prediction. The more information participants remembered on the free recall task, the more they remembered on the recognition task.
Additionally, it was found that inconsistent information was remembered better than consistent information on the free recall task and was also marginally significant on the recognition task. Thus, the inconsistent information was encoded and stored in long-term memory better than the consistent information. This result supports findings from other studies investigating the inconsistency resolution effect that inconsistent information is processed more extensively than consistent information (e.g., Bodenhausen et al., 1999; Brewer, 1988; Srull, 1981; see also Studies 1, 2a, and 2b in this dissertation). However, in the present study, this effect was independent of cognitive load. Central executive load should have impaired memory for inconsistent information more than memory of consistent information, resulting in equal or better memory performance for consistent versus inconsistent information (Haller, 2004).
The results are contradictory to the findings of Macrae and colleagues (1999) as well as Sherman and Frost (2000). These authors observed that central executive load impaired the
The results are contradictory to the findings of Macrae and colleagues (1999) as well as Sherman and Frost (2000). These authors observed that central executive load impaired the