Experimental Evidence That Demand Characteristics Do Not Play a Dominant Role in the Evaluative Conditioning Effect

Abstract

The evaluative conditioning (EC) effect has been documented in many experiments: Participants typically prefer stimuli that co-occurred with positive stimuli over stimuli that co-occurred with negative stimuli. The present research attempted to test whether demand characteristics are a dominant cause of the EC effect. In three experiments, we informed participants of the research hypothesis, sometimes indicating an expectation of a contrast effect, rather than an assimilative effect. That manipulation hardly moderated the EC effect. The manipulation influenced participants’ beliefs regarding the research hypothesis, although participants generally believed that an assimilative effect is a more plausible research hypothesis than a contrast effect. Even participants who believed that the researchers expected a contrast effect or assumed that stimulus co-occurrence typically causes a contrast effect still showed an assimilative effect. The results suggest that although demand characteristics might influence the EC effect, the overall influence of that factor is minor.

Keywords

evaluative conditioning demand characteristics evaluation attitudes evaluative learning

Objects in people’s surroundings, such as possessions, friends, and food, often elicit feelings and thoughts based on past experiences with them. One type of experience that might shape the (un)favorability of these feelings and thoughts is stimulus co-occurrence. People typically prefer objects (the conditioned stimulus; CS) that repeatedly co-occurred with favorable objects (unconditioned positive stimulus; US_pos) over those that co-occurred with unfavorable ones (US_neg); a phenomenon dubbed the evaluative conditioning (EC) effect (De Houwer et al., 2001; Moran et al., 2023).

The EC effect has been discovered and investigated in psychological experiments that utilized the EC procedure: exposing participants to pairs of stimuli. From the early days of EC research, there has been a concern that the EC procedure includes demand characteristics that cause the EC effect (Insko & Oakes, 1966; Page, 1969; Staats, 1969). Demand characteristics are cues that convey that research hypothesis to the participant (Orne, 1962). Demand characteristics can influence participants’ behavior in various ways (Corneille & Lush, 2023) due to intentional or unintentional processes. Participants can form correct or mistaken beliefs about the research hypothesis. Those beliefs may influence inferential processes, change the accessibility of specific concepts or memories, or motivate participants to comply with what they perceive as the researchers’ expectations, resist them, or even sabotage the experiment. The motivation to please researchers—perhaps the main driver of demand characteristics effects—can lead participants to feign that a manipulation influenced them as they believe the researchers intended. Alternatively, the motivation to please researchers may genuinely cause the change that the participants believe the researchers expect to find due to processes such as conforming with perceived social norms (Cialdini & Goldstein, 2004). The critical point is that demand characteristics may produce a causal relation between the manipulated and measured variables that would not have occurred if participants had not attributed the research hypothesis to the researchers.

In the present research, we tested a demand account for the EC effect (e.g., Insko & Oakes, 1966; Page, 1969). Figure 1 illustrates the account: participants notice the CS-US co-occurrence and guess that the researchers present those co-occurrences to bring participants’ evaluation of the CS closer to their evaluation of the US. In other words, participants guess that the researchers expect them to show an assimilative effect of the US valence on the CS evaluation. As a result, participants may falsely report evaluations that reflect an assimilative effect. Alternatively, participants may change their actual evaluations in accordance with that hypothesis, due to a motivation to please the researchers or due to other processes elicited by pondering the research hypothesis. If the EC procedure in EC experiments causes the EC effect due to its demand characteristics, those experiments fail to investigate the actual effect of stimulus co-occurrence in real life, where demand characteristics are rare. That would severely limit the usefulness of the knowledge that EC research has accumulated over the years.

Figure 1

The Demand Account for the EC Effect

Previous Research

Demand characteristics were initially ruled out as a dominant cause for the EC effect by evidence that the effect occurs even when participants are unable to report the CS-US co-occurrence (e.g., Baeyens et al., 1990; McGinley & Layton, 1973). Without co-occurrence awareness, participants cannot attribute any expectation regarding the effect of the co-occurrence. However, in recent decades, most studies found no EC effect without awareness of the CS-US co-occurrence, suggesting that co-occurrence awareness is a necessary condition for EC (Corneille & Mertens, 2020; Corneille & Stahl, 2019). If people know about the stimulus co-occurrence and can infer the researchers’ expectations from that co-occurrence, then demand characteristics might cause the EC effect.

Some studies attempted to test the role of demand characteristics with retrospective self-reported questionnaires that probed participants’ awareness of the research hypothesis (e.g., Insko & Oakes, 1966; McGinley & Layton, 1973; Page, 1969; Sweldens et al., 2010). These studies found conflicting results, perhaps due to the low validity of retrospective awareness measures. Such measures might cause awareness with their questions or might miss awareness due to forgetting or miscommunication (Gawronski & Walther, 2012).

A recent study (Ingendahl et al., 2023) examined the correlation of the EC effect with a behavioral measure of participants’ susceptibility to demand characteristics effects. In the behavioral measure, participants were informed that people tend to prefer photos that appear on one side of the screen (e.g., right) over the other side. The frequency of evaluating stimuli according to that expectation was considered a measure of demand compliance (i.e., compliance with the research hypothesis the participant attributes to the researcher). The authors found a modest but reliable correlation (r = .13) between performance on that behavioral measure and the size of the EC effect. That result is compatible with the possibility that demand characteristics play a small role in the EC effect. However, that role might appear smaller than it actually is due to a low validity of the demand compliance measure, failing to exhaustively capture individual differences in the tendency to comply with the research hypothesis that the participant attributes to the researchers. Another possibility is that the correlation underestimated the role of demand characteristics in EC because the influence of guessing the research hypothesis may vary by situational (rather than dispositional) factors. For example, showing an EC due to demand characteristics depends on noticing the co-occurrence and on correctly guessing the research hypothesis—two factors that are probably unrelated to individual differences in demand compliance. Alternatively, demand characteristics may play no causal role in EC because the correlation found in the previous study might reflect a noncausal relation between dispositional demand compliance and the EC effect.

One limitation of current knowledge about the likelihood of the demand account for EC is that it is missing the most useful evidence for causal inference: data from an experimental design. If knowledge about the research hypothesis is a dominant cause of the EC effect, then changing that knowledge would alter the effect. Based on that logic, in the present research, we manipulated participants’ beliefs about the research hypothesis. Although, typically, stimulus co-occurrence results in an assimilative effect, previous studies revealed that, under some circumstances, co-occurrence may result in a contrast effect (e.g., a preference of CS_neg over CS_pos; Alves & Imhoff, 2023). In line with this effect, we informed one group of participants in each of our experiments that we were investigating whether co-occurrence would result in a contrast effect. In Experiment 1, there were three other groups: a group that read that the research hypothesis was an assimilative effect, a group that read about both the assimilative and the contrast hypotheses, but without favoring one hypothesis over the other, and a group that, just like in a typical EC procedure, was not provided with any information about the co-occurrence and its possible results. Because Experiment 1 revealed that people typically attribute to the researchers the hypothesis of an assimilative effect, in Experiments 2 and 3, we included only two groups: those who read that the research hypothesis is a contrast effect and those who did not receive any information. Experiment 2 was a replication of Experiment 1 with a different population (paid participants rather than volunteers). In Experiment 3, we manipulated the research hypothesis with a different, more straightforward, text, than in Experiments 1 and 2.

In all the experiments, we examined whether the effect participants would reveal in their evaluation would be in line with the research hypothesis presented to them. If participants who believe that the research hypothesis is a contrast effect would show a contrast effect, that would be compatible with the possibility that demand characteristics are a dominant cause of the EC effect. On the contrary, if we find an assimilative effect of CS-US co-occurrence on CS evaluation even when the researchers are believed to expect a contrast effect, that will support the notion that demand characteristics are not the leading cause for the EC effect.

Method

Transparency and Openness

We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the study. All the materials, data, and pre-registrations are at https://osf.io/jhsra/. There were no deviations from the preregistered analyses, additional analyses are reported in the Supplement Online Materials (SOM).

Participants

The participants in Experiments 1 and 3 were volunteers on Project Implicit’s website (Nosek, 2005). In Experiment 2, participants came from the online commercial pool Prolific (Palan & Schitter, 2018; Peer et al., 2017), and paid €2 for a 10-min study. Table 1 shows all the information about the participants in each study and statistical power. In all experiments, the only preregistered exclusion rule was removing participants who did not rate all the CSs.

Table 1

Demographic Information and Statistical Power

Experiment	Started	Completed	Female	M Age (SD)	Achieved power (η²_p)
Exp. 1	1,521	1,234	61%	34.2 (14.6)	<0.01, 0.01
Exp. 2	697	648	46%	41.6 (13.3)	<0.01, 0.02
Exp. 3	966	787	68%	34.5 (15.5)	<0.01, 0.01
Residence	United States	United Kingdom	Canada	India	Australia	Other
Exp. 1	73.8%	3.9%	4.3%	1.5%	0.9%	15.6%
Exp. 2	5.1%	88.6%	0%	0.2%	0.2%	5.9%
Exp. 3	67.3%	5.2%	3.8%	1.8%	3.7%	18.2%
Race	White	Asian	Black	Hispanic	Multi-race	Other
Exp. 1	55.5%	9.3%	11.9%	13.0%	5.5%	4.8%
Exp. 2	82.7%	4.9%	6.8%	0%	4.0%	1.6%
Exp. 3	60.4%	11.2%	9.0%	9.5%	5.3%	4.6%

Note. Under achieved power, we indicate the size (η²_p) of the effects that could be detected with the final sample size with 80% and 95% probability, respectively. Power was computed using G*Power 3.1 (Faul et al., 2009) for the interaction in an ANOVA with US valence and the demand characteristics condition as factors predicting the CS explicit evaluations.

In Experiment 1, we determined the target sample size based on the power needed to detect specific effect sizes in a t-test that compared the EC effect across two of the four between-participants conditions. Specifically, we aimed for at least 85% power to detect a small effect size (d = 0.3) and 95% power to detect a medium effect size (d = 0.4). In Experiment 2 we aimed for a 95% of detecting the small moderation effect (η² _p = .021) of demand characteristics on the effect of the US valence and 95% to 97% power for detecting a US valence effect that had a range of d = 0.21 to 0.5 in various subsamples of participants in the contrast condition. In Experiment 3 we aimed for a 95% power for detecting a d = 0.3 effect of the manipulation and 95% power to detect a US valence effect of d = 0.17 in the contrast condition.

Procedure

Experiment 1 included the following components, ordered as follows: A short demonstration of the EC procedure, the demand characteristics manipulation, the full EC procedure, CS evaluation questionnaire, the self-reported perceived research hypothesis, and the evaluative priming task (EPT; Fazio et al., 1995). The demand characteristics manipulation assigned participants to one of four conditions: presenting the research hypothesis as an assimilative effect (the assimilative condition), a contrast effect (the contrast condition), presenting both hypotheses (the both condition), and introducing no research hypothesis (the control condition).

The EPT was administered because Project Implicit’s platform required including “implicit measures” in each study. We did not plan to make any inferences based on the EPT data. Details and results pertaining to the EPT are reported only in the SOM.

Experiment 2 was identical to Experiment 1, but conducted on Prolific, rather than Project Implicit and thus did not include an EPT at the end. Our purpose was to examine a different participant population—paid participants who might be more motivated to please the researchers. In Experiments 2 and 3, we included only a contrast and a control condition because results of Experiment 1 indicated that even without information about the research hypothesis, more than 80% of the participants reported assuming that the research hypothesis was an assimilative effect.

Experiment 3 was identical to Experiment 2 with the following modifications. To rule out the possibility that the short EC procedure before the demand characteristics manipulation in Experiments 1 and 2 already induced an assimilative effect and might have influenced participants’ beliefs regarding the research hypothesis, we did not provide participants an experience with the EC procedure before presenting the research hypothesis. Furthermore, we used a more straightforward manipulation of demand characteristics by stating only the exact prediction, rather than providing a long explanation for a possible mechanism of the contrast effect. In Experiment 3, the likelihood of random assignment to the contrast condition was 66% because we pursued reasonable power for detecting the EC effect in the contrast condition.

Design

Experiment 1 had the design 2 (US valence: positive, negative; within participants) × 4 (induced expectations: assimilative, contrast, both, control; between participants), whereas Experiments 2 and 3’s second factor had only the conditions contrast and control.

Materials

The CS materials and EC procedure were copied from previous research (Gawronski et al., 2015). The five CSs were randomly selected for each participant from a pool of ten abstract shapes. The USs were 32 free-to-use affective photos, 16 of each valence, from various sources (OASIS—Kurdi et al., 2017; BRIC—Segal-Doron & Bar-Anan, n.d.).

EC Procedure

Each of the 40 trials started with fixation (200 ms), followed by a CS-US pair or one CS_neu alone (1,000 ms), followed by a blank screen (300 ms). Each CS appeared in eight trials, four times on the left side of the screen and four times on the right side. The CS_neu appeared alone, whereas the two CS_pos and two CS_neg appeared always with the US. Each US appeared one time. In Experiments 1 and 2, participants were familiarized with the pairing task by observing a 10-trial version of this task (2 trials for each CS) before the demand characteristics manipulation, which was followed by the complete 40-trial EC procedure.

Demand Characteristics Manipulation

Experiments 1 and 2

Participants in the control condition were requested to continue to the next task in the experiment, which was the complete EC procedure. The other participants read what the research hypothesis was. We first informed participants that we were about to “explain more about why we are conducting the study” and that afterward we will ask a couple of questions to ensure they read and understood our explanation. After the example EC procedure, we provided the information on two consecutive pages. In the assimilative condition, participants read about the phenomenon “assimilation in liking”—“when things appear together, people start noticing the similarities between them and they start judging them similarly.” Therefore, “we are testing the prediction that people would like abstract photos that appeared together with positive photos more than abstract photos that appeared together with negative photos.” In the contrast condition, participants read about the phenomenon “contrast in liking”—“When things appear together, people start noticing the differences between them and they start judging them as opposites.” Therefore, “we are testing the prediction that people would like abstract photos that appeared together with negative photos more than abstract photos that appeared together with positive photos.” In the both condition, participants read both those texts and read that we are testing which of the phenomena occurs in the study. This additional control condition exposed participants to the possibility that CS-US co-occurrences would influence evaluation, without favoring one effect over another. In each condition, we included an example to increase comprehension. Before reading the comprehension questions, participants were allowed to read the information again, to prepare for those questions. If participants did not answer one or two of the comprehension questions correctly, we presented to them the information again and repeated the questions.

The first comprehension question was “According to our explanation, what will happen if we show people a relatively neutral-looking dog next to a very cute dog?” with the response options “(a) The relatively neutral-looking dog will seem cute.”, “(b) The relatively neutral-looking dog will seem ugly.”, “(c) It is difficult to know because there are findings that support (a) and findings that support (b).”, and “(d) It is difficult to know because I did not receive any information relevant to this question.” The second comprehension question was “Which of the following claims were included in our explanation?” with the response options “(a) When things appear together, people start seeing the similarities between them. As a result, people judge things that appear together similarly.”, “(b) When things appear together, people start seeing the differences between them. As a result, people judge things that appear together as opposites”, “(c) Both claims (a) and (b) were presented in the explanation.”, and “(d) None of the claims were presented.” In both questions, the first response options, (a) and (b), were randomly ordered for each participant.

Experiment 3

In the control condition, the EC procedure started after participants read the consent form. In the contrast condition, we briefly described the EC procedure, and explicitly stated the research hypothesis: “In this study, we are testing whether when things appear together, people start seeing the differences between them and judging them as opposites.” and “We suspect that if we present to you a specific abstract photo alongside negative photos, you would start liking that abstract photo because it is more pleasant than the negative photos.” We used a similar text to explain our contrast hypothesis regarding the CS_pos. If participants were wrong in the comprehension question (see below), we reminded them of our supposed research hypothesis.

We used only one comprehension question: “According to our explanation, what do we predict would happen in this experiment?” with the response options: “People would start liking the abstract photos that appear with pleasant photos, and would start disliking the abstract photos that appear with unpleasant photos.”, “People would start disliking the abstract photos that appear with pleasant photos, and would start liking the abstract photos that appear with unpleasant photos.”, “People would start liking the abstract photos that appear with pleasant photos, and the abstract photos that appear with unpleasant photos. ” and “People would start disliking the abstract photos that appear with pleasant photos, and the abstract photos that appear with unpleasant photos.”

Self-Reported Evaluations

Regarding each of the five CSs, we asked “How positive or negative are your feelings toward this image?” with the response options “Extremely negative”, “Moderately negative”, “Slightly negative”, “Neutral”, “Slightly positive”, “Moderately positive”, and “Extremely positive”.

Self-Reported Perceptions

We presented six statements with the question “Is the following statement true or false?” and the response options “False (certainly)”, “False (probably)”, “False (guess)”, “True (guess)”, “True (probably)”, and “True (certainly)”. The exact wording of the statements appears in Supplemental Figures S1 and S2 (in the SOM). Participants answered three randomly ordered statements to assess their perception of the researchers’ expectations (assimilative effect, contrast effect, or no expectations). In the final question, participants reported what the typical effect in this kind of study is (assimilative, contrast, or no effect). We used that measure to examine whether exposure to the research hypothesis influenced participants’ beliefs about the actual typical effect of the EC procedure. In Experiments 1-2, this questionnaire started with two randomly ordered statements regarding the CS-US contingency.

Results

The EC Effect

We report most of the results in sets of three numbers, for Experiments 1-3, respectively. Overall, participants rated the CS_pos (Ms = 4.95, 4.87, 4.76; SDs = 1.14, 1.11, 1.13) more favorably than the CS_neg (Ms = 3.21, 3.38, 3.47; SDs = 1.16, 1.18, 1.20), a clear assimilative effect of the EC procedure (ds = 0.89, 0.75, 0.68). In the 2 (paired valence) x 4 (demand characteristics induction condition; only 2 levels in Experiments 2-3) ANOVA, the main effect of valence was strong, Fs(1, 1227; 1, 646; 1, 785) = 991.16, 370.02, 353.25, ps < .001, η² _p = .447, .364, .310, and there was no main effect for the demand characteristics condition, Fs(3, 1227; 1, 646; 1, 785) < 1, η² _p = .002, < .001, < .001. The interaction, the focal effect in question, was always small, passing the statistical significance threshold only in Experiment 1, Fs(3, 1227; 1, 646; 1, 785) = 5.02, 2.29, 1.63, ps = .002, .130, .195; η² _p = .012, .004, .002.

In Experiment 1, a follow-up ANOVA with a subsample of participants from the contrast and control conditions revealed a main effect of valence F(1, 629) = 479.67, p < .001, η²_p = .433, qualified by a valence × demand characteristics condition interaction, F(1, 629) = 13.69, p < .001, η²_p = .021. The interaction reflected a stronger assimilative effect in the control condition than in the contrast condition (see Table 2). The second follow-up ANOVA, comparing the effect of valence in the assimilative, control, and both conditions, revealed only the main effect of valence, F(2, 915) = 816.83, p < .001, η²_p = .472, with no significant valence × demand characteristics condition interaction, F(2, 915) < 1, η²_p = .002. Thus, we obtained little evidence for a weak causal role for participants’ beliefs about the research hypothesis in the EC effect.

Table 2

The Preference for CS_pos Over CS_neg (the EC Effect) in Each Subsample in Experiments 1–3

	Experiment 1		Experiment 2		Experiment 3
Subsample	n	Mean EC (SD)	n	Mean EC (SD)	n	Mean EC (SD)
Overall	1231	1.74*** (1.95)	648	1.49*** (1.97)	787	1.29*** (1.90)
By condition
Assimilative	290	1.76*** (1.95)
Contrast	313	1.40*** (1.86)	328	1.37*** (1.93)	490	1.23*** (1.93)
Both	310	1.85*** (1.96)
Control	318	1.96*** (1.99)	320	1.60*** (2.01)	297	1.40*** (1.85)
Accurate hypothesis
Assimilative	167	2.02*** (1.93)
Contrast	62	0.86*** (1.29)	111	0.96*** (1.73)	185	1.00*** (1.85)
Accurate comprehension
Assimilative	130	2.01*** (1.96)
Contrast	184	1.34*** (1.80)	253	1.37*** (1.97)	242	1.18*** (1.79)
Both	88	1.98*** (1.82)
Assumed Contrast	70	0.93*** (1.43)	111	0.96*** (1.73)	185	1.00*** (1.85)

Notes. The accurate hypothesis subsample included only participants who rated the research hypothesis that was presented to them as the most likely hypothesis the researchers had; The accurate comprehension subsample included only participants who responded correctly to all the questions in the comprehension test that was a part of the manipulation (i.e., before the EC procedure); The assumed contrast subsample included only participants who chose the contrast effect as the typical EC effect.

***

p < .001.

As shown in Table 2, the paired valence had an assimilative effect in all the demand characteristics conditions in all the experiments. Non-preregistered Linear Mixed Model analyses, reported in the SOM, found the same results.

Participants’ Beliefs

Perceived Research Hypothesis

As the main manipulation check, we examined the effect of the demand characteristics condition on participants’ self-reported perceptions of the researchers’ hypothesis. As shown in Figure 2, only in the contrast condition in Experiment 3, most participants believed that it was more likely that the research hypothesis was a contrast effect than an assimilative effect. In all the studies, the overall pattern of differences between the conditions was compatible with the manipulation—the contrast condition had the smallest percentage of people (53%–67%, across studies) who attributed an assimilative hypothesis to the researchers and the largest percentage of people who attributed a contrasting hypothesis (54%–62%; note that participants were able to report attributing both hypotheses because these were separate questions).

Figure 2

The Effect of the Induced Demand Characteristics on Participants’ Beliefs About the Research Hypothesis

For the statistical test, we computed a score that reflected the participant’s belief that it was more likely that the research hypothesis was assimilative than a contrast effect: we subtracted each participant’s level of agreement with the statement that reflected the perception that the researchers expected a contrast effect from the level of agreement with the statement that reflected the perception that the researchers expected an assimilative effect. Figure 2 shows that the differences between the conditions in that score were compatible with the manipulation in all the experiments. In a one-way ANOVA on that score, with manipulation condition as the only factor, we found a large effect in all experiments, Fs(3, 1195; 1, 644; 1, 750) = 34.78, 79.76, 102.47, all ps = < .001, η² _p = .080, .110, .120. Table 3 shows the paired comparisons between each pair of conditions in Experiment 1, affirming that the manipulation was successful.

Table 3

Perceived Research Hypothesis: Paired Comparisons in Experiment 1

Group 1	Group 2	t	p	d
Assimilative	Both	3.05	.012	0.25
	Contrast	9.43	< .001	0.78
	Control	1.63	.364	0.13
Both	Contrast	6.44	< .001	0.52
	Control	–1.47	.455	–0.12
Contrast	Control	–7.97	< .001	–0.64

Note. df = 1,195.

The Perceived Typical EC Effect

The exposure to the research hypothesis influenced people’s beliefs regarding the typical EC effect. As shown in Figure 3, in all the conditions, most participants (a range of 51%–74%, across studies and conditions) assumed that the typical effect of the EC procedure is assimilative, and very few (3%–30%, across studies and conditions) assumed that it is a contrast effect. Yet, in each study, in comparison to all other conditions, participants in the contrast condition were the least likely to estimate that the effect is assimilative (51%–63%, across studies) and the most likely to estimate that it was contrast (20%–30%, across studies). We recoded people’s response to that question to −1 (a contrast effect), 0 (no effect) and 1 (an assimilative effect) and submitted that variable to an ANOVA with the manipulation as the only factor. The manipulation had a small-to-medium effect on people’s perception of the typical effect, Fs(3, 1199; 1, 644; 1, 748) = 11.45, 18.33, 38.38, ps < .001, η² _p = .028, .028, .049.

Figure 3

The Proportions of Participants’ Beliefs Regarding the Typical Effect of the EC Procedure in Each Condition for Each Experiment

The Relation Between the EC Effect and Participants’ Perceptions

To further explore the relation between the EC effect and participants’ beliefs about the research hypothesis, we tested the EC effect among participants who reported that the researchers’ most likely hypothesis was the one presented to them (i.e., a contrast effect in the contrast condition, and an assimilative effect in the assimilative condition in Experiment 1). The results, presented under the “Accurate hypothesis” row in Table 2, show that an assimilative EC effect always occurred even when participants in the contrast condition reported that a contrast effect was the more likely expectation that the researchers had. Table 2 also shows that the assimilative EC effect occurred in all the demand characteristics induction conditions, even among the subsample of participants who responded correctly to the comprehension test, immediately after reading the information about the research hypothesis. Finally, even participants who estimated that the typical effect of the EC procedure is a contrast effect, still showed an assimilative EC effect in all the experiments (bottom row of Table 2). These results are all incompatible with the assumption that demand characteristics are the main cause of the EC effect.

General Discussion

In three studies, we informed participants of the hypothesis that was tested regarding the effect of CS-US co-occurrence on the evaluation of CS. We examined whether manipulating participants’ belief about the hypothesis would influence the effect of the co-occurrence on their CS evaluation. Critically, some participants read that we tested whether a contrast effect would emerge (a preference for the CS_neg over the CS_pos), whereas other participants did not read anything about our hypothesis. In Experiment 1, two other groups of participants read that we tested whether an assimilative effect would emerge (a preference for the CS_pos over the CS_neg), or that we tested which of these hypotheses would be confirmed. That demand characteristics manipulation influenced the effect of the CS-US co-occurrence on CS evaluation only in Experiment 1, but even that effect was small (η ²_p = .021). Further, participants who reported that the researchers expected a contrast effect and participants who estimated that a contrast effect is the typical effect of CS-US co-occurrence, still showed an assimilative effect. On the other hand, although our manipulation successfully influenced participants’ reported perception of the research hypothesis, participants showed a bias for attributing to the researchers an assimilative effect hypothesis rather than a contrast effect hypothesis. These results may temper the certainty that our manipulation was sufficiently successful in influencing most people’s perceptions.

Because our manipulations did not always show a strong influence on participants’ beliefs regarding the research hypothesis, it is possible that a stronger or a different manipulation could reveal more robust influence of demand characteristics on the EC effect. That line of argument is bolstered by the fact that the manipulation did moderate the EC effect in one of the experiments (Experiment 1). On the other hand, the bulk of the evidence in this research is incompatible with the demand account. The moderating effect of the manipulation on the EC effect was small in Experiment 1 and completely absent from Experiments 2 to 3. Experiment 2 was a replication of Experiment 1 with a sample of paid participants who might be more susceptible to demand effects. Indeed, participants in the contrast condition in Experiment 2 were more likely to correctly answer the comprehension measure about the manipulation (77% in Experiment 2 vs. 59% in Experiment 1), and to assume that the researchers expected a contrast effect rather than an assimilative effect (34% in Experiment 2 vs. 20% in Experiment 1). In Experiment 3, we used a more straightforward demand characteristics manipulation, resulting in the strongest effect on participants’ perceived hypothesis. Only in Experiment 3, most participants in the contrast condition reported that it is more likely that the research hypothesis was a contrast effect rather than an assimilative effect. However, although the demand characteristics manipulation had a stronger effect on the manipulation check in Experiments 2 and 3, the results of those two experiments showed no evidence of moderation of the EC effect by the manipulation.

In line with the paucity of experimental evidence for a demand characteristics’ effect on EC in our research, our observational results are similarly incompatible with a demand account for EC. Critically, the stimulus co-occurrence effect was assimilative even among participants who believed that the researchers expected a contrast effect or who thought the typical effect of co-occurrence was contrast. This suggests that even when participants did not perceive the assimilative effect as the expected outcome, they still exhibited it. If demand characteristics were indeed the primary driver of the assimilative effect observed in the control condition, participants who believed that the research hypothesis is a contrast effect would have exhibited a contrast effect. The assimilative effect that these participants showed cannot be easily attributed to demand characteristics.

A further argument against the demand account for the EC effect is that our difficulty in changing participants’ perception of the researchers’ expectations from an assimilative to a contrast effect might suggest that most people view the assimilative effect as obvious and indisputable. Therefore, they reject the idea that researchers would anticipate a contrary outcome. The indisputability of the assimilative effect is compatible with the possibility that participants believe that valence assimilation is the correct inference from stimulus co-occurrence (De Houwer, 2018), and with the possibility that they have often witnessed this effect in their everyday life. These possibilities are contradictory to the assumption that demand characteristics drive EC. In summary, the evidence from three studies with straightforward experimental designs mostly aligns with the findings from prior observational studies (Ingendahl et al., 2023; Sweldens et al., 2010): Demand characteristics are not a major factor in the EC effect.

Supplemental Material

sj-docx-1-spp-10.1177_19485506241252306 – Supplemental material for Experimental Evidence That Demand Characteristics Do Not Play a Dominant Role in the Evaluative Conditioning Effect

Supplemental material, sj-docx-1-spp-10.1177_19485506241252306 for Experimental Evidence That Demand Characteristics Do Not Play a Dominant Role in the Evaluative Conditioning Effect by Yahel Nudler, May Zvi, Gal Levy and Yoav Bar-Anan in Social Psychological and Personality Science

Footnotes

Handling Editor: Christian Unkelbach

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by a grant from the Israel Science Foundation (ISF; grant No. 1684/21) to Y.B.A.

ORCID iD

Yoav Bar-Anan

Supplemental Material

The supplemental material is available in the online version of the article.

Author Biographies

Yahel Nudler holds a Master's degree from Tel-Aviv University. He is currently a PhD student at Tel-Aviv University.

Mai Zvi holds a Master's degree from Tel-Aviv University.

Gal Levy holds a Bachlor's degree from Tel-Aviv University. He is currently an MA student at the Academic College of Tel-Aviv Yaffo.

Yoav Bar-Anan is a full professor at Tel-Aviv University.

References

Alves

Imhoff

(2023). Evaluative context and conditioning effects among same and different objects. Journal of Personality and Social Psychology, 124(4), 735–753. https://doi.org/10.1037/pspa0000323

Baeyens

Eelen

Bergh

O. V. D.

(1990). Contingency awareness in evaluative conditioning: A case for unaware affective-evaluative learning. Cognition and Emotion, 4(1), 3–18. https://doi.org/10.1080/02699939008406760

Cialdini

R. B.

Goldstein

N. J.

(2004). Social influence: Compliance and conformity. Annual Review of Psychology, 55(1), 591–621. https://doi.org/10.1146/annurev.psych.55.090902.142015

Corneille

Lush

(2023). Sixty years after Orne’s American psychologist article: A conceptual framework for subjective experiences elicited by demand characteristics. Personality and Social Psychology Review, 27(1), 83–101. https://doi.org/10.1177/10888683221104368

Corneille

Mertens

(2020). Behavioral and physiological evidence challenges the automatic acquisition of evaluations. Current Directions in Psychological Science, 29(6), 569–574. https://doi.org/10.1177/0963721420964111

Corneille

Stahl

(2019). Associative attitude learning: A closer look at evidence and how it relates to attitude models. Personality and Social Psychology Review, 23(2), 161–189. https://doi.org/10.1177/1088868318763261

De Houwer

. (2018). Propositional models of evaluative conditioning. Social Psychological Bulletin, 13(3), 1–21. https://doi.org/10.5964/spb.v13i3.28046

De Houwer

Thomas

Baeyens

(2001). Association learning of likes and dislikes: A review of 25 years of research on human evaluative conditioning. Psychological Bulletin, 127(6), 853–869. https://psycnet.apa.org/doi/10.1037/0033-2909.127.6.853

Faul

Erdfelder

Buchner

Lang

A. G.

(2009). Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41, 1149–1160. https://doi.org/10.3758/BRM.41.4.1149

10.

Fazio

R. H.

Jackson

J. R.

Dunton

B. C.

Williams

C. J.

(1995). Variability in automatic activation as an unobtrusive measure of racial attitudes: A bona fide pipeline? Journal of Personality and Social Psychology, 69(6), Article 1013. https://psycnet.apa.org/doi/10.1037/0022-3514.69.6.1013

11.

Gawronski

Gast

De Houwer

(2015). Is evaluative conditioning really resistant to extinction? Evidence for changes in evaluative judgements without changes in evaluative representations. Cognition and Emotion, 29(5), 816–830. https://doi.org/10.1080/02699931.2014.947919

12.

Gawronski

Walther

(2012). What do memory data tell us about the role of contingency awareness in evaluative conditioning? Journal of Experimental Social Psychology, 48(3), 617–623. https://doi.org/10.1016/j.jesp.2012.01.002

13.

Ingendahl

Woitzel

Alves

(2023). Just playing the role of good study participants? Evaluative conditioning, demand compliance, and agreeableness. Social Psychological and Personality Science. 0(0). https://doi.org/10.1177/19485506231198653

14.

Insko

C. A.

Oakes

W. F.

(1966). Awareness and the “conditioning” of attitudes. Journal of Personality and Social Psychology, 4(5), 487–495. https://doi.org/10.1037/h0023874

15.

Kurdi

Lozano

Banaji

M. R.

(2017). Introducing the open affective standardized image set (OASIS). Behavior Research Methods, 49, 457–470. https://doi.org/10.3758/s13428-016-0715-3

16.

McGinley

Layton

(1973). Conditioning of evaluative meaning without contingency or demand awareness. The Journal of General Psychology, 89(1), 81–90. https://doi.org/10.1080/00221309.1973.9710820

17.

Moran

Nudler

Bar-Anan

(2023). Evaluative conditioning: Past, present, and future. Annual Review of Psychology, 74, 245–269. https://doi.org/10.1146/annurev-psych-032420-031815

18.

Nosek

B. A.

(2005). Moderators of the relationship between implicit and explicit evaluation. Journal of Experimental Psychology: General, 134(4), 565–584. https://doi.org/10.1037/0096-3445.134.4.565

19.

Orne

M. T.

(1962). On the social psychology of the psychological experiment: With particular reference to demand characteristics and their implications. American Psychologist, 17(11), 776–783. https://doi.org/10.1037/h0043424

20.

Page

M. M.

(1969). Social psychology of a classical conditioning of attitudes experiment. Journal of Personality and Social Psychology, 11(2), 177–186. https://doi.org/10.1037/h0027025

21.

Palan

Schitter

(2018). Prolific. ac—A subject pool for online experiments. Journal of Behavioral and Experimental Finance, 17, 22–27. https://doi.org/10.1016/j.jbef.2017.12.004

22.

Peer

Brandimarte

Samat

Acquisti

(2017). Beyond the Turk: Alternative platforms for crowdsourcing behavioral research. Journal of Experimental Social Psychology, 70, 153–163. https://doi.org/10.1016/j.jesp.2017.01.006

23.

Segal-Doron

Bar-Anan

(n.d.). The beyond reality image collection (BRIC) [unpublished].

24.

Staats

A. W.

(1969). Experimental demand characteristics and the classical conditioning of attitudes. Journal of Personality and Social Psychology, 11(2), 187–192. https://doi.org/10.1037/h0027026

25.

Sweldens

Van Osselaer

S. M.

Janiszewski

(2010). Evaluative conditioning procedures and the resilience of conditioned brand attitudes. Journal of Consumer Research, 37(3), 473–489. https://doi.org/10.1086/653656

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.51 MB