Mere Exposure Effects on Implicit Stimulus Evaluation: The Moderating Role of Evaluation Task,Number of Stimulus Presentations,and Memory for Presentation Frequency

Abstract

The mere exposure (ME) effect refers to the well-established finding that people evaluate a stimulus more positively after repeated exposure to it. So far, the vast majority of studies on ME effects have examined changes in explicit stimulus evaluation. We describe the results of three large-scale studies (combined N = 3,623) that examined ME effects on implicit stimulus evaluation. We looked at three moderators of these effects: the implicit evaluation measure, the number of stimulus presentations, and memory for presentation frequency. We observed ME effects on implicit stimulus evaluations as measured with an Implicit Association Test (IAT) and Affect Misattribution Procedure (AMP), but not an Evaluative Priming Task (EPT). ME effects were more robust when there were relatively few stimulus presentations and when participants had accurate memory for the presentation frequencies. We discuss how these findings relate to ME effects on explicit evaluations as well as theoretical and practical implications.

Keywords

mere exposure implicit evaluation frequency memory IAT evaluative priming

The mere exposure (ME) effect refers to the finding that people tend to prefer stimuli with which they have more experience (Zajonc, 1968). The ME effect is a robust (Bornstein, 1989) and ubiquitous finding in psychology. For example, ME effects have been observed in research on novel products (Janiszewski, 1993), food preferences (Pliner, 1982), and racial prejudice (Zebrowitz, White, & Wieneke, 2008). Whereas an abundant number of studies have examined ME effects on explicit stimulus evaluations as captured by self-reported liking and choice preference measures, only a handful of studies have investigated ME effects on automatic (i.e., implicit) stimulus evaluations as captured by implicit evaluation measures (see below, for an overview). This is a significant lacuna in ME research because implicit evaluation is often considered to be an important determinant of a wide range of behaviors in different domains of psychology such as consumer purchases, voting choices, or addictive behaviors. Indeed, several reviews and meta-analyses indicate that measures of implicit evaluation complement other measures of (explicit) evaluation and explain important additional variance in behavior under certain conditions (De Houwer, Teige-Mocigemba, Spruyt, & Moors, 2009; Eschenbeck, Heim-Dreger, Steinhilber, & Kohlmann, 2016; Hofmann, Gawronski, Gschwendner, Le, & Schmitt, 2005), especially in the context of more automatic or spontaneous behavior (e.g., Dovidio, Kawakami, & Gaertner, 2002; Eschenbeck et al., 2016; Friese, Hofmann, & Schmitt, 2009). Hence, the practical usefulness of ME for changing behavior might depend on whether, and under what circumstances, ME procedures influence implicit stimulus evaluations.

Furthermore, the question of whether ME can lead to changes in implicit stimulus evaluations is also important for evaluating theoretical accounts of ME effects such as the processing fluency/attribution account (e.g., Bornstein & D’Agostino, 1994). This account postulates that repeated exposure to a stimulus results in facilitated processing fluency. This fluency experience can be misattributed to stimulus properties that a participant is asked to rate (such as valence). In accordance with this account, ME effects have also been observed on rated stimulus dimensions other than valence, such as the prototypicality, truth, or brightness of a stimulus (e.g., Mandler, Nakamura, & Van Zandt, 1987). Importantly, according to certain interpretations of the processing fluency/attribution account, ME does not produce changes in a person’s genuine liking of a stimulus but only facilitates changes in overt reports of stimulus evaluation as a consequence of being asked to provide evaluative stimulus ratings (e.g., Whittlesea, 1993). Hence, there are reasons to suspect that ME effects might occur only when participants are required to complete measures of explicit evaluation (allowing for misattribution of the fluency experience to liking), but not when they complete measures of implicit evaluation.

In contrast, alternative accounts of ME effects assume that ME leads to an immediate change in liking that is not critically dependent on the (explicit) measurement of evaluation. For example, the hedonic-fluency account (e.g., Winkielman & Cacioppo, 2001) assumes that processing fluency is an inherently positive experience and, therefore, repeated exposure to a stimulus leads to a genuine change in the liking of a stimulus. Some have argued that this effect should be more easily observed on implicit evaluation measures because these measures are more sensitive to evaluations that arise from unconscious influences such as fluency experiences (e.g., Kawakami, 2012).

Propositional accounts of ME effects, which assume that ME effects depend on the acquisition of propositional knowledge about the frequency of exposure to a stimulus, also predict ME effects on implicit evaluation (Van Dessel, Mertens, Smith, & De Houwer, 2017). For example, when participants infer that a frequently occurring stimulus is positive (e.g., because such stimuli are safe and harmless; Zajonc, 2001), this newly acquired information may influence both explicit and implicit stimulus evaluations (see De Houwer, 2014). Such inferences might occur under certain conditions of automaticity (e.g., unaware or uncontrolled; see Van Dessel, Hughes, & De Houwer, 2018).

Relatively few studies have examined ME effects on implicit evaluation. First, three studies (i.e., Kawakami, 2012; Kawakami & Yoshida, 2015; P. K. Smith, Dijksterhuis, & Chaiken, 2008, Experiment 3) demonstrated subliminal ME effects on implicit measures (i.e., the Affect Misattribution Procedure [AMP], single-category Implicit Association Test [SC-IAT], and the Evaluative Priming Task [EPT], respectively). These studies involved 10 to 13 ms repeated presentations of either face stimuli or Nepalese signs, followed by a mask to prevent conscious reports of the stimuli. Because many recent studies have cast doubt on reported evidence for subliminal perception effects (e.g., Simmons, Nelson, & Simonsohn, 2011; Vadillo, Konstantinidis, & Shanks, 2016; Van Dessel, De Houwer, Roets, & Gast, 2016), we calculated Bayes Factors (BFs) for the reported tests in these studies. These BFs provide an indication of how strongly the data support either the null hypothesis (BF₀; reflecting the absence of an effect) or the alternative hypothesis (BF₁; reflecting the presence of an effect). BFs between 1 and 3, between 3 and 10, and larger than 10, respectively, designate “anecdotal evidence,” “substantial evidence,” and “strong evidence” for the tested hypothesis—most commonly the null (BF₀) or the alternative hypothesis (BF₁; Jeffreys, 1961). Overall, evidence in favor of the effects was low (BF_1s < 3), with the exception of one reported effect in Kawakami and Yoshida (2015; Experiment 2: BF₁ = 28.42).

Second, two supraliminal ME studies have also used implicit evaluation measures. However, these studies did not focus on the effect of ME on the evaluation of specific stimuli, but rather on the effects of ME on the evaluation of categories of stimuli (e.g., the implicit evaluation of Japanese writing systems following exposure to exemplars of words written in those writing systems; Kawakami, Sato, & Yoshida, 2010) and on general positive affect (e.g., the overall evaluation of artificial words following exposure to Chinese ideographs; Hicks & King, 2011).

Third, ME effects have been reported on psychophysiological measures such as facial electromyography (e.g., Harmon-Jones & Allen, 2001; Winkielman & Cacioppo, 2001; Winkielman, Halberstadt, Fazendeiro, & Catty, 2006; Witvliet & Vrana, 2007). However, it is unclear to what extent psychophysiological responses reflect implicit evaluation (De Houwer & Moors, 2010). Finally, a recent set of studies demonstrated that ME instructions (i.e., instructions about the number of upcoming presentations of stimuli in the absence of actual presentations) can influence implicit evaluations of individual stimuli (Van Dessel et al., 2017). However, it is not clear whether ME instruction effects rely on the same mechanisms as ME effects instantiated through actual stimulus presentations. Hence, currently there is only limited evidence that ME through actual stimulus presentations can influence the implicit evaluation of those stimuli.

In the current study, we investigated whether ME can influence implicit evaluations and additionally assessed three potential boundary conditions of ME effects on implicit evaluation. First, we examined whether ME effects depend on the task that is used to measure implicit stimulus evaluations. More specifically, we examined ME effects on implicit evaluations as measured with the IAT (Greenwald, McGhee, & Schwartz, 1998), EPT (Fazio, Sanbonmatsu, Powell, & Kardes, 1986), and AMP (Payne, Cheng, Govorun, & Stewart, 2005). These three tasks were chosen because (a) they constitute the most widely used tasks to measure implicit evaluation, (b) they are differentially sensitive to a number of factors other than the to-be-measured psychological construct of implicit evaluation (e.g., extra-personal knowledge: Olson & Fazio, 2004; salience asymmetries: Rothermund & Wentura, 2004) and are thus assumed to involve different underlying processes (De Houwer et al., 2009), and (c) they conform with important normative criteria of implicit evaluation measures to a different extent (De Houwer et al., 2009). For the sake of comparison, we have also included a measure of explicit stimulus evaluation (i.e., a self-reported liking rating scale).

Second, we manipulated the number of stimulus presentations in the ME task. Previous research has shown that this can be an important moderator of ME effects on explicit stimulus evaluations (Bornstein, 1989; Van den Bergh & Vrana, 1998). It is typically observed that a minimum number of stimulus presentations is needed to produce an ME effect, yet the ME effect also seems to decrease in size after a relatively small number of stimulus presentations (e.g., 10-20 presentations; Bornstein, 1989; Montoya, Horton, Vevea, Citkowicz, & Lauber, 2017; Zajonc, Shaver, Tavris, & van Kreveld, 1972). We examined whether this moderation is also observed for ME effects on implicit evaluations.

Third, we investigated whether ME effects depend on participants’ memory for the stimulus presentation frequencies. There is much debate about the importance of this factor for ME effects (Bornstein & D’Agostino, 1992; Newell & Shanks, 2007; Stafford & Grimes, 2012). Whereas some authors have stressed that memory for the presentation frequencies (and even conscious recognition of the stimuli at the time of exposure) is not necessary for, or could even hamper, ME effects (Bornstein, 1989; Monahan, Murphy, & Zajonc, 2000), other authors argue that memory for presentation frequencies is an important moderator of the ME effect (Brooks & Watkins, 1989; Newell & Shanks, 2007; Stafford & Grimes, 2012). This discrepancy may be due to the fact that some studies used small samples thus leading to more unreliable effects (Bornstein, 1989; Stafford & Grimes, 2012). In-line with Bar-Anan, De Houwer, and Nosek (2010), who investigated the relationship between memory of stimulus-stimulus contingencies and evaluative conditioning (EC) in a large sample, we recruited a large number of participants to investigate the relationship between memory of stimulus presentation frequencies and ME effects. To gain information about the strength of evidence for the presence or absence of ME effects for participants with either accurate or inaccurate presentation frequency memory, we supplemented traditional t-test analyses with Bayesian analyses (Dienes, 2011; Wagenmakers, 2007).

To address the above-mentioned questions, we conducted three large-scale experiments. All experiments used the same general procedure to manipulate the amount of exposure to different stimuli (Experiments 1 and 2: nonwords; Experiment 3: unknown brands). After this ME phase, implicit evaluations were measured with the IAT (Experiment 1, Experiment 2), EPT (Experiment 2), or AMP (Experiment 3).

Method

Participants

Participants were 892, 1,392, and 1,339 visitors to the Project Implicit research website (https://implicit.harvard.edu) in Experiments 1, 2, and 3, respectively. Prior to data collection, target sample size of Experiment 3 was preregistered together with the study design, data-analysis plan, and the described hypotheses. These preregistered plans as well as experiment scripts, stimuli, data, and analysis code of all experiments are available at https://osf.io/dnqcs/. In-line with standard procedures of data-reduction for Project Implicit data (e.g., C. T. Smith, De Houwer, & Nosek, 2013), we excluded data of participants who (a) did not complete all tasks (131 participants in Experiment 1, 14.7%; 184 participants in Experiment 2, 13.2%; 171 participants in Experiment 3, 12.8%), (b) had error rates above 30% when considering all critical IAT test blocks or above 40% for any one of these blocks (12 participants in Experiment 1, 1.6%; nine participants in Experiment 2, 0.8%), (c) had error rates in the EPT that exceeded the population mean by more than 2.5 standard deviations (eight participants in Experiment 2, 0.7%, population M = 7.2%, SD = 10.7%), or (d) used the same response key in the AMP for more than 90% of the trials (211 participants, 15.8%). The analyses were performed on the data of 749 participants (61.2% women, mean age = 35 years, SD = 13 years, range = 18-79 years) in Experiment 1, 1,191 participants (63.6% women, mean age = 32 years, SD = 13 years, range = 18-76 years) in Experiment 2, and 956 participants (58.1% women, mean age = 31 years, SD = 13 years, range = 18-77 years) in Experiment 3.

Procedure

ME phase

After participants gave informed consent, they were told that they would see one or more stimuli (words in Experiments 1 and 2, novel food brands in Experiment 3) presented on the screen sequentially, that is, one after the other. They were asked to pay close attention to the stimuli because this would be vital for the successful completion of the study. Participants then went through an ME phase in which they saw presentations of two non-existing words “FEVKANI” and “LOKANTA” (Experiments 1 and 2) or three novel brand names with logos (Empeya, Levida, and Witkap). The stimuli remained on the screen for 1,000 ms with a 1,000 ms inter-trial interval. The number of stimulus presentations was manipulated between-subjects such that for participants in Experiment 1 (a) one word was presented 2 times and the other word was never presented (0-2 condition), (b) one word was presented 3 times and the other word was presented once (1-3 condition), (c) one word was presented 6 times and the other word was presented once (1-6 condition), or (d) one word was presented 12 times and the infrequent word was presented once (1-12 condition). In Experiment 2, there were only two Stimulus Pair conditions: the 0-2 condition and the 1-12 condition. These conditions were selected because they were the conditions in which we had observed the strongest ME effects on implicit evaluation in Experiment 1. In Experiment 3, all participants saw one brand 12 times, one brand 2 times, and one brand was never presented (0-2-12 condition). Which specific word or brand was presented more often was counterbalanced across participants and Stimulus Pair conditions. The order of the stimulus presentations within the ME phase was randomized.

Explicit evaluation

For half of the participants of Experiments 1 and 3, the ME phase was followed by an explicit evaluation task, which was then followed by the implicit evaluation task. The other participants first completed the implicit evaluation task and then the explicit evaluation task. In Experiment 2, all participants first completed the implicit evaluation task and then completed the explicit evaluation task. In the explicit evaluation task, participants indicated liking ratings of each of the two nonwords (Experiments 1 and 2) or three brand names (Experiment 3) by selecting an option on a 9-point Likert-type scale (1 = not liked at all to 9 = completely liked) from a dropdown list on separate pages. The Likert-type scale for the different stimuli were presented in random order.

IAT (Experiments 1 and 2)

In Experiment 1, all participants completed the IAT; in Experiment 2, half of the participants completed an IAT and the other participants completed an EPT. In the IAT, participants were asked to sort stimuli by pressing either the “E” or the “I” on the keyboard. On each trial, a word was presented in the center of the screen until the participant pressed one of the two keys. If the response was correct, the word disappeared and the next word was presented 400 ms later. If the response was incorrect, the word was replaced by a red “X.” The next word appeared 400 ms after participants pushed the correct button. In the first block, participants categorized FEVKANI and LOKANTA as their respective names. To avoid classification of the target stimuli based only on simple perceptual features, the words were presented in different font types (Arial Black and Fixedsys), capitalizations (uppercase and lowercase), and sizes (16pt and 18pt), resulting in eight different stimuli for each nonword. Category labels were presented in the top left and right corner to aid classification. After 20 trials, participants categorized 10 attribute words as “Good” (wonderful, glorious, marvelous, success, peace) or “Bad” (nasty, failure, agony, unpleasant, evil) with the “E” and the “I” buttons for 20 trials. Next, participants completed 20 practice trials and 40 critical trials in which both attribute and target words were categorized and in which FEVKANI and positive stimuli shared the same response key and LOKANTA and negative stimuli shared the other response key (or vice versa). Participants then practiced sorting target words with the response key assignment reversed for 40 trials. Finally, participants completed 20 practice and 40 critical trials with the new response key assignment.

EPT (Experiment 2)

At the start of the EPT, participants were told that words would appear one after the other on the screen and that their task was to categorize the words as either “good” or “bad” using the “E” and “I” keys of a computer keyboard as quickly as possible, while making as few mistakes as possible. Participants were further told that they would see words presented before the positive and negative words and that they should not respond to those words. Participants were then shown a list of the 14 positive and 14 negative words that they would have to categorize. In-line with standard procedures (Spruyt, De Houwer, Hermans, & Eelen, 2007), a single trial consisted of a fixation cross presented in white for 500 ms, a blank screen for 500 ms, a prime for 200 ms, a post-prime pause for 50 ms, and the presentation of a target word in white font for 1,500 ms. The inter-trial interval was set to vary randomly between 500 ms and 1,500 ms. There were four types of trials: (a) trials with the word LOKANTA as prime and a positive word as target, (b) trials with the word LOKANTA and a negative target, (c) trials with the word FEVKANI and a positive target, and (e) trials with the word FEVKANI and a negative target. Each type of trial was presented on a quarter of the trials. Participants first completed eight practice trials (two of each of the four types of trials) and then completed 120 trials separated into three blocks of 40 trials, each containing 10 of the four types of trials, presented in random order.

AMP (Experiment 3)

In accordance with standard procedures (Payne et al., 2005), the AMP consisted of three blocks of 30 trials in which participants were presented with a prime stimulus for 75 ms, a blank screen for 125 ms, and a Chinese ideograph for 100 ms, which was then covered with a black-and-white pattern mask. The three brands Empeya, Levida, and Witkap served as prime stimuli. Each trial, participants indicated if they considered the Chinese ideograph more or less visually pleasant than average by pressing either “I” or “E,” respectively. Participants were asked to ignore the prime stimuli and respond only to the Chinese ideographs.

Stimulus frequency memory measurement

At the end of the experiment, participants were asked to indicate how many times they had seen each of the two words or the three brands during the first (ME) task. Participants could choose a number between 0 and 15 from a dropdown list for each stimulus. The order of the questions was randomized.

Task engagement measurement

In Experiment 3, we assessed task engagement for the ME phase by asking participants to rate their levels of boredom and attention in this task with two 10-point rating scales (short version of the Dundee Stress State Questionnaire; Helton & Naswall, 2015).

Results

Data Preparation

IAT ME scores were calculated using the D2-algorithm (Greenwald, Nosek, & Banaji, 2003), such that higher scores indicate a stronger ME effect (i.e., a stronger preference for the frequently presented word over the infrequently presented word). EPT ME scores (Experiment 2) were created by (a) subtracting the mean latencies on trials with a positive target and the frequent word prime from the mean latencies on trials with a negative target and the frequent word prime, (b) subtracting the mean latencies on trials with a positive target and the infrequent word prime from the mean latencies on trials with a negative target and the infrequent word prime, and (c) subtracting the second difference score from the first difference score. EPT ME scores were calculated on the basis of EPT trials that remained after exclusion of trials with an incorrect response (3.5%) and trials with reaction times that were at least 2.5 standard deviations removed from an individual’s mean for that type of trial (2.9%). Three AMP ME scores (Experiment 3) were calculated for each participant by subtracting the proportion of “pleasant” responses on (a) trials with the brand presented 0 times as prime from trials with the brand presented 12 times as prime (0-12 ME score), (b) trials with the brand presented 0 times as prime from trials with the brand presented 2 times as prime (0-2 ME score), and (c) trials with the brand presented 2 times as prime from trials with the brand presented 12 times as prime (2-12 ME score). The Spearman-Brown corrected split-half reliability was r(748) = .86 (Experiment 1) and r(590) = .87 (Experiment 2) for the IAT ME scores, r(597) = .43 for the EPT ME scores, and r(954) = [.51-.61] for the AMP ME scores.

Explicit rating ME scores were calculated by subtracting participants’ score rating for the infrequent word from their score rating for the frequent word (Experiments 1 and 2). For Experiment 3, three explicit rating ME scores were calculated by subtracting ratings for the infrequent brands from ratings for the frequent brands (0-12 ME score, 0-2 ME score, 2-12 ME scores). Explicit rating ME scores correlated significantly with IAT ME scores, r(747) = .32, p < .001 (Experiment 1), r(590) = .27, p < .001 (Experiment 2), and AMP ME scores, r(953) = .22 (0-12 ME scores), r(953) = .21 (0-2 ME scores), r(953) = .21 (2-12 ME scores), ps < .001 (Experiment 3), but not EPT ME scores, r(597) = .05, p = .18 (Experiment 2).

Stimulus frequency memory was coded as accurate for participants who correctly indicated that the frequent word was presented more often than the infrequent word (Experiment 1: 392 participants, 52.3%; Experiment 2: 690 participants, 57.9%; Experiment 3: 0-12 pair: 808 participants, 84.7%; 0-2 pair: 711 participants, 74.5%; 2-12 pair: 719 participants, 75.3%). It was coded as reversed for participants who indicated that the frequent word was presented less often than the infrequent word (Experiment 1: 116 participants, 15.5%; Experiment 2: 176 participants, 14.8%; Experiment 3: 0-12 pair: 90 participants, 9.4%; 0-2 pair: 164 participants, 17.2%; 2-12 pair: 127 participants, 13.3%) and as indiscriminate for participants who indicated that both words had been presented equally often (Experiment 1: 241 participants, 32.2%; Experiment 2: 325 participants, 27.3%; Experiment 3: 0-12 pair: 57 participants, 5.9%; 0-2 pair: 80 participants, 8.3%; 2-12 pair: 109 participants, 11.4%). We also created an index of subjective ME experience by subtracting the number of reported stimulus presentations for the infrequent word from the number of reported stimulus presentations for the frequent word. In Experiment 1, participants with accurate memory indicated smaller differences in the number of presentations for the 0-2 pair (M = 2.33, SD = 1.48) than for the 1-3 (M = 3.05, SD = 2.03), 1-6 (M = 5.97, SD = 3.62), and 1-12 pair (M = 9.21, SD = 4.17), ps < .014. In Experiment 2, participants with accurate memory indicated smaller differences for the 0-2 (M = 3.21, SD = 3.00) than for the 1-12 pair (M = 8.09, SD = 4.25), t(688) = 16.82, p < .001. In Experiment 3, participants with accurate memory indicated smaller differences for the 0-2 pair (M = 4.17, SD = 2.62) than for the 2-12 (M = 6.78, SD = 3.37) or 0-12 pair (M = 8.83, SD = 4.22), ps < .001.

Implicit Evaluation

IAT

In Experiment 1, IAT ME scores were significantly higher than zero, indicating an ME effect on IAT performance (M = 0.07, SD = 0.48), t(748) = 4.14, p < .001, d_z = 0.15, 95% confidence interval (CI) = [0.04, 0.11], BF₁ = 190.51. We performed an analysis of variance (ANOVA) on IAT ME scores that included Memory (accurate, indiscriminate, reversed), IAT Block Order (positive words and frequent word categorized with the same key in the first block, positive words and infrequent word categorized with the same key in the first block), Task Order (implicit evaluation task first, explicit evaluation task first), and Stimulus Pair (0-2, 1-3, 1-6, 1-12) as between-subject factors.¹ This revealed only a main effect of IAT Block Order, F(1, 701) = 5.21, p = .032, but not any other main or interaction effects, Fs < 3.11, ps > .078.² Planned one-sample t tests indicated that participants with accurate memory significantly preferred the frequent word (M = 0.10, SD = 0.48), t(391) = 4.21, p < .001, d_z = 0.21, 95% CI = [0.05, 0.15], BF₁ = 303.78. We did not observe an ME effect for participants with indiscriminate or reversed memory, ts < 1.49, ps > .13, d_zs < 0.10, BF₀s > 4.62. The ME effect for participants with accurate memory was significant only for the 0-2 pair (M = 0.16, SD = 0.44), t(77) = 3.25, p = .001, d_z = 0.37, 95% CI = [0.06, 0.26], BF₁ = 15.12, and the 1-12 pair (M = 0.13, SD = 0.50), t(130) = 2.91, p = .004, d_z = 0.25, 95% CI = [0.04, 0.21], BF₁ = 5.44, but not the 1-3 pair (M = 0.13, SD = 0.50), t(72) = 1.95, p = .055, d_z = 0.23, 95% CI = [0.00, 0.22], BF₀ = 1.33, or the 1-6 pair (M = 0.02, SD = 0.46), t(109) = 0.46, p = .64, d_z = 0.04, 95% CI = [–0.07, 0.11], BF₀ = 8.52. A summary of the t-test results is provided in Table 1.

Table 1.

Mean Implicit and Explicit Evaluation Scores Indicating a Preference for the Frequent Word Over the Infrequent Word in Experiment 1 as a Function of Stimulus Pair and Memory.

	Stimulus Pair	Memory	N	M (SD)	Test statistic	Effect size d_z	Bayesian t test
IAT ME score	0-2	accurate	78	0.16 (0.44)	t(77) = 3.25, p = .001	0.37	BF₁ = 15.12, Strong (H₁)
		indiscriminate	99	0.03 (0.48)	t(98) = 0.64, p = .52	0.06	BF₀ = 7.38, Substantial (H₀)
		reversed	26	0.00 (0.43)	t(25) = 0.06, p = .95	0.01	BF₀ = 4.82, Substantial (H₀)
	1-3	accurate	73	0.13 (0.50)	t(72) = 1.95, p = .055	0.23	BF₁ = 1.30, Anecdotal (H₁)
		indiscriminate	68	0.04 (0.47)	t(67) = 0.77, p = .45	0.09	BF₀ = 5.67, Substantial (H₀)
		reversed	40	0.05 (0.47)	t(39) = 0.66, p = .51	0.10	BF₀ = 4.78, Substantial (H₀)
	1-6	accurate	110	0.02 (0.46)	t(109) = 0.46, p = .64	0.04	BF₀ = 8.52, Substantial (H₀)
		indiscriminate	45	0.10 (0.50)	t(44) = 1.38, p = .17	0.21	BF₀ = 2.55, Anecdotal (H₀)
		reversed	34	0.00 (0.57)	t(33) = 0.01, p = .99	0.00	BF₀ = 5.44, Substantial (H₀)
	1-12	accurate	131	0.13 (0.50)	t(130) = 2.91, p = .004	0.25	BF₁ = 5.44, Substantial (H₁)
		indiscriminate	29	0.01 (0.48)	t(28) = 0.17, p = .87	0.03	BF₀ = 5.00, Substantial (H₀)
		reversed	16	0.08 (0.44)	t(15) = 0.75, p = .47	0.19	BF₀ = 3.07, Substantial (H₀)
Explicit rating ME score^a	0-2	accurate	44	0.55 (1.81)	t(43) = 2.00, p = .052	0.30	BF₁ = 1.00, Anecdotal (H₁)
		indiscriminate	56	0.66 (2.30)	t(55) = 2.15, p = .036	0.29	BF₁ = 1.22, Anecdotal (H₁)
		reversed	12	−0.25 (2.34)	t(11) = −0.37, p = .72	0.11	BF₀ = 3.28, Substantial (H₀)
	1-3	accurate	29	0.62 (2.87)	t(28) = 1.16, p = .25	0.22	BF₀ = 2.74, Anecdotal (H₀)
		indiscriminate	35	0.09 (1.14)	t(34) = 0.44, p = .66	0.07	BF₀ = 5.03, Substantial (H₀)
		reversed	23	0.43 (2.31)	t(22) = 0.90, p = .38	0.19	BF₀ = 3.18, Substantial (H₀)
	1-6	accurate	45	0.22 (1.94)	t(44) = 0.77, p = .45	0.11	BF₀ = 4.69, Substantial (H₀)
		indiscriminate	28	−0.04 (2.80)	t(27) = −0.07, p = .95	0.01	BF₀ = 4.98, Substantial (H₀)
		reversed	17	−0.59 (1.84)	t(16) = −1.32, p = .21	0.32	BF₀ = 1.92, Anecdotal (H₀)
	1-12	accurate	58	0.03 (1.98)	t(57) = 0.13, p = .89	0.02	BF₀ = 6.91, Substantial (H₀)
		indiscriminate	19	−0.05 (2.88)	t(18) = −0.08, p = .94	0.02	BF₀ = 4.20, Substantial (H₀)
		reversed	9	−0.89 (2.80)	t(8) = −0.13, p = .37	0.32	BF₀ = 2.15, Anecdotal (H₀)

Note. IAT = Implicit Association Test; ME = mere exposure; BF₁ = reflecting the presence of an effect; BF₀ = reflecting the absence of an effect.

For explicit rating ME scores, analyses were restricted to participants who completed the explicit evaluation task before the IAT.

In Experiment 2, IAT ME scores also indicated an ME effect on IAT performance (M = 0.09, SD = 0.49), t(591) = 4.41, p < .001, d_z = 0.18, 95% CI = [0.05, 0.13], BF₁ = 623.02. An ANOVA on IAT ME scores with Memory, IAT Block Order, and Stimulus Pair as factors revealed a significant main effect of IAT Block Order, F(1, 585) = 13.46, p < .001, but no other main or interaction effects, Fs < 2.09, ps > .12. We observed a significant ME effect for participants with accurate memory (M = 0.12, SD = 0.45), t(324) = 4.62, p < .001, d_z = 0.26, 95% CI = [0.07, 0.16], BF₁ = 1,754.95, but not for participants with indiscriminate or reversed memory, ts < 1.82, ps > .072, d_zs < 0.13, BF₀s > 2.50 (Table 2). Participants with accurate memory exhibited a significant ME effect for the 0-2 pair (M = 0.17, SD = 0.43), t(122) = 4.45, p < .001, d_z = 0.40, 95% CI = [0.10, 0.25], BF₁ = 813.57, and the 1-12 pair (M = 0.08, SD = 0.46), t(201) = 2.48, p = .014, d_z = 0.17, 95% CI = [0.02, 0.14], but the evidence for the latter effect was only anecdotal, BF₁ = 1.54.³

Table 2.

Mean Implicit and Explicit Evaluation Scores Indicating a Preference for the Frequent Word Over the Infrequent Word in Experiment 2 as a Function of Stimulus Pair and Memory.

	Stimulus Pair	Memory	N	M (SD)	Test statistic	Effect size d_z	Bayesian t test
IAT ME score	0-2	accurate	123	0.17 (0.43)	t(122) = 4.45, p < .001	0.40	BF₁ = 813.57, Strong (H₁)
		indiscriminate	129	0.05 (0.52)	t(128) = 1.18, p = .24	0.10	BF₀ = 5.17, Substantial (H₀)
		reversed	52	0.05 (0.54)	t(51) = 0.71, p = .48	0.10	BF₀ = 5.23, Substantial (H₀)
	1-12	accurate	202	0.08 (0.46)	t(201) = 2.48, p = .014	0.17	BF₁ = 1.54, Anecdotal (H₁)
		indiscriminate	58	0.10 (0.49)	t(57) = 1.51 p = .14	0.20	BF₀ = 2.39, Anecdotal (H₀)
		reversed	28	−0.03 (0.60)	t(27) = −0.24 p = .81	0.05	BF₀ = 4.86, Substantial (H₀)
EPT ME score	0-2	accurate	173	5.99 (83.22)	t(172) = 0.95, p = .35	0.07	BF₀ = 7.59, Substantial (H₀)
		indiscriminate	99	−35.99 (183.24)	t(98) = −1.95 p = .054	0.20	BF₀ = 1.45, Anecdotal (H₀)
		reversed	58	−0.69 (75.64)	t(57) = −0.07 p = .95	0.01	BF₀ = 6.95, Substantial (H₀)
	1-12	accurate	192	11.86 (104.36)	t(191) = 1.57, p = .12	0.11	BF₀ = 3.68, Substantial (H₀)
		indiscriminate	39	−29.26 (92.19)	t(38) = −1.98, p = .055	0.32	BF₀ = 1.00, Anecdotal (H₀)
		reversed	38	21.48 (188.41)	t(37) = 0.70, p = .49	0.11	BF₀ = 4.55, Substantial (H₀)
Explicit rating ME score	0-2	accurate	296	0.40 (2.07)	t(295) = 3.31, p = .001	0.19	BF₁ = 13.62, Strong (H₁)
		indiscriminate	228	0.15 (1.92)	t(227) = 1.21, p = .23	0.08	BF₀ = 6.60, Substantial (H₀)
		reversed	110	−0.03 (2.32)	t(109) = −0.12, p = .23	0.01	BF₀ = 9.39, Substantial (H₀)
	1-12	accurate	394	0.09 (1.98)	t(393) = 0.94, p = .35	0.05	BF₀ = 11.37, Strong (H₀)
		indiscriminate	97	−0.03 (2.30)	t(96) = −0.13, p = .90	0.01	BF₀ = 8.83, Substantial (H₀)
		Reversed	66	−0.15 (2.03)	t(65) = −0.61, p = .55	0.07	BF₀ = 6.21, Substantial (H₀)

Note. IAT = Implicit Association Test; ME = mere exposure; BF₁ = reflecting the presence of an effect; BF₀ = reflecting the absence of an effect; EPT = Evaluative Priming Task.

EPT

Overall, EPT ME scores in Experiment 2 did not differ significantly from zero (M = −1.02, SD = 120.87), t(598) = −0.21, p = .84, d_z = 0.01, 95% CI = [–10.72, 8.67], BF₀ = 21.30. An ANOVA on EPT ME scores that included Memory, Task Order, and Stimulus Pair as factors revealed a significant main effect of Memory, F(2, 593) = 3.98, p = .019. We observed a contrast ME effect for participants with indiscriminate memory (M = −34.09, SD = 162.43), t(137) = −2.47, p = .015, d_z = −0.21, 95% CI = [–61.43, –6.75], but evidence for this effect was only anecdotal, BF₁ = 1.75. We did not observe significant ME effects for participants with accurate or reversed memory, ts < 1.84, ps > .068, d_zs < 0.11, BF₀s > 3.26, or for participants with accurate memory for any of the stimulus pairs, ts < 1.58, ps > .11, d_zs < 0.12, BF₀s > 3.68 (Table 2). Additional between-subject t tests indicated that standardized IAT ME scores were significantly larger than standardized EPT ME scores for Experiment 2 participants with accurate memory for the 0-2 pair, t(294) = 3.00, p = .001, BF₁ = 18.04, but not the 1-12 pair, t(392) = 0.49, p = .31, BF₀ = 5.84.

AMP

AMP ME scores in Experiment 3 were significantly higher than zero, indicating an ME effect on AMP performance (M = 1.02%, SD = 18.33%), t(2867) = 2.97, p = .003, d_z = 0.06, 95% CI = [0.34%, 1.69%], BF₁ = 11.16. An ANOVA on AMP ME scores that included Stimulus Pair (0-12, 0-2, 2-12) as within-subject factor and Memory (accurate, indiscriminate, reversed) and Task Order (implicit evaluation task first, explicit evaluation task first) as between-subject factors revealed only a main effect of Task Order, χ²(1) = 4.22, p = .040, indicating bigger ME effects when participants started with the AMP, but not any other main or interaction effects, χ²s < 6.99, ps > .13. Planned follow-up t tests indicated that participants with accurate memory significantly preferred the frequent word (M = 1.29%, SD = 17.81%), t(2237) = 3.43, p < .001, d_z = 0.07, 95% CI = [0.55%, 2.03%], BF₁ = 52.57. We did not observe an ME effect for participants with indiscriminate or reversed memory, ts < 0.45, ps > .65, d_zs < 0.03, BF₀s > 2.98. The ME effect for participants with accurate memory was significant for the 0-12 pair (M = 1.79%, SD = 18.96%), t(807) = 2.68, p = .008, d_z = 0.09, 95% CI = [0.48%, 3.10%], BF₁ = 8.10, and the 0-2 pair (M = 1.31%, SD = 17.12%), t(710) = 2.04, p = .042, d_z = 0.08, 95% CI = [0.48%, 2.57%], BF₁ = 1.99, but not the 2-12 pair (M = 0.72%, SD = 17.14%), t(718) = 1.12, p = .26, d_z = 0.04, 95% CI = [–0.54%, 1.97%], BF₀ = 2.22 (Table 3).

Table 3.

Mean Implicit and Explicit Evaluation Scores Indicating a Preference for the Frequent Word Over the Infrequent Word in Experiment 3 as a Function of Stimulus Pair and Memory.

	Stimulus Pair	Memory	N	M (SD)	Test statistic	Effect size d_z	Bayesian t test
AMP ME score	0-2	Accurate	711	1.31 (17.12)	t(710) = 2.04, p = .042	0.08	BF₁ = 1.99, Anecdotal (H₁)
		Indiscriminate	81	−0.78 (21.95)	t(80) = −0.32, p = .75	0.04	BF₀ = 3.37, Substantial (H₀)
		Reversed	164	0.53 (21.46)	t(163) = 0.32, p = .75	0.02	BF₀ = 2.83, Anecdotal (H₀)
	0-12	Accurate	808	1.79 (18.96)	t(807) = 2.68, p = .008	0.09	BF₁ = 8.10, Substantial (H₁)
		Indiscriminate	58	0.80 (25.08)	t(57) = 0.24 p = .81	0.03	BF₀ = 2.08, Anecdotal (H₀)
		Reversed	90	−0.37 (20.79)	t(89) = −0.17 p = .87	0.02	BF₀ = 3.20, Substantial (H₀)
	2-12	Accurate	719	0.72 (17.14)	t(718) = 1.12, p = .26	0.04	BF₀ = 2.22, Anecdotal (H₀)
		Indiscriminate	110	1.48 (17.34)	t(109) = 0.90 p = .37	0.09	BF₀ = 1.44, Anecdotal (H₀)
		Reversed	127	−1.39 (15.97)	t(126) = −0.98 p = .33	0.09	BF₀ = 5.76, Substantial (H₀)
Explicit rating ME score	0-2	Accurate	711	0.19 (1.83)	t(710) = 2.83, p = .005	0.11	BF₁ = 12.43, Strong (H₁)
		Indiscriminate	81	0.05 (1.77)	t(80) = 0.25, p = .80	0.03	BF₀ = 2.31, Anecdotal (H₀)
		Reversed	164	0.05 (1.87)	t(163) = 0.38, p = .70	0.03	BF₀ = 2.69, Anecdotal (H₀)
	0-12	Accurate	808	0.15 (1.84)	t(807) = 2.35, p = .019	0.08	BF₁ = 3.67, Substantial (H₁)
		Indiscriminate	58	0.36 (2.18)	t(57) = 1.26, p = .21	0.17	BF₁ = 1.19, Anecdotal (H₁)
		Reversed	90	−0.11 (1.56)	t(89) = −0.68, p = .50	0.07	BF₀ = 4.27, Substantial (H₀)
	2-12	Accurate	719	0.04 (1.88)	t(718) = 0.52, p = .61	0.02	BF₀ = 4.43, Substantial (H₀)
		Indiscriminate	110	−0.24 (1.45)	t(109) = −1.71, p = .091	0.16	BF₀ = 7.32, Substantial (H₀)
		Reversed	127	−0.13 (1.86)	t(126) = −0.81, p = .42	0.07	BF₀ = 5.30, Substantial (H₀)

Note. AMP = Affect Misattribution Procedure; ME = mere exposure; BF₁ = reflecting the presence of an effect; BF₀ = reflecting the absence of an effect.

In-line with our preregistered data-analysis plan, we performed additional analyses excluding participants who reported low task engagement in the ME task (mean score < 3; overall M = 4.66, SD = 2.09). In these analyses, the same effects were significant as in the other analyses with the exception that we now observed a significant interaction effect of Task Order and Memory, χ²(2) = 10.63, p = .005, indicating a Memory main effect for participants who started with the explicit rating task, χ²(2) = 9.18, p = .010, but not for participants who started with the AMP, χ²(2) = 4.33, p = .11.⁴

Compound analysis

The performed ANOVA’s on implicit evaluation scores did not provide clear evidence for a moderation of ME effects by Memory or Stimulus Pair. To explore whether this might be due to a lack of statistical power in these analyses, we decided to perform additional ANOVA’s on standardized IAT and AMP ME scores (but not EPT ME scores) for participants in all experiments. These analyses also allowed us to compare ME effects on IAT and AMP scores.

First, an ANOVA that included Memory and Implicit Evaluation Task revealed a main effect of Task, F(1, 4203) = 6.07, p = .014, indicating stronger effects on IAT scores than on AMP scores. More importantly, we also observed a main effect of Memory, F(2, 4203) = 3.99, p = .019. There was a significant ME effect for participants with accurate memory (M = 0.11, SD = 0.97), t(2954) = 6.02, p < .001, d_z = 0.11, 95% CI = [0.07, 0.14], BF₁ > 10.000, or with indiscriminate memory (M = 0.08, SD = 1.06), t(676) = 2.06, p = .039, d_z = 0.08, 95% CI = [0.004, 0.165], BF₁ = 2.14, but not for participants with reversed memory (M = 0.01, SD = 1.07), t(576) = 0.18, p = .86, d_z = 0.01, 95% CI = [–0.08, 0.10], BF₀ = 5.44. Note that the evidence for an ME effect for participants with indiscriminate memory was only anecdotal. However, there is substantial evidence for the absence of a difference in effect sizes between participants with accurate and indiscriminate memory, BF₀ = 5.34.

Second, we performed an ANOVA for participants with accurate memory that included Implicit Evaluation Task (AMP, IAT) as well as two Stimulus Pair variables: (a) Infrequent Stimulus Presentation (whether the infrequent stimulus was presented or not) and (b) Difference in Number of Presentations for the frequent and infrequent stimulus (range 2-10). This revealed a main effect of Evaluation Task, F(1, 2950) = 17.35, p < .001, as well as a main effect of Infrequent Stimulus Presentation, F(1, 2951) = 4.01, p = .045. The ME effect was stronger when the infrequent stimulus was not presented (M = 0.14, SD = 0.98) than when it was presented (M = 0.04, SD = 0.96). Notably, however, ME effects were significant for both types of stimulus pairs, ps < .001, BF₁s > 79.

Explicit Evaluation

Experiments 1 and 2

Overall, explicit rating ME scores were significantly higher than zero, indicating an ME effect on explicit ratings (M = 0.13, SD = 2.01), t(1939) = 2.93, p = .003, d_z = 0.07, 95% CI = [0.04, 0.22], BF₁ = 11.82. An ANOVA on explicit rating ME scores that included Memory, Task Order, and Stimulus Pair as factors did not reveal any main or interaction effects, Fs < 1.71, ps > .16. The ME effect was significant for participants with accurate memory (M = 0.19, SD = 1.99), t(1081) = 3.12, p = .002, d_z = 0.09, 95% CI = [0.07, 0.31], BF₁ = 25.44, and nonsignificant for participants with reversed or indiscriminate memory, ts < 1.34, ps > .18, d_zs < 0.06, BF₀s > 1.52. We observed a significant ME effect for participants with accurate memory for the 0-2 pair (M = 0.38, SD = 2.03), t(373) = 3.63, p < .001, d_z = 0.19, 95% CI = [0.17, 0.59], BF₁ = 146.28, but not for any of the other pairs, ts < 1.06, ps > .29, d_zs < 0.13, BF₀s > 4.52 (Table 1).

Experiment 3

Explicit rating ME scores revealed an ME effect (M = 0.09, SD = 1.84), t(2867) = 2.73, p = .006, d_z = 0.05, 95% CI = [0.03, 0.16], BF₁ = 5.66. An ANOVA on explicit rating ME scores that included Memory, Task Order, and Stimulus Pair as factors did not reveal any main or interaction effects, χ²s < 5.67, ps > .059. The ME effect was significant for participants with accurate memory (M = 0.13, SD = 1.85), t(2237) = 3.28, p = .001, d_z = 0.07, 95% CI = [0.05, 0.20], BF₁ = 31.77, and nonsignificant for participants with reversed or indiscriminate memory, ts < –0.03, ps > .51, d_zs < 0.00, BF₀s > 4.41. The ME effect for participants with accurate memory was significant for the 0-12 Stimulus Pair (M = 0.15, SD = 1.84), t(807) = 2.35, p = .019, d_z = 0.08, 95% CI = [0.02, 0.28], BF₁ = 8.10, and the 0-2 Stimulus Pair (M = 0.19, SD = 1.83), t(710) = 2.83, p = .005, d_z = 0.11, 95% CI = [0.06, 0.33], BF₁ = 12.43, but not for the 2-12 Stimulus Pair (M = 0.04, SD = 1.88), t(718) = 0.52, p = .61, d_z = 0.02, 95% CI = [–0.10, 0.17], BF₀ = 4.43.

Compound Analysis on Implicit and Explicit Evaluation

An ANOVA on standardized implicit and explicit evaluation ME scores (excluding EPT scores) that included Memory, Task Order, Implicit Evaluation Task, and Type of Evaluation Task (implicit/explicit) revealed a significant main effect of Memory, χ²(2) = 6.67, p = .036, and a main effect of Type of Evaluation Task, χ²(1) = 4.41, p = .036, indicating stronger ME effects on implicit than explicit evaluation tasks.⁵

Discussion

In three experiments, we examined ME effects on implicit stimulus evaluations. Results showed that the frequent and infrequent presentation of non-existing words can produce significant changes in implicit evaluations of these words. The ME effect on implicit evaluations, however, depended on a number of boundary conditions. First, it was dependent on the task that was used to capture implicit evaluations. We obtained strong evidence for an ME effect on IAT scores (BF₁s > 190) and AMP scores (BF₁ > 11) and substantial evidence that ME does not influence EPT scores (BF₀ > 21). Second, we obtained evidence that the ME effect depends on participants’ memory of the ME experience. We observed a robust ME effect on IAT and AMP scores only when participants had accurate memory of which stimulus had been presented more often in the ME task. Third, our results indicate that the ME effect depends on the number of stimulus presentations. That is, we only observed a robust ME effect on IAT or AMP scores when the frequent word was presented 2 or 12 times and the infrequent word was never presented (i.e., for the 0-2/0-12 stimulus pairs). When the infrequent word was presented, ME effects were smaller overall, and significant effects were only found when the infrequent word was presented only once and the frequent word was presented more than 6 times (i.e., for the 1-12 pair but not the 1-3, 1-6, or 2-12 pairs).

ME Influences Implicit Evaluation

The presence of an ME effect on implicit stimulus evaluations is important for a number of reasons. First, due to the generality of the ME effect, the widespread application of ME procedures for changing stimulus evaluation, and the relevance of implicit evaluation for behavior, this finding may have practical importance. Second, it also has important theoretical implications. More specifically, it contrasts with an important assumption of the processing fluency/attribution account of ME effects that these effects depend on the explicit measurement of evaluation. In Experiments 1 and 3, ME influenced IAT scores independent of whether the IAT was performed before or after participants provided their explicit liking ratings. In Experiment 2, ME influenced IAT scores even though the IAT always preceded the explicit rating task. In other words, participants showed evidence of an ME effect on IAT scores without first actively reporting on the quality of their explicit evaluations. These findings accord with the assumption of other theoretical accounts that ME leads to an immediate change in stimulus liking even when participants do not have the task to rate their liking of the crucial stimuli (e.g., the hedonic/fluency account, propositional accounts). Our results also show that ME can influence implicit evaluations despite the fact that stimuli were exposed many times during the IAT, EPT, and AMP procedures. To further explore this issue, we performed additional analyses which showed that (a) effects on explicit evaluations were observed even for participants who first completed implicit evaluation tasks and (b) ME effects were observed on AMP scores in Experiment 3 even when excluding the first block of AMP trials. This resilience to re-exposure accords with the idea that propositional knowledge of stimulus frequencies during the ME phase drives ME effects rather than fluency experiences that result from repeated exposure (and therefore should not survive the following exposures).

Our results also suggested that ME effects were stronger on implicit evaluations (except for the EPT effects) than on explicit evaluations. In accordance, we found evidence for a full mediation of ME effects on explicit rating scores by effects on IAT scores. At first glance, these results are in-line with previous theorizing that ME effects are the result of implicit processes (e.g., in the sense of unaware) and that such evaluations can be more easily probed with implicit evaluation measures (Kawakami, 2012). However, this does not fit with findings indicating an important role for memory for stimulus frequencies in ME effects. An alternative explanation is that explicit evaluation measures emphasize validation which might lead some participants to refrain from using the frequency of stimulus presentations as a basis for evaluation. For example, participants who learn that there is a difference in the number of stimulus frequencies might easily infer liking on the basis of this regularity. However, when asked to explicitly report their liking they might refrain from using this information because they do not consider it a good enough reason for changing liking. As we discuss below, this idea also accords with our findings regarding the moderating role of the number of stimulus exposures on ME effects on implicit and explicit evaluation. Note, however, that the overall difference in effect sizes between effects on implicit and explicit evaluations was very small and evidence for this difference in effect sizes was only anecdotal.

Moderators of ME Effects on Implicit (and Explicit) Evaluation

The observation that an ME effect on implicit evaluations was observed only under certain conditions also raises many interesting issues. First, the dissociation between ME effects on the EPT versus AMP and IAT could be due to the fact that effects on the EPT tend to be smaller and more unreliable than those on IAT and AMP (De Houwer et al., 2009; Wittenbrink, 2007).⁶ Hence, the EPT might simply have failed to capture implicit evaluations. In-line with this idea, correlational analyses revealed only a nonsignificant correlation between EPT scores and explicit rating scores. This finding is noteworthy because especially for novel stimuli, such as unfamiliar non-existing words, stronger correlations between implicit and explicit evaluations are typically observed (Nosek, 2005). Another possibility is that EPT procedures hamper the observation of ME effects specifically. In accordance, it has been observed that although other evaluative learning procedures that include novel stimuli can lead to reliable EPT effects (e.g., approach-avoid learning; Van Dessel, De Houwer, Gast, & Smith, 2015), ME instructions do not (see Van Dessel et al., 2017, for a discussion). One potential explanation might be that evaluative priming effects result from semantic relatedness between target and prime stimuli rather than evaluative congruency (see Werner, von Ramin, Spruyt, & Rothermund, 2018). Whereas other evaluative learning procedures might allow novel stimuli to become semantically related to EPT targets (e.g., a novel stimulus that is repeatedly avoided might become related to target words such as “Loss” or “Lonely” which might allow for an evaluative priming effect), this might not be the case for ME (e.g., because presentation frequency does not readily relate to any of the EPT targets). Although further research is needed to test such explanations, it is clear that the current results highlight the importance of using multiple implicit measures of attitudes to avoid equating implicit evaluations with any one measurement procedure.

Another important finding of our studies is that ME produced robust effects only for participants with accurate memory. This result is at odds with the proposal by Bornstein (1989) that memory of stimulus presentations is an important inhibitor of ME effects and reduces the size of ME effects. Rather, these results are in-line with those of earlier studies showing that the ME effect necessarily involves conscious awareness of (a) the stimulus presentations (Brooks & Watkins, 1989; de Zilva, Vu, Newell, & Pearson, 2013; Newell & Shanks, 2007; Szpunar, Schellenberg, & Pliner, 2004) and (b) the frequency of occurrence of the stimuli (Stafford & Grimes, 2012). Our results extend these findings by showing that recognition memory also moderates ME effects on implicit evaluations. This contrasts with the assumption that ME effects observed on implicit evaluation measures more strongly reflect fluency-based processes that do not depend on conscious knowledge of stimulus frequencies (e.g., Kawakami, 2012). Moreover, results strongly resemble findings that contingency awareness is a potent moderator of EC and approach-avoidance effects on implicit and explicit evaluation (see Hofmann, De Houwer, Perugini, Baeyens, & Crombez, 2010; Van Dessel, De Houwer, & Gast, 2016).

One notable exception is that we did not observe reversed ME effects for participants with reversed memory, whereas reversed effects have been observed in EC and approach-avoidance research (e.g., Van Dessel et al., 2016). Although this could be due to a lack of power to observe (typically smaller) reversed effects in our studies, it could also indicate that ME effects have different characteristics compared with EC and approach-avoidance effects (e.g., with regard to automaticity features). Indeed, our results do not provide definitive evidence that ME effects are nonautomatic (e.g., in the sense of controlled or conscious). For example, we measured participants’ memory of which of two stimuli occurs most often with a single question that followed the ME task and evaluation tasks. It is possible that a third variable determines both ME effects and participants’ answer to this question such that more robust ME effects are observed when participants have accurate memory. For example, fluency experiences or familiarity feelings might not only facilitate stimulus liking but also a higher frequency response in the memory test. Hence, it is at least possible that participants exhibited ME effects and accurate memory yet were unaware of the number of presentations during evaluation. Another possibility is that participants with accurate memory were more attentive or engaged in the experiment, and this moderated both ME effects and memory test performance. In contrast with this attention explanation, however, Experiment 3 found a moderation of AMP ME effects by memory even for participants who reported high ME task engagement. In addition, Wang and Chang (2004) found that participants preferred a stimulus they were not familiar with—but that they mistakenly classified as being familiar—over a stimulus that they had been exposed to before—but that they mistakenly classified as unfamiliar. Thus, the judged old/new status of a stimulus was more important to determine the liking of a stimulus than the actual previous exposure to the stimulus. This result suggests that memory is an important causal factor for the ME effect, rather than merely being correlated with it. The current results thus add to the cumulating evidence that memory of the stimuli and the stimulus presentations is an important precondition, rather than a limiting factor, for the ME effect (see also Montoya et al., 2017).

Interestingly, our results provided (anecdotal) evidence for an ME effect for participants with indiscriminate memory and substantial evidence for the absence of a difference in effect sizes between participants with accurate and indiscriminate memory. Although this might indicate that a proportion of the ME effect is not dependent on frequency memory, it could also reflect an ME effect for (a) participants who were able to retrieve frequency memory during the evaluation task but not the memory task or (b) participants who misinterpreted the memory task.

Finally, our results also revealed another important boundary condition of ME effects, that is, the number of stimulus exposures. ME effects were generally bigger when the infrequent stimulus was never presented. Moreover, for the 0-2 pair, we consistently found a robust ME effect implicit and explicit evaluation, but not for pairs with larger differences in the number of stimulus presentations (except for a relatively smaller ME effect for the 0-12 and 1-12 pairs). This downturn in the frequency-affect curve has been observed in other ME studies as well (Harrison, 1977; Zajonc et al., 1972; see Montoya et al., 2017, for a review). One popular explanation of this observation is that participants may engage in a correction process and consciously revise their initial evaluation when they become more strongly aware of the differences in occurrences (Gilbert, 1989; Trope, 1986). In-line with this idea, we obtained initial evidence that the negative impact of exposure frequency on ME effects depends on the task that is used to measure evaluations. More specifically, we found robust evidence for ME effects on implicit evaluations not only for the 0-2 pair but also for the 1-6, 1-12, and 0-12 pair. In contrast, we only found strong evidence for an effect on explicit evaluations for the 0-2 pair. One possible explanation for this dissociation is that for the 0-2 pair participants have less motivation to control against changes in liking because the frequent stimulus is only presented on two occasions (which might, for example, prevent boredom due to overexposure; see Van den Bergh & Vrana, 1998). The current results thus suggest that implicit evaluation measures can be an important addition to explicit evaluation measures in the context of ME effects in that they might sometimes capture ME effects that are not registered with explicit evaluation measures (e.g., due to controlled correction processes related to overexposure; see Kawakami et al., 2010, for corroborative evidence). This accords with recent evidence that certain learning procedures such as approach-avoidance training sometimes influence implicit but not explicit evaluation when participants do not consider the learned regularity a good basis for their evaluation (Van Dessel, De Houwer, Gast, Smith, & De Schryver, 2016).

Limitations

An important limitation of our studies is that they do not provide clarity about the reasons why ME effects on implicit evaluations were influenced by certain moderators such as the number of stimulus presentations or the nature of the implicit evaluation task. Our study was originally set up with a propositional account of ME effects in mind, for which we found initial support in a study that revealed effects of ME instructions on implicit evaluations (Van Dessel et al., 2017). It is noteworthy that the current results seem to mirror those of the ME instruction study. In that study, effects of ME instructions were observed only on implicit evaluations measured with an IAT and AMP and not with an EPT. Moreover, ME instructions influenced evaluations only when participants could correctly report which of the two words would occur most often, and ME instruction effects were bigger on implicit than on explicit evaluation measures. These similarities might be viewed as indirect support for the idea that ME effects (in part) depend on similar (propositional) mechanisms as ME instruction effects. On the other hand, the observed similarities between ME and ME instruction effects could also arise because the specific moderators influence the (distinct) mechanisms underlying ME and ME instruction effects to a similar extent.

Another important limitation is that we did not probe the relation between ME effects on implicit evaluation and real-world behavior. This requires further study especially given recent evidence that changes in implicit evaluations (as measured with the IAT) sometimes do not mediate changes in other relevant behavior (e.g., Forscher et al., 2017; but see Friese et al., 2009). Finally, it is important to note that all the observed ME effects were of small effect size (all d_zs < 0.41). Of course that was not unique to effects on implicit measures; our data suggest that, if anything, effects are larger on implicit measures than on explicit measures. Hence, though our experiments were well-powered to find an overall ME effect, they had less statistical power to find robust evidence for effects of the different between-subject factors and their interactions.

The current results thus provide many clues for future research that might look into ME (and ME instruction) effects on (implicit) evaluation and further test moderation by (a) the number of presentation frequencies, (b) memory, and (c) evaluation (implicit vs. explicit) measurement tasks (e.g., in different domains). These studies will not only allow us to gain more insight into the mental processes underlying ME effects but also implicit (and explicit) evaluation, memory, or human cognition in general. For example, research examining why ME does not influence EPT effects can help us understand the cognitive underpinnings of priming-related mental processes whereas research on dissociative effects of the number of presentation frequencies on implicit and explicit evaluation might provide information about the controllability of ME and the (automaticity of) processes underlying reactance responses.

Importantly, however, the present studies do allow us to already make at least two important new conclusions with a high degree of confidence: (a) ME procedures can influence implicit evaluations as measured with an IAT and AMP, and (b) this effect can occur even in the absence of explicit evaluation. This is important information that might shed new light on the mental mechanisms underlying ME (which has proven difficult so far; see Montoya et al., 2017). Moreover, the fact that our data raise many new questions is important because those questions are likely to stimulate new research. We therefore hope that our studies will provide the basis for many important future discoveries.

Supplemental Material

Van_Dessel_onlineappendix – Supplemental material for Mere Exposure Effects on Implicit Stimulus Evaluation: The Moderating Role of Evaluation Task, Number of Stimulus Presentations, and Memory for Presentation Frequency

Supplemental material, Van_Dessel_onlineappendix for Mere Exposure Effects on Implicit Stimulus Evaluation: The Moderating Role of Evaluation Task, Number of Stimulus Presentations, and Memory for Presentation Frequency by Pieter Van Dessel, Gaëtan Mertens, Colin Tucker Smith and Jan De Houwer in Personality and Social Psychology Bulletin

Footnotes

Author’s Note

Gaëtan Mertens and Pieter Van Dessel have contributed equally to the article.

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: Pieter Van Dessel is supported by a postdoctoral fellowship of the Scientific Research Foundation, Flanders (FWO-Vlaanderen). Jan De Houwer is supported by Methusalem Grant BOF16/MET_V/002 of Ghent University and by the Interuniversity Attraction Poles Program initiated by the Belgian Science Policy Office (IUAPVII/33).

Notes

Supplemental Material

Supplemental material is available online with this article.

References

Bar-Anan

De Houwer

Nosek

(2010). Evaluative conditioning and conscious knowledge of contingencies: A correlational investigation with large samples. The Quarterly Journal of Experimental Psychology, 63, 2313-2335.

Bates

Maechler

Bolker

Walker

(2014). lme4: Linear mixed-effects models using Eigen and S4. R package version 1.0-6. http://CRAN.R-project.org/package=lme4

Bornstein

R. F.

(1989). Exposure and affect: Overview and meta-analysis of research, 1968-1987. Psychological Bulletin, 106, 265-289.

Bornstein

R. F.

D’Agostino

P. R.

(1992). Stimulus recognition and the mere exposure effect. Journal of Personality and Social Psychology, 63, 545-552.

Bornstein

R. F.

D’Agostino

P. R.

(1994). The attribution and discounting of perceptual fluency: Preliminary tests of a perceptual fluency/attributional model of the mere exposure effect. Social Cognition, 12, 103-128.

Brooks

J. O.

Watkins

M. J.

(1989). Recognition memory and the mere exposure effect. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 968-976.

De Houwer

. (2014). A propositional model of implicit evaluation. Social and Personality Psychology Compass, 8, 342-353.

De Houwer

Moors

. (2010). Implicit measures: Similarities and differences. In Gawronski

Payne

K. B.

(Eds.), Handbook of implicit social cognition: Measurement, theory, and applications (pp. 176-193). New York, NY: Guilford Press.

De Houwer

Teige-Mocigemba

Spruyt

Moors

. (2009). Implicit measures: A normative analysis and review. Psychological Bulletin, 135, 347-368.

10.

de Zilva

Newell

B. R.

Pearson

. (2013). Exposure is not enough: Suppressing stimuli from awareness can abolish the mere exposure effect. PLoS ONE, 8(10), e77726.

11.

Dienes

(2011). Bayesian versus orthodox statistics: Which side are you on? Perspectives on Psychological Science, 6, 274-290.

12.

Dienes

(2016). How Bayes factors change scientific practice. Journal of Mathematical Psychology, 72, 78-89.

13.

Dovidio

J. F.

Kawakami

Gaertner

S. L.

(2002). Implicit and explicit prejudice and interracial interaction. Journal of Personality and Social Psychology, 82, 62-68.

14.

Eschenbeck

Heim-Dreger

Steinhilber

Kohlmann

C.-W.

(2016). Self-regulation of healthy nutrition: Automatic and controlled processes. BMC Psychology, 4, Article 4.

15.

Fazio

R. H.

Sanbonmatsu

D. M.

Powell

M. C.

Kardes

F. R.

(1986). On the automatic activation of attitudes. Journal of Personality and Social Psychology, 50, 229-238.

16.

Forscher

P. S.

Lai

Axt

Ebersole

C. R.

Herman

Devine

P. G.

Nosek

B. A.

(2017). A meta-analysis of change in implicit bias. Retrieved from https://osf.io/awz2p/

17.

Friese

Hofmann

Schmitt

(2009). When and why do implicit measures predict behaviour? Empirical evidence for the moderating role of opportunity, motivation, and process reliance. European Review of Social Psychology, 19, 285-338.

18.

Gilbert

D. T.

(1989). Thinking lightly about others: Automatic components of the social inference process. In Uleman

J. S.

Bargh

J. A.

(Eds.), Unintended thought (pp. 189-211). New York, NY: Guilford Press.

19.

Greenwald

A. G.

McGhee

D. E.

Schwartz

J. L. K.

(1998). Measuring individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 74, 1464-1480.

20.

Greenwald

A. G.

Nosek

a Banaji

M. R.

(2003). Understanding and using the implicit association test: I. An improved scoring algorithm. Journal of Personality and Social Psychology, 85, 197-216.

21.

Harmon-Jones

Allen

J. B.

(2001). The role of affect in the mere exposure effect: Evidence from psychophysiological and individual differences approaches. Personality and Social Psychology Bulletin, 27, 889-898.

22.

Harrison

A. A.

(1977). Mere exposure. In Berkowitz

(Ed.), Advances in experimental social psychology (pp. 39-83). San Diego, CA: Academic Press.

23.

Helton

W. S.

Näswall

(2015). Short Stress State Questionnaire: Factor structure and state change assessment. European Journal of Psychological Assessment, 31(1), 20-30.doi:10.1027/1015-5759/a000200

24.

Hicks

J. A.

King

L. A.

(2011). Subliminal mere exposure and explicit and implicit positive affective responses. Cognition & Emotion, 25, 726-729.

25.

Hofmann

De Houwer

Perugini

Baeyens

Crombez

(2010). Evaluative conditioning in humans: A meta-analysis. Psychological Bulletin, 136, 390-421.

26.

Hofmann

Gawronski

Gschwendner

Schmitt

(2005). A meta-analysis on the correlation between the implicit association test and explicit self-report measures. Personality and Social Psychology Bulletin, 31, 1369-1385.

27.

Janiszewski

(1993). Preattentive mere exposure effects. Journal of Consumer Research, 20, 376-392.

28.

Jeffreys

(1961). Theory of probability. Oxford, UK: Oxford University Press.

29.

Kawakami

(2012). The implicit influence of a negative mood on the subliminal mere exposure effect. Perceptual and Motor Skills, 115, 715-724.

30.

Kawakami

Sato

Yoshida

(2010). Effects of mere exposure on category evaluation measured by the IAT and the GNAT. Shinrigaku Kenkyu, 81, 437-445.

31.

Kawakami

Yoshida

(2015). How do implicit effects of subliminal mere exposure become explicit? Mediating effects of social interaction. Social Influence, 10, 43-54.

32.

Mandler

Nakamura

Van Zandt

B. J.

(1987). Nonspecific effects of exposure on stimuli that cannot be recognized. Journal of Experimental Psychology: Learning, Memory, and Cognition, 13, 646-648.

33.

Monahan

J. L.

Murphy

S. T.

Zajonc

R. B.

(2000). Subliminal mere exposure: Specific, general, and diffuse effects. Psychological Science, 11, 462-466.

34.

Montoya

R. M.

Horton

R. S.

Vevea

J. L.

Citkowicz

Lauber

E. A.

(2017). A re-examination of the mere exposure effect: The influence of repeated exposure on recognition, familiarity, and liking. Psychological Bulletin, 143, 459-498.

35.

Newell

B. R. B.

Shanks

D. D. R.

(2007). Recognising what you like: Examining the relation between the mere-exposure effect and recognition. European Journal of Cognitive Psychology, 19, 103-118.

36.

Nosek

B. A.

(2005). Moderators of the relationship between implicit and explicit evaluation. Journal of Experimental Psychology: General, 134, 565-584.

37.

Olson

M. A.

Fazio

R. H.

(2004). Reducing the influence of extrapersonal associations on the implicit association test: Personalizing the IAT. Journal of Personality and Social Psychology, 86, 653-667.

38.

Payne

B. K.

Cheng

C. M.

Govorun

Stewart

B. D.

(2005). An inkblot for attitudes: Affect misattribution as implicit measurement. Journal of Personality and Social Psychology, 89, 277-293.

39.

Pliner

(1982). The effects of mere exposure on liking for edible substances. Appetite, 3, 283-290.

40.

Rothermund

Wentura

(2004). Underlying processes in the implicit association test: Dissociating salience from associations. Journal of Experimental Psychology: General, 133, 139-165.

41.

Simmons

J. P.

Nelson

L. D.

Simonsohn

(2011). False-positive psychology: Undisclosed flexibility in data collection and analysis allow presenting anything as significant. Psychological Science, 22, 1359-1366.

42.

Smith

C. T.

De Houwer

Nosek

B. A.

(2013). Consider the source: Persuasion of implicit evaluations is moderated by source credibility. Personality and Social Psychology Bulletin, 39, 193-205.

43.

Smith

P. K.

Dijksterhuis

Chaiken

(2008). Subliminal exposure to faces and racial attitudes: Exposure to Whites makes Whites like Blacks less. Journal of Experimental Social Psychology, 441, 50-64.

44.

Spruyt

De Houwer

Hermans

Eelen

(2007). Affective priming of nonaffective semantic categorization responses. Experimental Psychology, 54, 44-53.

45.

Stafford

Grimes

(2012). Memory enhances the mere exposure effect. Psychology & Marketing, 29, 995-1003.

46.

Szpunar

K. K.

Schellenberg

E. G.

Pliner

(2004). Liking and memory for musical stimuli as a function of exposure. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30, 370-381.

47.

Trope

(1986). Identification and inferential processes in dispositional attribution. Psychological Review, 93, 239-257.

48.

Vadillo

M. A.

Konstantinidis

Shanks

D. R.

(2016). Underpowered samples, false negatives, and unconscious learning. Psychonomic Bulletin & Review, 23, 87-102.

49.

Van den Bergh

Vrana

S. R

. (1998). Repetition and boredom in a perceptual fluency/attributional model of affective judgements. Cognition & Emotion, 12, 533-553.

50.

Van Dessel

De Houwer

Gast

. (2016). Approach-avoidance training effects are moderated by awareness of stimulus-action contingencies. Personality and Social Psychology Bulletin, 42, 81-93.

51.

Van Dessel

De Houwer

Gast

Smith

C. T

. (2015). Instruction-based approach-avoidance effects. Experimental Psychology, 62, 161-169.

52.

Van Dessel

De Houwer

Gast

Smith

C. T.

De Schryver

. (2016). Instructing implicit processes: When instructions to approach or avoid influence implicit but not explicit evaluation. Journal of Experimental Social Psychology, 63, 1-9.

53.

Van Dessel

De Houwer

Roets

Gast

. (2016). Failures to change stimulus evaluations by means of subliminal approach and avoidance training. Journal of Personality and Social Psychology, 110, e1-e15.

54.

Van Dessel

Hughes

De Houwer

(2018). How Do Actions Influence Attitudes? An Inferential Account of the Impact of Action Performance on Stimulus Evaluation. Manuscript invited for revision at Personality and Social Psychology Review. Prepint available at: https://osf.io/kb3wq/

55.

Van Dessel

Mertens

Smith

C. T.

De Houwer

. (2017). The mere exposure instruction effect: Mere exposure instructions influence liking. Experimental Psychology, 64, 299-314.

56.

Wagenmakers

E.-J.

(2007). A practical solution to the pervasive problems of p values. Psychonomic Bulletin & Review, 14, 779-804.

57.

Wang

M.-Y.

Chang

H.-C.

(2004). The mere exposure effect and recognition memory. Cognition & Emotion, 18, 1055-1078.

58.

Werner

von Ramin

Spruyt

Rothermund

(2018). Does sunshine prime loyal . . . or summer? Effects of associative relatedness on the evaluative priming effect in the valent/neutral categorisation task. Cognition & Emotion, 32, 222-230.

59.

Whittlesea

B. W. A.

(1993). Illusions of familiarity. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 1235-1253.

60.

Winkielman

Cacioppo

J. T.

(2001). Mind at ease puts a smile on the face: Psychophysiological evidence that processing facilitation elicits positive affect. Journal of Personality and Social Psychology, 81, 989-1000.

61.

Winkielman

Halberstadt

Fazendeiro

Catty

(2006). Prototypes are attractive because they are easy on the mind. Psychological Science, 17, 799-807.

62.

Wittenbrink

(2007). Measuring attitudes through priming. In Wittenbrink

Schwarz

(Eds.), Implicit measures of attitudes: Progress and controversies (pp. 17-58). New York, NY: Guilford Press.

63.

Witvliet

Vrana

(2007). Play it again Sam: Repeated exposure to emotionally evocative music polarizes liking and smiling responses, and influences other affect reports, facial EMG, and heart rate. Cognition & Emotion, 21, 3-25.

64.

Zajonc

R. B.

(1968). Attitudinal effects of mere exposure. Journal of Personality and Social Psychology, 9(2 Pt. 2), 1-27.

65.

Zajonc

R. B.

(2001). Mere exposure: A gateway to the subliminal. Current Directions in Psychological Science, 10, 224-228.

66.

Zajonc

R. B.

Shaver

Tavris

van Kreveld

(1972). Exposure, satiation, and stimulus discriminability. Journal of Personality and Social Psychology, 21, 270-280.

67.

Zebrowitz

L. A.

White

Wieneke

(2008). Mere exposure and racial prejudice: Exposure to other-race faces increases liking for strangers of that race. Social Cognition, 26, 259-275.

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

0.23 MB