Acta Psychologica 230 (2022) 103745 Contents lists available at ScienceDirect Acta Psychologica journal homepage: www.elsevier.com/locate/actpsy The attentional cost of comparisons: Evidence for a general comparison induced delay☆ Paul Barker a, Ron Dotsch b, Roland Imhoff a,c,* a Social Cognition Center Cologne, University of Cologne, Germany b Department of Psychology, Utrecht University, the Netherlands c Social and Legal Psychology, Johannes Gutenberg University Mainz, Germany A R T I C L E I N F O A B S T R A C T Keywords: The current work aimed to uncover the pattern of attention given to external comparison standards when Social comparison engaged in social judgments. In a series of 5 experiments (N = 463), a Modified Spatial Cueing Task provided Visuospatial attention evidence for a general Comparison Induced Delay (CID), but found no signs of visuospatial attention (Pilot, Study Spatial cueing 1 & 2). However, the CID did not occur if cues did not remain visually available throughout the trials (Study 3 & 4). Heterogeneity in results prompted the use of a single-paper meta-analysis including all secondary studies. A consistent CID effect was found across studies when standards remained visually available (K = 5), but not when they were masked (K = 2). No direct signs of visuospatial attentional bias were found. These results suggest that the attentional cost of engaging with external comparisons is mainly cognitive in nature, although a minor reoccurring visual component could not be excluded. 1. Introduction comparative in nature; either relying on internally held standards or external ones found in the current social environment (Kahneman & Whenever we meet someone, we instantly judge that person on a Miller, 1986; Mussweiler, 2003). For instance, knowing a student can solve variety of variables. The basis for answering these types of questions five math problems is not very diagnostic for a judgment of intelligence relies heavily on the selection of a standard to which we can compare until we observe his peers solving only two (an external standard), or if we our new acquaintance. When selecting these standards and forming a have prior knowledge that a good performance means solving at least three comparative judgment, some attentional resources are likely diverted to problems (internal standard). Furthermore, these standards can either be this process. These may include both the reallocation of cognitive re- higher than the target on the dimension of interest (upward comparisons), sources involved in the processing of the comparative information, but for instance when one judges an amateur to the standard of a professional, could also include visuospatial biases in the form of the initial inspection or can be in the opposite direction (downward comparison), when the and potential ongoing visual adhesion to the standards. The current professional is judged in comparison to the amateur. work will investigate this potential attentional cost of engaging with The largest amount of work in the social comparison literature has external comparison standards, and attempts to disentangle the visuo- focussed on understanding the effects that these upward and downward spatial and cognitive aspects. standards have on various comparative outcomes; such as ability esti- When making judgments about the self or others people often spon- mates, affect and self-esteem (for a recent meta-analysis of these effects taneously use some social standard as a comparison (Dunning & Hayes, see Gerber et al., 2018). Comparatively less work has investigated the 1996; Festinger, 1954). In fact, this process of comparing is such a attentional cost of engaging with these comparison standards, despite fundamental part of human cognition that presented comparison stan- the fact that this attention is likely an essential part of the unfolding of dards are used even when they are known to be irrelevant (Gilbert et al., the comparison process as a whole. 1995) or presented subliminally (Mussweiler et al., 2004a). This has led to Although internal standards are used extensively in the form of the assertion that most, if not all, social judgments are to some degree routine standards (Mussweiler, 2003) and egocentric judgments ☆ This project was supported by a DFG Grant (IM147/3-1) awarded to Roland Imhoff as part of the DFG Research Unit “Relativity in Social Cognition” (FOR 2150). The funding source was not involved during the execution of the project.Data files and additional materials can be found on the Open Science Framework page at: htt ps://osf.io/hkem9/ * Corresponding author at: Social and Legal Psychology, Johannes Gutenberg University Mainz, Binger Str. 14 - 16, 55122, Mainz, Germany. E-mail addresses: paul.l.barker@gmail.com (P. Barker), roland.imhoff@uni-mainz.de (R. Imhoff). https://doi.org/10.1016/j.actpsy.2022.103745 Received 27 April 2021; Received in revised form 14 August 2022; Accepted 12 September 2022 Available online 26 September 2022 0001-6918/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/). P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 (Dunning & Hayes, 1996), they are difficult to investigate directly, engaged, comparison standards should become a means to reach the especially in regards to the attentional cost they may cause. With only goal of an accurate assessment. As a result, pertinent comparison cues very distant measures capable of measuring the internal comparison that are presented subsequent to a social judgment task will gain a process, past investigations have resorted to measuring the efficiency in particular attentional advantage similar to those found for other goal- making subsequent comparative judgments in relation to the strength of relevant stimuli in countless other domains (Vogt et al., 2010; e.g. the assimilative effect resulting from a primed internal standard Hunger and food; Tapper et al., 2010; and Mate selection; Maner et al., (Mussweiler & Damisch, 2008). The results, thereby, rely heavily on 2007). These effects can be exploited by instructing individuals to esti- theoretical assumptions and can only offer distant estimates of the mate a judgment target's standing on a specific dimension in absolute cognitive engagement with the recruited standards. terms yet delaying the responses until after a pertinent comparison Investigations using external standards, on the other hand, can standard has been presented as a cue in the spatial cueing trial. This way directly measure the attention given to these comparison standards and the attention given to comparison standards can be assessed while do not need to rely on specific theoretical assumptions from one of the participants are still fully engaged in a comparison related judgment. many social comparison theories (e.g. SAM: Mussweiler, 2003; I/E Specifically, visuospatial attention given to the comparison standards Model: Bless & Schwarz, 2010). Instead, one needs only to assume that would be reflected in greater delays in completing the task during this process must logically start with at least minimal early visuospatial invalidly cued trials than validly cued ones when a pertinent comparison attention; in so far that one must direct their gaze to a stimulus in order standard is presented as the cue. An overall delay in responding, to assess its standing on the relevant dimension and to extract this in- regardless of cue position, when a comparison standard is presented formation for the act of comparison. For this reason, the current work would reflect non-visuospatial attention caused by the binding of will limit its investigation mainly to the attentional cost of engaging cognitive resources (for a differentiation of these two processes in the with these external comparison standards, although it acknowledges domain of erotic stimuli, see Imhoff et al. (2019a)). Thus, such an that internal standards can still play a role in any comparative judg- adjusted MSCT procedure should be able to disentangle the visuospatial ments and may affect response patterns. attention (the visual fixation on the comparison standard cue) and non- The scarce previous work that has investigated the time course of the visuospatial attention (reflecting cognitive preoccupation with the comparative process has shown that the first sign of cognitive engage- comparison information and process) and clarify the overall pattern of ment with external standards happens very early in the processing of attention to comparison standards. information (Ohmann et al., 2016). Therefore, comparative information seems to be encoded quickly and efficiently, suggesting any attentional 1.1. The present research bias is relatively automatic in nature. However, this work only required participants to indicate which of two simultaneously presented stimuli In order to investigate the pattern of attention given to external com- reflected a certain dimension to a greater degree, but did not request any parison standards and their informational content, the current work will absolute judgment of either on this judgment dimensions. Therefore, the use an adjusted MSCT task to measure the overall attentional cost of processes investigated in this work may be limited to the most funda- engaging in social comparative judgments (Pilot study, Study 1 and 2) and mental initial steps in the entire comparison process, i.e. assessing the attempt to disentangle visuospatial attention from non-visuospatial informational content of two stimuli and ranking them. Subsequent cognitive processes (Study 3 and 4). If comparison standards elicit visuo- integration of the comparative information in a way that ultimately spatial attentional adhesion (more or less deliberate continuation of leads to shifts in evaluative judgments might still require additional attention to comparison cues), we would expect longer latencies for attentional resources of some kind. This may be in the form of prolonged comparison relevant vs. irrelevant cues during invalidly cued trials, but visuospatial attentional adhesion to the external stimuli, or may be an not on validly cued trials. It is also conceivable that the effects on visuo- internal process that requires non-visuospatial resources, such as the spatial attention are relatively minor and short-lived, but that non- occupation of working memory capacity while integrating the extracted visuospatial attention related to the internal social comparison processes information into the final judgment. nevertheless results in a general cognitive delay related to the cognitive The many established measures of automatic attention can offer engagement with the comparison information. Such a comparison- some guidance regarding how to disentangle these two possible forms of induced delay (CID) might present itself in a general slowing of response attention, as they have been successfully implemented in a diverse set of times regardless of cue location when engaging in a social judgment task fields (for an overview see Fox et al., 2011). One of the best-established rather than a control task when pertinent standards are present. behavioral measures of automatic attention allocation is the Modified Throughout the presented studies, we report how we determined our Spatial Cueing Task (MSCT; developed by Posner et al., 1980; modified sample size, all data exclusions (if any), all manipulations, and all by Fox et al., 2001) and would seemingly provide an excellent basis for measures in the studies. In addition to the studies reported here directly, the development of a comparison focused paradigm. The MSCT manages an additional two studies were run as part of this project. These addi- to dissociate the reaction to relevant and irrelevant cues by presenting tional studies are described in detail in the additional materials and are them in separate trials as opposed to simultaneously (Fox et al., 2001). included in the final summative meta-analyses. Furthermore, all ano- The basic task involves identifying a neutral stimulus (probe target) that nymised raw1 and aggregated data and additional materials for all varies in its spatial location on the screen. Before the onset of the probe, studies can be found on the Open Science Framework page at: htt a supposedly irrelevant cue appears at either the same (valid cue) or the ps://osf.io/hkem9/ opposite side of the screen (invalid cue). Delays in response times for invalidly cued trials are considered indicative of automatic visuospatial 2. Pilot test attention given to the cue stimuli (Fox et al., 2011). To adapt this paradigm for the purpose of estimating attention to This initial pilot test implemented a first version of the modified comparison standards, a few adjustments are necessary since the cues spatial cueing paradigm to test the procedure and its general ability to themselves, which will form the comparison standards, are only one part detect social comparison related bias. More specifically, this study of the comparison process. The other half of the comparison relates to the target of the judgment that needs to be judged on a specific dimension. Indeed, simply asking participants to make an absolute 1 The variables of study area, gender and timestamps are not included in judgment about such a judgment target engages comparative process- these data to ensure the complete anonymity of participants, but are available ing, even when standards are presented subliminally and without upon request. In addition, the visual stimuli used in the studies are also not explicit prompting (Mussweiler et al., 2004a). When this process is provided publicly due to copyright concerns, but can be requested. 2 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 aimed to assess if latencies on the cueing task would be higher during the numerical response to the initial judgment question. presentation of social cues, compared to non-social cues, when partici- In this series of events, the social judgment task should induce social pants were engaged in a social judgment rather than non-social judg- comparative thinking resulting in the relevant social cues presented ment task. This was expected to be reflected in an interaction effect during the spatial cuing phase gaining an initial goal-relevant atten- between task type and cue type. Additionally, a three-way interaction tional advantage, possibly followed by later cognitive preoccupation also including the validity of the cue would additionally indicate vi- while the information is incorporated into the final judgment. Non- suospatial attentional bias specifically. social cues should not show this advantage for these trials. However, for the non-social task the non-social cues may gain such an advantage, 2.1. Method while here the social cues do not have any increased relevance though a slight general bias towards social cues may be present across trials. 2.1.1. Participants Twelve trials were run for each judgment condition (social vs. non- As no estimates of the size of the possible effects were known, this social judgment), for social and non-social cues (social vs. non-social pilot study recruited 59 German speakers on campus at the University of cues), and for upward or downward comparison standard (up vs. Cologne. Sensitivity analysis showed this sample was enough to detect a down), totalling 96 trials. Validly and invalidly cued trials each simple effect in a repeated measures design 80 % of the time with α 0.05 appeared in half the trials for each participant, as was the case for the = for effect sizes of η2 0.123 (determined in G*Power; Faul et al., 2007).2 probe position (left or right) and the probe identity (u or n). To ensure all p > Participants gave their informed consent after their rights and the combinations of factors occurred equally across the study these factors data collection practices had been explained. The final sample consisted were also counterbalanced between participants. of 28.8 % female participants and was aged between 19 and 39 years (M 25.16, SD 4.63). 2.1.3. Stimuli = = Twelve social judgments items such as “How much weight can this 2.1.2. Design man lift (in kg)?”, and twelve non-social items like, “How fast is this boat The software package Inquisit 3 was used in the setup and running of (in km/h)?” were created for this pilot study. These items were all studies. The main task consisted of trials made of three parts. The accompanied by 48 relevant neutral social and non-social images, and initial part of the trial included the presentation of an image of a social 48 comparison cues representing both upward and downward standards judgment target (e.g., men and women) or non-social judgment target in equal numbers. All images were selected off the internet to have clear (e.g., boats or cars), which participants were required to judge on an relevance to the judgment dimensions. open-ended question with a numerical estimate. An example of a social As previous research has noted individual differences in the extent to judgment item would be “How much weight can this man lift (in kg)?”, which people rely on comparative information, we explored this possi- while a non-social one was “How fast is this boat (in km/h)?”. Impor- bility by including the Iowa-Netherlands Comparison Orientation Scale tantly, these questions were not answered in this first step, but only at (INCOM; Gibbons & Buunk, 1999) consistent of 11 items (α = 0.85) to the end of the whole trial. assess individual differences in social comparison orientation. The second part of the trial consisted of a Spatial Cueing Task (Fox et al., 2001; Posner et al., 1980) in a modified form. This Modified 2.1.4. Procedure Spatial Cuing Task (MSCT) required participants to rapidly identify a Upon arrival, participants were sat in separate cubicles in front of a neutral stimulus, referred to as the “probe”, which was either randomly computer in a laboratory setting. Before starting the study, the general presented on the left or the right side of the screen. Responses were procedure of the study and data storage policy was explained to par- given by using the corresponding button on the keyboard, in this case ticipants prior to the signing of an informed consent form. Following the letters “u” and “n” were presented as a neutral probe stimulus to this, the general demographics, such as Sex, Age and Education level make position matching interferences less likely due to the vertically were recorded at the beginning of the study. In order to allow partici- inverse shape and location of these characters on a standard keyboard pants to become familiar with the main task, a detailed explanation was (for the superiority of identification over locations tasks see Imhoff et al. given followed by four practice trials. Participants were then given the (2019b)). Slightly before this probe appeared, a distracting cue was opportunity to ask any further questions regarding the task. With no presented on either the same side (Valid cue), or at the opposite side further questions, participants continued on to the main batch of 96 (Invalid cue) of the screen slightly beside where the probe was to appear. trials, which followed in random order. Finally, at the end of these trials These distractor cues comprised either social or non-social cues that the INCOM was administered, after which participants were debriefed represented upward or downward standards on the relevant dimension. and given their compensation. For instance, a picture of a bodybuilder might be presented as a social cue when the strength of the social judgment target was being consid- 2.1.5. Data treatment ered. Alternatively a rowing boat may be presented as a non-social cue in Trials with erroneous probe detections were removed (2.8 %), and the case of a non-social judgment about a boat's speed. Participants were the remaining trials were truncated under 200 ms (2 trials) and above explicitly instructed to ignore these stimuli and focus solely on the quick the Tukey criterion (1763 ms; 6.7 %). 3 This resulted in a total of 9.5 % of and accurate identification of the probe. In the current study the slight trials being removed. For use in the final analyses, these latencies were delay between the appearance of the cue and the probe, i.e., the stimulus then log transformed. Furthermore, using the raw truncated scores, onset asynchrony (SOA), was set to 250 ms to coincide with expected various difference scores were calculated for the social judgment con- early attentional processes. Both the cue and the probe remained dition and the non-social judgment control condition in an effort to gain onscreen until a response was given, see Fig. 1 for the layout of the indicators of individual differences in attentional bias not influenced by screen at the end of the trial. After giving their response in the spatial general interpersonal reaction speed variations. These scores, methods cueing phase, participants were asked at the end of the trial to give a of calculation, and descriptions are presented in Table 1. 2 Regretfully, the wrong version of η2 p was selected in the a priori power analysis (conducted in G*Power; Faul et al., 2007). This error was brought to our attention only after the project had concluded. Therefore, sensitivity ana- lyses are presented in this section as post-hoc power analyses are uninformative 3 This outlier criterion limit is calculated as 1.5 times the value range be- (Hoenig & Heisey, 2001). tween the first and third quartile above the 3rd, or Q3 + 1.5(Q3-Q1). 3 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 Fig. 1. Layout of a trial at the point of response, with the question remaining at the top of the page. The probe ‘u’ and cue are depicted on the left and right side respectively in this invalid trial. 2.2. Results current work will not investigate this effect any further, a speculative explanation for this may lie in the differential familiarity of the upward A 2 (Task type: Social vs Non-social) × 2 (Cue type: Social vs Non- and downward objects, which has also been found to influence social) × 2 (Comparison direction: Upwards vs Downwards) × 2 (Cue comparative judgments themselves (Häfner, 2009). validity: Valid vs Invalid) repeated measures ANOVA was used to analyse the data. As such, participants who did not provide any usable 2.2.1. Correlations trials, as per the exclusion criteria, for at least one of the factor levels The social and non-social Validity difference scores, Cue difference were not included in the analyse. scores, CA, and CID were used in a correlational analysis with the The presence of social cues compared to non-social cues significantly INCOM, age and sex variables. The Validity difference scores for com- increased the time to respond to the probe overall, F(1, 56) = 12.94, p = parison trials were found to correlate positively with sex, r = 0.273, p = .001, η2 p = 0.188, 90 % CI [0.056, 0.328]. Furthermore, in line with the .04, 95 % CI [0.013, 0.498], meaning being male compared to being expectations, this effect was significantly moderated by the task subjects female was associated with longer latencies for invalid compared to were engaged in, F(1, 56) = 6.90, p = .011, η2 p = 0.110, 90 % CI [0.014, valid cues when engaging in a comparison trials in these data. No other 0.243]. Only social comparisons were associated with a significant in- measures were significantly associated with any of the scales, r < 0.26. crease in reaction times in the presence of social cues (ΔM = 0.053, SE = 0.012, p < .001, 95 % CI [0.030, 0.076]), but no such difference appeared 2.3. Discussion for non-social comparisons (ΔM = 0.012, SE = 0.012, p = .278, 95 % CI [− 0.011, 0.038]). This suggests the presence of attentional bias resulting The results of this pilot study suggest that there is indeed attentional from a unique social comparison process that does not occur for non- bias towards social cues when engaging in social comparisons, which social comparisons. The expected 3-way interaction between task type, does not extend to comparisons involving non-social judgments and cue type and cue validity did not reach significance, however, F(1, 56) = relevant non-social cues. However, the presence of visuospatial atten- 1.13, p = .293, η2 p =0.020, 90 % CI [0.000, 0.112], meaning there were no tion to social comparison standards specifically was not detected. This signs of evidence for visuospatial attentional bias in these data. could be due to the limited scope of the pilot study, in terms of the The validity of the cue presentation did show a significant main ef- number of trials and stimuli that were included, which may have limited fect, F(1, 56) = 300.17, p < .001, η2 p = 0.843, 90 % CI [0.775, 0.878], the power of the analyses to find these more subtle effects. with invalid cues leading to longer reaction times. Lastly, task type also Furthermore, although the evidence in the current study does not showed a significant interaction with the direction of the comparison, F suggest that non-social comparisons induce a similar bias, this is in (1, 56) = 13.43, p = .001, η2 p = 0.193, 90 % CI [0.059, 0.334], as well as contrast to previous work which indicated that the processing of basic the related three-way interaction including cue type, F(1, 56) = 8.82, p social and non-social comparison information unfolded at similar speeds = .004, η2 p = 0.136, 90 % CI [0.026, 0.273]. Closer inspection showed (Ohmann et al., 2016) despite the more complex nature of social stimuli. that this effect of direction was limited to the non-social cues in a way Therefore, some form of comparative processing could still be present in where in social judgment trials upward non-social cues increased reac- the control condition. In addition, some judgment dimension may apply tion time, (ΔM = 0.048, SE = 0.013, p = .001, 95 % CI [0.021, 0.075]), to both social and non-social targets like speed for instance, which one while the opposite was the case for non-social judgments, (ΔM = might compare to non-social targets like animals or vehicles as well as − 0.045, SE = 0.017, p = .009, 95 % CI [− 0.079, − 0.012]). Although the other humans. Thus, it is possible that some comparative thinking could Table 1 The names, calculation methods and descriptions of the various difference scores that were calculated in the pilot study. Score name Calculation method Description For the social judgment condition and the control condition separately: Validity difference Socially cued invalid trials – Socially cued valid trials Reflect the amount of visuospatial attention granted to social cues with higher scores score indicating more visuospatial attention Cue difference scores Social cue trials – non-social cue trials Reflect the general amount of attentional bias caused by social cues vs. non-social cues regardless of validity Composite scores: Comparison attention Validity difference score in the Social judgment condition Capture increased visuospatial attentional bias for trials where social comparison information (CA) scores – Validity difference score in the Control condition is task relevant compared vs. not. Large positive scores indicate large bias Comparison induced Cue difference score in the social judgment condition – Captures the overall bias towards the social cue when engaging in social judgments compared delay (CID) Cue difference score in the Control condition to the control condition. Large positive scores indicate large bias in the social condition. 4 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 be present even when a social judgment is being made and a non-social also attempted to measure theoretically predicted shifts in judgment cue is present. Taken together these possibilities could cause the effect of estimates, assimilation or contrast, in the response patterns for the social comparative bias to appear smaller in comparison to the control con- judgments trials. In order to promote both consistent assimilation and dition. For these reasons, the MSCT used in the subsequent studies contrast in these responses, the extremity of the social cues were replaced non-social comparisons with a memory task and increased manipulated in addition to the direction of the standards as this has been power by including additional trials and stimuli. found to moderate the direction of the comparative outcomes (e.g. Mussweiler et al., 2004b). However, as these results are not the main 3. Study 1 focus and they did not show the theoretically expected patterns, they will not be described here in detail but can be found with the additional This study aimed to investigate whether the presence of relevant studies online (for a recent critical discussion regarding the heterogeneity comparison targets indeed biases visuospatial attention during a social of these effects across judgments and contexts, see Barker et al. (2020)). judgment task within the MSCT paradigm compared to a non- For this extensive design, a total of 320 trials were run including twenty comparative control task. More specifically, the expectation was that trials for each combination of the judgment task (social judgment vs. the invalidly cued stimuli would increase response latencies more than memory), cue direction (up vs. down), cue extremity (moderate vs. validly cued ones when participants were engaged in a social judgment extreme), and its validity (valid vs. invalid). The probe presentation (left or task compared to when they were engaged in a memory task. right) and probe identity (u or n) were presented in equal numbers to each participant and were also counterbalanced between participants so that each pairing was presented for each factor level in equal numbers overall. 3.1. Method 3.1.3. Stimuli 3.1.1. Participants A series of 30 questions were generated relating to a variety of di- A sample size of 166 German speakers were recruited on campus at the mensions and relevant behaviours and attributes. For each of these University of Cologne to participate in the study for a monetary reward of 6 2 questions 20 images were collected from various online sources that re-euros. This sample size is sufficient to detect an effect size of ηp > 0.046 flected the dimension of judgment. This collection of images was created with α = 0.05 and β = 0.20 according to a sensitivity analysis (determined with a large apparent variation in the extent to which the social target in G*Power; Faul et al., 2007). Furthermore, this sample size should pro- expressed traits related to the dimension of judgment. A pre-test was then vide reasonable stability for the exploratory correlational analyses, r > 4 conducted with 123 participants recruited at the University of Cologne, 0.20, w = 0.15, α = 0.05, β = 0.10 (see Schönbrodt & Perugini, 2013). who consisted of 49.2 % females and were aged between 18 and 50 years Prior to their participation, respondents' consent was obtained after (M = 24.97, SD = 6.48). Each participant was required to respond to the being informed about data collection and storage practices and their questions for a fourth of all the images, amounting to about 30 judgments right to withdraw from the experiment without consequence. The final per image. The ease of making the judgment and the clarity of the image sample in this study was 68.7 % female and was aged between 16 and 62 5 were also measured. Guided by these data, 20 questions were retained years (M = 24.17, SD = 5.72). based on their average ease of judgment and clarity scores. In addition, the presence of low intra-image variability (similar judgments of the 3.1.2. Design image on the relevant dimension) and high intra-question variability The main task again consisted of trials with three parts. However, the (different images representing a variety of extremity levels) were also initial part of the trial now only included the presentation of an image of considered when selecting the questions. This was done to ensure clear a social target, accompanied by one of two tasks. One was an open- images within the question groups that ranged substantially on, and re- ended question that required a numerical judgment about the target's flected distinct values of the judgment dimension. behaviours or skills (Social judgment task); while the other instructed Within these question groups, the four lowest scoring images on the participants to simply memorise the image (Memory task), see Fig. 2 for clarity and judgability scale were removed, leaving the 16 top rated an overview of possible trial sequences. The social judgment trials were images per question group. The eight images closest to the dimension similar to those used in the pilot study (see section 3.1.3 for a more average were used as the initial neutral judgment target in the first part detailed report), while the memory prompt was simply the instruction to of each trail, whereas the four highest and four lowest were selected to “Remember the image”. be used as upward and downward cues respectively within the MSCT. The second part of the trial was again an MSCT with an SOA of 250 These cues were further divided with the two highest and lowest judged ms, similar to that used in the pilot study. However, only relevant social images in each question group forming the extreme standards, while the stimuli were used as distractor cues in this investigation. After four left over cues formed the moderate standards. completing the spatial cueing phase, participants gave a numerical response to the initial social judgment question or, in the case of a 3.1.4. Procedure memory trial, were asked a specific question about the previously pre- Other than the increase in trials, the procedure was identical to the sented image, for instance “How many bags was the woman carrying?”. one used in the pilot study. As a result of this procedure, one would expect participants to only engage in the social comparative thinking during the social judgment 3.1.5. Data treatment trials making the distractors cues goal-relevant. In contrast, these cues Response times where the probe was misidentified were removed offer no such advantage in the memory trials and should not affect (4.1 %), with remaining latencies truncated below 200 ms (0.9 %) and attention in the same way. above the Tukey criterion (1539 ms; 5.8 %). 10.2 % of trials met at least Next to the attentional cost of the comparison cues, the current study one of these exclusion criteria.6 The remaining scores were then log transformed for use in the main analyses. 4 Furthermore, difference scores were again calculated in a similar The size and stability for the correlations of interest correspond to the manner. Firstly, validity difference scores for each judgment type were average size of correlations and standard deviations in social psychology respectively as determined in Richard et al. (2003). created by subtracting response times for valid trials from those of 5 The student population includes mature students and the campus is open to all. No exclusion criteria for age were defined, resulting in the broad age range seen here. However, as seen in the mean and standard deviation, the majority of 6 The distribution of these exclusions across the conditions for this and all the population was of a typically young age for a student population. following studies can be found in the supplemental materials online. 5 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 Fig. 2. MSCT trial in order; the social or memory trial is presented for 3000 ms; a short 500 ms pre-trial pause with fixation cross is presented; A Cue is presented followed by the probe 250 ms apart; finally, an open format response is given to the initial item. invalid trials with higher scores indicating more visuospatial attention. expected increased response times compared to the memory task for CA scores were calculated again by subtracting the validity difference invalidly cued trials in the pairwise comparisons (ΔM = 0.057, SE = scores from the memory condition from those of the social judgment 0.005, p < .001, 95 % CI [0.047, 0.066]), it also showed a similar, but condition, reflecting increased visuospatial attentional bias in trials larger effect for validly cued trials, (ΔM = 0.074, SE = 0.006, p < .001, where the cues were task relevant compared to when they were not. As 95 % CI [0.061, 0.087]), see Fig. 3. In fact, this discrepancy means the there were no non-social cues, no CID scores were calculated. valid-invalid difference score was larger for memory trials than social judgment trials, even though the social judgment task did increase re- 3.2. Results action times overall, F(1, 161) = 196.11, p < .001, η2 p = 0.549, 90 % CI [0.465, 0.613]. A 2 (Task: Social judgment vs. Memory) × 2 (Comparison direction: Unsurprisingly, the validity of the cue overall also showed a large Upward vs. Downward) × 2 (Comparison extremity: Moderate vs. significant main effect, F(1, 161) = 656.38, p < .001, η2 p = 0.803, 90 % CI Extreme) × 2 (Cue validity) factorial repeated measures ANOVA was [0.760, 0.833], with invalid cues increasing the response times compared conducted on the log transformed latencies. Again, participants without to valid ones. Lastly, a significant interaction was found between the task a single trial left in one of the factor levels, after application of the type and the direction of comparison, F(1, 161) = 4.39, p = .038, η2 p = exclusion criteria, could not be included in the analysis. 0.027, 90 % CI [0.001, 0.079]. This effect was marked by longer latencies The interaction effect of task type and validity of the cues, the focal for upward comparison cues during the social judgment task trials, (ΔM = test of the main hypothesis regarding visuospatial attention, was found 0.008, SE = 0.004, p = .038, 95 % CI [0.000, 0.015]), but not the memory to be significant, F(1, 161) = 8.23, p = .005, η2 p = 0.049, 90 % CI [0.009, trials, (ΔM = − 0.002, SE = 0.004, p = .648, 95 % CI [− 0.009, 0.006]). This 0.112]. However, although the social judgment condition showed the attentional bias towards upward comparisons specifically may reflect, or 6 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 be the underlying cause for, the preferential selection of upward over judgment task trials than during memory trials. An interaction between downward comparisons when no lateral comparison is provided, as re- the social nature of the cue and the task would be indicative of compar- ported in a recent meta-analysis (Gerber et al., 2018). ison related non-visuospatial attentional bias. Additionally, we expected visuospatial attentional bias specifically to be reflected in this effect 3.2.1. Correlational analyses being more pronounced when cues are invalidly cued than validly cued. No significant correlations were found between the MSCT scores and the INCOM or the demographics. 4.1. Method 3.3. Discussion 4.1.1. Participants A sample of 66 German speakers was recruited on campus at the In this study, the data provided counterintuitive evidence that the University of Cologne. The sensitivity analysis showed this sample size largest signs of attentional bias towards comparison standards is to be was sufficient to detect an effect of η 2 p > 0.111 with α = 0.05 and β = found not in the invalid trials (as would be expected of visuospatial 0.20 (determined in G*Power; Faul et al., 2007). attention), but in the validly cued ones. Although surprising, these Consent was obtained prior to the experiment, in which the partic- findings seem to indicate the external standards did receive some limited ipants were informed about the collection and storage of data in an visuospatial attention, but that the extracting of the relevant social in- anonymised form as well as their right to withdraw from the experiment formation is so short lived and efficient that only the extremely fast valid at any time without consequence. Participants received a monetary trials are affected. The overall longer latencies of the social comparison compensation of 6 euros for their participation. The sample consisted of trials compared to the memory trials could indicate that, once the social 66.7 % female participants and was aged between 18 and 38 years (M = information has been internalised, the largest part of the attentional cost 22.54, SD = 3.45). might reside in an extended cognitive process for which no lingering visuospatial attention is necessary. 4.1.2. Design However, the current set-up cannot clearly disentangle this possi- The current study used an identical three-step MSCT set-up as in the bility from simple task demands that might not involve the presented previous study. However, the standing of the comparison standards on cues at all. For instance, it is conceivable that being faced with the the judgment dimension was no longer considered explicitly, but rather judgment task itself was sufficient for some participants to recruit an the same stimuli from the previous study were randomly presented internal comparison standard and ignored the external standard all across trials. In addition to these social cues, non-social ones were together. This scenario would still lead to an overall lengthening of re- added, which did not offer any comparative information relevant to the action times for social judgment trials regardless of the presented cues. social judgment. As a result, only the social cues represented pertinent In order to distinguish between the external and possible internal comparison standards for the judgment task. comparison related effect, or other unrelated task demands, Study 2 will Twenty trials were run for each combination of the judgment task include a condition that displayed non-social cues that do not offer any (social judgment vs. memory), cue type (social vs. non-social), and its social information in addition to those that do. Any difference between validity (valid vs. invalid), for a total of 160 trials. The probe presen- these conditions would reflect an attentional bias that is strictly the tation (left or right) and probe identity (u or n) were presented in equal result of the informational content that the external standards provide numbers to each participant and counterbalanced between participants during the social judgment task, beyond the possible influence of other to ensure equal presentation overall. task demands that could be present in both situations. 4.1.3. Stimuli 4. Study 2 The pre-tested questions and stimuli were used in this study as in Study 2, with the exception of the non-social cues which were taken In this study, the MSCT paradigm was implemented with the addition from the pilot study. This time no distinction was made between mod- of presenting non-social cues in order to disentangle non-visuospatial erate, extreme social comparisons, as the judgments themselves would attention bound by the content of external comparison cues from the not be analysed further since no consistent assimilation and contrast potential use of internal standards and other unspecified task related effects were found in the initial study. cognitive demands. We expected the presence of social cues, compared to non-social cues, to increase reaction times to a larger extent during 4.1.4. Scales In addition to the main task, we again included scales hypothesized to be related to the comparison process. For exploratory reasons we included at the end of the study the INCOM (Gibbons & Buunk, 1999), the BFI-10 (Rammstedt & John, 2007), the Rosenberg self-Esteem scale (Rosenberg, 1965), the Benign and Malicious Envy Scale (BeMaS; Lange & Crusius, 2015), and the Body Dissatisfaction subscale (Garner et al., 1983). For a more complete description of these scales, refer to the supplemental ma- terials online. As no correlations were found between these scales and the MSCT scores, results for the exploratory correlational analyses will not be described in text, but can be found in Table A1 in the Appendix A. 4.1.5. Procedure The procedure was identical to Study 1 with the exception that there were only 160 trials as part of the main task. Furthermore, in addition to the INCOM, all described scales were administered at the end of the study, after which participants were debriefed and given their compensation. Fig. 3. Means and standard errors of log transformed latencies by Task type 4.1.6. Data treatment and Cue validity. Note: Standard errors in all figures were calculated using the formula described by Jarmasz and Hollands (2009). Response times for erroneous trials (3.6 %) were excluded, latencies were further truncated below 200 ms (1.9 %) and above the Tukey 7 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 criterion (1479 ms; 4.5 %). Due to co-occurrence of exclusion criteria, cues did not moderate this effect in any way as would be expected if the total amount of excluded latencies amounted to 9 %. The remaining visuospatial attention played an extensive role. scores were subsequently log transformed for use in the main analyses. Considering these results in addition to the previous findings, it Again, the truncated raw scores were used to create a number of seems likely that the increased latencies during the MSCT trials is indeed difference scores for use in correlational analyses. These scores were driven in large part by prolonged non-visuospatial attentional bias calculated similarly to those presented in Table 1, but with the control resulting from the comparison process. As opposed to the idea that condition now consisting of a memory task. If the comparison process participants allow their gaze to linger on the comparison standards does not rely heavily on visuospatial attention, but instead mainly binds throughout the comparison process, the current pattern is more indic- cognitive resources without the need for continued visual inspection of ative of cognitive engagement with the social information that is effi- to the cues, one would expect longer reaction time for social judgment ciently extracted from the cues after brief visual inspection in the first trial regardless of cue location when comparison information is present. few hundred milliseconds of presentation. This pattern would suggest This should be reflected in differences in the cue difference scores for the that the very early extraction of social information and ranking that was two conditions, i.e. larger CID scores which capture the overall atten- found by Ohmann et al. (2016) might be all that is needed to form tional bias that is caused by the presents of pertinent comparison stan- sufficient mental representations of the standard to complete the entire dards when engaged in a social judgment task. comparison process. After this extraction, no additional visuospatial attention is necessary for the full comparison process to unfold, although 4.2. Results cognitive engagement with these representations seems to continue to inhibit task performance for longer periods. This would also explain how A 2 (Validity) × 2 (Task: social judgment vs. memory) × 2 (Cue type: even subliminally presented comparison standards can have an effect on Social vs. non-social) factorial repeated measures ANOVA was con- the final comparative judgments (Mussweiler et al., 2004a). ducted on the log transformed truncated latencies. If no trials remained If this is the case and no prolonged visuospatial attention is necessary in one of the factor levels, after the exclusion criteria were applied, the for the comparison process to unfold, removing the cue from view at the participant could not be included in this analysis. The results of this onset of the probe should leave the CID effect intact. Therefore, the next analyses showed that the data were in line with the main prediction study will investigate if the non-visuospatial attentional bias remains concerning non-visuospatial attention, as seen in the significant Task present when stimuli are covered by a mask after initial exposure. type and Cue type interaction, F(1, 64) = 4.76, p = .033, η2 p = 0.069, 90 % CI [0.003, 0.183], see Fig. 4. The prediction regarding visuospatial 5. Study 3 attentional was, however, not supported, with the three-way interaction including cue validity failing to reach significance, F(1, 64) = 0.05, p = As both previous studies found no evidence of visuospatial attention, .819, η2 p = 0.001, 90 % CI [0.000, 0.032]. this third study limited the possibility of prolonged visuospatial attention A significant main effect for the type of task, F(1, 64) = 41.74, p < completely by removing the cues at the probe onset. This allowed partic- .001, η2 p = 0.395, 90 % CI [0.239, 0.513], revealed that trials with social ipants to extract the social information initially with the brief visuospatial judgments overall showed increased response times for probe identifi- attention necessary to internalise it, but limited later attentional bias to be cation compared to trials with a memory task. Furthermore, validly cued purely cognitive in nature. As in the previous study, we expected that, trials overall were associated with far smaller latencies compared to compared to trials with non-social cues, trials with social cues would in- invalidly cued trials as can be seen in the very large significant main crease reaction times for social judgment trials more so then for the effect of cue validity, F(1, 64) = 353.71, p < .001, η2 p = 0.847, 90 % CI memory trials. Such an effect would be indicative of a non-visuospatial and [0.786, 0.880]. The last of the main effects, concerning the social nature likely purely cognitive attentional bias. However, as prolonged visuospa- of the cue, did not meet the conventional standards of significance in tial attention was no longer possible after probe onset, we did not expect these data, F(1, 64) = 3.44, p = .068, η2 p = 0.051, 90 % CI [0.000, 0.158]. the validity of the cue to have any influence on this effect. 4.3. Discussion 5.1. Method In this second study, the reaction time data revealed that social cues entailed a higher attentional cost during the social judgment task than 5.1.1. Participants during the memory tasks, in line with the expectations regarding a Similar to the previous study, 60 German speakers were recruited on general CID effect. However, in contrast to the theoretical expectations the University of Cologne campus, which sensitivity analysis showed as well as to the surprising results found in Study 1, the validity of the would be sufficient to detect an effect size of η2 p > 0.121 with α = 0.05 and β = 0.20 (determined in G*Power; Faul et al., 2007). These re- spondents offered their informed consent about the collection and storage of their data and their right to withdraw from the experiment without consequence at any time. Six euros were offered for their participation in this study. The sample was 35 % female and aged be- tween 16 and 29 years (M = 22.15, SD = 3.01). 5.1.2. Design The general design was identical to Study 2 with the exception that the cues no longer remained on the screen once the probe appeared, making sustained visuospatial attention impossible, but still allowing any cognitive processes that do not rely on visuospatial inspection to occur. Each combination of the judgment task (social judgment vs. memory), cue type (social vs. non-social), and its validity (valid vs. invalid) was measured in 20 trials, totalling 160 trials. The probe pre- sentation (left or right) and probe identity (u or n) were again presented Fig. 4. Means and standard errors of log transformed latencies by Task type in equal numbers to each participant and counterbalanced between and Cue type. participants so all combinations occurred equally often overall. 8 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 5.1.3. Stimuli onset. The former allowed both visuospatial attention and cognitive Stimuli were identical to those used in Study 2. Since no strong as- attention to affect reaction times, while the latter did not allow sustained sociations were found between the additional scales and the reaction visuospatial attention, isolating only the processes that can occur time measures in Study 2, the additional scales removed with the without the prolonged visual availability of the cue. exception of the INCOM. However, as the exploratory correlational For unmasked trials, we expected to find a CID effect similar to Study analyses in this Study once again did not reveal any consistent correla- 2; with social cues increasing reaction times compared to non-social cues tions, they will not be reported in detail. for judgment trials more so than for memory trials. In this effect, both cognitive and complex visuospatial attentional components could be at 5.1.4. Procedure play. As a last test of more classical pattern of visuospatial attention, cue The procedure was identical to Study 2, except that the INCOM was validity would be expected to moderate this effect in the unmasked the only scale administered at the end of the study. condition. However, if purely cognitive processes of comparison can occur without any type of prolonged visuospatial attention, masked 5.1.5. Data treatment trials would be expected to show a similar CID effect as unmasked trials. Once again erroneous trials were removed (4.7 %), with remaining latencies truncated below 200 ms (0.6 %) and above the Tukey criterion 6.1. Method (1539 ms; 3.8 %). In 8.6 % of trials met at least one of these exclusion criteria. The remaining scores were then log transformed for use in the main 6.1.1. Participants analyses. Again, the truncated raw scores were used to create the same Considering the addition of the masked and unmasked conditions, difference scores presented in Table 1 for use in the correlational analyses. sample size was doubled from the previous two studies to include 118 German speakers recruited at the University of Cologne campus. Sensi- 5.2. Results tivity analysis showed this sample size was sufficient to detect an effect of η2 p > 0.064 with α = 0.05 and β = 0.20 (determined in G*Power; Faul The log transformed truncated latencies were used in a 2 (Task: So- et al., 2007). Respondents' consent was obtained after being informed cial judgment vs. Memory) × 2 (Cue type: Social vs. Non-social) × 2 about data collection and storage practices and their right to withdraw (Validity) factorial repeated measures ANOVA. Counter to the expec- from the experiment without consequence. A monetary reward of 6 tations, the interaction between Task type and Cue type did not produce euros was offered for participation. 63.6 % of the final sample was fe- a significant interaction, F(1, 59) = 0.23, p = .637, η2 p = 0.004, 90 % CI male and was aged between 16 and 64 years (M = 24.88, SD = 7.56). [0.000, 0.066], with the three-way interaction also failing to reach significance, F(1, 59) = 2.15, p = .15, η2 p = 0.035, 90 % CI [0.000, 6.1.2. Design 0.138]. These findings, thus, offer no evidence of attentional bias when a The design combined the ones used in Study 2 and 3 with both trials mask was present. In addition to these main results, a main effect of the where cues remained on screen until a response was given as well as type of task was again found. Social judgment trials showed significantly trials where cues were masked once the probe appeared. In order to increased latencies compared to ones with a memory task, F(1, 59) = reduce stress on participants, only 16 trials were included for each 48.94, p < .001, η2 p = 0.453, 90 % CI [0.292, 0.566], again pointing combination of the judgment task (social judgment vs. memory), cue towards higher cognitive demands for the social judgment task that may type (social vs. non-social), validity (valid vs. invalid), and masking be the result of internally produced comparison standards being condition, totalling 256 trials. The probe presentation (left or right) and generated. Finally, the analysis again showed that participants identi- probe identity (u or n) were again presented in equal numbers to each fied the probe significantly more quickly when paired with a valid cue participant and counterbalanced between participants to ensure all than an invalid cue, F(1, 59) = 276.96, p < .001, η2 p = 0.824, 90 % CI combinations occurred equally overall. [0.752, 0.863]. However, the social nature of the cue did not signifi- cantly affect reaction times compared to non-social ones in these data, F 6.1.3. Procedure (1, 59) = 0.04, p = .845, η2 p = 0.001, 90 % CI [0.000, 0.027]. The procedure was identical to that of Study 3. 5.3. Discussion 6.1.4. Data treatment Latencies for all trials where the probe was not correctly identified In the current study, where the visual availability of cues was were removed (3.7 %), and then were truncated below 200 ms (0.6 %) restricted to the onset of the probe, no detectable signs of comparison and above the Tukey criterion (1488 ms; 4.2 %). One or more of these related attentional bias were found. Contrary to the predictions and the exclusion criteria were met in 8 % of trials. These remaining scores were lack of evidence for validity effects in the first two studies, it seems that then log transformed for use in the main analyses. the visual availability of the cues might yet be necessary for the CID to The non-transformed truncated raw scores were again used to create occur even though no signs were detected previously. Therefore, the the same difference scores as in previous studies, now for masked and type of visuospatial attention that occurs during the comparative pro- unmasked trials separately, see Table 1. cess might not be limited to the fast initial inspection nor simple visual lingering. Instead, it may include a more complex pattern of gazes that 6.2. Results the standard validity manipulation in the MSCT is not sensitive enough to detect consistently. However, based on the current study's absence of A 2(Task: Social judgment vs. Memory) × 2 (Cue type: Social vs. evidence this is too strong a claim to be supported by these data alone. Non-social) × 2 (Validity) × 2 (Mask) factorial repeated measures Therefore, the next study will include both masked and unmasked trials ANOVA was conducted on the log transformed truncated latencies. to replicate the previous findings as well as test whether the masking Counter to the expectations, the interaction between task type and cue indeed prohibits the CID effect. type did not reach significance in this study overall, F(1, 117) = 0.46, p = .501, η2 p = 0.004, 90 % CI [0.000, 0.043], and the presence of an effect 6. Study 4 was also not moderated by the masking of the cues, F(1, 117) = 1.29, p = .258, η2 p = 0.011, 90 % CI [0.000, 0.061]. Follow up analyses looking This fourth and final study combined aspects of the two previous at the unmasked and masked trials separately showed that neither studies, by presenting trials where the cues remain on screen until a unmasked, F(1, 117) = 1.71, p = .194, η2 p = 0.014, 90 % CI [0.000, response was provided, and trials where the cues were masked at probe 0.068], nor masked trials, F(1, 117) = 0.278, p = .599, η2 p = 0.002, 90 % 9 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 CI [0.000, 0.037], showed this interaction to be significant in these data. 6.3. Discussion The three-way interaction between task type, cue type and cue val- idity did not reach the standard level of significance in this sample The results of this final study somewhat contradict the evidence either, F(1, 117) = 3.82, p = .053, η2 p = 0.032, 90 % CI [0.000, 0.099], found in the initial two studies and the pilot study concerning the indicating there was not enough evidence to support the hypotheses presence of a CID effect for unmasked trials, and found no evidence of regarding visuospatial attention in these data, see Fig. 5. This effect was such an effect in masked trials either. Consequently, there was no sign also not moderated by the presence of a mask, F(1, 117) = 0.65, p = that masked and unmasked trials differed in the size of a potential CID in .420, η2 p = 0.006, 90 % CI [0.000, 0.048]. Follow up analyses separated the main analyses. Lastly, there was once again no evidence of classical by mask condition did not show enough evidence that classical signs of signs of visuospatial attention to the comparison cues, nor that such an visuospatial attention were present for unmasked, F(1, 117) = 3.330, p effect might be influenced by the masking condition. = .071, η2 p = 0.028, 90 % CI [0.000, 0.092], or masked trials in these However, the exploratory correlational analyses did reveal a nega- data, F(1, 117) = 0.573, p = .451, η2 p = 0.005, 90 % CI [0.000, 0.046]. tive association between the masked and unmasked CID effects, sug- The social judgment task was again found to produce significantly gesting engagement with the social stimuli did diverge systematically as larger latencies than the memory task was, F(1, 117) = 118.78, p < .001, η2 p a result of the masking procedure during the social judgment trials. = 0.504, 90 % CI [0.398, 0.583], reaffirming the undefined higher Nevertheless, the results so far do not clearly support a CID with visual cognitive demand of the social judgment task found throughout the studies. component, nor one that is purely cognitive in nature. This final study, The main effect of the validity of the cue did also reach significance in this far from presenting a clearer image of the attentional pattern, showed no sample, F(1, 117) = 689.05, p < .001, η2 p = 0.855, 90 % CI [0.816, 0.880], obvious support for any of the findings in the initial studies. Although with valid trials being associated with faster responses than invalid ones. this lack of evidence is itself not evidence that the CID does not exist, it Lastly, an interaction was found between the presence of a mask and may cast some doubt on the size and robustness of CID effect. Therefore, the validity of the cue overall, F(1, 117) = 4.33, p = .040, η2 p = 0.036, 90 the next section will summarise all findings in a single paper meta- % CI [0.001, 0.105]. Surprisingly, in these data masking the cues analyses for a clearer overview. actually significantly increased response times compared to leaving the cues visible only for valid trials (ΔM = 0.009, SE = 0.004, p = .026, 95 % 7. Meta-analyses CI [0.001, 0.017]), but not for invalid trials (ΔM = -0.001, SE = 0.004, p = .753, 95 % CI [− 0.009, 0.006]). Although a rather small and unex- Considering the variation of findings across this set of studies, a pected effect that should not be given too much weight, it could single paper meta-analysis was performed to gain a clearer overview of potentially be a sign that a general inhibition on return effect could the robustness of the various effects and their sizes. Even though there result from the mask being present for valid trials. are substantial differences across the studies in terms of methodology and stimulus selection, it is important to include all relevant studies that 6.2.1. Correlations form part of this project in order to gain the most comprehensive Finally, the extent to which the corresponding masked and overview the effects. Therefore, another study in the project that is not unmasked scores correlate in these data can help shed some light on directly reported here will also be included.7 The main analyses will which indicators are consistent across the masked conditions and thus estimate the effect of the social judgment task, the social nature of the unaffected by the prolonged visual availability of the cue. Therefore, cues, the cue difference scores, and the validity effects for unmasked only these corresponding scores will be considered here, though all trials found within this project. As only two instances for estimating the other correlations can be found in Table A2 in the Appendix A. The effects of masked trials exist, the results for these studies are described in validity difference score for masked and unmasked social judgment a shorter summative overview included at the end. It is, also, important trials showed a significant positive correlation, r = 0.443, p < .001, 95 to note that these analyses should be treated as summative of the find- % CI [0.285, 0.577], showing the visuospatial attentional bias for social ings in this paper rather than a direct separate test of the hypotheses judgment trials was relatively consistent whether the cue was masked or themselves due to the small number of selected studies and variation not. Masked and unmasked CID scores, however, showed a significantly among methods. All means, standard deviations and correlations used in negative association, r = − 0.218, p = .018, 95 % CI [− 0.383, − 0.039], the meta-analyses can be found in the supplemental materials online, indicating that the amount participants show a social judgment task and all corresponding forest plots can be found in the Appendix. specific bias for social cues, regardless of cue location, for unmasked trials was inversely related to their performance when a mask was present. No other scores were significantly correlated with each other. 7.1. Unmasked trials 7.1.1. Task Throughout the studies, engaging in social judgments seemed to be more cognitively demanding than performing the memory task. This effect of the judgment task itself will be investigated first. For each of the five studies the log-transformed latencies were averaged for both task types in unmasked trials. Separate repeated measures t-tests were con- ducted using the truncated average log latencies for social judgment verses control trials for each study to estimate Cohens dz effect sizes of 7 All studies that are not directly reported here are available on the project's OSF page. The choice to not directly describe study S1 was made in the sake of clarity and brevity as the reported line of studies focuses more on quick auto- matic attention, while this study uses an extended SOA (1000 ms). We felt this would further complicate the narrative of an already heterogeneous set of re- sults. Furthermore, in the interest of transparency, a second line of research that explored self-relevant comparisons using a similar paradigm is also described as Fig. 5. Means and standard errors of log transformed latencies by Task type, Study S2, but is not included in the meta-analyses as it was not deemed relevant Cue type, and Cue validity. since the focus of this project is specifically on other-related judgments. 10 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 task type. The Cohens dz's were then used in a meta-analysis (utilizing 2.87, p = .004, indicating increased attentional bias towards social cues the metafor package in R; Viechtbauer, 2010), which suggested the when engaged in a social judgment with no significant heterogeneity in presence of an overall effect of task type on the average log latencies, dz the effect sizes, Q(3) = 1.71, p = .63, I2 = 0 %. This summation shows = 0.78, 95 % CI [0.49, 1.07], Z = 5.27, p < .001. However, the Q was across studies there is evidence that a consistent CID did occur, highly significant, Q(4) = 794.64, p < .001, and the I2 was high (99.29 reflecting the presence of a distinct attentional cost associated with %), indicating very large amounts of heterogeneity in effect sizes likely engaging in external social comparisons. However, the exact nature of reflective of the different control task used in the pilot study, variations this attentional bias (be it visuospatial, purely cognitive, or a combi- in SOA, and stimulus effects. Individual studies were left out one by one nation of the two) cannot be assessed based on this effect alone. to probe the main cause of the heterogeneity and the robustness of the Therefore, the next analyses will isolate the evidence for classic signs of effect, with the largest reduction in heterogeneity being achieved by visuospatial attention by looking at the validity difference scores. excluding the pilot study increasing the estimate, dz = 0.94, 95 % CI = [0.84, 1.04], Z = 17.92, p < .001, although heterogeneity remained very 7.1.4. Visuospatial bias large, I2 (92.68 %). Estimates in all analyses varied between dz = 0.70 to The question remains whether there is a visuospatial aspect to the 0.94. Notwithstanding these issues, all estimates remained positive, CID described above. If it does, the validity of the cue in the MSCT indicating that across studies an overall increase in response latencies paradigm is usually expected to play a moderating role in this bias (Fox was recorded for the social judgment task, which was highly sensitive to et al., 2011). However, only Study 1 showed any effect related to val- inter-study variation. This indicates that regardless of the validity and idity, and even this was in the opposite direction as would be expected. cue type, the social judgment task requires more cognitive resources To investigate if on aggregate there is any evidence visuospatial atten- leading to longer latencies. This is presumably due to the fact that the tional bias to the comparison cues when engaging in a social judgment, a comparison process can take place internally with the use of internal final meta-analysis will compare the validity difference scores for social standards even when no external stimuli are presented (Kahneman & judgment versus control trials. Miller, 1986; Mussweiler, 2003). However, as noted in the introduction, For each of the five studies log transformed latencies from all the exact nature of such internal processes are not easily assessed and unmasked valid trials were subtracted from those of unmasked invalid remain speculative within the limits of the current data as they may still trials separately for social judgments and control tasks only for trials be related to other unspecified task demands. with social cues to create the validity difference scores. Multiple repeated measures t-tests were then used to estimate Cohen's dz for each 7.1.2. Social cues study. A meta-analysis was then conducted using these effect sizes, The second main effect of interest is the possible bias towards all cues showing a non-significant overall effect, dz = − 0.04, 95 % CI [− 0.22, of a social nature independent of the task type. To summarise the effect 0.13], Z = − 0.49, p = .62, that had signs of significant heterogeneity, Q the presence of social cues had on response latencies in the unmasked (4) = 14.91, p = .005, I2 = 68.11 %. As with previous cases of hetero- trials across studies, we included only the four studies that included geneity, separate analyses were run in which individual studies were left unmasked social as well as non-social trials. For each of these studies, the out one by one to test the robustness of the effect and cause of hetero- average truncated log latencies of unmasked social trials and non-social geneity. The greatest reduction in heterogeneity was achieved by trials were included in paired sample t-tests to calculate Cohen's dz ef- removing Study S1 in which the stimulus onset asynchrony was fect sizes. A meta-analysis of the resulting effect sizes showed a significant increased, dz = − 0.12, 95 % CI [− 0.26, 0.02], Z = − 1.67, p = .094, I2 = meta-analytic effect of social cues on the average log latencies, dz = 0.21, 46.10 %. This might suggest that any validity effects are sensitive to 95 % CI [0.01, 0.41], Z = 2.06, p = .040. There seemed to be significant differences in SOA, in line with the notion that social comparisons take heterogeneity in this effect across the studies, Q(3) = 206.52, p < .001, place early on in processing (Ohmann et al., 2016), although in these and the I2 was high (97.84 %), again possibly caused by the differences in data such an effect still remained non-significant and in the opposite the stimuli used across the studies. For this reason, individual analyses direction as would be expected. Indeed, most effect size estimates in were again run excluding individual studies one by one to probe the main these analyses were largely in the opposite direction, ranging from dz = cause of the heterogeneity and the robustness of the effect. Heterogeneity 0.02 to − 0.12, and remained non-significant. Therefore, this summative was most reduced by excluding Study 4, dz = 0.29, 95 % CI [0.10, 0.48], analysis suggests the current line of studies did not provide evidence to Z = 3.02, p = .003, although heterogeneity remained large, I2 (95.70 %). support the hypotheses regarding visuospatial attentional bias towards Across the analyses, estimates remained positive, dz = 0.12 to 0.29. The comparison standards. positive nature of these summative results speaks to the presence of a small general bias towards social cues across the studies, indicating social 7.2. Masked trials stimuli themselves are more engaging than non-social images regardless of the task that was presented. Data treatment for all analyses with masked trials was identical to the ones described for unmasked trials described above. The task effect for 7.1.3. CID the two studies that included masked trials, the effect of task type on the The main goal of this project was to investigate the presence or average log latencies also reached significance, dz = 0.91, 95 % CI [0.87, absence of attentional bias towards social cues when engaged in a social 0.95], Z = 46.92, p < .001. Overall the effect of the social nature of the judgment task as a sign of comparative thinking. Some studies indeed cue was in the opposite direction of the effect found in the unmasked showed signs of such a general CID, hinting at a slowdown in reaction trials and did not reach significance, dz = − 0.02, 95 % CI [− 0.05, 0.00], Z times resulting from the comparison process. The following meta- = − 1.68, p = .09. Similarly, no evidence for a CID effect was found for analyses will estimate the aggregated effect of this CID across all rele- masked trials as reflected in the non-significant difference between the vant studies by comparing the cue difference scores of social judgment two masked cue difference scores, dz = − 0.05, 95 % CI [− 0.24, 0.14], Z = versus control tasks. − 0.55, p = .58. In sum, evidence for both cue related effects was only Cue difference scores were calculated for each of the four relevant present in trials where the stimuli remain visually available throughout studies by subtracting the log transformed latencies of all unmasked the trials and was not found when the cues were masked. Despite the trials with non-social cues from those with social cues for social judg- theoretical contradiction of visuospatial attention for masked stimuli, a ment and control trials separately. These scores were then used in meta-analysis was run for the masked validity difference scores for multiple repeated measures t-tests to estimate the Cohens dz effect sizes. comparison purposes. Estimates for unmasked trials were similar to those These scores were again used in a meta-analyses, which showed a sig- found in unmasked trials and non-significantly different from zero, dz = nificant positive aggregate effect, dz = 0.20, 95 % CI [0.06, 0.33], Z = − 0.10, 95 % CI [− 0.27, 0.06], Z = − 1.23, p = .22. 11 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 8. General discussion previous findings concerning the processing of comparative information (Ohmann et al., 2016). The rest of the process might well continue at a Across the presented studies, the social comparison task itself and the largely cognitive level once the visual information has been internalised presence of a social cue were both associated with increased latencies for in the first few hundred milliseconds of exposure. This would explain the unmasked trials, though these effects proved quite sensitive to meth- occurrence of a CID effect in the later studies even without any of the odological variations. The fact that the task related effects also extended classic signs of visuospatial attention within the current MSCT design. to the masked trials, and those with non-social stimuli implies the However, it fails to account for the difficulty of finding a CID effect once forming of a social judgment is in and of itself somehow cognitively cues were masked, leaving the possibility of a more complex form of demanding. Theoretically, this could be explained by the ubiquity of the visuospatial attention being part of the CID. comparison process that often occurs completely internally with pre- Nonetheless, there are some important limitations to the evidence existing standards when external standards are not available (Kahne- presented for such an explanation. Most obviously, the proposed early man & Miller, 1986; Mussweiler, 2003). In this line of reasoning, a visual component that would be necessary to initially assess the com- comparison process occurs with internally selected standards in the parison information was not detected using this standard spatial cueing absence of external ones when asked to make a social judgment. The set-up and is thus only speculatively based on previous findings internal comparison process is one that exerts its own cognitive demands (Ohmann et al., 2016). Although it is theoretically unlikely that a CID and, therefore, impedes performance on parallel tasks, although the effect could occur without any kind of visuospatial attention, it seems current findings cannot exclude other explanations related to unspeci- the sensitivity of the current paradigm proved insufficient to detect and fied task demands. investigate it. Similarly, the lack of evidence for a CID in masked trials is More interestingly for the current project and research question, only suggestive of some type of prolonged visual aspect of the effect, but unmasked social cues were found to exert a higher attentional cost this could not be determined definitively within the current paradigm. during the social judgment task than the control tasks, indicating a CID Whether prolonged visuospatial attentional in some form is part of the caused by the increased engagement with the external comparison CID or if cognitive processes exclusively drive the effect is, therefore, standards during the comparison process. Although this overall effect still an open question. Future work using measures that are more sen- does not help us clearly distinguish the visuospatial from non- sitive might still yield robust cueing effects, though they likely occur visuospatial cognitive attention, the lack of evidence for the CID very early on and are very small in size, if present at all. across studies with masked stimuli is an initial indication that it might Furthermore, given the lack of clear evidence for a substantial visual include some form of prolonged visual component. However, the single aspect to the CID, the current MSCT paradigm falls short in its capacity direct test in this study did not confirm this assertion, and the simple to gain deeper insight into non-visuospatial component of the effect. A absence of a significant CID effect in masked trials, while suggestive, is more complex procedure would need to be implemented in order to not evidence that it does not exist. Indeed, the complete lack of CID in determine directly which step or steps of the comparison process are a masked trials is surprising, as previous work has reported that even main driver in the CID effect itself. However, some initial suggestions subliminally presented comparison standards can directly affect judg- can be made based on the current findings. Namely, given the overall ment outcomes (Mussweiler et al., 2004a). This discrepancy may be due longer latencies in social judgment trials and the theoretical ubiquity of to the novel nature of the standards presented as cues in the current the comparison process (Dunning & Hayes, 1996; Festinger, 1954; work, where previous literature has relied on famous individuals about Kahneman & Miller, 1986; Mussweiler, 2003), it could be assumed that whom participants already had sufficient internal knowledge. For even in the social judgment trials without external standards a number instance, novel standards might require extended or reoccurring visuo- of comparative processes are taking place. In fact, all steps of the spatial attention, where known standards need merely be activated. comparative processes other than the initial visual inspection, infor- In any case, the current work has shown that the whole comparative mation extraction and information processing of the external compar- process, at least when it comes to forming actual social judgments, is one ative information can theoretically occur for internal standards as well. that extends far beyond the efficient initial inspection and ranking of Hence, these steps, certain to be unique to the socially cued trials, could stimuli that is described in the earlier work done by Ohmann et al. be the most likely causes behind the CID, at least if social comparisons (2016). Instead, the current findings show that after the initial pro- are indeed as ubiquitous as suggested in the literature. cessing, the formation of the social comparative judgments require some Notwithstanding these issues and need for further investigation, the type of further engagement with the comparative information in order to current work demonstrates a basic CID effect across all studies as evi- integrate the acquired information. In further contrast to this previous dence for attentional preoccupation with the comparison relevant work, the social comparisons did require a significantly longer pro- stimuli that extends long past initial visual inspection. The pattern is, cessing time than non-social comparisons as seen in the pilot study. instead, suggestive of an initial very early short-lived visuospatial Hence, despite the similar timeframe associated with processing of the component, followed largely by non-visuospatial attention in the form of social and non-social comparative information, the later stages of the cognitive engagement with the extracted social information. These comparative process associated with the formation of judgment esti- finding can form the basis for future work looking at attentional patterns mates appear to be increasingly affected by the more complex nature of in the early comparison process that aim to expand and clarify the social stimuli (Dahlgren, 1985; Tversky, 1977). questions raised in this underdeveloped area of comparisons research. Furthermore, although the CID effect necessarily implies some minimal level of visuospatial attention in the form of the initial gaze and Data availability statement internalisation of the comparative information, no direct evidence was found for this visuospatial bias. On the contrary, Study 1 actually All raw and aggregated data, and additional materials for all studies showed a validity effect in the opposite direction. The summative results can be found on the Open Science Framework under: https://osf. of the meta-analyses, however, did not find these results to be robust io/hkem9/ across the studies. The overall lack of validity effects might suggest that visual engagement with comparative information occurs very early on in information processing and is completed before the probe onset at 250 Declaration of competing interest ms, a notion that is bolstered by the fact that a change in the SOA was the largest contributor to effect size heterogeneity in the meta-analysis. The authors declare that they have no known competing financial Hence the necessary initial visual inspection seems to be very short interests or personal relationships that could have appeared to influence lived and efficient in extracting social information in congruence with the work reported in this paper. 12 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 13 Appendix A Table A1 Means, standard errors and correlations for all measures and scores used in Study 2. N M SD 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 1. Age 61 22.54 3.45 2. Sex 66 1.33 0.48 0.11 3. INCOM 66 3.37 79.22 − 0.07 − 0.01 4. Self-esteem scale 66 29.62 5.54 − 0.08 − 0.16 ¡0.25* 5. Body dissatisfaction 66 2.25 0.65 0.05 ¡0.29* 0.07 ¡0.39** scale BeMaS 6. Benign envy 66 3.86 1.15 − 0.11 0.04 0.61** − 0.22 0.22 7. Malicious envy 66 2.35 1.18 − 0.18 0.07 0.37** − 0.36** 0.25* 0.42** BFI-10 8. Extraversion 66 3.51 1.11 − 0.18 − 0.30* − 0.17 0.60** − 0.12 − 0.16 − 0.33** 9. Agreeableness 66 3.52 0.71 0.12 0.11 − 0.02 0.08 − 0.12 − 0.07 − 0.24 0.02 10. Conscientiousness 66 3.49 1.01 − 0.13 − 0.32** − 0.16 0.37** − 0.02 − 0.05 − 0.18 0.18 − 0.10 11. Neuroticism 66 3.08 1.03 0.01 − 0.21 0.33** − 0.43** 0.34** 0.29* 0.36** − 0.36** − 0.29* 0.05 12. Openness to 66 3.51 1.06 − 0.18 − 0.08 − 0.05 0.32** ¡0.28* 0.03 − 0.28* 0.24 0.09 − 0.11 − 0.23 experience 13. Validity difference 66 123.17 72.56 0.23 − 0.09 0.09 − 0.19 0.24 0.08 0.01 − 0.17 0.12 − 0.07 0.23 − 0.15 score (memory) 14. Validity difference 66 117.92 64.96 − 0.08 0.05 − 0.17 − 0.06 0.08 0.07 0.05 − 0.06 0.22 0.06 − 0.03 0.20 0.06 score (judgment) 15. CA 66 − 5.25 94.64 − 0.23 0.11 − 0.18 0.11 − 0.13 − 0.01 0.03 0.09 0.06 0.10 − 0.19 0.25* − 0.73** 0.64** 16. Cue difference 65 8.46 84.52 − 0.03 0.15 0.07 − 0.03 − 0.13 − 0.07 − 0.07 0.00 0.22 − 0.05 − 0.17 − 0.10 0.12 0.23 0.07 score (memory) 17. Cue difference 66 22.99 121.64 0.17 0.21 0.17 − 0.11 − 0.28* 0.14 0.07 − 0.25* 0.30* − 0.19 − 0.08 0.17 − 0.19 0.15 0.25* 0.19 score (judgment) 18. CID 65 25.91 104.43 0.10 0.06 0.02 0.00 − 0.05 0.07 0.04 − 0.14 − 0.01 − 0.01 0.14 0.01 − 0.17 − 0.19 0.00 − 0.66** 0.61** Note. * p < .05, ** p < .01 (printed in bold). P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 14 Table A2 Means, standard errors and correlations for all measures and scores used in Study 4. N M SD 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 1. Age 117 24.88 7.56 2. Sex 118 1.36 0.48 0.10 3. INCOM 118 3.37 0.78 ¡0.29** ¡0.25** Un-masked: 4. Validity difference score (memory) 118 123.41 74.98 0.15 0.06 − 0.03 5. Validity difference score (judgment) 118 109.98 84.07 0.07 0.01 − 0.08 0.28** 6. CA 118 − 13.43 95.74 − 0.05 − 0.03 − 0.04 ¡0.54** 0.66** 7. Cue difference score (memory) 118 − 0.958 83.46 − 0.04 0.23* 0.03 0.00 − 0.02 − 0.02 8. Cue difference score (judgment) 118 11.29 98.17 0.05 0.10 − 0.07 0.03 0.16 0.12 0.14 9. CID 118 20.87 119.40 0.07 − 0.08 − 0.08 0.03 0.15 0.11 ¡0.58** 0.72** Masked: 10. Validity difference score (memory) 118 115.48 68.61 0.23* 0.00 0.03 0.07 0.31** 0.22* − 0.09 0.20* 0.23* 11. Validity difference score (judgment) 118 102.85 72.85 0.14 0.04 − 0.10 0.26** 0.44** 0.19* 0.00 0.05 0.04 0.25** 12. CA 118 − 12.63 86.79 − 0.07 0.04 − 0.07 0.16 0.13 − 0.01 0.07 − 0.12 − 0.14 ¡0.58** 0.64** 13. Cue difference score (memory) 118 1.80 84.49 − 0.11 − 0.01 0.03 − 0.12 ¡0.22* − 0.10 − 0.05 − 0.03 0.01 0.05 − 0.11 − 0.13 14. Cue difference score (judgment) 118 − 3.50 85.87 0.01 0.15 0.11 0.00 − 0.12 − 0.11 0.24** − 0.15 ¡0.29** 0.08 − 0.06 − 0.11 0.02 15. CID 118 − 5.30 119.40 0.08 0.11 0.06 0.09 0.07 − 0.01 0.21* − 0.09 ¡0.22* 0.02 0.03 0.01 ¡0.70** 0.71** Note. *p < .05, **p < .01 (printed in bold). P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 Appendix B. Appendix Task: Unmasked Masked Social: Unmasked Masked Cue difference: Unmasked Masked Validity difference: Unmasked Masked (continued on next page) 15 P. Barker et al. A c t a P s y c h o l o g i c a 230 (2022) 103745 (continued ) Validity difference: Unmasked Masked References Jarmasz, J., & Hollands, J. G. (2009). Confidence intervals in repeated-measures designs: The number of observations principle. Canadian Journal of Experimental Psychology, Barker, P., Dotsch, R., & Imhoff, R. (2020). Assimilation and contrast in spontaneous 63, 124–138. https://doi.org/10.1037/a0014164 comparisons: Heterogeneous effects of standard extremity in facial evaluations. Kahneman, D., & Miller, D. T. (1986). Norm theory: Comparing reality to its alternatives. International Review of Social Psychology, 33, 11. https://doi.org/10.5334/irsp.402 Psychological Review, 93, 136. https://doi.org/10.1037/0033-295X.93.2.136 Bless, H., & Schwarz, N. (2010). Mental construal and the emergence of assimilation and Lange, J., & Crusius, J. (2015). Dispositional envy revisited: Unraveling the motivational contrast effects: The inclusion/exclusion model. Advances in Experimental Social dynamics of benign and malicious envy. Personality and Social Psychology Bulletin, 41, Psychology, 42, 319–373. https://doi.org/10.1016/S0065-2601(10)42006-7 284–294. https://doi.org/10.1177/0146167214564959 Dahlgren, K. (1985). The cognitive structure of social categories. Cognitive Science, 9, Maner, J. K., Gailliot, M. T., Rouby, D. A., & Miller, S. L. (2007). Can't take my eyes off 379–398. https://doi.org/10.1016/S0364-0213(85)80004-5 you: Attentional adhesion to mates and rivals. Journal of Personality and Social Dunning, D., & Hayes, A. F. (1996). Evidence for egocentric comparison in social Psychology, 93, 389. https://doi.org/10.1037/0022-3514.93.3.389 judgment. Journal of Personality and Social Psychology, 71, 213. https://doi.org/ Mussweiler, T. (2003). Comparison processes in social judgment: Mechanisms and 10.1037/0022-3514.71.2.213 consequences. Psychological Review, 110, 472. https://doi.org/10.1037/0033- Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). G* Power 3: A flexible statistical 295X.110.3.472 power analysis program for the social, behavioral, and biomedical sciences. Behavior Mussweiler, T., & Damisch, L. (2008). Going back to Donald: How comparisons shape Research Methods, 39, 175–191. https://doi.org/10.3758/BF03193146 judgmental priming effects. Journal of Personality and Social Psychology, 95, 1–295. Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7, https://doi.org/10.1037/a0013261 117–140. https://doi.org/10.1177/001872675400700202 Mussweiler, T., Rüter, K., & Epstude, K. (2004a). The man who wasn't there: Subliminal Fox, E., Russo, R., Bowles, R., & Dutton, K. (2001). Do threatening stimuli draw or hold social comparison standards influence self-evaluation. Journal of Experimental Social visual attention in subclinical anxiety? Journal of Experimental Psychology: General, Psychology, 40, 689–696. https://doi.org/10.1016/j.jesp.2004.01.004 130, 681. https://doi.org/10.1037/0096-3445.130.4.681 Mussweiler, T., Rüter, K., & Epstude, K. (2004b). The ups and downs of social Fox, E., Derakshan, N., & Standage, H. (2011). The assessment of human attention. comparison: Mechanisms of assimilation and contrast. Journal of Personality and Cognitive Methods in Social Psychology, 15–47. Social Psychology, 87, 832. https://doi.org/10.1037/0022-3514.87.6.832 Garner, D. M., Olmstead, M. P., & Polivy, J. (1983). Development and validation of a Ohmann, K., Stahl, J., Mussweiler, T., & Kedia, G. (2016). Immediate relativity: EEG multidimensional eating disorder inventory for anorexia nervosa and bulimia. reveals early engagement of comparison in social information processing. Journal of International Journal of Eating Disorders, 2, 15–34. https://doi.org/10.1002/1098- Experimental Psychology: General, 145, 1512. https://doi.org/10.1037/xge0000222 108X(198321)2:2 15::AID-EAT2260020203 3.0.CO;2-6 Posner, M. I., Snyder, C. R., & Davidson, B. J. (1980). Attention and the detection of < > Gerber, J. P., Wheeler, L., & Suls, J. (2018). A social comparison theory meta-analysis 60 signals. Journal of Experimental Psychology: General, 109, 160. https://doi.org/ + years on. Psychological Bulletin, 144, 177. https://doi.org/10.1037/bul0000127 10.1037/0096-3445.109.2.160 Gibbons, F. X., & Buunk, B. P. (1999). Individual differences in social comparison: Rammstedt, B., & John, O. P. (2007). Measuring personality in one minute or less: A 10- Development of a scale of social comparison orientation. Journal of Personality and item short version of the Big Five Inventory in English and German. Journal of Social Psychology, 76, 129. https://doi.org/10.1037/0022-3514.76.1.129 Research in Personality, 41, 203–212. https://doi.org/10.1016/j.jrp.2006.02.001 Gilbert, D. T., Giesler, R. B., & Morris, K. A. (1995). When comparisons arise. Journal of Richard, F. D., Bond, C. F., Jr., & Stokes-Zoota, J. J. (2003). One hundred years of social Personality and Social Psychology, 69, 227. https://doi.org/10.1037/0022- psychology quantitatively described. Review of General Psychology, 7, 331–363. 3514.69.2.227 https://doi.org/10.1037/1089-2680.7.4.331 Häfner, M. (2009). Knowing you, knowing me: Familiarity moderates comparison Rosenberg, M. (1965). Rosenberg self-esteem scale (RSE). In , 61. Acceptance and outcomes to idealized media images. Social Cognition, 27, 496–508. https://doi.org/ commitment therapy. Measures package. 10.1521/soco.2009.27.4.496 Schönbrodt, F. D., & Perugini, M. (2013). At what sample size do correlations stabilize? Hoenig, J. M., & Heisey, D. M. (2001). The abuse of power: The pervasive fallacy of Journal of Research in Personality, 47, 609–612. https://doi.org/10.1016/j. power calculations for data analysis. The American Statistician, 55, 19–24. https:// jrp.2013.05.009 doi.org/10.1198/000313001300339897 Tapper, K., Pothos, E. M., & Lawrence, A. D. (2010). Feast your eyes: Hunger and trait Imhoff, R., Barker, P., & Schmidt, A. F. (2019a). To what extent do erotic images elicit reward drive predict attentional bias for food cues. Emotion, 10, 949. https://doi. visuospatial vs. cognitive attentional processes? Consistent support for a (non- org/10.1037/a0020305 spatial) Sexual Content Induced Delay. Archives of Sexual Behavior. https://doi.org/ Tversky, A. (1977). Features of similarity. Psychological Review, 84, 327–352. https://doi. 10.1007/s10508- 019-01512-0 org/10.1037/0033-295X.84.4.327 Imhoff, R., Barker, P., & Schmidt, A. F. (2019b). To what extent do erotic images elicit Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor package. visuospatial vs. cognitive attentional processes? Consistent support for a (non- Journal of Statistical Software, 36(3), 1–48. https://doi.org/10.18637/jss.v036.i03 spatial) Sexual Content Induced Delay. Archives of Sexual Behavior. https://doi.org/ Vogt, J., De Houwer, J., Moors, A., Van Damme, S., & Crombez, G. (2010). The automatic 10.1007/s10508- 019-01512-0 orienting of attention to goal-relevant stimuli. Acta Psychologica, 134, 61–69. https://doi.org/10.1016/j.actpsy.2009.12.006 16