Abstract
Previous work has shown that compared to passive perceivers who view preselected information about target persons, active perceivers are less confident in their impressions, do not show increased confidence with increased amounts of information, and like targets less. The authors now explain these findings, postulating that perceivers without control over the amount of information they receive should be motivated to form impressions earlier, altering their information-processing strategies. Study 1 predicted and found that content-only active perceivers who control the content, but not the amount, of information show the same positive relationship between confidence and amount of information as passive perceivers, as well as the same reading-time patterns and level of liking. Study 2 used clearly valenced target stimuli and found support for the hypothesis that passive perceivers form more extreme early impressions, leading to greater liking when early information is positive but less liking when it is negative.
Person perception studies almost invariably place participants in a passive role, forming impressions of target individuals based on a fixed, preselected body of information such as photos, videos, or written behavior descriptions. Although these perceivers can actively attend to and cognitively interpret the available information, perceivers in everyday life exert considerably more control over the content and amount of information they receive about impression targets (Smith & Collins, 2009). For example, when reading about a potential new beau on a dating website you can decide what type of information you want to seek out (is he a dog or cat person? Democrat or Republican?) and how much information you need to form an adequate impression.
Waggoner, Smith, and Collins (2009) recently examined differences between impressions formed under active versus passive conditions. In their study, active and passive participants formed impressions of target individuals based on information from Facebook profiles. Active perceivers chose which items from Facebook they wanted to view, as well as how many, before reporting their impression. Yoked-passive perceivers had no control but were shown the same items chosen by the active perceivers, in the same order. Although active and passive perceivers did not differ in the accuracy of their judgments, their impressions did differ in a number of ways. Active perceivers reported lower confidence and ease in their judgments, compared to passive perceivers. The active perceivers failed to exhibit increasing confidence with increased information, while the passive perceivers became more confident in their impressions as they obtained more information. The latter pattern replicates previous studies, which (as is typical) used passive perceivers (Gill, Swann, & Silvera, 1998).
Finally, active perceivers reported liking the targets less than passive perceivers. Several potential explanations for this difference were ruled out by the data. The first possibility was that the active perceivers engaged in valence-dependent sampling, seeking additional information about liked targets, while breaking off the information search for disliked targets (Denrell, 2005). Such a process would create a negativity bias for active but not passive perceivers. However, it would also generate a positive relationship between liking and amount of information viewed for the active perceivers, which was not found, making this explanation untenable. A second possibility was that the passive perceivers misattributed their judgmental fluency (confidence and ease) as liking (Winkielman, Schwarz, & Nowak, 2002). However, the difference in liking was not mediated by confidence or ease, and it persisted even when ease and confidence were statistically controlled. Finally, passive perceivers (without the additional cognitive load of choosing information) might be more sensitive to the actual likeability of the profiles. Because people are generally motivated to appear likeable on Facebook, this could result in greater liking by the passive perceivers. However, the difference in liking between active and passive perceivers was consistent across the range of profile likeability—rather than being larger for the more likable profiles as predicted by this differential accuracy hypothesis. Thus, our previous study ruled out several possibilities but failed to document a reason for the greater liking by the passive perceivers.
We now propose an explanation for all these findings. The key idea is that passive and active perceivers use different information processing strategies. Passive perceivers, uncertain about how much information they are going to receive, should be motivated to construct an impression early, because any given piece of information might be their last. Once they form an impression, they would enter a confirmatory processing mode, assimilating any additional information into this already-formed impression—assuming this is possible, for example, when the information is ambiguous or consistent with the existing impression. This assumption parallels research on deliberative versus implemental mindsets by Gollwitzer and colleagues (e.g., Gollwitzer, 1990; Gollwitzer, Heckhausen, & Steller, 1990). Once passive perceivers have made a decision about their impression of this target, they shift from relatively open, deliberative processing in which they carefully consider the presented information, to biased postdecisional processing that treats further information as confirmatory. Such processing would increase confidence without necessarily improving the accuracy of impressions, resulting in a positive relationship of confidence to the number of items viewed for passive perceivers.
In contrast, active perceivers who control the amount of information would not feel pressure to form an impression early on, for they can always collect more information if needed. Rather than assimilating new information into an already-formed impression, they would use later items to adjust and refine their impressions, remaining in a predecisional mindset for a longer period of time. Thus, early information would not have the extra weight it has for passive perceivers. Active perceivers would continue to view information about each target until reaching their own subjective threshold level of confidence in that impression (the sufficiency principle; Chaiken, Liberman, & Eagly, 1989). Since this threshold would be expected to be constant across all targets, ratings of confidence should be relatively similar regardless of the number of items viewed. For example, whether a perceiver’s threshold level of confidence is reached after viewing 3 items or 12 for a given target, the final level of confidence would be roughly the same, resulting in a null relationship between confidence and amount of information.
Our theory implies that control over the amount of information, rather than the content of information, drives the observed differences between active and passive perceivers. Of course, control over the content of information could also influence information-processing strategies. For example, active but not passive perceivers can use a hypothesize-test strategy (“I bet she likes romantic comedies—I’ll look at her Movies section next.”). But such a strategy might increase confidence among hypothesis-testing active perceivers as they gathered increasingly more evidence that they see as relevant to their hypotheses—inconsistent with our findings.
Thus, our assumption about divergent information-processing strategies resulting from differences in control over information amount can account for the increase in confidence as more items are viewed for passive (but not active) perceivers. If one additional assumption is added, this postulated difference in processing strategies can also explain the differences in liking. Passive perceivers’ greater weighting of early information means that their liking for targets (compared to that of active perceivers) would be based more strongly on the valence of the early encountered information about each target. Since people are motivated to present themselves favorably on Facebook (Strano, 2008; Zhao, Grasmuck, & Martin, 2008), such an effect could explain why passive perceivers like Facebook targets more than active perceivers. Importantly, perceivers tend to look at more detailed, information-rich Facebook items first (such as About Me, an open-ended self-description, the profile picture, and lists of interests and activities; Waggoner et al., 2009). These items are precisely those that one would expect to be interpreted most positively, compared to less meaningful later-viewed items such as the number of wall posts the person has. Thus, the greater liking of the targets by passive perceivers may be specific to largely positive Facebook stimuli, rather than reflecting a general tendency for passive perceivers to like targets more regardless of the nature of the stimuli. This issue is examined directly in study 2, where we use experimenter-constructed and clearly valenced stimuli rather than Facebook materials.
Study 1
Study 1 was designed to test three hypotheses based on our theoretical reasoning, as well as to replicate our previous findings. We predict that control over the amount rather than the content of information is responsible for the differences between active and passive perceivers. Thus, besides replicating our previous active and passive conditions, the study included a new condition: content-only active perceivers, who control information content but not amount. We predicted first that content-only active perceivers should show the same increase of confidence with information amount as passive, rather than full-active, perceivers. Second, we predicted that the liking difference previously observed with full-active versus passive perceivers would be eliminated for the content-only active perceivers. Third, to provide direct evidence for our hypothesized information-processing difference, we measured the amount of time spent reading each item. We predicted that perceivers without control over the amount of information should spend less time viewing later items than perceivers with control over the amount.
Method
Participants
A total of 258 Indiana University undergraduates (155 female) participated for experimental credit. Active participants were randomly assigned to the full-active or content-only active condition. Passive participants were randomly yoked to an active participant, creating 63 full active–passive pairs and 66 content-only active–passive pairs.
Targets and Stimuli
Volunteer targets consented to research use of their Facebook profiles (including photos but without names or other identifying information), and completed self-report questionnaires in exchange for a chance to win a $50 prize. We used the same 30 profiles as Waggoner et al. (2009), and each profile was divided into 15 segments: About Me (an open-ended self-description), activities, books, current status, groups, interests, movies, music, number of friends, number of photos, number of status updates, number of wall posts, profile picture, quotes, and TV shows. These 30 profiles were divided into three sets of 10, and each active participant was randomly assigned to one of these sets.
Procedure
After providing informed consent, participants completed the study individually. Instructions familiarized participants with the information items available for each target. Active participants were told that they would see a menu of these items for each target. Full-active perceivers were told they would be able to view as many or as few of the items as they wished (with a minimum of one item), in any order, before providing their impression of the target. Content-only active perceivers were told that they would be able to choose which item they wished to view first (and second, and so on), but they would not know in advance how many items they would get to view for each target. For these participants, the number of items shown for each target was randomly selected by the computer (constrained to match the distribution from Waggoner et al., 2009). All passive participants were told that they would see items whose number and order would vary from target to target, before providing their impression.
After providing demographic information participants began with their first target. Active participants viewed up to 10 targets (the session was terminated after an hour). Passive participants viewed the same targets and items, in the same order, as their yoked-active participant. Both passive and active perceivers were allowed to view each item for as long as they wished, and we recorded their viewing time. 1 For each target, participants completed the measures described below. After rating all targets, participants were debriefed, thanked, and dismissed.
Measures
Big Five
Targets were rated on the Big Five personality traits (extraversion, agreeableness, conscientiousness, neuroticism, and openness) using the 10 Item Personality Inventory (Gosling, Rentfrow, & Swann, 2003). Targets had previously rated themselves on the same inventory.
Politics and religion
Participants estimated each target’s political ideology on a scale from 1 (strongly conservative) to 7 (strongly liberal). Participants also rated how important they felt religion was to the target on a scale from 1 (extremely important) to 4 (not at all important). The targets also completed these items.
Other ratings
Participants indicated how confident they were in their judgments of the target, how easy it was for them to form their impression of the target, how much they liked the target, and how similar they were to the target, all on 1–7 scales. (No differences were predicted or found on the similarity measure, which will not be mentioned further.)
Results
Items Viewed
Full-active participants viewed a mean of 5.9 items (SD = 3.16) out of 15. Content-only active participants viewed a mean of 8.3 items (SD = 4.16) out of 15 (this distribution was set to approximate that found in Waggoner et al., 2009, which had a mean of 8.04). It is unclear why full-active participants chose to view fewer items. However, the types of items chosen were similar across conditions and replicated findings from Waggoner et al. (2009). For example, About Me was selected the most frequently, and often first.
Accuracy
Our analyses of accuracy followed Gosling, Gaddis, and Vazire (2007). Mean ratings of each profile were calculated for participants in both active and corresponding passive conditions, and correlated (across the 30 profiles) with the targets’ self-reports. As shown in Table 1, these correlations fell within the same range for each trait across our previous study and both conditions of this study, although the smaller Ns in this study mean that some correlations are not significant. As predicted, participants were most accurate for extraversion, political affiliation, and religiosity. Most importantly, and replicating Waggoner et al. (2009), there are no notable differences in accuracy between active and passive perceivers in any condition; the mean correlations are .11 for actives and .14 for passives.
Correlation Across Profiles (N = 30) of Target Self-Reports With Aggregated Participant Ratings
*p < .05.
**p < .01.
***p < .001.
Judgmental Confidence and Ease
We first compared the full-active and yoked-passive participants to see whether results for confidence and ease replicated those found in previous research. Paired t-tests (with the subject pair as the unit of analysis) comparing full-active and passive participants ratings of judgmental confidence (active: M = 4.61; passive: M = 4.65) and ease (active: M = 4.53; passive: M = 4.45) failed to reach significance (both ps > .05). However, the relationship between confidence and amount of information replicates the pattern found in our previous work. We used a hierarchical linear model analysis with targets as Level 1 and perceivers as Level 2. As shown in Figure 1, the number of items viewed has virtually no relationship to confidence for active perceivers, b = –.03, t(525) = –1.24, ns, but a positive relationship for passive perceivers, b = .08, t(527) = 3.24, p < .01, and the difference between these two slopes is significant, b = –.10, t(525) = 2.90, p < .01. A similar pattern is found for ease of judgment: number of items viewed is negatively related to ease for active perceivers, b = –.06, t(525) = –2.35, p < .02, but has a significant positive relationship for passive perceivers, b = .08, t(527) = 3.19, p < .01, and the difference between these two slopes is highly significant, b = –.14, t(525) = 4.00, p < .001. These relationships for confidence/ease replicate the pattern found in Waggoner et al. (2009). The smaller number of items viewed in this study (averaging 5.9 vs. 8.3) explains why there is no overall difference between the active and passive perceivers.
Relationship between confidence and the number of items viewed for each perceiver condition.
Our main interest was in comparing content-only active and yoked-passive participants. Compared to the content-only active perceivers (confidence: M = 4.19; ease: M = 4.21), passive perceivers were more confident in their impressions, M = 4.57, t(655) = 4.81, p < .0001, and found them easier to form, M = 4.52, t(655) = 3.87, p < .001. However, as predicted, the relationship between confidence and ease and the number of items viewed changes, when control over the amount of information is removed. As can be seen in Figure 1, there is a positive relationship between confidence and number of items viewed for both the content-only active participants, b = .09, t(560) = 6.75, p < .0001, and the passive participants, b = .05, t(561) = 4.42, p < .0001. The same pattern is found for ease (content-only active: b = .11, t(560) = 8.55, p < .0001; passive: b = .07, t(561) = 5.75, p < .0001). Thus, consistent with our first hypothesis, content-only active participants resembled passive participants in showing a positive relationship of confidence/ease to the number of items viewed, demonstrating that control over the amount rather than content of information is responsible for this active–passive difference.
Liking
Paired t-tests comparing the full-active and yoked-passive participants replicated our previous finding that the passive perceivers (M = 4.33) liked the targets more than the active perceivers, M = 4.06, t(617) = 3.73, p < .001. Consistent with our second hypothesis, this difference in liking disappears, when comparing the content-only active (M = 4.08) and yoked-passive participants, M = 4.0, t(655) = 1.19, ns. As in our previous research, liking was not significantly related to the number of items viewed for any type of perceiver.
Item Viewing Time
Our third hypothesis is that because content-only active and passive participants switch to a relatively shallow, confirmatory processing mode, they should spend relatively less time viewing later items than the active participants. We examined the relationship between reading time for the last item viewed and the number of items viewed. As shown in Figure 2, this relationship was flat for the active perceivers (b = –13.67, ns), and negative for the three groups of participants who did not control information amount; passive: b = –402.12, t(518) = –3.49, p < .001; content-only active: b = –161.32, t(560) = –7.73, p < .001; content-only passive: b = –248.51, t(561) = –4.62, p < .001. Additionally, there is a significant difference between the slopes of the active and passive participants, b = 394.81, t(516) = 3.11, p < .01, but not between content-only active and yoked-passive participants (b = 91.14, ns). Thus, as predicted, those without control over the amount of information spent less time viewing the final item with increasing numbers of items viewed, a pattern that is not shown by those with control over the amount of information.
Relationship between the time spent viewing the last item and the number of items viewed for each perceiver condition.
For both sets of passive participants, we were also able to examine the relationship between the number of items viewed and the mean viewing time for all items (see Note 1). As predicted, the relationship is negative in both cases, passive: b = –312.28, t(518) = –3.73, p < .01; content-only passive: b = –82.84, t(561) = –2.47, p = .01, demonstrating that these perceivers spend less time viewing later items, and providing further support that they shift to a quick, cursory mode of information processing.
Discussion
Results from the replication conditions are generally consistent with Waggoner et al. (2009), with passive (but not active) perceivers exhibiting increased confidence and ease as they view more items, and passive perceivers liking the targets more than active perceivers. Although the mean difference in confidence/ease between the active and passive perceivers failed to reach significance due to a smaller number of items viewed overall, the relationship to number of items viewed replicates the pattern from the previous study. More important, results from the new conditions (with active perceivers who control information content but not amount) show that control over the amount of information is key. The content-only active perceivers appear to be forming impressions as passive perceivers do, showing increased confidence and ease with greater amounts of information and reporting similar levels of liking toward the targets, consistent with our first two hypotheses. In addition, content-only active and passive perceivers spend less time reading later-viewed items, supporting our third hypothesis that these perceivers shift to a quick, confirmatory type of processing, when presented with additional information.
As we argued above, for generally positive Facebook stimuli, the shift to confirmatory processing by perceivers without control over the amount of information may explain the liking difference between active and passive conditions. Passive perceivers would be expected to give more weight to early information relative to later information, so compared to the active perceivers, their impressions should be more strongly influenced by early-encountered positive information.
Study 2
Study 2 was designed to directly test a fourth hypothesis deriving from the theory that passive perceivers weight early information more heavily in their judgments: passive perceivers should show greater liking when early information is positive but less liking when it is negative. Rather than Facebook stimuli, participants viewed clearly valenced behavioral sentences for each target. Active perceivers controlled the number of sentences they viewed for each target (but not their content). For each target, the initial behaviors were either positive or negative, and then the valence reversed. We predicted that passive participants would judge the positive-first targets more positively and negative-first targets more negatively than the active participants but only when the number of items viewed is small. When larger number of items is viewed, we expected the liking difference to disappear, because the later behavioral information differs unambiguously in valence from the initial information (unlike our studies with Facebook stimuli) and thus cannot readily be assimilated to an early-formed impression.
Method
Participants
A total of 62 Indiana University undergraduates (26 female) participated for experimental credit and were randomly assigned to either the active or passive condition. One active–passive pair was removed due to a data-collection error, leaving 30 active–passive pairs.
Targets and Stimuli
Perceivers saw a photo and read manipulated behavioral information for eight targets (four male, four female). A total of 15 sentences were constructed for each target. Each was unambiguously positive or negative (e.g., Matt volunteered at the hospital), and behaviors for each target reversed in valence after the first three to five sentences. For half of the targets, positive information was presented first, and for half of the targets negative information was presented first. Participants also viewed two neutral filler targets, which were not analyzed.
Procedure
Active participants were told that they would be able to view as many short behavioral descriptions as they wished (up to 15) before reporting their impression; a menu screen helped them keep track of how many behaviors they had viewed. Passive perceivers were yoked to an active perceiver and were informed that there were 15 possible behaviors available for each target, but that the number of behaviors they would see would vary from target to target. Reading time was fixed at 5 s per behavior for all participants. After viewing the items for each target, participants completed the same impression measures as Study 1.
Results
Data were analyzed using multilevel regression, with target photo and subject entered as random factors. Perceiver condition (active or passive) was a between-subjects factor and target type (positive-first or negative-first) a within-subjects factor. Number of items viewed (determined by the active perceivers) also varied within subjects. This continuous variable was recoded into “three or fewer items” (N = 149) versus “four or more items” (N = 332) and entered as a dichotomous factor in the analyses below. This split ensured that those categorized as viewing a small number of items for a particular target were seeing only items that were evaluatively consistent.
The key dependent measure in this study is liking for the targets. We regressed liking onto perceiver condition, target type, and the number of items viewed (three or fewer vs. four or more). As shown in Table 2, the three-way interaction was significant, t(472) = –1.99, p < .05. The pattern of means for the interaction was as predicted (see Figure 3): The difference in liking for the positive-first and negative-first targets is larger for the passive participants (difference = 2.13) than for the active participants (difference = 1.44) but only when the number of items viewed is small. When more items are viewed the differences are both trivial (.09 for passives, .21 for actives).
Regression Coefficients (and Standard Errors) for all Main Effects and Interactions of Perceiver Type, Target Type, and Number of Items Viewed
*p < .05.
**p < .01
***p < .001.
Mean liking toward the targets across all conditions.
Discussion
As predicted, when viewing only the early evaluatively consistent items, passive perceivers form more extreme judgments, liking positive targets more and negative targets less than active perceivers. However, when evaluatively mixed information is received about a target, passive perceivers are unable to assimilate later information into their impressions and liking differences disappear. Active perceivers in this study control only the amount of information they receive, providing further evidence that control over the amount, and not the content, of information drives the observed differences between the active and passive perceivers.
General Discussion
Two experiments illuminated the mechanisms underlying the previously demonstrated differences between the judgments made by active and passive perceivers (Waggoner et al., 2009). We argued that differences in perceived confidence and ease were driven by control over the amount (rather than the content) of information, which leads perceivers to adopt distinct information-processing strategies. Passive perceivers (who lack any control) and content-only active perceivers (who control information content but not amount), should try to form impressions early and then switch to more superficial, confirmatory processing for subsequent stimulus items (cf. Gollwitzer, 1990). This difference in processing modes is supported by the successful test of our first hypothesis: for all except full-active perceivers, impression confidence increases as more items are viewed. The second hypothesis was also supported: the liking difference between active and passive conditions disappeared, when comparing content-only active and passive perceivers. For our third hypothesis, the analysis of item viewing times provided direct evidence that passive and content-only active perceivers devoted less time to each item as the number of items increased, while active perceivers did not. Finally, a second study using clearly valenced behavioral stimuli supported a fourth hypothesis: passive perceivers weight early information more heavily in their judgments, resulting in greater liking if early information is positive but less liking if the information is negative (compared to active perceivers).
Confirmatory Processing and Confidence
Previous work has shown that perceivers tend to feel more confident about their impressions if they receive greater amounts of target information, as this allows perceivers to form relatively richer mental representations (Gill et al., 1998). However, this previous work used a standard impression formation paradigm with passive perceivers lacking control over the target information they received. Our work shows that perceivers who can choose the amount of information they receive about a target may not exhibit such increases in confidence. Since increased confidence is not associated with increased accuracy (in our studies, or in Gill et al., 2001), this implies that perceivers are miscalibrated. Either perceivers with control over information amount are underconfident in their impressions or perceivers lacking such control (notably including passive perceivers in typical person-perception studies) are overconfident.
Methodological Implications
This work demonstrates that one cannot always generalize directly from lab studies employing passive experimental paradigms to less constrained person perception processes in everyday life. Previously documented relationships of confidence to amount of information vanish, when perceivers are given control over the amount of information they view, and active and passive perceivers show different calibrations of confidence with judgmental accuracy. Even the degree of liking based on exactly the same person information can differ. In our studies using Facebook materials, we find that passive perceivers like targets more than active perceivers. However, when using experimenter-constructed sentences, we predicted and found that the direction of the liking difference depends on the valence of the early information received.
Theoretical Implications
With few exceptions (e.g., Snyder & Swann, 1978), little research examines active information search and selection in person perception. However, active person perception opens the door for many interesting processes. For example, perceivers who can control information amount may engage in valence-dependent sampling (ending interactions with dislikeable targets), which can create a negativity bias in impressions (Denrell, 2005). Control over other aspects of information collection (such as the content or sources of information) may also have important consequences for impression formation, although these processes remain largely unexplored in the literature (Smith & Collins, 2009).
Conclusion
Person impressions formed by active and passive perceivers differ in important ways, many depending on control over the amount of information received about a target. In everyday life, we often control the amount of information we receive about someone—we can leave the conversation, turn off the TV, stop following the person on Twitter, and so on. But in the lab, how often do researchers tell participants to “read as much as you want?” These findings emphasize the importance of exploring the processes that operate as impression formation unfolds in everyday life.
Footnotes
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by NSF grant BCS094911.