Attitudes and Social Cognition: Vertical Positions as Perceptual Symbols of Power PDF

ATTITUDES AND SOCIAL COGNITION Your Highness: Vertical Positions as Perceptual Symbols of Power...

ATTITUDES AND SOCIAL COGNITION Your Highness: Vertical Positions as Perceptual Symbols of Power Thomas W. Schubert University of Jena Metaphorically, power equals up. Drawing on embodied theories of cognition, the author argues that thinking about power involves mental simulation of space and can be interfered with by perception of vertical differences. Study 1 assessed image schemas for power and found a shared vertical difference This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. metaphor. Studies 2, 3, and 4 showed that the judgment of a group’s power is influenced by the group’s This document is copyrighted by the American Psychological Association or one of its allied publishers. vertical position in space and motor responses implying vertical movement. Study 5 ruled out that the influence of vertical position on power judgments is driven by valence differences. Study 6 showed that vertical position also influences the power judgment result itself. The evidence suggests that the concept of power is partly represented in perceptual form as vertical difference. Keywords: power, embodiment, perceptual symbols, metaphors When we talk about power, we often use metaphors about up ingly abstract concepts like power has to do with perception and and down. Examples abound: When someone has a high status, or experience. is up in the hierarchy, he or she has control over and can oversee others who have lower status. One can look up to those who rose Perceptual Symbols to the height of their power or look down on underlings. When a Perceptual content is usually neglected in current theories of picture of a hierarchy is drawn, the most powerful person is usually mental representation. Instead, mental representations are typically at the top, and the subordinates are drawn below. All these met- described as nodes that are abstract and amodal, which are con- aphors are cases of the “control is up, lack of control is down” nected to each other via associative links. The idea is that knowl- metaphor (Lakoff, 1987; Lakoff & Johnson, 1980). In short, power edge is represented in the nodes and their connections. Perceptual is metaphorically described as a vertical dimension in physical content is only an early input to the representations of concepts in space. these models and not regarded as important for conceptual think- In psychological research, however, power is usually defined as ing itself. As an alternative to such amodal models of mental the potential to influence others and to promote one’s own goals representation, recent theorizing on the nature of human knowl- (Ellyson & Dovidio, 1985; Keltner, Gruenfeld, & Anderson, 2003; edge, categories, and cognition proposes that concepts are a lot less McClelland, 1975). Although spatial positions have been consid- abstract than previously thought. Embodied theories of cognition ered insofar as they are cues to the power of others in the envi- argue that concepts include a lot more perceptual content than a ronment (Argyle, 1988; Mehrabian, 1972), we do not know what clear divide between perception, cognition, and behavior would role they play in thinking about power. Are we missing something suggest (Barsalou, 1999; Glenberg, 1997). Empirical work, both important? Is it the case that metaphors about the vertical dimen- from cognitive psychology and neuropsychology, supports this sion are used only to talk about power but not to think about it? idea. In a review of evidence from several domains, Barsalou Should social psychology leave metaphors to linguistics? The (1999) found that mental representations of concepts are still tied following arguments and data will try to show that by taking to their perceptual basis. Barsalou concluded that knowledge con- metaphors literally, we can discover what thinking about seem- sists not of amodal propositions but of modality-specific or modal representations, which he calls perceptual symbols. Perceptual symbols are thought of as schematized perceptual experiences This research was partly funded by a research grant awarded to Thomas involving all senses, including proprioception, introspection, and W. Schubert by the Friedrich-Schiller-Universität Jena. I thank Markus motor programs. These schematized perceptual representations are Brauer, Steffen Gießner, Arthur Glenberg, Jonathan Haidt, Johann Jacoby, then used in cognitive processes, such as perception, categoriza- Thomas Kessler, Sander Koole, Amélie Mummendey, Kai Sassenberg, tion, and judgment. These processes are made possible by using Beate Seibt, and Sven Waldzus for helpful discussions and comments on earlier versions of this article, and Babett Voigt, Tobias Raabe, Susann the perceptual symbols to construct and run simulations, similar to Schmidt, Lisa Schubert, and Anna Steudel for collecting data. mental models (Johnson-Laird, 1983). In other words, thinking is Correspondence concerning this article should be addressed to Thomas W. argued to involve perceptual simulation. Schubert, Department of Psychology, University of Jena, Humboldtstraße 26, It is interesting to note that an embodied view of cognition is 07743, Jena, Germany. E-mail: [email protected] compatible with recent findings and theoretical developments in Journal of Personality and Social Psychology, 2005, Vol. 89, No. 1, 1–21 Copyright 2005 by the American Psychological Association 0022-3514/05/$12.00 DOI: 10.1037/0022-3514.89.1.1 1  2 SCHUBERT social psychology (cf. Barsalou, Niedenthal, Barbey, & Ruppert, Yaxley, 2003; see also Richardson, Spivey, Barsalou, & McRae, 2003; Niedenthal, Barsalou, Winkielman, Kraut-Gruber, & Ric, in 2003), perception interacts with these mental representations. press). For example, the activation of a social stereotype has been shown to cause the unintended mimicry of behavior that is strongly Perceptual Simulation of Space for Abstract Concepts associated with the stereotyped group (e.g., mimicry of the slow walking associated with the category of the elderly; Bargh, Chen, These results certainly encourage embodied views of knowl- & Burrows, 1996; Dijksterhuis & Bargh, 2001). Such automatic edge; however, they are all related to concrete objects that can behavior is explained by assuming that the mental representation actually be observed in the environment. It is interesting to note of a group stereotype contains sensory-motor representations. This that there is also evidence for perceptual simulation of space in theoretical framework complements quasi-verbal associative net- thinking about abstract concepts, especially about valence and time works with modal content. Further examples are findings showing (for further concepts, see Talmy, 1988). that the processing of valence information is hindered by perform- Studying the embodiment of valence, Meier and Robinson ing behavior that does not fit the valence (e.g., approach vs. (2004) showed that a vertical spatial dimension underlies valence representations. They tested this by letting participants evaluate This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. avoidance movements, or nodding vs. head shaking; Förster & This document is copyrighted by the American Psychological Association or one of its allied publishers. Stepper, 2000; Förster & Strack, 1996; Neumann, Förster, & words that appeared at either the top or the bottom of a computer Strack, 2003). Such bodily feedback effects are explained by screen. As predicted from a perceptual symbols account, positive arguing that motor representations interact with the cognitive words were evaluated quicker when they appeared at the top of the processing involved in the cognitive tasks, which can be explained screen compared with the bottom of the screen, whereas the only if cognitive processing involves perceptual simulation. In opposite was true for negative words. Apparently, judging valence sum, both theoretical developments and empirical evidence in involves simulation of a vertical spatial dimension, on which good cognitive and social psychology suggest that modal content plays is up and bad is down. a role in conceptual thinking. This encourages the search for The hypothesis that time is thought about by imagining it as a space perceptual content in such social concepts as power. was already put forward by Jaynes (1976) and is supported by the existence of metaphors that we use to talk about time. The best experimental evidence comes from a set of ingenious experiments by Perceptual Simulation of Space for Concrete Concepts Boroditsky and colleagues (Boroditsky, 2000, 2001; Boroditsky & Ramscar, 2002; Gentner, Imai, & Boroditsky, 2002). Time is imag- As the current goal is to investigate the spatial metaphor of ined as a horizontal, not vertical, space, at least in Western cultures. power, most interesting to the present discussion are results that Consequently, Boroditsky (2001) could show that judgments about illustrate perceptual simulation of space in thinking about con- temporal facts (e.g., whether January comes before June) are made cepts. Indeed, there is evidence that points to a simulation of space faster when a preceding judgment has to be made about a horizontal in conceptual thinking. In one instructive study, Borghi, Glenberg, spatial dimension than when it has to be made about a vertical spatial and Kaschak (2003) asked their participants to verify whether a dimension. Time, an abstract dimension, is thought about in terms of given object (e.g., a car) had a certain part (e.g., a roof). To answer, spatial—that is, perceptual—terms. participants had to press one of two keys on a vertically mounted These data support the notion that abstract concepts, like con- keyboard. Affirmative answers had to be given either with a key crete concepts, are at least partially represented by perceptual that required an upward movement of the participant’s arm or with symbols that relate the concepts to perceptual content. The con- a key that required a downward movement. Results showed that if clusion is that conceptual thinking involves the simulation of this there was a fit between the position of the part relative to the object content and can therefore be influenced by priming or concurrent (i.e., roof is at the top of the car) and the movement of the arm (i.e., presentation of perceptual input. upward), reactions were quicker. This supports the idea that con- ceptual thinking involves perceptual simulation: When we want to Perceptual Simulation of Space for Power verify the property of a category, the concept is simulated mentally on the basis of perceptual knowledge, and the result is read off. These examples show that thinking about both concrete and Actual motor movement can either interfere with or facilitate this abstract concepts can be influenced by spatial information that is simulation (cf. Glenberg & Kaschak, 2003). canonically included in the construction of these concepts— either If a visual simulation of the words’ referents underlies these in the real environment or in metaphorical thinking. This supports effects, then spatial information provided by vision should also be the idea that power, an abstract social concept, also includes spatial able to interfere with or facilitate the simulation. This is what information about the vertical dimension, as the metaphors sug- Zwaan and Yaxley (2003) showed. They presented two words gest. In other words, the hypothesis is that when we think of power simultaneously, and the participants’ task was to judge whether the differences, we actually think of spatial differences. two were related or not (e.g., root and branch). The words were Barsalou (1999) proposed that a perceptual symbol is derived presented above each other, and the crucial manipulation was that from multiple sources of direct experience. He exemplified this for their order either followed the canonical arrangement (i.e., branch anger: To represent it, experiences of goal blocking, intense affec- above root) or contradicted it (i.e., root above branch). As pre- tive states, and behavioral responses are schematized. The same dicted by a simulation account, relatedness affirmations were reasoning can be applied to mental representations of power. Here, quicker when the arrangement of the words followed the canonical direct physical experience of vertical differences might be sche- arrangement of the objects. By providing additional perceptual matized into a perceptual symbol of power. Indeed, the linguists input, either motor (Borghi et al., 2003) or visual (Zwaan & Lakoff and Johnson (1980) listed this image as one of the “met-  VERTICAL POSITION AND POWER 3 aphors we live by,” and argued that the experiential basis lies in the preconditions for the schematization of a perceptual symbol out of fact that “physical size typically correlates with physical strength, a multitude of perceptual events and for the development of a and the victor in a fight is typically on top” (p. 15). As they keenly strong association between vertical position and power (Barsalou, observed, there are two different spatial correlates of power: ver- 1999). tical size (height) and vertical position. The association of power The logic behind this schematizing of experiences can be made and height is important for most mammals that are involved in clearer by juxtaposing metaphors that refer to a perceptual symbol physical fights: The larger animal is typically more powerful, or as with those that do not. Not every metaphor is still “alive” and tied Freedman (1979) noted, “throughout nature the rule is the bigger, to actual experiences; some are learned and used without referring the more dangerous” (p. 29). For humans, size matters a lot for to perceptual content. One example of the latter type is the meta- negotiating power relations, especially during childhood and ado- phor that conservative parties are “right,” whereas liberal parties lescence. Children learn that their taller parents are more powerful are “left.” These labels originally referred to an actual seating and that taller siblings or other taller children are able to coerce order in the 1789 French National Assembly—that is, an actual them physically. Schwartz, Tesser, and Powell (1982) noted that a experience of a spatial order. Although many modern parliaments This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. “universal association of statural superiority and parental domi- maintain this seating order, laypersons today probably do not share This document is copyrighted by the American Psychological Association or one of its allied publishers. nance” exists, which leads to an “invariant use of elevation sym- this experience enough for it to be schematized, and therefore it is bolism in the representation of social dominance as a generaliza- likely that today only the spatial labels “left” and “right” are tion of this elementary facet of experience” (p. 119; cf. Argyle, associated with conservatism and liberalism, but not the actual 1967; Spiegel & Machotka, 1974). This carries over into adult- spatial representation. If the current hypothesis on power is cor- hood, where it is still the rule that taller persons use their physical rect, however, the powerful ⫽ up and powerless ⫽ down percep- advantage to gain power: Big people hit little people (Felson, tual symbol is different from the conservative ⫽ right and lib- 2002), and taller persons enjoy gaining higher wages, reaching eral ⫽ left metaphor in that in the case of power, everybody shares higher status occupations, and (at least on average) winning pres- the underlying experience that then becomes schematized. If ver- idential elections (Young & French, 1996; for a review, see Judge & Cable, 2004). Finally, according to Keltner and Haidt (2003), tical positions are perceptual symbols of power, thinking about size or vastness is a central feature of the emotion awe, which, in power should be influenced by the perception of vertical spatial its prototypical form, is felt toward powerful others. Perceiving differences. This hypothesis is tested in the following studies, something that is “much larger than the self, or the self’s ordinary examining judgments of power as one type of thinking about level of experience or frame of reference,” (Keltner & Haidt, 2003, power. p. 303), together with the experience that one needs to accommo- date one’s concepts in response to the event, leads to experiences Overview of the Current Research of awe (see also Haidt, 2003). For a relation between power and vertical position, there is The following six studies were designed to test the hypothesis plenty of evidence from the anthropological literature, although the that the concept of power involves a perceptual simulation of difference between vertical position and vertical size is sometimes vertical differences in space. Study 1 assessed image schemas held fuzzy. Often, vertical position acts as a surrogate of physical size. by laypersons for propositions that describe power and powerless- People construct power as vertical difference in language, in ness. Study 1’s goal was to verify that these metaphors are sche- nonverbal communication, and in physical manifestations. Fiske matic and shared. Studies 2, 3, and 4 investigated whether per- (1992, 2004) has described extensively how authority ranking, the ceived vertical differences interfere with judgments of power. In social and cognitive construction of who has power over whom, these studies, predictions of the perceptual symbols hypothesis relies on vertical spatial difference. He observed that in order to were tested by investigating whether judgmental speed and accu- differentiate between persons with and without power, people racy can be influenced by the spatial position of the groups to be typically use metaphors of spatial order and magnitude. Citing judged. In Study 2, pairs of groups were used as stimuli, and both evidence from diverse cultures, Fiske concluded that virtually all visual input and the motor response varied such that they were cultures use vertical markers for authority ranks in their language compatible or incompatible with the power relations. Study 3 (for further references, see Schwartz, 1981; Schwartz et al., 1982). Furthermore, verticality embodies power almost everywhere in the isolated the motor response, and Study 4 isolated the visual input, domains of posture, housing, and furniture (Hewes, 1955). Many with the common goal of testing whether each alone can interfere of these manifestations use height and size simultaneously or with power judgments. By providing pretest and manipulation interchangeably, such as when the powerful have the largest house check data on the valences of powerful and powerless groups, with the highest tower, sit at an elevated seat during meals, and are Studies 2– 4 also addressed an alternative explanation, namely, that addressed as “Your Highness.” powerful agents are associated with up because they are positive Taken together, extensive evidence from diverse sources sug- and positive valence is associated with up (Meier & Robinson, gests that vertical space easily affords the linear ordering necessary 2004). In addition, Study 5 tested experimentally whether valence for the creation of a hierarchy. Human children are surrounded by predicts interference judgments on power judgments or only on a world in which power is over and over again correlated with valence judgments. Finally, Study 6 went beyond influencing the vertical positions, between people, in language, and in artifacts. speed and accuracy of judgments and shows that the amount of Humans may even be evolutionarily prepared to pick up associa- power attributed to an agent can be influenced by perceiving tions of power and spatial positions (Fiske, 2004). These are ideal vertical differences.  4 SCHUBERT Analytic Strategy Concerning Response Times and Errors case with missing values, 78 participants, 12 of them male, remained in the as Indicators of Interference sample. The mean age was 21.2 years (SD ⫽ 2.2). Stroop-like interference effects can typically be observed on Materials and Procedure both response latencies and accuracy. Although the latter index has received less empirical attention, both are equally good indicators, Each proposition (see Appendix), with the agent as a black circle and the and both deserve equal attention in any two-choice reaction task, patient as a white circle (or reversed), was depicted above a picture with the because both are outcomes of the same processes (Ratcliff & eight alternative positions (shown in Figure 1). On the paper-and-pencil Smith, 2004). In many interference paradigms, the effect on re- questionnaire, participants were instructed to mark for each of the 18 propositions the picture that best fit their image of the proposition. The sponse latencies is more pronounced because the stimuli are se- order of propositions was determined by random and kept constant across lected such that their judgment should be unambiguous and, as a participants. At the end, participants indicated their age and gender and consequence, the total number of errors is low. Effects on accuracy were debriefed and thanked. show up only when external factors lead to a higher number of errors (e.g., the requirement to answer within a short response This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. window) or when the stimuli themselves become harder to judge Results This document is copyrighted by the American Psychological Association or one of its allied publishers. (MacLeod, 1991). If this is not the case, many more trials are Figure 2 provides an overview of the data by showing the necessary to get reliable results on accuracy (Ratcliff & Smith, frequencies of each of the eight angles summed separately for 2004). In the following studies, effects on both response latencies powerful, powerless, and horizontal propositions. The figure and error frequencies will be reported and combined across all shows that for both powerful and powerless propositions, primarily studies in a meta-analysis at the end. Effects are primarily pre- vertical angles were chosen, whereas horizontal propositions were dicted to occur on response latencies, unless other factors lead to indeed primarily horizontal in their angles. more difficult judgments. To compute the angle of the propositions, it was necessary to ignore whether the agent was left or right of the patient. Each Study 1: Surveying Power answer was scored as seen in Figure 1: Horizontal angles were In the introduction, several examples for a vertical spatial met- scored as 0°, vertical angles with the agent on top were scored as aphor of power were cited. Still, intuition notwithstanding, it has 90°, vertical angles with the agent at the bottom were scored as first to be shown that there really is a shared power ⫽ up schema. ⫺90°, and the diagonal lines were scored as in between. Then, In Study 1, I sought to test this hypothesis. To do so, I based Study angles were averaged for each cell of the Proposition Type ⫻ 1 on a method developed by Richardson and colleagues (Richard- Agent Color design. Collapsed across agent color, powerful prop- son, Spivey, Barsalou, & McRae, 2003; Richardson, Spivey, Edel- ositions had a mean angle of 65.6° (SD ⫽ 17.7); that is, the angle man, & Naples, 2003). They identified both horizontal verbs (e.g., was vertical, with the agent over the patient. A one-sample t test pull) and vertical verbs (e.g., sink) by asking participants to asso- confirmed that it was significantly larger than 45°, t(77) ⫽ 10.29, ciate a verb with one of several tilted lines. Study 1 assessed p ⬍.001. In contrast, powerless propositions had an angle of whether being powerful and being powerless is associated with ⫺56.4° (SD ⫽ 16.4); that is, the angle was also vertical, but the high and low positions in space, respectively. agent was below the patient. This angle was significantly smaller than ⫺45°, t(77) ⫽ 6.15, p ⬍.001. The horizontal propositions Method were indeed horizontal (M ⫽ 0.1°, SD ⫽ 15.0). The angles were submitted to a 2 (agent color) ⫻ 3 (proposition Overview and Design type) general linear model (GLM) with repeated measures on the For 18 propositions, participants answered which one of eight pictures second factor. Proposition type had a significant effect, F(2, best represented the proposition. In 6 of the propositions, the agent (rep- 152) ⫽ 926.90, p ⬍.001, ␩2P ⫽.92.1 Both powerful and powerless resented by a small black circle) was more powerful than the patient proposition angles differed significantly from the horizontal prop- (represented by a small white circle): for example, “ has influence on X.” ositions angle, F(1, 76) ⫽ 510.13, p ⬍.001, ␩2P ⫽.87, and F(1, In another 6 propositions, the agent was less powerful than the patient: for 76) ⫽ 598.61, p ⬍.001, ␩2P ⫽.89, respectively. No other effect example, “ is weaker than X.” Finally, 6 propositions described horizontal reached significance.2 relations between agent and patient: for example, “ pulls X.” The eight pictures depicted eight possible angles between agent and patient (see Figure 1). Because the agent’s color (black or white) was counterbalanced Discussion across participants, the study had a 2 (agent color) ⫻ 3 (proposition type: Study 1 confirmed that there is a shared metaphor that links powerful vs. powerless vs. horizontal) design with repeated measures on power relations to a vertical schema, in which the powerful agent the second factor. The following hypotheses were tested: Both powerful and powerless propositions were expected to be vertical rather than hori- is on top of the powerless one. At first sight, this result might seem zontal; that is, the angle of a powerful proposition should be larger than both 45° and the horizontal proposition’s angle, and the angle of a pow- 1 ␩2P denotes the effect size estimate partial ␩-squared as computed by erless proposition should be lower than ⫺45° and the horizontal proposi- SPSS Version 12. tion’s angle. 2 Figure 2 also suggests that powerful agents are more frequently located left of powerless patients and that powerless agents are more frequently Participants located right of powerful patients. It may be that power is also represented Students of an introductory psychology course at the University of Jena as left of powerlessness, which might be a more general case of a schema (Jena, Germany) answered the questionnaire voluntarily. After deleting one of causation that flows from left to right (Maass & Russo, 2003).  VERTICAL POSITION AND POWER 5 Figure 1. Depictions of eight different angles between agent and patient, and angle values scored for each picture, as used in Study 1 (values were not shown in the original questionnaires). trivial. But in fact, power is defined as having influence on others’ two groups were related and that a clear power difference between outcomes, although the results support the idea that in addition to the groups existed. An additional pretest made sure that high this definition, it is understood as a vertical schema, or in the form power was not confounded with positive valence. For each pair, of a power ⫽ up perceptual symbol. Of course, the results do not participants had to decide as quickly as possible which of the two This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. allow strong conclusions about the mental representations. It could groups was the more powerful one. The crucial point was that the This document is copyrighted by the American Psychological Association or one of its allied publishers. be that the participants merely succeeded in applying a metaphor groups were presented above each other on the screen, such that and aligning it with the graphical representation but that this the powerful one was at either the top or the bottom. It was schema is not usually evoked when one thinks about power. Study expected that participants would more quickly identify a powerful 2 was designed to collect more conclusive evidence that power group when it was at the top rather than when it was at the bottom involves the perceptual simulation of vertical difference. of a pair. As this confounds the answering key with power, in a second condition, participants had to find the less powerful group. Study 2: Vertical Difference Between Two Group Names Here, it was expected that participants would more quickly identify Influences Power Judgments the powerless group when it was at the bottom, compared with when it was at the top. In sum, an interaction of the group type and How spontaneously is this power ⫽ up schema used, and how the vertical position was predicted. deeply are thinking about power and thinking about vertical spatial difference intertwined? In other words, is the power ⫽ up schema indeed a perceptual symbol or just a metaphor that can be repro- Method duced when asked for (as in Study 1)? One way to show that it is indeed a perceptual symbol is to show that thinking about power Overview and Design interacts with available perceptual content on the vertical dimen- sion. This follows directly from the interference paradigm studies In each trial, participants saw labels of two social groups above each cited above, in which compatibility between motor schema and other on a screen and had to decide either which one was the powerful or cognitive tasks (Förster & Strack, 1996) or between visual schema which one was the powerless group; this was manipulated between partic- ipants and formed the first factor task. Each of the pairs was presented and cognitive tasks (Zwaan & Yaxley, 2003) served as evidence twice, once with the powerful group at the top and once with the powerful for the inclusion of perceptual symbols in the cognitive group at the bottom. This formed the second factor position (top vs. representation. bottom), which was manipulated within subjects. Answers had to be given In Study 2, I implemented such an interference paradigm for the with the cursor up and down keys. Thus, whatever group the participants verticality of power. To do so, I adapted the task developed by had to find, they had to press the up key when it was at the top and the Zwaan and Yaxley (2003). In each trial of a reaction time task, down key when it was at the bottom. In sum, the experiment had a 2 (task: participants saw two group labels together on the screen (e.g., find powerful vs. find powerless, between) ⫻ 2 (position: top vs. bottom, master and servant). The group pairs were selected such that the within) design. Figure 2. Frequencies and percentages of angles chosen for propositions with powerful and powerless agents, and horizontal relations (Study 1). Each line’s length depicts frequency of its angle if the patient is at the center of the circle and the agent is at the outer end of the line. Percentages are given in the corresponding numbers. The outer circle corresponds to 65%.  6 SCHUBERT Participants score. Note that top answers were required when the task was to find the powerful group and it was at the top (compatible trials), Ninety-one participants took part in the study. Eleven of them were but they were also required when the task was to find the power- excluded because they answered wrong or too slowly in more than 10 less group and it was at the top (incompatible trials). Likewise, trials, leaving 80 in the sample. Of these, 47 were female and 28 were male (missing data for 5 cases). The average age was 21.5 years (SD ⫽ 2.7). bottom answers were required when the task was to find the powerless group and it was at the bottom (compatible trials), but they were also required when the task was to find the powerful Materials group and it was at the bottom (incompatible trials). It was hy- Altogether, 24 pairs of social groups were used. For each pair, a pretest pothesized that for both tasks, the reactions in the compatible trials had shown that the one group was almost unanimously judged to be more would be faster than in the incompatible trials. powerful than the other group (pairs are listed in the Appendix). In a further To test this, the two scores were entered into a 2 (task: find pretest, 34 participants were presented with the group pairs and were asked powerful vs. find powerless) ⫻ 2 (position: top vs. bottom) GLM to decide which of the two groups they liked more on a scale from 1 (more with repeated measures on the second factor. The means in Table liking for the powerful group) to 5 (more liking for the powerless group).3 1 show the expected pattern: When the task was to find the This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. The mean of the averaged ratings (M ⫽ 3.21, SD ⫽.29) differed signifi- This document is copyrighted by the American Psychological Association or one of its allied publishers. cantly from the scale midpoint (3), t(33) ⫽ 4.23, p ⬍.001, indicating that powerful group, reactions were indeed faster when it was at the the powerful groups were on average liked less than the powerless groups. top, compared with when it was at the bottom. Simple effects In the reaction time task, each pair was presented twice, resulting in 48 analyses confirmed that this difference was significant, F(1, 78) ⫽ trials. Order of presentation was randomized, and it was counterbalanced 11.91, p ⫽.001, ␩2P ⫽.13. The opposite was the case when the for each pair whether it first appeared with the powerful at the top or at the task was to find the powerless group. Here, reactions were faster bottom. Each trial started with a blank screen for 750 ms and a fixation when it was at the bottom, compared with when it was at the top. cross (⫹) in the middle of the screen, which disappeared after 250 ms, This difference was less pronounced but still significant, F(1, followed by a pair of groups above each other, centered vertically and 78) ⫽ 4.12, p ⫽.046, ␩2P ⫽.05. Together, these differences horizontally on the screen, with five blank lines between them. (Thus, the resulted in a significant interaction, F(1, 78) ⫽ 14.82, p ⬍.001, ␩2P fixation cross was located directly in the middle between the two labels.) ⫽.16. In addition, there was a marginal main effect of task: When no answer was given after 2 s, the program continued. Feedback on too long or wrong answers was not given. Words appeared in black Answers were slower when the task was to find the powerless 10-point Arial letters on a white background. The study was run on laptops group than when it was to find the powerful group, F(1, 78) ⫽ with 14-in. displays (resolution 1024 ⫻ 768) and programmed in DMDX 3.71, p ⫽.058, ␩2P ⫽.05. (Forster & Forster, 2003). The laptops stood on tables, at which the participants were seated. Thus, the participants had to look down at the Error Frequencies laptop screen. The same analytic procedure was repeated for the number of Procedure errors (not counting too long or not given answers). Each partic- ipant committed on average 2.95 (SD ⫽ 2.09), or 6.1%, errors. Participants were approached at the campus of the University of Jena Although the means showed the predicted pattern, a GLM follow- and asked to participate in a computer study in exchange for chocolate (a ing the above design found no Task ⫻ Position interaction, F(1, value of about $1). When they agreed, they completed the experiment in 78) ⫽ 0.95, p ⫽.330, ␩2P ⫽.01. groups of up to 3 persons in cubicles set up at the campus. The first screen informed them that the study investigated reactions to verbal stimuli, explained the task to which they were randomly assigned (find powerful vs. Discussion find powerless), and asked them to work both as quickly and as accurately The goal of Study 2 was to test whether vertical differences as possible. The instructions simply asked participants to find the more actually play a role in mental representations of power and thereby powerful (or powerless) group, without defining power more explicitly. (None of the participants complained about a lack of clarity.) After they influence power judgments. To test this, participants had to decide completed the task, they were debriefed, thanked, and given their which of two groups presented together on the screen was, in the chocolate. first condition, the more powerful or, in the second condition, the powerless group. Furthermore, whether the group that was to be Results found was above or below the other group varied. The results show that this position on the screen influenced how quickly the task Response Latencies could be solved. Finding the powerful group was faster when it was at the top than when it was at the bottom. Finding the The grand mean of all response latencies was 1,095 ms. Fol- powerless group was faster when it was at the bottom than when lowing the recommendations of Bargh and Chartrand (2000), I it was at the top. In sum, answers were facilitated when the groups planned to exclude all response latencies longer than three stan- were where the metaphor suggests: the powerful at the top and the dard deviations above the mean; however, this criterion equaled powerless at the bottom. 1,980 ms in this study, probably because each trial was terminated The simple effects analyses showed that the effect was some- automatically after 2,000 ms anyway. Consequently, no response what stronger when the powerful group had to be found compared latencies were excluded in this study, and the maximum response latency was increased in the following studies. Latencies of responses in which the to-be-found group was at 3 The original German question asked which of the two groups partici- the top were averaged to one score, and reactions in which the pants had more “Sympathie” for. This term differs in its meaning from the to-be-found group was at the bottom were averaged to a second English sympathy and is best translated as liking.  VERTICAL POSITION AND POWER 7 Table 1 As cited above, evidence from Borghi et al. (2003) that supports Mean Response Latencies (in ms) to Find the Powerless or this hypothesis shows that judgments about concrete objects are Powerful Group, Depending on Its Position on Screen (Study 2) influenced by the motor response necessary to make the judgment. Apparently, imagining an object is facilitated by compatible move- Position of target group on screen ment and interfered with by an incompatible movement. Given the Top Bottom parallel evidence for thinking about concrete objects and meta- phoric mappings cited so far, it seems likely that thinking about the Task M SD M SD abstract domain of power is also influenced by vertical move- Find powerful 1050 134 1093 138 ments. This hypothesis is also compatible with the theory of event Find powerless 1140 142 1113 124 coding (Hommel, Müsseler, Aschersleben, & Prinz, 2001), which argues that both perception and action codes are represented at a common, commensurable level—that of external events. Up and down motor responses would therefore be represented as vertical This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. with the condition when the powerless group had to be found. movements of an external object (hand or the affected object) and This document is copyrighted by the American Psychological Association or one of its allied publishers. However, this effect has to be interpreted with caution because the thus at the same level as external visual stimuli (for a similar line answers use different keys: the up key when the powerful group of reasoning, see Neumann & Strack, 2000). was at the top and the down key when the powerful group was at The up and down cursor keys used in Study 2 are not actually the bottom. In principle, it is possible that the up key can be hit above each other in space. Thus, it would be more precise to say faster or that the group appearing at the top can always be found that they imply different imagined movements or motor imagery of more quickly because it is read first. These two reasons might have an up and down movement. Given most students’ extensive use of resulted in a slight overall increase in the speed of top answers and computer software in which a press of the up cursor results in an might thus have distorted the simple comparisons. The interaction up movement of a cursor or another object, it seems lik