Running head: CONSTRUAL OF A SOCIAL TARGET
The effect of level of construal on social distance, and the role of action valence.
Bregje J. C. C. Sweep, 10254587
Supervisor: Dr. Michael L. W. Vliek Universiteit van Amsterdam
Abstract
In this field experiment conducted in the train, the effects of level of construal and action valence are explored on interpersonal and intergroup social distance. Combining mechanisms of Construal Level Theory and Linguistic Intergroup Bias, participants’ (N = 200)
interpersonal and intergroup social distance towards a social target was measured after a construal level (high, low) and action valence (negative, neutral, positive) manipulation. The results suggest that the more negative an action is, the more interpersonal social distance people experience. Intergroup social distance show mixed results on action valence. The hypothesized interaction effect between level of construal and action valence on interpersonal and intergroup social distance was not found. Further exploring the results, level of construal has only a significant effect on interpersonal social distance for people who do not have a friend with the same name as the target. Theoretical and practical implications are discussed.
The effect of level of construal on social distance, and the role of action valence. A good neighbor is worth more than a distant friend. This well-known saying emphasizes the importance of actual distance between us and the people we are interacting with. But how distant we feel towards the people we are interacting with–independent of actual distance–is also an important determinant in everyday social interaction (Stephan, Liberman & Trope, 2010). Social closeness might determine which neighbor we ask for a favor, which colleague we ask for a ride home or which friend we choose as Master of Ceremonies for our wedding. The present paper explores factors that influence subjective social distance. More specifically, this study examines the effect of a concrete or abstract representation on how close or distant someone is perceived, and explores the role of action valence on that effect. To that end, first Construal Level Theory, Action Identification and Linguistic Intergroup Bias are described in the light of social distance. Then, three predictions resulting from these theories will be described and tested in a field study.
Construal Level Theory
Think about what you are going to do next Saturday evening. Maybe it is having a barbecue at 7.30 pm with four friends at your house? Then think about what you are going to do Saturday evening next year. Maybe it is hanging out with some friends? These two Saturday evenings are both in the future, but the Saturday evening next week is represented, or construed, much more concrete and detailed than the Saturday evening next year. The same mechanism occurs with people who are psychologically close or distant to you: there are represented differently.
The differences in representation of both Saturday evenings occurs because the
amount of available concrete information tends to diminish when the event occurs in the more distant-future. Thus the event naturally has to be perceived in more abstract ways, because there is simply no concrete information available. The phenomenon that distant-related cues
(e.g. ‘next year’) evoke abstract descriptions (e.g. ‘hanging out’) is exactly what Construal Level Theory (CLT) entails (Bar-Anan, Liberman, Trope, & Algom, 2007; Trope & Liberman 2003; 2010). The effect of CLT on social distance will be addressed in this study.
High and Low. Events or objects can be described at different levels of abstraction. In
other words, events or objects can be described using different levels of construal. A description using a low level of construal is a more concrete, unstructured, superficial, and complex description. A high level of construal is a description that is more abstract,
structured, profound, and simple. (Trope & Liberman, 2003). Whether a stimulus (an event or object) is construed using high or low levels depends on the psychological distance from the perceiver to that stimulus. The more a stimulus is psychologically distant from a perceiver’s direct experience (for example: Saturday evening next year), the higher level of construal will be used (for example: hanging out) to represent that stimulus (Ban-Anan et al., 2007; Stephan et al., 2010). For instance, when an expected social interaction is more distant in time, that person will be described using higher level of construal than when the social interaction is closer in time (Stephan et al., 2010).
Social distance. Psychological distance has four dimensions: temporal (how much
time is between ‘now’ and the event), spatial (how much distance in space is between ‘here’ and the object), social (how different is the target from the self, e.g. friend vs. stranger), and hypotheticality (how likely is the event to happen). The present research will only focus on social distance and will investigate the effects of level of construal and action valence on social distance.
Social distance can be operationalized as interpersonal similarity (see Liviatan et al., 2008 for an overview): the more a social target is close, similar, or familiar, the less distant is a social target from the self. In this paper, social distance and social dissimilarity will be used interchangeably.
CLT predicts that a psychologically close social target (e.g. a friend) is construed using lower levels than a psychologically distant social target (e.g. an enemy). That is indeed what research found: socially similar people are evaluated with more weight on subordinate and secondary features, that is, with lower levels of construal, than dissimilar people
(Liviatan, Trope, & Liberman, 2008). For example, people based the grading of a physics essay of dissimilar others mostly on the primary (high level) aspect of the task: the story’s quality. In contrast, the grading of physics essays of similar others was influenced by
secondary (low level) information about the target’s physics abilities (Experiment 4, Liviatan et al., 2008).
Bi-directional effect. Thus, previous research suggest that socially distant people are
represented using higher levels of construal. Yet this paper investigates whether a high level of construal can lead to more social distance. This would imply that CLT has a bi-directional effect: not only does social distance affect the subsequent construal, but also does construal affect the subsequent social distance. That would imply that the construal of a social target could influence the perceived social distance towards that target. Evidence for a bi-directional effect of CLT was found on temporal distance and level of construal. An event that was construed using high levels was perceived to be in the more distant future, and distant-future events were construed using higher levels than near-future events (Liberman, Trope, McCrea, & Sherman, 2007). Because the four dimensions of psychological distance have similar and overlapping effects (Bar-Anan et al., 2007; Trope & Liberman, 2010), the findings on temporal distance should also apply on social distance.
A bi-directional effect of CLT on social distance has not been studied extensively. Liberman and Förster (2009) are the first to show that level of construal can influence social distance. Participants’ level of construal was primed with a global (high level) vs. local (low level) processing style. It was found that the subsequent effect on social distance followed the
same pattern as predicted by CLT: a global processing style increased the perceived social distance, compared to a local processing style. So people felt less close to others when they were primed with a global (high level) processing style (Liberman & Förster’s study 3, 2009). Additionally, research by Stephan, Liberman and Trope (2011) showed that people who used high levels to construe a social target reported less feelings of familiarity and allocated fewer resources (so, more social distance) to that social target than people who construed the social target using lower levels of construal.
In line with previous research, this paper proposes that there is a bi-directional effect of CLT on social distance: level of construal can affect the perceived social distance towards a social target.
Action Identification
There is, however, some debate in the literature about whether socially similar people are indeed represented in a more concrete, detailed way as CLT predicts. Compelling
evidence from research on action identification, that is, the amount of abstraction used in identifying and describing what a target person does, shows that also the opposite effects of CLT can occur. When people are asked to describe others’ actions, they describe actions from people they like at a higher level than people that they do not like (Kozak, Marsh, & Wegner, 2006). An abstract description takes people’s goals and intentions into consideration, and we are more motivated to do that with people we like (Kozak et al., 2006). In that way, liking covaries the effect of level of construal on action identification, in that people are more motivated to explore underlying causes of behavior for people they like, which results in construing the actions of a liked target using higher levels than actions of a disliked target. So in that case, social similarity leads to higher levels of construal instead of lower levels of construal predicted by CLT.
Linguistic Intergroup Bias
The results of CLT and action identification seem contradicting, but this paper attempts, for the first time, to explain and combine them both in the light of Linguistic Intergroup Bias (LIB; Maass, Salvi, Arcuri, & Semin, 1989). LIB entails a subtle, systematic bias in language use while describing positive or negative actions from an in-group member or an out-group member. More specifically, it describes the tendency to communicate positive in-group behavior and negative out-group behavior in more abstract terms and communicate negative in-group behavior and positive out-group behavior in more concrete terms. An abstract description conveys the impression that the behavior is due to stable dispositions, whereas a concrete description suggest the behavior is caused by contextual cues (Reitsma-van Rooijen, Semin, & Leeuwen (Reitsma-van, 2007). In that way, in-group favoritism and out-group derogation is prolonged by describing the respectively positive and negative actions in an abstract manner. This effect occurs in both interpersonal and intergroup communication (Maass, Milesi, Zabbini, & Stahlberg, 1995; Maass, 1999). For instance, a tipping in-group member is ‘generously and friendly’, whereas a tipping out-group member ‘left some extra money’. A cheating in-group member, on the other hand, ‘kissed someone else’; a cheating out-group member is ‘a cheater and a liar’.
Although CLT predicts that actions of a (socially dissimilar) out-group member are described more abstractly than actions of a (socially similar) in-group member, LIB shows that it depends on the valence of the action whether a (dis)similar social target is represented in abstract or concrete terms. In other words, how positive or negative an action is, determines whether the member of the in-group or out-group is described abstractly or concretely. The present study tries to incorporate these mechanisms of LIB and CLT by adding positive and negative actions into the CLT paradigm. The valence of the action has–to my knowledge–not been taken into consideration while studying the effects of level of construal on social
distance. This is remarkable, because an often-used paradigm to induce a high level or low level of construal is to describe why (high level) or how (low level) a social target performs an action (developed by Freitas, Gollwitzer, & Trope, 2004; used in e.g. Kanten, 2011; Liberman et al., 2007; Stephan et al., 2011; Wakslak & Trope, 2009). Where the
aforementioned studies only used neutral actions, this paper uses the same manipulation technique, but adds positive and negative actions into the design.
Bi-directional effect. In order to investigate the possible effect of level of construal
and action valence on perceived social distance, it is necessary to explore a bi-directional effect of LIB. LIB predicts that social distance and action valence influence the abstractness of representation (Maass et al., 1989). A bi-directional effect would entail that the level of abstraction and action valence also influence social distance. A large body of research confirms the assumption of a bi-directional LIB effect. Different lines of research labeled it differently (for instance, Semin & Fiedler’s (1988) Linguistic Category Model, or Wigboldus, Semin, & Spears’s (2000) Linguistic Expectancy Bias), but the main idea is that people can draw inferences about other people, purely based on the level of the abstraction of a
description (e.g. Semin & Fiedler, 1988). For example, research on written messages (feedback to the participant) found that people who received a positive, abstract message perceived more proximity towards the sender than people who received a positive, concrete message. The reverse was found for negative messages (Reitsma-van Rooijen et al., 2007). It is important to note that the proximity effect was not simply due to the valence of the written message they received. Interestingly, a significant part of proximity towards the sender was uniquely explained by the abstraction level of the message: the sender of concrete messages was seen as being more proximate than the sender of abstract messages, when controlled for the valence of the message (Reitsma-van Rooijen et al., 2007), which is again in line with CLT.
In contrast to the aforementioned body of research, this paper explores the effects of self-written messages on social distance. Here, participants will write their own concrete or abstract representation of the target’s actions instead of reading other-written messages. It will allow to investigate the effects of construal of a social target on social distance, instead of investigating the evaluation of the sender of an abstract message (for instance in Reitsema-van Rooijen et al., 2007).
Intergroup Social Distance
CLT and Action Identification mostly focus on interpersonal social distance: the distance between the self and a social target. LIB, however, also takes intergroup social distance into account. It entails whether a social target is categorized as being an in-group member or out-group member. An in-group is a broad term for a group one psychologically identifies with and feels like being a member of. People are more socially close to in-group members than out-group members (Tajfel et al., 1971). In order to examine social distance as complete as possible, both interpersonal and intergroup social distance will be addressed in the present study.
Bi-directional effect. Following the same mechanisms as interpersonal social
distance, a bi-directional LIB effect of intergroup social distance would suggest that the action valence and the abstraction of the description can influence whether the social target is being categorized as an in-group member. This is an unexplored effect of LIB. Aside from the theoretical contribution to LIB, this effect could have important practical implications, too. LIB was originally designed to explore the effect of language on the transmission and maintenance of prejudice and stereotypes (Maass, 1999). The possibility of a bi-directional effect of LIB on intergroup social distance will give us means in trying to reduce prejudice and stereotypes by manipulating the level of abstraction while describing a positive or negative action from an out-group member.
Age
While exploring factors that can influence social distance, age must also be
considered. The effect of age on the selection and pursuit of social interaction is described in the socioemotional selectivity theory (Carstensen, Isaacowitz, & Charles, 1999). It entails a life-span theory of motivation in which older people place more importance on the quality of social interactions and they are more selective in their choice in social partners than younger people. Because of their limited time, older people are more focused on present-orientated goals, rather than future-orientated goals and they prefer positive over negative information (Carstensen et al., 1999). In terms of the present study on social distance, this would imply that the older people are, the less social distance they experience in general. In order to control for this effect, age will be added as covariate.
Present Study
In the present study, the aim is to expand the findings of Construal Level Theory and its effect on social distance in the light of Linguistic Intergroup Bias. This study explores the bi-directional effect of LIB in order to examine the main question: what is the effect of level of construal and action valence of a social target on the social distance from that social target?
First, this paper hypothesizes that level of construal of a social target influences the social distance from that social target. More specifically, it is expected that a high level of construal of neutral actions leads to more social distance than a low level of construal of neutral actions, as expected by CLT.
Second, it is hypothesized that action valence has an effect on social distance. I predict that a social target’s positive actions result in less social distance than a social target’s
negative actions.
Finally, combining these two hypotheses, I predict that level of construal of a social target and the social target’s action valence have interacting effects on social distance. As
mentioned above, LIB predicts that people tend to describe in-group members’ positive actions more abstractly than positive actions of out-group members (Maass et al., 1989). Therefore, I predict that an abstract description of positive behavior will be attributed as being behavior from an in-group member and therefore lead to less social distance. The same
possible explanation applies–in reverse–for negative actions: an abstract description of
negative actions will be attributed as being behavior from an out-group member and therefore lead to more social distance. Therefore, I expect that a social target’s positive actions that are construed using high levels will result in less social distance than positive actions that are construed using low levels. The reverse is expected for negative actions: negative actions that are construed using high levels will lead to more social distance than negative actions that are construed using low levels.
In order to test these three hypotheses also for intergroup social distance, the extent to which people are likely to categorize the social target as being an in-group member will be measured. The same prediction as described for interpersonal social distance apply for intergroup social distance: (1) a high level of construal will decrease the likelihood the social target belongs to one’s in-group compared to a low level of construal, (2) positive actions will increase the likelihood the social target belongs to one’s in-group compared to negative actions, and (3) abstract positive actions and negative concrete actions are more likely to be seen as behavior form an in-group member than concrete positive actions and abstract negative actions.
In sum, the following three hypotheses will be tested:
Hypothesis 1: Level of construal of a social target has an effect on interpersonal and
intergroup social distance. I predict that an abstract representation of a social target results in more interpersonal and intergroup social distance than a concrete representation of a social target if the actions of the social target are neutral.
Hypothesis 2: Action valance of a social target has an effect on interpersonal and intergroup social distance. I predict that a social target who performs neutral actions will be perceived with less interpersonal and intergroup social distance than social target who performs negative actions, but with more interpersonal and intergroup social distance than a social target who performs positive actions.
Hypothesis 3: Level of construal and action valance of a social target have interacting effects on interpersonal and intergroup social distance. I predict that a negative abstract
representation of a social target results in more interpersonal and intergroup social distance than a negative concrete representation of a social target. Furthermore, I predict that a positive abstract representation of a social target results in less interpersonal and intergroup social distance than a positive concrete representation of a social target.
Finally, I predict that age has a covariating effect on the hypotheses described above, in that the older people are, the less social distance they experience.
Method Participants and Design
Participants were approached in the Dutch train on different days and different routes, covering the north, south, east, and west of the Netherlands. On Saturday December 7th 2013 between 10:15h and 19:00h, participants were approached on the route Tilburg-Leeuwarden and on Sunday December 29th 2013 between 13:00h and 20:30h, the route Tilburg-Schiphol was used to collect participants. The selection criteria for asking whether people were willing to participate in the study were if they looked older than 18 years, were sitting down in the train, and spoke Dutch sufficiently. Two hundred and twenty seven people initially agreed to participate in the study (Mage = 36.22 years, SDage = 17.57 years, age range: 15–80 years, 57%
female). The reward for participation was a chocolate bar, given after completion of the questionnaire.
Participants were randomly assigned to a 2 (Level of Construal: high, low) x 3 (Action Valence: negative, neutral, positive) between subjects factorial design. Gender of the target person in the study (Tom/Iris) was randomly distributed across participants.
Procedure and Materials
After a brief introduction to the study, anonymity guarantee, and signing an informed consent, participants started the study. They all read three sentences of three actions that either ‘Tom’ (male target) or ‘Iris’ (female target) performed. Depending on the condition, these three actions were all positive (e.g. helping someone move), all negative (e.g. cheating), or all neutral (e.g. painting a room). The neutral actions were selected from the Behavioral Identification Form (Vallacher & Wegner, 1989), and were washing clothes, filling a cavity, and painting a room. The positive actions were taking care of an ill friend, volunteering with elderly, and helping someone move; the negative actions were cheating, stealing money, and wrecking a bus stop. These actions were chosen based on their valence evaluation in a pilot study1. The mean valence scores of the pilot study are shown in Appendix A. ‘Tom’ and ‘Iris’ are common Dutch surnames with both a prevalence of 0.3% in the Dutch population in 2010 (Meertens Instituut, 2013).
Next, participants were asked either to write down why Tom/Iris performed these actions to induce a high level of construal or how Tom/Iris performed these actions to induce a low level of construal. This procedure has been successfully used in previous studies to manipulate participants’ level of construal (e.g. Freitas et al., 2004; Kanten, 2011; Liberman et al., 2007; Stephan et al., 2011; Wakslak & Trope, 2009). So half of the participants
represented the target in an abstract, global way; half of the participants represented the target 1 In the pilot study, participants (N = 22, 85% female, M
age = 22.1, SDage = 1.5) indicated on a 7-point
Likert-scale (1 = negative, 4 = neutral, 7 = positive) the valence of 18 actions. The nine actions closest to a mean evaluation of 1, 4, and 7 were chosen for the respectively negative, neutral, and positive conditions. Two paired sample t-tests showed that the mean valence for the three negative actions (Mneg = 1.1, SDneg = 0.2) was lower
than the mean valence for the three neutral actions (Mneu = 4.4, SDneu = 0.6); t(21) = -22.8, p < 0.001; and the
mean valence for the three neutral actions (Mneu = 4.4, SDneu = 0.6) was lower than the mean valence for the
three positive actions (Mpos = 6.2, SDpos = 0.6); t(21) = -11.4, p < 0.001. Age and gender revealed no significant
effect. See Appendix A for the 18 actions and their valence scores.
in a concrete, detailed way. After the level of construal and action valence manipulation, participants were asked about their social distance towards Tom/Iris. They indicated on a 7-point Likert-scale how close (hecht), familiar (vertrouwd), and similar (lijken op) Tom/Iris felt to them (1 = not at all, 7 = very much). The ‘close’ evaluation is after Liberman and Förster (2009) who, after a processing style manipulation, asked participants how close (hecht) they felt towards their family doctor. The ‘familiar’ and ‘similar’ evaluations were used in the third study of Stephan and her colleagues (2011), in which they asked participants after a temporal distance manipulation how familiar and similar they felt towards a social target.
Participants also indicated with the Inclusion of Other in the Self scale (Aaron, Aaron, & Smollan, 1992) their closeness to Tom/Iris. This scale provides a single-item measurement of interpersonal interconnectedness by having participants choose one out of seven Venn diagrams, each varying the amount of distance between the self and the other (Tom/Iris). This scale was also used by Reitsema van Rooijen and her colleagues’ (2007) research on written messages. If this four dependent measures (close, familiar, similar, and the Inclusion of Other in the Self scale) achieve satisfactory reliability (Cronbach’s Alpha ≥ 0.7), the data will be aggregated to one ‘Interpersonal Social Distance’ dependent measure, with higher scores indicating more interpersonal social proximity.
Next, participants indicated with three questions on a 7-point Likert-scale the
likelihood that they categorize Tom/Iris as a member of their in-group (1 = not at all likely, 7 = very likely). People belong to different in-groups, but it was measured whether Tom/Iris could be a colleague at work, a citizen of the same town, or have the same nationality. Furthermore, the Inclusion of Other in the Self Scale was adjusted to provide a single-item measurement for how the target fits into one’s in-group. Participants chose one out of seven Venn diagrams, each varying the distance between the target Tom/Iris and a typical group of
coworkers. If this four dependent measures (colleague, citizen, nationality, and the adjusted Inclusion of Other in the Self scale) achieve satisfactory reliability (Cronbach’s Alpha ≥ 0.7), the data will be aggregated to one ‘Intergroup Social Distance’ dependent measure, with higher scores indicating more intergroup social proximity. After providing some demographic data (gender, age, level of education, having a ‘Tom/Iris’ in one’s circle of friends, earnestly while answering) participants were debriefed, thanked, and given a chocolate bar.
Manipulation check. Participants were also asked to indicate the valence of the
actions they described on a 7-point Likert-scale (1 = negative, 4 = neutral, 7 = positive) and how much they liked Tom/Iris (1 = not at all, 7 = very much).
Results Manipulation check
The two items concerning the valence of the target’s action and how much they liked the target were aggregated into one Affective Evaluation scale (Cronbach’s Alpha = .89), with higher scores indicating more positive affect. The affective evaluation was significantly different for positive, neutral, and negative actions, so the manipulation of action valence was successful. The results are reported in Table 1. The effects of gender and having a friend with the same name as the target will be addressed in the explorative analyses.
Table 1.
Mean Affective Evaluation of the Target
Action Valence Gender Having a friend with the
same name as the target Negative 2.06 (1.02)a Male 4.35 (2.09)a Yes 5.75 (1.36)a Neutral 5.20 (1.31)b Female 4.73 (2.00) No 3.31 (2.00) Positive 6.38 (0.71)c
Note: mean scores in the same column with different superscripts differ significantly from each other, using p < 0.05.
Higher scores indicate more positive affect.
Participants
Participants were included for analyses if their descriptions on the manipulation were content-wise in line with the condition they were assigned to and if they indicated they answered earnestly. Excluding the 27 participants who did not provide a satisfactory manipulation (N = 23) or indicated they did not answer earnestly (N = 4), left 200 valid participants. Twenty one participants did not sign the informed consent attached to the front of the questionnaire. They did, however, answered the questions, therefore indicating indirect consent, so their data was included for analyses2. Outlier analysis with standardized item scores revealed no deviations more than three standard deviations. The age range is 15-75 years, Mage = 34.66 years, SDage = 16.34 years. One hundred and fifteen participants were
female (58%), 81 male, and 4 unknown. The distribution of the highest completed education level is displayed in Table 2. On the question whether they had a friend with the same name as the target (Tom/Iris), 52% of the participants answered positive.
Initial GPower analyses revealed that 158 participants should participate in order to reveal a difference with a power of .80 and an uncertainty level of 5% in a multivariate 2 When the 21 participants who completed the study without signing their informed consent were excluded from
analyses, the results remained virtually the same. These participants were therefore included for analyses.
analysis (ANOVA). A power of .90 would be obtained with 206 participants. With this number of participants, an effect size of 0.25 (medium), with six groups: 2 (Level of Construal: high, low) x 3 (Action Valence: positive, negative, neutral) and two numerator degrees of freedom: (2-1)*(3-1) = 2 would be obtained. See Appendix B for the GPower analyses. With 200 valid participants, a power of .80 is obtained.
Table 2.
Highest Completed Education.
Frequency Percent Valid Percent
Cumulative Percent
Valid Primary school 2 1,0 1,0 1,0
VMBO 19 9,5 9,7 10,7 HAVO 21 10,5 10,7 21,4 VWO 21 10,5 10,7 32,1 MBO 26 13,0 13,3 45,4 HBO 47 23,5 24,0 69,4 WO 60 30,0 30,6 100,0 Total 196 98,0 100,0 Missing 999 4 2,0 Total 200 100,0 Variable selection
Missing values were excluded list wise, because Little’s MCAR test revealed that the missing values were missing completely at random: χ2 (49) = 47.03, p = .55.
The four items referring to interpersonal social distance (to what extent does the target feel close, familiar, similar, and the Inclusion of Other in the Self scale) achieved satisfactory reliability (Cronbach’s Alpha = 0.9) and were aggregated into one Interpersonal Social
Distance scale, with higher scores indicating more social proximity. The four items measuring intergroup social distance items (how likely is it that Tom/Iris is a colleague, is a citizen of one’s town, has the same nationality, fits into a typical group of coworkers) did not reach a
satisfactory reliability (Cronbach’s Alpha = .62), so these variables will be analyzed separately3.
Participant’s gender, target’s gender (Tom or Iris), and education level did not reveal any significant effects on social distance and were excluded from analyses. Gender
(in)congruency with participant’s gender and gender of the target (Tom/Iris) was taken into account, but did also not reveal any significant effects4.
An Alpha level of .05 is used for all statistical tests, unless otherwise stated.
Confirmatory analysis
Hypothesis 1 and 2 predicted a main effect of respectively level of construal and action valence on interpersonal and intergroup social distance. The third hypothesis would be supported by an interaction effect of level of construal and action valence on interpersonal and intergroup social distance. A covariating effect of age was also predicted.
In order to test these hypotheses, a multivariate analysis of variance (MANOVA) was conducted with level of construal and action valence as independent factors. Interpersonal social distance and the four items measuring intergroup social distance served as dependent variables, and age was added as covariate. Simple contrasts for the two independent variables were added a priori in order to test the hypotheses with an Alpha level of .05/2 = .025.
The results show that, controlled for age as a marginal significant multivariate covariate (Wilks’ Lambda = .95, F(5,178) = 2.02, p = .078, η2 = .05), there is no interaction effect of level of construal and action valence on social distance (Wilks’ Lambda = .96, F(10,356) = .72, p = .70). The multivariate main effect of construal level is also not significant (Wilks’ Lambda = .97, F(5,178) = 1.23, p = .30).
3 The aggregated Intergroup Social Distance variable did not reveal any significant results on Construal Level,
Valence, the interaction between the independent variables, or age. P values varied between .19 and .91.
4 A multivariate ANOVA with Interpersonal Social Distance and the four intergroup social distance items as
dependent variables, and participant’s gender, gender of the target, and gender congruence as independent variables revealed no significant multivariate effects for participant’s gender (F(5,177) = .23, p = .95), target’s gender (F(5,177) = 1.77, p = .54), nor gender congruence (F(5,177) = .65, p = .66).
Only action valence reveals a significant multivariate main effect on social distance (Wilks’ Lambda = .38, F(10,156) = 22.19, p < .001, η2 = .38), supporting the second hypothesis that there is an effect of action valence on social distance. This effect, however, was not found in every dependent variable. The univariate analysis shows an effect of action valence only for interpersonal social distance (F(2,182) = 112.5, p < .001, η2 = .55), whether the target could be a colleague (F(2,182) = 20.69, p < .001, η2 = .19) and whether the target fits into a typical group of coworkers (F(2,182) = 29.08, p < .001, η2 = .24). A representation of the mean social distance per action valence is shown in Table 3.
The univariate analysis also shows that the covariating age effect only occurs for interpersonal social distance (B = .015, t(2.92) = .004) and not for any of the four items
measuring intergroup social distance. This indicates that, as expected, the older people are, the less interpersonal social distance they experience towards the target.
The first hypothesis specifically predicted that there would be a main effect of level of construal on social distance if the target performs neutral actions. It was expected that a high level of construal leads to more social distance than a low level of construal if the actions of the target are neutral. This would replicate the findings of Construal Level Theory. In order to Table 3.
Mean Social Distance with Action Valence Interpersonal
Social Distance*
Colleague** Citizen** Nationality** Group of coworkers** Negative 1.87 (0.89)a 3.34 (1.72)a 5.79 (1.39) 5.89 (1.46) 2.32 (1.42)a Neutral 4.28 (1.48)b 4.94 (1.64) 5.31 (1.58) 5.73 (1.48) 4.45 (1.72) Positive 4.90 (1.12)c 5.11 (1.66) 5.56 (1.58) 5.37 (1.74) 4.14 (1.87) Note: mean scores in the same column with different superscripts differ significantly from each other, using p < .025.
* Higher scores indicate more social proximity.
test this hypothesis, the aforementioned MANOVA was executed again, but now only with neutral actions. No effects of level of construal on interpersonal or intergroup social distance were found with neutral actions. P values varied from .19 to .97.
Explorative analysis
It is interesting to further explore the results, because–as shown in the confirmatory analysis–there seems to be some variation in the effects on interpersonal and intergroup social distance of several variables (e.g. age, action valence). Interpersonal and intergroup social distance behave apparently somewhat different, so these concepts will tested separately. This could shed light on the specific effects the different variables have. Furthermore, it would be unwise to disregard the findings in Table 1, which showed effects of gender and whether the participant has a friend with the same name as the target on the affective evaluation of the target. It was not hypothesized that the name of the target could have an effect, but this explorative analysis will test if the factors whether the participants has a Tom/Iris in their circle of friends, and gender also have covariating effects on interpersonal and intergroup social distance.
Interpersonal social distance. In order to test the explorative hypothesis that gender
and having a friend with the same name as the target could influence interpersonal social distance, an univariate analysis of variance (ANOVA) was conducted. Interpersonal Social Distance served as dependent variable, and level of construal and action valence as
independent variables. Age, gender, and having a friend with the same name as the target were added as covariates.
The results show that for interpersonal social distance, age (B = .015, t(3.1) = .002) and having a namesake of the target in one’s circle of friends (B = -1.06, t(-5.62) < .001) served as significant covariates, but gender did not.
Interestingly, when circle of friends is accounted for in the analysis, a non-significant marginal effect of construal level appears: F(1,182) = 2.69, p = .10, η2 = .02. People feel more socially close to a target that they described using low levels (M = 3.84, SD = 1.83) than to a target that they described using high levels (M = 3.61, SD = 1.71). Although it is only a marginal effect and should be treated with caution, the results point in the same direction as CLT predict and support the first hypothesis. The results on action valence and the interaction between level of construal and action valence are virtually the same as reported in the
confirmatory analysis: there is an effect of action valence (F(2,182) = 56.07, p < .001, η2 = .38), in that all three valences differ from each other in the predicted direction, but there is no interaction effect between level of construal and action valence on interpersonal social distance (F(2,182) = .64, p = .53).
Having a target’s namesake in one’s circle of friends. Having a friend with the same name as the target has a large effect on interpersonal social distance (Cohen’s d = .86): people who have a Tom/Iris in their circle of friends report overall more interpersonal social
proximity (M = 4.74, SD = 1.28) than people who do not have a namesake of the target in their circle of friends (M = 2.61, SD = 1.53; t(10.57) < .001). In order to explore the exact effects of having a friend with the same name as the target on interpersonal social distance, a new univariate analysis of variance (ANOVA) was conducted. Interpersonal Social Distance was still the dependent measure, but ‘having a friend with the same name as the target’ served now as independent measure, along with level of construal and action valence. Age served as covariate. The results were separated for the two groups (with and without having a friend with the same name as the target) in order to explore the different effects on interpersonal social distance.
The results show now a clear effect of level of construal, but only for people who do not have a friend with the same name as the target (F(1,84) = 6.27, p = .014, η2 = .07). This
effect of level of construal does not appear for people who have a Tom/Iris in their circle of friends. It suggest that–if you do not know a Tom/Iris–a high level of construal (M = 2.92, SD = .16) leads to more social distance than a low level of construal (M = 3.46, SD = .14). This is in line with Hypothesis 1. The results on action valence are virtually the same for both groups and in line with the confirmatory analysis.
The hypothesized interaction between level of construal and action valence shows a non-significant trend (F(2,84) = 2.1, p = .13, η2 = .05) for people who do not have a target’s namesake in one’s circle of friends. This non-significant trend does not appear for people who do have a friend with the same name as the target. As shown in Figure 1, the amount of interpersonal social distance seem to be dependent of both action valence and level of construal.
Another difference on interpersonal social distance between people with and without a friend with the same name as the target was found on the covariate age. Only for people who have a target’s namesake in their circle of friends served age as significant covariate (B = .03, t(3.69) < .001). This covariating effect was not found for people who do not have a Tom/Iris in their circle of friends.
Intergroup social distance. In order to explore the possible covariating effects of
gender and having a Tom/Iris in one’s circle of friends on intergroup social distance, a multivariate analysis of variance (MANOVA) was conducted. The four items measuring intergroup social distance (how likely is it that Tom/Iris is a colleague, is a citizen of one’s town, has the same nationality, fits into a typical group of one’s coworkers) served as dependent variables, level of construal and action valence as independent variables, and age, gender and having a friend with the same name as the target as covariates. Simple contrasts for the two independent variables were added a priori with an Alpha level of .05/2 = .025.
As found in the confirmatory analysis, age is not a significant covariate on intergroup social distance. Neither is gender, but having a friend with the same name as the target is (Wilks’ Lambda = .88, F(4,177) = 5.82, p < .001, η2 = .12). Adding this significant covariate
results in virtually the same non-significant results of level of construal and the interaction between action valence and level of construal as found in the confirmatory analysis. On action valence, however, adding the covariate makes the results more profound. All four items show a significant effect of action valence (compared to only ‘colleague’ and ‘group of coworkers’ in the confirmatory analysis), with P values varying from <.001 to .05. The results of action valence per item are displayed in Table 4. Overall, there seems to be only a significant effect of negative actions on intergroup social distance, compared to neutral or positive actions. This effect, however, has different directions for different variables. For ‘colleague’ and ‘group of coworkers’, the likelihood the target is part of that in-group decreases when the target
performed negative actions. For ‘citizen’ and ‘nationality’, on the other hand, the likelihood the target is part of that in-group increases when the target performed negative actions.
Table 4.
Mean Intergroup Social Distance with Action Valence
Colleague Citizen Nationality Group of
coworkers Negative 3.33 (1.73)a 5.77 (1.44)a 5.87 (1.47)a 2.28 (1.39)a Neutral 4.92 (1.63) 5.29 (1.58)b 5.71 (1.49)ab 4.46 (1.71) Positive 5.11 (1.66) 5.56 (1.58)ab 5.37 (1.75)b 4.14 (1.87) Note: mean scores in the same column with different superscripts differ significantly from each other, using p < .025.
Higher scores indicate that it is more likely the target is part of that in-group.
Discussion
In this study, it was hypothesized that interpersonal and intergroup social distance is influenced by level of construal (Hypothesis 1) and action valence (Hypothesis 2), and by an interaction between level of construal and action valence (Hypothesis 3). The results suggest an effect of level of construal on interpersonal social distance, but only under specific
interpersonal and intergroup social distance. The hypothesized interaction of level of construal and action valence on social distance was not found.
Confirmative Results
Hypothesis 1. This study was unable to find an effect of level of construal on
interpersonal or intergroup social distance, predicted in Hypothesis 1, in the confirmatory analysis. Seemingly, it does not matter how concretely or abstractly people describe someone on the amount of social distance people experience towards someone. This bi-directional effect of CLT on social distance has not been studied extensively (Liberman & Förster, 2009) and this study could not directly add evidence to the assumption that a bi-directional effect of CLT on interpersonal and intergroup social distance exist. The forthcoming discussed
explorative analyses, however, revealed some specific conditions under which an effect of level of construal on interpersonal social distance occurs. Therefore, it is in my opinion premature to disregard an effect of level of construal on social distance.
Hypothesis 2. The second hypothesis predicted that action valence has an effect on
interpersonal and intergroup social distance. The results suggest that the valence of an action indeed determines the amount of interpersonal and intergroup social distance one experiences. For interpersonal social distance, the three action valences all differ significantly from each other in the predicted direction. This indicates that people perceive someone who performs negative actions with more interpersonal social distance than someone who performs neutral actions, and someone who performs neutral actions with more interpersonal social distance than someone who performs positive actions. As predicted, this finding suggest that we feel less familiar, close, and similar to people who behave negatively in comparison to people who behave positively.
For intergroup social distance, the results point in the same direction, but were not found on every variable measuring intergroup social distance. Someone is less likely to be
categorized as being a member of the in-groups ‘colleague’ and ‘group of coworkers’ when the actions are negative, compared to neutral or positive actions. For the in-groups ‘citizen’ and ‘nationality’, however, this effect was not found in the confirmatory analysis.
Interestingly, the action valence effect on ‘colleague’ and ‘coworker’ was only found for negative actions: the likelihood of being an in-group member decreases when someone acts negatively, but the likelihood of being an in-group member does not increase when someone acts positively. This could be because of the negativity bias: negative information has a greater impact than positive information on almost all psychological phenomena (Baumeister, Bratslavsky, Finkenauer, & Vohs, 2001; Rozin & Royzman, 2001), so negative actions of the target have more impact on social distance than positive actions.
Hypothesis 3. The hypothesized interaction between level of construal and action
valence on interpersonal and intergroup social distance was not found. The interaction could tie the findings of CLT and Action Identification, using mechanisms of bi-directional LIB effect. A large body of work (e.g. Linguistic Category Model (Semin & Fiedler, 1988) and Linguistic Expectancy Bias (Wigboldus et al., 2000)) seem to support a bi-directional effect of LIB, in that people draw inferences about people based on the level of abstraction of a message and action valence. These effects, however, does not seem to interact with each other. It is my personal opinion that a bi-directional effect of LIB provides a plausible explanation for the seemingly contradicting effects of CLT and Action Identification.
Age. As predicted by the socioemotional selectivity theory (Carstensen et al., 1999),
age covaries the effects on social distance. The older people are, the less social distance they experience, independent of the condition they are in. The effect of age was only found in interpersonal social distance and not intergroup social distance. A possible explanation could be that over time, people’s in-groups change. Groups of coworkers, friends and neighbors change, so by the time people are older, they are perhaps not as committed to their in-groups
as they are to other interpersonal relationships. Another possible explanation could be some missing variables in ‘colleague’ and ‘group of coworkers’, because participants were too old for having colleagues. Perhaps older people did not report more intergroup social proximity towards these groups, because they themselves did not belong to that in-group anymore.
Explorative Results
Having a target’s namesake in one’s circle of friends. In the explorative analyses,
having a friend with the same name as the target was taken into account. People who
indicated they have a friend with the same name as the target reported significantly less social distance than people who indicated they do not have a friend with the same name as the target, independent of the condition they were assigned to. This suggests that having a friend with the same name as the target could serve as (unconscious) point of reference while completing the study. It could be that people read the target’s name and immediately and unconsciously pictured their friend. This entails participants completing the study with their friend (unconsciously) in mind and not the target depicted. That could explain why they evaluated the target as more socially close than people who do not have a friend with the same name as the target: it was purely based on prior experience with their (socially close) friend and not based on the induced level of construal and/or effects of action valence.
I believe the use of a friend as point of reference did not stem from a priming-effect, because the question whether people have a Tom/Iris in their circle of friend was asked at the end of the questionnaire. It could, however, be possible that people read ahead and saw the question before they completed the manipulation. They had to turn the page of the
questionnaire to do that, but the lack of control in this field experiment makes this probable.
Level of construal. The findings in the explorative analyses show that level of
construal influences the experienced interpersonal social distance, but only for people who do not have a friend with the same name as the target. If people do not have a point of reference
for the name of the target, a high level of construal results in more interpersonal social distance than a low level of construal. This finding supports Hypothesis 1 and follows the predicted direction. This finding also provides some support for a bi-directional effect of CLT, in that not only more social distance leads to a representation in higher levels, but also that a higher level of construal leads to more social distance. Previous research on temporal distance (e.g. Liberman et al., 2007) established this bi-directional effect of CLT. The present study shows the bi-directional effect of CLT also holds up in interpersonal social distance, in line with Liberman and Förster’s third study (2009) and Stephan and her colleagues (2011).
It is, however, important to emphasize that the effect is only found within the group participants who do not have a friend with the same name as the target. It suggests that people experience less social distance towards someone whose name does not ring a bell, when they describe them in a concrete, detailed, superficial manner. A possible explanation could be that their representation of the target is not tainted with prior experience, so people are more perceptible for a construal manipulation than people who have a Tom/Iris in their circle of friends.
The effect of level of construal on interpersonal social distance does not occur with people who indicated they have a friend with the same name as the target. A possible explanation for this finding could be that they construed Tom/Iris in a certain way, because the name sounded familiar and they know a Tom/Iris, independent of the condition they were in. It is possible the manipulation was not strong enough to ‘overwrite’ the (unconscious) existing construal level with the construal level of the condition they were assigned to. For instance: Tom/Iris is a close friend, so he/she is construed using low levels, as predicted by CLT. The present manipulation, high or low, could not undo this pre-existing level of construal. Again, having a point of reference for the name of the target seems to determine
whether someone is susceptible for a level of construal manipulation and its subsequent effect on social distance.
Action valence. On interpersonal social distance, adding ‘having a friend with the
same name as the target’ as covariate does not change the above mentioned effects of action valence on interpersonal social distance: people experience more social distance as the target’s actions become more negative, independent whether they have a point of reference.
On intergroup social distance, however, some interesting shifts in results occur. For the four variables measuring intergroup social distance, all negative actions affects the likelihood a target is categorized as being an in-group member, compared to neutral or positive actions. But for two variables (‘colleague’ and ‘group of coworkers’), the likelihood decreases with negative actions, and for two variables (‘citizen’ and ‘nationality’) the
likelihood increases with negative actions. These ambiguous results could have three possible explanations.
First, the results could be skewed, because there are some abnormalities in the
distribution in action valence and having a friend with the same name as the target. Originally, the negative (N = 65), neutral (N = 70) and positive (N = 65) condition were distributed
equally. Yet, the group who indicated they have a Tom/Iris in their circle of friends (N = 103) contains only 10 participants in the negative condition and respectively 45 and 48 participants in the neutral and positive condition. A Pearson’s chi-square test shows that the distribution of action valence is not at random: χ2 (2, N = 199) = 50.57, p < .001. I think this is due to a systematic bias, with people less likely to indicate they have a Tom/Iris in their circle of friends when they were just confronted with negative actions performed by that person. It is likely that people in the negative condition maybe have a Tom/Iris in their circle of friends, but that friend does not resemble the Tom/Iris they just described, so they checked ‘no’. Another explanation could be that the question was open for multiple interpretations, in that
people thought they ought to indicate whether they have someone in their circle of friends who resembles the target they just described. The question could be misconstrued: instead of asking whether the participant has actually a Tom/Iris in their circle of friends, it was seen as “do I have someone in my circle of friends who behaves like the Tom/Iris I just described”. Both reasons could explain why the negative condition with people who have a Tom/Iris in their circle of friends was underrepresented. The valence results on intergroup social distance in the explorative analysis should therefore be treated with caution, because the data could be skewed due sample abnormalities.
Second, the reverse results on negative actions appear in whether the target could live in the same city or could have the same nationality. An explanation could be that both
variables seem to suffer from a ceiling effect. On the 7-point Likert-scale of ‘citizen’ and ‘nationality’, respectively 39% and 44% of the participants choose option 7: it is very likely the target lives in one’s city or shares one’s nationality. The target names Tom and Iris sounded Dutch, so the target could be Dutch or living in one’s city, independent of the target’s action valence. These items are therefore probably not sensitive enough to measure intergroup social distance.
Finally, the results of action valence on intergroup social distance could be
ambiguous, because another social psychological construct influences the results, namely Social Comparison Theory (Festinger, 1956). It could be that the positive actions of the target were seen as so good or noble that one’s self was threatened. Therefore, the participant used upward social contrast to maintain its positive view of self: the unbelievable good and moral Tom/Iris could impossibly be someone who belong in their in-group. This could explain why for instance on ‘nationality’ positive actions lead to more social distance than negative actions.
However, due to the sample abnormalities and ceiling effects, I think the results of action valence on intergroup social distance in the explorative analyses should be treated with caution.
Age. In the explorative analyses, a covariating effect of age on interpersonal and
intergroup social distance was only found for people who do have a friend with the same name as the target. A possible explanation for this covariating effect of age could be that the longer people know their Tom/Iris, the less social distance they experience towards them. Considering the aforementioned possibility that participants have completed this study with their friend (unconsciously) in mind, it seems logical that the older they are, the less social distance towards Tom/Iris they experience, because they probably know their friend longer. This is again in line with the socioemotional selectivity theory (Carstensen et al., 1999). The absence of a covariating age effect on interpersonal and intergroup social distance with people who do not have a Tom/Iris in their circle of friends could again emphasize the influence of having a point of reference while indicating perceived social distance.
Bi-directional LIB Effect
Because the predicted interaction between level of construal and action valence in Hypothesis 3 was not found, the bi-directional effect of LIB was not established. However, some interesting main effects of action valence and–under certain circumstances–level of construal on social distance were found. Therefore, I believe it is premature to dismiss the proposed link between Linguistic Intergroup Bias and Construal Level Theory. As Reitsema-van Rooijen and her colleagues (2007) point out, LIB is a very subtle bias in language use. It is likely that this subtle bias does not have an effect when people have a pre-existing point of reference for the target. Roughly half of the participants in the present study had a friend with the same name as the target, so this point of reference could ‘overrule’ the effects of LIB. That is indeed what the explorative analyses, controlling for the effect of having a target’s
namesake in one’s circle of friends, found. Without a friend with the same name as the target, both level of construal and action valence are significant, but also a non-significant interaction trend appears. That would imply that it depends on both level of construal and action valence how distant a stranger feels. Although this trend could provide ground to further explore the bi-directional effect of LIB on people without a point of reference for the target, I must make three important side notes.
First, this interaction was found with the sample abnormalities as described above, so the results could be skewed. Second, the interaction has a P value of .13, which is too high for any generalizations or strong conclusions, so it should be treated with caution. Finally, and perhaps most importantly, the interaction does not follow the hypothesized direction. As shown in Figure 1, people experience more social distance with a positive abstract description than a positive concrete description. Based on LIB, the opposite was expected in Hypothesis 3. It is therefore too premature to draw any conclusion from this findings.
Theoretical and Practical Implications
Having an (unconscious) point of reference with the target’s name could have
important implications for further research in the CLT domain, and in daily life. For research in CLT, an often used paradigm is to describe how or why someone performs a neutral action (developed by Freitas, Gollwitzer, & Trope, 2004; used in e.g. Kanten, 2011; Liberman et al., 2007; Stephan et al., 2011; Wakslak & Trope, 2009). The different pattern of results of level of construal displayed in the explorative analyses between people who have someone in their circle of friends with the same name as the target and people who have not, should add caution to the field of CLT. Most CLT research uses the target name ‘Ron’ and does not mention anything regarding effects of the name of the target, which is, in my opinion, an oversight.
Also, in day to day life the finding that people are less susceptible to construal level manipulation when they have a point of reference for the name could have some implications. It could imply that when you encounter someone new with the same name as a friend, your perception of that stranger could be influenced by your perception of his namesake. The namesake stranger could be construed using lower levels like your friend, which subsequently could lead to more social proximity than if the stranger did not have the same name as a friend. The inferences people draw from someone’s first name differ for common and uncommon first names (Leier, Hamilton, & Carpender, 1982). So, when you encounter someone with the same name as a friend, that name is common for you, which again can influence the perception of that stranger.
The present study established a bi-directional CLT effect on social distance for people who do not have a point of reference for the target, indicating that an abstract representation of a stranger leads to more interpersonal social distance. Besides the opportunities for further exploration in research on Construal Level Theory, this finding has practical implications, too. Apparently, it matters how one describes a person before encountering that person. People form (unconscious) implications and images of a stranger, partly based on the level of abstractness of the description, and evaluate that stranger as being socially close or not. The bi-directional effect of level of construal on interpersonal social distance has some theoretical implications, too. It adds support to the field of LIB, which explores the effect of language on the transmission and maintenance of prejudice and stereotypes (Maass, 1999). This study shows that an abstract (and in this case: linguistic) representation of someone’s actions results in less feelings of similarity, familiarity, and closeness. Furthermore, research by Reitsma-van Rooijen and her colleagues (2007) established that senders of abstractly written messages were evaluated with more interpersonal social distance than senders of
concretely written messages. This study suggests that this effect also holds up for self-written messages about a target.
An important practical implication of the findings in this study could be found in performance evaluation. If you have to write an evaluation of a presentation or performance of a stranger, the level of abstraction used in the evaluation could influence your social distance towards the assessed target. If you describe how someone acted, instead of why someone acted, you will probably feel more social proximity, because a low level of construal is used to represent the target. Also, writing an evaluation of a bad performance could lead to more interpersonal and intergroup social distance than writing an evaluation of a good performance.
Strengths and Limitations
Research in Social Psychology is known for its use of undergraduate Psychology students. The biggest strength and the biggest limitation of the present study is the use of a more random sample of the population, namely Dutch train passengers. It allows for a more diverse sample, covering participants from all around The Netherlands, with a wide age range and education range5. This sample allows for a more confident generalization of the found effects, because a more representable sample of the Dutch population was used, which makes this study more ecological valid.
Also the sample size (N = 227) with a more or less equal distribution of male (42%) and female (58%) participants adds to the strength of this study. The associated limitation of this field study is the lack of direct control while participants completed the study, and the little experience most participants have with a psychological questionnaire. Some participants consulted with their spouse during the completion of the test, or completed the questionnaire
5 A participant ‘warned’ me that my data could be biased, because I only used train passengers who were sitting
in the train on a Saturday. She thought that must have some confounding effects on my data, because it is a ‘type of person’ who sits in the train on Saturday. I thanked her for her concerns and explained what sample
population usually is used in psychological experiments.
in a random order. However, these side effects in my opinion do not outweigh the strengths of using train passengers. I think it is always important to test psychological constructs outside the borders of a University, with a more representative sample of the target population.
In this study, cheating was chosen as a negative action that participants ought to describe in a how or why manner. In practice, it entailed asking strangers in a train to describe how or why a certain person would cheat. Some participants asked whether they really had to answer that question and some people commented (social desirably) on how they had no idea how or why to do that. It suggests that another action might be more suitable for this study conducted in the train. Also, some missing values are due to the fact that some participants were too old for having colleagues.
Another limitation of this study is the weak measurement of intergroup social distance. The combined four variables measuring intergroup social distance could not be used as one reliable measure. Furthermore, two of the variables (‘citizen’ and ‘nationality’) suffered from a ceiling effect and are probably not sensitive enough to measure intergroup social distance. Although a moderate correlation (r = .5), ‘colleague’ and ‘group of coworkers’ seem to have overlap in what they are measuring: how likely it is the target is a colleague. Another example of an in-group instead of coworkers could perhaps provide a more diverse measurement of intergroup social distance. These limitations could explain the ambiguous or absent results in intergroup social distance, where the results in interpersonal social distance were clear or present. I think the different results in interpersonal and intergroup social distance are therefore due to the weak measurement of intergroup social distance and not actually differences between interpersonal and intergroup social distance.
It is a strength of this study that the question whether people have a friend with the same name as the target was included in the questionnaire. As shown in the explorative analyses, using common surnames without checking could contaminate the data. In the
present study, that cofounding effect could statistically be controlled for, thus providing a more internally valid study.
Further Research
Other research could build upon the interesting patterns of the non-significant interaction trend between level of construal and action valence on interpersonal social distance. The non-significant trend in an unexpected direction is too premature to interpret, but I think this bi-directional effect of LIB is plausible. Perhaps, this new hypothesized but undetected interaction effect could be explored in a more controlled experimental setting, for instance in a lab with undergraduate Psychology students. When doing so, a more sensitive measurement of intergroup social distance should be used. That ensures a more valid
measurement of the concept ‘social distance’ and allows to interpret the possible differences between interpersonal and intergroup social distance with confidence.
Furthermore, I propose that all further research using the CLT manipulation of
describing why or how someone performs an action should include questions about the name of the target. As shown in the explorative analyses, having a point of reference influences the results significantly. A question about whether the participant knows someone with the same name as the target, and a statistical comparison between both groups could shed more light on this unexplored–but important–side effect in CLT. The present study suggests that people who have an (unconscious) point of reference for the target could be less susceptible for construal level manipulation than people who do not have that point of reference. This is, however, a tentative hypothesis that should be tested experimentally.
Finally, selecting which (un)common first name to use in the manipulation should be grounded in theories about the effects of one’s first name (e.g. Leier et al., 1982). The present study explorative showed that the actual name of the target could have many additional effects on interpersonal and intergroup social distance.
Acknowledgements
First, I would like to thank dr. Michael Vliek of the Universiteit van Amsterdam for his supervision, his insights, his guidance and his patience. His enthusiasm about the subject reinforced my enthusiasm. Furthermore, I have to thank the anonymous NS-employees who let me conduct this study in the train while pretending to look the other away. Also, I would like to thank my friends and classmates for their interest and support. I deeply appreciate Nadine, Puck and Susan for their incredible friendship, encouragement and love. Finally, I owe a lot of my sanity, motivation, inspiration and strength to Femke. Thank you so much.
Bregje Sweep Tilburg, June 2014
