SEARCH

SEARCH BY CITATION

Keywords:

  • Similarity-Attraction;
  • When-Similarity;
  • Connectedness;
  • Intimacy;
  • Real-time Web

Abstract

  1. Top of page
  2. Abstract
  3. Study 1: The Effects of When-Similarity
  4. Study 2: When-Similarity and Who-Similarity
  5. General Discussion
  6. References
  7. Biographies

The feeling of connectedness experienced in computer-mediated relationships can be explained by the similarity-attraction effect (SAE). Though SAE is well established in psychology, the effects of some types of similarity have not yet been explored. In 2 studies, we demonstrate similarity-attraction based on the timing of activities—“when-similarity.” We describe a novel experimental paradigm for manifesting when-similarity while controlling for the activities being performed (what-similarity). Study 1 (N = 24) shows when-similarity attraction in the evaluation of connectedness with others. Study 2 (N = 42) identifies an interaction between who-similarity—similarity in personal backgrounds—and when-similarity. Both studies show that real-time computer-mediated interaction can lead to greater feelings of connectedness between people when there is an opportunity to discover when-similarity.

In their early stages, social networking sites (SNSs) focused primarily on sharing personal information via user profiles, enabling users to discover similarities in demographics, interests, and attitudes. A great deal of empirical evidence indicates that when users discover these types of similarities, even when this discovery is mediated, they become more attracted to each other (Montoya et al., 2008)—the so-called similarity-attraction effect (Byrne, 1971). Previous similarity-attraction effect manipulations include personality traits (Banikiotes & Neimeyer, 1981; Bleda, 1974), attitudes (Yeong Tan & Singh, 1995), ethnic backgrounds (Hu et al., 2008), facial features (Bailenson et al., 2008), and voice features (Nass & Brave, 2005), among others.

Similarity-attraction has been a topic of investigation for more than half a century, but the recent growth of mediated communication inspires new questions regarding similarity and interpersonal attraction. In this paper, we explore a previously unstudied type of interpersonal similarity: similarity in the timing of activities. That is, beyond knowing that someone else holds the same attitudes as oneself, shares a common background, or performs the same activities, is similarity-attraction amplified when two people know that they are doing the same activity at the same time? With the dramatic growth of real-time applications—such as Facebook and Twitter—one can be presented with contemporaneity information without actively interacting with the other person. Does this computer-mediated simultaneity influence people's feelings of connectedness and belonging? Does similarity-attraction extend to similarity in the timing of activities?

Background on The Similarity-Attraction Effect

The empirical evidence for similarity-attraction is so compelling that Byrne and Rhamey (1965) early on labeled the positive relationship between respondents' similarity and the attraction between respondents the Law of Attraction. After numerous replications in multiple domains using different similarity manipulations, Berger (1975) proclaimed that similarity-attraction is “one of the most robust relationships in all of the behavioral sciences.”

The explanations for the origin of similarity-attraction are, however, multifold and often disputed. Initially, no explanations for the effect were offered, and similarity-attraction was regarded to be self-evident—as we see even today by the lack of explanations for similarity-attraction in many social psychology textbooks (Heine et al., 2009). When explanations are proposed, one of the most popular views is based on people's innate desire to be consistent with societal norms and values. This explanation assumes that the discovery of interpersonal similarity leads to the validation of one's own characteristics and views by providing consensus support (Byrne & Clore, 1970). The validation then leads to a higher perceived appropriateness of one's current beliefs, attitudes, behaviors, and traits.

Byrne (1971) and Clore and Byrne (1974) extended this explanation in their formulation of the reinforcement-affect model. This model is based on the assumptions that: (a) people experience stimuli as rewarding or punishing and seek out those that are rewarding, (b) positive feelings—affect—are associated with rewarding stimuli, and (c) other people are liked or disliked according to their association with rewarding or punishing stimuli. That is, we learn to associate positive feelings with people that are linked to rewards. Instances of interpersonal similarity function as rewarding stimuli, which leads people to associate positive feelings with similar others, which in turn leads people to be more attracted to similar others.

Another popular explanation for similarity-attraction is the emergence of positive feelings stemming from smooth and rewarding interactions, which are more likely to arise when communicating with people who are similar to oneself (Berscheid & Walster, 1978) than when communicating with dissimilar others. While this explanation is similar to the explanation of Clore and Byrne (1974) in its emphasis on rewards experienced by positive stimuli, Berscheid & Walster (1978) specifically focus on the smooth interactions that are likely to occur with similar others as opposed to the positive feelings stemming from discovering instances of interpersonal similarity. Through these smooth interactions with others, similarity-attraction partly satisfies a person's “Need to Belong” (Baumeister, 1995).

Other explanations put forward for similarity-attraction include enhanced reciprocal liking towards similar others (Condon & Crano, 1988) or a desire to satisfy implicit egoism (Jones et al., 2004). One final explanation for the similarity-attraction effect inverts these arguments: A person's default state is to like everyone, and dissimilarity leads to repulsion (Rosenbaum, 1986). Rosenbaum supports this view by showing several experiments in which the standard experimental paradigm of the SAE is extended by adding a “no-interaction” control group. His work shows that attraction ratings in the no-interaction and in the similar other groups are comparable, while the ratings in the nonsimilar other group are significantly lower.

Despite the overwhelming empirical evidence supporting the existence of similarity-attraction as well as the numerous plausible mechanisms identified to explain it, several questions have been raised regarding the importance and integrity of the effect. Some authors have discounted the effect as merely resulting from demand characteristics operating in experiments (Sunnafrank, 1991) or other methodological flaws (Bochner, 1991). Morry (2007) questions the claimed causality of similarity-attraction, and Sunnafrank and Miller (1981) demonstrate that similarity-attraction is diminished greatly when allowing for initial interactions between participants in laboratory studies. Overall, these results lead critics to conclude that similarity-attraction only exists in a laboratory setting using ad-hoc dyads and is of no practical importance in a real-world setting. Meta-analysis of numerous SAE studies both within and outside of the laboratory indeed shows that while similarity manipulations have strong effects on attraction in laboratory settings, these effects are limited in real long-term relationships (Montoya et al., 2008). In the next section, we will explain why these criticisms do not discredit similarity-attraction as an important psychological effect in real-time computer-mediated interactions.

Types of Similarity in New Media

The most common laboratory paradigm in similarity-attraction research is the phantom-other technique (Smith, 1957; Byrne, 1961). In laboratory experiments using the phantom-other technique, the target is often unknown to participants. Participants are then presented with details about the nonpresent target, such as age, personality, judgments, or social status. This phantom-other technique leads to the strongest similarity-attraction effects.

Coincidentally, these laboratory characteristics frequently hold for a wide variety of computer-mediated communications that take place in new media and SNSs: A large number of “friends” in people's social networks are relatively unfamiliar, not physically present, and their profiles present details not commonly discovered in face-to-face interactions. Thus, while similarity-attraction has been hard to replicate outside of the laboratory, the situation is now reversed: Real life, through SNSs and other mediated communication, has replicated the laboratory conditions in which similarity-attraction was first observed. It is thus plausible that the laboratory studies that support the similarity-attraction effect will possess high external validity in these new media contexts.

There are many types of similarity encountered by people when using new media like SNSs. Initially, SNSs enabled people to explore and experience “who-similarity": similarity in demographic features such as ethnic background or religious affiliation. As personal profiles and information streams on SNSs grew, people were also able to discover “what-similarity": similarity in attitudes, activities, and hobbies. The use of social GPS tracking, as is done by applications like Foursquare, has even increased the salience of “where-similarity": similarity in location.

Recent advancements—specifically, real-time social technologies—have created the ability to discover a new type of similarity that has not been previously examined either in the laboratory or in the field: when-similarity. While it was already possible to connect with remote others in real-time since the emergence of chat rooms and instant messaging (see e.g. Baker, 2008), experiencing co-occurrence of (remote) activities in time without an explicit conversation or shared activity is relatively novel. In this paper, we create a situation in which co-occurrence of remote activities is experienced without introducing additional confounds that would naturally arise during an explicit conversation.

Real-time services like Facebook and Twitter enable users to experience similarity with others in the timing of activities. With millions of people broadcasting their current activity status in real-time repeatedly during the day, it is highly likely users will experience some form of when-similarity: Users discover that at the point in time that they are carrying out a specific activity, someone in their extended social network is carrying out that same activity, without an active conversation or conscious joint activity of the two parties involved. Due to the growing prominence of real-time services both on the web as well as on other devices such as mobile phones, it is worthwhile to explore whether this new type of similarity also enhances attraction and thus supports the similarity-attraction effect.

Overview of the Studies

In this paper, we describe two studies in which when-similarity (Study 1) and both when- and who-similarity (Study 2) are manipulated. The effects on the participant's perception of a target other are measured. Based on the strong evidence that supports similarity-attraction in settings similar to those experienced by users of real-time services, we hypothesize that similarity-attraction will also hold for this new type of similarity. That is, we expect that users who discover when-similarity with a target through computer-mediated communication channels will feel closer to the target. This expectation is already supported by a number of sociological investigations into the effects of the timing of events such as religious festivities: When a group of individuals perform an activity or ritual at the same time, simultaneity plays a key role in the formation of feelings of connectedness within the social group. (Durkheim, 1912 / 2008; Horton, 1967; Lee & Liebenau, 2000; Zerubavel, 1982).

In both of the presented studies, we use a novel method to manipulate when-similarity while controlling for other types of similarity. We are the first to isolate the effects of the timing of activities from confounds such as the type of activity being performed. The mediated nature of the interactions in the experiment enables us to manipulate the timing of activities while keeping the pattern of activity types constant and while avoiding confounds caused by physical proximity. Our manipulation of when-similarity differs from traditional experimental manipulations in that the two experimental groups—those high and low in when-similarity—are created dynamically based on the behavior of participants. We explain this method in detail.

Study 1: The Effects of When-Similarity

  1. Top of page
  2. Abstract
  3. Study 1: The Effects of When-Similarity
  4. Study 2: When-Similarity and Who-Similarity
  5. General Discussion
  6. References
  7. Biographies

Method

This study tests the effects of when-similarity—while controlling for other types of similarity—on the participants' evaluations of others. Interpersonal attraction, after a weeklong intervention manipulating when-similarity, was measured using both social connectedness and intimacy scales. When-similarity was manipulated between participants.

Participants

Twenty-four United States college students participated in this balanced, between-participants experiment. Thirteen (54.2%) participants were female, and gender was balanced as much as possible across the conditions. The average age of respondents was 20.1 (SD = 1.61) years. Participants received partial course credit for their participation.

Procedure

First, participants were asked to complete a brief personal profile, which asked for their gender, age, area(s) of academic study, and favorite pastimes. Participants were then ostensibly partnered with another participant of the same gender. Participants were provided with the name of their partner for the duration of the experiment, and no other information about the partner was disclosed. In reality, participants' partners were not actual participants, and partners' names were selected to provide a perceived gender match with the participant.

Next, participants received six text messages per day on their mobile phones over the course of five days. Each message contained the question: “What are you doing right now?” Participants were instructed to reply with a number from 1 to 5, which represented different behavioral categories: 1: Eating, 2: Studying, 3: Physical Activity, 4: Relaxing, and 5: Working. These categories were pretested to resonate with the participant population and were found exclusive as well as exhaustive: a 2-day pretest showed that for each moment in the day that they were queried, students (N = 11) were able to pick exclusively one of the five provided categories as their current activity.

A few minutes after responding to the text message, participants received a follow-up message stating the activity their partner was ostensibly performing at that same point in time, and whether their own activities matched their partner's. Participants were not provided with any information about their partner other than these activity messages. After 5 days of text messages, summing to 5 × 6 = 30 messages, participants answered an online questionnaire to evaluate their partner and their perceived relationship with their partner.

We chose text messaging as our medium, as opposed to an SNS, since respondents would have direct access to their devices to be able to report on their activities at the moment the messages were received. Furthermore, the text messages allowed us to fully control the conversation with the ostensible other with less risk of directed online searches to get in contact via other means. Thus, text messaging was chosen mainly for methodological ease. We do not feel our results are restricted to text messaging but rather are representative of a much broader class of mediated real-time interactions.

When-Similarity Manipulation

To manipulate when-similarity while controlling for other types of similarity, half of the participants were assigned to the Similar Timing condition and half were assigned to the Dissimilar Timing condition. In both conditions, to keep constant what the partner was doing, the ostensible partner responded in such a way that after all six text messages had been sent for a specific day, the partner had studied twice and performed each of the other behavioral categories only once. This constraint on the activity pattern of the ostensible partner was imposed to prevent the actual activity from influencing the perceptions of the partner. By providing the same “activity profile” in both of our conditions, we minimize the effect of what-similarity.

The two conditions differed only in when the participants were told their partners were performing these activities. In the Dissimilar Timing condition, the response, when possible, consisted of a different behavioral category than the one performed by the participant at that point in time. For example, if the participant indicated she was “eating” at the time she received a message, the response message would be any (random) behavioral category other than eating—unless all other behavioral categories had already been exhausted that day, leaving “eating” to be the only valid remaining response. In the Similar Timing condition, the response was, when possible, in the same behavioral category. Since no background information about the participants' ostensible partners was presented, this experimental setup also controlled for possible confounds of who- or where-similarity (beyond the matching gender).

Validity of the When-Similarity Manipulation

Something to keep in mind is that ostensible partner responses generated by our algorithm depend on the activities that are performed by our participants. Thus, to evaluate the validity of our when-similarity manipulation, we need to determine whether there would always be a difference between the number of simultaneous activity occurrences for the Similar and Dissimilar Timing groups regardless of whether our participants' overall activity pattern were different. We addressed this by conducting simulations of our response algorithm for different possible participant activity patterns. For N = 20 participants per simulation (M = 1000), we took six draws from a five-category multinomial—corresponding to the six activity messages sent daily. By changing the initial probabilities of the five activity categories, we were able to test the consistency of our algorithm's responses to different participant activity patterns (i.e., what are the generated responses from the ostensible partner if the participant performs the same activity at all times?).

Figure 1 shows the simulation results for the actual observed activity probabilities in our study (Row 1), a flat activity pattern in which the simulated participants performed each of the activities equally often (Row 2), and a severely skewed distribution of activities in which simulated participants spent far more time studying than performing any of the other activities (row 3). The distribution of responses by the ostensible partner (Column 2) is the same for all simulations due to the restricted daily activity pattern implemented to control for what-similarity (i.e., study twice a day and perform all other categories only once). Column 3 shows the distribution of the percentage of simultaneous activities in the Similar Timing and the Dissimilar Timing conditions—that is, the times in the simulation when the activities of the participant and the ostensible partner would match. In each of the simulated scenarios, including the severely skewed scenario (Row 3, Column 3), the number of matches in activity responses produced by the algorithm differed significantly between the Dissimilar Timing and Similar Timing conditions. Thus, our when-similarity manipulation creates distinct numbers of activity matches between the Similar Timing and the Dissimilar Timing groups while controlling for what-similarity even in situations where the activities of the participant are skewed1.

image

Figure 1. Evaluation by means of simulation of the when-similarity algorithm for different true population distributions of activities. The column on the right shows the distribution of the number of matches over all simulated experiments. Rows show different participant activity patterns: Row 1 shows the actual activity pattern of our participants, Row 2 shows an evenly distributed activity pattern, and Row 3 shows a highly skewed pattern. Even in the last case, the response generated by our matching algorithm led to a significant difference in the number of matches between the two conditions.

Download figure to PowerPoint

In this study, the activity of the ostensible partner matched that of the participant 50.6% of the time in the Similar Timing condition and 3.0% of the time in the Dissimilar Timing condition. Thus, the actual number of instances of experienced when-similarity in the two conditions is in accordance with those produced in the simulations.

Measures

Participants evaluated their partner after the one-week manipulation using three rating scales. The first scale was a 6-item, 7-point Social Connectedness Scale (Cronbach's α = 0.94) (Van Bel et al., 2009). The endpoints of the items were labeled “(1) Totally disagree” to “(7) Totally agree.” This social connectedness scale consisted of items addressing the feelings of closeness and shared thoughts between the participant and their partner (e.g. “I often know what my partner feels” and “I feel that my partner often knows what I think”).

The second scale was the Inclusion of the Other in the Self (IOS) Scale. The IOS measures perceived intimacy (Aron et al., 2006) using a single, 7-point pictorial item. Each of the pictures shows two circles labeled “You” and “Your Partner.” In each picture, the circles overlap more and more—from nontouching to almost fully overlapping. Participants are asked which of the pictures most closely represents the relationship with their ostensible partner.

Finally, participants were asked to state how much they agreed with the statement “I would like to meet my partner” on a 7-point scale. The end-points of this scale—which was specifically designed for the purposes of this experiment—were labeled “(1) Completely disagree” to “(7) Completely agree.”

Results

A MANOVA with when-similarity as an independent factor and the three attitudinal measures as dependent factors showed a strong multivariate main effect of when-similarity-attraction on the overall partner evaluations, F(3,20) = 10.58, p < 0.001, η2 = 0.61. Table 1 presents the mean scores of each dependent variable and the outcomes of separate t-tests for the effect of when-similarity. For each of the three dependent measures, participants in the Similar Timing condition scored significantly higher than participants in the Dissimilar Timing condition.

Table 1. Comparisons of attitudes towards the ostensible partner in the Dissimilar Timing and Similar Timing conditions. N = 24
MeasureMDissimilar Timing (SD)MSimmilar Timing (SD)t (22)p
Connectedness1.74 (0.76)3.54 (0.94)5.190<.001
Intimacy1.33 (0.49)2.42 (0.99)3.377<.01
“Would like to meet…”2.67 (1.23)3.92 (1.51)2.227<.05

Discussion

Study 1 examined the effects of similarity in the timing of activities—when-similarity—on participants' evaluations of others using a method that controlled for other types of similarity. This experimental manipulation of when-similarity produced a large difference in the number of matching simultaneous activities between our experimental conditions while controlling for the patterns of activities performed during the day. The latter constraint controls for what-similarity but also ensures a realistic activity pattern performed by the ostensible partners of participants in this study. The ostensible partner-generated activity patterns are very close to the activity patterns performed by the participants themselves (compare Figure 1, column 1, row 1 & 2).

Participants in the Similar Timing condition evaluated their partners more positively than those in the Dissimilar Timing condition: Participants that experienced a similar timing of activities felt more connected to their partner, felt more intimate, and were more eager to meet their partner than those who did not experience timing similarity. These findings support our hypothesis that similarity-attraction holds for this previously unstudied type of similarity. The finding extends similarity-attraction to when-similarity in a computer-mediated setting that is a common experience for people using real-time web services.

Besides the practical importance of showing that when-similarity can have positive effects on people's evaluations of others in SNSs, these results and the proposed method can be further used to study the interactions between different types of similarity and even to evaluate possible similarity-attraction explanations. Historically, similarity-attraction explanations were concerned largely with who- or what-similarity. The existence of a positive when-similarity attraction effect in cases for which who- and what-similarity is absent could partially invalidate explanations that are solely or heavily dependent on who people are and/or what they do. To explore this further and test the robustness of the when-similarity effect, Study 2 combines a manipulation of when-similarity with a more traditional manipulation of who-similarity.

Study 2: When-Similarity and Who-Similarity

  1. Top of page
  2. Abstract
  3. Study 1: The Effects of When-Similarity
  4. Study 2: When-Similarity and Who-Similarity
  5. General Discussion
  6. References
  7. Biographies

In Study 2, we used a similar experimental protocol as detailed in Study 1 to examine the possible interaction between when-similarity and who-similarity. This interaction is of practical importance because these two types of similarity are often experienced together when using SNSs as well as other forms of computer-mediated communication. In addition to their practical importance, interactions between different types of similarity are also theoretically interesting: Hypotheses about these interactions would differ based on which explanation of similarity-attraction to which one subscribes. When following the popular explanation that similarity-attraction is caused by a desire to validate one's own beliefs, when-similarity would have no effect (which is not consistent with Study 1). If, however, an explanation of similarity-attraction is not necessarily tied to this type of similarity and rather encompasses a holistic view in which any type of additional similarity increases attraction, one would expect separate main effects of both when- and who-similarity.

Method

Study 2 examined the effects of both when-similarity and who-similarity simultaneously. A 2 (when-similarity: Similar Timing vs. Dissimilar Timing) × 2 (who-similarity: Similar Profile vs. Dissimilar Profile) between-participants experiment was created to test the effects of both when-similarity and who-similarity on the evaluations of others.

Participants

Participants in this study consisted of 17 male (38.6%) and 27 female (61.4%) United States college students with an average age of 20.9 (SD = 1.5) years. Gender was evenly balanced as much as possible across conditions, and participants again received partial course credit for their participation in this study. None of the participants had participated in Study 1.

Procedure

The procedure in this study was very similar to the procedure in Study 1 with one minor change: after participants were told that they would be paired with another person—during the introduction questionnaire—they were shown a profile of their ostensible partner. The profile contained information about their partner's gender, age, academic focus, and three favorite pastimes. These variables were chosen to link directly to the types of information presented in profiles that are typically used in SNSs.

When-Similarity Manipulation

The when-similarity manipulation was similar to that described in Study 1. In this study, activities matched temporally 52.2% of the time in the Similar Timing condition and 2.5% of the time in the Dissimilar Timing condition. Again, keeping the ostensible partner's activity pattern constant within each day of the experiment controlled for what-similarity.

Who-Similarity Manipulation

Based on the participant's profile, their partner's profile was dynamically generated as employed in the phantom-other technique (Smith, 1957). The partner's profile was adjusted to implement the two experimental conditions. In the Dissimilar Profile condition, the participant's partner was randomly 3 to 4 years older or younger, pursued a different academic focus, and had at least two different pastimes. In the Similar Profile condition, the partner was randomly one year older or younger, pursued the same academic focus, and had two similar pastimes. Who-similarity was thus manipulated on multiple dimensions. In all cases, the partner's gender matched the participant's gender.

This manipulation of who-similarity in which the difference in age, academic focus, and favorite pastimes are all simultaneously manipulated may seem overstated. However, to examine the robustness of when-similarity, it was important to have a strong manipulation of who-similarity and -dissimilarity. The manipulation was not unrealistic: frequently on SNSs, people are exposed to profiles of other people with whom they have a single or very few commonalities (e.g., a shared alma mater) but who are, in most other respects (e.g., their academic focus, age, and favorite pastimes), different from themselves.

Measures

Study 2 used the same dependent variables as Study 1: Connectedness (Cronbach's α = 0.93), Intimacy, and Willingness to Meet. Different than in Study 1, participants answered these questions both directly after the who-similarity manipulation when they received their partner's profile—prior to the experience of when-similarity—and at the end of the text messaging intervention period—after the experience of when-similarity.

Results

Pre-When-Similarity Manipulation

Participants were asked to rate their initial impressions of their partner directly after reading their profile and thus before the when-similarity manipulation. A MANOVA with who-similarity as an independent factor and the three attitudinal measures (i.e. Connectedness, Intimacy, and Willingness to Meet) as dependent factors showed a strong multivariate main effect of who-similarity on the overall partner evaluations, F(1,42) = 7.02, p < 0.001, η2 = 0.36. Table 2 shows separate t-tests for each of the dependent measures. Participants in the Similar Profile condition felt more connected to and more intimate with their partner, consistent with previous findings using the phantom-other technique. The who-similarity manipulation failed to influence whether or not participants wanted to meet their partner, although the difference was in the expected direction.

Table 2. Comparisons of attitudes in the Dissimilar Profile and Similar Profile conditions after reading the profile. N = 44
MeasureMDissimilar Profile (SD)MSimilar Profile(SD)t (42)p
Connectedness3.26 (1.28)4.69 (1.18)3.847<.001
Intimacy2.09 (0.87)3.23 (0.92)4.209<.001
“Would like to meet…”3.77 (1.23)4.09 (1.51)1.059.30
Post-When-Similarity Manipulation

Participants also rated their impressions of their partner after the weeklong text messaging intervention—the experimental manipulation of when-similarity. In the 2x2 MANOVA, there was a significant multivariate main effect of when-similarity, F(3,38) = 6.61, p < 0.001, η2 = 0.34, such that participants who worked with ostensible partners who had similar timing of activities were much more attracted to their partners than those participants who had a partner with dissimilar timing. There was no multivariate main effect of who-similarity, F(3,38) = 1.26, p > 0.05, η2 = 0.09, although the means were in the expected direction.

There was a significant interaction effect between who- and when-similarity, F(3,38) = 7.46, p < 0.001, η2 = 0.37, presented in Figure 2. The effect of when-similarity is strong and positive when the partner is initially perceived as dissimilar based on the presented profile information. If, however, attraction is already established based on a similar profile, the additional when-similarity does not increase attraction.

image

Figure 2. Estimated marginal means and standard errors of the similarity-attraction scores for the four experimental groups.

Download figure to PowerPoint

Table 3 presents separate univariate results to provide a more detailed look at the data. Participants' feelings of Connectedness and Perceived Intimacy with their partners were influenced by the when-similarity manipulations. Again, the Willingness to Meet question—as was the case after the who-similarity manipulation—does not exhibit an effect of the similarity manipulations, although the means were in the expected direction.

Table 3. Results of both the Who- and When-similarity manipulations
MeasureMeans (SE)F (1, 40)η2p
EffectDissimilarSimilar
ConnectednessWho3.1 (0.23)3.4 (0.23)0.8870.022.352
 When2.7 (0.23)3.8 (0.23)9.3300.189.004
 Interaction  11.3970.222.002
IntimacyWho1.8 (0.19)2.3 (0.19)2.8250.066.101
 When1.5 (0.19)2.6 (0.19)19.0960.323.001
 Interaction  0.1130.003.739
“Would like to meet…”Who3.0 (0.28)3.6 (0.28)1.8700.045.179
 When3.1 (0.28)3.5 (0.28)0.4680.012.498
 Interaction  0.8310.020.367

Conclusion

The results of the who-similarity manipulation directly after the presentation of the partner profiles replicate the laboratory-setting finding that who-similarity affects people's evaluations of others positively: Sharing of a common background leads to more attraction when no other information about the partner is provided.

In Study 2, no results were found regarding the willingness to meet the ostensible partner based on both who- and when-similarity, but this item has limited validity and reliability: It queries a behavioral intention of the participants instead of a judgment about the ostensible partner and is a single item. After a week of interactions geared for the discovery of when-similarity, while controlling for who- and what-similarity, the main effect of who-similarity was weakened. This is probably due to the fact that throughout the week, the ostensible partner only matched the participant's activities 50% of the time, leading to a weakening of the initial manipulation of extremely high similarity.

The reinforcement-affect model provides an interesting means for understanding these results. The discovery of similarity or dissimilarity functions as a reward or punishment, which in turn leads to a change in affect towards similar or dissimilar others. A classical finding in the operant conditioning literature shows that rewards or punishments are not additive but rather elicit satiety: the effectiveness of reinforcement reduces once an individual's need for that reinforcement has been satisfied (see, e.g., Guttman, 1953). Under this view, while who-similarity acts as a reward and in turn produces positive affect, any potentially positive effects of additional similarity are limited.

The results of Study 2 do however indicate that a weak attraction to others based on encountering a dissimilar profile in SNSs can be overcome by the synchronous timing of activities. In cases of initial dissimilarity (a dissimilar profile), the reward of discovering when-similarity leads to more positive evaluations.

General Discussion

  1. Top of page
  2. Abstract
  3. Study 1: The Effects of When-Similarity
  4. Study 2: When-Similarity and Who-Similarity
  5. General Discussion
  6. References
  7. Biographies

The opportunity for people to experience when-similarity—without experiencing other types of similarity simultaneously—has emerged due to the recent growth of real-time technology. Services like Twitter enable people to discover what people who are essentially strangers are doing at exactly that moment in time. The mediated nature of these services enables a separation of when-similarity from what-similarity (and physical proximity) that was virtually unimaginable before the existence of real-time mediated interactions. Current communication practices, mediated through computers, mobile phones, and other devices, affect the nature and structure of our encounters with interpersonal similarity—in some respects bringing them closer to the encounters that occurred in past laboratory studies examining similarity-attraction. Our two studies show that when no other information about the other is provided (Study 1) or when the other is regarded as dissimilar based on a number of demographics and interests (Study 2), when-similarity leads to attraction and a more positive evaluation of the other. Greater feelings of connectedness towards the other are reported, and the perceived intimacy between people is increased upon encountering when-similarity.

We introduce a new method to study when-similarity in a setting that has high external validity. The presented manipulation enables control over other types of similarity. In both of our studies, this manipulation of when behaviors occurred while controlling for daily behavioral patterns.

When-Similarity Manipulation and Interactions

The when-similarity manipulation presented in this paper is qualitatively different from most manipulations found in experimental psychology or communication studies. While participants were randomly assigned to a fixed condition, the stimuli they received in these respective conditions were not deterministic. Due to our imposed constraint, which controlled for what-similarity—and thus ensured that participants' partners had the same activity pattern in both conditions—the generated responses differed based on the activity patterns of our participants. However, through simulations including severely skewed activity distributions for participants, we showed that the manipulation consistently led to differing numbers of time-matched activities between the two conditions. We hope that this manipulation can be useful in further research into the effects of similar timing of activities within the similarity-attraction paradigm.

Our when-similarity manipulation also created a clear distinction between different types of similarity that can be experienced by people on SNSs. While within the similarity-attraction literature, numerous moderating variables have been investigated (Pilkington & Lydon, 1997), there is no work detailing interactions between the different types of similarity as introduced in our studies. However, these interactions are of great interest because they enable studying a possible additive effect of increased similarity and can also be used to further examine the explanations offered for similarity-attraction.

Explanations for Similarity-Attraction

As noted in the introduction, a popular explanation for similarity-attraction put forward by Byrne and Clore (1970) is based on people's need for accuracy: Finding people who share similar attitudes or background corroborates one's own beliefs and as such, positively enforces one's feeling of accuracy. While this explanation is very plausible for what-similarity—similarity based on shared attitudes or beliefs—given that finding similar others indeed reinforces the accuracy of consciously chosen and held beliefs, it is debatable that this explanation holds for every type of similarity. Already for who-similarity, in which characteristics that are not consciously chosen or selected by respondents are manipulated, it is harder to argue that, for example, being of the same gender enforces one's motivation towards accuracy. In Study 2, the who-similarity manipulation included several factors and thus its effectiveness does not necessarily invalidate the motivation-for-accuracy explanation. However, for the when-similarity manipulation employed in this study, it is not very plausible that the need for accuracy is an appropriate explanation for the established effect: One would have to subscribe to the assumption that people strive for accuracy in their timing of relatively trivial activities—such as eating—and that finding someone eating at some other time threatens their own accuracy perception. Future work, through manipulating different similarity types, can test conflicting hypotheses elicited from the different similarity-attraction explanations. In this way, the new types of similarity emerging in computer-mediated communication can aid our understanding of important social phenomena.

Our results are best explained not by a need for accuracy but rather by the predictions one would make based on the reinforcement-affect model. In this model, similarity-attraction is explained by the idea that similarity (or dissimilarity) functions as a reward (or punishment). Rewards are then associated with positive feelings which in turn lead to both positive affect towards similar others as well as a tendency to actively seek out similar others. The literature on operant conditioning however shows that both rewards and punishment are not additive: Once a person's need for a certain reward has been satisfied, the effectiveness of that reward reduces. This explains the interaction between who- and when-similarity observed in Study 2: Both who- and when-similarity can increase people's evaluations of others; however, once one is established, the additional effect of more similarity (or different types of similarity) is reduced.

Future Work

In this article, we presented when-similarity and the idea that this particular type of similarity is emergent in mediated communication. Next, we described a method to study this type of similarity while controlling for other types of similarity and have shown the method's effects on interpersonal evaluations. However, the results presented in this article still require further explanation: Possible conflicting explanations for the origins of the interaction between different types of similarity need to be examined in more detail. In particular, a more in-depth examination of the interactions between other types of similarity (e.g. what-similarity and who-similarity) is clearly called for. A better understanding of the effects demonstrated in this paper would also emerge if the exact nature of the reinforcing process that is likely in play when experiencing when-similarity is clarified: Does similar timing reinforce one's own previous choices? Or does similar timing create a bond through more subtle means identical to the sharing of physical space (e.g. Gibson, 1984)?

Finally, we believe it is also important for future research to look at the cognitive processes that are activated when experiencing when-similarity: these processes might provide a more developed explanation for the interaction effects observed in Study 2.

Practical Implications

Web services like Twitter and Facebook show the activities of other users at a given moment. Critics of the impact of social real-time technologies have argued that people's urge to post status updates to SNSs can be explained merely by an egocentric urge for self-exhibition. Our two studies of the effects of when-similarity, the type of similarity that is primarily experienced using services like Twitter, suggest that another effect could be in play: people's need to belong (Baumeister, 1995) is partly satisfied by the discovery of when-similarity. Hence, real-time services can serve a very social goal: enhancing connectedness by emphasizing when-similarity.

Note
  1. 1

    The exact algorithm to generate the replies is available in [R] (used for the simulation study) or PHP (used for the empirical studies) upon request.

References

  1. Top of page
  2. Abstract
  3. Study 1: The Effects of When-Similarity
  4. Study 2: When-Similarity and Who-Similarity
  5. General Discussion
  6. References
  7. Biographies
  • Aron, A., Aron, E. N., & Smollan, D. (2006). Inclusion of other in the self scale and the structure of interpersonal closeness. Journal of Personality and Social Psychology, 63, 596612.
  • Bailenson, J. N., Iyengar, S., Yee, N., & Collins, N. A. (2008). Facial similarity between voters and candidates causes influence. Public Opinion Quarterly. 72(5), 935961.
  • Baker, A. J. (2008). Down the rabbit hole: The role of place in the initiation and development of online relationships. In A. Barak (Ed.), Psychological aspects of cyberspace: Theory, research, applications. Cambridge, UK: Cambridge University Press.
  • Banikiotes, P. G., & Neimeyer, G. J. (1981). Construct importance and rating similarity as determinants of interpersonal attraction. British Journal of Social Psychology, 20, 259263.
  • Baumeister, R. F. (1995). The need to belong: Desire for interpersonal attachments as a fundamental human motivation. Psychological bulletin, 3, 497529.
  • Berger, C. R. (1975). Task performance and attributional communication as determinants of interpersonal attraction. Speech Monographs, 40, 280286.
  • Berscheid, E., & Walster, E. H. (1978). Interpersonal attraction. McGraw-Hill College.
  • Bleda, P. R. (1974). Toward a clarification of the role of cognitive and affective processes in the similarity-attratction relationship. Journal of Personality and Social Psychology, 29, 368373.
  • Bochner, A. P. (1991). On the paradigm that would not die. Communication Yearbook, 14 484491.
  • Byrne, D. (1961). Interpersonal attraction and attitude similarity. Journal of Abnormal and Social Psychology, 62, 713715.
  • Byrne, D. (1971). The attraction paradigm. Academic Press.
  • Byrne, D., & Clore, G. L. (1970). A reinforcement model of evaluative responses. Personality: An International Journal, 1, 103128.
  • Byrne, D., & Rhamey, R. (1965). Magnitude of positive and negative reinforcements as a determinant of attraction. Journal of Personality and Social Psychology, 2, 884889.
  • Clore, G. L., & Byrne, D. A reinforcement-affect model of attraction. In T. L. Huston (Ed.), Foundations of interpersonal attraction. New York: Academic Press, 1974.
  • Condon, J., & Crano, W. D. (1988). Inferred evaluation and the relation between attitude similarity and interpersonal attraction. Journal of Personality and Social Psychology, 54, 789797.
  • Durkheim, E. (1912/2008). The elementary forms of religious life. trans. by C. Cosman. New York: Oxford University Press.
  • Gibson, W. (1984). Necromancer. New York: Ace.
  • Guttman, N. (1953). Operant conditioning, extinction, and periodic reinforcement in relation to concentration of sucrose used as reinforcing agent. Journal of Experimental Psychology, 46(4), 213224.
  • Heine, S. J., Foster, JA. B., & Spina, R. (2009). Do birds of a feather universally flock together? Cultural variation in the similarity-attraction effect. Asian Journal of Social Psychology, 12, 247258.
  • Horton, J. (1967, April). Time and cool people. Society, 4(5), 512.
  • Hu, C., Thomas, K. M., & Lance, C. E. (2008). Intentions to initiate mentoring relationships: Understanding the impact of race, proactivity, feelings of deprivation, and relationship roles. The Journal of Social Psychology, 148(6), 727744.
  • Jones, J. T., Pelham, B. W., M., C., & Mirenberg, M. C. (2004). How do I love thee? let me count the ways: Implicit egotism and interpersonal attraction. Journal of Personality and Social Psychology, 87, 665683.
  • Lee, H., & Liebenau, J. (2000). Time and the internet at the turn of the millennium. Time and Society, 9(1), 4356.
  • Montoya, R. M., Horton, R. S., & Kirchner, J. (2008). Is actual similarity necessary for attraction? A meta-analysis of actual and perceived similarity. Journal of Social and Personal Relationships, 25(6), 889922.
  • Morry, M. M. (2007). Relationship satisfaction as a predictor of perceived similarity among cross-sex friends. Journal of Social and Personal Relationships, 24, 117138.
  • Nass, C. & Brave, S. (2005). Wired for speech: How voice activates and advances the human-computer relationship. Cambridge, MA: MIT Press.
  • Pilkington, N. W., & Lydon, J. E. (1997). The relative effect of attitude similarity and attitude dissimilarity on interpersonal attraction: Investigating the moderating roles of prejudice and group membership. Personality and Social Psychology Bulletin, 23(2), 107122.
  • Rosenbaum, M. E. (1986). The repulsion hypothesis: On the nondevelopment of relationships. Journal of Personality and Social Psychology, 51, 11561166.
  • Smith, A. J. (1957). Similarity of values and its relation to acceptance and the projection of similarity. Journal of Psychology, 43, 251260.
  • Sunnafrank, M. J. (1991). Interpersonal attraction and attitude similarity: A communication-based assessment. In Communication yearbook 14. Sage.
  • Sunnafrank, M. J., & Miller, G. R. (1981). The role of initial conversations in determining attraction to similar and dissimilar strangers. Human Communication Research, 8, 1625.
  • Van Bel, D. T., Smolders, K. C. H. J., IJsselsteijn, W. A., & De Kort, Y. A. W. (2009). Social connectedness: concept and measurement. In V. Callaghan, A. Kameas, A. Reyes, D. Royo, & M. Weber (Eds.), Proceedings of the 5th international conference on intelligent environments. (p. 6774). Amsterdam: IOS Press.
  • Yeong Tan, D. T., & Singh, R. (1995). Attitudes and attraction: A developmental study of the similarity-attraction and dissimilarity-repulsion hypotheses. Personality and Social Psychology Bulletin, 21(9), 975986.
  • Zerubavel, E. (1982, July). The standardization of time: A sociohistorical perspective. The American Journal of Sociology, 88(1), 123.

Biographies

  1. Top of page
  2. Abstract
  3. Study 1: The Effects of When-Similarity
  4. Study 2: When-Similarity and Who-Similarity
  5. General Discussion
  6. References
  7. Biographies
  • Maurits Kaptein is an assistant professor at the Department of Methodology and Statistics at the Tilburg school of Social and Behavioral Sciences. Maurits is also Chief Scientist at PersuasionAPI.

    Address: Department of Methodology and Statistics, Room P1.106. Tilburg University. PO Box 90153, 5000LE Tilburg, the Netherlands.

  • Deonne Castaneda is a User Experience Designer in California. She received an M.S. in Computer Science and a B.A. in Sociology from Stanford University.

    Address: Department of Computer Science, 353 Serra Mall, Stanford, CA 94305

  • Nicole Fernandez is a User Experience Researcher at Google in New York. She received her Masters in Human-Computer Interaction from Carnegie Mellon University and a B.S. in Symbolic Systems from Stanford University.

    Address: Google Inc, 76 Ninth Avenue, 4th Floor, New York, NY 10011

  • Clifford Nass is the Thomas M. Storke Professor at Stanford University; he has been a professor at Stanford since 1986. His primary appointment is in Communication, but he also has appointments by courtesy in Computer Science, Education, Law, and Sociology, and is affiliated with the programs in Science, Technology, and Society and Symbolic Systems (cognitive science).

    Address: Department of Communication Room 300E, McClatchy Hall Stanford University Stanford, CA. 94305–2050, USA.