Contributed equally to this work with: Simon Gächter, Chris Starmer, Fabio Tufano * E-mail: simon.gaechter@nottingham.ac.uk Affiliations Centre for Decision Research and Experimental Economics, School of Economics, University of Nottingham, University Park, Nottingham, United Kingdom, CESifo Network, Munich, Germany, IZA Network, Bonn, Germany ⨯
Contributed equally to this work with: Simon Gächter, Chris Starmer, Fabio Tufano Affiliation Centre for Decision Research and Experimental Economics, School of Economics, University of Nottingham, University Park, Nottingham, United Kingdom ⨯
Contributed equally to this work with: Simon Gächter, Chris Starmer, Fabio Tufano Affiliation Centre for Decision Research and Experimental Economics, School of Economics, University of Nottingham, University Park, Nottingham, United Kingdom ⨯
Understanding the nature and influence of social relationships is of increasing interest to behavioral economists, and behavioral scientists more generally. In turn, this creates a need for tractable, and reliable, tools for measuring fundamental aspects of social relationships. We provide a comprehensive evaluation of the 'Inclusion of the Other in the Self' (IOS) Scale, a handy pictorial tool for measuring the subjectively perceived closeness of a relationship. The tool is highly portable, very easy for subjects to understand and takes less than 1 minute to administer. Across our three online studies with a diverse adult population (n = 772) we show that six different scales designed to measure relationship closeness are all highly significantly positively correlated with the IOS Scale. We then conduct a Principal Component Analysis to construct an Index of Relationship Closeness and find that it correlates very strongly (ρ = 85) with the IOS Scale. We conclude that the IOS Scale is a psychologically meaningful and highly reliable measure of the subjective closeness of relationships.
Citation: Gächter S, Starmer C, Tufano F (2015) Measuring the Closeness of Relationships: A Comprehensive Evaluation of the 'Inclusion of the Other in the Self' Scale. PLoS ONE 10(6): e0129478. https://doi.org/10.1371/journal.pone.0129478
Academic Editor: MariaPaz Espinosa, University of the Basque Country, SPAIN
Received: January 8, 2015; Accepted: May 9, 2015; Published: June 12, 2015
Copyright: © 2015 Gächter et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: Funding was received from the European Research Council (http://erc.europa.eu/) Advanced Investigator Grant COOPERATION 295707 and the UK Economic and Social Research Coincil (http://www.esrc.ac.uk/) grant ES/K002201/1 (Network for Integrated Behavioural Science). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
An important fact of social life is that people have social relationships that vary in closeness. Most people have some close relationship to a romantic partner, a few friends and family, somewhat looser relationships with other friends and even looser ones with numerous acquaintances. If such relationships are a fact of social life, how then do relationships matter for social behavior and social preferences? We suggest these important issues provoke a prior question of how the perceived closeness of a relationship, which is a subjective judgment, can be measured. In this paper, we hope to make some progress in answering this prior question by evaluating in three studies a simple pictorial measurement instrument, called the 'Inclusion of the Other in the Self' Scale, by Aron et al.[1]. The question, of how social relationships matter for social behavior, is the ultimate motivation underlying our interest, as behavioral scientists, in the measurement of relationships. Here is why.
For a long time many economists and other social scientists used the homo economicus assumption to explain social behavior. Homo economicus conceives of social relations purely from the point of view of an individual’s self interest, and from this perspective relationships only matter indirectly, to advance own well-being. Behavioral research in the last three decades has changed one aspect of this picture profoundly by demonstrating that many people have strong other-regarding preferences [2–5]. However, much of this research abstracts from the social relationships people actually have and measures social preferences towards unidentified, anonymous other people. Thus, relationship closeness is disregarded even if other-regarding motives matter strongly. Similarly, behavioral investigations of strategic thinking largely disregard the psychological nature of the relationships among individuals and, in experiments, usually only study anonymous agents and their interactions.
Thus, in our view, a significant open research question is how social preferences or strategic thinking change with the nature of real, non-anonymous relationships. Some evidence suggests that the degree to which individuals can identify each other matters for social preferences (e.g., [6], [7]), but this research does not directly investigate the role of relationships. Some contemporary research measures a concept akin to relationships, namely the network structures in which people are embedded (people are asked to list the names of their friends or people with whom they interact) and then this network structure is related to a variable of interest, e.g., altruistic sharing [8, 9] or diffusion of information [10]. These are excellent tools to advance our understanding of the importance of social relationships. However, network structures highlight who is linked to whom and who is "central" and do not consider the psychology of relationships, that is, how "close" people feel to be to a specific other person.
In this paper we are motivated by a complementary strategy for advancing our understanding of relationships, which involves borrowing tools from social psychology to measure the closeness of bi-lateral personal and social relationships between individuals. The tool is the "Inclusion of the Other in the Self" (IOS) Scale, developed by Aron, Aron and Smollan in a highly cited 1992 paper in the Journal of Personality and Social Psychology ([1], henceforth AAS). It is a simple pictorial tool, which is very easy to implement making it a potentially highly useful instrument. We illustrate our version of the IOS task in Fig 1 (exactly as it was seen by the participants of our three studies); further details are given in the next section. The IOS task asks respondents (“You” in our version) to assess their relationship with a specific individual (referred to as "X" in our figure) by selecting one out of seven pairs of increasingly overlapping circles. In each pair of circles, one circle refers to the respondent and the other circle to X. Respondents are asked to select the pair of circles that best describes their relationship with X. For example, if a respondent feels unrelated to X, it would be natural to select the first pair of still disjoint circles; if a respondent feels very close to X, he or she may choose the almost completely overlapping set of circles.
PowerPoint slide larger image original image Fig 1. The 'Inclusion of the Other in the Self' (IOS) task.Adapted and changed for our online implementation from AAS ([1], Fig 1, p. 597). Respondents are asked to select the pair of circles that best describes their relationship with X. AAS speak of ‘Self’ and ‘Other’, whereas we use the terminology ‘You’ and ‘X’.
The IOS Scale is extremely simple to use and, we believe, highly intuitive. It takes less than a minute to administer. It is also neutral about the reason why people feel close or distant; it just measures an individual’s own subjective perception of their degree of closeness to another person. AAS ([1], p. 598) argue that the "IOS Scale is hypothesized to tap people's sense of being interconnected with another. That sense may arise from all sorts of processes, conscious or unconscious […] The IOS Scale is intended to capture something in the respondent's perception of a relationship that is consistent with many theoretical orientations." Our evaluation of the IOS Scale is inspired by Aron et al.'s intention that the IOS Scale is "consistent with many theoretical orientations" and therefore involves linking several conceptually different scales that all intend to measure relationship closeness.
Our evaluation of the IOS Scale is motivated by two developments since AAS. First, AAS, in their original evaluation of the IOS Scale, were mostly interested in measuring close relationships but for behavioral scientists non-close relationships such as friends and acquaintances are often also of substantial interest. In response to such demand, Starzyk, Holden, Fabrigar and MacDonald ([11], henceforth, SHFM) developed a new questionnaire, the 'Personal Acquaintance Measure' (PAM) that allows measuring all types of relationships, close and non-close. Hence, we will also evaluate how the IOS Scale is related to the PAM Scale and compare our results to those of SHFM. Second, AAS as well as SHFM relied on student subjects and paper-based questionnaires administered in the classroom. Many potential applications of this tool, however, will nowadays use non-student subject pools and/or internet-based data collection methods and these facts provide a central motivation for the new evaluation of the IOS Scale we report here: Do the psychometric properties of the IOS Scale extend to non-students and the use of an internet-based survey tool?
We report three studies; all conducted with the help of Amazon.com's labor market Mechanical Turk (MTurk). Participants on MTurk are typically older and come from much more varied backgrounds than undergraduates. Study 1 replicates AAS and the psychometric evaluation reported in that paper. One important claim from AAS is that the IOS Scale captures the essence of a much more complex and comprehensive scale, the influential Relationship Closeness Inventory (RCI) developed by Berscheid, Snyder and Omoto ([12], henceforth BSO). As part of the evaluation of the IOS Scale, Study 1 will therefore also implement the RCI. Our results confirm all psychometric properties of RCI found in BSO and AAS for the new and more diverse population from which we sample. Most importantly, we confirm AAS's conclusions about the validity of the IOS Scale.
The RCI is predominantly about "close" relationships, such as close friendships and in particular romantic relationships. However, many relationships that are of special interest to behavioral social scientists are not romantic and not particularly close, but still quite important; think of workplace relationships as an example. Studies 2 and 3 will, therefore, also explore the IOS Scale for non-close relationships. In particular, in addition to close relationships, Study 2 will elicit measurements on the IOS scale for two more categories of relationships: friends who are more than an acquaintance, but not a person's most intimate relationship; and acquaintances who are closer than strangers, but less close than friends. We find that the IOS measurements vary strongly and in the expected directions across the three classes of relationship.
Study 3 elicits the IOS measurements for close relationships, friends and acquaintances and compare them to results based on a set of conceptually related questionnaires, such as, again, the RCI, the Liking and Loving Scale by Rubin [13], and the PAM Scale. Details of the conceptual background and implementation of the RCI and PAM will be given in the introduction and methods sections of the respective studies. Our findings confirm the results from Studies 1 and 2 as well as the most important findings of SHFM with regard to the PAM Scale.
One central observation of Study 3 is that the IOS Scale is highly significantly and strongly correlated with several scales that all measure dimensions of relationship closeness: among them, the RCI, the Loving and Liking Scales, and the PAM Scale. In our final section we use the fact that several different measures of relationship closeness correlate with IOS to perform a Principal Component Analysis to derive an "Index of Relationship Closeness" (IRC). We show that IRC correlates very highly (ρ = .85) with IOS. We conclude, therefore, that the IOS Scale is measuring important aspects of relationship closeness in a compact and highly reliable way, which coupled with the simplicity and portability of the tool, renders it a very useful instrument for any researcher interested in measuring the closeness of relationships.
The Research Ethics Committee of the School of Economics of the University of Nottingham approved this research. The research was conducted using Amazon Mechanical Turk (http://www.mturk.com). Mturk is a crowd sourcing internet-based market place that facilitates hiring of "workers" to do short online jobs (called a HIT—"Human Intelligence Tasks") [14–16]. MTurk is popular for conducting research because it has a vast subject pool, which is more varied than the typical undergraduate subject pool. Although it is not fully representative of the (American) population at large, it is more representative than purely undergraduate subject pools and provides a very cost effective route to large data sets [17–20]. And, most importantly, it provides quality data, according to several studies that compared MTurk samples to more traditional lab samples where experimental control is typically stronger than on MTurk [16, 19–21].
Informed consent was obtained online from all participants (before they finally decided to take part participants were told the following: "By accepting this HIT you give us informed consent that we can use your answers in anonymized form for research purposes only.").
After they had accepted our HIT, participants read the following introductory text: "In this HIT we will ask you to respond to a questionnaire on the nature of interpersonal relationships. Our interest is entirely scientific. All answers will be treated confidentially and will only be reported in aggregated statistical form. There are no right or wrong answers in this survey; we are only interested in your honest assessment. If you feel uncomfortable answering some questions you will have opportunities to select 'prefer not to answer' as an answer."
A total of 772 people, all recruited on MTurk and residents of the United States, participated in our three studies. The studies were conducted between February and July 2014. The average age of our participants was 34 years (s.d. 11) and about 46% were female. In contrast, the participants in our reference studies were undergraduates with an average age of about 19 years. Across the three studies, the percentage of females varied between 52% and 78%. After deciding to participate, subjects were directed to the survey questions on an external webpage. The survey was coded using the survey software Qualtrics (http://www.qualtrics.com/). The survey questions are available as part of the Supporting Information to this paper (S1 Text). Participants received a flat payment for participation (details are given in the description of studies).
Participants were first introduced to the topic of the research and told that it is about understanding the nature of relationships (see above). They were then asked to focus on one specific individual (the exact rules varied across the studies we report here). Participants were asked to identify this individual by the initial of their first name. Participants were then told the following:
"In the following figure we ask you to consider which of these pairs of circles best describes your relationship with [This Individual] in all questions that follow. In the figure "X" serves as a placeholder for [This Individual], that is, you should think of "X" being [This Individual]. By selecting the appropriate number please indicate to what extent you and [This Individual] are connected."
On the screen the IOS task was presented exactly as in Fig 1. Participants had to select the answer with a simple click but they could also indicate that they preferred not to answer.
We also elicited two more measures proposed in the literature to measure relationship closeness. First, following Cialdini et al. [22], right after eliciting the IOS Scale rating we added a "We Scale". The exact wording was as follows: "Please, select the appropriate number below to indicate to what extent you would use the term “WE” to characterize you and [This Individual]. Answers were on a 7-point scale (1 = "not at all"; 7 = "very much so "). The average between the IOS Scale and the "We- Scale" forms a measure that Cialdini et al. [22] call "oneness".
Answers were on a 7-point scale, where 1 = "not close at all", and 7 = "very close". The SCI is simply the sum of scores.
The IOS Scale, the We Scale, and the SCI Scale are used in all three studies. Study 1 benchmarks them against the RCI; Study 2 looks for differences in levels of relationship closeness; and Study 3 investigates further benchmarks, in particular the PAM Scale. The raw data are available as Supporting Information file S1 Dataset (in Excel format).
We close our comprehensive evaluation of the IOS Scale with a novel Principal Component Analysis we use to derive an "Index of Relationship Closeness". This index will be our comprehensive benchmark for the IOS Scale.
The main purpose of Study 1 was to replicate the RCI and in addition to see whether the IOS Scale is correlated with RCI, the We Scale, and the SCI Scale and how the correlations and the scores of the scales compare between our study and previous studies [1],[12]. The two most important instruments of Study 1 are the RCI and the IOS Scale. The RCI provides the conceptual background to the IOS Scale and therefore we describe it in more detail now.
The conceptual foundations for RCI are due to Kelley et al. who argue that a close relationship is characterized by high “interdependence” [23], which manifests itself in interconnected activities, where people have frequent impact on each other; the degree of impact is strong; and impact is based on diverse activities people undertake together. Based on this conceptual framework, the RCI is a 69-item self-report to measure the frequency with which partners see each other; how many diverse activities partners undertake together, and the strength of influence a partner has on the respondent. Answers are then aggregated into three subscales (RCI Frequency, RCI Diversity, and RCI Strength) and a total RCI Scale. We describe the questionnaires used to measure the RCI and its subscales in the methods section.
The RCI takes about 10–15 minutes to complete and is often too detailed for many research purposes. Therefore, AAS developed the IOS Scale (as per Fig 1), to have a handy and compact instrument to measure relationship closeness. AAS argue that the IOS is not linked only to the RCI but is consistent with several approaches to relationship closeness in social psychology (a claim we will evaluate in Study 3). Based on previous work [24], AAS ([1], p. 598) argue that "in a close relationship the individual acts as if some or all aspects of the partner are partially the individual's own" and that "in close relationships the individual may perceive the self as including resources, perspectives, and characteristics of the other".
We recruited 200 volunteers on MTurk who received a flat payment of $3 for a survey that took most participants about 30 minutes. Our sample size is in the same ball park as our comparison studies: AAS, [1] had 208 participants (Primary Study), and BSO [12] had 241 participants (in their main RCI study).
The survey consisted of two parts. Part I included, in this order, the IOS and two related questions, and the RCI. The order of questionnaires was the same as in Part I of AAS. Part II consisted of a set of hypothetical decision problems, which we do not report here. When answering Part I questions, participants were not informed about the nature of questions in Part II; hence, Part II cannot influence the responses to IOS and RCI. Answering the questions of Part I took about 15 minutes.
We used the exact same texts and questions as BSO and only changed any wording related to the fact that ours was an online implementation of BSO's questionnaire. (Their exact wording can be found in BSO [12], Appendix A, pp. 806–807). All participants read the following introductory text (p. 806, and S1 Text):
"We are currently investigating the nature of interpersonal relationships. As part of this study, we would like you to answer the following questions about your relationship with another person. Specifically, we would like you to choose the one person with whom you have the closest, deepest, most involved, and most intimate relationship, and answer the following questions with regard to this particular person. For some of you, this person may be a dating partner or someone with whom you have a romantic relationship. For others of you, this person may be a close, personal friend, family member, or companion. It makes no difference exactly who this person is as long as she or he is the one person with whom you have the closest, deepest, most involved, and most intimate relationship. Please select this person carefully since this decision will affect the rest of this study. With this person in mind, please respond to the following questions."
Participants were then asked to provide the initial of the first name of the person they had in mind (here, we will refer to this person as "X"). The initial they provided was automatically inserted into all questions that referred to this specific individual. Participants were then presented with the IOS task, in exactly the same way as illustrated in Fig 1, followed by the "We Scale". Participants were also asked about the gender and age of X, as well as their own age and gender, how long they had known X, and whether X is a family member, a friend, someone from work, or a romantic partner. After these introductory questions, participants proceeded to the core questions that constitute the RCI.
The first block of questions concern the frequency of interactions in the past week, where participants are asked to assess how many hours they had spent with X alone in the morning, the afternoon, and in the evening. The stated times are translated into minutes and then given a score between 1 and 10, where the score increases in the frequency of interactions. This score constitutes the RCI Frequency score.
The second block of questions measures the diversity of activities that the participant and X undertook together in the past week. Participants were presented with a list of 38 activities and asked to check all that apply. The number of joint activities is translated into a score between 1 and 10, where the score increases in the diversity of activities. This score is the RCI Diversity score.
The third block of 34 questions measures the strength of influence that X has on the participants on a range of attitudes, time spent with friends or relatives, financial expenditures, leisure activities etc. A set of questions in this block also asks to what extent X influences the participant's future plans in a variety of domains from vacation plans to marriage plans. Participants had to answer on a 7-point scale (1 = strongly disagree; 7 = strongly agree). The sum of scores (which theoretically can be between 34*1 = 34 and 34*7 = 238) was then given a RCI Strength score, which ranges between 1 and 10 (higher score means higher strength).
The Total RCI score is then simply the sum of the RCI Frequency, RCI Diversity and RCI Strength scores. By construction, the Total RCI score takes values between 3 and 30. All specific details of the scoring rules can be found in BSO ([12], Appendix B, p. 807).
Our main purpose in Study 1 is to compare the psychometric properties of the IOS Scale and the RCI Scale as reported by AAS [1] and BSO [12]), respectively, with our data. The nature of the relationship with X differs predictably between our sample and the much younger student samples of BSO and AAS, respectively. In our sample, 68% report their closest relationship to be a romantic one, whereas in BSO and AAS this was the case for only 47% and 44%, respectively. In our sample, family and friends constitute 10.5% and 16.5%, respectively. For many student participants in BSO and AAS, their closest relationship was a friend (36% and 37%, resp.) or a family member (14% and 16%, resp.). The longevity of close relations was also substantially bigger among our participants (around 9 years for non-family members; 32 years for family members; compared to about 4.5 years for non-family and 19 years for family among students), which is unsurprising given the age difference between our sample and the students of previous samples.
Next we compare the internal consistency of the scales [25]. We have two comparison sets here, BSO and AAS, because in their validation of the IOS Scale, AAS also replicated BSO. Internal consistency (as measured by Cronbach's α) is remarkably similar across the three studies and all scales. For example, for RCI Total, our α = .65; for BSO it was α = .62; and for AAS it was α = .66. See S1 Table for further details.
Given our particular interest in the IOS Scale the most important comparison for our purposes is to look at the correlations between the IOS Scale, the RCI Scale, and the SCI Scale. The comparison data are from AAS ([1], Table 1, p. 600). The top panel of Table 1 reports the results for our sample and the bottom panel documents the respective results of AAS. Before we look at correlations we notice that the mean of the IOS Scale is 5.30 (sd 1.41) in our data and 4.74 (sd 1.48) in the AAS data. Thus, our participants report somewhat closer relationships as measured by the IOS Scale than the student participants of AAS. The mean scores of all scales are a bit higher in our sample compared to AAS.