This is an open access article distributed under the terms of the Creative Commons
Attribution License (Early relationships affect later relationships. According to attachment theory (Ainsworth, Blehar, Waters, & Wall, 1978), responsive, caring, and warm parents produce “securely attached” infants who feel secure enough to explore their environment. Whereas, parents who are inconsistent in their responses produce “anxious-ambivalent” children who are anxiously preoccupied with parental attention, and less inclined to explore. Parents who are rejecting and lacking in tender, loving care produce “avoidant” children who gradually learn to reject parental contact in favor of “the neutral world of things” (Ainsworth et al., 1978, p. 310).
These attachment styles have long-lasting effects on children’s friendships and romantic relationships (Greenberger & McLaughlin, 1998; Raby et al., 2015). Indeed, children’s early attachment styles tend to mirror their later attachment styles as adults (Hazan & Shaver, 1987). Insecurely attached children are apt to form less secure, more troubled, emotionally distant, and abusive relationships as adults (Dutton, Saunders, Starzomski, & Bartholomew, 1994; Feeney & Noller, 1990; Levitt, Silver, & Franco, 1996; see Mikulincer & Shaver, 2007).
In the current research, we argue that insecurely attached people may be especially likely to form parasocial relationships as adults (i.e., where person A is attracted to person B, but person B is unaware of the existence of person A; Horton & Wohl, 1956; Rubin, Perse, & Powell, 1985). Parasocial relationships – particularly those between celebrities and their fans – might be especially appealing to an insecurely attached person because such relationships make few demands and have little risk of rejection. Indeed, people with anxious-ambivalent-attachments, for example, tend to form positive attitudes of others but negative attitudes of themselves (Bowlby, 1969) and as a result, fear rejection from others. Because celebrities are nonthreatening, they may become idealized surrogate attachment figures (Greenwood & Long, 2011). The fan does not need to possess good social skills, and does not run the risk of being criticized or rejected (Ashe & McCutcheon, 2001) unless the fan tries to contact the celebrity (McCutcheon, Aruguete, McCarley, & Jenkins, 2016; McCutcheon, Scott, Aruguete, & Parker, 2006).
Attachment Styles and Celebrity Attitudes
Cohen (1997) studied the relationship between attachment styles and parasocial interaction in single people, some of whom were dating. Cohen used attachment categories (close, depend, anxiety) – slightly different from secure, anxious-ambivalent, and avoidant – and found relations only for those people who were currently dating. He found that anxious men were more likely to engage in parasocial interaction than anxious women, but securely attached women were more likely to engage in parasocial interaction than securely attached men.
In another study, university students completed a Parasocial Interaction Scale (PIS) with a favorite television personality as the target person, along with two measures of attachment style (Cole & Leets, 1999). Hazan and Shaver’s (1987) forced-choice descriptions of the avoidant, anxious-ambivalent, and secure attachment styles were used to self-classify participants into one or the other of the three categories.i The result was that anxious-ambivalents engaged in more parasocial interaction than avoidants, with secures falling in between. Further analysis revealed that among those classified as secure, attachment to a favorite TV celebrity was significantly related to the avoidant subscale score, such that secure people who tended to mistrust/avoid others also tended to be more strongly attracted to their favorite TV celebrity as measured by the Parasocial Interaction Scale (Cole & Leets, 1999).
McCutcheon et al. (2006) explored the link between attachment and choice of a favorite celebrity, using the Celebrity Attitude Scale (CAS), a measure that differs from the Parasocial Interaction Scale (PIS) in at least two ways. The PIS targets a favorite television celebrity, whereas the CAS permits the choice of a favorite celebrity from 14 categories, including television, music, athletics, and motion pictures. Secondly, the CAS has three subscales, Entertainment-Social (ES), Intense-Personal (IP), and Borderline-Pathological (BP). The first subscale is relatively benign; high scorers are interested in a favorite celebrity primarily because that celebrity’s performances are entertaining and they provide something to discuss with like-minded friends.
However, there is a dark side to celebrity admiration that is largely untapped by the PIS. According to the Absorption-Addiction Model (McCutcheon, Maltby, Houran, & Ashe, 2004) a few ES people move to a second level (IP) by becoming intensely and neurotically involved (Maltby, Houran, & McCutcheon, 2003) with the personal lives of their favorite celebrity. Further, a few IP individuals eventually become so obsessed with details of their celebrity’s life that support for their celebrity borders on the pathological (BP). These level three worshippers are more likely to endorse irresponsible attitudes and behaviors than those who are on levels one or two (McCutcheon, Wong, Black, Maynard, Frey, & Rich, 2014), and to admit that they would break laws on behalf of their favorite celebrity (McCutcheon, Lange, & Houran, 2002; McCutcheon et al., 2004).
McCutcheon and colleagues found that all three CAS subscales were negatively related to two measures of attachment, resulting in six correlation coefficients. However, these correlations ranged between -.08 and -.11. None were statistically significant.
Greenwood and Long (2011) gave measures of attachment and imagined intimacy with celebrities to a large sample of undergraduates. Attachment anxiety predicted imagined intimacy with favorite celebrities of the same gender. Attachment anxiety predicted imagined intimacy with favorite celebrities of the opposite gender only for those undergraduates who were not in a romantic relationship. In other words, an opposite gender celebrity serves as a surrogate attachment figure for those anxiously attached people who currently are without a romantic partner.
More recently, Coddaire (2015) administered measures of parasocial interaction and attachment to a large sample of adults ranging in age from 19 to 73. Controlling for loneliness she found a significant correlation (.24) between anxious attachment scores and attraction to one’s favorite celebrity.
Current Research
We firmly believe that a clearer understanding of the relation between attachment style and celebrity attraction is a timely, much needed, and relevant extension of the previous studies’ mixed results. Therefore, we propose to replicate some of the features of the aforementioned studies for three reasons. First, there has been considerable passage of time since some of these studies were published. Secondly, we would like to determine if the results of the study by Cole and Leets (1999) will generalize to a different measure of celebrity admiration, namely the Entertainment-Social subscale from the Celebrity Attitude Scale (CAS). Third, we are also interested in how the two problematic subscales of the CAS are related to attachment style.
An inspection of the 15 Parasocial Interaction Scale items used by Cole and Leets suggests that most of them fit roughly into the ES framework (“I like hearing the voice of my favorite TV personality in my home,” “My favorite TV personality makes me feel comfortable, as if I am with friends,” “I look forward to watching my favorite TV personality’s show”). Furthermore, the parasocial attraction scales used by Greenwood and Long (2011) and Coddaire (2015) also consisted of items that are similar to the items found on the PIS. What these parasocial interaction measures lack are subscales that measure the Intense-Personal and Borderline-Pathological attitudes and behaviors that characterize people who are absorbed and addicted to their favorite celebrities.
We are intrigued by the finding that a subset of securely attached people are especially likely to be attracted to celebrities when they have high levels of avoidance of others (Cole & Leets, 1999). As reflected in the items used by Cole and Leets, avoidance is similar to lacking trust. One of their three avoidance items reads “I find it easy to trust others” (reverse scored). We intend to use the Trusting Others subscale of the Predisposition to Trust Scale (PTS; Ashleigh, Higgs, & Dulewicz, 2012) to see if an additional measure of trust, one with nine items, will result in a stronger relationship between lack of trust and celebrity admiration for securely attached people.
Hypotheses. We predict that anxious-ambivalent participants, as defined by Self-classified Attachment Style choice will score higher on both measures of relatively benign celebrity attraction (CAS-ES & PIS) than avoidant participants, with securely attached participants in between (McCutcheon et al., 2006).
We expect that securely attached participants (as defined by Self-classified Attachment Style choice) who mistrust others (i.e. have high avoidant scores on the avoidant attachment subscale & low scores indicating mistrust on the Trusting Others subscale) will tend to have high levels of benign celebrity attraction (CAS-ES & PIS).
Method
Participants
The participants were 50 men (mean age = 35.34 years, SD = 12.35) and 61 women (mean age = 40.21 years, SD = 12.08). The participants in this sample were 69.4% white, 11.7% Black, 4.5% Hispanic, 9.0% Asian, 1.8% American Indian, & 3.6% other. All were recruited via Amazon’s Mechanical Turk (MTurk) – an online crowdsourcing platform. MTurk is commonly used by researchers in the social sciences to recruit diverse and high-quality samples of participants (see Bates & Lanza, 2013; Buhrmester, Kwang, & Gosling, 2011). Participants received $0.75 USD as remuneration.
Eleven participants were removed who spent less than 200 seconds completing all the measures. We decided to use that time as a cutoff because it would be nearly impossible to meaningfully respond to each item in each measure in such a brief period of time. When we did so, the alphas for each measure improved slightly as compared to the alphas we obtained when we included all 122 participants.
Measures
Celebrity Attitude Scale
Celebrity worship was measured using the Celebrity Attitude Scale (CAS). This scale consists of 23 items on a five-point Likert-type scale, with responses ranging from strongly disagree (1) to strongly agree (5). High scores indicate a greater interest in one’s favorite celebrity. The CAS has been found to have adequate reliability and validity in numerous studies (for examples see Griffith, Aruguete, Edman, Green, & McCutcheon, 2013; McCutcheon et al., 2004). The CAS consists of three subscales. Entertainment-social (ES) is captured in agreement with items like “My friends and I like to discuss what my favorite celebrity has done,” A second level of celebrity worship is characterized by more Intense-personal (IP) feelings, defined by items like “I have frequent thoughts about my celebrity, even when I don’t want to.” The third level, labeled Borderline-pathological (BP), is shown in items like: “If I were lucky enough to meet my favorite celebrity, and he/she asked me to do something illegal as a favor I would probably do it.” Cronbach’s alpha in the present study for the total CAS was .95. Alpha for each subscale was CAS ES = .89, CAS IP = .93, CAS BP = .72).
Trust Scales
Trust was measured by the Trusting Others subscale of the Predisposition to Trust Scale (TO; Ashleigh, Higgs, & Dulewicz, 2012). Responses are made on a Likert-type scale with “strongly disagree” equal to 1 and “strongly agree” equal to 7. The authors reported a Cronbach alpha over .70, and significant correlations with scales measuring happiness, life satisfaction, meaning in life, and total well-being. All nine items (“Other people lie to get ahead,” “Other people let you down”) are reverse scored, thus low scores indicate mistrust. Cronbach’s alpha in the present study was .94.
Attachment Scales
The Attachment Style Scale (ASS; Cole & Leets, 1999) contains six items (“I often worry that my partner doesn’t really love me,” & “I find that others are reluctant to get as close as I would like”) designed to measure the strength of one’s anxious-ambivalent attachment. High scores on this subscale suggest an anxious-ambivalent attachment. The secure subscale has five items (“I find it relatively easy to get close to others,” “I don’t often worry about someone getting too close to me”). High scores indicate a secure attachment. Three items (“I find it difficult to depend on others,” “I find it easy to trust others,” and “I feel comfortable depending on other people” the latter two reverse scored) attempt to measure the strength of the avoidant attachment subscale. High scores indicate mistrust. The authors found reliabilities of .93, .96, and .86, respectively. Cronbach’s alphas in the present study were .74, .78, and .78, respectively.ii The Self-classified Attachment Style developed by Hazan and Shaver (1987) is a forced-choice description of the avoidant, anxious-ambivalent, and secure attachment styles. Participants are requested to circle the one attachment style that best describes them. In the Cole and Leets study 57.4% self-described as secure, 24.3% self-described as avoidant, and 18.3% self-described as anxious-ambivalent. They also reported a close match between the Self-classified Attachment Style choice and scores on the Attachment Style Scale. In the present study 49.5% self-described as secure, 47.7% self-described as avoidant, and 2.7% self-described as anxious-ambivalent.iii
Parasocial Scale
The 15-item Parasocial Interaction Scale (PIS) was used to determine the strength of the parasocial relationship with one’s favorite celebrity. The items on this scale (“When I’m watching the program my favorite TV personality is on, I feel as if I am part of the group,” “I like to compare my ideas with what my favorite TV personality says,” “My favorite TV personality makes me feel comfortable, as if I am with friends”) were measured on a 5-point Likert scale with “strongly disagree” equal to 1 and “strongly agree” equal to 5. High scores indicate a strong attraction to a favorite TV celebrity. Cole and Leets obtained a mean score of 49.05 and an alpha of .87. Cronbach’s alpha in the present study was .91.
Procedure
After obtaining IRB approval, we created an online survey with several different orders of presentation of the various measures to reduce the probability of a systematic order effect. Then, participants from MTurk were provided with a URL link to the survey titled, “Self and Celebrity Attitudes Survey.” Participants read that the purpose of the study was for them to answer several questions about themselves and how they felt towards celebrities. After providing informed consent, participants completed the Celebrity Attitude Scale, Trusting Others subscale from the Predisposition to Trust Scale, the Attachment Style Scale, the Self-classified Attachment Style, and the Parasocial Interaction Scale in counterbalanced order and basic demographic questions (e.g., age, gender, ethnicity). Finally, participants were debriefed and compensated for their participation.
Results
All variables used in the statistical analyses had approximately normal distributions and there were no significant outliers found.iv Scatterplots indicated linear relationships among the scales. Descriptive statistics for all the scales are presented in Table 1. Pearson product moment correlations (r) were performed to examine relationships among these nine scale variables. See Table 2 for these correlations.
Table 1
Means, Standard Deviations, and Ranges for all Measures Used in the Study
| Measure | M | SD | Min-Max |
|---|---|---|---|
| 1. CAS – ES | 29.56 | 8.11 | 10-45 |
| 2. CAS – IP | 19.11 | 7.91 | 9-43 |
| 3. CAS – BP | 9.57 | 3.49 | 4-18 |
| 4. Trusting Others subscale | 34.32 | 10.81 | 9-63 |
| 5. ASS – Anxious-ambivalent | 11.97 | 3.38 | 6-24 |
| 6. ASS – Secure | 13.55 | 3.18 | 6-20 |
| 7. ASS – Avoidant | 7.52 | 2.13 | 3-12 |
| 8. Parasocial Interaction Scale (PIS) | 49.81 | 10.55 | 18-70 |
Table 2
Correlations Among the Measures Used in the Study (n = 111)
| Measure | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| 1. CAS – ES | — | .70** | .73** | -.15 | .34** | .03 | -.22* | .70** |
| 2. CAS – IP | — | .77** | -.19* | .47** | -.05 | -.18 | .56** | |
| 3. CAS – BP | — | -.20* | .29** | .09 | -.23* | .54** | ||
| 4. Trusting Others subscale | — | -.30** | .31** | -.59** | -.15 | |||
| 5. ASS – Anxious-ambivalent | — | -.40** | .12 | .28** | ||||
| 6. ASS – Secure | — | -.48** | -.05 | |||||
| 7. ASS – Avoidant | — | -.02 | ||||||
| 8. Parasocial Interaction Scale (PIS) | — |
*p < .05. **p < .01.
Our first hypothesis was the prediction that anxious-ambivalent participants, as defined by Self-classified Attachment Style choice, would score higher on both measures of relatively benign celebrity attraction (CAS-ES & PIS) than avoidant participants, with securely attached participants in between.
There were 55 participants who self-identified as secure, 53 as avoidant and 3 as anxious-ambivalent. Due to the small number of anxious-ambivalent, we decided to run analyses only on the first two groups since they were similar in size. Independent t-tests were performed on CAS ES and PIS scores. Since we hypothesized a specific direction, one-tailed tests using an alpha of .05 were used. There was no significant difference found between the secure group (M = 48.61, SD = 9.75) and the avoidant group (M = 50.47, SD = 11.06) on the CAS ES, t(106) = -0.52, p = .604, r2 = .01. There was also no significant difference between the secure group (M = 29.82, SD = 6.83) and the avoidant group (M = 29.00, SD = 9.34) on PIS scores, t(106) = -0.92, p = .357, r2 = .01. See Table 3 for these analyses.
Table 3
Means and SDs for the Two Largest Self-Classified Groups on the Scales and Subscales
| Measure | Secure (n = 55)
|
Avoidant (n = 53)
|
t | r2 | ||
|---|---|---|---|---|---|---|
| M | SD | M | SD | |||
| 1. CAS – ES | 29.82 | 6.83 | 29.00 | 9.34 | 0.52 | .002 |
| 2. CAS – IP | 18.91 | 8.23 | 18.92 | 7.67 | -0.01 | .001 |
| 3. CAS – BP | 9.70 | 3.10 | 9.24 | 3.83 | 0.67 | .004 |
| 4. Trusting Others | 39.20 | 10.12 | 29.89 | 9.03 | 5.04** | .193 |
| 5. ASS – Anxious-ambivalent | 10.96 | 3.10 | 12.68 | 3.07 | -2.89** | .073 |
| 6. ASS – Secure | 15.47 | 2.62 | 11.43 | 2.22 | 8.62** | .412 |
| 7. ASS – Avoidant | 6.30 | 1.48 | 8.64 | 2.16 | -6.36** | .276 |
| 8. Parasocial Interaction Scale (PIS) | 48.62 | 9.75 | 50.47 | 11.06 | -0.92 | .008 |
**p < .01.
Our second hypothesis was the expectation that securely attached participants (as defined by Self-classified Attachment Style choice) who mistrust others (i.e. have low scores on the Trusting Others subscale) will tend to have higher levels of benign celebrity attraction (CAS-ES & PIS) compared to those secures who trust others (high scores on Trusting others subscale). Because the avoidant subscale from the ASS also measures trust (“I find it easy to trust others,” reversed), we also expected that securely attached participants who mistrust others (high scores on the ASS-avoidant subscale) will tend to have higher levels of benign celebrity attraction.
Only those participants who selected the securely attached option were used for these analyses (n = 55). Univariate regressions were performed on PIS and CAS_ES scores using the continuous Trust variable as the predictor. Trust was found to not be a significant predictor of PIS scores (β = -210), t(53) = 1.56, p = .125, r2 = .04. However, for the CAS ES scores, the Trusting variable did significantly predict CAS ES scores (β = -354), t(53) = -2.75, p = .008, r2 = .12.
We repeated these analyses on PIS and CAS_ES scales using the ASS-avoidant (trust) subscale scores as the predictor. The ASS-avoidant scores did not significantly predict PIS scores (β = .040), t(53) = .29, p = .773, r2 = .01. Further, ASS-avoidant scores did not significantly predict CAS_ES scores (β = -097), t(53) = -0.71, p = .481, r2 = .01.
Discussion
Our first hypothesis was that securely attached people would score higher than avoidant attached people on both measures of benign celebrity admiration. Surprisingly, neither hypothesis was supported. Interestingly, the original study by Cole & Leets showed that although securely attached people scored higher on the Parasocial Interaction Scale, they did not score significantly higher. The significant difference in their study was between people with anxious-ambivalent and avoidant attachment styles; the securely attached people’s scores were in between. Therefore, the current null findings tend to confirm those of McCutcheon et al. (2006) who found that neither of the two attachment measures – where the insecure attachment style categories were combined – was significantly correlated with CAS ES scores.
Our second hypothesis was that securely attached participants who mistrust others - as indicated by low scores on the Trusting Others subscale and high scores on the ASS-avoidant subscale - will tend to have higher levels of benign celebrity attraction (CAS-ES & PIS) compared to those secures who trust others. This hypothesis was confirmed only for those with lower high scores on the Trusting Others subscale and only on the CAS ES. Our finding tends to cast some doubt on the robustness of the results obtained by Cole and Leets (1999).
On the other hand, our correlational data suggest a modest link between trust, as measured by the Trusting Others subscale, and scores on two of the three subscales of the CAS. These significant correlations, shown in Table 2, range between -.19 and -.20. We interpret this finding to mean that there exists a weak trend for the mistrusting of others to be associated with greater admiration for one’s favorite celebrity. However, that trend is reversed (-.22 & -.23) when the ASS-avoidant subscale (low scores = high trust) is correlated with the three CAS subscale scores, and no significant correlations were found between PIS scores and either of the two trust measures. Additionally, it is also possible that, as a reviewer brought to our attention, low trust scores may be not necessarily reflect mistrust.v In short, additional research is clearly needed to clarify the links between (mis)trust, attachment styles, and celebrity attitudes.
In an exploratory fashion, our findings provide additional support for the validity of both measures of celebrity attraction. Earlier we argued that the PIS is most similar to the ES subscale of the CAS, inasmuch as both seem to be measuring a relatively benign form of celebrity admiration. This argument was supported by the finding that PIS was more strongly correlated with CAS ES (.69) than with the less benign forms of celebrity admiration (.56 for CAS IP; .54 for CAS BP).
Strengths, Limitations, and Future Directions
Our study has several strengths, as well as limitations, which warrant discussion. First, the current study sought to clarify the (somewhat) mixed findings from previous research regarding the relation between various attachment styles and celebrity attitudes. Unfortunately, our sample precluded a deeper exploration among people with an anxious-ambivalent attachment style. Furthermore, Hispanics were underrepresented in our sample. Although we did not hypothesize any difference as a function of ethnicity, previous research shows that Hispanics tend to score in between Black and White people on the CAS (McCutcheon, Aruguete, Jenkins, McCarley, & Yockey, 2016). A larger, more representative sample would have allowed us to further explore such differences in an exploratory fashion.
Second, our findings raised important questions about the role of (mis)trust as it relates to the strength of parasocial bonds with a favorite celebrity. Our research cast doubt on the robustness of Cole and Leet’s finding that securely attached people who have low levels of trust tended to have higher levels of benign celebrity attraction, as compared to securely attached people with high levels of trust. Additional research is certainly needed to determine why one of our measures of mistrust tended to support the findings of Cole and Leets but the other measure of mistrust did not. Additional psychometric data on our measures and a large-scale, representative study may better explain why our findings occasionally (dis)confirm the findings of Cole and Leet.
Finally, our study has larger implications for why people may (or may not) seek out parasocial relationships with celebrities. Although we do not want to interpret non-significant findings, it appears that interpersonal relationships – particularly feelings of trust and attachment style – may not have as robust a relationship with people’s “fandoms” (i.e., extreme celebrity attitudes and behaviors) as previously believed. Perhaps longitudinal studies which assess people’s attachment styles, peer-relationships, and general changes in need to belong throughout life may shed more light on some people’s varying attitudes towards celebrities later in life.