Differentiation and Eating Disorder Symptoms Among Males

Differentiation and eating disorders

While researchers have started to examine the link between differentiation of self and eating disorder symptoms among male participants, additional investigation is warranted. In particular, previous studies have focused mainly on the interpersonal elements of differentiation and have focused less on the intrapersonal components. For example, Levy and Hadley (1998) examined two factors related to differentiation which have been understood as interpersonal areas of differentiation (Skowron et al., 2003). Levy and Hadley assessed fusion in family of origin relationships and personal authority in one’s family of origin and found that male participants with high levels of bulimic symptoms had increased fusion in their family relationships. Blackmer, Searight, and Ratwick (2011) also assessed interpersonal elements of differentiation and eating disorder symptoms. These authors measured autonomy and intimacy in one’s family of origin. Reporting results for the participants as a whole (i.e., not separated by gender), they found that participants with less autonomy in their families of origin had increased eating disorder symptoms. Thus, while preliminary evidence suggests a link between fusion in family relationships and eating disorder symptoms, less is known about the link between other elements of differentiation and eating disorder symptoms among males. Specifically, the association between the intrapersonal aspects of differentiation (Skowron et al., 2003) and eating disorder symptoms among males deserves increased investigation.

Schwartz, Grammas, Sutherland, Siffert, and Bush-King (2010) did use a comprehensive measure of differentiation when they examined the link between differentiation of self and body image attitudes of male college student participants. Participants with less ER reported reduced concern with attaining low body fat; participants with increased levels of IP reported less concern about both being muscular and having minimal body fat. The other differentiation elements (i.e., fusion and EC) were not significantly linked with male participant concerns for low body fat or muscularity. These authors did not, however, assess eating disorder symptomatology, and our study thus represents a next step in extending this literature.

Current Study

In the present investigation, we assessed four domains of differentiation of self which comprised both the interpersonal and intrapersonal components of differentiation: IP, ER, EC, and FO (Skowron et al., 2003; Skowron & Schmitt, 2003). In order to extend previous research around differentiation of self in relation to body image (Schwartz, Grammas, Sutherland, Siffert, & Bush-King, 2010) and bulimia (Levy & Hadley, 1998) among male participants, we asked several research questions.

Specifically, we asked: (a) what is the relationship between an IP and eating disorder symptoms? (b) What is the relationship between ER and eating disorder symptoms? (c) What is the relationship EC and eating disorder symptoms? (d) What is the relationship between FO and eating disorder symptoms? In addition, we investigated the relative influence of each differentiation of self-component, while accounting for the influence of all other differentiation factors. For this analysis, we asked (e) what is the relative contribution of ER, IP, EC, and FO in explaining eating disorder symptoms?

Participants

Participants were 349 college students who self-identified as male and reported an age range of 18–31 (M = 20.11; one participant declined to answer). Participants represented a range of undergraduate academic class statuses: 24.6% were first-year students; 26.6% were sophomores; 20.9% were juniors, and 27.5% were seniors (one chose not to answer). The majority of participants identified as Caucasian/White (80.2%) with the remainder identifying as Asian/Pacific Islander (9.5%), Hispanic/Latino (4.6%), African American (2.6%), multiethnic/multiracial (2.6%), American Indian/Alaskan Native (0.3%), and Other (0.3%). Participants primarily reported their sexual identity as heterosexual (94%), 4.6% identified as gay, and 1.4% identified as bisexual. Participants reported their economic background as lower class (2.5%), working class (11.5%), middle class (54.2%), upper middle class (29.8%), and upper class (2%). Only one participant identified as an international student (0.3%) and one participant chose not to answer this item. Participants reported body mass indexes (BMIs) that were underweight (below 18.5; 4.9%), normal (18.5–24.99; 60.5%), overweight (25.0–29.99; 25.2%), and obese (30.0 and up; 8.6%); participants reported a mean BMI of 24.14, which is a normal weight (World Health Organization, 2012). We could not calculate BMI for two participants, due to missing or implausible height and weight information.

Procedures

After obtaining institutional review board approval, we sent out an e-mail invitation to the campus e-mails of students at a public mideastern college; an administrative office at the college assisted us with this process. For this study, 2,761 students (n = 2,755 males; n = 6 no gender information on file at the college) were initially e-mailed as potential participants. This e-mail contained a survey link through which participants could access the informed consent and the online survey (via Qualtrics^©). Potential participants were sent an initial e-mail and two e-mail reminders (each 1 week apart) and also were invited to enter a random drawing to win one of the two US$100 VISA^© gift cards. Our total participants were 349 male students. This is a 12.6% response rate, which is comparable to response rates reported by other researchers who surveyed male college students using Internet data collection methods (Utpala-Kumar & Deane, 2010).

Instruments

For the purposes of the current study, participants completed demographic questions and two assessment instruments. In particular, participants completed the DSI-R; Skowron & Schmitt, 2003) and the Eating Attitudes Test-26 (EAT-26; Garner, Olmsted, Bohr, & Garfinkel, 1982). All instruments were self-report measures.

The DSI-R (Skowron & Schmitt, 2003) is a revision of the DSI (Skowron & Friedlander, 1998), wherein the authors altered the FO subscale for reliability and validity purposes. The DSI-R is a 46-item scale with four subscales: ER, IP, EC, and FO. Participants answer items on a 6-point Likert-type scale (1 = not at all true of me; 6 = very true of me) and the scale authors noted that higher scores indicated increased differentiation. ER (11-item subscale) measures one’s “tendency to respond to environmental stimuli on the basis of autonomic emotional responses, emotional flooding, or lability” (Skowron & Schmitt, 2003, p. 212). Higher scores indicate higher levels of differentiation of self and lower levels of ER. IP (11-item subscale) refers to “a clearly defined sense of self and the ability to thoughtfully adhere to one’s convictions even when pressured to do otherwise” (Skowron & Schmitt, 2003, p. 212). Higher scores on this subscale indicate increased differentiation of self and IP. EC (13-item subscale) assesses “fears of intimacy or engulfment in relationships, and the accompanying behavioral defenses against those fears” (Skowron & Schmitt, 2003, p. 212). Higher scores reflect higher levels of differentiation of self and less EC. FO (12-item subscale) refers to “emotional over-involvement with others… heavy reliance on others in decision making… and holding few constant beliefs of one’s own” (Skowron & Schmitt, 2003, p. 214). Higher scores reflect higher levels of differentiation of self and less FO. Previous researchers have established the validity and acceptable reliability of these subscales (e.g., Skowron & Friedlander, 1998; Skowron & Schmitt, 2003). In the present study, reliability for the four subscales was as follows: ER (α = .87); IP (α = .82), EC (α = .83), FO (α = .76).

The EAT-26 measures general eating disorder symptomatology (Garner et al., 1982). The scale is composed of 26 questions on 6-point Likert-type scale (1 = never; 6 = always); these items inquire into disordered thoughts and behaviors (e.g., binge eating, self-induced vomiting, food restriction). Higher scores indicate higher levels of eating disorder symptomatology. The EAT-26 has three subscales (dieting, oral control, and bulimia and food preoccupation), but authors often use this scale as a total score to assess general eating disorder symptomatology among males (e.g., Blackmer, Searight, & Ratwick, 2011; Gan, Mohd Nasir, Zalilah, & Hazizi, 2011; Reyes-Rodriguez et al., 2011; Tao, 2010). This instrument was originally developed with female participants and was reported to have appropriate reliability and validity (Garner et al., 1982), but researchers have utilized the EAT-26 with male participants and have reported sufficient reliability (Gan et al., 2011; Reyes-Rodriguez et al., 2011; Tao, 2010). With the current study participants, reliability for the EAT-26 was α = .88.

The EAT-26 does not establish a clinical eating disorder diagnosis, but it does serve as a screening tool to identify individuals who may be exhibiting concerning symptoms (Garner et al., 1982). Garner (n.d.) discussed the use of a score of 20 or above on the EAT-26 as an indicator of potentially problematic eating disorder symptoms that would warrant further assessment to determine a clinical diagnosis. Among our participants, the mean EAT-26 score was 54.35, which is above the cutoff score of 20. In fact, all participants scored above 20 on the EAT-26 (range of 26–111).

Data Analysis

Prior to data analysis, we examined the main continuous variables of interest for approximation of normality. The four subscales of DSI-R had skew and kurtosis values that have been considered acceptable (Miles & Shevlin, 2001). The EAT-26 variable evidenced potentially concerning deviations from normality (Miles & Shevlin, 2001). Thus, we carried out a logarithm transformation on the EAT-26 variable, which improved both its skew and kurtosis values and suggested an increased estimation of a normal curve (Tabachnick & Fidell, 2007).

We used Statistical Package for the Social Sciences statistical program for our data analyses. Our data analysis procedures included an initial examination of links between demographic variables and eating disorder symptoms. Then, we carried out Pearson correlation analyses and a hierarchical regression, in order to investigate the relationships between the differentiation of self-variables (i.e., ER, IP, EC, and FO) and eating disorder symptoms. In the data analysis procedures described in the following section, we used the transformed EAT-26 scores for parametric statistics, that is, bivariate Pearson correlations and the hierarchical regression. When reporting results from nonparametric statistics (i.e., Kruskal–Wallis and Mann–Whitney U tests), which do not assume normality of distributions (Pallant, 2007), we used the nontransformed EAT-26 scores.

Pearson Correlations and Hierarchical Regression

In order to answer the research questions posed in this study—the relationships between four differentiation of self-variables and eating disorder symptoms—we carried out Pearson correlation analyses and a hierarchical regression procedure. The bivariate correlations provided initial information about the relationships between these variables (Research Questions 1–4), while the hierarchical regression provided a more statistically rigorous test of the link between these variables (Research Question 5).

After carrying out Pearson correlations, we found statistically significant associations between the four differentiation of self-variables and eating disorder symptoms (Table 1). In all cases, participants with increased levels of differentiation of self had reduced rates of eating disorder symptoms. Participants with less ER had reduced rates of eating disorder symptoms (r = −.29, p = .000). Participants with higher levels of an IP had reduced rates of eating disorder symptoms (r = −.26, p = .000). Participants who reported less emotionally cutoff relationships had reduced incidence of eating disorder symptoms (r = −.20, p = .000). In addition, participants who reported less FO reported less eating disorder symptoms (r = −.15, p = .006). All of these correlation findings were small effect sizes (Cohen, 1992).

OPEN IN VIEWER

Table 1.

Means, Standard Deviations, Observed and Possible Ranges, and Correlations for the EAT-26 and the Four DSI Subscales (n = 349).

	M	SD	Observed Range	Possible Range	1	2	3	4	5
1. EAT-26	54.35	15.78	26–111	26–156
2. ER	42.47	10.42	12–65	11–66	−.286**
3. IP	47.95	8.74	22–66	11–66	−.263**	.574**
4. EC	52.36	10.17	21–72	13–76	−.195**	.348**	.409**
5. FO	42.41	8.9	56–14	12–72	−.146*	.692**	.432**	.296**

Note. DSI = Differentiation of Self-Inventory; EAT-26 = Eating Attitudes Test-26.

Eating disorder symptomatology as measured by the EAT-26; ER = emotional reactivity as measured by the DSI-R subscale; IP = “I” Position as measured by the DSI-R subscale; EC = emotional cutoff as measured by the DSI-R subscale; FO = fusion with others as measured by the DSI-R subscale. The M, SD, and range values are reported using the nontransformed EAT-26 variable; the correlations are reported using the transformed EAT-26 variable.

*p <.01. **p <.001.

In order to carry out a more stringent test of the association between differentiation of self-factors and eating disorder symptoms, we conducted a hierarchical regression (Table 2). In the first step of the regression, we entered BMI and economic class, thus including a total of 347 participants in this regression, due to missing BMI data for two participants. Explaining 5.7% of the variance in eating disorder symptoms, this step was statistically significant, ▵F(2, 344) = 10.37, p = .000. As characterized by Cohen (1992), this was a small effect size (f ² = .06; Soper, 2013b). In the second step of the regression, we entered the four differentiation of self-variables: ER, IP, EC, and FO. Together, these variables accounted for an additional 11.4% of the variance in eating disorder symptoms. This step in the regression was also statistically significant, ▵F(4, 340) = 11.69, p = .000. The effect size for this step (f ² = .064; Soper, 2013a) has been described as a small effect (Cohen, 1992).

OPEN IN VIEWER

Table 2.

Unstandardized Regression Coefficients (B) and Intercept, the Standardized Regression Coefficient (β), t Values, and p Values, for Variables as Predictors of Eating Disorder Symptomatology (n = 347).

		B	SE	β	t	Significance
1	(Constant)	1.605	.037		43.587	.000
	BMI	0.006	.001	.201	3.845	.000
	Economic	−0.033	.014	−.129	−2.465	.014
2	(Constant)	1.792	.050		35.743	.000
	BMI	0.006	.001	.228	4.594	.000
	Economic	−0.029	.013	−.113	−2.280	.023
	ER	−0.003	.001	−.267	−3.521	.000
	IP	−0.002	.001	−.133	−2.128	.034
	EC	−0.001	.001	−.087	−1.586	.114
	FO	0.002	.001	.113	1.638	.102

Note. BMI = body mass index; ER= emotional reactivity as measured by the DSI-R subscale; IP = “I” Position as measured by the DSI-R subscale; EC = emotional cutoff as measured by the DSI-R subscale; FO = fusion with others as measured by the DSI-R subscale; dependent variable = eating disorder symptomatology, as measured by the EAT-26.

Upon examination of the relative impact of each differentiation of self-variable, however, we found that only two variables reached the threshold of statistical significance. After controlling for BMI and economic background and accounting for the impact of the other differentiation of self-variables, EC and FO were no longer significantly linked to eating disorder symptoms. ER (β = −.267, B = −.003, p = .000) and IP (β = −.133, B = −.002, p = .034) retained statistical significance. The β values of the two variables revealed that ER was more strongly linked with eating disorder symptoms than IP.

ER and IP

We found that ER and IP, the intrapersonal elements of differentiation (Skowron et al., 2003), were significantly associated with eating disorder symptoms. In addition, we found that these two aspects of differentiation had the strongest links with eating disorder symptoms when considered alongside the interpersonal elements of differentiation. These findings correspond to the results reported by Schwartz et al. (2010). While these authors did not measure eating disorder symptoms, they did measure body image attitudes (i.e., a desire for reduced body fat and increased muscularity) and reported that lower levels of an IP were linked with an increased aim for both less body fat and additional muscularity. These researchers also reported that higher levels of ER were associated with an increased desire for less body fat. Researchers have found that college student male participants with eating disorder symptoms reported more dissatisfaction with their bodies than male participants without eating disorder symptomatology (Ousley, Cordero, & White, 2008). Thus, displeasure with one’s body may be one pathway by which males develop and continue engaging in eating disorder symptoms (Ousley et al., 2008). Given our findings and the Schwartz et al. (2010) results, it may be that, for some males, elements of differentiation which involve developing a strong sense of self (high IP) and not being unduly impacted by emotions (low ER) are related to both a diminished desire for less body fat and decreased eating disorder symptoms. Potentially, individuals with a high IP adhere to an internal value system which is not impacted by body image standards proliferated in society and that these individuals thus experience less body dissatisfaction and reduced eating disorder symptoms (Schwartz et al., 2010). In addition, individuals who are less emotionally reactive, namely, less susceptible to emotional reactions and being impacted by the emotional states of others (Skowron & Schmitt, 2003) may have less body dissatisfaction and eating disorder symptoms. Such individuals may be able to avoid emotional investment in body image ideals, thus diminishing body dissatisfaction and eating disorder symptoms.

Alternately, other variables may be involved. For example, individuals with a high IP and less ER may also evidence effective coping skills to manage stress (Bowen, 1978; Kerr & Bowen, 1988; Murdock & Gore, 2004). Researchers have reported that poor emotion regulation skills are linked to increased eating disorder symptoms among males, thus suggesting that males may use eating disorders as way to cope with unpleasant emotions (Lavender & Anderson, 2010). Thus, individuals with a high IP, less ER, and solid coping skills, would not use unhealthy symptoms, such as eating disorder behaviors, as a way to deal with negative affect.

FO and EC

We also found a correlation between FO and EC, the interpersonal elements of differentiation (Skowron & Schmitt, 2003), and eating disorder symptoms among males. Corroborating previous findings (Blackmer et al., 2011; Levy & Hadley, 1998), we found that males with increased levels of FO had increased eating disorder symptoms. Our findings contrast, however, with the Schwartz et al. (2010) study. These researchers reported no link between FO and ECs and a desire for reduced body fat or increased muscularity. In part, this difference may be due to the improved FO subscale we used. Skowron and Schmitt (2003) revised the FO subscale to improve reliability and validity in the DSI-R; Schwartz et al. (2010) used the original FO subscale in the DSI (Skowron & Friedlander, 1998). On the other hand, as Schwartz et al. (2010) studied body attitudes and not eating disorder symptomatology; thus, it may be that FO is related to eating disorder symptoms, but not related to a desire for less body fat or increased muscularity among males. Similarly, our findings suggest that ECs are also correlated with eating disorder symptoms, despite the Schwartz et al. (2010) results wherein EC was nonsignificantly associated with a body image concerns.

Our findings suggest that some males who struggle with high levels of fusion and ECs in relationship with others may engage in eating disorder symptoms. Given the previously discussed Schwartz et al. (2010) findings, these individuals may struggle with eating disorder symptoms for reasons separate from body dissatisfaction. Termed the interpersonal elements of differentiation (Skowron et al., 2003), FO and EC involve an individual’s ability to be independent from, while also bonded with, others. An individual who is cutoff from others evades connection and closeness and an individual who is fused with others cannot separate his own self from that of others (Kerr & Bowen, 1988; Skowron & Schmitt, 2003). Such characteristics suggest problematic relational functioning (Kerr & Bowen, 1988) and may set up individual to be vulnerable to eating disorder behaviors and attitudes. Such an individual likely lacks a sense of competency in relationships and is likely to have unfulfilling, troubling, and/or nonexistent relationships with others (Kerr & Bowen, 1988). Researchers have pointed to a connection between depressive symptoms and a lack of satisfying relationships among male college student participants (Givertz & Safford, 2011; Oliffe et al., 2010). Researchers have also linked depression with increased unhealthy dieting behaviors (Boyes, Feltcher, & Latner, 2007) and disordered eating behaviors (Gan et al., 2011) among male college students. Thus, males who experience failures in intimate relationships may experience depression and consequently engage in eating disorder symptoms.

When considered in tandem with the intrapersonal elements of differentiation, however, EC and FO no longer significantly contributed to the explanation of eating disorder symptoms. Previous researchers who reported findings on the link between fusion in family relationships and eating disorder symptoms did not examine the complete construct of differentiation, but rather focused on the interpersonal elements of differentiation (Blackmer et al., 2011; Levy & Hadley, 1998). Thus, while FO and ECs in relationships may be associated with eating disorder symptoms, these constructs appear to be less salient in comparison to the intrapersonal domains of differentiation.

Family Counseling Implications

Family counselors may want to attend to a male client’s level of differentiation of self, particularly if that client reports eating disorder symptoms. Our study results suggest that four elements of differentiation (IP, ER, FO, and EC) may be pertinent to male eating disorder symptomatology. In particular, ER and IP were especially relevant to male eating disorder behaviors and attitudes, with ER evincing the strongest relationship. Family counselors may thus want to focus clinical interventions on a male client’s sense of self and particularly on a client’s susceptibility to emotional reactions. In the following section, we provide recommendations for assessment and intervention related to differentiation of self.

Strengths and Limitations

Our study is strengthened by our attention to several deficits in the literature on eating disorders and differentiation among males by, for example, assessing four domains of differentiation. Prior researchers have emphasized the interpersonal elements of differentiation (Blackmer et al., 2011; Levy & Hadley, 1998); our study examined the intrapersonal domains as well. Our study is also limited in several ways. The response rate may be a limitation, in addition to the demographic traits of participants, who were predominantly White/Caucasian and heterosexual. This lack of diversity may reduce the generalizability of findings. Another limitation involves the intended participant age range for the DSI-R, which is at least 25 (Skowron & Friedlander, 1998; Skowron & Schmitt, 2003). Schwartz et al. (2010) used the original DSI, also intended for a participant age of at least 25 (Skowron & Friedlander, 1998), and noted the potential limitations involved with a younger participant age. Also, as noted previously, the EAT-26 was developed with female participants (Garner et al., 1982). Previous researchers have used the EAT-26 with male participants (e.g., Gan et al., 2011; Reyes-Rodriguez et al., 2011; Tao, 2010), but this nonetheless represents a limitation of the study.