Fear of Fraudulence: Graduate School Program Environments and the Impostor Phenomenon

ABSTRACT

Those who suffer from impostorism experience feelings of fraudulence, worrying that they are fooling others about their abilities and that they will eventually be exposed. While prior research emphasizes the trait-like durability of impostor personalities, we argue that impostorism is sensitive to experiences in proximate social environments, such as graduate school programs. The authors examine the relationship between perceived characteristics of graduate school program environments and students’ impostor feelings using survey data from a large university (N = 1,476). Results demonstrate that students’ perceptions of lower-quality mentorship, increased competition, and increased isolation are associated with more frequent impostor fears. The authors discuss the consequences of impostorism in academia and review implications for program policies and future research.

KEYWORDS:

• Impostorism
• social psychology
• education
• mental health

Acknowledgments

The data used in this article were collected by the Indiana University Mental Health Working Group. Versions of this article were presented at meetings of the North Central Sociological Association and the American Sociological Association. The authors sincerely thank Clem Brooks, Jessica Calarco, Jessica Collett, Long Doan, Bianca Manago, and Brian Powell for their valuable comments on earlier drafts.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. For variety, we interchange IP, impostorism, feelings of fraudulence, and impostor fears throughout the article. IP is also referred to as “impostor syndrome” and “perceived fraudulence” (Kolligian and Sternberg 1991). We adopt the term impostor phenomenon used in the original clinical description (Clance and Imes 1978).

2. Further establishing the relevance of social context to IP is a related line of research questioning whether impostorism is a stable personality trait at all (McElwee and Yurak 2007). Following Goffman (1959), this literature argues that impostorism is an impression management strategy adopted to mitigate negative judgments when individuals experience failure (Ferrari and Thompson 2006; Leary et al. 2000). This line of research suggests that situational cues embedded in social contexts, rather than internalized traits, shape reported feelings of fraudulence.

3. The relation among stereotypes, impostor fears, and performance is a fascinating issue. However, there are a number of important distinctions between the current literature on stereotype threat and the current literature on IP. Stereotype threat research demonstrates that being subject to negative stereotypes can depress academic performance (Spencer, Steele, and Quinn 1999; Steele and Aronson 1995) because anxiety about fulfilling the negative stereotype leads to expenditure of cognitive energy on self-monitoring and emotional regulation rather than on the performance task itself (Schmader, Johns, and Forbes 2008). In contrast, IP is said to be an internalized feeling of fraudulence, and those who suffer from it are likely to perform very well despite their negative self-perception (Clance and O’Toole 1988). Scholars have not yet brought these two strands together to discover the role of stereotypes in perceived fraudulence, although there are many documented reports of students from underrepresented groups feeling like impostors (Boyd 2012; Gardner and Holley 2011; Granfield 1991).

4. Sensitivity analyses indicate that the subsample of cases with any missing data does not differ demographically from the subsample with complete information, with one exception. Based on two-sample tests, cases with any missing values are significantly more likely be a race other than white. Consequently, our results may not be generalizable to the population of graduate students of color. However, we estimated ordinary least squares models predicting IP using poststratification survey weights for gender, race, and degree type; see Table A1 for details. The weighted and unweighted analyses produce substantively similar results with respect to the size and significance of effects for our explanatory variables. In addition, results are nearly identical whether we multiply impute or delete missing data, including for students of color. Results for models based on survey weights or imputed data are available on request.

5. Sakulku and Alexander (2011) recommended restricting usage of impostorism to refer to the small subgroup who exhibit high levels of IP. However, we choose a continuous IP variable that gives us more flexibility in our statistical analyses. We conduct robustness checks using a cutoff point to designate a subgroup of high-scoring impostors (i.e., those above the 75th percentile; see Holmes et al. 1993). Our results are substantively similar whether using the continuous or dichotomous outcome.

6. Additional descriptive statistics for the impostorism items and scale are in Table A2. Our second impostorism item directly measures negative social comparisons. While this is less commonly used for measuring impostorism, inter-item reliability of the scale suggests that it is a worthwhile addition (Cronbach’s α = .86). Furthermore, we conducted supplemental analyses in which we use mixed effects ordinal regression to model each impostorism item individually (Table A3). We obtain substantively similar results for each item, which suggests that no single item is driving our results. Items included in the well-known Clance Impostorism Scale that are similar to our items include, “I often worry about not succeeding with a project or on an examination, even though others around me have considerable confidence that I will do well”; “I’m afraid people important to me may find out that I’m not as capable as they think I am”; “Sometimes I’m afraid others will discover how much knowledge or ability I really lack”; “I often compare my ability to those around me and think they may be more intelligent than I am.” Similar items from the Harvey Impostor Phenomenon Scale include, “People tend to believe I am more competent than I really am”; “Sometimes I am afraid I will be discovered for who I really am.” Typically, distorted attributions for success are also considered to be an integral part of impostorism (Clance and Imes 1978). Unfortunately, our measures cannot capture this dimension.

7. Advising typically refers to responsibilities such as providing informational resources. Mentoring covers a broader set of responsibilities related to professional development, career decision-making, and psychosocial support (Rose 2005). We conducted several robustness checks but ultimately do not distinguish between these roles in our results. Throughout the text, we generally refer to “mentorship” or “mentor,” though this is meant to include students with advisors as well.

8. Descriptive statistics for the mentorship quality items are in Table A4.

9. Due to lack of detail in the responses and our relatively small sample size, we cannot further disaggregate these fields. Table A5 reports the average values for our program variables for each of the 10 fields of study.

10 Respondents could select any number of the following mental health problems/traumatic experiences: depression, bipolar, or other mood disorder; posttraumatic stress disorder, obsessive-compulsive disorder, panic disorder, or other anxiety disorder; anorexia, bulimia, or other eating disorder; attention/deficit hyperactivity disorder, autism spectrum disorder, learning disability, or other developmental, learning, or attention disorder; schizophrenia or other psychotic disorder; alcohol or drug abuse/dependence; suicidal thoughts or attempt; self-harm; sexual assault; and, other emotional or stress-related problem.

11. In our sample, nearly 64% of the students of color are international students. The overlap between our race and international student indicators might suggest that both variables should not be included in our models. However, results are the same when we control for just race, just international student status, or both variables. We include both because each factor is substantively meaningful. Evidence suggests that underrepresented minorities may experience impostor fears more often than white students (Ewing et al. 1996). Though some research suggests that the impostor phenomenon is a cross-cultural phenomenon (Sakulku and Alexander 2011), it is not clearly established that this is the case and thus it is important to take into account students’ cultural contexts.

12. We also conducted exploratory factor analysis (EFA) to quantify distinct dimensions of mentorship. Consistent with prior research (Noy and Ray 2012), EFA revealed instrumental and affective factors. Supplementary analyses demonstrate that instrumental and affective mentorship have an identical impact on impostorism; we present results for the combined scale for parsimony.

13. Because of concerns that our impostorism measure is a proxy for anxiety more generally, we conduct two robustness checks. First, we estimate models replacing the mental health binary variable with the more detailed K6 scale, which measures serious mental illness (Kessler et al. 2010). The program effects are substantively identical. Second, we introduce a control measuring how often a student “felt nervous” in the past 30 days. In these models, funding competition no longer significantly affects impostor fears, but other program effects remain unchanged. We have noted our concerns about contemporaneous mental health measurement in text; in light of these concerns, we present results using the binary variable for prior mental health problems.

14. We tested interaction effects between program environment variables and several controls—gender, race, international student status, and field type. We found no differences in the effects of program environment variables on impostor scores across these groups. That is, all interactions were nonsignificant (p > 0.10) according to omnibus tests.

15. We conducted several additional robustness checks. First, we re-estimate all models using clustered ordinary least squares with standard errors clustered by field. Our results are consistent when using this method. Second, results in the first panel of Table 2 are the same whether using the full estimation sample or the restricted sample used in the third panel of Table 2.

Additional information

Funding

This work was supported by the Office of the Dean of the University Graduate School at Indiana University Bloomington.

Notes on contributors

Emma D. Cohen

Emma D. Cohen is a Ph.D. candidate in the Department of Sociology at Indiana University Bloomington. Her research focuses on educational stratification and has been published in Social Science Research, Race and Social Problems, and Socius.

Will R. McConnell

Will R. McConnell is an Assistant Professor of Sociology at Florida Atlantic University. His research focuses on social networks, aging, mental health, and culture. His work appears in the Journal of Health and Social Behavior, Advances in Medical Sociology, Network Science, and other journals.