Peer-reviewed journal articles and unpublished dissertations were included in the review if they were an original quantitative research study that developed, validated and/or investigated the psychometric properties of a self-report measure of trait impostorism and sampled an adolescent or adult population. Only studies published in the English language were included which also included studies conducted on non-English speaking samples, as long as the research was based on trait impostor phenomenon measures.

Studies were excluded in the review if a child population was utilized, were non-peer reviewed journal articles, conference proceedings, non-psychometric studies and not written in the English language. It was noted, there are currently no evidence-based interventions for the impostor phenomonenon and as a result, comparators or outcomes in the literature to be accounted for by this systematic review. Therefore, this review has been limited to comprehensively defining the populations of interest and specific study designs in the inclusion and exclusion criteria.

All studies provided adequate evidence of the measurement aim, target population, and concepts being measured. The assessment framework (Terwee et al., 2007) classifies item selection as relevant criteria for content validity, however, this review only considered item selection as a compulsory criterion for original scale development studies of which there were only three studies. Harvey's (1981) study was rated positively for content validity because item selection was driven by theoretical and therapeutic observations, and reported item analysis statistics. Kolligian and Sternberg's (1991) study also received a positive rating for sufficient item selection information however, Leary et al.'s (2000) article was rated indeterminate overall for content validity as Study 1 provided adequate evidence, however, Study 2 did not provide item selection information for the LIS development. The PFS and LIS were newly developed impostor phenomenon measures at the time and scale development data was only available upon request from the first authors. Responses from first authors had not been received by the time of submission for this publication. Two other studies were also allocated an indeterminate rating for content validity. Simon and Choi (2018) and Brauer and Wolf's (2016) studies provided brief measurement aims, explanations for the constructs of interest and little to no justification for the target populations sampled.

Three studies received positive ratings for internal consistency. These studies conducted factor analysis on an adequate sample size, with appropriate design and method, and reported Cronbach's alphas between 0.70 and 0.95 for each dimension and overall. The remaining studies were allocated indeterminate ratings for gaps in addressing one or more criterion for internal consistency. Item Response Theory analysis was also an acceptable form of analysis for the internal consistency criterion, however, none of the selected studies in the review sample carried out this form of analysis.

Sixteen studies reported Cronbach alpha scores of adequate magnitude for the impostor phenomenon measures. Among the 11 CIPS studies, overall Cronbach alphas ranged from 0.85 to 0.96. Seven of these studies examined the factorial structure of the CIPS and only three reported the subscale reliability statistics (French et al., 2008; McElwee and Yurak, 2010; Brauer and Wolf, 2016). Cronbach alphas were presented for factors in a theoretically preferred three factor model for the CIPS—Fake (0.84), Discount (0.79), and Luck (0.70), compared to a statistically better fitting two factor model, however, subscale reliabilities were not reported for the two factor model (French et al., 2008). Similarly, a three factor model was replicated for the CIPS with subscale reliabilities ranging from 0.74 to 0.89 (McElwee and Yurak, 2010). A third study validated the German CIPS (0.87–0.89) utilizing exploratory and confirmatory factor analysis with two samples (Brauer and Wolf, 2016). A three factor model resulted in the best fit statistics and Cronbach alphas for each factor Fake (0.84), Discount (0.73), and Luck (0.69). This three factor structure aligned to the typical three characteristics of the impostor phenomenon presented by Clance (1985) and equivalent to the English version of the CIPS—feeling like a fake, discounting achievement, and attributing success to luck.

Five studies primarily examined the HIPS with overall Cronbach alphas ranging from 0.34 to 0.85, in addition to a study comparing the CIPS and HIPS (α = 0.91) (Holmes et al., 1993) and a second study comparing the PFS to HIPS (α = 0.64) (Kolligian and Sternberg, 1991). Three HIPS studies explored the factorial structure of the measure. One study proposed a three factor model with subscale reliabilities between 0.65 and 0.81 (Edwards et al., 1987). In comparison, a HIPS four-factor model was presented with moderate correlations however subscale alphas were not reported (Fried-Buchalter, 1992). In an adolescent sample, Hellman and Caselman (2004) reported an alpha of 0.70 for the original 14 items. However, following factor analysis, an alpha of 0.80 was reported for a better fitting 11-item two factor model (self-confidence and impostor phenomenon) for the HIPS. The subscale reliabilities were not presented.

The original PFS validation study proposed a two factor model with an overall Cronbach alpha of 0.94 and subscale reliabilities for Inauthenticity (0.95) and Self-deprecation (0.85) (Kolligian and Sternberg, 1991). Similarly, a CIPS validation study also reported an alpha of 0.94 for the PFS. However, when the Spearman-Brown equation was applied to the 51-item PFS to reduce it to the 20-item CIPS equivalent, the estimated internal reliability of the PFS was 0.57 (Chrisman et al., 1995). Leary et al.'s (2000) Study 2 reported a Cronbach's alpha of 0.87 for the unidimensional LIS.

A clear “gold standard” measure of the impostor phenomenon was not ascertained. Most studies did not compare the impostorism measure utilized to a “gold standard” and, if the measure was compared to an alternate impostorism measure, limited convincing rationale was provided to establish the measure as a “gold standard.” Overall, all studies in this review were allocated a “Currently Not Possible” (CNP) rating for criterion validity. Of the reviewed studies, four studies utilized two or more impostor phenomenon measures and reported correlation coefficients. Holmes et al. (1993) reported a coefficient of .89 between the CIPS and HIPS, while Chrisman et al. (1995) reported a coefficient of 0.78 between the CIPS and PFS. Kolligian and Sternberg (1991) reported a correlation of 0.83 between the PFS and HIPS. Leary et al.'s (2000) third study reported correlation coefficients between the LIS and the HIPS, CIPS, and PFS ranging from 0.70 to 0.80 and noted the LIS “showed strong evidence on construct validity” (Leary et al., 2000, p. 735).

The construct validity criteria require studies to formulate theory-driven specific hypotheses and at least 75% of the results are in correspondence with these hypotheses to obtain a positive rating (Terwee et al., 2007). Based on these strict criteria, six studies were evaluated with a positive rating and achieved the maximum score on this property. These studies presented specific theoretically derived hypotheses that highlighted the extent to which scores on the particular impostor phenomenon measure related to other measures in a consistent manner. Among the positively rated studies examining the CIPS, HIPS, and PFS consistent yet discriminant relationships were established with other constructs. Correlations ranged from 0.34 to 0.69. Higher impostorism was associated with constructs such as low self-esteem, low confidence, high self-monitoring, higher depressive symptoms, higher anxiety, and higher negative self-evaluations than their lower impostorism counterparts (e.g., Topping, 1983; Kolligian and Sternberg, 1991; Chrisman et al., 1995; Rohrmann et al., 2016).

However, the majority of studies were intermediately rated and two studies were poorly rated because despite formulating specific hypotheses, less than 75% of the results were in accordance with the hypotheses presented (Topping, 1983) and did not consider the impostor phenomenon in relation to other measures (Edwards et al., 1987). Studies were often allocated an intermediate rating for reasons such as not formulating hypotheses rather exploratory research questions and, not considering the extent to which impostorism measures related to other measures in a manner consistent with theoretically driven hypotheses.

The agreement criterion is defined as the extent scores on repeated measures are close to each other (absolute measurement error) (Terwee et al., 2007). The 18 studies in the review sample did not examine repeated measures of the impostor phenomenon, therefore no information was available for agreement. Thus, all studies were assigned an “NR” rating on this measurement property as longitudinal data was not collected.

Reliability (test-retest) refers to the extent individuals are distinguishable from each other, despite measurement errors (Terwee et al., 2007). Among the 18 studies, changes over time in impostor phenomenon scores were not investigated and therefore allocated not reported ratings.

Responsiveness is considered a measurement property of longitudinal validity and defined as the ability of a measure to detect clinically important changes over time (Terwee et al., 2007). Since none of the studies selected examined longitudinal changes of the impostorism measures and the construct is not classified as a clinical condition, not reported ratings were applied for all studies.

Overall, four studies reported information noting equal to or less than 15% of respondents achieved the highest or lowest possible scores on the impostor phenomenon measures utilized, or provided sufficient statistics to calculate this information and therefore, received the maximum score on this criteria (Topping, 1983; Edwards et al., 1987; Holmes et al., 1993; Brauer and Wolf, 2016). Topping (1983) reported floor and ceiling results for three faculty groups, while Holmes et al. (1993) examined the CIPS and HIPS and presented the number of participants who scored the minimum and maximum across clinical and non-clinical groups. Brauer and Wolf (2016) provided skew statistics and minimum/maximum data and Edwards et al. (1987) also reported relevant statistics to ascertain whether floor and ceiling effects were present. The remaining papers did not provide information on floor and ceiling effects and subsequently allocated low ratings.

Three studies were appraised positively for providing sufficient descriptive statistics for at least four relevant subgroups. Holmes et al. (1993) provided means and standard deviations for clinician referred and non-clinical subgroups on the CIPS, while Topping (1983) comprehensively reported descriptive statistics by gender and faculty rank on the HIPS. Jöstl et al. (2012) also reported descriptive statistics by gender and faculty membership. However, the remaining studies were allocated indeterminate or poor ratings due to inadequate information to assign qualitative meaning to quantitative findings. It was also noted, several studies did not comprehensively report descriptive statistics for sample characteristics such as mean age, standard deviations and gender breakdown (refer to Table 2).

The majority of selected studies provided adequate information for content validity and internal consistency. However, gaps were evident on several criteria. Based on the study ratings for all nine measurement properties, establishing the internal dimensionality of the impostor phenomenon could not be reached due to mixed results from the selected papers. Four impostorism scales were identified—CIPS, HIPS, PFS, and LIS—which demonstrated moderate to high internal consistency, with the exception of two HIPS studies (Edwards et al., 1987; Kolligian and Sternberg, 1991).

Seven studies utilized factor analysis to develop or validate impostor phenomenon scales. The English and German CIPS were factor analyzed resulting in a three factor theoretically preferred model aligned to Clance's (1985) original conceptualization of the impostor phenomenon as Fake, Luck and Discount (Holmes et al., 1993; Chrisman et al., 1995; Brauer and Wolf, 2016). However, a two factor model was shown to have a better statistical fit when compared to the three factor solution (Chrisman et al., 1995). Of interest, the English CIPS studies excluded four items from the 20-item CIPS to achieve a three factor model and the rationale was based on the results of a frequently referenced (but currently unobtainable) unpublished dissertation by Kertay et al. (1992). Despite the use of a 16-item CIPS which resulted in a theoretically preferred three factor model, the 20-item one factor model of the CIPS continues to be used and a total score is calculated for individuals who complete the measure. By comparison, the HIPS has been validated thrice using factor analysis for scale validation resulting in a two factor model (Edwards et al., 1987; Hellman and Caselman, 2004) and a third time, for construct validation against the fear of success/failure resulting in a four factor model (Fried-Buchalter, 1992). Similar to the CIPS, an overall total score continues to be calculated on the HIPS rather than subscale scores to reflect the multidimensional definition of the measure as postulated by Harvey (1981). Further to this, validation of the PFS has resulted in a two factor-solution of Inauthenticity and Self-deprecation as dimensions of Perceived Fraudulence, yet scoring of the measure involves an overall total score much like the CIPS and HIPS. Since the LIS is a unidimensional measure of the impostor phenomenon it was considered logical to calculate an overall impostorism score.

These results raise points for discussion about the conceptualization and dimensionality of impostorism. The CIPS, HIPS and PFS are based on multidimensional definitions of the impostor phenomenon, however, the measures are operationalized in research and applied settings by calculating overall scores, rather than subscale scores based on the corresponding factors. This evidence suggests a definitive conclusion about the dimensionality of the impostor phenomenon is unclear. Scoring of these measures appears to reflect a unidimensional conceptualization of the construct despite factor analysis results that indicate multiple dimensions. Clearer and convincing rationale would provide greater clarity about this methodology in scoring. Furthermore, additional evidence from future validation studies will aid in establishing item homogeneity or heterogeneity to ascertain the dimensionality of the impostor phenomenon.

Criterion validity was problematic in the selected studies because a clear ‘gold standard’ measure could not be ascertained. The quality appraisal tool required convincing arguments for a gold standard to be “gold” and a correlation coefficient with the gold standard equal to or greater than 0.70 (Terwee et al., 2007). This review has highlighted a gold standard is yet to be established and this is due to a number of factors relating to chronology, dimensional clarity and scale popularity.

In earlier studies, it was not possible to provide gold standard comparisons since the HIPS (Harvey, 1981) was the first impostor phenomenon measure developed and the CIPS was not constructed until 1985 for comparison to be possible. Five of the selected studies compared various impostor phenomenon measures to each other and reported consistently high correlation coefficients exceeding .70. A comparison of the CIPS and HIPS utilizing an ANCOVA, appeared to identify the CIPS as more sensitive in distinguishing between high and low impostorism in the non-clinical sample when CIPS scores were held constant (Holmes et al., 1993). It was also argued the CIPS had reduced incidents of false positives and false negatives when establishing cut-off scores. Researchers acknowledged that participants were not randomly assigned to groups, however, it was suggested the CIPS is the instrument of choice for use with the general population due to its sensitivity and reliability (Holmes et al., 1993). A comparison of the CIPS and PFS suggests the CIPS has better utility because of its brevity in comparison to the 51-item PFS. However, problematic interpretations of correlation coefficients did not provide convincing arguments to establish discriminant validity of the impostor phenomenon with other constructs and to assert the strengths of the CIPS over the PFS (Chrisman et al., 1995). Furthermore, instances of low internal consistency for the HIPS and the questionable stability of impostorism as measured by 14 items in a one factor model, rather than 11 items in a better fitting two factor solution did not provide convincing arguments for the HIPS being the gold standard either (Hellman and Caselman, 2004). These findings did not provide sufficient evidence to establish the CIPS nor HIPS as a gold standard measure of the impostor phenomenon. In addition, the PFS and LIS also require further validation to establish its strengths and utility as a potential gold standard measure. The limited number of PFS and LIS studies in this review demonstrated the need for further exploration of these measures.

An argument could be made the CIPS is the ‘gold standard’ measure by virtue of it being the most commonly cited and utilized measure by practitioners and impostor phenomenon researchers. However, popularity is not necessarily a reflection of higher quality. It would be premature to classify the CIPS as the gold standard measure of the impostor phenomenon in light of the results from this review. There remains to be questions about the dimensionality of the impostor phenomenon and its operationalization in measures such as the CIPS, HIPS, and PFS. Despite being based on multidimensional definitions of the construct, these measures calculate overall total scores and do not define subscale scores. Scoring of these measures appears to contradict the theoretical conceptualization of the impostor phenomenon. Without robust and consistent validation results and conceptual clarity, it is currently premature to select a ‘gold standard’ measure within the context of an evidence base that is still growing.

The search strategy did result in the inclusion of two unpublished dissertations. Systematic reviews typically limit searches to peer-reviewed journal articles, however, due to the limited number of validation studies of impostor phenomenon measures a consensus decision was made by the researchers to include these dissertations. Following examination of reference lists and noticing frequent citations of these studies, it was justified to include the original scale development study for the HIPS and an often cited dissertation examining the construct validation of the HIPS (Harvey, 1981; Topping, 1983).

The quality assessment framework used to appraise the measurement properties of the selected studies was originally designed for evaluating self-report questionnaires of officially diagnosable health conditions (Terwee et al., 2007). As expected, assessment criteria for specific methodological and psychometric properties are strict to distinguish between quality measures as they relate to patient reported health outcomes. However, the impostor phenomenon is not an official clinical condition, yet its measures are similar to self-report health measures and used to identify people and the degree to which they experience this phenomenon. Since impostorism is also linked to poorer well-being and poorer mental health outcomes (e.g., Chrisman et al., 1995; Sonnak and Towell), it was deemed appropriate to use this rigorous assessment framework and to tailor the application of specific criteria to studies to better reflect the current state of impostor phenomenon research. For example, the criteria to calculate Cronbach's alpha per dimension for internal consistency was only applicable to studies that conceptualized impostorism as a multidimensional construct. Despite these adjustments, specific psychometric properties did not necessarily receive higher scores and in some instances, it was not possible, however, it is emphasized these lower ratings are not necessarily a reflection of a poor study, but rather an absence of data or a current gap in the overall impostor phenomenon literature. The comprehensive reporting of essential data was a crucial element for studies to be appraised with higher scores based on the measurement criteria. A recommendation for improved reporting of data is made for future research. Also, the limited number of studies identified in this review, highlight the opportunity for increased research validating measures of impostorism and utilizing rigorous criteria to aid in research design and reporting, similar to established clinical conditions.

An alternative assessment framework was also considered for the present review—the Consensus-Based Standards for the Selection of Health Measurement Instruments (COSMIN) checklist—however, the criteria in this appraisal tool was considered too strict to assess the methodological quality of current impostor phenomenon studies (Mokkink et al., 2010; Terwee et al., 2012). The COSMIN checklist was also originally designed for the evaluation of Health-Related Patient-Reported Outcomes measures. Since impostorism is not an established diagnosable health condition and a definitive conclusion about its dimensionality has not been ascertained, applying an appraisal tool such as the COSMIN checklist may have resulted in much lower ratings that were not truly representative of the quality of current impostorism validation research.

This review focused on trait measures of the impostor phenomenon. Therefore, measures based on definitions that conceptually assume consistencies in thoughts, feelings and behaviors across situations and time. Despite this theoretical assumption, research is yet to examine the longitudinal variability of impostorism measures, as was the case for all studies in this review. As a result, each study was allocated “not reported” ratings for reproducibility properties. Of interest, the HIPS, CIPS and PFS were examined in relation to “reflected” vs. “self-appraisals” (Leary et al., 2000). However, despite examining the test-retest reliability of the appraisal measures at the 6 to 8-week follow-up, the impostor phenomenon measures were not revisited to obtain longitudinal data. Rationale was not provided for this research design and highlights a consistent gap in the current evidence base. Similarly, another study examining undergraduates' impostorism feelings, self-esteem, personality, and predicted marks were collected at the beginning of the semester. However, at follow-up 1 week prior to an exam and after exam grades were released, longitudinal CIPS data was not collected alongside measures of self-esteem and questions about attributions of success or failure (Cozzarelli and Major, 1990). It is evident there is scope to improve research design in order to explore changes over time detected by trait impostor phenomenon scales. Currently, the developmental trajectory of the impostor phenomenon is unknown. Validation studies are yet to examine the longitudinal variability of impostorism scores. It is recommended future research explores the longitudinal stability and intensity of such scores. It is also suggested that researchers who examine change scores at a minimum, ensure means, standard deviations, and sample sizes are reported to enable sample size calculations.

Based on the quality assessment framework, measurement gaps were identified in each study resulting in indeterminate or poor ratings across one or more criteria. Terwee et al. (2007) assert measures may be given many indeterminate or low ratings across measurement properties that are yet to be evaluated, especially if they are new questionnaires or due to less comprehensively reported validation studies. This subsequently results in lower ratings that are not necessarily due to poor questionnaire design or performance, but rather an absence of data or existing research. This was the case for several studies in this review, for example, the majority of studies demonstrated appropriate content validity and adequate internal consistency. However, due to doubtful design, no Item Response Theory or factor analysis, studies were then allocated indeterminate ratings for the internal consistency criterion. To better understand the dimensionality of the construct and therefore establish conceptual clarity of the impostor phenomenon, it is suggested statistical analyses utilizing Classical Test Theory and/or Item Response Theory are applied to impostorism measures with appropriate sample sizes. The comprehensive reporting of essential psychometric data will enable meaningful conclusions.

The floor and ceiling effects criterion, along with interpretability were also indeterminate, poor or largely unavailable for most studies since descriptive statistics were often omitted. The reporting of means and standard deviation scores from the measure of interest for at least four relevant subgroups is stipulated for interpretability, in order to assign qualitative meaning to quantitative data (Terwee et al., 2007). However, a limited number of studies provided this information. It was also common among the studies reviewed for sample descriptive statistics such as age not being reported or only age range (Harvey, 1981; Topping, 1983; Cozzarelli and Major, 1990; Fried-Buchalter, 1992; Leary et al., 2000). Gender distribution was also omitted in Harvey's (1981) original scale development study. It is assumed such data was collected, however, without the publication of this information, it is challenging for researchers and practitioners to make meaningful interpretations of findings and to select the most appropriate measure for specific purposes and populations. In the absence of essential data where practitioners require robust psychometrics to evaluate the utility of a measure, for example, for therapeutic settings, there is an increased risk of false positives and misidentifying those experiencing the impostor phenomenon.

Comprehensive reporting of psychometric data is necessary to assess the hypothesized structure of a construct and to evaluate the adequacy of its operationalization in measurement scales. Conducting this review also highlighted challenges in accessing research from the impostor phenomenon literature. While it was acknowledged comprehensive scale development data for some measures were only accessible upon request from first authors (Kolligian and Sternberg, 1991; Kertay et al., 1992; Leary et al., 2000), in most instances, attempts to obtain research data and/or papers in a timely manner were unsuccessful. This highlighted an important point about improving accessibility to scale development and validation information. Increasing the collection and publication of results will allow the scientific community to easily evaluate the quality and utility of impostor phenomenon instruments for future investigative and applied purposes, as will the trend in making research datasets publicly available.

This review has identified different conceptualizations of the impostor phenomenon and the measures associated with these definitions. There is a great deal of variability in the methodological quality and findings available to establish the dimensionality of the construct and therefore questions around what it is, what characteristics it is made up of and whether a gold standard instrument exists. For the purposes of research and applied settings, a clear purpose, target population and definition of the construct is necessary to select the most appropriate measure for its intended use. Availability of an established psychometrically sound gold standard measure of impostorism is also likely to be useful in related clinical areas of research where patient populations fear not meeting an inferred “audience” standard that they assume will result in negative evaluation; such populations also discount corrective information including the non-occurrence of feared outcomes. This is particularly the case for anxiety disorders featuring social threat fears such as social anxiety disorder (Rapee and Heimberg, 1997) and generalized anxiety disorder (American Psychiatric Association, 2013), though beliefs about being negatively evaluated by others also features in depression (Swann et al., 1992). A rigorous review of psychometric properties and justification for the use of a specific scale is always necessary. However, since the developmental trajectory of the impostor phenomenon is currently unknown, the use of such measures especially with child, adolescent, and older populations requires theoretically driven justification even more so. At least until greater understanding is established of the longitudinal variability of impostorism scores across the ages.