In comparison to physiological approaches, psychological assessments in professional sports are rather seldom, particularly in soccer. Yet measuring the psychological aspects of athletes is rapidly gaining popularity and many of the well-established measurements from domains external to sport (i.e., cognitive and differential psychology) are now being applied in sporting domains. Examples of assessments that have already made this transition is the measurement of athletes' executive functions (Beavan et al., 2020), emotional behaviors (Schilaty et al., 2016), and of particular interest for this research, personality (Smith, 2008; Zhang et al., 2019). In the area of differential psychology, the classification of the Five Factor Model of Personality (FFM) is well-established (McCrae and Costa, 2008). The FFM assesses personality of an individual on five traits: Neuroticism (N), Extraversion (E), Openness (O), Agreeableness (A), and Conscientiousness (C); hence being commonly referred to as “the big five.” Despite some criticism of whether there are less (Eysenck and Eysenck, 1985; Gray, 1991; Zuckermann, 2005) or more than five dimensions (Paunonen and Jackson, 2000), the FFM is based on a general consensus in modern research and is widely accepted (O'Connor, 2002; de Moor et al., 2012; Allen et al., 2013; Bircher et al., 2017).

The NEO Personality Inventory (NEO-PI) and its subsequent revised version (NEO-PI-R) were developed as a measure of the FFM of personality (McCrae and Costa, 1985; Costa and McCrae, 1992). The NEO-PI-R contains 240 items that are grouped accordingly to one of the five personality dimensions. Despite the NEO-PI-R measuring narrower personality traits with more scales, the time commitment to complete the questionnaire is a limitation for many time-scarce populations, requiring up to 60 min to complete. In high performance settings where practitioners may be allocated limited time access to players, the use of the shorter version of the NEO-PI-R may be more appropriate (Egan et al., 2000). The NEO Five-Factor Inventory (NEO-FFI) is a shortened version of the NEO-PI-R, consisting of only 60 items that take between 10 and 15 min to complete (McCrae and Costa, 1987). It has demonstrated validity and utility in several different contexts and languages and is one of the most used instruments to assess the FFM (Egan et al., 2000; Zillig et al., 2002).

Reliability measures of the NEO-FFI has first been mentioned by Costa and McCrae (1992) in an English speaking sample, which has shown internal consistencies of 0.89 (N), 0.79 (E), 0.76 (O), 0.74 (A), and 0.84 (C) that were later supported (Holden and Fekken, 1994). Egan et al. (2000) similarly reported ranges between 0.72 (O) and 0.87 (N) on a large and diversified British cohort. Results from Eastern Europe and Iran also confirmed similar findings (Hrebíčková et al., 2002; Anisi, 2012). Of particular importance for the current study, the translation by Borkenau and Ostendorf (1993) found reliability indexes between 0.71 and 0.85 in a German population sample that were later on verified (Schmitz et al., 2001). Furthermore, test-retest reliability separated by a 2-week break has also been positively supported, ranging from 0.89 (N), 0.86 (E), 0.88 (O), 0.86 (A), and 0.90 (C) (Robins et al., 2001). Although the test-retest stability of the NEO-FFI is high, concerns for the factor structure of the NEO-FFI have been highlighted in the literature. Holden and Fekken (1994) conducted a factor analysis on the NEO-FFI using Canadian female students, reporting low loadings of ≥0.30 in 55 of 60 items. Various studies further reported some items not loading highly on their corresponding component when using factor analyses (Rolland et al., 1998; Egan et al., 2000; Aluja et al., 2005).

As a solution to the loading concerns, McCrae and Costa (2004) replaced 14 items of the original NEO-FFI with newer items taken from the NEO-PI-R in order to improve the instruments factor structure. The selection criteria were: (1) minimized effects of acquisition responding, (2) increased NEO-PI-R factor score correlation, and (3) diversification of item content in favor of underrepresented facets of remaining items of the scales. The newest version of the NEO-FFI showed correlations from 0.56 (O) to 0.62 (N) in self report adjective factors, 0.39 (C) to 0.53 (O) in NEO-PI factors in spouse and 0.34 (C) to 0.59 (O) in peer ratings (Costa and McCrae, 2008). Convergent correlation ranges from 0.34 to 0.62. The newest version of the NEO-FFI scales account for about 75–85% variance of the original NEO-PI factors for convergent criteria. However, remaining concerns about the loading were once more highlighted by Aluja et al. (2005) who compared the old and revised NEO-FFI version in a Swiss (n = 1,090) and Spanish (n = 1,006) sample with unsatisfying results, observing that 10 items did not fit into a perfect five-factor structure because of loadings being lower than 0.30.

Although the NEO-FFI has been described as a quick and effective inventory to measure the FFM and may be the preferable choice in time-scarce populations, the data on populations such as high-performance sport is insufficient. Practitioners and researchers alike would benefit from not having to rely on generalizations sourced from normative data outside sporting populations. Despite the wide range of studies using the NEO-FFI, there are less studies published that have specifically focused on physically active subjects (Allen et al., 2013; Wilson and Dishman, 2015; Piepiora, 2020). Moreover, no studies currently exist that target a large sample of athletes in elite-level populations, and specifically in soccer. Therefore, the current study aimed to examine the internal consistency, test-retest reliability, and trait- respective item-level analysis of the factor structure of the NEO-FFI on a sample of German national and international high-level soccer athletes. Additionally, a second aim was to analyse differences between male and female athletes in responding style and traits. It is hypothesized that the NEO-FFI will demonstrate to be reliable and suitable measure of the FFM in high-level athletes, consistent with the literature that has used this inventory in non-sporting populations.

In settings with limited time access to participants, unable or unwilling compliance for long assessments, short and reliable instruments are necessary to measure relevant information for both research and practice. In the area of sport psychological diagnostics, the NEO-FFI delivers the possibility of rapid screenings to assess the big-five personality traits. In high-level sports and, respectively, soccer, researchers, coaches, and sports psychologists alike require information about quality criteria of the employed instrument such as internal and test-retest reliability, factor structure, and stability in their specific field to avoid over-generalizing or misinterpreting results based on non-comparable populations. As research of the FFM in high-performance athletes is lacking, the present study aimed to fill this gap and focus on the suitability of the NEO-FFI to measure personality traits specifically within elite-level soccer athletes.

Analysis of internal consistency across all athletes showed similar values in comparison to other results outside sport in the English (Caruso, 2000), French (Rolland et al., 1998) and German (Borkenau and Ostendorf, 2008) versions: N and C had the highest Cronbach's alpha levels (0.83 and 0.83, respectively), whereas E, O, and A all shared similar albeit lower alpha levels (0.69, 0.68, and 0.68, respectively). More specifically, females had higher internal consistency outcomes across each trait than males. This is in line with the large scale by Caruso (2000) who combined 51 samples in which the NEO Instruments PI, PI-R and FFI were combined, also reporting that females displayed higher alpha levels for every personality trait. Our finding could be explained by higher preciseness during the answering process in the female group, which we perceived a lot in the assessments. Opposingly, not all research reports gender differences for internal consistencies in the NEO questionnaires. For example, some studies reported reliability measures to be similar across genders (Egan et al., 2000; Borkenau and Ostendorf, 2008). Together, the analogous findings of the components from the NEO-FFI in high performance sporting populations with other general populations supports the transfer and use of the NEO-FFI into professional soccer.

Additionally, results of test-retest reliability are again in a similar range to other studies, although the present study used a different interval between the assessments. Studies with longer intervals like two (Borkenau and Ostendorf, 1993) or four (Robins et al., 2001) years found lower reliabilities, and shorter two-week intervals found comparable results (Robins et al., 2001). A 6-week interval was chosen to reduce impacts of item or answer remembrance (which maybe occur in a 2-week interval) and have a more realistic view of stability and reflect true change with a minimized measurement error (Becker, 2000; Schmidt et al., 2003; Watson, 2004). The findings of the current study show a high level of robustness, without biases of different occasions separated by an interval where no rapid personality changes could be expected. It must also be noted that most of the long interval studies mentioned above are not made with the intention of giving a view into potential applicable instruments in certain setting; rather they focused on longitudinal changes in personality traits. That leads to a lack of information concerning studies with a theoretical background and a specific aim for test-retest data.

In the current study, intercorrelations for the total sample appeared largely in line with similar studies, but with slight differences. For instance, in comparison to Egan et al. (2000) and Borkenau and Ostendorf (2008), the present study reported intercorrelations between all traits except for a positive rather than a negative association between O and C. Furthermore, apart from the correlations between N and E (for both the total and male sample only) and N and C (for female sample only), all correlation coefficients were higher in the present study than what is reported by Borkenau and Ostendorf (2008), whereas only four out of ten coefficients were higher in comparison to Egan et al. (2000).

Even if the non-orthogonality of the factors suggests an oblique rotation, an oblimin rotation was also conducted to determine differences in rotation-methods and have a comparison to previous studies. As expected, the oblique rotation showed better but not exceeding results than the oblimin. Only N and C scales appear to homogenously measure the traits as they should. Yet E, O, and A show more heterogeneity and variance amongst the factors. This is in line with previous findings, where also N and C show the best homogeneity (Egan et al., 2000; Aluja et al., 2005). The current study also replicated the pattern of weak and, respectively, missing loading of items on their intended factor (Egan et al., 2000; McCrae and Costa, 2004; Aluja et al., 2005; Borkenau and Ostendorf, 2008). These aspects of heterogeneous loadings may be attributed to the intercorrelations between the scales and their classification in two higher order factors. Principle component analysis (PCA) confirmed the consensus in literature, that the big five can be assigned to two higher order factors (Digman, 1997; Markon et al., 2005; Chang et al., 2012). PCA of the 60 items and the five-factor solution showed similar explanations of variances as previous studies (Egan et al., 2000; Aluja et al., 2005; Borkenau and Ostendorf, 2008).

A limitation of our study is linked to the misunderstanding of terms. We had several cases where athletes asked for the meanings of different words like “depressed,” “abstract,” “poetry,” or whole statements like “I believe letting students hear controversial speakers can only confuse and mislead them.” The origin of those misunderstanding problems lie in the population, respectively, samples which were chosen for development and evaluation of the NEO-FFI (Bodner, 2006). Those samples mostly exist of well-educated subjects and might not reflect the average type of elite-level soccer athletes. This case leads to difficulties when such questionnaires get applied without an immediately available consultant. As many instruments nowadays are applied using online software that can be answered from everywhere, future instruments should aim to prevent abstract and difficult expressions to understand. A second limitation lies in the biases like social desirability or role thinking when answering these questions. Athletes may often report what they believe is the right answer within a sporting context despite the instrument being a measure of non-sporting specific questions. Such mindsets may alter the way in which they answer the questions, as personality characteristics may be contextually dependent (i.e., different on and off the field). For instance, a team-captain get asked about leadership, and he/she may immediately think about their role in the team, despite them being not highly into leadership from a trait point of view. Additionally, the sample size could be one limitational aspect that influences the divergent factor loadings in our and similar studies with smaller or specific niche samples.

The current study implemented the NEO-FFI to measure the personality traits of a large sample of high-level soccer athletes and to examine the suitability of the use of NEO-FFI as a measure of the FFM for elite soccer players. The results demonstrated that the NEO-FFI had similar findings for (test-retest-) reliability, factor structure and stability in the elite-level soccer environment as previously reported in various other general populations. This study supports the use of the NEO-FFI as a time-efficient and reliable personality instrument that can inform staff, players, and researchers alike on the unique personality characteristics of each athlete. It would be beneficial for more studies to continue to investigate the NEO-FFI in various other high-performance sports in order to better generalize the findings of this study.

The datasets presented in this article are not readily available because the dataset includes information about national and international elite-athletes. Therefore, the dataset is under restrictions. Requests to access the datasets should be directed to jan.spielmann@tsg-researchlab.de.

The studies involving human participants were reviewed and approved by Universität des Saarlandes Ethikkommission der Fakultät HW Campus A1 3 66123 Saarbrücken Ethics Approval Number: 19/19. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.

JS: conceptualization, methodology, formal analysis, investigation, writing—original draft, writing—review & editing, and project administration. AB: writing—original draft and writing—review & editing. JM: conceptualization, resources, and supervision. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.