The dominant model of personality structure amongst adults specifies that individual differences in personality attributes can be optimally described in terms of five factors. Whereas the labels and definitions of the traits varies somewhat across countries, instruments and authors, the five traits are commonly labelled as extraversion, neuroticism, conscientiousness, agreeableness and openness/intellect (De Raad, 2000). The five factors present a satisfactory balance between bandwidth and fidelity, which means that the factors provide an economical description of personality, yet allow for meaningful prediction of important outcomes (Ones & Viswesvaran, 1996), such as health, education and work performance (e.g. Cheng, Weiss, & Siegel, 2015; Judge & Zapata, 2015).

The Big Five structure has been shown to be replicable across different cultures amongst adults (Rolland, 2002). Personality psychologists have also demonstrated that the Big Five factors are also useful in personality description of adolescents (De Fruyt, Mervielde, Hoekstra, & Rolland, 2000; McCrae, Martin, & Costa, 2005; Mervielde & De Fruyt, 2002; Parker & Stumpf, 1998; Shiner & Caspi, 2003; Soto & Tackett, 2015). Indeed, the so-called Little Six model of personality in childhood and adolescence includes the Big Five traits with the addition of activity as a sixth trait (Soto & John, 2014; Soto & Tackett, 2015).

There has been some debate in the literature as to whether the Big Five model of personality is an adequate representation of the structure of personality in childhood and adolescence. Soto, John, Gosling and Potter (2008) postulated that the factor structures of personality measures ‘should be recovered less clearly in the responses of children and adolescents than in those of adults’ (p. 720). This is related to an increasing awareness of identity and differentiation of self-concept with age, which should lead to more clearly defined structures amongst adults. In this respect it appears that self-report ratings indeed become more consistent, and factors are better differentiated with an increase in age (Soto et al., 2008; Soto & John, 2014).

A related issue is the consistency with which persons respond to evaluations of their own behaviour and attributes and how this relates to age. In this respect Soto et al. (2008) suggested that older adolescents would likely respond more consistently to personality items than younger adolescents, which would manifest in better reliability coefficients of measures of the traits for older adolescents. They suggest that the reason for this is because older adolescents have more developed self-concepts and better ability to evaluate the issues of logical consistency when rating their own behaviour.

Soto, John, Gosling and Potter (2011), highlighted two principles with respect to youth personality development. The first principle states that that youth personality traits are organised hierarchically in a similar fashion to adult traits (Soto & John, 2014) that is higher-order traits (e.g. agreeability) subsuming narrow, lower-order ones (e.g. modesty and generosity). The cumulative-continuity principle states that changes in personality traits occur during the transition from childhood to adulthood and that traits reach their highest levels of stability in adulthood (Roberts & DelVecchio, 2000; Soto & Tackett, 2015).

The BTI (Taylor & De Bruin, 2006) was developed using a combined emic–etic approach to test development. From an etic perspective, the Big Five taxonomy was used to inform the five-factor structure of the inventory. From an emic perspective, the items were developed keeping the multilingual and multicultural South African context in mind. Two versions are available, namely an English and an Afrikaans version. The items are brief, require a low reading level and avoid cultural particularities. In other words, the development philosophy was that the items should be comprehensible for persons who complete it in a language other than their first language and that the items should include content that would be relevant to most adults in South Africa. For a more detailed description of the development of the BTI, please consult Taylor (2004, 2008) or Taylor and De Bruin (2006).

The BTI measures five factors, namely extraversion, neuroticism, conscientiousness, openness to experience and agreeableness. Each factor has a number of sub-factors (varying between four and five), called facets, which measure narrow aspects of the broader factors and provide potentially rich interpretive information (e.g. De Vine & Morgan, 2020). The five-factor structure of the BTI has been replicated with adults across gender groups, language groups and cultural groups in numerous studies (Ramsay et al., 2008; Taylor, 2008; Taylor & De Bruin, 2006). The reliability of the five-factor scores is consistently higher than 0.85 and most facets consistently demonstrate reliability coefficients of 0.70 and above (Taylor & De Bruin, 2006, 2013).

Against the background of the lack of personality inventories suitable for the use with adolescents and emerging evidence that the Big Five structure is replicable and provides adequate descriptions of personality amongst adolescents (Cheng et al., 2015; McCrae et al., 2005; Wu et al., 2008), the present study examined the factor structure and reliability of the BTI for South African adolescents. The BTI was deemed appropriate given the steps that were taken in its development to keep the content brief and simple, which could facilitate its use with adolescents. It is necessary to demonstrate the structural validity and reliability of the BTI with adolescents given the cautions that have been expressed regarding the differentiation and consistency of personality in childhood and adolescence

In particular, the present study examined (1) the similarity of the factor structure of the BTI, and (2) the reliability of the Big Five traits and their facets across younger adolescents (12–15 years), older adolescents (16–18 years) and the normative adult sample (18–72 years).

We used the psych package (Revelle, 2017) in R (R Core Team, 2016) to subject the correlations of the 24 BTI facets of the pooled adolescent data set to an unrestricted unweighted least squares factor analysis. We decided the number of factors to retain with reference to Velicer’s MAP test, the empirical Bayesian Information Criterion (eBIC), Horn’s parallel analysis, Cattell’s scree plot, the Root Mean Squared Residual (RMSEA) and the Standardised Root Mean Squared Residual (SRMR). The maximum a-posteriori (MAP) criterion and the eBIC reached their respective minima with five factors. Parallel analysis evidenced that only the first five roots had eigenvalues that exceeded those of random data, and the scree plot revealed a clear elbow in the plot of the eigenvalues after the fifth root. With five factors extracted, the RMSEA = 0.05 and the SRMR = 0.02 suggested satisfactory fit with the observed data. Against this background, and theoretical expectation, we retained five factors.

Next, we obtained separate unweighted least squares five-factor solutions for adolescents of 15 years and younger (labelled the younger adolescents) on the one hand and adolescents 16 year and older (labelled the older adolescents) on the other hand. For each group the factor solution was rotated to a target structure based on the theoretical structure of the BTI. The factors were allowed to freely correlate.

In both groups the obtained factor pattern matrices corresponded closely with the theoretical structure of the BTI.¹ On the basis of the pattern of high and low loadings of the 24 facets the factors were labelled as follows: Factor 1 = conscientiousness; Factor 2 = neuroticism; Factor 3 = openness to experience; Factor 4 = agreeableness and Factor 5 = extraversion.

Next, we examined the similarity of the empirically obtained factor pattern matrices of the two groups. Congruence coefficients (Tucker’s phi coefficient) of corresponding factors show that the factors manifested very similarly across the two groups of adolescents (conscientiousness, Φ = 0.98; neuroticism, Φ = 0.99; openness, Φ = 0.97; agreeableness, Φ = 0.98; and extraversion, Φ = 0.96; Table 1). These coefficients indicate that the corresponding factors of the younger and older adolescents can be considered similar (Lorenzo-Seva & Ten Berge, 2006).

We also compared the empirical target rotated factor pattern matrices of the two adolescent groups with the rotated factor pattern matrix of the adult standardisation sample (as reported in Table 8 of the BTI manual [Taylor & De Bruin, 2006]). For both adolescent groups each facet’s primary factor loading corresponded with the pattern of loadings observed for adults. The coefficients of congruence of the corresponding factors of the younger adolescents and adults were as follows: conscientiousness, Φ = 0.97; neuroticism, Φ = 0.99; openness, Φ = 0.97; agreeableness, Φ = 0.97; and extraversion, Φ = 0.96. In turn, the coefficients for the older adolescents and the adults were as follows: conscientiousness, Φ = 0.97; neuroticism, Φ = 0.98; openness, Φ = 0.97; agreeableness, Φ = 0.97; and extraversion, Φ = 0.97. These coefficients indicate that the corresponding factors of the adolescents and the working adults can be considered similar.

On the basis of the high levels of factor similarity across the younger and older adolescents we obtained a target rotated solution for the pooled adolescent data set. The factor pattern matrix is presented in Table 2, which shows that each facet had a salient loading (> 0.30) on its target factor. Two facets had cross-loadings that just exceeded the |0.30| criterion on non-target facets (i.e. ascendance on the conscientiousness factor (λ = 0.33) and excitement-seeking on the conscientiousness factor [λ = −0.33]). The factor correlations are given in Table 3, which shows small to medium sized correlations between the five factors. Somewhat lower correlations between the factors were observed for the adolescents compared with the adults.

The reliability coefficients (Cronbach’s alpha) of the five factors and the facets across the two groups and the working adult group are given in Table 4, which shows that the personality scales yielded scores with similar levels of measurement precision across the three groups. Across the five traits the reliability coefficients of the three groups were within |0.01| of each other, with coefficients ranging from 0.86 to 0.95 across the five traits. Overall, the reliability coefficients indicate a high level of measurement precision for each of the five scales.

Against the background of a dearth of suitable tools for the measurement of personality traits amongst adolescents, we examined the construct validity of the BTI amongst South African adolescents. As a whole the results support the replicability of the factor structure of the BTI (and therefore its construct validity) and indicate that the scales yield highly reliable scores for adolescents. In the paragraphs that follow we discuss these results in more detail.

Results show that the theoretical Big Five structure of the BTI was replicated in the empirical correlations of the 24 facets amongst younger (13–15 years) and older (16–18 years) adolescents. These results underline the robustness of the Big Five factors (i.e. extraversion, neuroticism, conscientiousness, openness for experience and agreeability) across age groups and support the construct validity of the BTI scales for adolescents. The Big Five factors manifested almost identically amongst the younger and older adolescents. These factors were also almost identical with the factors of the normative working adult group, which suggest that the adolescents and adults attached similar meaning to the content of the BTI items. McCrae et al. (2005) and Wu et al. (2008) similarly demonstrated that the factor structures of the NEO-PI-3 (McCrae et al., 2005) and NEO-PI-R (Wu et al., 2008), respectively, were replicable across adolescents and adults. The reliabilities of scores yielded by the Big Five scales were uniformly high across the two adolescent groups and similar to the reliabilities reported for adults by Taylor and De Bruin (2006). Hence, adolescents and adults responded with similar consistency to the items. These results suggest that the structure and coherence of personality, as reflected in responses to the BTI, might be established amongst adolescents as young as 13 years.

The replicable factor structure and high reliabilities of the five scales suggest that the BTI holds promise as a measure of the Big Five traits in adolescents, which opens possibilities for future personality research amongst this group in the South African context. Qualitative analysis revealed a small number of items (about 3%) with content that do not directly apply to adolescents, for example, items related to work, owning a house and contributing towards charity. The psychometric analyses suggested that the inclusion of these items do not detract from the measurement quality (i.e. the factor structure or reliability) of the scales, but it is necessary to revise these items for future applications.

We adopted a top-down approach, where a measure that was developed for adults is examined with respect to its utility for measuring personality traits in adolescents [(see De Fruyt et al., 2000; McCrae et al., 2005 for studies where a similar approach was adopted]. Whereas the evidence in support of the replicability of the BTI factor structure amongst adults support the construct validity of these factors, it does not necessarily mean that the particular set of BTI facets are optimal for the description of personality amongst adolescents. Indeed, it is possible that a bottom-up approach may yield a different set of facets as indicators of the broader Big Five traits. In this respect it is perhaps useful to emphasise that there is no such thing as ‘the correct set of facets’ for the Big Five factors. Ultimately, as long as the facets are proper indicators of the factors, it is the utility of the chosen facets that matters and in this respect studies that examine the predictive validity of the BTI scales with respect to educational, health and social outcomes represent a fruitful area of further research.

A second limitation is that the study focussed on structural similarity and reliability of the BTI for adolescents at the scale or trait level only. Whereas these results were supportive of the construct validity of the scales it is possible that some items may function less than optimally for adolescents. Further research should examine the quality of individual items when used with adolescents and whether the items function equivalently for younger and older adolescents.

The results indicate that the BTI, which was developed for adults, holds promise as a measure of the Big Five personality traits amongst adolescents. The BTI appears to be a potentially useful tool to track the development of personality in adolescence. In addition, practitioners who are interested in the role of personality traits in educational and career counselling with adolescents might fruitfully employ the BTI in these contexts. As a whole, these results add to the growing body of evidence that supports the validity and usefulness of the Big Five personality traits in South Africa and the validity of the BTI as a measure of these traits.