The impact of language on personality assessment with the Basic Traits Inventory

Abstract

Personality psychology became an identifiable discipline in the social sciences in the 1930s when Allport (1937) published an article on the psychological interpretation of personality. The field of personality traditionally emphasised the study of the whole person, the dynamics of human motivation and the identification and measurement of individual differences (McAdams, 1997). Since the publication of Allport’s article, personality has been extensively researched and several theories exist that attempt to organise and explain the differences in human behaviour. Personality instruments are based on personality theories and aim to assist psychologists with the prediction of human behaviour. Psychologists use personality instruments as part of a selection battery to assist organisations with the screening and selection of individuals who have the potential to be successful within a specific work environment. The Basic Traits Inventory (BTI), a personality instrument that is based on the Five-Factor model, was developed in South Africa by Taylor and De Bruin (2006). The BTI is a valid and reliable personality instrument as indicated by results from research by Taylor (2004), Taylor and De Bruin (2006) and Taylor (2008) on its utility within the multicultural and multilingual environment of South Africa. Taylor (2008) nevertheless identified some problematic items when she analysed the construct, item, and response bias of the BTI across cultures for three language groups – Afrikaans, English and indigenous African languages – and consequently indicated the need for further research in this regard. The current study therefore explores the possible bias of the items of the BTI, and uses a sample large enough to analyse each of the eleven of the official languages of South Africa separately. The study focuses on the impact of the eleven official languages of South Africa on assessment of the Big Five personality factors with the BTI. The actual sample consisted of 105 342 respondents, resulting in the sub-samples per official language group being larger than 1 000. Each of the eleven official South African language groups could therefore be analysed separately, which has not been possible in previous studies. Analysis of the responses of the total sample to the BTI items generally yielded high reliability in terms of Cronbach alpha coefficients (α) and the Person Separation Index (PSI). The results were reported as follows: Extraversion (α=.86; PSI=.85); Neuroticism (α=.89; PSI=.86); Conscientiousness (α=.93; PSI=.88); Openness to experience (α=.90; PSI=.84); Agreeableness (α=.94; PSI=.86); and Social desirability (α=.72; PSI=.70). MANOVA results indicated statistically significant differences between the mean values of each of the BTI factors for the different language groups. Rasch analysis methods were used to further analyse the differences in terms of item responses for each of the eleven official language groups in South Africa. Respondents generally interpret and endorse the items of a personality instrument according to their intrinsic personality characteristics and their interpretation of the words used in the items of the personality instrument. In order to assess the respondents’ understanding of the administration language, English, two English proficiency tests were administered together with the BTI. The combined English proficiency scores were used to differentiate between respondents who understood English very well (top 25% – high English proficiency group) and those who struggled to understand English terminology (bottom 25% – low English proficiency group). Rasch analysis techniques were used to analyse the data for the whole sample as well as for the high and low English proficiency groups. Some items showed statistically significant differences for the language groups, indicating item bias in the BTI. Contrary to expectation, a larger number of biased items were indicated for the higher English proficiency group than for the total group or for the low English proficiency group. Due to the number of biased items for the high English proficiency group, it was concluded that the differences between the eleven official language groups may be a result of the differences in the intrinsic personality characteristics of the respondents, rather than measurement errors or item bias of the BTI instrument. Further research in this regard was recommended. The current study confirms that home language and English proficiency, as indicators of the level of understanding of the language in which the personality instrument was administered, undeniably influence the individual’s response pattern. Far fewer items indicated bias than those identified in Taylor’s (2008) study, where a smaller sample was used and language groups were combined. Despite some BTI items showing bias, the conclusion was reached that this personality instrument can be used with confidence to assess personality traits in persons speaking any of the eleven official South African languages.