Abstract
M.Com. (Industrial Psychology)
This study was conducted to examine the dimensionality of the Neuroticism scale of the
Basic Traits Inventory in order to contribute to the larger project of developing computer
adaptive testing for the five scales of the Basic Traits Inventory. The Neuroticism scale
comprises the general factor of Neuroticism and four group factors or facets, namely,
Affective Instability, Depression, Self-Consciousness, and Anxiety (Taylor & De Bruin,
2006; Taylor & De Bruin, 2013).
Personality constructs and scales are complex as they typically comprise heterogeneous items
for reliability and validity purposes (Reise, Morizot, & Hays, 2007). It is therefore to be
expected that some studies present evidence of a scale that measures an overarching single
dimension – unidimensionality – whilst other studies find evidence of multiple dimensions –
multidimensionality (Gibbons et al., 2007; Immekus & Imbrie, 2008; Reise, Moore, &
Haviland, 2010). Consequently, specification of a strictly unidimensional model is often not
realistic for complex personality constructs and scales. However, test-developers and
researchers aspire to unidimensionality (Zinbarg, Yovel, Revelle, & McDonald, 2006) as
unidimensionality allows for the unambiguous interpretation of scale scores (Reise et al.,
2010), is considered a precondition to reliability and validity, encourages the use of a total
score (Netemeyer, Bearden, & Sharma, 2003), meets the unidimensional assumption of item
response theory models, and is the prerequisite for the development of computer adaptive
testing (Thomas, 1990; Weiss & Gibbons, 2007).
Against this background, the key question that was posed in this study was not whether the
Neuroticism scale was strictly unidimensional or not, but rather if the scale was
unidimensional enough to allow for the meaningful interpretation of a total score (Morizot,
Ainsworth, & Reise, 2009). This study focussed on attaining evidence of the presence of a
strong general factor which would justify the interpretation of a total score and for applying
unidimensional item response theory models to the scale (Reise, 2012).
The Basic Traits Inventory was completed by 1 966 participants from various industries and
organisations within South Africa. Statistical analyses were performed on the responses to the
Neuroticism scale. Confirmatory factor analysis was used to investigate the
unidimensionality and strength of the general factor of the scale, while evaluating the
influence of multidimensionality on the interpretation of the general factor or the total score.
This dimensionality investigation of the scale was achieved by comparing three confirmatory
factor models using the goodness-of-fit indices and the standardised factor loadings of the
models. The competing structural models were: a one-factor or unidimensional model which
represented the general factor and interpretation of a total score for the scale; a four-factor or
multidimensional model which represented the multiple factors of the scale and therefore the
interpretation of the group factors or facets; and a bifactor model which allowed for the
examination of the strength of the general factor in comparison to that of the group factors of
the scale.
The results of this study confirmed that the Neuroticism scale is multidimensional as the scale
comprises multiple group factors or facets. However, the findings of the study revealed the
strong presence of a general factor that runs through all the items of the scale, and
consequently suggest that efforts to make use of and interpret separate subscale scores lack
the needed empirical evidence to do so. In essence, the findings of the study demonstrated
that the presence of unidimensionality in the scale is prominent despite the
multidimensionality of the scale, which confirmed that the interpretation of the total score for
the scale is justified and preferred irrespective of the multidimensionality of the scale.
Overall, the findings and evidence of this study proposed that the Neuroticism scale is
unidimensional enough to meet the unidimensional assumption of item response theory
models. Accordingly, these results support the advancement of adapting the Neuroticism
scale of the Basic Traits Inventory for computer adaptive testing (Reise, 2012; Thomas, 1990;
Weiss & Gibbons, 2007).
This study further confirmed that when developing a multifaceted scale, it is essential to
analyse the dimensionality of the scale in order to determine what the scale is actually
measuring. The bifactor model can be employed in the development, exploration and
confirmation of the dimensionality of the scale to examine the strength of the general factor
relative to the multiple group factors of the scale.