Pm. Wall et C. Messier, Ethological confirmatory factor analysis of anxiety-like behaviour in the murine elevated plus-maze, BEH BRA RES, 114(1-2), 2000, pp. 199-212
The elevated plus-maze has been used in animal research to measure anxiety
since 1985 and is currently the most widely used animal model of anxiety. S
ince this paradigm has been the subject of several principal components ana
lyses, it is well qualified for confirmatory factor analysis research. The
current report builds on the substantial theoretical knowledge and empirica
l data obtained from these structural analyses with a view to obtain furthe
r progress in the evolution of our understanding of animal anxiety in the e
levated plus-maze. The purpose of the present report was two-fold: (a) to t
est if the a piori imposition of a 3-factor model, or a competing 2-factor
elevated plus-maze model, would fit our sample (n = 200 CD-1 mice) data in
each of two trials within an inferential confirmatory factor analytic frame
work; (b) provide a well-fitting model that confers indicator variables tha
t can most effectively and parsimoniously measure underlying constructs of
elevated plus-maze behaviour. Multiple model-fitting criteria were used, an
d issues related to data non-normality, outliers, replicability of the mode
l, sampling error and error of approximation ill the estimation of final mo
del fit were addressed. The final 2-factor model, with estimated error cova
riance between two different pairs of indicator variables, was a good fit o
n the trial-1 data, although it was necessary to allow unprotected stretch
attends to non-significantly cross-load on factor-2. A 2-factor model also
fit the trial-2 data from the present analysis, although it was necessary t
o allow closed arm time ratio to negatively cross-load on factor-1. These r
esults indicate that inferential hypothesis testing and model building proc
edures within a confirmatory factor analysis framework produces interpretab
le animal anxiety indices in the elevated plus-maze. Moreover, a 2-factor,
rather than a 3-factor model, parsimoniously and unambiguously explained th
e underlying constructs of anxiety-like mouse behaviour in the elevated plu
s-maze in the present study. Taken together, a reduction in the growing num
ber of behavioural indices reported in elevated plus-maze pharmacological s
tudies is suggested. (C) 2000 Elsevier Science B.V. All rights reserved.