The test-retest reliability of four behavioural tests of fearfulness for quail: a critical evaluation

Tests of animal fearfulness are being used in a broad range of basic and applied scientific fields. However, rigorous assessment of the test-retest reliability of these tests is relatively uncommon. The importance of such assessment is illustrated by this study of the responses of Japanese quail during four tests of fearfulness. Forty-six adult quail were subjected to an emergence test, novel object test, novel food test, and predator surprise test on two consecutive days to assess test-retest reliability. Multiple behavioural measures were made during each test. One test was performed per day. There were discrepancies in the reliabilities of the various measures made within each test. For example, latency to extend the head into the emergence arena was poorly correlated on two consecutive days (partial r = .19, n = 46, P = .29), but the number of times the birds extended their heads into the arena was much more reliable (partial r = .73, n = 46, P < .001). Similarly, latency to try a novel food was a relatively unreliable measure (partial r = .47, n = 46, P < .01) as compared to the amount of novel food eaten (partial r = .81, n = 46, P < .001). This study also revealed the value of using a partial correlation coefficient when examining the reliability of animal fearfulness, rather than the more commonly reported Pearson's coefficient. Factors such as sex and pen location inflated the latter but were partialled out by the former, leading to lower coefficients on most measures (e.g. for latency to orient towards a novel object on two consecutive days, partial r = .19, Pearson's r = .52). Finally, the statistically significant but small correlation coefficients for many measures in this study and other published fear studies, many less than .4, indicate that some common behavioural measures of fearfulness are not always reliable.