level higher, species-typical behaviours may not only arise through differences in the
regulation of different behaviours by internal regulating mechanisms such as personality
traits (trait manifestations), but also because the mechanisms themselves may differ due to
selection on these mechanisms. Whether the selection eventually acts on traits or on
behavioural measurements, remains largely unknown. This gap in our knowledge is
mainly caused by the lack of a systematic way of choosing behaviours and the absence of a
systematic trait generation strategy. An important limiting factor hereby is that studies
interested in the adaptive significance of variation in personality follow individuals in their
natural environment. The intensive observations needed for approaches like the
‘behavioural repertoire approach’ are therefore nearby impossible without taking
individuals into captivity for a longer time. ‘Knowing thy species’ is, however, an
important aspect in animal personality research, and the ‘behavioural repertoire approach’
is a good tool to achieve that knowledge.
Species of Nonhuman Personality Assessment
ALEXANDER WEISS
1
and MARK JAMES ADAMS
2
1
Department of Psychology, The University of Edinburgh, UK
2
Institute of Evolutionary Biology, The University of Edinburgh, UK
alex.weiss@ed.ac.uk
Abstract
The target paper sets out an ambitious approach to animal personality research. This
approach uses bottom-up measures and behavioural repertoires. This method holds
promise for understanding personality trait expression in a given species; however, it has
limited applications for comparing personality between species because it does not
specify a measurement of interspecific similarity or difference and does not consider
phylogenetic relationships between organisms. Our research led us to develop a modified
top-down trait approach which makes comparing species feasible, is sensitive to species-
level differences and is consistent with evolutionary thinking. Copyright # 2008 John
Wiley & Sons, Ltd.
You can’t eat the orange and throw the peel away—a man is not a piece of fruit! (Miller,
1994/1949, p. 59)
We are pleased to see new empirical studies and methodological innovations in the
study of animal personality. Methodological approaches from cross-cultural psychology
are relevant for animal personality research, and showing how personality is reflected in
responses to situations may clarify personality trait expression. However, we question the
utility of Uher’s method for comparative personality research because it neither provides a
common metric for determining similarities or differences between species nor considers
phylogenetic relationships between organisms.
472 Discussion
Copyright# 2008 John Wiley & Sons, Ltd. Eur. J. Pers. 22: 457–474 (2008)
DOI: 10.1002/per

Uher suggests that bottom-up approaches are superior because they involve ‘study[ing]
personality variation from inside the species, largely uninformed by findings from other
species’. While interesting and informative, this approach would not lend itself to
comparative personality research because it assumes little can be learned about the
personality of one species from studying the personality of another. As species are not
types (Mayr, 1985), disregarding other species ignores that species are related, structured
hierarchically and thus not independent (Felsenstein, 1985). So, while it is reasonable to
take an ‘agnostic’ view when investigating the personality of evolutionarily distant
species, for example squid (Sinn et al., 2006), this would be ill-advised when comparing
closely related species.
Incorporating situations is also problematic. By definition, personality traits are stable
across situations (Allport, 1961). In fact, Mischel (1968) criticised early personality
researchers for failing to demonstrate cross-situational stability. Also, no metric for
categorising behaviour-trait signatures as being more or less similar in meaning exists and
there is no way to measure the extent of differences. Finally, while subjective ratings
involve assessing individuals’ personality across a wide-range of situations over long
periods of time, studying behavioural responses to particular situations restricts this
natural variation. The result would therefore be a reduction in ecological validity.
A top-down approach to comparative personality research is not without limitations,
including the possibility of implicit personality theory or its ill-behaved pet,
anthropomorphism (see Gosling, 2001, for a discussion). While implicit personality
theory was a concern in human personality research (Passini & Norman, 1966), cross-trait/
cross-twin correlations have since shown that it is unlikely to explain human personality
structure (McCrae, Jang, Livesley, Riemann, & Angleitner, 2001; Rowe, 1982). There are
no animal twin registries, but, using Kenny’s social relations model, Kwan, Gosling, and
John (2008) found no evidence of anthropomorphic projection in ratings of dog
personality.
Like those studying personality across cultures (Church, 2001), Uher is wary that, if a
top-down approach is used, important traits may be missed. Before addressing this, we
wish to correct a misperception of research conducted by ourselves and our colleagues
(e.g. King, Weiss, & Farmer, 2005; Weiss, King, & Hopkins, 2007). We do not use a strict
top-down approach, but a modified version of the top-down approach that developed
(bottom-up) over time. This approach involves using as broad a set of indicators as
possible to determine the personality structure of a particular species. Because the Big
Five or Five-Factor Model (Goldberg, 1990; McCrae & Costa, 2003) have extensively
catalogued human personality traits, they are a natural starting point. However, because
trait expression differs among species, we include clarifying sentences to capture these
differences. For example, fearful in zoo chimpanzees and zoo orangutans is defined as
‘Subject reacts excessively to real or imagined threats by displaying behaviours such as
screaming, grimacing, running away or other signs of anxiety or distress’. (King &
Figueredo, 1997; Weiss, King, & Perkins, 2006). For horses, this same trait was defined as
‘Subject reacts excessively to real or imagined threats by displaying behaviours such as
side-stepping, shying, running away or other signs of anxiety or distress’. (Fairholm, in
preparation). In addition, for horses, it was deemed desirable to split the trait friendly into
friendly to humans and friendly to horses. (Fairholm, in preparation).
This approach is also reflected in our choice of analyses and how we interpret results.
We do not impose the Five-Factor Model or any other structure via confirmatory factor
analysis and instead use exploratory factor analyses such as principal components
Discussion 473
Copyright# 2008 John Wiley & Sons, Ltd. Eur. J. Pers. 22: 457–474 (2008)
DOI: 10.1002/per