Consistent individual variation in behaviour (often referred to as ‘temperament’ or ‘personality’) has increasingly been shown to be relevant to understanding patterns of behaviour that may influence survival and reproductive success. Although observer ratings on expert-nominated behavioural descriptors have been successfully used to identify and measure temperament traits in non-human primates and other species, some researchers argue that these measurements may be contaminated by observer bias both in the nomination of descriptors and in ratings of individual animals by the observers. Instead, it has been suggested that a better means of measuring temperament relies on a bottom-up approach based on direct, quantitative measurements of a species’ behavioural repertoire. In this study, we compare the utility of a trait-nomination-and-ratings approach and an approach using direct behavioural observation for identifying and describing the patterning of temperament traits a captive population of cockatiels (Nymphicus hollandicus). Our data show that the nomination-based and the behavioural repertoire-based approaches yield very similar pictures of the patterning of temperament traits in cockatiels, although trait scales generated using the nomination-and-ratings approach have higher internal consistency. Additionally, trait scores on both the behavioural repertoire-based trait scales and the nomination/ratings-based trait scales show reasonable power to predict observed behaviour, consistent with findings in both humans and non-human animals. The nomination/ratings approach seems to be more powerful in predicting birds’ behaviour towards novel objects. While further validation is certainly needed, the present results suggest that both methods may have some utility for describing temperament in cockatiels.