Contemporary zoos house polar bears (Ursus maritimus) to serve the conservation efforts of these institutions. However, evidence of behavioural problems in captivity has highlighted the need for methods to assess polar bear welfare, why valid and feasible animal-based welfare indicators are a prerequisite. Qualitative Behaviour Assessment (QBA) has been used to assess emotional state in various species. This study therefore aimed to assess the ability of QBA to discern emotional state in zoo-housed polar bears, and further to investigate its validity through associations to behavioural and health-related indicators, as well as intra- and inter-day consistency of QBA scores. A list of 24 terms was developed by reviewing the literature and on-site assessments. Reliability of the list was tested through Kendall's W between two assessors (Inter-OR), as well as over time (Intra-OR), based on videos. QBAs and behavioural observations were collected on 22 polar bears from nine zoos in four European countries, repeated within days (three times) and across days (four days). Behavioural observations were summarized as percentage of time spent in the respective categories, whereas behavioural diversity was expressed by the Shannon Index. Health-related indicators were expressed as an overall dichotomous value (presence/absence). Principal component analysis revealed two main components categorised as valence (PC1, 29.5%) and arousal (PC2, 15.1%). Reliability was met for both PCs, with a W of 0.82 for PC1 and 0.68 for PC2 for Inter-OR, and 0.92 for PC1 and 0.89 for PC2 for Intra-OR. PC1 was significantly positively associated with behavioural diversity (p = 0.006), environmental interaction (p = 0.006) and rest (p = 0.005), and inversely with stereotypic behaviour (p < 0.001) and general activity level (p = 0.003). PC2 was significantly positively associated with stereotypic behaviour (p < 0.001), environmental interaction (p = 0.040) and general activity level (p < 0.001), and inversely with awake inactivity (p = 0.002). Valence scores revealed no significant effect of time of day nor day, however a significant effect was found between early to mid-day on arousal scores (p = 0.015). The results provide evidence of sensitivity and some evidence of validity by associations of QBA with other welfare indicators. QBA may therefore potentially serve to assess emotional state in polar bears, needing only to be carried out once in the short term. Although the proposed list may benefit from additional validation and reliability testing, this study takes the first step towards a standardised QBA for polar bears.