The ecological relevance of behavioural syndromes is little studied in cooperative breeding systems where it is assumed that the behavioural type might influence individual decisions on helping and dispersal (e.g. shy, nonaggressive and nonexplorative individuals remain philopatric and helpful, whereas bold, aggressive, explorative individuals compete for vacancies outside their group and disperse). We measured the behavioural type of 19 subordinates in the cooperatively breeding cichlid fish Neolamprologus pulcher in their natural environment by quantifying six behavioural traits up to four times ('trials') in three different contexts, by presenting them with a conspecific intruder, a predator or nothing inside a tube. We found only moderate within-context repeatability (intraclass correlation coefficients) of the focal individual's behaviour, except for attacking either the conspecific or the predator inside the tube. The focal individual's attack rate of the tube was also positively affected by its group size. Averaging traits per context removed the between-trial variation, and consequently the across-context repeatability was very high for all six traits, except for territory maintenance. Trait values depended significantly on the context, except for territory defence. Consequently, individuals could be classified into different behavioural types based on their reaction towards the tube, but surprisingly, and opposite to laboratory studies in this species, ranging propensity and territory maintenance were not included in this behavioural syndrome. We suggest that more studies are needed to compare standardized focal personality tests (e.g. exploration propensity) with actual behaviour observed in nature (e.g. ranging and dispersal).