The steroid hormone testosterone (T) plays a central role in the regulation of breeding in males, because many physiological, morphological and behavioural traits related to reproduction are T dependent. Moreover, in many seasonally breeding vertebrates, male plasma T levels typically show a pronounced peak during the breeding season. While such population-level patterns are fairly well worked out, the sources and the implications of the large variability in individual T levels within the seasonal cycle remain surprisingly little understood. Understanding the potential sources of individual variation in T levels is important for behavioural and evolutionary ecologists, for at least two reasons. First, in 'honest signalling' theory, T is hypothesized to play a critical role as the assumed factor that enforces honesty of the expression of sexually selected quality indicators. Second, T is often considered a key mediator of central life-history trade-offs, such as investment in survival versus reproduction or in mating versus parental care. Here, we discuss the patterns of within- and between-individual variation in male plasma T levels in free-living populations of birds. We argue that it is unclear whether this variability mainly reflects differences in underlying individual quality (intrinsic factors such as genetic or maternal effects) or in the environment (extrinsic factors including time of day, individual territorial status and past experience). Research in avian behavioural endocrinology has mainly focused on the effects of extrinsic factors, while other sources of variance are often ignored. We suggest that studies that use an integrative approach and investigate the relative importance of all potential sources of variation are essential for the interpretation of data on individual plasma T levels.