Objective Previous studies on cognitive deficits in acute and remitted states of old‐age depression have shown mixed findings. The episodic nature of depression makes repeated assessment of cognitive performance important in order to address reversibility and stability of cognitive deficits. Methods Dementia‐free older participants (≥60 years) from the population‐based Swedish National Study on Aging and Care in Kungsholmen who completed neuropsychological testing at baseline (T1) and follow‐up (T2) formed the basis of the study sample. Participants were grouped according to depression status at T1 and T2: depressed–remitted (n = 32), remitted–depressed (n = 45), and nondepressed–depressed (n = 29). These groups were compared with a group of randomly selected and matched (age, gender, education, and follow‐up time) healthy controls (n = 106) over a period of maximum 6 years. Results Mixed ANCOVAs, controlling for age and gender, revealed depression‐related deficits for processing speed, attention, executive function, and category fluency. In remitted states, only processing speed and attention were affected. However, these deficits were attenuated after exclusion of persons using benzodiazepine medications. A general pattern of cognitive decline was observed across all groups for processing speed, executive function, category fluency, and episodic and semantic memory; persons transitioning from a nondepressed to depressed state tended to show exacerbated cognitive decline. Conclusions The results support the notion that cognitive deficits in depression may be more transient than stable. Consequently, cognitive deficits in depression might be regarded as potential treatment targets rather than stable vulnerabilities. As such, repeated assessment of cognitive functioning may provide an additional marker of treatment response.