This paper presents a concise review of the development of limited capacity metaphors for explaining performance limits. Capacity views have been mainly inspired by limited capacity computers of the sixties and seventies, which are now replaced by parallel distributed processors with a virtually unlimited capacity. This is among the reasons that the view is no longer popular, while, in addition, the predictions of the notion of a simple unidimensional pool of resources have failed to be confirmed. The alternative assumption of multiple resources runs the risk of becoming too loose a set of faculties without internal coherence. There is a similar problem, with Fleishman's mental abilities approach which is based on factor-analysis. Yet, this approach has had some success as a taxonomy for complex tasks and deserves, therefore, more close scrutiny. At present, the theoretical emphasis is on connectionist notions of confusion and interference, which are more in line with stage-like views on information processing. However, so far, they suffer from unsufficiently detailed constraints to predict performance in concrete dual tasks. This may be the reason that, as yet, they have not led to much new experimental research.