Inspired by a dynamic approach to recognition memory (Cox & Shiffrin, 2012), we present results from a recognition memory experiment in which the time at which diagnostic information arrives is unconsciously varied. Contrary to the predictions of most models, performance improves when diagnostic information is available later, rather than earlier. These results are accounted for by a dynamic model of recognition, where the time at which information starts to be accumulated for a recognition decision can vary independently of when features are available to be sampled from the test display. The same model is shown to be able to reproduce the priming results of Jacoby & Whitehouse (1989), originally attributed to a fluency heuristic. The ability to account for such seemingly disparate results with a single model illustrates the utility of a dynamic approach to recognition.