Abstract: I present results from two wug-tests of parts of the English stress system, which yield quite different patterns of productivity.  One experiment was a typical wug-test while the other involved a concurrent memory task. While the typical wug-test yielded probabilistic behavior matching statistical details of the English lexicon, when the memory task was introduced participants (1) diverged more from each other in behavior, (2) behaved more categorically as individuals, and (3) matched only broad generalizations about the English stress system as a whole, and did not match the same details they matched in the first experiment, either as individuals or in aggregate.  I discuss possible explanations for these apparent effects of the memory task on participants’ ability to use their English stress knowledge, and possible future directions for learning more.