The inadvertent use of prior knowledge in a generative cognitive task

In four experiments with 332 participants, participants were asked to generate novel nonwords for English categories. When participants were shown examples embedded with regular orthographic structures, participants' nonwords tended to conform orthographically to the examples, despite instructions to avoid using features of the examples. The effect was found with immediate testing (Experiments 1) and delayed testing (Experiment 2). The effect was also found with arbitrary features (Experiments 1-4), as well as with naturally occurring orthographic regularities (Experiment 4). Participants had difficulty avoiding the use of this prior knowledge, despite being able to list the features they were asked to avoid (Experiment 3). The results are discussed in terms of the inadvertent use of prior knowledge in generative cognitive tasks.