The space gets warped bythe existence of an exception. The model predicts that exceptions will moor poorly generalize compared to regularized words. Live" slows reaction time when it comes to disambiguating and considering a pronunciation). In this example, pint warps the otherwise regularized grid. Kint rhymes with pint; apply it to fint (rhyme with pint: regularizations tested for nonword neighbours (probes = ex. Druff: results, anchors/probes: regulars are always generalized no matter condition. Anchors: before training: ambiguous terms 75% regularized, exceptions regularized, after training: ambiguous terms 25% generalized, very little regularized of exceptions (simulated human is learning them, pronunciation confined to small space not extended/generalized. Probes: before training: ambiguous regularized decreases slightly compared to anchors learned, after training: decreased to 50%, dramatic increase of regularization. Humans: multi-day experiment, monday: exposure repeat the following process twice, listen to 30-word anchors a few times, engage in active tasks, spelling to sound: ex.