1. Experiments were carried out to test the effectiveness of mimicry at different proportions of models and mimics with Starlings (Sturnus vulgaris) as predators and mealworms (Tenebrio molitor) as artificial models and mimics. 2. A band of green cellulose paint was used to mark the "model" mealworms which were made unpalatable by being dipped into a 66 per cent solution of quinine dihydrochloride. "Mimic" mealworms were identically banded, but dipped into distilled water, as were orange-banded "edible" mealworms. 3. It was found that birds which received ten per cent, 30 per cent, and even 60 per cent mimics (the balance being models) rejected about 80 per cent of the mimics by associating their color band with that of the unpalatable model. Mimicry is thus effective when mimics are more frequent than models. 4. Two birds which were given 90 per cent mimics (only ten per cent models) nevertheless rejected 17 per cent of the mimics, indicating that when a model is highly unpalatable, and a mimic relatively "perfect," mimicry may be effective even when a mimic outnumbers its model by as much as nine to one. 5. Under the conditions of the experiment, the selective advantage of mimicry is far greater than believed by previous investigators.