Taiwanese voiced stops are followed by non-nasalized vowels, while nasals are followed by nasalized vowels. This study uses a concept formation paradigm to test whether native speakers treat the voiced stops and nasals as allophones or separate phonemes. There are three experiments in the study. In the training session of experiment A, monosyllabic morphemes are presented. Subjects receive a positive reinforcement when a syllable beginning with /b/ is presented. In the test session of experiment A, syllables beginning with /m/ are introduced. In experiments B and C disyllabic phrases are presented. In the training session of test B, the subjects are reinforced to treat /b/ and /m/ as separate categories. In the training session of test C, the subjects are reinforced to treat /b/ and /m/ as the same category. The results show that /m,b/ are grouped into the same category. In experiment B, subjects cannot differentiate between /b/ and /m/. In experiment C, when /b/ and /m/ are reinforced to be of the same category, the subjects perform well. According to the interview responses after each test, subjects describe the category that is being reinforced positively as nasals.