Abstract: | Two experiments with a total of 104 undergraduates tested chunk frequency explanations of artificial grammar learning, which hold that classification performance is dependent on some metric derived from the frequency with which certain features occur within the letter string stimuli. Exp 1 revealed that classification performance was affected by close graphemic similarity between specific training (e.g., MXRVXT) and test strings (e.g., MXRMXT), despite the fact that similar strings did not contain frequently occurring features (e.g., bigrams or trigrams). This effect was replicated in Exp 2a, and Exp 2b demonstrated that substituting letters to make the consonant strings pronounceable (e.g., substituting X, R, and T, in the consonant string MXRMXT with Y, A, I, to produce MYAMYI) affected classification performance, despite the fact that objective measures of feature frequency were not altered. It is argued that models of classification that focus entirely on the frequency of features within the literal stimulus are insufficient, and that some allowance must be made for how the stimulus is encoded. (PsycINFO Database Record (c) 2010 APA, all rights reserved) |