Equivalence class formation as a function of the pronounceability of the sample stimulus