A number of models of lexical access were developed in the past years. The major disagreement between models focuses on the issue of how lexical information is accessed. Does it use a strategy involving decomposing of elements of speech, or are they processed as a unit? No single model has yet been devised to settle such disagreement because both decomposed morphemes and whole words appear to exist in the access to word representation. The conflicting evidence in the literature suggests that hidden factors influence the organization of access representation. Surface frequency and transparency are two commonly mentioned factors (see a review by Sandra, 1994). Three experiments were designed to test for such factors in Chinese two-morpheme words. Results confirmed the presence of two factors. However, both interactive-activation models (Taft, 1994) and AAM (augmented addressed morpheme) models (Caramazza, Laudanna, & Romani, 1988) as well as other empirical research failed to account for our results. So, an IIC (Inter/Intra Connection) model was proposed and adapted to the problem of word processing.