Mild cognitive impairment (MCI) is characterized by subjective and objective memory impairment in the absence of dementia. MCI is a strong predictor for the development of Alzheimer’s disease (AD). A standard task used in the diagnosis of MCI is verbal fluency, where participants produce as many items from a specific category (e.g., animals) as possible. This task is typically analyzed by counting the number of items produced. However, analysis of the actual pattern of items produced can provide valuable additional information. Here we describe a computational model that uses multiple types of lexical information in conjunction with a standard memory search process. The model parameters were able to successfully differentiate changes in semantic memory processing for individuals who develop MCI, even when behavioral variables could not.