Previous multinomial models that aimed at separating storage and retrieval processes typically focused on associative memory for (semantic) word pairs (e.g., knife – fork). However, we are interested in storage and retrieval for individual items, so for single words that are not necessarily clusterable and are therefore less likely to induce associative strategy use. To this end, we developed a free-recall-then-recognition paradigm and multinomial processing tree model with parameters that represent storage, retrieval in free recall, retrieval in recognition, and guessing. Experiment 1 revealed interesting effects of test-format expectancy. Participants who expected a free recall test (n = 38) showed better storage and recall retrieval than those who expected a recognition test (n = 38). While this pattern reversed for the guessing parameter, the recognition retrieval parameter was, somewhat surprisingly, not affected. Before further interpreting these effects, it is necessary to systematically validate all parameters to understand whether they correspond to our hypothesized memory processes. In Experiments 2 and 3, which are currently underway, we aim at selectively influencing the guessing parameter by manipulating the base rates of distractor items in the recognition test and the recall retrieval parameter by cueing participants with the first letters of the study words. The results of these experiments will be discussed with respect to the validity of the proposed measurement model.