Habitual behavior is characterized by responses elicited by stimuli without deliberation or reliance on the predicted value of the outcome. We developed a reinforcement-learning task based on computational models suggesting that habit strength is proportional to the frequency of the behavior. We analyzed data of 220 participants who practiced our task on each of five consecutive days, with a test phase on the fifth day. Linear mixed-effects models revealed that choice frequency during training significantly increased choice probability and decreased response times during the test session (Nebe et al., under review). Recently, a competing model was formulated that could explain the results by an effect of recency instead (Giesen et al., 2020). In a secondary analysis of this dataset, we tested effects of recency against this effect of frequency. Additional mixed-effects models revealed statistically significant effects of recency on choice and reaction times: participants responded faster and had a higher probability of choosing a stimulus if they chose it at its last occurrence, but neither stimulus location nor number of trials since its last occurrence were meaningfully associated with choice or reaction times. Combining the effects of recency and frequency in the same model did not change the results of separate models of either effect. Thus, both choice frequency during training and choice at its last occurrence during testing influenced choice and response times in the same direction, but surprisingly these effects were uncorrelated and did not interact. These results suggest separate effects of past frequency and recency.