This article reports on two studies, testing how three different types of input (skewed first, skewed random and balanced) affect second language (L2) learning of English present counterfactual (IF-Is) and past counterfactual (IF-IIs) conditionals, two constructions differing in complexity. The experiment included a proficiency test, a pretest, a posttest and a delayed posttest for both IF-Is and IF-IIs. Results suggest that the three types of input did not affect learning gains within the two complexity conditions. However, there was a difference in learning gains between the two conditions, namely that more gains were made in the simpler condition (i.e. IF-Is). In a follow-up study, effects of intralingual influence were tested. Participants who had studied IF-Is were taught the more complex construction, IF-IIs. Again, there were three different types of input (skewed first, skewed random and balanced). Results suggest that participants who received skewed input had higher gains and were less influenced by IF-Is they had learned before. Based on this, we concluded that skewed first input was able to effectively reduce the influence of IF-Is on IF-IIs during L2 learning.