The benefit for spaced compared to massed presentation of to-be-learned items on delayed post-tests, known as the spacing effect, is one of the oldest findings in cognitive psychology. However, despite the robustness of findings in studies investigating distributed practice with paired-associate learning, such as the rote-learning of L2 vocabulary, the findings for studies that have investigated L2 learning under incidental learning conditions are more mixed. Over two studies, I investigated aspects of the temporal distribution of the presentation of L2 grammar and vocabulary when learned under two different incidental learning paradigms. In the first study, I investigated the role that distributed and massed practice play in the learning of an artificial language with nouns, verbs, adjectives and case markers, bound by a verb-final word order under incidental cross-situational learning conditions, and the role that five individual differences in memory (visual and verbal declarative memory, procedural memory, working memory capacity and phonological short-term memory) affected learning and retention. Results from study 1 showed that there was no significant difference in delayed post-test results between massed and distributed practice schedules. However, results suggest that lags may result in a shift in attention to different aspects of the language (from verbs to nouns) for those with strong declarative memory. Building on these findings, in study 2 I investigated whether several factors (intentional vs incidental learning conditions; items that were presented in training vs items that require a generalisation of rules; and declarative memory) influence the optimal lag for a 35-day retention interval when learning form-meaning connections (animacy and distance) of four artificial determiners. Results of study 2 mirrored study 1 in that, under incidental conditions, there was no difference between massed and distributed schedules. For the intentional aspect of the form-meaning connection, distributed practice schedules outperformed massed, with no one optimal lag.