The purpose of this exploratory longitudinal study was to evaluate the efficacy of two new forms of recasts (i.e., elaborated and paraphrased recasts), each of which was designed to be more in accordance with contested views of input processing. The effectiveness of the two new forms of recasts was compared to that of conventional standard recasts. To this end, an experienced language teacher engaged four intermediate-level learners of English as a foreign language (EFL) in weekly, one-on-one, English conversation sessions under four instructional conditions (elaborated recast, paraphrased recast, standard recast, and no recast), one learner under each condition. A diachronic, time-series research design was used to assess the effects of recasts on each learner's performance of past tense over a nine-week, instructional period. The results showed that all three types of recasts proved effective in assisting the learners to improve their accuracy on English past tense; all things considered, elaborated and especially paraphrased recasts proved to have a distinct advantage over standard recasts. The results suggest that recasts, in general, and more specifically paraphrased and elaborated recasts, are effective when employed in less controlled, spoken conversational practice, where there is free turn-taking, topic shifts, and the conversational content is related to the learner's personal experiences and background knowledge.