This study partially replicates that of Braidi (2002). Like the original, it investigates occurrences and use of recasts, that is, reformulations of incorrect utterances, in relation to the number of grammatical errors in conversations of NSEs (Native Speakers of English) and NNSEs (Nonnative Speakers of English) and examines whether different types of tasks can affect the occurrences and use of recasts. In contrast to many previous studies, I examine here the occurrences of recasts in non-instructional interactions between NSEs and NNSEs.;Five dyads of American teaching assistants and Korean graduate students at Iowa State University participated in performing two different types of tasks, a picture description and a spot-the-difference task. Their performances were audio/video recorded and transcribed with Transana 2.0, a transcription software program.;For data analysis, frequency of recasts was counted and NNSE responses to recasts were also counted, according to the classification of Braidi (2002) and a chi-square test and a Fisher's exact test were used to find statistical significance. The results showed similarities to those of Braidi's (2002) study in that recasts occurred less than non-recasts in the frequencies of recasts predicted by the level of grammaticality. However, in contrast to Braidi's result showing a significant relationship between the level of grammaticality and occurrences of recasts, this study showed non-significant relationship between them. Furthermore, it appears that task types did not affect the occurrence of recasts. However, the study reveals considerable variation between dyads regarding both the occurrences of recasts and NNSE responses to recasts.