The optimal timing of feedback on formative assessments is an open question, with the cognitive processing window theory (Doughty, 2001) underlying the interaction approach suggesting that immediate feedback may be most beneficial for language acquisition (e.g., Gass, 2010; Polio, 2012) and two educational psychology hypotheses conversely suggesting that delayed feedback may be superior for error correction (dual-trace hypothesis, Kulik & Kulik, 1988; interference-perseveration hypothesis, Kulhavy & Anderson, 1972). To explore the effects of varied feedback timing on both item learning and rule generalization, 118 intermediate ESL students were randomly assigned to item-by-item or end-of-test computerized feedback conditions. Within each timing group, half of the students received feedback that indicated the correct answer and whether they had answered correctly or incorrectly (without metalinguistic feedback). The other students received additional feedback that stated a rule that applied to the item (metalinguistic feedback). A pretest, two treatments, a 5-minute-delayed posttest, and a 1-week-delayed posttest were administered. Each treatment contained 17 multiple-choice items that were followed by item-by-item or end-of-test feedback. The pretest and both posttests included all items from the treatment (to test item learning) plus 10 new multiple-choice items to test generalization of rules. The data were analyzed using mixed-design ANOVAs.The item-by-item metalinguistic feedback group had higher gain scores than the other feedback groups on the treatment items on both posttests, although no significant main effects were found for either feedback timing or type. This suggests that item-by-item metalinguistic feedback is better for item learning. On the items that did not appear on the treatment, the item-by-item groups outperformed the end-of-test groups, with a marginally significant main effect of feedback timing, F(1, 108) = 3.61, p = .06, η2part = .032. This suggests that item-by-item feedback may be better for learning to generalize. In addition, the groups that received item-by-item feedback spent significantly less time reading the feedback than did the groups who received end-of-test feedback, F(1, 108) = 4.14, p = .044, η2part = .037. These combined results suggest that item-by-item metalinguistic feedback may be more effective and efficient for language learners for both item learning and learning to generalize, although the small effects sizes indicate that providing this type and timing of feedback should be only one of many interventions to improve instruction. In addition, these results lend support to the cognitive processing window theory and attention-based theory underlying the interaction approach.
Read
