Investigation of the generation effect on memory and metamemory through semantic and perceptual cues

Abstract

The generation effect is an experimental finding that self-generated information produces higher memory performance than reading. The effect can be obtained through various semantic and perceptual generation manipulations. Despite numerous studies investigating semantic and perceptual generation tasks separately, studies have not compared the effectiveness of the two formats directly. For Experiment 1, participants read or generated words from rhyming words or highly associated words, comprising a 2 (level of processing: perceptual, semantic) x 2 (generation status: generation, read) within-subjects design. The results showed that participants had higher memory performance for generated words than for words that were read. Moreover, an interaction revealed that memory performance was higher for the semantic generation task than for the perceptual generation task. Experiment 2 aimed to investigate participants’ memory predictions for perceptual and semantic generation tasks. Moreover, we investigated whether making predictions modifies memory performance. Experiment 2 incorporated judgments-of-learning (JOL) and no-JOL groups, yielding that participants accurately predicted and performed better on memory tasks involving generation and semantic manipulations. Additionally, the cued-recall retrieval phase produced higher memory performance for the JOL-group than the no-JOL group, suggesting that predicting one's memory performance enhances actual memory performance. Experiment 3 aimed to see the importance of the test type for the metamemory reactivity. Like Experiment 2, the JOL group still had a higher memory performance than the no-JOL group. The group difference was only observable during a semantic cued-recall test, implying the test type's importance.