Generation disrupts memory for intrinsic context but not extrinsic context

Abstract

Although generation typically enhances item memory, the effect is subject to a number of theoretically important limitations. One potential limitation concerns context memory but there has been debate about whether generation actually enhances or disrupts memory for contextual details. Five experiments assessed the effect of generation on context memory for perceptual attributes of the study stimulus (intrinsic context). The results indicate that despite enhancing item memory, generation disrupts memory for intrinsic context. This result generalized over different intrinsic details (print colour and font), different generation manipulations, and several methodological factors identified in earlier research as potential moderators of this negative generation effect. Despite the negative generation effect on intrinsic context, contextual details external to the target item (extrinsic context) are not disrupted. In a sixth experiment, a visually based generation task (letter transposition) enhanced context memory for the extrinsic detail of location. Coupled with earlier research, the results indicate that generation disrupts context memory for intrinsic details but either does not harm or enhances memory for extrinsic context.

Keywords:

• Memory
• Context memory
• Episodic memory
• Generation effect

Notes

¹ Spencer and Raz (1995) used the terms stimulus-bound versus spatio-temporal context for this same distinction.

² Control experiments ruled out the alternative possibility that the effect is due to the total amount of the coloured surface or font present in the target stimulus. For example, in the generate condition of Mulligan (2004, Experiment 5), the underscores were changed to solid blocks of coloured print (e.g., hot–c █ █ █ █ █) so that the amount of coloured surface was greater in generate than in read trials. The negative generation effect in context memory persisted. Mulligan et al. (2006, Experiments 3a and 3b) provided complementary demonstrations, showing that the negative effect on context memory is not due to the total number of letters presented in the critical colour or font.

³ Given that the combination of covert response and a nongenerative distractor task produced a negative generation effect for print colour (the same result found with written response coupled with either a nongenerative or a generative distractor task), there is no reason to suspect that covert response coupled with a generative distractor task would produce any difference in result (see Marsh, 2006).

⁴ The fact that nonwords were recognized as well or better than words might seem a bit surprising, although this result is often found when nonwords are compared to high-frequency words (as in the present experiment). In particular, some studies report equal recognition accuracy for nonwords and words (e.g., Johns & Swanson, 1988; Mulligan, 2002), others show higher accuracy for nonwords (e.g., Nairne et al., 1985; Nairne & Widner, 1987), and still other studies show lower accuracy for nonwords (e.g., Payne et al., 1986). Recall, on the other hand, is virtually always greater for words than for nonwords.

⁵ Aside from individual extrinsic details, generation may enhance memory for more global aspects of spatio-temporal context, such as the room in which a word was encoded (Koriat, Ben-Zur, & Druch, 1991; Marsh et al., 2001). However, the exact nature of the enhanced information is unclear. It could be that generation improved memory for temporal information (which covaried with room), memory for perceptual details in the room, or memory for associations between the word and some aspect(s) of the room. It is also not known whether the effects of generation on room memory generalize over some of the methodological factors and generation tasks that moderate generation effects with location memory.