Interfering with memory for faces: The cost of doing two things at once

Abstract

We inferred the processes critical for episodic retrieval of faces by measuring susceptibility to memory interference from different distracting tasks. Experiment 1 examined recognition of studied faces under full attention (FA) or each of two divided attention (DA) conditions requiring concurrent decisions to auditorily presented letters. Memory was disrupted in both DA relative to FA conditions, a result contrary to a material-specific account of interference effects. Experiment 2 investigated whether the magnitude of interference depended on competition between concurrent tasks for common processing resources. Studied faces were presented either upright (configurally processed) or inverted (featurally processed). Recognition was completed under FA, or DA with one of two face-based distracting tasks requiring either featural or configural processing. We found an interaction: memory for upright faces was lower under DA when the distracting task required configural than featural processing, while the reverse was true for memory of inverted faces. Across experiments, the magnitude of memory interference was similar (a 19% or 20% decline from FA) regardless of whether the materials in the distracting task overlapped with the to-be-remembered information. Importantly, interference was significantly larger (42%) when the processing demands of the distracting and target retrieval task overlapped, suggesting a processing-specific account of memory interference.

Keywords:

• Divided attention
• Episodic memory
• Interference
• Configural
• Featural

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

¹ An a priori threshold was set to exclude participants scoring below 0.4 in accuracy. With scores this close to floor, any further reduction from a DA condition is unlikely to be observed, thereby preventing a reliable estimate of interference effects. One participant fell below this threshold and a new participant was run in their place. The excluded participant was greater than 2.0 units of standard deviation below the final sample’s mean for accuracy.

² The assumption of sphericity was tested using Mauchly’s test for all ANOVAs, and the Greenhouse-Geisser corrections were applied when warranted.

³ As in Experiment 1, an a priori threshold was set to exclude participants scoring below 0.4 in accuracy. Fourteen participants fell below this cutoff and new participants were run in their place. The excluded participants were greater than 2.0 units of standard deviation below the final sample's grand mean for accuracy. To be conservative, analyses were also conducted with these participants included, and the results followed the same pattern as reported.

⁴ See Footnote 2.

⁵ Tone classification data were also analyzed for the raw number of tones correctly identified, as well as response times. These data followed the exact same pattern, such that baseline performance was better and faster than all dual-task conditions, which were not different from one another (See Table 3 for means).

Additional information

Funding

This work was funded by an NSERC Discovery grant awarded to author Myra A. Fernandes.