Face working memory deficits in developmental prosopagnosia: Tests of encoding limits and updating processes

- Poor face WM in developmental prosopagnosia (DP) is not just an encoding deficit.
- Additional comparison difficulties at retrieval are proposed.
- The recency effect in WM for faces shown sequentially is attenuated in DP.
- Resource demands of inefficient face processing in WM are discussed.
- The nature of task demands at encoding and retrieval should be better considered.

Developmental prosopagnosia (DP) is a condition in which individuals experience life-long problems recognising faces. In recent years, unpacking the nature of the impairments of this population has been the focus of numerous studies. One focus has been on the nature of face-based memory impairments for such individuals, with the onus being mainly on long-term memory deficits. Far fewer have considered the nature of face-based working memory (WM) impairments for DP cases, and the current study seeks to address this. One recent WM study (Shah et al., 2015) reported that the maintenance of faces over time in WM was spared among DPs, and argued instead that face encoding was limited in some way. Here we further explore the nature of face-based WM impairments in DP across two experiments designed to probe encoding limits (Experiment 1) and WM updating processes (Experiment 2). In Experiment 1 we manipulated the number of faces (1−4) to encode into WM and presented these simultaneously. We reasoned that if face encoding among DPs was inefficient or imprecise, then increasing encoding demands (WM load) would disproportionately impair WM accuracy compared to controls. However, we found that DP cases were consistently poorer than controls across all face load conditions, suggesting that front-end encoding problems are only part of the deficit. In Experiment 2, to measure updating four faces were shown sequentially for encoding into WM and accuracy was analysed as a function of whether the test face had been presented first, second, third or last in the encoding sequence. DPs had significantly poorer WM than controls for later faces but not the first face encoded in the sequence, and showed an attenuated recency effect. To account for these findings, we discuss the potential role of comparison processes at retrieval, impairments in configural face processing, and the impact of noise in the face identification system of individuals with DP.