Faces and the Extralist Feature Effect

Background

The extralist feature effect describes an effect in short-term memory, in which stimuli containing features which did not appear on the study list are more easily recognised as novel when they contain features which were not presented on the study list (Mewhort & Johns, 2000). It has been previously found to occur in stimuli whose dimensions are processed independently of one another, but not those whose dimensions are processed as a single whole (Osth et al., 2023). Processing of face stimuli is proposed to be holistic, meaning that faces are processed as unified wholes rather than as a collection of parts or features. Much research on holistic processing in faces has used composite faces, which consist of the upper and lower halves of different faces aligned to form the illusion of a novel face. Previous research using composite faces has suggested that participants struggle to attend to one half of a composite face without interference from an aligned distractor half (Young et al., 1987). In assessing whether the extralist feature effect occurs in aligned composite faces, we can better understand whether composite face halves can be processed independently of one another.

Research Questions / Hypotheses

Does the extralist feature effect occur in short-term recognition memory of composite faces?

Participants

Ninety participants completed the study through the REP. To exclude participants with low sensitivity, d' values were calculated to compare the hit rates of targets and the false alarm rates for lures whose halves were not presented anywhere on the study list. Fifteen participants were excluded due to d' values of under 0.4.

Methods

Participants completed a series of short recognition memory trials, during which they viewed four composite faces and were subsequently asked to identify if a test face was among the faces they just studied. On half of the trials, participants were presented with one of the study list items, while on the other half they were presented with a face which did not appear among those they studied - a lure. The lures in this study varied on how often their halves were presented among the four studied faces. Two lure types were of particular interest to the research question: 1:1 lures, whose halves both appeared once each on the study list, and 2:0 lures, which consisted of one half which appeared twice on the study list, and one which did not appear at all. As both lure types consist of face halves that appear a total of two times on the study list, any differences in accuracy on each lure type may reveal whether extralist halves make faces easier to identify as novel.

Results

A Bayesian paired-samples t-test comparing participants’ accuracy on 1:1 and 2:0 lures revealed evidence suggesting that 1:1 lures were erroneously endorsed more often than 2:0 lures, with mean error rates of .42 and .37 respectively. However, we also conducted further exploratory analyses to understand whether this effect held regardless of whether the upper or lower half of the 2:0 composite face lure was the novel half (i.e., extralist). The results indicated that accuracy for 2:0 lures was only higher than that of 1:1 lures when the upper half of the face was extralist; an extralist lower half did not grant 2:0 lures any rejection advantage.

Implications

Although the overall rejection advantage for 2:0 lures over 1:1 lures would suggest an extralist feature effect in composite faces, the finding that the rejection advantage for 2:0 lures only held for those whose upper half did not appear on the study list indicates that participants attended preferentially to the upper halves of the composite faces. This is unlike previous research investigating the extralist feature effect in stimuli whose dimensions can be processed independently of one another, as in these studies, participants made fewer errors with 2:0 lures regardless of which dimension contained the extralist feature (Mewhort & Johns, 2000; Osth et al., 2023). These findings suggest a need to assess the extralist feature effect in faces using multiple dimensions which are equally salient and diagnostic of identity. These results will be addressed in an honours-level thesis and may be considered for future publication.