From Global-to-Local? Uncovering the Temporal Dynamics of the Composite Face Illusion Using Distributional Analyses

Abstract

It is widely believed that faces are processed holistically such that their facial features or parts are represented as global wholes rather than independent entities. But how does their holistic representation evolve in time? According to the global-to-local hypothesis, the initial representation of faces is holistic and coarse at the outset but is becoming progressively detailed and analytic. The current study set to test this global-to-local hypothesis by applying fine-grained methods of response time analyses to the composite face illusion – a traditional marker of holistic face processing. The analyses included the delta plots and conditional accuracy functions. These tools move beyond the mean RT and accuracy to provide detailed analysis of the temporal dynamics of the composite face effect. The methodologies converged on the conclusion that the composite effect is minimal for fast RTs but becomes progressively larger as RT gets slower. This pattern is inconsistent with a global-to-local dynamics. The implications of these results to the study of face perception are discussed.

Decomposing the composite face effect: Evidence for non-holistic processing based on the ex-Gaussian distribution

“…What can be deduced from this pattern? First, such a pattern increases our belief in the view that partial and complete measure tap into common processes (Fitousi, 2019b). Second, since τ is strongly associated with attentional and working memory processes (Schmiedek et al, 2007) and/or object-based attention processes (Spieler, 2001), the results support attentional- and object-based accounts of the composite face effect.…”

“…However, these researchers based their conclusions only on the signal detection theory (SDT) indices. When Fitousi (2019b) measured the correlation in terms of mean RTs (MRTs) and accuracy rates, he found strong (in accuracy) and medium (in RT) significant correlations. Fitousi’s (2019b) results support the relatedness of the two types of measures, and also raise further questions regarding the nature of the (holistic?)…”

“…Also, little is known about the relations between holistic and attentional mechanisms. Moreover, the popular concepts of “holistic” or “configural” processing have been recently challenged (Burton et al, 2015; Fitousi, 2013, 2015, 2016, 2019b). And their operational definitions have become a matter of debate (Richler & Gauthier, 2014; Rossion, 2013).…”

There's a Time and a Face: The Time Course of Composite Face Processing

“…The second explanation is that holistic processing might take time to emerge. Processing might start as analytic or independent but through a gradual widening of attention, might become holistic (Fitousi, 2019). On the other hand, processing might start out as holistic and over time become more independent (cf., White, Ratcliff, & Starns, 2011).…”

“…However, these investigations have relied on distributional measures (Fitousi, 2019) or have varied encoding time but not decision time (Richler, Mack, Gauthier, & Palmeri, 2009;Richler, Mack, Palmeri, & Gauthier, 2011), thus preventing direct comparison between tasks (e.g., the composite face task and SFT, Cheng et al, 2018).…”