Time Course of Processes and Representations Supporting Visual Object Identification and Memory

Abstract

Event-related potentials (ERPs) were used to delineate the time course of activation of the processes and representations supporting visual object identification and memory. Following K. Srinivas (1993), 66 young people named objects in canonical or unusual views during study and an indirect memory test. Test views were the same or different from those at study. The first ERP repetition effect and earliest ERP format effect started at approximately 150 msec. Multiple ERP repetition effects appeared over time. All but the latest ones were largest for same views, although other aspects of their form specificity varied. Initial ERP format effects support multiple-views-plus-transformation accounts of identification and indicate the timing of processes of object model selection (frontal N350 from 148-250 to 500-700 msec) and view transformation via mental rotation (posterior N400/P600 from 250-356 to 700 msec). Thereafter, a late slow wave reflects a memory process more strongly recruited by different than same views. Overall, the ERP data demonstrate the activation of multiple memory processes over time during an indirect test, with earlier ones (within 148-400 msec) characterized by a pattern of form specificity consistent with the specific identification-related neural process or representational system supporting each memory function.

Neurophysiological evidence for two processing times for visual object identification

“…While the Ncl (∼232 ms) 2 has been linked to completion processes that compensate for partial visual information (i.e. perceptual closure), this seems dubious since complete pictures, for which a closure function is not needed, also modulate the frontal N350 [45,48]. Rather we propose that these components reflect object selection processes.…”

“…Hence, identification may be expected to entail ∼100 ms of further processing, estimating identification time at ∼275 ms, within the 232-300 ms range based on the Ncl/N350. The N350 is likely a subcomponent of a frontal negativity peaking ∼400 ms found previously for non-objects [20,46] and related to a frontal N400 (N380, "Ny") evoked by pictures of real objects or nonsense figures [39,48,56]. Being larger for more complex images [39,56], the N400 was hypothesized to index access to long-term memory [56].…”

“…Stuss et al [55] showed people fragmented line drawings of objects at four levels of completeness and found a broadly-distributed late positive complex (LPC) that diverged as a function of naming accuracy by 550-650 ms. Pietrowsky et al [37] replicated the LPC finding but also observed an earlier effect; identified landscapes first differed from unidentified and scrambled scenes by ∼250-300 ms in a frontal negativity peaking between 325 and 400 ms (N350) [37,45,48]. Presence or absence of the frontal N350, thus, distinguishes the timing of visual identification between these two studies.…”

View-sensitive ERP repetition effects indicate automatic holistic processing of spatially unattended objects

“…In combination with the prediction from the attentional cueing paradigm (as described above), the use of ERP correlates of object recognition in this study will therefore help to determine the nature of shape representations underlying object recognition. Schendan and Kutas (2003) observed repetition effects as early as the vertex P150 component (140-250 ms) up to 700 ms post-probe onset, and these were viewdependent: Objects repeated in the same view elicited enhanced amplitudes compared to new objects or objects repeated in a different view. Schendan et al (2010) distinguished processing before and after 200 ms post-probe onset as perceptual or post-perceptual, respectively.…”

“…Some theories posit that object recognition is mediated by viewspecific representations (see e.g. Schendan & Kutas, 2003), possibly by some type of interpolation across several 2D views of an object (Bülthoff & Edelman, 1992;Logothetis, 1994;Poggio & Edelman, 1990;Tarr et al, 1998;Ullman, 1989;Ullman 5 & Basri, 1991) or via a distributed neural representation across view-tuned neurons (Perrett, Oram & Ashbridge, 1998).…”