Overt and covert identification of fragmented objects inferred from performance and electrophysiological measures.

Abstract

The authors investigated visual processing leading to object identification by manipulating the number of fragments and nature of the study. During the study, participants either named or drew objects in Experiment 1 and drew them all in Experiment 2. During the test, participants made an identification judgment at each of 6 different fragmentation levels for studied and new objects. Fewer fragments were needed to identify studied than unstudied objects. Reaction times were faster for studied than unstudied objects both at identification and at the preceding level. Event-related brain potentials (ERPs) to unidentified objects were characterized by a late negativity in contrast to a positivity to identified objects. ERPs to studied but not to new objects contained a smaller and later version of the identification positivity at level just prior to identification, which was not due to differential response confidence. Much covert visual analysis and even object identification may precede overt identification, depending on the nature of prior experience. At present, there is no single psychological or neural account of how people identify the objects that they see. Of the handful of different theories of visual object identification that have been proposed, all describe the identification process as some function of the visual input (retinal image) and the structure and content of the processor (memory). The identification of an object is presumed to emerge from a dynamic interplay between bottom-up, data-driven, sensory processes and top-down, conceptually driven, mnemonic processes. The various theories differ primarily in the finer details of how this happens. Although there is a consensus that retinal images must be interpreted, presumably by processes that compare visual input against representations in memory until a match (i.e., identification) emerges, there is less agreement as to either the nature or the number of memory traces involved. In part, this is due to the changing views of what constitutes memory, which now include distinctions between memories that are easy to consciously access and talk about (declarative or explicit memories) and those that manifest their existence through an individual's actions or preferences (procedural or implicit memories). Other points of contention concern the specifics of the (a) matching process, (b) nature of the sensory representation(s), (c) nature or number of object representations in memory, (d) relative contributions of bottom-up versus

Neurophysiological evidence for two processing times for visual object identification

“…Within the same people, we found that ERP identification effects occur earlier for recoverable objects (Levels 4-5 in Block 1; Level 5 in Block 3) than for non-recoverable ones (Level 2 in Block 2). Indeed, part recoverability can explain all extant findings: Identification begins by ∼300 ms for recoverable images, landscape scenes [37] and fragmented pictures (Level 5) [62], but not until 544 ms for non-recoverable images, fragmented pictures (Level 2) [55].…”

The influence of contour fragmentation on recognition memory: An event-related potential study

“…The second subset covered the centro-parietal area and included the electrodes Cz, C4, C3, Cp4, Cp3, Pz, P4, P3 (Stuss et al, 1992). The third subset covered the frontal area and included the electrodes Fp2, Fp1, F4, and F3 (Viggiano & Kutas, 1998;Viggiano & Kutas, 2000). The negativity found in each of these three subsets had different latencies.…”

“…The negativity to fragmentation was identified as the ERP component that was more negative for the fragmentation conditions as compared to the complete condition between 180 and 450 ms. Because our stimuli were visually simpler, this time-window started slightly earlier than those used by Stuss and colleagues (1992) and by Viggiano and Kutas (1998), Viggiano and Kutas (2000). The analyses were conducted on three subsets of electrodes, each of which covered an area where modulation to fragmentation has been reported in the literature.…”

“…A visual inspection of the ERPs depicted in their Fig. 8 (p. 114) (Viggiano & Kutas, 2000), however, suggests that the effect peaked somewhere between 200 and 300 ms.…”

“…Their results indicate that between 250 and 450 ms, at Cz, the amplitude was more negative to stimuli that were the most incomplete. Taking a different approach, Viggiano and Kutas (1998), Viggiano and Kutas (2000) compared the ERPs evoked at the identification level with those at one higher level with more fragments added to the stimulus. Consistent with the results of Stuss and colleagues (Stuss et al, 1992), the amplitude was more negative for the identification level, thus for the most fragmented level.…”