Object maintenance beyond their visible parts in working memory

Abstract

Completion of a partially occluded object requires that a representation of the whole is constructed based on the information provided by the physically specified parts of the stimulus. Such processes of amodal completion rely on the generation and maintenance of a mental image that renders the completed object in visual working memory (VWM). The present study examined this relationship between VWM storage and processes of object completion. We recorded event-related potentials to track VWM maintenance by means of the contralateral delay activity (CDA) during a change detection task in which composite objects (notched shapes abutting an occluding shape) to be memorized were primed to induce either a globally completed object or a noncompleted, mosaic representation. The results revealed an effect of completion in VWM despite physically identical visual input: change detection was more accurate for completed compared with mosaic representations when observers were required to memorize two objects, and these differences were reduced with four memorized items. At the electrophysiological level, globally completed (vs. mosaic) objects gave rise to a corresponding increase in CDA amplitudes. These results indicate that although incorporating the occluded portions of the presented shapes requires mnemonic resources, the complete object representations thus formed in VWM improve change detection performance by providing a more simple, regular shape. Overall, these findings demonstrate that mechanisms of object completion modulate VWM, with the memory load being determined by the structured representations of the memorized stimuli. NEW & NOTEWORTHY This study shows that completion of partially occluded objects requires visual working memory (VWM) resources. In the experiment reported, we induced observers to memorize a given visual input either as completed or as noncompleted objects. The results revealed both a behavioral performance advantage for completed vs. noncompleted objects despite physically identical input, and an associated modulation of an electrophysiological component that reflects VWM object retention, thus indicating that constructing an integrated object consumes mnemonic resources.

Amodal Completion of a Target Template Enhances Attentional Guidance in Visual Search

“…However, search for a global interpretation of the composite object was still less efficient than search for a simple global target, which potentially resulted from a perceptual difference between these two types of stimuli (e.g., due to the additional, occluding square in composite objects, see Figure 1(a)). Despite these perceptual differences, our results show that the context provided by trials presenting simple objects (in a given part of the experiment) engendered some form of “priming” that affected the interpretation of the concurrent composite objects (Chen et al., 2016, 2018; Plomp & Van Leeuwen, 2006). In line with previous findings in a working memory task (Chen et al., 2016, 2018), priming by the simple-object context led to an increased search efficiency for the globally completed (composite) objects—indicating that the completion of a symmetrical global object can facilitate search.…”

“…From this perspective, our results would indicate that, given a predictive simple-object context, a globally completed template representation is set up preferentially and this regular, symmetric specification of the target then in turn expedites performance by directly guiding focal attention to the (partially matching) target object—which, as a result of the top-down projection, is then actually interpreted or “perceived” as globally completed. This is consistent with previous findings that have established a link of object completion to working memory processes (Chen et al., 2016, 2018) and suggests that some imagery process completes the missing, occluded information, thereby generating an effective, perceptually simple template (a global “Gestalt”) that in turn expedites target detection. Although a symmetric global target representation facilitated search, representations of mosaic-type or locally completed targets were found not to be comparably efficient in providing search guidance—likely because they are more difficult to construct and maintain as a template in visual working memory.…”

“…Of note in this context, recent work has shown that visual working memory tends to represent occluded objects in terms of globally completed wholes, rather than local completions (Chen, Müller, & Conci, 2016; Chen, Töllner, Müller, & Conci, 2018; for stimuli, see Figure 1). A benefit of global over local completions was also reported in studies that used an implicit primed matching task (Sekuler, Palmer, & Flynn, 1994; R. Van Lier, 1999).…”

Neuroimaging Findings on Amodal Completion: A Review

“…Zhou, Friedman, & von der Heydt, 2000), electroencephalography (EEG) with humans (Caputo, Romani, Callieco, Gaspari, & Cosi, 1999; J. Chen, Liu, Chen, & Fang, 2009; S. Chen, Töllner, Müller, & Conci, 2017; Hazenberg et al., 2014; Hazenberg & van Lier, 2016; Johnson & Olshausen, 2005; Makin, Poliakoff, & El-Deredy, 2009; Murray, Foxe, Javitt, & Foxe, 2004), magnetoencephalography (MEG) with humans (de Wit, Bauer, Oostenveld, Fries, & van Lier, 2006; Liu, Plomp, van Leeuwen, & Ioannides, 2006; Plomp, Liu, van Leeuwen, & Ioannides, 2006; Rajaei, Mohsenzadeh, Ebrahimpour, & Khaligh-Razavi, 2018), and functional magnetic resonance imaging (fMRI) in humans (Ban et al., 2013; J.…”

Object-based grouping benefits without integrated feature representations in visual working memory

“…This indicates that grouping operations increase the amount of information that can be stored in VWM. Moreover, there is evidence that VWM not only uses object grouping to compress the stored information to save capacity, but perceptual completion may conversely also improve the resolution of a memorized object-for example, when an occluded part of an object is completed behind an occluder to increase the fidelity of the object maintained in memory (Chen, Müller, & Conci, 2016;Chen, Töllner, Müller, & Conci, 2018b).…”