Predicting the Time Course of Individual Objects with MEG

Abstract

To respond appropriately to objects, we must process visual inputs rapidly and assign them meaning. This involves highly dynamic, interactive neural processes through which information accumulates and cognitive operations are resolved across multiple time scales. However, there is currently no model of object recognition which provides an integrated account of how visual and semantic information emerge over time; therefore, it remains unknown how and when semantic representations are evoked from visual inputs. Here, we test whether a model of individual objects—based on combining the HMax computational model of vision with semantic-feature information—can account for and predict time-varying neural activity recorded with magnetoencephalography. We show that combining HMax and semantic properties provides a better account of neural object representations compared with the HMax alone, both through model fit and classification performance. Our results show that modeling and classifying individual objects is significantly improved by adding semantic-feature information beyond ∼200 ms. These results provide important insights into the functional properties of visual processing across time.

MEG sensor patterns reflect perceptual but not categorical similarity of animate and inanimate objects

“…It should be noted that previous studies provided some evidence for a contribution of semantic object properties to the MEG signal even at relatively early latencies (<250 ms) (Clarke et al, 2015;Coggan et al, 2016;Kaiser et al, 2016b). These studies differed from the current study in several important ways.…”

“…Replicating previous reports, MEG sensor patterns could be used to accurately decode individual objects (Carlson et al, 2011;Carlson et al, 2013;van de Nieuwenhuijzen et al, 2013;Cichy et al, 2014;Isik et al, 2014;Clarke et al, 2015;Ritchie et al, 2015;Coggan et al, 2016;Kaiser et al, 2016a). Using representational similarity analysis, we then related MEG neural similarity to the objects' perceptual and categorical similarity.…”

“…First, none of these studies controlled for perceptual similarity using human judgments or task performance. It is likely that the semantic factors in these studies (e.g., the property "has legs"; Clarke et al, 2015) would still express in perceptual similarity differences in tasks like the oddball visual search task used here. Second, two of these studies included human faces (Coggan et al, 2016) or bodies (Kaiser et al, 2016b).…”

Segregation of anterior temporal regions critical for retrieving names of unique and non-unique entities reflects underlying long-range connectivity

“…Binder et al, 2009). Studies using EEG and MEG have identified the timing of ATL involvement within the network, revealing recurrent interactions between the left ATL, a putative integrative hub, and more posterior regions involved in perceptual processes (Clarke et al, 2011; van Ackeren MJ and Rueschemeyer SA; Clarke et al, 2014). It is also informative that stimulation of long-range fiber tracts connecting the ATL and posterior and frontal regions has been shown to disrupt semantic knowledge retrieval (Shinoura et al, 2010; Papagno et al, 2011; Von der Heide, 2013).…”

Visual object naming in patients with small lesions centered at the left temporopolar region

“…Recent studies have demonstrated that visual object recognition follows an increasingly complex featural processing (Tyler et al 2004) during which conceptual knowledge evolves from a coarse-to-fine grained representation (Cichy et al 2014;Clarke et al 2011Clarke et al , 2013Clarke et al , 2014Chan et al 2011;Peelen and Caramazza 2012). This progress takes place along the hierarchical architecture of the ventral visual pathway (Liu et al 2009), which is a recurrent network that links early visual areas with lateral and medial aspects of the anterior temporal lobe (ATL), and, most rostrally, the ventral temporal pole (Kravitz et al 2013).…”