Causal neural mechanisms of context-based object recognition
- Donders Institute for Brain, Cognition and Behaviour, Radboud University, Netherlands
- Department of Biomedical Engineering, University of Minnesota, United States
Figures
Figure 1
Example of context-based object recognition.
At night (top panels), the truck is easily recognized by participants when placed in context (left) but not when taken out of context (right). With sufficient light (bottom panels), the truck is easily recognized also when presented in isolation.
Figure 2
Overview of task and stimulation methods.
(a) Schematic overview of a trial. Two TMS pulses (40 ms apart) were delivered on each trial at one of three time windows relative to stimulus onset (60–100 ms, 160–200 ms, and 260–300 ms). The three TMS timings occurred in random order within each block. (b) Examples of each of the eight categories shown in the experiment, in the isolated object condition (left) and the context-based object condition (right). Note that the local degradation of the objects in the context-based object condition is not clearly visible from these small example images. This degradation strongly reduces object recognition when the degraded objects are presented out of scene context (see Brandman and Peelen, 2017). These conditions were presented in random order and participants performed the same categorization task on all stimuli. (c) Overview of the three TMS sites and the three time windows of stimulation. Shaded background colors indicate presumed time windows of inhibition for double-pulse TMS. TMS, transcranial magnetic stimulation.
Figure 3
Predictions and results.
(a) We hypothesized that isolated object recognition (top row) would be causally supported by EVC at 60–100 ms (early time point in right plot), followed by LOC at 160–200 ms (middle time point in central plot), reflecting feedforward processing of intact object features (Cichy et al., 2014). Scene-selective OPA (left plot) was not expected to contribute to isolated object recognition at any time point (Dilks et al., 2013; Wischnewski and Peelen, 2021). Similar to isolated object recognition, we hypothesized that context-based object recognition (middle row) would be causally supported by EVC at 60–100 ms and by LOC at 160–200 ms, reflecting feedforward processing. In contrast to isolated object recognition, we hypothesized that OPA would causally support context-based object recognition at 160–200 ms (middle time point in left plot), reflecting scene processing. Crucially, scene-based expectations were hypothesized to reach LOC later in time, disambiguating object representations at 260–300 ms (late time point in central plot; Brandman and Peelen, 2017). TMS over LOC at this time point should thus selectively disrupt context-based object recognition. EVC was hypothesized to receive feedback from LOC at 160–200 ms (Camprodon et al., 2010; Koivisto et al., 2011; Murray et al., 2002; Wokke et al., 2013), which we expected to be most important for context-based object recognition, in which the object needs to be segregated from the background scene (Korjoukov et al., 2012; Lamme and Roelfsema, 2000; Scholte et al., 2008). Finally, OPA was predicted to causally support scene-alone recognition at 160–200 ms (bottom row). (b) Results of three TMS experiments. Predictions were largely confirmed, except for feedback effects in EVC (at 160–200 ms), which were specific to isolated object recognition rather than context-based object recognition. *p<0.05, **p<0.01, ***p<0.001, with error bars reflecting the SEM. EVC, early visual cortex; LOC, lateral occipital cortex; OPA, occipital place area; TMS, transcranial magnetic stimulation.
-
Figure 3—source data 1
Individual participant means (accuracy and RT).
- https://cdn.elifesciences.org/articles/69736/elife-69736-fig3-data1-v1.xlsx
Figure 4
Schematic summarizing results.
Distinct cortical routes causally support isolated object recognition and context-based object recognition. Isolated object recognition (top row) was supported by EVC early in time (60–100 ms), reflecting initial visual encoding. This was followed by LOC at 160–200 ms, reflecting higher-level object processing. At this time window, EVC was still required for isolated object recognition, presumably reflecting feedback processing. Similar to isolated object recognition, context-based object recognition (bottom row) was supported by EVC at 60–100 ms, followed by LOC at 160–200 ms. However, context-based object recognition additionally required OPA at 160–200 ms, reflecting scene processing. Finally, context-based object recognition causally depended on late processing (260–300 ms) in LOC, reflecting contextual disambiguation (Brandman and Peelen, 2017). Note that the arrows do not necessarily reflect direct connections between brain regions. EVC, early visual cortex; LOC, lateral occipital cortex.
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Causal neural mechanisms of context-based object recognition
eLife 10:e69736.
https://doi.org/10.7554/eLife.69736
