Top-down feedback processes are engaged by unreported visible, but not invisible, changes
EliseRowe1
442 views
13 slides
Dec 03, 2018
Slide 1 of 13
1
2
3
4
5
6
7
8
9
10
11
12
13
About This Presentation
Detecting changes in the environment is fundamental for survival, as these may indicate potential rewards or threats. In our everyday lives, many changes occurring in our environment do not pose any threat and largely go unnoticed. Recent work has shown that the visual system can unconsciously detec...
Detecting changes in the environment is fundamental for survival, as these may indicate potential rewards or threats. In our everyday lives, many changes occurring in our environment do not pose any threat and largely go unnoticed. Recent work has shown that the visual system can unconsciously detect and represent these changes without generating any awareness, however our understanding of the network-level modulations and the contribution of the prefrontal cortex to these processes remains unclear.
Here, we developed a novel no-report visual oddball paradigm, where participants engaged in a central letter task during EEG recordings while presented with task-irrelevant high- or low- coherence background random dot motion. Critically, once in a while, the dots changed in direction.
After the EEG session, we behaviourally confirmed that changes in motion direction at high- and low-coherence were visible or invisible, respectively, for each participant. We found that unreported task-irrelevant changes in motion direction elicited event-related potentials (ERP) in visible and even in invisible conditions.
To better understand the neural mechanisms underlying these motion changes, we applied Dynamic Causal Modelling (DCM) to the source-localized EEG data. We focused on the core nodes of: primary visual cortex, middle temporal visual area (MT+), inferior temporal gyrus and orbitofrontal cortex. Both ERP and DCM analysis revealed left and right hemispheric lateralisation of visible and invisible motion direction changes, respectively. Furthermore, winning DCMs comprised only feedforward interactions for invisible motion direction changes, whereas both feedforward and feedback modulations were recruited for visible changes specifically between prefrontal cortex and MT+.
Our findings challenge studies implicating a critical involvement of the prefrontal cortex only for report responses. Instead we propose that unreported, yet reportable, changes in visual motion direction can engage the prefrontal cortex which then modulates MT+ via top-down connections.
Size: 28.89 MB
Language: en
Added: Dec 03, 2018
Slides: 13 pages
Slide Content
Top-down feedback processes are engaged by unreported visible changes in motion direction, but not by invisible Elise G. Rowe, Naotsugu Tsuchiya & Marta I. Garrido
Predictive coding theory Detecting changes in the environment is fundamental for survival Predictive coding error Surprise event! Mismatch between sensory information & prior beliefs Friston , 2005; Garrido et al., 2007; Rao & Ballard, 1999; Friston & Stephan, 2007; Hohwy , 2013; Clark, 2013
Prediction errors without awareness Some changes r each awareness … Whilst others r emain unseen WHY? Bernat et al., 2001; Berti, 2011; Czigler, Weisz & Winkler, 2007; Kogai et al., 2011; Jack et al., 2016; Jack et al., 2015; van Rhijn et al., 2013 Network-level?
The current study: EEG (N = 19) Visual prediction errors Mismatch between standard and surprising (deviant) events Motion coherence (%) for perceived global motion direction Approaching 0%, discrimination of overall direction near impossible Using a no-report paradigm Britten et al., 1992
Visible changes Invisible changes High coherence motion (50%) Low coherence motion (5%) Two motion coherence conditions (roving oddball design):
PE = prediction error First, behaviourally confirm for EACH INDIVIDUAL: High coherence changes VISIBLE and Low coherence changes INVISIBLE Spatiotemporal scalp-level results
Spatiotemporal scalp-level results PE = prediction error
Source-level results: ITG
Dynamic Causal Modelling (DCM): E ffective connectivity LITG LOFC Input ROFC RITG LV1 RV1 Input RMT+ LMT+ 8 Models Tested: 2 x Connections (Direct or indirect) + 4 x Direction of modulation (Forwards or Recurrent, W/ Wout Intrinsic) Nodes extracted from lower t hresholded source images and known visual processing regions
DCM results: E ffective connectivity Visible change visual PE Invisible change visual PE
What does all this mean? Visual PE occur when aware and unaware of changes ERP results support invisible visual PE’s earlier and at lower-levels Visible changes: LEFT ITG Invisible changes RIGHT ITG Difference in aware and unaware changes = PFC feedback Visible changes visual PE recruit feedback process from PFC I nvisible changes only activate forwards and self-connections of V1 and MT+
Take home message … Visual PE to unseen changes: G enerated within early visual areas (FF, right dominant) V isual PE to seen changes: R ecruit top-down processes from prefrontal back to early visual areas (FF/FB, left dominant) Support PFC required for conscious awareness of a change Not just reporting in a task
Thank you! Acknowledgements: This work was funded by University of Queensland Fellowship (2016000071) and the ARC (Australian Research Council) Centre of Excellence for Integrative Brain Function (ARC Centre Grant CE140100007) to MIG as well as ARC Future Fellowship (FT120100619) to NT, and ARC Discovery Project: DP180104128 to MIG and NT, DP180100396 to NT.