Event-related gamma-band activity in visual object representation: the coding of object features Jasna Martinovic Thomas Gruber * Matthias M. Müller * School of Psychology, University of Liverpool * Department of Experimental Psychology, University of Leipzig X ICON, Bodrum, Turkey, 2008
Outline Introduction Visual object representation Coding of object features Methods Experimental section 1) surface detail (contours, texture, colour) 2) visual complexity (low or high complexity) 3) colour typicality in colour diagnostic objects Conclusions
Visual Object Representation EXTRACTION OF OBJECTS FROM IMAGES Low-level: Edges Texture Colour IN VISUAL PROCESSING Mid-level: surfaces In 2.5/3D High-level: categorical Knowledge City; Buildings; Trees, Lampposts; Liver Building; Pierhead; Liverpool
Coding of object features Object representation can be achieved ultra-rapidly based on very few features in images of natural scenes (Simon Thorpe’s group). VanRullen & Thorpe, 2001, JoCN Early object-related effects reflect systematic low-level differences while late effects reflect the decision process; this is likely to be based on feed- forward processing of magnocellular input. Animal vs. vehicle 80 ms Target vs. non-target 150 ms
Coding of texture and colour? Object representations are task-relevant as well as context-dependent – they can also rely on other features (Tanaka et al., 2001). Animal Superordinate Bird Basic Bluebird Subordinate Shape Texture Colour Entry-level Entry-level is the level of identification of objects in everyday experience, wherein their names are being accessed.
Neural markers of visual object representation Cortical object representation occurs through formations of distributed synchronously-firing neural assemblies, coding for each of the various object features (Tallon-Baudry and Betrand, 1999) blue feathery texture shape flies about High-frequency oscillatory activity (gamma-band of the EEG) Evoked GBA – feature representation; induced GBA – object representation
Experimental design Experiment 3: Colour typicality Typical Atypical Naor-Raz et al., 2003 Experiment 2: Visual complexity Low complexity High complexity Ellis & Morrison, 1998 Line drawing Texture & shading Colour Rossion & Pourtois, 2004 Experiment 1: Surface detail
Procedure x Fixation ( ms) Stimulus (650 ms) Fixation (1650 ms) End of trial; blink (900 ms) [time for response: 2300 ms ] Grammatical gender decision task Martinovic et al., 2008, Brain Research
Gamma-band analysis EVOKED + INDUCED: Average of single trials in the freq.domain EVOKED: Average of single trials (ERP) 128 electrode Biosemi ActiveTwo Amplifier System SCADS artifact correction (Junghoefer et al., 2000) incorrect trials excluded average trial rejection rate <30% average of trials remaining average reference used EEG RECORDING:
Hypotheses EVOKED GBA: Sensory processing of image features (Karakas & Basar, 1998) INDUCED GBA: Cortical object representation (Tallon-Baudry & Bertrand, 1999) Surface detail enhancement no modulation Visual complexity enhancement Colour typicality no modulation EVOKED GBA: INDUCED GBA: AMPLITUDE MODULATIONS
Reaction Times Experiment 3: Colour typicality Experiment 2: Visual complexity Experiment 1: Surface detail (No effects across participants) Line Texture Colour Colour diagnostic p< LowHigh Conceptual complexity p< TypicalAtypical p<0.01
Evoked GBA – Surface Detail -0.1 [μV] [ms] [Hz] Line drawing Texture Colour ms [μV] Oz O2 O Line drawingTextureColour [μV] F(2,16)=3.50 p< [μV] line drawing texture coloured P1 ***
Evoked GBA – Visual Complexity ms [μV] [μV] [ms] [Hz] Low High [μV] Oz O1 O2 POz PO3 PO Low visual complexity High visual complexity t (17) = p< low comp. high comp. [μV] P1 ***
Evoked GBA – Colour Typicality ms [μV] Oz O1O TypicalAtypical t (25) = [μV] [ms] Typical Atypical [Hz] [μV] typical atypical P1
ERPs Experiment 3: Colour typicality Experiment 2: Visual complexity Experiment 1: Surface detail [μV] [ms] N [ms] [μV] N [μV] [ms] N350 Shorter latencies for coloured objects Amplitude increase for more complex objects Amplitude increase for atypically coloured objects
Induced GBA [ms] [Hz] [μV] [Hz] [ms] [μV] [Hz] [ms] Line drawing Texture Colour [μV] Typical Atypical Low comp. High comp. Experiment 1: Surface detail Experiment 2: Visual complexity Experiment 3: Colour typicality
Evoked GBA primarily reflects the quantitative processing of object features Early synchronizations may be sufficient for relatively crude discrimination of objects in visual scenes (Frϋnd et al. 2008) but cannot support entry-level categorization. Induced GBA remained unmodulated in amplitude by object features. It seems to reflect the communality of object-related conceptual processes and is not so reliant on the coding of basic perceptual features. Conclusions blue feathery texture shape flies about Superordinate Identification (animal) Entry-level Identification (bird) ?
Acknowledgments Renate Zahn Sophie Trauer Søren Andersen Tobias Forderer Research has been funded by the DFG and DAAD J.M. is supported by an ESRC Postdoctoral Fellowship
Induced GBA Waveforms Experiment 3: Colour typicality Experiment 2: Visual complexity Experiment 1: Surface detail Hz, Grand Means [μV] low comp. high comp. typical atypical line drawing texture coloured