High-Level Visual Perception, Category Selectivity, and Mechanisms of Directed Attention
High-Level Visual Perception and Object Recognition
Transition from Low-Level Features to High-Level Perception
Visual perception progresses from low-level feature extraction (e.g., orientation, luminance, spatial frequency, color) in primary visual cortex to complex object representations in high-level extrastriate regions.
Perception of whole objects requires integration mechanisms beyond individual, isolated features.
Repetition Attenuation (Adaptation / Suppression / Neural Priming)
Definition: Repetition attenuation (also termed repetition suppression, adaptation, or neural priming) is the phenomenon wherein repeated presentation of identical visual stimuli leads to a reduction in the second neural response relative to the first.
Functional Utility: Used as a methodological probe to test whether a given cortical region treats two consecutive stimuli as functionally distinct or identical.
Differential Selectivity across Visual Regions:
Region A (High Selectivity): Exhibits distinct non-adapted neural responses when switching between different object types (e.g., hammers versus screwdrivers), demonstrating sensitivity to object categories.
Region B (Category Invariance): Demonstrates neural adaptation when a screwdriver follows a hammer, treating distinct object categories as functionally equivalent.
Shape and Form Selectivity in the Lateral Occipital Complex (LOC)
Kourtzi & Kanwisher (2001) Study: Investigated whether the Lateral Occipital Complex (LOC) responds selectively to high-level visual shape/form while remaining invariant to low-level contour changes.
Experimental Findings: Viewing two objects sharing the same high-level shape causes significant repetition suppression in LOC, even when low-level visual contours differ (e.g., occluded versus unoccluded boundaries, luminance versus texture-defined contours).

Signal Change Timecourse in LOC:
Identical stimuli evoke the strongest signal attenuation (peak \tildesymbol\approx 0.15\% signal change).
Same shape stimuli with differing contours evoke equivalent adaptation (peak \tildesymbol\approx 0.16\% signal change).
Completely different shapes evoke unattenuated, high-amplitude signals (peak \tildesymbol\approx 0.20\% to signal change).
Category-Selective Neural Regions in the Ventral Visual Pathway
Neuropsychological Evidence: Patient CK exhibits visual object agnosia with preserved face recognition (e.g., perceiving human faces in Arcimboldo fruit/vegetable portraits when upright, but failing to recognize individual constituent vegetables).
Fusiform Face Area (FFA): Located on the fusiform gyrus. Kanwisher et al. (1997) demonstrated that the FFA shows significantly greater fMRI signal change for human faces () compared to non-face objects ().

Parahippocampal Place Area (PPA): Located on the parahippocampal gyrus; selectively engaged during the perception of scenes, landscapes, houses, and spatial environments.
Extrastriate Body Area (EBA): Selectively responds to human body parts and headless human bodies.
Visual Word Form Area (VWFA): Selectively engaged during the visual perception of written words and letter strings.

Data-Driven Discovery of Ventral Specializations: Khosla, Murty, & Kanwisher (2022) applied computational data-driven methods to correlate ventral visual activity across extensive image databases, re-discovering FFA, EBA, PPA, and identifying a distinct Food Area.
Theoretical Debates: Modularism versus Distributed Representations
Modularism: Hypothesizes that the ventral visual pathway is divided into discrete, domain-specific functional processing modules optimized for specialized stimulus categories (advocated by Kanwisher, Brodmann, Broca, Chomsky, and early phrenological lineages).
Distributed Coding: Hypothesizes that object categories are encoded via broad, overlapping population patterns across ventral temporal cortex rather than encapsulated modules (advocated by Haxby, McClelland, Rumelhart, and connectionist frameworks).
Spatial Attention and the Biased Competition Model
Attentional Selection and Visual Capacity Constraints
Visual perception is capacity-limited; only a small subset of incoming visual input can be processed in fine detail at any given moment.
Covert spatial attention operates voluntarily to selectively process information from discrete spatial locations without overt eye movements.
Hierarchy of Receptive Field Sizes along the Ventral Pathway
Receptive Field (RF) Definition: The specific region of the visual field within which visual stimuli trigger a response in a given neuron.
RF sizes systematically expand at successive visual processing stages:
Primary visual cortex (V1): Small receptive fields ( near fovea).
Intermediate extrastriate cortex (V2, V4): Moderate receptive fields.
Inferotemporal cortex (IT, including FFA and PPA) and Prefrontal Cortex (PFC): Broad receptive fields covering entire quadrants or whole visual fields.


The Biased Competition Model of Attention
Mechanistic Principle: Due to increasing receptive field sizes, multiple neurons from early visual cortex converge on single neurons at higher visual stages. When multiple stimuli fall within a single higher-order RF, they compete for representation in a mutually suppressive fashion.
Role of Spatial Attention: Directing attention to a specific stimulus location biases the neural competition, driving the neuron's response as if only the attended stimulus were present inside its receptive field.
Electrophysiological Proof in Non-Human Primates (Moran & Desimone, 1985)
Single-unit recordings from area V4 neurons in macaque monkeys presented with effective (preferred) and ineffective (non-preferred) stimuli inside their receptive field.
Both Stimuli Inside RF:
Attending to the effective stimulus inside the RF yields a strong, unsuppressed firing response.
Attending to the ineffective stimulus inside the same RF suppresses the response to the effective stimulus, reducing the neuron's spike rate.
One Stimulus Inside RF, One Stimulus Outside RF:
Attending to a stimulus outside the RF does not reduce the response to the effective stimulus inside the RF, proving that competitive suppression operates specifically within shared receptive fields.

Feature- and Category-Based Attention
Non-Spatial Attentional Selection
Attention can target specific low-level features (e.g., color, shape, orientation, velocity) or high-level categories (e.g., faces, scenes, body parts) throughout the visual field, independent of fixed spatial locations.
Cortical Loci of Feature Attention (Corbetta et al., 1990)
fMRI/PET subtractive methodology comparing selective feature attention (detecting changes in color, shape, or velocity individually) minus divided attention (detecting changes across any feature).
Color Attention: Enhances activation in the dorsal lateral occipital gyrus (dlo) and lingual gyrus (lin).
Coordinates: dlo L (, magnitude 31; , magnitude 33; , magnitude 34; , magnitude 35), dlo R (, magnitude 30), lin L (, magnitude 39), lin R (, magnitude 30).
Shape Attention: Enhances activation in the lingual gyrus (lin), fusiform gyrus (fus), parahippocampal gyrus (ph), parieto-occipital sulcus (pos), and superior temporal sulcus (sts).
Coordinates: lin L (, magnitude 28; , magnitude 29), fus L (, magnitude 32), fus R (, magnitude 25), ph L (, magnitude 31), ph R (, magnitude 27; , magnitude 33; , magnitude 29), pos L (, magnitude 33), sts L (, magnitude 26), sts R (, magnitude 39).
Velocity/Motion Attention: Enhances activation in area MST / top (, magnitude 32).

Object-Based and Category-Selective Attention (O'Craven et al., 1999)
Stimuli comprised transparently superimposed face and house images (one moving, one stationary) occupying the exact same spatial location.
Attending to the face attribute selectively increases FFA signal change (\tildesymbol\approx 2.3\% versus \tildesymbol\approx 1.3\% when attending scene).
Attending to the house attribute selectively increases PPA signal change (\tildesymbol\approx 1.0\% versus \tildesymbol\approx 0.4\% when attending face).

Object-Based Selection Rule: Attending to a single feature of an object (e.g., the motion of a moving face) automatically causes attentional selection of all other bound features of that object, elevating activity in corresponding category-selective areas (e.g., FFA). Integrated objects serve as the fundamental unit of visual selection.
Integration of Feature and Spatial Attention: Feature-Similarity Gain Model (Treue et al., 1999)
Global Feature Facilitation: Feature-based attention acts globally across the entire visual field, enhancing the response of neurons tuned to the attended feature regardless of location.
Additive Spatial Enhancement: Spatial attention provides targeted, localized enhancement at the specific focus of attention.

Electrophysiological Measurements: Featural attentional modulation alone increases response across recorded cells by a geometric mean of (out of 131 cells). Combined spatial and featural modulation produces a geometric mean response of .
Neural Mechanisms of Directed Attention in Human Extrastriate Cortex (Kastner et al., 1998)
Study Overview and Methodological Parameters
Investigators: Sabine Kastner, Peter De Weerd, Robert Desimone, and Leslie G. Ungerleider (1998).
Objective: Measure suppressive sensory interactions among simultaneous visual stimuli without directed attention (Experiment 1) and test whether spatially directed attention counteracts these suppressive interactions in human extrastriate cortex (Experiment 2).
fMRI Acquisition Parameters: 1.5 Tesla GE magnet, gradient echo EPI, repetition time () = , echo time () = , flip angle = , 14 contiguous coronal -thick slices starting from the posterior pole, in-plane resolution = .

Retinotopic Meridian Mapping and Region Identification
Visual areas identified in the left hemisphere during right upper quadrant stimulation:
Calcarine sulcus (BA 17): Area V1 (mean Talairach coordinates: ).
Lingual gyrus (BA 18): Area V2 (mean Talairach coordinates: ).
Fusiform gyrus (BA 19 and 37): Area V4 (mean Talairach coordinates: ) and putative area TEO (mean Talairach coordinates: ).

Experiment 1: Suppressive Sensory Interactions in the Absence of Directed Attention
Stimulus Conditions: Four complex images () presented in the right upper visual quadrant at to 10^^\tildesymbol\circ eccentricity in blocks.
Sequential (SEQ): Stimuli shown individually across four locations ( exposure, blank per location, rate = per location).
Simultaneous (SIM): Stimuli shown together across all four locations ( exposure, blank, rate = ).
Attentional Control Task at Fixation: Participants performed a high-load letter discrimination task (counting T's or L's of size presented at in random orientations; average performance correct) at the fixation point to divert attention away from peripheral stimuli.
fMRI Response Suppression Results: SIM presentations evoked significantly weaker responses than SEQ presentations across all regions ( for V1, V2, TEO; for V4).
Sensory Suppression Index (SSI):
Scaling with Receptive Field Size: Significant interaction between cortical area and presentation condition (). SSI magnitude increased progressively from V1/V2 to V4/TEO ( for V1/V2 versus V4/TEO), matching receptive field expansion along the ventral stream.
Control Experiments in Experiment 1
Presentation Rate Control: Evaluated a single stimulus () presented at near the horizontal meridian in the upper visual field (UVF) alone versus together with three stimuli in the lower visual field (LVF). Response in V4's UVF representation was significantly greater for the single stimulus ( signal change) than when presented with LVF stimuli ( signal change; paired t-test), confirming suppressive interactions independent of rate.
Spatial Separation Control: Separating four smaller stimuli () by 6^^\tildesymbol\circ abolished suppressive interactions in V2, strongly reduced them in V4, but maintained them in TEO, confirming that suppression depends on stimuli falling within shared receptive fields.

Experiment 2: Attentional Modulation of Suppressive Interactions
Factorial Design: Presentation Condition (SEQ versus SIM) Directed Attention Condition (Unattended versus Attended).
Attended Task: Covertly attend to the array stimulus location closest to fixation and count occurrences of a designated target stimulus (target accuracy: in SEQ, in SIM).
Attentional Enhancement Rates:
In V4: Attending elevated fMRI signal by for SIM versus for SEQ.
In TEO: Attending elevated fMRI signal by for SIM versus for SEQ.
Interaction Effects: Significant interaction between attention and presentation condition in area V4 () and area TEO (); just missed significance in V2 (). Attending to the location partially canceled out the suppressive influence of surrounding visual clutter, marked by reductions in SSI values in V4 and TEO.
Main Attentional Effect: General increase in neural activity affecting both SEQ and SIM conditions in V2 (), V4 (), and TEO (), but not V1 (); main effect ; interaction with area .
Expansion of Activated Volumes: Directed attention expanded activated cortical volumes in V4 ( increase) and TEO ( increase; ).