Experimental setup for sharpening of the neuronal tuning curve.
- Author
- Frederik Beuth
Data: Martinez-Trujillo, Treue (2004), Fig. 4A
Significant results (p. 747, right column):
- Feature-based attention increases response for the preferred motion (positive modulation) and decreases it for anti-preferred motion (negative modulation). The modulation ratio changes monotonic from the preferred motion to the anti-preferred one, resulting in a curve with significant negative slope (Fig. 4B).
- Fig. 4A,4B illustrate the slope of a single cell, but the negative slope is also valid for the population data (Fig. 5).
- The point of no modulation is at the half distance between preferred and anti-preferred motion, i.e. at 90° angular distance of preferred and stimulus motion.
Setup:
- Stimuli have the same size as RFs (area MT), resulting in sizes from 1° to 12°. => We model the stimulus as large as the RF.
Calibration of the fit:
- The tuning curve of feature-based attention is similarly wide as the input tuning curve, but does not have a baseline.
- Input tuning curve is a broad exponential function with a baseline. Baseline is necessary to simulate the response towards anti-preferred stimuli.
- In the simulation, suppression results from two sources: suppression from anti-preferred features, and the standard amount of surround suppression.
- We use full contrasted stimuli: the original authors specify 55cd/m^2. However, it is difficultly to relate this value to a proportional contrast, so we follow Reynolds and Heeger (2009) which uses full contrasted stimuli.