Table 1. Equations and parameter values for each of the components of the two models.

Model component	Equation	Description
Suppressive drive	$S_L=K_S*(A_x{A}_v{E}_L^n)$	* denotes convolution. K s is the suppression kernel. The attention gain factors, A x and A v, and the excitatory drive $E_L^n$ were point-by-point multiplied.
Suppression kernel	$K_S(x,{\theta }_i)=e^{\frac{-x^2}{2\sigma {({\theta }_i)}^2}}$ $\sigma (\theta )={\sigma }_{sx}{e}^{\frac{-\theta }{\tau }}$	σ sx = 6°: width of the spatial suppressive field for iso-orientation stimuli. τ = 20°: orientation selectivity of surround suppression. Suppression was broadly tuned (all orientations) for stimuli within a neuron’s receptive field while surround suppression was narrowly tuned to the neuron’s preferred orientation. See Fig 1B.
Stimulus-driven attentional gain factors	$A_x=w_x(a_{\theta }{a}_x^T)+1$	w x: free parameter that determined the strength of stimulus-driven attentional modulation. aθ and ax: vectors representing the extent of feature-based and spatial attention. The attentional gain factors had a baseline of 1. Values larger/smaller than 1 indicated increases/decreases in gain.
Goal-driven attentional gain factors	$A_v=w_v(a_{\theta }{a}_x^T)+1$	w v: free parameter that determined the strength of goal-driven attentional modulation. Both A x and A v involved computation of aθ and ax but the attended orientation and attended position were different for stimulus-driven and goal-driven attention (see Feature-based attention and Spatial attention below).
Feature-based attention	$a_{\theta }=e^{k\left\{\text{cos}\right\{2(\theta -{\theta }_a)\}-1\}}-0.5$	A circular Gaussian (von Mises) profile scaled to range from -0.5 and 0.5. θ a: attended orientation. k: extent of feature-based attention. We used k = 3, which resulted in a 40° bandwidth (FWHM), similar to that used in previous simulations [37]. For stimulus-driven attention, θ a was the orientation of the competitor in the FS model while a θ was uniform (= 1) in the ES model. For goal-driven attention, θ a was the orientation of the target in both FS and ES models.
Spatial attention	$a_x=\frac{1}{{\sigma }_{ax}^p\sqrt{2\pi }}{e}^{\frac{-{(x-x_a)}^2}{2{\sigma }_{ax}^2}}$	p: tradeoff between the spatial extent and the magnitude of attentional modulation. If p = 1, spatial attention had a fixed volume, decreasing in magnitude when its spatial extent increased. If p = 0, the spatial extent was independent of the magnitude of the attentional gain at the attended position x a. σ ax: spatial extent of spatial attention. For goal-driven attention, these parameters were fixed: σ ax = 60° and p = 0. For stimulus-driven attention, p was a free parameter and σ ax was equal to the size of the competitor.