An integrated microcircuit model of attentional processing in the neocortex

Abstract

Selective attention is a fundamental cognitive function that uses top-down signals to orient and prioritize information processing in the brain. Single-cell recordings from behaving monkeys have revealed a number of attention-induced effects on sensory neurons, and have given rise to contrasting viewpoints about the neural underpinning of attentive processing. Moreover, there is evidence that attentional signals originate from the prefrontoparietal working memory network, but precisely how a source area of attention interacts with a sensory system remains unclear. To address these questions, we investigated a biophysically based network model of spiking neurons composed of a reciprocally connected loop of two (sensory and working memory) networks. We found that a wide variety of physiological phenomena induced by selective attention arise naturally in such a system. In particular, our work demonstrates a neural circuit that instantiates the "feature-similarity gain modulation principle," according to which the attentional gain effect on sensory neuronal responses is a graded function of the difference between the attended feature and the preferred feature of the neuron, independent of the stimulus. Furthermore, our model identifies key circuit mechanisms that underlie feature-similarity gain modulation, multiplicative scaling of tuning curve, and biased competition, and provide specific testable predictions. These results offer a synthetic account of the diverse attentional effects, suggesting a canonical neural circuit for feature-based attentional processing in the cortex.

Figure 1.

Scheme of the loop architecture (red is excitation, and blue is inhibition). Two kinds of motion stimuli are considered (random-dot patterns; yellow arrows indicate signal motion directions): single (left) and transparent (right) motion. WM, Working memory.

Figure 2.

Feature-based attention in model simulations for single motion stimulus. A, Network activity for an unattended (left) and an attended trial (right). x-Axis, Time; y-axis, neurons labeled by preferred direction θ_pref. Activity is color-coded. C, Cue period; D, delay period; T, test period. Calibration, 1 s. B, Activity of a neuron with θ_pref = θ_S. Top, Sample membrane potential; middle, spike trains in several trials; bottom, trial-averaged activity (red, attended; black, unattended trials) (calibration: time, 1 s; voltage, 50 mV; rate, 40 Hz). C, Selective enhancement of MT population activity. The scheme (top) depicts how the curves (attended in red; unattended in black) were generated: for fixed test stimulus θ_S and attended feature θ_A, the activity of all neurons (blue arrows) were measured. D, Smoothed modulation ratio (firing rate with attention divided by that without attention) plotted against the distance between attended feature and neuronal preference.

Figure 3.

Multiplicative modulation of tuning curves. A, Tuning curve of an MT neuron (blue arrow in top scheme) computed from responses to the distributed range of test stimuli θ_S (black solid arrows) as attention is focused onto the preferred direction (att pref, squares) or diametrically opposed (att null, circles) or is not presented (no att, diamonds). B, Rescaling of tuning curves (factors, 0.73 for att pref, 1.28 for att null) reveals multiplicative scaling. C, Power-law relationship between the response R and the external input (sum of bottom-up I_S and top-down I_A). Data for the three cases (no att, att pref, att null) were fitted with r = a(I_S + I_A + c)^b (blue curves); a and b were found to be independent of the attentional condition (a = 1.11 Hz/nA^b, b = 3.97), whereas c was smaller for attention than for nonattention conditions (0.42 and 0.59 nA, respectively).

Figure 4.

Attentional processing with transparent motion: the model conforms with the feature-similarity gain principle. A, MT population activity for trials with transparent motion presented during the test period. B, Test period MT activity in the attended condition (green) is well predicted by the product of the unattended activity (gray) and the modulation ratio (Fig. 2D). C, Same as B for baseline activity in the delay period (no sensory stimulus). D, Modulation ratio for single motion (red), transparent motion (green), and lack of sensory stimulation (orange).

Figure 5.

Neural responses to attended transparent motion: the model shows biased competition phenomenology. A, Dynamics of the activity of a cell at the onset of unattended test stimuli (black, preferred single motion; light gray, nonpreferred single motion; dark gray, preferred plus nonpreferred transparent motion). B, Same for responses to transparent motion in the unattended case (dark gray, same as in A) and in the attended cases (dark green, attention on preferred; light green, attention on nonpreferred). C, Mean activity during the test period for the cell under the conditions in A shows sensory competition. D, Same for the conditions in B reveals attentional bias of competitive responses.

Figure 6.

Mechanisms of the feature-similarity gain principle and biased competition in the model. Left, MT population activity for single and transparent motion in unattended condition (inset, activity of a single neuron; colors are as in Fig. 5C). Middle, MT population activity for single motion in attended and unattended conditions (inset, single neuron responses to transparent motion; colors are as in Fig. 5D). Right, Rescaled MT tuning curves as in Figure 3B. Modulation ratios for single (red), and transparent (green) motion in left inset. A, Control data (from Figs. 2–5). B, Reduced external Poisson input (right inset, f–I curves turn linear). Parameter modifications were G_EE^PFC→MT = 0.012 nS, G_EI^PFC→MT = 0.005 nS, I₀^E = 2 nA, I₁^E = 0.05 nA, I₀^I = 0, I₁^I = 0, g_ext,E = 1 nS, and g_ext,I = 1 nS. C, Blockade of bottom-up and local excitatory inputs to MT interneurons. Parameter modifications were G_EEPFC→MT = 0.134 nS, I₀^E = 0.225 nA, I₀^I = 0, I₁^I = 0, G_EI,AMPA^MT = 0, G_EI,NMDA^MT = 0, and g_ext,I = 7.12 nS. D, Blockade of top-down and local excitatory inputs to MT interneurons. Parameter modifications were G_EE^PFC→MT = 0.061 nS, G_EI^PFC→MT = 0, I₀^E = 0.9 nA, I₁^E = 0.18 nA, G_EI,AMPA^MT = 0, G_EI,NMDA^MT = 0, and g_ext,I = 7.12 nS. When not indicated, axis scales are as in corresponding control case (A). Framed graphs show departure from these qualitative behaviors.