jake_slides (1)

jake_slides (1) - Attention improves performance primarily...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Attention improves performance primarily by reducing interneuronal correlations Marlene R Cohen & John H R Maunsell Visual attention can improve behavioral performance by allowing observers to focus on the important information in a complex scene. Attention also typically increases the firing rates of cortical sensory neurons. Rate increases improve the signal-to-noise ratio of individual neurons, and this improvement has been assumed to underlie attention-related improvements in behavior. We recorded dozens of neurons simultaneously in visual area V4 and found that changes in single neurons accounted for only a small fraction of the improvement in the sensitivity of the population. Instead, over 80% of the attentional improvement in the population signal was caused by decreases in the correlations between the trial-to-trial fluctuations in the responses of pairs of neurons. These results suggest that the representation of sensory information in populations of neurons and the way attention affects the sensitivity of the population may only be understood by considering the interactions between neurons. The responses of sensory neurons are variable, and laboratory studies typically deal with this variability by averaging responses to many stimulus presentations. In the real world, however, people and animals must respond to individual stimulus events, and the brain is thought to compensate for neuronal variability by encoding sensory information in the responses of large populations of neurons. To understand the way sensory information guides behavior in everyday life, we need to understand the way information is encoded in populations of neurons. One way to identify the important aspects of a population code is to look at the differences between the neuronal representation of a sensory stimulus when it is used to guide behavior and when it is behaviorally irrelevant. Tasks that control attention provide a powerful way to manipulate behavioral relevance. Attention allows observers to select the most important stimuli and greatly improves perception of the attended location or feature. Attention modulates the firing rates of sensory neurons, typically increasing responses to attended stimuli1–4. This increased rate of firing acts to improve the 5,6 the population is read out8,9, but its effect can be far greater than the effect of independent variability of single neurons. If the noise in individual neurons is independent, averaging the responses of many neurons will lead to a very accurate estimate of the mean, no matter how noisy the individual neurons are. If, however, there are positive correlations in the trial-to-trial fluctuations of the responses of pairs of neurons, then the shared variability can never be averaged out, leading to a more variable (and less accurate) estimate of the mean activity in the population10–12. We found that attention adaptively decreased correlated variability in a population of neurons in visual area V4 in a change-detection task. Furthermore, we found that this decrease accounted for the vast majority of the attentional improvement in the amount of sensory information encoded by the population and is probably the major contributor to the improved psychophysical performance. These results indicate that studies that focus on a single neuron, which necessarily ignore interactions between neurons, miss the most critical aspect of the way sensory information is encoded in populations Noise in a Population of Neurons • Independent Noise • A Problem? • Shared Variability • Effects on population sensitivity? Center of Receptive Fields Performance Curve Good Things To Know MISomething = Attended In - Attended Out Sum Fano Factor = Variance . Mean Synchrony • Is attention related to synchrony levels? • In this experiment, only chance levels of synchrony in both attended and non-attended response. • Synchrony was not different in attended and unattended conditions. (Paired t Test P = .46) Attentional effects are larger on strongly responsive neurons Population Sensitivity Discriminability between the distributions of responses to the original and changed orientations. d’ is the difference in the means divided by their root mean square s.d. Attentional Modulation Simulation • Method - Use attentional modulation of the factor of interest with values observed in the unattended conditions for the other two factors. • Results • Noise Correlation - 79% • Firing Rate - 9% • Independent Noise - 4% COGS 272 Week 3:V4 .27:611-647. Downloaded from arjournals.annualreviews.org fornia - San Diego on 04/02/10. For personal use only. 618 REYNOLDS CHELAZZI Figure 3 Contrast-dependent response modulations. (A) Contrast-response functions for a Somers et al. 1995 The Journal Feedforward Inhibitory of Neuroscience, August 1995, 7~178) 5449 Recurrent Figure 1. Models o f visual cortical orientation selectivity. a, In feedforward models all “first-order” cortical neurons (triangle, excitatory; hexagon, inhibitory) receive converging input (gray arrow) from a population o f LGN neurons that cover a strongly oriented region o f visual space. The bandwidth or sharpness o f a cortical cell’s orientation tuning is determined b y the aspect ratio o f its LGN projection. b, Many inhibitory models employ a mild feedforward bias to establish the initial orientation preference o f cortical neurons and utilize inhibitory inputs (white arrows), from cortical neurons preferring different orientations, to suppress nonpreferred responses. Here, we present a model, c, in which recurrent cortical excitation (hluck arrows)among cells preferring similar orientations, combined with iso-orientation inhibition from a broader range o f orientations, integrates and amplifies a weak thalamic orientation bias, which is distributed across the cortical columnar population. (Ferster, 1986; Douglas et al., 1991a; for a differing view, see Pei et al., 1994). Such inhibitory tuning conflicts with crossorientation inhibitory models (Bishop et al., 1973; Morrone et al., 1982), and strong iso-orientation suppression poses problems even for models that use other forms of hyperpolarizing inhi- bition to sharpen thalamocortical input (e.g., Wiirgotter and Koch, 1991). Furthermore, recent results from our laboratory layer IV circuitry under a 1700 pm by 200 p,m patch of cortical surface and was composed of more than 3000 spiking neurons with over 180,000 synapses.Cortical excitatory and inhibitory neurons were modeled separately using intracellular parameters from regular-spiking and fast-spiking neurons (Connors et al., 1982; McCormick et al., 1985). Ret- Somers et al. 1998 Figure 6. Local circuit module construction. Behavior of a local population M in the full model is estimated by constructing a local circuit module composed of excitatory and i nhibitory populations. (A) The module s implifies analysis of RF integration by capturing nonlinear local cortical interactions over a radius R. The module is constructed by replacing local inputs (from the gray shaded area between M and R with local inputs from within M. (B) Activity in the module depends o nly on thalamocortical and long-range intracortical excitatory inputs to the module. All s ources of long-distance input (possibly including feedback from higher cortical areas) are summed together linearly, but inputs to e xcitatory and inhibitory cells remain segregated. Simulations of the local circuit module are much faster than those of the full model and thus permit exploration of local circuit responses to all possible input values. odule construction. Behavior of a local population M in the full model is estimated by constructing a local circuit module composed of excitatory and The module simplifies analysis of RF integration by capturing nonlinear local cortical interactions over a radius R. The module is constructed by replacing shaded area between M and R with local inputs from within M. (B) Activity in the module depends only on thalamocortical and long-range intracortical dule. All sources of long-distance input (possibly including feedback from higher cortical areas) are summed together linearly, but inputs to excitatory and regated. Simulations of the local circuit module are much faster than those of the full model and thus permit exploration of local circuit responses to all .27:611-647. Downloaded from arjournals.annualreviews.org fornia - San Diego on 04/02/10. For personal use only. 618 REYNOLDS CHELAZZI Figure 3 Contrast-dependent response modulations. (A) Contrast-response functions for a Troyer et al. 1998 Figure 6. ! Behavior of model using correlation-based connectivity. Schematic representing behavior of the model in response to preferred (A) and null (B) stimuli. The excitatory cell described in Results is in the top left; its inhibitory antiphase partner is in the bottom right. E, Excitatory cells; I, inhibitory cells. Solid lines represent excitation and depolarization; open lines represent inhibition and hyperpolarization. Line thickness and size of RF icon represent magnitude of activity. Dashed lines represent correlation-based excitation, which is included in the complete computational model only (see Figs. 8-11). Some simulations were performed without cortical excitatory projections onto inhibitory neurons (gray dashed lines), but this did not substantially affect network behavior (see Fig. 13B). .27:611-647. Downloaded from arjournals.annualreviews.org fornia - San Diego on 04/02/10. For personal use only. 618 REYNOLDS CHELAZZI Figure 3 Contrast-dependent response modulations. (A) Contrast-response functions for a Carandini et al. 1997 Fig. 1. B, The normalization model extends the linear model by adding a divisive stage. The linear stage feeds into a circuit composed of a resistor and a capacitor in parallel (RC circuit). The conductance of the resistor grows with the pooled output of a large number of cortical cells. This effectively divides the output of the linear stage. Fig. 6. The ratio of Gaussians (RoG) model. We constructed the RoG model from independent and spatially stable center and surround components. We modeled each component as a Gaussian envelope of sensitivity incorporating the spatiotemporal tuning characteristics of a neuron. Each component had an independently controlled gain, and the surround affected the model cell's response through divisive suppression. The model is instantiated in Eq.!9 through 11. Cavanaugh et al. 2002 .27:611-647. Downloaded from arjournals.annualreviews.org fornia - San Diego on 04/02/10. For personal use only. 618 REYNOLDS CHELAZZI Figure 3 Contrast-dependent response modulations. (A) Contrast-response functions for a Reynolds et al. 1999 Figure 2. ! Model circuit diagram. The circle on top represents the neuron being recorded. The variable y is the firing rate of this neuron. The two circles at the bottom of the diagram represent populations of "input" neurons that respond to the reference (left) and probe (right) stimuli and that project to the measured cell. The average response of the ith input population is designated xi. Black lines indicate the excitatory projections from each input population to the measured cell, and gray lines indicate the inhibitory projections, which are assumed to depend on inhibitory interneurons (not shown). The variable wi+ is the magnitude, or weight, of the excitatory projection from the ith input population, and wi is the weight of the inhibitory projection from the ith input population. For a complete description of the model, see Materials and Methods. .27:611-647. Downloaded from arjournals.annualreviews.org fornia - San Diego on 04/02/10. For personal use only. 618 REYNOLDS CHELAZZI Figure 3 Contrast-dependent response modulations. (A) Contrast-response functions for a ...
View Full Document

{[ snackBarMessage ]}

Ask a homework question - tutors are online