Attention modulates the strength of neuronal responses

The most prominent effect of attention on the activity of neurons in visual cortex is a change in the strength of their response to a stimulus. Typically, a neuron's response to its optimal stimulus is stronger when the subject pays attention to it. Most observations suggest that attention alters the sensitivity of neurons without affecting their stimulus preferences or selectivity, although exceptions will be discussed.

Range of Attentional Effects within and between Cortical Areas

Attention has been found to modulate neuronal responses in every visual area examined, from V1 to the latest stages in the cortical hierarchy. However, the magnitude of attentional modulation varies greatly within and between areas and appears to be related to several factors.

Samples of neurons from individual visual areas invariably show a range of modulation by attention, even when the stimulus is optimized for each neuron. Some neurons respond robustly when the stimulus is attended and not at all when it is ignored. Others show no detectable effect of attention. Furthermore, occasional neurons have been found to have significant negative modulation by attention, so that they respond more strongly when a stimulus is ignored.

The reason for this range of effects is not understood. It is possible that neurons with response properties best suited to the current behavioral task are modulated most by attention. For example, when subjects perform a color discrimination task, neurons with greater color selectivity, or with color preferences well matched to the stimuli, might be more modulated by spatial attention than other neurons. However, no such correlation has been described, and the reason for the range of attentional modulations within cortical areas remains an important unanswered question.

Modulation by attention also differs between visual areas. The average strength of modulation is generally stronger in later stages of visual cortex. In V1, the average effect is weak, typically about a 10% increase in response (Crist et al., 2001; Haenny and Schiller, 1988; Luck et al., 1997; McAdams and Maunsell, 1999a; Mehta et al., 2000; Motter, 1993). Stronger effects are found in extrastriate areas. It is difficult to compare modulations from different studies quantitatively because they differ in stimulus conditions and task difficulty (see below), but some investigations have measured attentional modulation in different visual areas in the same subjects using consistent task conditions. Those studies typically show greater modulations in successive stages of cortical processing. Figure 103.1 plots the averages for attentional modulation from experiments that measured responses in more than one visual area under equivalent conditions. Each symbol represents data from a different experiment. More modulation was found in later levels of visual cortex.

Figure 103.1..  

Attentional enhancement of neuronal responses increases as a function of position in the cortical hierarchy. Average attention modulation was taken from experiments that measured responses in at least two cortical levels using the same subjects and the same task. Open squares are data from McAdams and Maunsell (1999a); open circles are from Treue and Maunsell (1999); filled squares are from Ferrera et al. (1994); and filled circles are from Cook and Maunsell (2002). The cortical hierarchy levels are spaced according to the levels defined by Felleman and Van Essen (1991). (From Cook and Maunsell, 2002.)

It is not known why attentional modulation should be stronger in later stages of visual cortex. One possibility is that attention acts to insert additional spikes at each level of visual cortex and that this additional activity accumulates, so that attentional modulation is greater at later stages. An alternative explanation is that attention acts to insert activity at the highest levels of visual cortex, and that this activity diminishes as it descends along feedback pathways. While either explanation might account for stronger modulations in later stages of cortex, they seem unlikely because they suggest that signals must grow or diminish uncontrollably as they propagate through cortex. No evidence suggests that sensory signals grow or diminish uncontrollably in this way. It seems more likely that there is an optimal amount of attentional modulation for sensory signals and that different optima exist for earlier and later sensory representations. Whatever the source, the reason for different degrees of modulation in different cortical areas is one of the most important questions to be answered about attentional modulations.

One consequence of the differences in attentional modulation between visual areas is to encourage investigators to focus more on certain areas. For example, area V4 has been extensively examined, in part because it is well positioned. It is early enough in cortical processing that its neurons respond well to relatively simple stimuli that are easily controlled and manipulated, such as oriented bars or patches of color. It is also sufficiently advanced in cortical processing that attention modulations are fairly prevalent and easy to demonstrate. However, while V4 is a frequent subject for study in attention experiments, there is little reason to believe that it plays a special role in processes related to attention.

Effect of Task Difficulty on Attentional Modulation

Another factor affecting the magnitude of attentional modulation is task difficulty. Stronger responses are generally seen when animals must pay more attention to a stimulus (Spitzer et al., 1988). For example, responses of inferotemporal neurons grow progressively stronger as an animal goes from ignoring a stimulus, to attending it to detect a dimming, to attending it to discriminating its pattern (Spitzer and Richmond, 1991).

Changes in responses associated with task difficulty may be associated with vigilance, and therefore may be less specific than effects seen with attention to spatial locations or stimulus features. However, task difficulty can influence the magnitude of selective modulations associated with spatial attention. Modulation by spatial attention can be much stronger when animals perform a difficult version of a task in which they must detect changes in orientation (Boudreau and Maunsell, 2001).

Differences in a subject's level of effort cannot easily explain differences in attentional modulation between different cortical areas because in most relevant cases the same stimulus and task have been used in both areas, leaving no obvious basis for the subject to devote different levels of effort when recordings were made from different areas. Differences in effort might explain some of the range of modulations seen within areas, because the subject's effort might differ while data are collected from different neurons. However, an experiment that compared neuronal and behavioral effects of spatial attention found that the modulation of individual neurons by attention was uncorrelated with attentional modulation of behavior during the period when the neuron's responses were recorded (Cook and Maunsell, 2002). This suggests that the range of neuronal modulation within areas comes largely from sources other than vigilance.

Because task difficulty affects attentional modulation, the values reported by various studies cannot be taken as absolute, and comparisons between different experiments are problematic. Although the data in Figure 103.1 suggest that different experiments will find comparable average enhancement for a given visual area, those data are not representative of the differences between experiments because they were collected in the same laboratory using tasks of similar difficulty.

The influence of the subject's effort also raises the question of whether much stronger modulations might be seen with particularly challenging tasks. Although some attention experiments have used near-threshold stimuli, many have used stimuli that were unlikely to have required much effort from the subjects. It is notable that particularly strong modulations were found by Mountcastle and his colleagues (1981), who compared responses during a visual task with responses to the same stimuli while the animals sat idle. Little work has been done to explore the limits of attention modulation. For example, responses to a visual stimulus might be much stronger when that stimulus is the subject of a challenging discrimination than to when it is irrelevant and attention is directed to a difficult auditory or somatosensory discrimination.

Effect of Attention When No Stimulus Is Present

The effects of attention are often examined in terms of responses to visual stimuli, but attention can also affect neuronal activity when no stimulus is present. In early stages of visual cortex, modest increases in the spontaneous activity of neurons have been described when attention is directed to the part of the visual field containing a neuron's receptive field (Luck et al., 1997). This increased activity associated with spatial attention appears equivalent to increases in activity reported in the later stages of visual cortex, which are usually discussed in the context of short-term memory. For example, some inferotemporal neurons maintain an elevated level of activity during periods when the subject remembers a particular form or color, and not during periods when other stimuli must be remembered (Fuster, 1990; Fuster and Jervey, 1982; Miyashita et al., 1993; Yakovlev et al., 1998). Similarly, neurons in the lateral intraparietal area (LIP) remain active when animals pay attention to the motion of a stimulus that is briefly hidden from view (Assad and Maunsell, 1995; Eskandar and Assad, 1999). Thus, attentional modulations apply to undriven activity as well as stimulus-driven activity.

Some studies have not detected changes in undriven activity associated with attention. Because these changes are usually small (typically a few spikes per second), this may be a matter of statistical power. Alternatively, aspects of task design may make these changes less likely. McAdams and Maunsell (1999a) examined attentional modulation of driven and undriven activity in V4 and found that while responses to stimuli were modulated, activity immediately before the appearance of the stimulus was not affected by the location to which attention had been directed. The absence of modulation of undriven activity in this case may have been due to the timing of the stimulus presentations. Because the stimuli always appeared at predictable times and stayed on for relatively long periods, the animal had little motivation to direct its attention to the cued location until after the stimulus appeared. Consistent with this, average attentional modulations increased during the stimulus presentation. Motter (1994) examined how quickly animals redirected their attention to different parts of the visual field after an instruction was given, and found that neuronal responses changed within a few hundred milliseconds. The dynamics with which attention can shift have not been well characterized, and there is no reason to believe that attention will be fixed during a behavioral trial or even during the course of an individual stimulus presentation.