Cholinergic Enhancement and Increased Selectivity of Perceptual Processing During Working Memory

See allHide authors and affiliations

Science  22 Dec 2000:
Vol. 290, Issue 5500, pp. 2315-2319
DOI: 10.1126/science.290.5500.2315


Using functional magnetic resonance imaging, we investigated the mechanism by which cholinergic enhancement improves working memory. We studied the effect of the cholinesterase inhibitor physostigmine on subcomponents of this complex function. Cholinergic enhancement increased the selectivity of neural responses in extrastriate cortices during visual working memory, particularly during encoding. It also increased the participation of ventral extrastriate cortex during memory maintenance and decreased the participation of anterior prefrontal cortex. These results indicate that cholinergic enhancement improves memory performance by augmenting the selectivity of perceptual processing during encoding, thereby simplifying processing demands during memory maintenance and reducing the need for prefrontal participation.

Working memory (WM) (1) is mediated by a widely distributed neural system in the human brain. Modulation of cholinergic neurotransmission alters memory function, including WM. Performance on WM tasks is improved by pharmacologic agents that enhance cholinergic function and is impaired by agents that block cholinergic function (2–4). The mechanism by which cholinergic modulation alters WM, however, is unclear.

Different regions in the distributed neural system for WM support dissociable cognitive subcomponents of this complex function. With functional magnetic resonance imaging (fMRI) the neural activity associated with WM can be decomposed into task subcomponents that are separated in time and space, namely the responses during perceptual encoding, activity during memory delays, and the responses during recognition testing (5, 6).

Prefrontal cortex plays a central role in maintaining and manipulating the contents of WM (7, 8), but previous work suggests this region is not the site where cholinergic enhancement improves processing efficiency. Cholinergically mediated improvement in WM performance is correlated with reduced activity in right prefrontal cortex (9, 10). Reduced activity in this region likely reflects reduced WM load or task difficulty (11, 12) that is a consequence of increased efficiency in other regions of the WM system.

Here, we show that cholinergic enhancement of WM is associated with increased selectivity of responses during perceptual processing in visual extrastriate cortices, particularly during encoding, and redistributed memory delay activity, suggesting that processing demands for WM maintenance were simplified by the production of a more robust visual percept during encoding.

Seven young healthy subjects participated in a double-blind, placebo controlled, cross-over study with two fMRI sessions on separate days, one during a steady-state infusion of physostigmine and one during an infusion of saline (13, 14). During scanning, subjects performed a task that consisted of alternating face WM and sensorimotor control items (Fig. 1).

Figure 1

The WM for faces task. For each scan series, subjects performed a task that alternated between a sensorimotor control item and a WM item. For each WM item, a picture of a face was presented for 3 s, followed by a 9-s delay, and by a 3-s presentation of two faces. Subjects indicated which of the two faces they had seen previously by pressing response buttons with the right or left thumb. For each sensorimotor control item, identical scrambled faces were presented in the same spatial and temporal manner as in the WM item. Scrambled faces were used to control for spatial frequency, brightness, and contrast. Subjects were informed that there was no memory component for this portion of the task and instructed to press both buttons simultaneously when shown two scrambled faces. Different scrambled faces were used across trials. Items were separated by 9-s intertrial intervals.

We measured cortical responses to different subcomponents of the task and identified voxels with significant responses to visual stimuli or significant memory delay activity (5, 15). Visual responses were classified as face-selective if the response to faces was greater than the response to control stimuli and were classified as encoding-selective if the response during face encoding was greater than the response during recognition testing.

Performance of this WM task during saline infusion evoked activity in a widely distributed, bilateral set of cortical regions (16). Figure 2illustrates the locations of the posterior regions in one subject and shows the mean time series, averaged over subjects (n = 7, for all regions), hemispheres, and trials, during physostigmine and saline infusions. Table 1 shows the size of each of the effects of interest in these ventral and dorsal visual regions. The magnitude of each contrast was measured for the subset of voxels within each region that showed significance on that contrast during either session. The percentage of each posterior region that showed the response of interest is indicated in Table 2.

Figure 2

Examples of ventral and dorsal extrastriate visual areas that were activated in the WM task and time series data from the two experimental sessions. Three axial slices from a single representative subject are shown at the top of the figure with the voxels that showed a significant response to any component of the WM task shown in color. Arrows indicate the locations of ventral occipital (A), ventral temporal (B), dorsal occipital (C), and intraparietal (D) regions. The four panels on the bottom of the figure show time series averaged across subjects, hemispheres, and all trials for the voxels that showed significant face-selectivity [(A), (B), and (D)] or encoding-selectivity (C). The figures show percent change in signal from baseline. The light gray bars indicate when the control stimuli (scrambled faces) were presented and the dark gray bars illustrate when the memory stimuli (faces) were presented. Data acquired during placebo (red) and during physostigmine (blue) are shown in each panel.

Table 1

Effect sizes and the significances of differences between effect sizes during placebo and drug infusion in dorsal and ventral extrastriate regions. Effects are contrasts between responses to different subcomponents, expressed as differences in percent response: Nonselective = responses to visual stimuli minus responses to a blank screen; Face-selective = responses to faces minus responses to control stimuli; Encoding-selective = responses during face encoding minus responses during face retrieval; Memory delay = activity during the memory delays minus activity during the control delays. Asterisks indicate effects that were significantly altered by cholinergic enhancement: *, P < 0.05; **,P < 0.01.

View this table:
Table 2

Effect of cholinergic enhancement on region volumes in ventral occipital, ventral temporal, and dorsal occipital cortices. For each region, the volumes (cm3) of cortex (±SE) showing a significant overall experimental effect and the percent of the total combined region volumes (±SE) showing significant face-selective, encoding-selective, or memory delay effects are shown for each of the two experimental sessions. Total combined regions (cm3) consisted of voxels that showed a significant overall experimental effect on either placebo or drug and were as follows: ventral occipital, 19.6 ± 3.3; ventral temporal, 20.2 ± 3.8; and dorsal occipital, 12.3 ± 4.1.

View this table:

Infusion of physostimine resulted in enhanced neural processing in visual cortical areas and a trend toward faster reaction times (RT) (P = 0.07) during WM (17). In ventral occipital cortex (Fig. 2A and Table 1), both the face-selectivity and the encoding-selectivity of responses to stimuli were significantly enhanced by physostigmine (P < 0.01 in both cases), as indicated by larger differences between the responses to faces and control stimuli (face-selectivity) and between the responses during encoding and recognition testing (encoding-selectivity). Responses in ventral temporal cortex (Fig. 2B and Table 1) also showed significant enhancement of face-selectivity during physostigmine infusion (P < 0.01) and a nonsignificant trend toward enhanced encoding-selectivity (P = 0.09). Responses in dorsal extrastriate regions of occipital and parietal cortices were of smaller amplitude (compare Fig. 2, A and B, to Fig. 2, C and D) and involved smaller cortical volumes (Table 2) than those observed in ventral extrastriate cortices. Nonetheless, responses in dorsal occipital cortex (Fig. 2C and Table 1), like ventral occipital cortex, showed significant enhancement of both face-selectivity and encoding-selectivity during physostigmine infusion (P < 0.05 in both cases), and responses in the intraparietal sulcus (Fig. 2D and Table 1) showed significant enhancement of face-selectivity (P < 0.05). The increased face-selectivity of responses in ventral occipital cortex resulted from both an increased response to faces (P < 0.05) and a decreased response to control stimuli (P < 0.05), whereas increased face-selectivity in the other regions was due primarily to increased responses to faces.

Enhancement of visual processing by physostigmine also was evident as changes in the volume of cortex that showed significant selectivity or memory delay activity (Table 2) (18). The volume of cortex that showed face-selectivity increased by approximately 50% in ventral occipital, dorsal occipital, and ventral temporal cortex (P < 0.05). Similarly, the volume of cortex that showed encoding-selectivity increased over twofold in these same regions (P < 0.05). These changes were not due to a greater overall volume of activated cortex with physostigmine as the volume of cortex in visual areas that was activated by any component of the face WM task did not change (P > 0.2). The number of voxels in the ventral occipital and temporal cortices that showed significant memory delay activity increased nearly twofold with cholinergic enhancement (P < 0.05), despite the fact that the magnitude of sustained activity over memory delays in these voxels did not increase significantly (Table 1). The volumes of cortex in the intraparietal sulcus showing face-selectivity, encoding-selectivity, and memory delay activity did not change with cholinergic enhancement.

In prefrontal cortex, significant responses were observed bilaterally in the inferior frontal gyrus (n = 4), in the anterior and posterior middle frontal gyri (n = 5 andn = 7, respectively), and in the superior frontal gyrus (n = 3). These regions were bilateral in most cases. When unilateral, there were an equivalent number of right-sided (n = 4) and left-sided (n = 3) regions. Physostigmine induced a reduction of activity during WM in dorsal anterior prefrontal regions (anterior middle frontal and superior frontal gyri, defined by voxels that showed significant memory delay activity, P < 0.01) that was not specific to individual subcomponents but was a general effect across the WM task. By contrast, physostigmine induced a similarly nonspecific increase of activity during WM (P < 0.05) in the inferior prefrontal cortex. In the posterior middle frontal region, physostigmine had no effect on activity during WM task performance. Thus, enhanced processing during physostigmine infusion is associated with reduced participation of dorsal anterior but not inferior prefrontal regions.

The results indicate that enhancement of cholinergic activity improves WM by focusing perceptual processing in extrastriate visual areas on relevant stimuli, particularly during encoding. The difference between responses to memory and control visual stimuli increased in both ventral and dorsal visual extrastriate cortices, with larger increases during encoding. Cholinergic enhancement also increased the volume of cortex showing memory delay activity in ventral extrastriate cortex, although there was no significant effect on the amplitude of memory delay activity. Consistent with our previous findings (9), activity during perceptual processing and memory maintenance in dorsal anterior prefrontal cortex was reduced during cholinergic enhancement.

Cholinergic modulation is known to influence the signal-to-noise ratio for the responses of single neurons in cortex (19, 20). We demonstrated this effect at a cortical systems level, by the increased face-selectivity and encoding-selectivity of neural responses during physostigmine infusion. Signal-to-noise ratio can be improved by increasing signal or by decreasing noise. Our data show increased responses to memory stimuli (i.e., signal), and, in early visual processing areas, decreased responses to control stimuli (i.e., noise;Fig. 2A).

Cholinergic projections of the nucleus basalis of Meynert to the neocortex influence visual attention (21, 22). In single-unit recording studies, enhanced selective attention is demonstrated as an enhanced tuning of neural responses to attended stimuli (23). In a single-unit study in cats, cholinergic enhancement augmented the selectivity of visual neurons to stimulus orientation (24). In functional imaging studies the effect of selective attention is evident as increased responses in extrastriate visual areas that process the attended information (25, 26). Our results indicate that cholinergic enhancement augments selective attention, as evidenced by increased selectivity of perceptual responses. Our results do not identify the mechanism by which attention is modulated. Cholinergic neurotransmission could influence selective attention through a direct effect on systems that control attention or by modulating the effect of input from the control systems on local neural activity in perceptual areas. Specific cellular mechanisms have been suggested (20) by which acetylcholine could modulate the effect of input from attention control systems by increasing the response to afferent input and reducing background activity. Because the effect of physostigmine infusion was an increase in the selectivity of responses to task-relevant stimuli rather than increased responses to all stimuli, our results indicate that the effect of cholinergic enhancement is not a simple increase in alertness or arousal.

Cholinergic enhancement of perceptual processing has a greater effect on memory at the time of encoding than at the time of retrieval. Consistent with this finding, cholinergic antagonists, such as scopolamine, selectively interfere with encoding of new information, but not with retrieval (27–29).

Cholinergic enhancement also redistributed sustained activity during memory delays. In prefrontal cortex, memory delay activity was reduced in dorsal anterior regions and increased in inferior regions. In ventral extrastriate cortex, the volume of cortex that showed significant memory delay activity doubled. Enhanced encoding could account for this redistribution of memory delay activity. Improved encoding can produce a more vivid or distinct visual percept that is easier to maintain in WM as a simple image, mediated by activity in ventral temporal and inferior frontal cortex, as opposed to a representation that requires more executive function to construct and maintain, mediated by activity in dorsal anterior prefrontal cortex (30). Alternatively, cellular mechanisms have been proposed by which cholinergic input could increase sustained neural firing directly (31).

Deterioration of the cholinergic system contributes to memory failure and cognitive decline in Alzheimer's disease and also may play a role in the more benign memory changes associated with healthy aging (32, 33). The main pharmacological agents used to counteract cognitive dysfunction in patients with Alzheimer's disease are cholinesterase inhibitors, similar to physostigmine but with a longer plasma half-life (34). Our results suggest that the enhancement of memory performance induced by these cholinergic agents is mediated primarily by more selective perceptual processing during encoding.

  • * To whom correspondence should be addressed. E-mail: furey{at}


View Abstract

Navigate This Article