Pain Affect Encoded in Human Anterior Cingulate But Not Somatosensory Cortex

See allHide authors and affiliations

Science  15 Aug 1997:
Vol. 277, Issue 5328, pp. 968-971
DOI: 10.1126/science.277.5328.968


Recent evidence demonstrating multiple regions of human cerebral cortex activated by pain has prompted speculation about their individual contributions to this complex experience. To differentiate cortical areas involved in pain affect, hypnotic suggestions were used to alter selectively the unpleasantness of noxious stimuli, without changing the perceived intensity. Positron emission tomography revealed significant changes in pain-evoked activity within anterior cingulate cortex, consistent with the encoding of perceived unpleasantness, whereas primary somatosensory cortex activation was unaltered. These findings provide direct experimental evidence in humans linking frontal-lobe limbic activity with pain affect, as originally suggested by early clinical lesion studies.

Affective aspects of pain, such as perceived unpleasantness, have been classically considered to be distinct from the simple sensory dimensions of pain, which include the perception of location, quality, and intensity of noxious stimulation (1). Largely on the basis of indirect evidence, separate neuronal pathways have been postulated to underlie these different aspects of the pain experience (2). For example, involvement of frontal lobe regions, particularly the anterior cingulate cortex (ACC), in pain affect is suggested by clinical reports that patients with frontal lobotomies or cingulotomies sometimes still feel pain but report it as less distressing or bothersome (3). On the other hand, primary and secondary somatosensory cortices (SI and SII) have been considered plausible candidates for the processing of sensory-discriminative aspects of pain, on the basis of their anatomical connections to subcortical and spinal regions, which encode discriminative properties of somatosensory stimuli (4). Recent neuroimaging studies in humans documented pain-related activation in limbic sites, such as ACC and rostral insula (IC), and in the primary sensory regions SI and SII (5). In addition, anatomical and electrophysiological data show that these regions receive direct nociceptive input in the monkey (6). However, the extent to which these different cortical structures contribute to specific dimensions of the human pain experience is largely unknown and untested.

In the present study we used hypnosis as a cognitive tool to reveal possible cerebral mechanisms of pain affect in normal human volunteers. A perceptual dissociation of sensory and affective aspects of the pain experience was achieved with hypnotic suggestions to both increase and decrease pain unpleasantness, without changing the perceived intensity of the pain sensations (7). Cerebral cortical activity related to this perceptual dissociation was measured by positron emission tomography (PET) (8).

PET scans were conducted during conditions of alert control, hypnosis control, and hypnotic suggestion for increased unpleasantness (↑UNP) or decreased unpleasantness (↓UNP) (9). During each scan tonic stimuli were presented to the left hand by passive immersion in “neutral” (35°C) or “painfully hot” (47°C) water (10). After each scan, the perceived intensity and unpleasantness of the stimulation were rated by the participant (11).

Regional cerebral blood flow (rCBF) was measured with three-dimensional high-resolution PET after H2 15O bolus injection (12). Each participant also received a high-resolution magnetic resonance imaging (MRI) anatomical brain scan that was used for alignment and transformation of PET volumes into the Talairach coordinate system (13). To obtain volumes of pain-related changes in rCBF for each participant, we subtracted normalized PET data recorded during the “neutral” condition from those of the “painfully hot” condition. Resulting volumes of pain-related changes in rCBF were averaged across sessions, and statistical activation maps were derived on the basis of the methods of Worsleyet al. (14). Directed searches of rCBF increases were conducted on right (contralateral to stimulus) SI, SII, ACC, and IC to confirm pain-related activation of these structures and to test the hypothesis that changes in pain unpleasantness modulate activity only within limbic regions thought to be involved in affective processes. The threshold for statistical significance was corrected for multiple comparisons (15).

Results of “painfully hot” versus “neutral” subtractions from scans taken during the alert control condition support previous findings of significant pain-related activations in SI, SII, IC, and ACC (Table 1). After hypnotic induction, but before suggestions of increased or decreased unpleasantness, painful heat again activated these four cortical areas (Table 1), indicating little influence of hypnotic induction itself on pain-related activation. Similarly, the hypnotic induction had no significant effect on psychophysical ratings of either pain intensity or unpleasantness (alert control compared with hypnosis control: intensity, 77.6 ± 14.7 and 75.0 ± 14.0, and unpleasantness, 61.4 ± 28.8 and 54.8 ± 25.8, respectively).

Table 1

Pain-related activation sites within SI, SII, ACC, and IC. Coordinates are given in Talairach space (13); Lateral, anterior, and superior are relative to midline, anterior commisure, and commissural line, respectively (positive values are right, anterior, and superior). A t statistic of 2.55 is equivalent to P = 0.05 (15).

View this table:

Hypnotic suggestions for increased or decreased unpleasantness, on the other hand, altered both the perception of pain affect and the activation within some but not all of these pain-related cortical regions. A comparison of rCBF changes between hypnotic suggestion and hypnosis control conditions revealed significant pain-related activations in SI, ACC, and IC, during both ↑UNP and ↓UNP conditions. Within the vicinity of SII, however, no significant pain-evoked activity was observed in either the ↑UNP or ↓UNP conditions (Table 1). One possible explanation for this absence of pain-evoked activity in SII is that the mental effort or attention demanded by these suggestions may suppress such activation. Alternatively, there may have been an habituation of SII activity with repeated stimulation.

The effectiveness of hypnotic suggestions in selectively altering pain affect is demonstrated by the significant difference observed only in the participants' ratings of unpleasantness during the ↑UNP and ↓UNP conditions [unpleasantness: 81.4 ± 14.6 and 45.0 ± 25.8, respectively, analysis of variance (ANOVA) P < 0.001; intensity: 78.0 ± 14.6 and 71.2 ± 18.2, respectively, statistically not significant]. In parallel with this modulation in pain affect, direct volume-of-interest (VOI) comparisons (16) of the three pain sites activated during the hypnotic suggestion conditions revealed significantly greater activation during ↑UNP scans, compared with that observed during ↓UNP, only in ACC (P < 0.02; see Fig. 1). In SI, pain-related rCBF was actually lower (nonsignificantly) in the ↑UNP condition than in the ↓UNP condition, indicating no tendency for increased activation in this area related to increased unpleasantness.

Figure 1

Changes in pain-related activity associated with hypnotic suggestions of high and low unpleasantness (left and right images, respectively) are revealed by subtracting PET data recorded during the neutral/hypnosis control condition from those of the painfully hot/↑UNP and painfully hot/↓UNP conditions. PET data, averaged across 11 experimental sessions, are illustrated against an MRI from one person; horizontal and saggital slices through SI and ACC, respectively, are centered at the activation peaks observed during the relevant suggestion conditions; red circles indicate the location and size of VOIs used to analyze activation levels across the two conditions (16).

To test the strength of the relation between pain affect and activation within the ACC, we did regression analyses of unpleasantness ratings and rCBF levels across all participants and all scans taken during the hypnotic suggestion condition for each pain activation site. After removing effects due to interperson variability, perceived intensity, and scan session, the residual variance in rCBF [analysis of covariance (ANCOVA)] demonstrates that only activation levels within the ACC (Fig. 2) are consistent with the encoding of the perceived unpleasantness of these noxious stimuli (ACC: Pearson's correlation coefficient, r, = 0.419, P = 0.005; IC: r = 0.245,P = 0.134; SI: r = –0.224,P = 0.149).

Figure 2

Activation levels (as measured by residual rCBF) observed within the ACC during high (red) and low (yellow) unpleasantness conditions are significantly correlated with ratings of pain unpleasantness (ANCOVA: r = 0.42,P = 0.005). The coordinates of this r-value peak (inset: saggital section of PET regression volume—lateral, +7; anterior, +20; superior, +29) lie precisely within the region of ACC that was activated in both ↑UNP and ↓UNP conditions (see Fig. 1 and Table 1). Line shows linear best fit.

These results demonstrate a modulation of pain-related activity in ACC that closely parallels a selective change in the perceived unpleasantness of painful stimuli. The absence of changes in the sensory component of pain perception and the lack of similar modulation within other pain-related cortical structures argue for a significant involvement of the ACC in the affective component of pain. Such findings support earlier proposals that the anterior cingulate gyrus is integrally involved in pain and emotions (3, 5, 6), but our findings go beyond these general ideas by providing direct evidence of a specific encoding of pain unpleasantness in the ACC.

We propose that pain-related activation in ACC reflects a nociceptive input from a highly modifiable pain pathway (17), and that the level of pain-evoked ACC activation is determinant in the individual's emotional and behavioral reactions to pain. The proximity of the nociceptive, motor, and attentional regions of ACC (18) suggests possible local interconnections that might allow the output of the ACC pain area to command immediate behavioral reactions. Similarly, the ACC pain area might participate in the substantial interconnections between the ACC and the “fight or flight” regions of the midbrain periaqueducal gray matter (19).

The anatomical connections between ACC, IC, SI, and SII (20) suggest that these regions do not function independently in encoding different aspects of pain but are highly interactive. Such interactions are reflected in the experiences of pain itself. For example, pain intensity, location, and quality (sensory features) are major factors in determining unpleasantness (21). Nevertheless, despite these associations, there appears to be at least a partial segregation of function between pain affect and sensation, with ACC activity possibly reflecting the emotional experience that provokes our reactions to pain.

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


Stay Connected to Science

Navigate This Article