Posthypnotic Effects on Value-Based Decision-Making
Posthypnotic Effects on Value-Based Decision-Making
Bids per condition are shown in Figure 2A. For all subjects together, there was a main effect of snack type, F = 16.095, P < 0.001, η = 0.29, as participants bid less for target snacks than for non-target snacks. There was also a main effect of cue, F = 17.686, P < 0.001, η = 0.02, as participants bid less for snacks during cueON-runs than during cueOFF-runs. Finally, there was an interaction of cue × snack type, F = 22.953, P < 0.001, η = 0.04, as the difference between bids for target snacks and non-target snacks was bigger during cueON-runs than during cueOFF-runs. Furthermore, there were no interactions of group × snack type, group × cue or group × snack type × cue (all P-values >0.50), and also no main effect of group (P = 0.16). Hence, the effects for bids did not differ between groups.
(Enlarge Image)
Figure 2.
Behavior and self-report. (A) Bids. (B) Postexperimental disgust ratings. (C) Postexperimentally reported degree of physicality of disgust, experience of self-control (as opposed to automaticity) regarding the feeling of disgust, pretending of disgust and conscious recall of suggestion/instruction during the experiment. Ninety-five percent confidence interval (CI) of the mean are shown, adjusted for within-subject designs where appropriate Loftus and Masson, (1994). T: target snacks (sweet or salty). NT: non-target snacks. CueON: cue color shown in the background. CueOFF: neutral background color. ME: Main effect. IE: Interaction effect. ***P < 0.001, **P < 0.01, *P < 0.05; n.s., not significant.
Pairwise comparisons for all subjects together further showed that bids for targets-cueON were lower than those for each of the other three conditions (all P-values < 0.001; all P-values reported here are Bonferroni-corrected.). There was no difference between non-targets-cueON and non-targets-cueOFF (P > 0.50). Unexpectedly, bids for targets-cueOFF were also lower than bids for non-targets-cueON (P < 0.05), and tended to be lower than those for non-targets-cueOFF (P < 0.10). In sum, both hypnosis and autosuggestion successfully devalued target snacks during cue presentation, as measured by bidding behavior.
RTs for bid responses (Table 1) did not differ between groups or conditions (P-values > 0.10 for all main effects and interactions).
Results concerning the postexperimental questionnaires are shown in Figure 2B and C (disgust ratings and other questions, respectively). For disgust ratings, there was a main effect of snack type, F = 17.177, P < 0.001, η = 0.18, of cue, F = 48.661, P < 0.001, η = 0.20 and an interaction of cue × snack type, F = 23.236, P < 0.001, η = 0.08. There were no interactions of group × cue, group × snack type or group × cue × snack type (all P-values >0.15). However, there was a main effect of group, as the hypnosis group reported higher average disgust than the autosuggestion group, F = 10.76, P = 0.003, η = 0.26. As predicted, t-tests for all participants together showed that reported disgust was higher for targets-cueON than for each of the other conditions (all P-values <0.001, Bonferroni-corrected). The remaining three conditions did not differ from each other (all corrected P-values >0.40).
Thus, disgust ratings per condition indicated that both interventions were equally successful in devaluing target snacks experientially. However, participants in the hypnosis group reported their disgust to be more bodily, t = 2.250, P = 0.03, r = 0.38 and less self-controlled (i.e. more automatic) than participants in the autosuggestion group, U = 22.500, P < 0.001, r = −0.70 (two values were missing here). Moreover, participants in the autosuggestion group reported merely having pretended to feel disgust to a stronger degree than participants in the hypnosis group, t(20.033) = −2.499, P = 0.02, r = 0.49 (degrees of freedom adjusted due to unequal variances). In terms of recalling the suggestion or instruction during the experiment, the groups did not differ, U = 91.500, P = 0.16, r = −0.25.
In the fMRI analysis, we first tested whether vmPFC correlated with subjective value, as shown by previous studies. Thus, we determined whether any voxels within our vmPFC ROI (Supplementary Figure S1A http://scan.oxfordjournals.org/content/9/9/1281/suppl/DC1) correlated with bids independent of condition in all 32 participants. As expected, rACC and parts of mOFC correlated with bids at P < 0.05, corrected (Figure 3).
(Enlarge Image)
Figure 3.
Correlation with bids independent of condition for participants of both groups analyzed together (n = 32; peak: − 6, 35, − 2, t = 5.09). Results are masked by the a priori defined vmPFC ROI (Supplementary Figure S1A). For visualization purpose, activations are shown at P < 0.005, k = 10. The beta-plot visualizes the results (mean beta ± 1 s.e.). The plot was constructed using a procedure that ensures independency from the main analysis (see Supplementary Data; Litt et al., 2011).
Effects on vmPFC. Our next goal was to corroborate our behavioral findings across the four conditions with corresponding activation patterns in vmPFC. We tested for the effects of cue (OFF > ON), of snack type (non-target > target) and for the interaction of cue × snack type (stronger effect for non-targets > targets during cueON-runs than during cueOFF-runs) within our vmPFC ROI (Supplementary Figure S1A http://scan.oxfordjournals.org/content/9/9/1281/suppl/DC1). Diminished attractiveness of snacks should be reflected in diminished blood oxygen level-dependent (BOLD) signal in vmPFC. We initially analyzed the fMRI data for all subjects together and then assessed group differences.
Indeed, for all these contrasts we found BOLD signal changes in vmPFC (Figure 4). The effects of cue and snack type were both significant at P < 0.05, corrected, while the interaction of cue × snack type was significant at P < 0.001, uncorrected. Interestingly, activations for the three effects were found in distinct but partly overlapping regions of vmPFC. Note that it is not problematic that the cue × snack type interaction was non-significant after correction. This interaction might also be encoded by the pattern of activation across the two vmPFC subregions showing effects for cue and snack type. Thus, there does not necessarily need to be a subregion representing the interaction as such.
(Enlarge Image)
Figure 4.
Effect of the cue (yellow), of snack type (cyan) and of the interaction of cue × snack type (red) in vmPFC for participants of both groups analyzed together (n = 32; peaks within ROI for cue: −6, 35, −2, t = 4.01; snack type: 6, 26, −17, t = 4.29; cue × snack type: 0, 32, −2, t = 3.63). Results are masked by the vmPFC ROI. For visualization, activations are shown at P < 0.005, k = 10. beta-plots visualize the results (mean beta ± 1 s.e.) and were constructed in a way that ensures independency from the main analysis (see Supplementary Data). T: target snacks (sweet or salty). NT: non-target snacks. CueON: cue color shown in the background. CueOFF: neutral background color.
The effects of cue, snack type and of their interaction (Figure 4) overlapped with the region that correlated with bids 'within' the four conditions (Figure 3). This indicates that hypnosis and autosuggestion indeed altered 'valuation' of food stimuli. Hence, when analyzing all subjects together, the results of the fMRI analysis corroborated the behavioral results as vmPFC was responsive to the experimental manipulations in the expected directions.
When comparing the two groups, we further found that the effect of the cue on rACC (a subregion of the vmPFC ROI) was stronger in the hypnosis group than in the autosuggestion group, at P < 0.05, corrected (Figure 5A). The effect of snack type and the interaction of cue × snack type, in contrast, did not differ significantly between groups within the vmPFC ROI.
(Enlarge Image)
Figure 5.
Group differences. (A) There was a stronger effect of the cue on vmPFC activation in the hypnosis group (peak: 3, 32, 4, t = 4.73). (B). There was also a stronger interaction effect of cue × snack type on the precuneus in the hypnosis group (peak: 9, −58, 46, t = 4.00). Results are masked by the vmPFC ROI in panel A and by the precuneus ROI in panel B (Supplementary Figure S1 gives a depiction of the ROIs). For visualization, activations are shown at P < 0.005, k = 10. beta-plots visualize the results (mean beta ± 1 s.e.) and were constructed in a way that ensures independency from the main analysis (see Supplementary Data). T: target snacks (sweet or salty); NT: non-target snacks. CueON: cue color shown in the background. CueOFF: neutral background color.
Effects on the Precuneus. If the precuneus was functionally involved in the effects of hypnosis or autosuggestion, we would expect finding an effect of cue, snack type or of their interaction in this region. As above, we first tested for these three effects in all subjects together (n = 32) in all voxels of the precuneus ROI (Supplementary Figure S1B http://scan.oxfordjournals.org/content/9/9/1281/suppl/DC1). There was no effect of cue or snack type, but there was an interaction of cue × snack type at P < 0.05, corrected (peak: 6, −55, 52, t = 4.25).
The cue × snack type interaction in the precuneus was stronger in the hypnosis group than in the autosuggestion group, at P < 0.05, corrected. Figure 5B shows mean β-values of the interaction voxels per condition. The pattern shows that the group × cue × snack type interaction arises because only in the hypnosis group, precuneus differentiated between targets/non-targets during cueON-runs but not during cueOFF-runs. This indicates that the precuneus was indeed functionally involved in the effects of the hypnotic suggestions because it encoded relevant information concerning the hypnotic suggestions. Finally, we also found stronger activation of the precuneus for all trials (against baseline) for the hypnosis group compared with the autosuggestion group, at P < 0.05, corrected (peak: 0, −61, 34, t = 3.89).
Whole-brain Analysis. Our exploratory whole-brain analyses revealed effects of snack type and of the interaction of cue × snack type on several regions outside of our ROIs when all participants were analyzed together. Among other findings, the right anterior insula was more active for target snacks than for non-target snacks. An interaction of cue × snack type was, for example, found in the fusiform gyrus, the posterior cingulate cortex and the parahippocampal gyrus (Supplementary Table S3 http://scan.oxfordjournals.org/content/9/9/1281/suppl/DC1 shows the complete results). The analyses further showed that group differences between hypnosis and autosuggestion were largely specific to vmPFC and precuneus (Supplementary Table S4 http://scan.oxfordjournals.org/content/9/9/1281/suppl/DC1).
Results
Bids
Bids per condition are shown in Figure 2A. For all subjects together, there was a main effect of snack type, F = 16.095, P < 0.001, η = 0.29, as participants bid less for target snacks than for non-target snacks. There was also a main effect of cue, F = 17.686, P < 0.001, η = 0.02, as participants bid less for snacks during cueON-runs than during cueOFF-runs. Finally, there was an interaction of cue × snack type, F = 22.953, P < 0.001, η = 0.04, as the difference between bids for target snacks and non-target snacks was bigger during cueON-runs than during cueOFF-runs. Furthermore, there were no interactions of group × snack type, group × cue or group × snack type × cue (all P-values >0.50), and also no main effect of group (P = 0.16). Hence, the effects for bids did not differ between groups.
(Enlarge Image)
Figure 2.
Behavior and self-report. (A) Bids. (B) Postexperimental disgust ratings. (C) Postexperimentally reported degree of physicality of disgust, experience of self-control (as opposed to automaticity) regarding the feeling of disgust, pretending of disgust and conscious recall of suggestion/instruction during the experiment. Ninety-five percent confidence interval (CI) of the mean are shown, adjusted for within-subject designs where appropriate Loftus and Masson, (1994). T: target snacks (sweet or salty). NT: non-target snacks. CueON: cue color shown in the background. CueOFF: neutral background color. ME: Main effect. IE: Interaction effect. ***P < 0.001, **P < 0.01, *P < 0.05; n.s., not significant.
Pairwise comparisons for all subjects together further showed that bids for targets-cueON were lower than those for each of the other three conditions (all P-values < 0.001; all P-values reported here are Bonferroni-corrected.). There was no difference between non-targets-cueON and non-targets-cueOFF (P > 0.50). Unexpectedly, bids for targets-cueOFF were also lower than bids for non-targets-cueON (P < 0.05), and tended to be lower than those for non-targets-cueOFF (P < 0.10). In sum, both hypnosis and autosuggestion successfully devalued target snacks during cue presentation, as measured by bidding behavior.
RTs for bid responses (Table 1) did not differ between groups or conditions (P-values > 0.10 for all main effects and interactions).
Self-report
Results concerning the postexperimental questionnaires are shown in Figure 2B and C (disgust ratings and other questions, respectively). For disgust ratings, there was a main effect of snack type, F = 17.177, P < 0.001, η = 0.18, of cue, F = 48.661, P < 0.001, η = 0.20 and an interaction of cue × snack type, F = 23.236, P < 0.001, η = 0.08. There were no interactions of group × cue, group × snack type or group × cue × snack type (all P-values >0.15). However, there was a main effect of group, as the hypnosis group reported higher average disgust than the autosuggestion group, F = 10.76, P = 0.003, η = 0.26. As predicted, t-tests for all participants together showed that reported disgust was higher for targets-cueON than for each of the other conditions (all P-values <0.001, Bonferroni-corrected). The remaining three conditions did not differ from each other (all corrected P-values >0.40).
Thus, disgust ratings per condition indicated that both interventions were equally successful in devaluing target snacks experientially. However, participants in the hypnosis group reported their disgust to be more bodily, t = 2.250, P = 0.03, r = 0.38 and less self-controlled (i.e. more automatic) than participants in the autosuggestion group, U = 22.500, P < 0.001, r = −0.70 (two values were missing here). Moreover, participants in the autosuggestion group reported merely having pretended to feel disgust to a stronger degree than participants in the hypnosis group, t(20.033) = −2.499, P = 0.02, r = 0.49 (degrees of freedom adjusted due to unequal variances). In terms of recalling the suggestion or instruction during the experiment, the groups did not differ, U = 91.500, P = 0.16, r = −0.25.
Correlation Between vmPFC Activation and Bids Independent of Condition
In the fMRI analysis, we first tested whether vmPFC correlated with subjective value, as shown by previous studies. Thus, we determined whether any voxels within our vmPFC ROI (Supplementary Figure S1A http://scan.oxfordjournals.org/content/9/9/1281/suppl/DC1) correlated with bids independent of condition in all 32 participants. As expected, rACC and parts of mOFC correlated with bids at P < 0.05, corrected (Figure 3).
(Enlarge Image)
Figure 3.
Correlation with bids independent of condition for participants of both groups analyzed together (n = 32; peak: − 6, 35, − 2, t = 5.09). Results are masked by the a priori defined vmPFC ROI (Supplementary Figure S1A). For visualization purpose, activations are shown at P < 0.005, k = 10. The beta-plot visualizes the results (mean beta ± 1 s.e.). The plot was constructed using a procedure that ensures independency from the main analysis (see Supplementary Data; Litt et al., 2011).
Effects on vmPFC. Our next goal was to corroborate our behavioral findings across the four conditions with corresponding activation patterns in vmPFC. We tested for the effects of cue (OFF > ON), of snack type (non-target > target) and for the interaction of cue × snack type (stronger effect for non-targets > targets during cueON-runs than during cueOFF-runs) within our vmPFC ROI (Supplementary Figure S1A http://scan.oxfordjournals.org/content/9/9/1281/suppl/DC1). Diminished attractiveness of snacks should be reflected in diminished blood oxygen level-dependent (BOLD) signal in vmPFC. We initially analyzed the fMRI data for all subjects together and then assessed group differences.
Indeed, for all these contrasts we found BOLD signal changes in vmPFC (Figure 4). The effects of cue and snack type were both significant at P < 0.05, corrected, while the interaction of cue × snack type was significant at P < 0.001, uncorrected. Interestingly, activations for the three effects were found in distinct but partly overlapping regions of vmPFC. Note that it is not problematic that the cue × snack type interaction was non-significant after correction. This interaction might also be encoded by the pattern of activation across the two vmPFC subregions showing effects for cue and snack type. Thus, there does not necessarily need to be a subregion representing the interaction as such.
(Enlarge Image)
Figure 4.
Effect of the cue (yellow), of snack type (cyan) and of the interaction of cue × snack type (red) in vmPFC for participants of both groups analyzed together (n = 32; peaks within ROI for cue: −6, 35, −2, t = 4.01; snack type: 6, 26, −17, t = 4.29; cue × snack type: 0, 32, −2, t = 3.63). Results are masked by the vmPFC ROI. For visualization, activations are shown at P < 0.005, k = 10. beta-plots visualize the results (mean beta ± 1 s.e.) and were constructed in a way that ensures independency from the main analysis (see Supplementary Data). T: target snacks (sweet or salty). NT: non-target snacks. CueON: cue color shown in the background. CueOFF: neutral background color.
The effects of cue, snack type and of their interaction (Figure 4) overlapped with the region that correlated with bids 'within' the four conditions (Figure 3). This indicates that hypnosis and autosuggestion indeed altered 'valuation' of food stimuli. Hence, when analyzing all subjects together, the results of the fMRI analysis corroborated the behavioral results as vmPFC was responsive to the experimental manipulations in the expected directions.
When comparing the two groups, we further found that the effect of the cue on rACC (a subregion of the vmPFC ROI) was stronger in the hypnosis group than in the autosuggestion group, at P < 0.05, corrected (Figure 5A). The effect of snack type and the interaction of cue × snack type, in contrast, did not differ significantly between groups within the vmPFC ROI.
(Enlarge Image)
Figure 5.
Group differences. (A) There was a stronger effect of the cue on vmPFC activation in the hypnosis group (peak: 3, 32, 4, t = 4.73). (B). There was also a stronger interaction effect of cue × snack type on the precuneus in the hypnosis group (peak: 9, −58, 46, t = 4.00). Results are masked by the vmPFC ROI in panel A and by the precuneus ROI in panel B (Supplementary Figure S1 gives a depiction of the ROIs). For visualization, activations are shown at P < 0.005, k = 10. beta-plots visualize the results (mean beta ± 1 s.e.) and were constructed in a way that ensures independency from the main analysis (see Supplementary Data). T: target snacks (sweet or salty); NT: non-target snacks. CueON: cue color shown in the background. CueOFF: neutral background color.
Effects on the Precuneus. If the precuneus was functionally involved in the effects of hypnosis or autosuggestion, we would expect finding an effect of cue, snack type or of their interaction in this region. As above, we first tested for these three effects in all subjects together (n = 32) in all voxels of the precuneus ROI (Supplementary Figure S1B http://scan.oxfordjournals.org/content/9/9/1281/suppl/DC1). There was no effect of cue or snack type, but there was an interaction of cue × snack type at P < 0.05, corrected (peak: 6, −55, 52, t = 4.25).
The cue × snack type interaction in the precuneus was stronger in the hypnosis group than in the autosuggestion group, at P < 0.05, corrected. Figure 5B shows mean β-values of the interaction voxels per condition. The pattern shows that the group × cue × snack type interaction arises because only in the hypnosis group, precuneus differentiated between targets/non-targets during cueON-runs but not during cueOFF-runs. This indicates that the precuneus was indeed functionally involved in the effects of the hypnotic suggestions because it encoded relevant information concerning the hypnotic suggestions. Finally, we also found stronger activation of the precuneus for all trials (against baseline) for the hypnosis group compared with the autosuggestion group, at P < 0.05, corrected (peak: 0, −61, 34, t = 3.89).
Whole-brain Analysis. Our exploratory whole-brain analyses revealed effects of snack type and of the interaction of cue × snack type on several regions outside of our ROIs when all participants were analyzed together. Among other findings, the right anterior insula was more active for target snacks than for non-target snacks. An interaction of cue × snack type was, for example, found in the fusiform gyrus, the posterior cingulate cortex and the parahippocampal gyrus (Supplementary Table S3 http://scan.oxfordjournals.org/content/9/9/1281/suppl/DC1 shows the complete results). The analyses further showed that group differences between hypnosis and autosuggestion were largely specific to vmPFC and precuneus (Supplementary Table S4 http://scan.oxfordjournals.org/content/9/9/1281/suppl/DC1).
Source...