Fig 2 - uploaded by James Rowe
Content may be subject to copyright.
Dynamic Causal Modeling (DCM) neural network. The amygdala (AMY) and the ventral anterior cingulate cortex (ACC) are the regions of interest (ROIs). The faces (regardless of the emotion) were entered as driving input (DCM matrix C values) directly in both the ROIs according with electrophysiological data. The intrinsic connectivity between ROIs (DCM matrix A values) was modelled as bidirectional (see DCM methods section for further details). ‘ Angry faces ’ was used as bilinear modulator (DCM matrix B values) of connectivity in both pathways (from ventral ACC to AMY and from AMY to ventral ACC).
Source publication
For some people facial expressions of aggression are intimidating, for others they are perceived as provocative, evoking an aggressive response. Identifying the key neurobiological factors that underlie this variation is fundamental to our understanding of aggressive behaviour. The amygdala and the ventral anterior cingulate cortex (ACC) have been...
Contexts in source publication
Context 1
... intrinsic connections (connectivity regardless the main effect of the task, DCM Bilinear matrix A value; see Fig. 2) were modelled as bidirectional in accord with anatomical evidence showing that the ventral ACC projects to the amygdala and vice versa ( Aggleton et al., 1980;Amaral and Price, 1984;Ghashghaei et al., 2007;McDonald and Mascagni, ...
Context 2
... modulation by the emotion of faces was included as a bilinear effect expressing the contextual moderator (i.e. anger versus neutral context; DCM Bilinear matrix B value; see Fig. 2). A significant effect of the bilinear variable on connectivity indicates a PPI ( Friston et al., ...
Context 3
... presentation of faces regardless of the emotional expres- sion) were 'injected' into different parts of the network (Supplemen- tary Fig. 1). This 'injection' determines the origin of perturbation of the network, from which other points in the network will be activated according to the pattern of connectivity. For the first "parallel" model ( Fig. 2), face driving inputs (i.e., all faces, regardless of emotional expressions) were 'injected' into both amygdala and the ventral ACC. This was considered the more neurobiologically plausible model according to electrophysiological literature in animals ( Leonard et al., 1985;Rolls, 2007) and humans (Eimer and Holmes, 2007;Eimer et al., ...
Context 4
... every participant we found very strong evidence (see DCM results and Supplementary Fig. 2), in favour of the "parallel" model 1. Therefore using this model we further analyzed the impact of the driving input (all faces) (DCM matrix C value: effect of faces regardless of the expression) on both ROI activities, the intrinsic connectivity between the ROIs (DCM matrix A value: connectivity regardless the main effect of the task), and the modulatory effect by emotion (DCM matrix B value: angry vs. neutral context) on specific bidirectional intrinsic connections (from amygdala to the ventral ACC and from the ventral ACC to amygdala) in each participant at a fixed-effects level. ...
Context 5
... the Bayesian model selection procedure ( Penny et al., 2004), we found strong evidence that the parallel model 1 was associated with the highest probability to explain the data (see Supplementary Fig. 2). Therefore we used model 1 to test for modulatory influence of anger (vs. ...
Context 6
... intrinsic connections (connectivity regardless the main effect of the task, DCM Bilinear matrix A value; see Fig. 2) were modelled as bidirectional in accord with anatomical evidence showing that the ventral ACC projects to the amygdala and vice versa (Aggleton et al., 1980;Amaral and Price, 1984;Ghashghaei et al., 2007;McDonald and Mascagni, ...
Context 7
... modulation by the emotion of faces was included as a bilinear effect expressing the contextual moderator (i.e. anger versus neutral context; DCM Bilinear matrix B value; see Fig. 2). A significant effect of the bilinear variable on connectivity indicates a PPI ( Friston et al., ...
Context 8
... (i.e. presentation of faces regardless of the emotional expression) were 'injected' into different parts of the network (Supplementary Fig. 1). This 'injection' determines the origin of perturbation of the network, from which other points in the network will be activated according to the pattern of connectivity. For the first "parallel" model ( Fig. 2), face driving inputs (i.e., all faces, regardless of emotional expressions) were 'injected' into both amygdala and the ventral ACC. This was considered the more neurobiologically plausible model according to electrophysiological literature in animals ( Leonard et al., 1985;Rolls, 2007) and humans (Eimer and Holmes, 2007;Eimer et al., ...
Context 9
... every participant we found very strong evidence (see DCM results and Supplementary Fig. 2), in favour of the "parallel" model 1. Therefore using this model we further analyzed the impact of the driving input (all faces) (DCM matrix C value: effect of faces regardless of the expression) on both ROI activities, the intrinsic connectivity between the ROIs (DCM matrix A value: connectivity regardless the main effect of the task), and the modulatory effect by emotion (DCM matrix B value: angry vs. neutral context) on specific bidirectional intrinsic connections (from amygdala to the ventral ACC and from the ventral ACC to amygdala) in each participant at a fixed-effects level. ...
Context 10
... the Bayesian model selection procedure (Penny et al., 2004), we found strong evidence that the parallel model 1 was associated with the highest probability to explain the data (see Supplementary Fig. 2). Therefore we used model 1 to test for modulatory influence of anger (vs. ...
Context 11
... see Supplementary Table 1). Again, the time-series for each participant was computed by using the fi rst eigenvariate from all voxels ’ time series in this common left amygdala ROI. Regardless of the approach used to extract the time-series in the source region we obtained highly consistent results (see PPI GLM results section). The BOLD time series for each participant was deconvolved to estimate a ‘ neuronal time series ’ for this region (Gitelman et al., 2003). The psycho-physiological interaction term (PPI regressor) was calculated as the element-by-element product of the left amygdala neuronal time series and a vector coding for the main effect of task (1 for angry faces, − 1 for neutral faces, and 0 for null events). This product was re-convolved by the canonical hemodynamic response function (hrf). The model also included the main effects of task convolved by the hrf, and the movement regressors as effects of no interest. Participant speci fi c PPI models were run, and contrast images generated for positive and negative PPIs. The identi fi ed regions have greater or lesser connectivity with the source region according the context of angry vs. neutral face presentation. The 21 contrast images were entered into second level GLM analyses for contrasts of interest, and SPM-maps generated using Gaussian Random Field theory to make statistical inferences (Friston et al., 1995). To test regions with changes in connectivity with the source region following angry vs . neutral faces context in the whole sample (regardless of any personality dimension) we used a one sample t test. To identify regions for which the changes in connectivity with the source region (following angry vs. neutral faces context) were correlated with the individual variation in reward – drive score, we employed a regression model within SPM. Using distinct regression models, we also explored any correlations with the fun seeking and reward responsiveness subscales, although these dimensions have been associated with measures of aggression to a lesser extent (Carver, 2004; Putman et al., 2004). Moreover, as the neural response to angry faces has also been found to be in fl uenced by anxiety and depression (Ewbank et al., submitted; Leyman et al., 2007; Phan et al., 2006), we investigated any potential effect of the participants' STAI (state and trait anxiety) scores, FNE (fear negative evaluation) scores and CES-D (depression) scores on the connectivity between the amygdala and other potential brain ‘ target ’ regions (regression models). Two approaches to statistically threshold maps were applied. First, for small volume corrections (svc) within a priori regions of interest (ROI), the threshold was set at p b .05 Family Wise Error (Worsley et al., 1996). For the effect of reward – drive, we de fi ned a 15-mm sphere in the ventral ACC ROI using as center the local maxima derived from our previous study ( x − 15, y 36, z − 12) (Beaver et al., 2008). For the effect of anxiety the ROIs comprised the dorsal ACC (MNI local maxima: x − 2, y 12, z 40) and the ventrolateral prefrontal cortex (VLPFC) (MNI local maxima: x − 36, y 16, z − 6), de fi ned from previous work (Bishop et al., 2004; Bishop et al., 2007). Second, to explore other possible regions which were not predicted, a threshold of p b .001, uncorrected was used. To understand further the effective connectivity between the amygdala and the ventral ACC (the two regions showing a higher order interaction with the reward – drive personality, see the PPI GLM results section) we used dynamic causal modelling (DCM) (Friston et al., 2003). DCM enables an alternative method of analysis of psychophysiological interactions within a hypothesis driven anatomical model. More speci fi cally, the DCM explains regional effects in terms of changing patterns of connectivity amongst regions according to experimentally induced contextual modulation of connection strengths. The principal advantage of DCM over the GLM implementa- tion of PPI analysis is the ability to make inferences about the directionality of causal connections. The DCM anatomical model was built from speci fi c hypotheses about the amygdala and the ventral ACC as key neural structures for the processing and recognition of emotional faces (Adolphs et al., 1999; Broks et al., 1998; Calder et al., 2001; Hornak et al., 2003). A 10 mm sphere ROI was created in the left amygdala by using the local maximum for each participant that was the nearest voxel to the activation peak in the left amygdala de fi ned by the whole group cluster (Supplementary Table 1). For the ventral ACC, a 15 mm sphere ROI was de fi ned using the local maxima for each subject that was the nearest voxel to the activation peak identi fi ed by the PPI analysis (see Results section, Analysis of effective connectivity 1: PPIs in the General Linear Model). The intrinsic connections (connectivity regardless the main effect of the task, DCM Bilinear matrix A value; see Fig. 2) were modelled as bidirectional in accord with anatomical evidence showing that the ventral ACC projects to the amygdala and vice versa (Aggleton et al., 1980; Amaral and Price, 1984; Ghashghaei et al., 2007; McDonald and Mascagni, 1996). The modulation by the emotion of faces was included as a bilinear effect expressing the contextual moderator (i.e. anger versus neutral context; DCM Bilinear matrix B value; see Fig. 2). A signi fi cant effect of the bilinear variable on connectivity indicates a PPI (Friston et al., 2003). We fi rst tested three different DCM models in which the driving inputs (i.e. presentation of faces regardless of the emotional expression) were ‘ injected ’ into different parts of the network (Supplementary Fig. 1). This ‘ injection ’ determines the origin of perturbation of the network, from which other points in the network will be activated according to the pattern of connectivity. For the fi rst “ parallel ” model (Fig. 2), face driving inputs (i.e., all faces, regardless of emotional expressions) were ‘ injected ’ into both amygdala and the ventral ACC. This was considered the more neurobiologically plausible model according to electrophysiological literature in animals (Leonard et al., 1985; Rolls, 2007) and humans (Eimer and Holmes, 2007; Eimer et al., 2003; Kawasaki et al., 2001; Oya et al., 2002) suggesting that the amygdala and ACC respond very quickly and within approximately the same time-scale window ( ∼ 110 – 220 ms) to faces. The two last “ serial ” models differ from the fi rst one with respect to where the driving inputs were injected: only in the amygdala (model 2) or only in the ventral ACC (model 3). When comparing all models using Bayesian model selection implemented within SPM5 software we assumed that all of them were equally likely a priori (Penny et al., 2004). We used the selection procedure which estimates the probability of each model given the data using Akaike's information criterion (AIC) and Bayesian's information criterion (BIC) approximations to each model's log-evidence or marginal likelihood (Penny et al., 2004). For every participant we found very strong evidence (see DCM results and Supplementary Fig. 2), in favour of the “ parallel ” model 1. Therefore using this model we further analyzed the impact of the driving input (all faces) (DCM matrix C value: effect of faces regardless of the expression) on both ROI activities, the intrinsic connectivity between the ROIs (DCM matrix A value: connectivity regardless the main effect of the task), and the modulatory effect by emotion (DCM matrix B value: angry vs. neutral context) on speci fi c bidirectional intrinsic connections (from amygdala to the ventral ACC and from the ventral ACC to amygdala) in each participant at a fi xed-effects level. One sample t tests were performed on the A-, B- and C-DCM matrix values to enable inference about the whole group (irrespective of any personality scores). Finally, individual B-, and C-DCM matrix values were entered into simple regression models with reward – drive, STAI (state or trait) anxiety, FNE (fear negative evaluation), and CES-D (depression) scores as main regressors in order to identify any speci fi c modulation by the reward-drive personality, anxiety or depression. The scores on the BAS/BIS subscales, on the Spielberger State and Trait anxiety (STAI), on the Fear of Negative Evaluation scale (FNE-Brief), and on the Center for Epidemiologic Studies Depression Scale (CES-D) were as follows: BAS reward-drive, range 6 to14 (0.5 – 80 percentile of the normal population), mean = 10.09, SD = 2.02; BAS- reward responsiveness, range 12 to 20 (0.4 – 87 percentile), mean = 16.66, SD = 1.85; BAS-fun seeking, range 7 to15 (0.8 – 87 percentile), mean = 11.66, SD = 2.29; BIS, range 17 to 28 (14 – 98 percentile), mean = 21.71, SD = 3.54; State anxiety, range 20 to 43 (1.7 – 94 percentile), mean = 30.71, SD = 6.76; Trait anxiety, range 21 to 52 (2.5 – 99 percentile), mean = 36.52, SD = 8.39; FNE range 0 to 42 (0 – 94 percentile), mean = 20.45, SD = 12.73; CES-D, range 1 to 24 (17 – 96 percentile), mean = 11, SD = 6.20. We found a borderline negative correlation between the reward – drive and BIS measures ( r = − .40, p = .06). Thus, to exclude any contribution from BIS scores, they were included as a covariate of no interest in the general linear model (GLM). Across the whole group, reaction times (RT) during the gender decision task were longer for angry (mean RT = 717 ms, SD = 88) than for neutral face trials (mean RT = 696 ms; SD = 74; t (20) = 2.46, p b .02). The difference in RT between angry and neutral face trials correlated negatively with reward – drive scores ( r = − .51, p b .02), re fl ecting faster gender categorization of angry faces with increasing reward – drive. This accords with previous research showing increased attention to angry faces in high reward – drive individuals (Putman et al., 2004). Again, to exclude any confounding effect of ...
Context 12
... series ’ for this region (Gitelman et al., 2003). The psycho-physiological interaction term (PPI regressor) was calculated as the element-by-element product of the left amygdala neuronal time series and a vector coding for the main effect of task (1 for angry faces, − 1 for neutral faces, and 0 for null events). This product was re-convolved by the canonical hemodynamic response function (hrf). The model also included the main effects of task convolved by the hrf, and the movement regressors as effects of no interest. Participant speci fi c PPI models were run, and contrast images generated for positive and negative PPIs. The identi fi ed regions have greater or lesser connectivity with the source region according the context of angry vs. neutral face presentation. The 21 contrast images were entered into second level GLM analyses for contrasts of interest, and SPM-maps generated using Gaussian Random Field theory to make statistical inferences (Friston et al., 1995). To test regions with changes in connectivity with the source region following angry vs . neutral faces context in the whole sample (regardless of any personality dimension) we used a one sample t test. To identify regions for which the changes in connectivity with the source region (following angry vs. neutral faces context) were correlated with the individual variation in reward – drive score, we employed a regression model within SPM. Using distinct regression models, we also explored any correlations with the fun seeking and reward responsiveness subscales, although these dimensions have been associated with measures of aggression to a lesser extent (Carver, 2004; Putman et al., 2004). Moreover, as the neural response to angry faces has also been found to be in fl uenced by anxiety and depression (Ewbank et al., submitted; Leyman et al., 2007; Phan et al., 2006), we investigated any potential effect of the participants' STAI (state and trait anxiety) scores, FNE (fear negative evaluation) scores and CES-D (depression) scores on the connectivity between the amygdala and other potential brain ‘ target ’ regions (regression models). Two approaches to statistically threshold maps were applied. First, for small volume corrections (svc) within a priori regions of interest (ROI), the threshold was set at p b .05 Family Wise Error (Worsley et al., 1996). For the effect of reward – drive, we de fi ned a 15-mm sphere in the ventral ACC ROI using as center the local maxima derived from our previous study ( x − 15, y 36, z − 12) (Beaver et al., 2008). For the effect of anxiety the ROIs comprised the dorsal ACC (MNI local maxima: x − 2, y 12, z 40) and the ventrolateral prefrontal cortex (VLPFC) (MNI local maxima: x − 36, y 16, z − 6), de fi ned from previous work (Bishop et al., 2004; Bishop et al., 2007). Second, to explore other possible regions which were not predicted, a threshold of p b .001, uncorrected was used. To understand further the effective connectivity between the amygdala and the ventral ACC (the two regions showing a higher order interaction with the reward – drive personality, see the PPI GLM results section) we used dynamic causal modelling (DCM) (Friston et al., 2003). DCM enables an alternative method of analysis of psychophysiological interactions within a hypothesis driven anatomical model. More speci fi cally, the DCM explains regional effects in terms of changing patterns of connectivity amongst regions according to experimentally induced contextual modulation of connection strengths. The principal advantage of DCM over the GLM implementa- tion of PPI analysis is the ability to make inferences about the directionality of causal connections. The DCM anatomical model was built from speci fi c hypotheses about the amygdala and the ventral ACC as key neural structures for the processing and recognition of emotional faces (Adolphs et al., 1999; Broks et al., 1998; Calder et al., 2001; Hornak et al., 2003). A 10 mm sphere ROI was created in the left amygdala by using the local maximum for each participant that was the nearest voxel to the activation peak in the left amygdala de fi ned by the whole group cluster (Supplementary Table 1). For the ventral ACC, a 15 mm sphere ROI was de fi ned using the local maxima for each subject that was the nearest voxel to the activation peak identi fi ed by the PPI analysis (see Results section, Analysis of effective connectivity 1: PPIs in the General Linear Model). The intrinsic connections (connectivity regardless the main effect of the task, DCM Bilinear matrix A value; see Fig. 2) were modelled as bidirectional in accord with anatomical evidence showing that the ventral ACC projects to the amygdala and vice versa (Aggleton et al., 1980; Amaral and Price, 1984; Ghashghaei et al., 2007; McDonald and Mascagni, 1996). The modulation by the emotion of faces was included as a bilinear effect expressing the contextual moderator (i.e. anger versus neutral context; DCM Bilinear matrix B value; see Fig. 2). A signi fi cant effect of the bilinear variable on connectivity indicates a PPI (Friston et al., 2003). We fi rst tested three different DCM models in which the driving inputs (i.e. presentation of faces regardless of the emotional expression) were ‘ injected ’ into different parts of the network (Supplementary Fig. 1). This ‘ injection ’ determines the origin of perturbation of the network, from which other points in the network will be activated according to the pattern of connectivity. For the fi rst “ parallel ” model (Fig. 2), face driving inputs (i.e., all faces, regardless of emotional expressions) were ‘ injected ’ into both amygdala and the ventral ACC. This was considered the more neurobiologically plausible model according to electrophysiological literature in animals (Leonard et al., 1985; Rolls, 2007) and humans (Eimer and Holmes, 2007; Eimer et al., 2003; Kawasaki et al., 2001; Oya et al., 2002) suggesting that the amygdala and ACC respond very quickly and within approximately the same time-scale window ( ∼ 110 – 220 ms) to faces. The two last “ serial ” models differ from the fi rst one with respect to where the driving inputs were injected: only in the amygdala (model 2) or only in the ventral ACC (model 3). When comparing all models using Bayesian model selection implemented within SPM5 software we assumed that all of them were equally likely a priori (Penny et al., 2004). We used the selection procedure which estimates the probability of each model given the data using Akaike's information criterion (AIC) and Bayesian's information criterion (BIC) approximations to each model's log-evidence or marginal likelihood (Penny et al., 2004). For every participant we found very strong evidence (see DCM results and Supplementary Fig. 2), in favour of the “ parallel ” model 1. Therefore using this model we further analyzed the impact of the driving input (all faces) (DCM matrix C value: effect of faces regardless of the expression) on both ROI activities, the intrinsic connectivity between the ROIs (DCM matrix A value: connectivity regardless the main effect of the task), and the modulatory effect by emotion (DCM matrix B value: angry vs. neutral context) on speci fi c bidirectional intrinsic connections (from amygdala to the ventral ACC and from the ventral ACC to amygdala) in each participant at a fi xed-effects level. One sample t tests were performed on the A-, B- and C-DCM matrix values to enable inference about the whole group (irrespective of any personality scores). Finally, individual B-, and C-DCM matrix values were entered into simple regression models with reward – drive, STAI (state or trait) anxiety, FNE (fear negative evaluation), and CES-D (depression) scores as main regressors in order to identify any speci fi c modulation by the reward-drive personality, anxiety or depression. The scores on the BAS/BIS subscales, on the Spielberger State and Trait anxiety (STAI), on the Fear of Negative Evaluation scale (FNE-Brief), and on the Center for Epidemiologic Studies Depression Scale (CES-D) were as follows: BAS reward-drive, range 6 to14 (0.5 – 80 percentile of the normal population), mean = 10.09, SD = 2.02; BAS- reward responsiveness, range 12 to 20 (0.4 – 87 percentile), mean = 16.66, SD = 1.85; BAS-fun seeking, range 7 to15 (0.8 – 87 percentile), mean = 11.66, SD = 2.29; BIS, range 17 to 28 (14 – 98 percentile), mean = 21.71, SD = 3.54; State anxiety, range 20 to 43 (1.7 – 94 percentile), mean = 30.71, SD = 6.76; Trait anxiety, range 21 to 52 (2.5 – 99 percentile), mean = 36.52, SD = 8.39; FNE range 0 to 42 (0 – 94 percentile), mean = 20.45, SD = 12.73; CES-D, range 1 to 24 (17 – 96 percentile), mean = 11, SD = 6.20. We found a borderline negative correlation between the reward – drive and BIS measures ( r = − .40, p = .06). Thus, to exclude any contribution from BIS scores, they were included as a covariate of no interest in the general linear model (GLM). Across the whole group, reaction times (RT) during the gender decision task were longer for angry (mean RT = 717 ms, SD = 88) than for neutral face trials (mean RT = 696 ms; SD = 74; t (20) = 2.46, p b .02). The difference in RT between angry and neutral face trials correlated negatively with reward – drive scores ( r = − .51, p b .02), re fl ecting faster gender categorization of angry faces with increasing reward – drive. This accords with previous research showing increased attention to angry faces in high reward – drive individuals (Putman et al., 2004). Again, to exclude any confounding effect of the response time, the differences in RT between angry and neutral faces trials were factored out in the PPI GLM model. There were no correlations between differences in RT and other BAS/BIS subscales, measures of anxiety, or depression scores ( r s b .23, p s N .15). In addition, task accuracy was consistently high across participants (mean accuracy = 94.5%, SD = 2.39) with no statistically signi fi cant correlations ...
Context 13
... speci fi c PPI models were run, and contrast images generated for positive and negative PPIs. The identi fi ed regions have greater or lesser connectivity with the source region according the context of angry vs. neutral face presentation. The 21 contrast images were entered into second level GLM analyses for contrasts of interest, and SPM-maps generated using Gaussian Random Field theory to make statistical inferences (Friston et al., 1995). To test regions with changes in connectivity with the source region following angry vs . neutral faces context in the whole sample (regardless of any personality dimension) we used a one sample t test. To identify regions for which the changes in connectivity with the source region (following angry vs. neutral faces context) were correlated with the individual variation in reward – drive score, we employed a regression model within SPM. Using distinct regression models, we also explored any correlations with the fun seeking and reward responsiveness subscales, although these dimensions have been associated with measures of aggression to a lesser extent (Carver, 2004; Putman et al., 2004). Moreover, as the neural response to angry faces has also been found to be in fl uenced by anxiety and depression (Ewbank et al., submitted; Leyman et al., 2007; Phan et al., 2006), we investigated any potential effect of the participants' STAI (state and trait anxiety) scores, FNE (fear negative evaluation) scores and CES-D (depression) scores on the connectivity between the amygdala and other potential brain ‘ target ’ regions (regression models). Two approaches to statistically threshold maps were applied. First, for small volume corrections (svc) within a priori regions of interest (ROI), the threshold was set at p b .05 Family Wise Error (Worsley et al., 1996). For the effect of reward – drive, we de fi ned a 15-mm sphere in the ventral ACC ROI using as center the local maxima derived from our previous study ( x − 15, y 36, z − 12) (Beaver et al., 2008). For the effect of anxiety the ROIs comprised the dorsal ACC (MNI local maxima: x − 2, y 12, z 40) and the ventrolateral prefrontal cortex (VLPFC) (MNI local maxima: x − 36, y 16, z − 6), de fi ned from previous work (Bishop et al., 2004; Bishop et al., 2007). Second, to explore other possible regions which were not predicted, a threshold of p b .001, uncorrected was used. To understand further the effective connectivity between the amygdala and the ventral ACC (the two regions showing a higher order interaction with the reward – drive personality, see the PPI GLM results section) we used dynamic causal modelling (DCM) (Friston et al., 2003). DCM enables an alternative method of analysis of psychophysiological interactions within a hypothesis driven anatomical model. More speci fi cally, the DCM explains regional effects in terms of changing patterns of connectivity amongst regions according to experimentally induced contextual modulation of connection strengths. The principal advantage of DCM over the GLM implementa- tion of PPI analysis is the ability to make inferences about the directionality of causal connections. The DCM anatomical model was built from speci fi c hypotheses about the amygdala and the ventral ACC as key neural structures for the processing and recognition of emotional faces (Adolphs et al., 1999; Broks et al., 1998; Calder et al., 2001; Hornak et al., 2003). A 10 mm sphere ROI was created in the left amygdala by using the local maximum for each participant that was the nearest voxel to the activation peak in the left amygdala de fi ned by the whole group cluster (Supplementary Table 1). For the ventral ACC, a 15 mm sphere ROI was de fi ned using the local maxima for each subject that was the nearest voxel to the activation peak identi fi ed by the PPI analysis (see Results section, Analysis of effective connectivity 1: PPIs in the General Linear Model). The intrinsic connections (connectivity regardless the main effect of the task, DCM Bilinear matrix A value; see Fig. 2) were modelled as bidirectional in accord with anatomical evidence showing that the ventral ACC projects to the amygdala and vice versa (Aggleton et al., 1980; Amaral and Price, 1984; Ghashghaei et al., 2007; McDonald and Mascagni, 1996). The modulation by the emotion of faces was included as a bilinear effect expressing the contextual moderator (i.e. anger versus neutral context; DCM Bilinear matrix B value; see Fig. 2). A signi fi cant effect of the bilinear variable on connectivity indicates a PPI (Friston et al., 2003). We fi rst tested three different DCM models in which the driving inputs (i.e. presentation of faces regardless of the emotional expression) were ‘ injected ’ into different parts of the network (Supplementary Fig. 1). This ‘ injection ’ determines the origin of perturbation of the network, from which other points in the network will be activated according to the pattern of connectivity. For the fi rst “ parallel ” model (Fig. 2), face driving inputs (i.e., all faces, regardless of emotional expressions) were ‘ injected ’ into both amygdala and the ventral ACC. This was considered the more neurobiologically plausible model according to electrophysiological literature in animals (Leonard et al., 1985; Rolls, 2007) and humans (Eimer and Holmes, 2007; Eimer et al., 2003; Kawasaki et al., 2001; Oya et al., 2002) suggesting that the amygdala and ACC respond very quickly and within approximately the same time-scale window ( ∼ 110 – 220 ms) to faces. The two last “ serial ” models differ from the fi rst one with respect to where the driving inputs were injected: only in the amygdala (model 2) or only in the ventral ACC (model 3). When comparing all models using Bayesian model selection implemented within SPM5 software we assumed that all of them were equally likely a priori (Penny et al., 2004). We used the selection procedure which estimates the probability of each model given the data using Akaike's information criterion (AIC) and Bayesian's information criterion (BIC) approximations to each model's log-evidence or marginal likelihood (Penny et al., 2004). For every participant we found very strong evidence (see DCM results and Supplementary Fig. 2), in favour of the “ parallel ” model 1. Therefore using this model we further analyzed the impact of the driving input (all faces) (DCM matrix C value: effect of faces regardless of the expression) on both ROI activities, the intrinsic connectivity between the ROIs (DCM matrix A value: connectivity regardless the main effect of the task), and the modulatory effect by emotion (DCM matrix B value: angry vs. neutral context) on speci fi c bidirectional intrinsic connections (from amygdala to the ventral ACC and from the ventral ACC to amygdala) in each participant at a fi xed-effects level. One sample t tests were performed on the A-, B- and C-DCM matrix values to enable inference about the whole group (irrespective of any personality scores). Finally, individual B-, and C-DCM matrix values were entered into simple regression models with reward – drive, STAI (state or trait) anxiety, FNE (fear negative evaluation), and CES-D (depression) scores as main regressors in order to identify any speci fi c modulation by the reward-drive personality, anxiety or depression. The scores on the BAS/BIS subscales, on the Spielberger State and Trait anxiety (STAI), on the Fear of Negative Evaluation scale (FNE-Brief), and on the Center for Epidemiologic Studies Depression Scale (CES-D) were as follows: BAS reward-drive, range 6 to14 (0.5 – 80 percentile of the normal population), mean = 10.09, SD = 2.02; BAS- reward responsiveness, range 12 to 20 (0.4 – 87 percentile), mean = 16.66, SD = 1.85; BAS-fun seeking, range 7 to15 (0.8 – 87 percentile), mean = 11.66, SD = 2.29; BIS, range 17 to 28 (14 – 98 percentile), mean = 21.71, SD = 3.54; State anxiety, range 20 to 43 (1.7 – 94 percentile), mean = 30.71, SD = 6.76; Trait anxiety, range 21 to 52 (2.5 – 99 percentile), mean = 36.52, SD = 8.39; FNE range 0 to 42 (0 – 94 percentile), mean = 20.45, SD = 12.73; CES-D, range 1 to 24 (17 – 96 percentile), mean = 11, SD = 6.20. We found a borderline negative correlation between the reward – drive and BIS measures ( r = − .40, p = .06). Thus, to exclude any contribution from BIS scores, they were included as a covariate of no interest in the general linear model (GLM). Across the whole group, reaction times (RT) during the gender decision task were longer for angry (mean RT = 717 ms, SD = 88) than for neutral face trials (mean RT = 696 ms; SD = 74; t (20) = 2.46, p b .02). The difference in RT between angry and neutral face trials correlated negatively with reward – drive scores ( r = − .51, p b .02), re fl ecting faster gender categorization of angry faces with increasing reward – drive. This accords with previous research showing increased attention to angry faces in high reward – drive individuals (Putman et al., 2004). Again, to exclude any confounding effect of the response time, the differences in RT between angry and neutral faces trials were factored out in the PPI GLM model. There were no correlations between differences in RT and other BAS/BIS subscales, measures of anxiety, or depression scores ( r s b .23, p s N .15). In addition, task accuracy was consistently high across participants (mean accuracy = 94.5%, SD = 2.39) with no statistically signi fi cant correlations with any of the BAS/BIS subscales, measures of anxiety, or depression scores ( r s b .12, p s N .48). The PPI GLM showed a borderline negative connectivity between the left amygdala (source region) and the left ventral ACC across all participants (regardless any personality dimension) that did not meet the a priori threshold for signi fi cance ( x − 8, y 44, z 4; t = 2.89; p b .005, uncorrected) (one sample t test). However, our hypothesis was that the magnitude of this effect might re fl ect systematic individual differences in ...
Context 14
... the response time, the differences in RT between angry and neutral faces trials were factored out in the PPI GLM model. There were no correlations between differences in RT and other BAS/BIS subscales, measures of anxiety, or depression scores ( r s b .23, p s N .15). In addition, task accuracy was consistently high across participants (mean accuracy = 94.5%, SD = 2.39) with no statistically signi fi cant correlations with any of the BAS/BIS subscales, measures of anxiety, or depression scores ( r s b .12, p s N .48). The PPI GLM showed a borderline negative connectivity between the left amygdala (source region) and the left ventral ACC across all participants (regardless any personality dimension) that did not meet the a priori threshold for signi fi cance ( x − 8, y 44, z 4; t = 2.89; p b .005, uncorrected) (one sample t test). However, our hypothesis was that the magnitude of this effect might re fl ect systematic individual differences in reward – drive personality, representing a higher-order PPI. As predicted, the statistical parametrical map (SPM) of this higher-order PPI again identi fi ed the left ventral ACC and was highly signi fi cant ( x − 10, y 42, z − 10; t = 6.03; p b .005, Family Wise Error (FWE), small volume corrected (svc); Fig. 1 C). Moreover, it is striking that the left ventral ACC was one of only three regions that showed connectivity with the amygdala as a function of the anger context and the reward – drive personality ( x − 10, y 42, z − 10; t = 6.03; p b .001, uncorrected). The others regions were the dorsolateral prefrontal cortex ( x 26, y 34, z 34; t = 4.10; p b .001, uncorrected) that has been also implicated in aggression although to a lesser extent (Davidson et al., 2000), and the parietal cortex ( x 12, y − 52, z 74; t = 4.68; p b .001, uncorrected). Anger-related changes in the connectivity between the amygdala and the ventral ACC were highly correlated with the reward – drive scores ( r = .77, p b .001), ranging from more negative connectivity with lower reward – drive scores to less negative connectivity for higher reward – drive scores (Fig. 1 D, see also Supplementary Fig. 3). In the previous analysis, the time series for the source region was extracted from a participant-speci fi c local maximum in the left amygdala ROI, consistent with previous studies (Stephan et al., 2003). However, it is of note that we also obtained consistent results using a different approach in which the time series for each participant was extracted using the same center of the 10 mm sphere ROI for all participants (see PPI-GLM methods for details). Again, the left ventral ACC showed connectivity with the amygdala as a function of both the anger context and the reward – drive personality ( x − 12, y 40, z − 10; t = 4.80; p b .02, FWE svc). In both PPI analyses we did not fi nd any higher-order PPI between the amygdala – ventral ACC connectivity and other BAS/BIS measures (reward responsiveness, fun seeking or BIS) (no suprathreshold voxels even reducing the threshold at p b .005, uncorrected). We also tested for brain ‘ target ’ regions showing changes in connectivity with the amygdala as a function of the trait or state anxiety and we identi fi ed different areas, including the dorsal ACC ( x − 10, y 24, z 42; t = 4.27; p b .05, FWE, svc) and the ventrolateral prefrontal cortex (VLPFC) ( x − 32, y 24, z − 12; t = 4.11; p b .05, FWE, svc), which have been previously associated with individual differences in anxiety (Bishop et al., 2004; Bishop et al., 2007). The anger-related changes in the connectivity between the amygdala and the dorsal ACC and the VLPFC were positively correlated with state anxiety ( r = .69, p b .001 for the dorsal ACC; r = .65, p b .001 for the VLPFC), ranging from more negative connectivity with lower anxiety scores to less negative connectivity for higher anxiety scores. No signi fi cant higher-order PPI was found with the FNE (fear negative evaluation) scale (no suprathreshold voxels even reducing the threshold at p b .005, uncorrected). However, we identi fi ed the left ventral putamen as a region coupled with the left amygdala as a function of the CES-D (depression) score ( x − 12, y 8, z − 10; t = 5.21; p b .0001, uncorrected; r = .63, p b .005). In summary, the anger-related modulation of connectivity between the amygdala and the ventral ACC was strongly correlated with individual differences in reward – drive but not with other emotional dimensions (fun-seeking, reward-responsiveness, BIS, state and trait anxiety, FNE, and depression). Moreover, by covarying out any contribution by RT's and BIS scores, we have demonstrated that these variables do not account for the statistically signi fi cant effect of the reward – drive personality on amygdala-ventral ACC PPI. Using the Bayesian model selection procedure (Penny et al., 2004), we found strong evidence that the parallel model 1 was associated with the highest probability to explain the data (see Supplementary Fig. 2). Therefore we used model 1 to test for modulatory in fl uence of anger ( vs. neutral) expression on each speci fi c pathway (from ventral ACC to amygdala and from amygdala to ventral ACC). As predicted, face presentation (regardless of the emotional expression: DCM matrix C) had a very strong in fl uence on the neural activity of both amygdala and ventral ACC ROIs in all participants ( p s b .00001) (Fig. 3). Similarly, the intrinsic connectivity (DCM matrix A) between the ROIs in the whole sample was highly signi fi cant in both directions (from the ventral ACC to amygdala and vice versa ) ( p s b .00001) (Fig. 3). The main effect of the emotional expression (angry vs. neutral context: DCM Bilinear matrix B) enhanced the connectivity from ventral ACC to amygdala ( p b .002) but not vice versa ( p N .2) across individuals (one sample t test, irrespective of the reward – drive personality) (Fig. 3). The direct comparison between the DCM Bilinear matrix B's (B values from ventral ACC to amygdala vs. B values from amygdala to ventral ACC) con fi rmed that the anger-related changes in connectivity between the ventral ACC and amygdala were not symmetrical (paired t test, t (20) = 2.54, p b .02). Of particular interest, we found a signi fi cant higher-order interaction between the reward – drive scores and the subject-speci fi c effect of emotion on connectivity (DCM Bilinear matrix B: angry versus neutral context). There was a negative correlation between the reward – drive and the DCM bilinear moderator term (DCM Bilinear matrix B) for the connectivity in the speci fi c pathway from the ventral ACC to the amygdala ( r = − .50, p b .02) (Fig. 4). The reverse connection (from the amygdala to the ventral ACC) did not correlate with reward – drive ( r = .02, p N .90). Furthermore, there was no signi fi cant in fl uence of reward – drive on the driving inputs (presentation of faces regardless of the emotional expression: DCM Bilinear matrix C) to the amygdala or the ventral ACC (Fig. 4) and we found no correlations between STAI (state or trait anxiety), FNE (fear negative evaluation) or the CES-D (depression) scores and the B- or C-DCM Bilinear matrix values ( r s b .21, p s N .36). The speci fi city of reward – drive effect in the pathway from the ventral ACC to the amygdala was con fi rmed by comparing the r value derived from the correlation with reward – drive with the r values derived from the correlations with other personality dimensions ( t s N 2.7, p s b .05, Hottelling's t tests). This means that the anger-related effective connectivity from the ventral ACC to amygdala is the only critical pathway in fl uenced by the reward – drive personality but not by other emotional dimensions, such anxiety or depression. Group statistics for DCM results are summarized in Figs. 3 and 4 (one sample t tests and correlation with reward – drive scores). We have shown that individual differences in reward – drive (appetitive motivation) strongly modulate the neural connectivity between the ventral ACC and amygdala while viewing facial signals of aggression. This was addressed using two complementary methods — Psychophysiological Interactions (PPI) in a general linear model (GLM) and Dynamic Causal Modelling (DCM). PPI analyses have the advantage of being anatomically unconstrained, providing an objec- tive, data-driven approach. The PPI analysis in the whole sample (regardless of individual differences in reward – drive) identi fi ed the ventral ACC as showing a weak change in connectivity with the amygdala according to whether the faces displayed angry or neutral expressions (more negative connectivity for angry faces), although this effect did not meet our criteria for signi fi cance. However, when the individual differences in reward – drive were taken into consideration, we found that the change in connectivity between the same regions (ventral ACC and amygdala) was highly correlated with reward – drive, ranging from more negative to less negative values with increasing reward – drive. DCM analysis supported and extended these fi ndings by showing that this latter effect was restricted to connectivity from the ventral ACC to the amygdala, but not vice versa . Note also, that our results provide no evidence that connectivity between the amygdala and ventral ACC was modulated by the different measures of anxiety, other reward processing dimensions, or depression. Our previous research showed that regional activation within the amygdala and ventral ACC in response to viewing angry (relative to neutral or sad) expressions is correlated with individual differences in reward – drive (Beaver et al., 2008), a dimension consistently linked to the tendency to display hostile and aggressive behaviour (Carver, 2004; Cooper et al., 2008; Cornell et al., 1996; Diefendorff and Mehta, 2007; Putman et al., 2004; Smits and Kuppens, 2005). The present study goes signi fi cantly further by showing that it is speci fi cally ...
Context 15
... and depression (Ewbank et al., submitted; Leyman et al., 2007; Phan et al., 2006), we investigated any potential effect of the participants' STAI (state and trait anxiety) scores, FNE (fear negative evaluation) scores and CES-D (depression) scores on the connectivity between the amygdala and other potential brain ‘ target ’ regions (regression models). Two approaches to statistically threshold maps were applied. First, for small volume corrections (svc) within a priori regions of interest (ROI), the threshold was set at p b .05 Family Wise Error (Worsley et al., 1996). For the effect of reward – drive, we de fi ned a 15-mm sphere in the ventral ACC ROI using as center the local maxima derived from our previous study ( x − 15, y 36, z − 12) (Beaver et al., 2008). For the effect of anxiety the ROIs comprised the dorsal ACC (MNI local maxima: x − 2, y 12, z 40) and the ventrolateral prefrontal cortex (VLPFC) (MNI local maxima: x − 36, y 16, z − 6), de fi ned from previous work (Bishop et al., 2004; Bishop et al., 2007). Second, to explore other possible regions which were not predicted, a threshold of p b .001, uncorrected was used. To understand further the effective connectivity between the amygdala and the ventral ACC (the two regions showing a higher order interaction with the reward – drive personality, see the PPI GLM results section) we used dynamic causal modelling (DCM) (Friston et al., 2003). DCM enables an alternative method of analysis of psychophysiological interactions within a hypothesis driven anatomical model. More speci fi cally, the DCM explains regional effects in terms of changing patterns of connectivity amongst regions according to experimentally induced contextual modulation of connection strengths. The principal advantage of DCM over the GLM implementa- tion of PPI analysis is the ability to make inferences about the directionality of causal connections. The DCM anatomical model was built from speci fi c hypotheses about the amygdala and the ventral ACC as key neural structures for the processing and recognition of emotional faces (Adolphs et al., 1999; Broks et al., 1998; Calder et al., 2001; Hornak et al., 2003). A 10 mm sphere ROI was created in the left amygdala by using the local maximum for each participant that was the nearest voxel to the activation peak in the left amygdala de fi ned by the whole group cluster (Supplementary Table 1). For the ventral ACC, a 15 mm sphere ROI was de fi ned using the local maxima for each subject that was the nearest voxel to the activation peak identi fi ed by the PPI analysis (see Results section, Analysis of effective connectivity 1: PPIs in the General Linear Model). The intrinsic connections (connectivity regardless the main effect of the task, DCM Bilinear matrix A value; see Fig. 2) were modelled as bidirectional in accord with anatomical evidence showing that the ventral ACC projects to the amygdala and vice versa (Aggleton et al., 1980; Amaral and Price, 1984; Ghashghaei et al., 2007; McDonald and Mascagni, 1996). The modulation by the emotion of faces was included as a bilinear effect expressing the contextual moderator (i.e. anger versus neutral context; DCM Bilinear matrix B value; see Fig. 2). A signi fi cant effect of the bilinear variable on connectivity indicates a PPI (Friston et al., 2003). We fi rst tested three different DCM models in which the driving inputs (i.e. presentation of faces regardless of the emotional expression) were ‘ injected ’ into different parts of the network (Supplementary Fig. 1). This ‘ injection ’ determines the origin of perturbation of the network, from which other points in the network will be activated according to the pattern of connectivity. For the fi rst “ parallel ” model (Fig. 2), face driving inputs (i.e., all faces, regardless of emotional expressions) were ‘ injected ’ into both amygdala and the ventral ACC. This was considered the more neurobiologically plausible model according to electrophysiological literature in animals (Leonard et al., 1985; Rolls, 2007) and humans (Eimer and Holmes, 2007; Eimer et al., 2003; Kawasaki et al., 2001; Oya et al., 2002) suggesting that the amygdala and ACC respond very quickly and within approximately the same time-scale window ( ∼ 110 – 220 ms) to faces. The two last “ serial ” models differ from the fi rst one with respect to where the driving inputs were injected: only in the amygdala (model 2) or only in the ventral ACC (model 3). When comparing all models using Bayesian model selection implemented within SPM5 software we assumed that all of them were equally likely a priori (Penny et al., 2004). We used the selection procedure which estimates the probability of each model given the data using Akaike's information criterion (AIC) and Bayesian's information criterion (BIC) approximations to each model's log-evidence or marginal likelihood (Penny et al., 2004). For every participant we found very strong evidence (see DCM results and Supplementary Fig. 2), in favour of the “ parallel ” model 1. Therefore using this model we further analyzed the impact of the driving input (all faces) (DCM matrix C value: effect of faces regardless of the expression) on both ROI activities, the intrinsic connectivity between the ROIs (DCM matrix A value: connectivity regardless the main effect of the task), and the modulatory effect by emotion (DCM matrix B value: angry vs. neutral context) on speci fi c bidirectional intrinsic connections (from amygdala to the ventral ACC and from the ventral ACC to amygdala) in each participant at a fi xed-effects level. One sample t tests were performed on the A-, B- and C-DCM matrix values to enable inference about the whole group (irrespective of any personality scores). Finally, individual B-, and C-DCM matrix values were entered into simple regression models with reward – drive, STAI (state or trait) anxiety, FNE (fear negative evaluation), and CES-D (depression) scores as main regressors in order to identify any speci fi c modulation by the reward-drive personality, anxiety or depression. The scores on the BAS/BIS subscales, on the Spielberger State and Trait anxiety (STAI), on the Fear of Negative Evaluation scale (FNE-Brief), and on the Center for Epidemiologic Studies Depression Scale (CES-D) were as follows: BAS reward-drive, range 6 to14 (0.5 – 80 percentile of the normal population), mean = 10.09, SD = 2.02; BAS- reward responsiveness, range 12 to 20 (0.4 – 87 percentile), mean = 16.66, SD = 1.85; BAS-fun seeking, range 7 to15 (0.8 – 87 percentile), mean = 11.66, SD = 2.29; BIS, range 17 to 28 (14 – 98 percentile), mean = 21.71, SD = 3.54; State anxiety, range 20 to 43 (1.7 – 94 percentile), mean = 30.71, SD = 6.76; Trait anxiety, range 21 to 52 (2.5 – 99 percentile), mean = 36.52, SD = 8.39; FNE range 0 to 42 (0 – 94 percentile), mean = 20.45, SD = 12.73; CES-D, range 1 to 24 (17 – 96 percentile), mean = 11, SD = 6.20. We found a borderline negative correlation between the reward – drive and BIS measures ( r = − .40, p = .06). Thus, to exclude any contribution from BIS scores, they were included as a covariate of no interest in the general linear model (GLM). Across the whole group, reaction times (RT) during the gender decision task were longer for angry (mean RT = 717 ms, SD = 88) than for neutral face trials (mean RT = 696 ms; SD = 74; t (20) = 2.46, p b .02). The difference in RT between angry and neutral face trials correlated negatively with reward – drive scores ( r = − .51, p b .02), re fl ecting faster gender categorization of angry faces with increasing reward – drive. This accords with previous research showing increased attention to angry faces in high reward – drive individuals (Putman et al., 2004). Again, to exclude any confounding effect of the response time, the differences in RT between angry and neutral faces trials were factored out in the PPI GLM model. There were no correlations between differences in RT and other BAS/BIS subscales, measures of anxiety, or depression scores ( r s b .23, p s N .15). In addition, task accuracy was consistently high across participants (mean accuracy = 94.5%, SD = 2.39) with no statistically signi fi cant correlations with any of the BAS/BIS subscales, measures of anxiety, or depression scores ( r s b .12, p s N .48). The PPI GLM showed a borderline negative connectivity between the left amygdala (source region) and the left ventral ACC across all participants (regardless any personality dimension) that did not meet the a priori threshold for signi fi cance ( x − 8, y 44, z 4; t = 2.89; p b .005, uncorrected) (one sample t test). However, our hypothesis was that the magnitude of this effect might re fl ect systematic individual differences in reward – drive personality, representing a higher-order PPI. As predicted, the statistical parametrical map (SPM) of this higher-order PPI again identi fi ed the left ventral ACC and was highly signi fi cant ( x − 10, y 42, z − 10; t = 6.03; p b .005, Family Wise Error (FWE), small volume corrected (svc); Fig. 1 C). Moreover, it is striking that the left ventral ACC was one of only three regions that showed connectivity with the amygdala as a function of the anger context and the reward – drive personality ( x − 10, y 42, z − 10; t = 6.03; p b .001, uncorrected). The others regions were the dorsolateral prefrontal cortex ( x 26, y 34, z 34; t = 4.10; p b .001, uncorrected) that has been also implicated in aggression although to a lesser extent (Davidson et al., 2000), and the parietal cortex ( x 12, y − 52, z 74; t = 4.68; p b .001, uncorrected). Anger-related changes in the connectivity between the amygdala and the ventral ACC were highly correlated with the reward – drive scores ( r = .77, p b .001), ranging from more negative connectivity with lower reward – drive scores to less negative connectivity for higher reward – drive scores (Fig. 1 D, see also Supplementary Fig. 3). In the previous analysis, the time series for ...
Similar publications
Background
Maternal sensitive behavior depends on recognizing one’s own child’s affective states. The present study investigated distinct and overlapping neural responses of mothers to sad and happy facial expressions of their own child (in comparison to facial expressions of an unfamiliar child).
Methods
We used functional MRI to measure dissocia...
Citations
... In addition, to identify "target" regions for which their changes in connection with the source regions varied as a function of the behavioral framing effect, a whole-brain regression model (or "higher-order" PPI) was also performed in which the behavioral framing effect [(P sure + N gamble ) -(P gamble + N sure )] was used as a vector to identify regions functionally connected with the source regions (e.g., amygdala and striatum) based on the PPI.ppi contrast (Passamonti et al., 2008(Passamonti et al., , 2009). ...
Human decision-making can be influenced by whether the option is framed as positive or negative, known as the framing effect. Neuroeconomic studies have shown the engagement of the amygdala underlies the framing effect by automatically incorporating emotional (or intuitive) information into the decision process, thus leading to "irrational" decisions or decision biases (De Martino et al. 2006). However, these studies focus only on the gain domain in which the initial state is appetitive, leaving the mechanisms of the framing effect in the loss or aversive domain less understood, which could be equally important or even more dominant for human survival. In the present study, we first replicated De Martino et al.’s research on the framing effect in the gain domain. We then checked the commonality and distinctiveness of the framing effect across gain and loss domains using a similar experimental design combined with functional magnetic resonance imaging (fMRI). Behaviorally, participants showed comparable framing effects in both domains, suggesting a domain-general decision bias. Neuronally, the amygdala represented the framing effect for the gain domain, and its connectivity to vmPFC was positively modulated by framing bias, consistently shown in De Martino et al.’s study. Moreover, the striatum represented the framing effect for the loss domain, and its connectivity to dmPFC was subject to framing bias, suggesting distinctive neural substrates across two domains. Our study accentuates the importance of distinguishing decision processes across appetitive or aversive domains and highlights the role of the cortical-striatal-limbic network underlying the framing effect.
... Two GLMs were performed separately with PPI regressors of (1) the main physiological effect of the striatum for an increasing mean, or the insula for an increasing SD; (2) the mean reward and reward SD, and (3) the interaction of the striatum or insula with the target regions, respectively, corresponding to PPI.Y, PPI.P, and PPI.ppi in the design matrix. Using PPI, target regions (from the PPI.ppi) could be identified because their activity was functionally correlated with activations in the source regions (e.g., striatum or insula) under the modulation of the mean or SD, which were the main effects of the PPI analysis ( Passamonti et al., 2008 ;Passamonti et al., 2009 ). ...
When faced with uncertainty, individuals’ value-based decisions are influenced by the expected rewards and risks. Understanding how reward and risk are processed and integrated at the behavioral and neural levels is essential for building up utility theories. Using a modified monetary incentive delay task in which the mean of two possible outcomes (expected reward) and the standard deviation (SD) of the possible outcomes (risk) were parametrically manipulated and orthogonalized, we measured eye movements, response times (RTs), and brain activity when participants seek to secure a reward. We found that RTs varied as a function of the mean but not the SD of the potential reward, suggesting that expected rewards are the main driver of RTs. Moreover, the difference between gazes focused on high vs. low value rewards became smaller when the magnitude of the potential reward (mean of possible outcomes) was larger and when risk (SD of possible outcomes) became smaller, highlighting that reward and risk have different effects on attention deployment. Processing the mean reward activated the striatum. The positive striatal connectivity to the amygdala and negative striatal connectivity to the superior frontal gyrus were correlated with individuals’ sensitivity to the expected reward. In contrast, processing risk activated the anterior insula. Its positive connectivity to the ventromedial prefrontal cortex and negative connectivity to the anterior midcingulate cortex were correlated with individual differences in risk sensitivity, further suggesting the functional dissociation of reward and risk at the neural level. Our findings, based on several different measures, delineate the distinct representations of reward and risk in non-decision contexts and provide insight into how these utility parameters modulate attention, motivation, and brain networks.
... Both vmPFC and vlPFC are classically considered important regulatory regions; in particular, when considering anger, they seem to be respectively involved in cognitive control and in the evaluation of the behavioural outcomes [12,58,115]. Further, the interaction between the vmPFC and amygdala during the presentation of angry faces is associated with individual differences in behavioural approach [116], a specific feature of anger expression [58]. ...
Anger can be broken down into different elements: a transitory state (state anger), a stable personality feature (trait anger), a tendency to express it (anger-out), or to suppress it (anger-in), and the ability to regulate it (anger control). These elements are characterized by individual differences that vary across a continuum. Among them, the abilities to express and suppress anger are of particular relevance as they determine outcomes and enable successful anger management in daily situations. The aim of this study was to demonstrate that anger suppression and expression can be decoded by patterns of grey matter of specific well-known brain networks. To this aim, a supervised machine learning technique, known as Kernel Ridge Regression, was used to predict anger expression and suppression scores of 212 healthy subjects from the grey matter concentration. Results show that individual differences in anger suppression were predicted by two grey matter patterns associated with the Default-Mode Network and the Salience Network. Additionally, individual differences in anger expression were predicted by a circuit mainly involving subcortical and fronto-temporal regions when considering whole brain grey matter features. These results expand previous findings regarding the neural bases of anger by showing that individual differences in specific anger-related components can be predicted by the grey matter features of specific networks.
... Both vmPFC and vlPFC are classically considered important regulatory regions; in particular, when considering anger, they seem to be respectively involved in cognitive control and in the evaluation of the behavioural outcomes [12,58,115]. Further, the interaction between the vmPFC and amygdala during the presentation of angry faces is associated with individual differences in behavioural approach [116], a specific feature of anger expression [58]. ...
... Two GLMs were performed separately with PPI regressors of (1) the main physiological effect of the striatum for an increasing mean, or the insula for an increasing SD; (2) the mean reward and reward SD, and (3) the interaction of the striatum or insula with the target regions, respectively, corresponding to PPI.Y, PPI.P, and PPI.ppi in the design matrix. Using PPI, target regions (from the PPI.ppi) could be identified because their activity was functionally correlated with activations in the source regions (e.g., striatum or insula) under the modulation of the mean or SD, which were the main effects of the PPI analysis ( Passamonti et al., 2008 ;Passamonti et al., 2009 ). ...
... Conversely, ACC-amygdala FC only mediated alcoholrelated AB. Interactions between these two brain regions modulate responses to emotional stimuli [108][109][110] and may also underlie motivation for rewards [111]. The unique association of this connection with alcohol AB, but not generalized reward AB, suggests that alcohol cues become imbued with distinct emotional and motivational qualities beyond their ability to predict reward. ...
Individuals who abuse alcohol often show exaggerated attentional bias (AB) towards alcohol-related cues, which is thought to reflect reward conditioning processes. Rodent studies indicate that dopaminergic pathways play a key role in conditioned responses to reward-and alcohol-associated cues. However, investigation of the dopaminergic circuitry mediating this process in humans remains limited. We hypothesized that depletion of central dopamine levels in adult alcohol drinkers would attenuate AB and that these effects would be mediated by altered function in frontolimbic circuitry. Thirty-four male participants (22-38 years, including both social and heavy drinkers) underwent a two-session, placebo-controlled, double-blind dopamine precursor depletion procedure. At each visit, participants consumed either a balanced amino acid (control) beverage or an amino acid beverage lacking dopamine precursors (order counterbalanced), underwent resting-state fMRI, and completed behavioral testing on three AB tasks: an alcohol dot-probe task, an alcohol attentional blink task, and a task measuring AB to a reward-conditioned cue. Dopamine depletion significantly diminished AB in each behavioral task, with larger effects among subjects reporting higher levels of binge drinking. The depletion procedure significantly decreased resting-state functional connectivity among ventral tegmental area, striatum, amygdala, and prefrontal regions. Beverage-related AB decreases were mediated by decreases in functional connectivity between the fronto-insular cortex and striatum and, for alcohol AB only, between anterior cingulate cortex and amygdala. The results support a substantial role for dopamine in AB, and suggest specific dopamine-modulated functional connections between frontal, limbic, striatal, and brainstem regions mediate general reward AB versus alcohol AB. Neuropsychopharmacology (2021) 0:1-11; https://doi.
... Interestingly, both ACC and AMY are thought to gate information flow to and from the hippocampus via EC connections to enhance hippocampal dependent learning (review:Barbas et al., 2012;Canto et al., 2008;de Curtis & Pare, 2004;Killian et al., 2012;Koganezawa, Gisetstad, Husby, Doan, & Witter, 2015;Paz et al., 2007;Paz et al., 2006;Witter et al., 2000). In addition, direct vACC projections to the AMY play a role in modulating emotional learning and memory(Etkin, Egner, Peraza, Kandel, & Hirsch, 2006;Ghashghaei, Hilgetag, & Barbas, 2007;John, Bullock, Zikopoulos, & Barbas, 2013;Passamonti et al., 2008;Toyoda et al., 2011) (review: Etkin et al., 2011. In monkeys and humans, activity in the vACC has been specifically associated with autonomic arousal, pain, and negative valence(Critchley, Tang, Glaser, Butterworth, & Dolan, 2005;Ichikawa et al., 2011;Mohanty et al., 2007;Monosov & Hikosaka, 2012;Onoda et al.;Rudebeck et al., 2014; ...
The anterior cingulate cortex (ACC) is important for decision‐making as it integrates motor plans with affective and contextual limbic information. Disruptions in these networks have been observed in depression, bipolar disorder, and post‐traumatic stress disorder. Yet, overlap of limbic and motor connections within subdivisions of the ACC is not well understood. Hence, we administered a combination of retro and anterograde tracers into structures important for contextual memories (entorhinal cortex), affective processing (amygdala) and motor planning (dorsal premotor cortex) to assess overlap of labeled projection neurons from (outputs) and axon terminals to (inputs) the ACC of adult rhesus monkeys (Macaca mulatta ). Our data show that entorhinal and dorsal premotor cortical (dPMC) connections are segregated across ventral (A25, A24a) and dorsal (A24b,c) subregions of the ACC, while amygdalar connections are evenly distributed across subregions. Among all areas, the rostral ACC (A32) had the lowest relative density of inter‐connections with all three regions. In the ventral ACC, entorhinal and amygdalar inter‐connections strongly overlap across all layers, especially in A25. In the dorsal ACC, outputs to dPMC and the amygdala strongly overlap in deep layers. However, dPMC input to the dorsal ACC was densest in deep layers, while amygdalar inputs predominantly localized in upper layers. This is consistent with roles of the dorsal ACC in motor evaluation and the ventral ACC in affective and contextual memory. Further, distinct laminar patterns suggest unique interactions within specific ACC compartments that are likely important for the temporal integration of motor and limbic information during flexible goal‐directed behavior.
... To achieve the goal mentioned above, we further conducted a dynamic causal modelling analysis to test directional information flow between the brain regions involved in phantom inferior and phantom superior conditions (Friston et al. 2003). DCM estimates three sets of parameters, including the intrinsic connections between regions independent of experimental context, the driving effect of external stimuli on specific regions, and the modulatory effect of stimuli on the effective connectivity between regions (Ewbank et al. 2011;Passamonti et al. 2008). Consistent with common approaches (Gandolla et al. 2014;Passamonti et al. 2008;Rothkirch et al. 2018), the nodes in our models were identified based on brain regions identified in our previous network connectivity analysis. ...
... DCM estimates three sets of parameters, including the intrinsic connections between regions independent of experimental context, the driving effect of external stimuli on specific regions, and the modulatory effect of stimuli on the effective connectivity between regions (Ewbank et al. 2011;Passamonti et al. 2008). Consistent with common approaches (Gandolla et al. 2014;Passamonti et al. 2008;Rothkirch et al. 2018), the nodes in our models were identified based on brain regions identified in our previous network connectivity analysis. In the phantom inferior condition, we create a 6-mm sphere ROI in the left caudate and the left vACC (see Tables 1 and 2). ...
Rational decision theories posit that good choices should be based solely on information that is relevant to the choice at hand. However, introducing an inferior option that would never be chosen can influence choices among other relevant options, known as decoy effect. We used functional magnetic resonance imaging (fMRI) combined with a simple gambling task to investigate the neural signature of decision-making in or against the influence of the decoy effect in inferior and superior phantom decoy conditions. The fMRI results show that compared with choosing against the influence of the dominated phantom inferior option, choosing in the influence of the same option was associated with stronger activation in bilateral caudate and weaker functional connectivity between the left ventral anterior cingulate cortex (vACC) and the left caudate. Phantom inferior effect selectively enhanced the connectivity from the caudate to the vACC but not vice versa. Choosing in the influence of the dominated phantom superior option engaged greater activity in the left dorsal ACC and stronger functional connectivity between the left dACC and bilateral anterior insula. Furthermore, the direction of the phantom superior effect was restricted from left dACC to the anterior insula, but not vice versa. Our findings suggest that a phantom inferior decoy may boost the value of the target via the reward network, whereas a phantom superior decoy may diminish the value of the target option via the aversion network. Our study provides neural evidence to support that valuation is context dependent and delineates differential neural networks underlying the influence of unavailable inferior and superior decoy options on our decision-making.
... Prior studies on emotion-dependent brain connectivity have typically constrained the analysis to a small numbers of regions of interest and suffered from the low power of conventional event-related and boxcar designs in connectivity analyses (e.g. Eryilmaz et al., 2011;Passamonti et al., 2008;Tettamanti et al., 2012). A naturalistic stimulation setup in turn offers a high-powered alternative for tapping fear-dependent connectivity in the brain. ...
Fear protects organisms by increasing vigilance and preparedness, and by coordinating survival responses during life-threatening encounters. The fear circuit must thus operate on multiple timescales ranging from preparatory sustained alertness to acute fight-or-flight responses. Here we studied the brain basis of sustained and acute fear using naturalistic functional magnetic resonance imaging (fMRI) enabling analysis of different time-scales of fear responses. Subjects (N = 37) watched feature-length horror movies while their hemodynamic brain activity was measured with fMRI. Time-variable intersubject correlation (ISC) was used to quantify the reliability of brain activity across participants, and seed-based phase synchronization was used for characterizing dynamic connectivity. Subjective ratings of fear were used to assess how synchronization and functional connectivity varied with emotional intensity. These data suggest that acute and sustained fear are supported by distinct neural pathways, with sustained fear amplifying mainly sensory responses, and acute fear increasing activity in brainstem, thalamus, amygdala and cingulate cortices. Sustained fear increased ISC in regions associated with acute fear, and also amplified functional connectivity within this network. The results were replicated in an independent experiment with a different subject sample and stimulus movie. The functional interplay between cortical networks involved in sustained anticipation of, and acute response to, threat involves a complex and dynamic interaction that depends on the proximity of threat, and the need to employ threat appraisals and vigilance for decision making and response selection.
... The PPI first-level contrast (positive and negative PPIs) for each source region in stress and control condition were fed into 2 s-level GLM analyses: (1) exploring the brain regions whose connectivity with source regions for SS versus SE context varied under stress and control condition (paired t-test); (2) identifying regions of which the stress-related changes in connectivity with the source regions in SS -SE context varied as a function of individual differences in SSRT between stress and control condition ("higher-order" PPI). For higher-order PPI analysis, we performed a whole-brain regression with "SSRT under stress -SSRT under control" as the independent variable and stress-induced changes in PPI (positive PPI contrast under stress-positive PPI contrast under control) in SS -SE context as the dependent variable (Passamonti et al., 2008(Passamonti et al., , 2009. ...
Life stress has been shown to impact cognitive functions, including inhibitory control. However, the immediate effects of acute stress on inhibitory control and the underlying neural mechanisms remain unclear. In a behavioral pilot study (N = 30) and a within-subject functional magnetic resonance imaging study (N = 30), we examined how acute stress induced by Trier Social Stress Test influenced inhibitory control in a stop signal task. Behavioral results across two studies showed that stress consistently improved inhibitory control. Shorter stop signal reaction time (SSRT) in stress as compared with control condition was associated with stronger connectivity between the superior/middle frontal gyrus (SFG/MFG) and striatum. Dynamic causal modeling revealed distinct best models under stress and control condition, with an enhanced interaction between the SFG/MFG and the striatum after stress exposure. This research identified the SFG/MFG-striatum network as a key circuit underlying acute stress-elicited enhancement of inhibitory control in a stop signal task.
