3.1 Clinical measures
Internalizing scores showed no significant sex difference, and men vs. women showed higher externalizing scores (Table 1).
Internalizing score showed a significant correlation with externalizing score (all: r = 0.57, p < 0.001; men: r = 0.55, p < 0.001; women: r = 0.60, p < 0.001) with age and sex as covariates for all, and with age as a covariate for men and women separately. In slope tests men and women did not differ in the slope of the regressions (Z = -1.17, p = 0.242).
Table 1
Demographics and ASR Measures of Men and Women
Characteristic | Men (n = 473) | Women (n = 508) | t | p value* |
Age (years) | 27.9 ± 3.6 | 29.6 ± 3.6 | -7.05 | 0.000 |
Internalizing | 10.6 ± 9.1 | 10.8 ± 9.1 | -0.98 | 0.329^ |
Externalizing | 10.3 ± 7.4 | 7.7 ± 6.0 | 5.18 | 0.000^ |
Note: Values are mean ± SD; *two-sample t test (^with age as a covariate). Because men and women differed significantly in age, age was included as a covariate in all analyses.
3.2 Behavioral measures of the gambling task
To characterize how individuals reacted to wins and losses, we computed individual RT of trials following loss (post-loss RT) and win (post-win RT) each for reward and punishment blocks. The results of a 3-way (trial ⋅ block ⋅ sex) repeated measures ANOVA are shown in Supplementary Table S1. There was a significant trial (F = 4.86, p = 0.028), block (F = 132.96, p < 0.001) and sex (F = 29.29, p < 0.001) main effects. Simple effects analyses showed shorter RT in post-loss than in post-win trials (407 ± 112 vs. 423 ± 108 ms; t = -9.86, p < 0.001, paired t test), shorter RT in punishment than in reward blocks (431 ± 106 vs. 447 ± 106 ms; t = -10.92, p < 0.001, paired t test), and shorter RT in men than in women (422 ± 104 vs. 455 ± 100 ms; t = -5.13, p < 0.001, two-sample t test). There was also a significant trial ⋅ block (F = 25.32, p < 0.001) interaction, with greater differences in RT between post-win and post-loss trials in reward than in punishment blocks (12 ± 77 vs. -4 ± 74 ms; t = 5.03, p < 0.001, paired t test). No other interaction effects, including those involving sex as a factor, were significant.
In linear regressions with age and sex as covariates for all and with age only for men and women separately, the difference in RT between post-win and post-loss trials was not correlated with internalizing score across all subjects (r = -0.02, p = 0.452), men (r = -0.01, p = 0.871), or women (r = -0.04, p = 0.380) alone. A slope test did not show significant sex difference in the correlation (Z = 0.50, p = 0.617). The difference in RT was not correlated with externalizing score across all (r = 0.02, p = 0.463), men (r = 0.08, p = 0.106), or women (r = -0.03, p = 0.482), and a slope test did not show sex difference (Z = 1.66, p = 0.097). Thus, the extent of faster RT in post-loss vs. post-win trials was not related to internalizing or externalizing trait.
3.2 VS activations to win and loss
We extracted the contrast of “reward – baseline” and “punishment – baseline” of the VS (VS-win and VS-loss β’s) for individual subjects. The results of a valence ⋅ sex repeated measures ANOVA showed a significant valence (F = 496.30, p < 0.001) and sex (F = 7.63, p = 0.006) main effect but not valence ⋅ sex interaction effect (F = 0.50, p = 0.482). VS-win β did not show significant sex difference (t = 1.93, p = 0.054), but VS-loss β was significantly higher in men than in women (-0.14 ± 0.28 vs. -0.20 ± 0.32; t = 2.79, p = 0.005, Fig. 1B).
We also computed the β estimates of “face – shape” of the VS (VS-NegEmo β’s) for individual subjects. Men (0.12 ± 0.26) and women (0.10 ± 0.26) did not differ significantly in VS-NegEmo β (t = 1.05, p = 0.293; Fig. 1B).
3.3 VS reactivity to win/loss: the influences of Internalizing and externalizing trait
VS-win β was not correlated with internalizing score across all subjects (r = -0.01, p = 0.798) with sex, age and externalizing score as covariates, or in men (r = 0.01, p = 0.866) or women (r = -0.03, p = 0.554) alone with age and externalizing score as covariates. Likewise, VS-win β was not correlated with externalizing score across all subjects (r = 0.03, p = 0.366) with sex, age and internalizing score as covariates, or in men (r = 0.01, p = 0.910) or women (r = 0.06, p = 0.213) alone with age and internalizing score as covariates.
VS-loss β was not correlated with internalizing score across all subjects (r = -0.04, p = 0.201) with sex, age and externalizing score as covariates, or in men (r = -0.03, p = 0.460) or women (r = -0.03, p = 0.474) alone with age and externalizing score as covariates. VS-loss β was not correlated with externalizing score across all subjects (r = 0.06, p = 0.057) with sex, age and internalizing score as covariates, or in women (r = -0.02, p = 0.719), but significantly correlated with externalizing score in men (r = 0.13, p = 0.004) alone with age and internalizing score as covariates.
We conducted a slope test to examine the difference between the correlation each of VS-win β and of VS-loss β vs. externalizing score. The results showed no difference across all subjects (Z = -1, p = 0.317, Fig. 2A) or in women (Z = 1.62, p = 0.105, Fig. 2C). However, men showed significant difference in the slope of regressions at a threshold without correction for multiple comparisons (Z = -2.75, p = 0.006, Fig. 2B). No significant differences in slope were found for the correlation of VS-win β as compared to VS-loss β vs. internalizing score across all, men, or women (all p’s > 0.303). Likewise, no significant differences in slope were observed for the correlation of VS-win β vs. externalizing as compared to VS-win β vs. internalizing score across all, men, or women (all p’s > 0.066). The correlation of VS-loss β vs. externalizing and of VS-loss β vs. internalizing showed significant differences in slopes across all subjects (Z = 3.19, p = 0.001, Fig. 3D), and in men (Z = 3.59, p < 0.001, Fig. 3E) but not women (Z = 0.36, p = 0.719, Fig. 3F) alone.
3.4 Sex differences in the influences of Internalizing and externalizing trait on VS reactivity to win/loss:
Across all correlations, only VS-loss β was significantly correlated with externalizing score in men alone (r = 0.13, p = 0.004), and slope test confirmed the sex difference (Z = 2.31, p = 0.021, Fig. 4C). No other correlations were significant across all subjects, men or women (all p’s > 0.057) or showed significant sex differences in the slopes (all p’s > 0.424, Fig. 4A, B and D).
3.5 Findings on the dorsal striatum and its subnuclei
We conducted the same analyses with the DS and the subregions of DS as seeds and showed the findings in the Supplement Table S2-S5. Briefly, the caudate nucleus response to punishment (Caudate-loss β) showed significant correlation with externalizing score in men alone (r = 0.11, p = 0.014). Men showed significant difference between the correlation each of Caudate-win β and of Caudate-loss β vs. externalizing score (Z = -2.01, p = 0.044), and significant difference between the correlation of Caudate-loss β vs. externalizing and of Caudate-loss β vs. internalizing (Z = 2.14, p = 0.032). However, these findings were no longer significant with correction for multiple comparisons (p < 0.05/4 = 0.0125). No other correlations were significant with putamen, pallidum or DS (all p’s > 0.055).
3.6 Accounting for the effects of alcohol use
The HCP comprises largely neurotypical populations and candidates with a neuropsychiatric condition requiring treatment were excluded from enrollment. However, individuals may use alcohol to varying extents, which is known to influence striatal activities. Thus, we investigated whether the “positive” findings described earlier, as shown in 3D and 3E, remained significant after accounting for alcohol use. HCP evaluated alcohol use with multiple questions and, as in our earlier work (Li et al., 2022; Li et al., 2021c), we conducted a principal component analysis of all drinking-related measures and identified a single, principal component (PC1) with an eigenvalue (7.42) > 1 and explaining 49.47% of the variance. With PC1 as an additional covariate to reflect the severity of alcohol use, the results showed that VS-loss β was correlated with externalizing score across men (r = 0.11, p = 0.017), but not all subjects (r = 0.04, p = 0.167) or women (r = -0.03, p = 0.516), and VS-loss β was not correlated with internalizing score across all (r = -0.03, p = 0.291), men (r = -0.03, p = 0.592) or women (r = -0.03, p = 0.560). The correlation of VS-loss β vs. externalizing and of VS-loss β vs. internalizing showed significant differences in slopes across all subjects (Z = 2.44, p = 0.015), and in men (Z = 2.94, p = 0.003) but not women (Z = -0.07, p = 0.944) alone. Thus, the findings remained after considering alcohol use severity as a covariate.
3.7 Loss processing vs. negative emotion exposure
We evaluated whether the relationship between externalizing trait and VS loss reactivity broadly reflects individual differences in neural processing of events of negative valence. To this end we took advantage of the imaging data collected of a (angry or fearful) face vs. shape identification task in the HCP. We extracted the β estimates of “face – shape” of individual participants, which reflected negative emotion processing (VS-NegEmo β) and performed the same regressions against externalizing and internalizing traits across all subjects and in men and women separately. The results showed that VS-NegEmo β was positively correlated with externalizing score across all subjects (r = 0.08, p = 0.024) and women (r = 0.12, p = 0.013) but not men (r = 0.05, p = 0.319), and negatively correlated with internalizing score across all subjects (r = -0.08, p = 0.012) and men (r = -0.12, p = 0.015) but not women (r = -0.06, p = 0.197). The correlation of VS-NegEmo β vs. externalizing and of VS-NegEmo β vs. internalizing score showed significant differences in slopes across all subjects (Z = 4.76, p < 0.001, Fig. 5A) and in men (Z = 3.42, p < 0.001, Fig. 5B) and women (Z = 3.77, p < 0.001, Fig. 5C) alone. We also included the regressions of VS loss reactivity for this sample as 5D, E, F for comparison. Across all subjects and in men only, the patterns of correlation between VS-loss and VS-NegEmo were similar.
No sex differences were found in the correlations between VS-NegEmo β and externalizing score (Z = -1.03, p = 0.303) or between VS-NegEmo β and internalizing score (Z = -0.82, p = 0.412).
Likewise, we examined whether the findings remained significant with alcohol use severity (PC1) included as an additional covariate. The results showed that VS-NegEmo β was correlated with externalizing score across all subjects (r = 0.08, p = 0.013), women (r = 0.11, p = 0.018) but not men (r = 0.07, p = 0.160), and correlated with internalizing score across all (r = -0.09, p = 0.009), men (r = -0.13, p = 0.008) but not women (r = -0.06, p = 0.201). The correlation of VS-NegEmo β vs. externalizing and of VS-NegEmo β vs. internalizing still showed significant differences in slopes across all (Z = 5.12, p < 0.001), men (Z = 4.07, p < 0.001), and women (Z = 3.67, p < 0.001) alone. Thus, the findings remained the same after we controlled for alcohol use severity.
We conducted the same analyses with the DS and its subregions as seeds. The results showed no significant correlations of DS, caudate, putamen, or pallidum response to negative emotion with externalizing or internalizing score across all subjects or in men or women alone (all p’s > 0.058).