We have assessed the relationship of several psychological tests related to anxiety phenotypes with physiological indices, both at baseline and during various phases of experimental fear conditioning. While we were unable to find correlations with extinction learning or extinction recall, we were able to explain 24 to 45% of variance in SCR responses during conditioning and renewal, both indicators of the strength of the CS-US association.
The predictive power of our model for conditioning (45%) is consistent with findings from Otto and colleagues, who were able to account for 28% of CS + variance in conditioning, also using multiple regression . This is interesting given that the two models employed different measures. While our model included only conscientiousness and physiological variables associated with reactivity, Otto et al. (2007) found a predictive model that included the Penn State Worry Questionnaire (PSWQ), the Fear Questionnaire (FQ), and the Anxiety Sensitivity Index (ASI). Conscientiousness can be related to PSWQ scores, in that it assesses thoughtfulness and carefulness  , and was previously reported to be correlated with conditioning . Thus, together with Otto et al. (2007), we suggest that approximately 28-45% of SCR variance in conditioning may be due to individual characteristics that can be revealed with self-report scales and measures of physiological reactivity. The greater variance explained in our sample was probably due to our taking into account response to the shock and baseline skin conductance.
In contrast to conditioning, we were unable to find significant correlations with extinction learning or extinction recall, using either linear or multiple regression. This differs from Rauch et al. (2005) who reported that recall of extinction was correlated with extraversion (r = 0.77) and neuroticism (r = −0.61) . The experimental protocols used at the two sites were designed to be identical. Factors that might account for the discrepancy include the larger number of subjects in the current study (46 vs. 14 in Rauch et al.), or an ethnic factor affecting the relationship between personality and extinction (Puerto Rican vs. a largely Caucasian sample in the Boston area). A lack of correlation of extinction with psychological and neuropsychological indices suggests that extinction may be less determined by trait variables, and may depend more on the subject’s current emotional state. In support of this idea, Milad et al. (2008) showed that extinction retention deficits in PTSD patients were acquired rather than familial, based on a twin design . Additionally, Vriends et al. (2011) reported decreased extinction in healthy participants who first watched a film intended to induce an anxious emotional state . Our state measure (STAI) showed no correlation with extinction, likely due to the low state anxiety levels of our healthy subjects in the absence of a provocation such as that used in Vriends et al. (2011).
Our most novel finding was the ability to successfully model contextual renewal. Renewal of extinguished fear with a contextual shift is thought to model relapse after treatment with exposure therapy . It has been suggested that the ability to successfully inhibit fear responses when challenged by contextual shifts is the best predictor of clinical response . Extraversion was negatively correlated with renewal, consistent with a protective effect. Individuals with high extraversion scores are more likely to seek stimulation, undertake activities with unknown consequences, and experience positive emotions [20, 29]. Extraversion is also positively correlated with the volume of the medial oribitofrontal cortex , an area implicated in fear inhibition  and reward . Thus, the presence of extraversion may reflect the engagement of structures that modulate fear responses under ambiguous conditions. In addition, sex was also a significant factor in the model (males showed higher renewal). This is an interesting finding given that anxiety disorders are more prevalent in females [8, 33]. We are now conducting additional studies on the sex differences we observed in our sample.
We propose several follow-up steps to extend these findings and to address the limitations of our study. In order to state that personality characteristics can predict physiological responses, the predictive power of these models should be tested using predictive model analysis with new subjects. The applicability of these results to subjects with anxiety disorders should also be evaluated. It will be particularly interesting to determine the extent to which conditioning and renewal models are able to predict the severity of anxiety disorders, response to treatment, and/or post-treatment relapse, in longitudinal studies. Factors that could be predictive of renewal, such as low extraversion in our sample, may be used for identifying patients at risk for relapse, who might benefit from preventive measures such as delivering therapy in multiple contexts or providing reminders of the therapeutic context .