Abstract
The use of Pavlovian-to-instrumental transfer (PIT) in addiction research is on the rise as a means of assessing an individual's susceptibility to interference between Pavlovian and instrumental control over behaviour. However, the reliability of PIT tasks has been rarely assessed.The present study provides an investigation of the reliability of PIT, both on a behavioural and neural level, examining split-half as well as test-retest reliability. We assessed two different samples: (1) a mixed detoxified alcohol-dependent sample (with controls) comprising 119 behavioural and 69 fMRI datasets assessed twice within 3-4 weeks and (2) a developmental sample with 117 behavioural and 91 fMRI datasets assessed twice after 3 years. We computed two behavioural parameters of PIT (interference and motivational PIT effects) that were used in our previous studies. The interference PIT effect was assessed as the difference in error rates between the congruent and incongruent trials of the PIT task. The motivational PIT effect was assessed as the linear relationship between the number of button presses influenced by Pavlovian-conditioned environmental cues. We further assessed the reliability of four predefined fMRI PIT contrasts.On the behavioural level, our results revealed excellent split-half reliability for both the interference (r = 0.92-0.97) and motivational (r = 0.94-0.98) aspects of PIT in both clinical and developmental samples. In comparison, test-retest reliability after 3 weeks was lower (clinical sample: ICC = 0.53) and again lower yet still significant despite neurodevelopmental brain maturation after 3 years (developmental sample: ICC = 0.26-0.29). In the fMRI analysis, regions of interest showed acceptable ICCs in the incongruent and congruent contrasts (split-half: 0.59-0.80; test-retest: 0.13-0.51). Global overlap assessments using Jaccard coefficients revealed individual-level variability in neural responses (split-half: 47%-51% overlap; test-retest: 29%-35% depending on the sample and contrast). All fMRI reliability coefficients for the motivational PIT effect were below 0.17.Overall, behavioural PIT reliability was good, especially from the split-half perspective. For neuroimaging, the incongruent contrast seems best suited for predicting individual outcomes, while the neural motivational PIT effect seems to represent more changeable current states.