Figure 1
Schematic of go-nogo task. After the presentation of a fixation cross (250 ms), participants were shown a fractal image (1000 ms). After a brief pause (100ms) a fixation cross was presented (250ms). This was followed by another pause (100ms), after which participants were shown a fractal image (1000ms). After another pause (100ms) fixation cross (250ms) and another pause (100ms) a circle was presented. After another fixation cross (250 ms), a circle was presented, on either the left or the right side of the screen. On each trial, participants could either respond (Go) or not respond (No-go) to the presentation of the circle. If they chose to respond, this response consisted of pressing the ‘a’ keyboard key if the circle was on the left of the screen, or the ‘l’ keyboard key of the circle was on the right of the screen. Note that these keys were not randomised, as the response keys (‘a’ for the left side, and ‘l’ for the right side) was designed to be consistent with a standard UK/US keyboard layout. This screen automatically timed-out after 1000 ms, even if the participant had not responded. Subsequently, feedback was shown on the screen for 1000 ms. A ‘reward’ consisted of the presentation of a happy face and text saying ‘+10 points’, a ‘neutral’ outcome was the presentation of a horizontal yellow bar with the text ‘0 points’, and a ‘punishment’ outcome was an unhappy face with the text ‘–10 points’.
Figure 2
Schematic of ambiguous midpoint task. Participants were shown a circle for 1000 ms, followed by a fixation cross for 750 ms. Participants responded using the ‘m’ and ‘z’ keyboard keys. There were three sizes of circles – small and large – which were each associated fully deterministically with one key and one size of reward – and one medium sized circle which was associated 50% with the ‘m’ key and corresponding reward, and 50% with the ‘z’ key and corresponding reward. If participants responded correctly, they were shown the message ‘Correct, win $4’ or ‘Correct, win $1’ depending on whether the stimulus presented was associated with a larger or smaller reward. This feedback was presented for 750 ms. If participants responded incorrectly, they were shown the message ‘Timeout for incorrect response’, which lasted for 3250 ms. If they failed to respond in time, they were shown the message ‘Too late, timeout!’ which also lasted for 3250 ms.
Figure 3
Pearson’s correlation coefficients and Intra-Class Correlation Coefficients for summary statistic measures from each affective bias task: a) Go-nogo task, b) Ambiguous midpoint task. The X axis shows the relevant summary statistics from the specific task, y axis shows the estimate and the 95% confidence interval around them. The black line at y = 0 represents a correlation or ICC of 0 – confidence intervals that do not cross this represent significant estimates. Points above the light grey line at y = 0.4 represent estimates that show reliability of moderate or above; points below this line indicate poor reliability.
Table 1
Pearson’s correlation coefficients and Intra-Class Correlation Coefficients for summary statistic measures from each affective bias task. Values greater than 0.4 are often described as moderate-good, less than 0.4 are poor.
| TASK | MEASURE | PEARSON’S CORRELATION | ICC(A,1) | ICC(C,1) | |||
|---|---|---|---|---|---|---|---|
| COEFFICIENT | p-VALUE | COEFFICIENT | p-VALUE | COEFFICIENT | p-VALUE | ||
| go-nogo | Go to win accuracy | 0.439 | 0.0006 | 0.430 | 0.0002 | 0.440 | 0.0002 |
| Go to avoid accuracy | 0.495 | <0.0001 | 0.496 | <0.0001 | 0.495 | <0.0001 | |
| No-go to win accuracy | 0.179 | 0.1786 | 0.180 | 0.0886 | 0.178 | 0.0888 | |
| No-go to avoid accuracy | 0.462 | 0.0003 | 0.461 | 0.0001 | 0.457 | 0.0001 | |
| Ambiguous midpoint | P (high|mid) | 0.484 | 0.0001 | 0.488 | <0.0001 | 0.484 | <0.0001 |
Figure 4
Pearson’s correlation coefficients and Intra-Class Correlation Coefficients for parameters from different models fit to the affective bias tasks. X axis shows parameter, y axis shows the estimate and the 95% confidence interval around it for the three different measures of test-retest reliability used. The black line at y = 0 represents a correlation or ICC of 0 – confidence intervals that do not cross this represent significant estimates. Points above the grey line at y = 0.4 represent estimates that show reliability of moderate or above; points below this line indicate poor reliability. A) The model that best fit choice data on the go-nogo task, B) a simplified drift diffusion model fit to the ambiguous midpoint task, C) the 4-parameter drift diffusion model fit to the ambiguous midpoint task.
Figure 5
Posterior predictive performance of model parameters, displayed using scatterplots (one individual is represented by each point) and error bars showing the standard error around the mean. Each individual’s model parameters from different sessions were used to calculate the trialwise likelihood of their actual choices made during that session for the go-nogo task (a) and the ambiguous midpoint task (b). To check whether better-than-chance prediction was due to shrinkage resulting from the partial pooling used in hierarchical Bayesian model fitting, the mean parameters from each session were used to predict each individual participant’s choices in the go-nogo task (c) and the ambiguous midpoint task (d). The first number in all x axis names refers to the session from which the parameters were drawn; the second refers to the session on which performance was predicted, thus ‘s2s1’ refers to the use of parameters from session 2 to predict session 1 performance. The grey lines represent the values against which t-tests were performed (i.e. chance).
Figure 6
Pearson’s correlation coefficients and estimates of the correlations when these are embedded within a generative model, for parameters from different models fit to the affective bias tasks. X axis shows parameter, y axis shows the estimate of each measure of test-retest reliability and the 95% confidence interval around it. The black line at y = 0 represents a correlation of 0 – confidence intervals that do not cross this represent significant estimates. Points above the grey line at y = 0.4 represent estimates that show reliability of moderate or above; points below this line indicate poor reliability. A) The model that best fit choice data on the go-nogo task, B) the 4-parameter drift diffusion model fit to data on the ambiguous midpoint task.