Figures and data in Catecholaminergic modulation of meta-learning

Figures
Tables
Additional files

8 figures, 6 tables and 1 additional file

Figures

Figure 1 with 1 supplement

Download asset Open asset

Task design.

Participants selected between two bandits (blue and green boxes) in order to win points. On each trial, participants saw the direct sources (boxes, 1–4 s), subsequently either the blue or green box was highlighted with a red frame (the indirect source, 1–4 s). Participants were instructed that the frame represented either the most popular choice made by a group of participants who had completed the task previously (Social Group) or the choice from rigged roulette wheels (Non-social Group). After participants had responded, their selected option was framed in grey. After 0.5–2 s participants received feedback in the form of a green or blue box between the two bandits. A new trial began after 1–3 s. Success resulted in the red reward bar progressing towards the silver and gold goals. The probability of reward associated with the blue and green boxes and the probability that the red frame surrounded the correct box varied according to probabilistic schedules which comprised stable and volatile phases.

Figure 1—figure supplement 1

Download asset Open asset

Probabilistic schedules.

Solid blue lines show the probability of blue being the correct choice, dashed red lines show the probability of the red frame information being correct. For the win-stay, lose-shift analysis the interaction between drug, learning type and volatility remained significant when probabilistic schedule group (1,2,3,4) was included as a factor in the ANOVA (F(1,94) = 7.935, p = 0.006). There was no significant main effect of schedule (F(3,94) = 0.906, p = 0.441) and no interactions involving both drug and schedule (all p>0.05). For the learning rate analysis, the interaction between drug, learning type and volatility remained significant when schedule (1,2,3,4) was included as a factor in the ANOVA (F(1,94) = 7.500, p = 0.007). There was no significant main effect of schedule (F(3,94) = 0.850, p = 0.470) and no drug x learning type x volatility x group x schedule interaction (F(3,94) = 0.404, p = 0.751). There was a significant drug x learning type x volatility x probabilistic schedule interaction (F(3,94) = 2.725, p = 0.049). Although the pattern of results was comparable, the interaction between drug x learning type x volatility was only statistically significant for schedule 2 (F(1,23) = 12.231, p = 0.002), and did not reach significance for schedules 1 (F(1,24) = 4.015, p = 0.057), 3 (F(1,22) = 2.881, p = 0.104) and 4 (F(1,25) = 0.501, p = 0.486), likely due to a lack of power with these reduced group sizes.

Figure 2 with 1 supplement

Download asset Open asset

Drug effects on win-stay, lose-shift beta values.

(A) Win-stay, lose-shift betas, for experienced-value learning, in stable and volatile periods, under MPH (purple) and PLA (green). (B) Win-stay, lose-shift betas, for inferred-value learning, in stable and volatile periods, under MPH and PLA. (C) There was a significant interaction between drug and volatility for experienced-value learning, but not for inferred-value learning. (D) Drug effect on win-stay, lose-shift betas for volatile minus stable blocks, for experienced-value (orange) and inferred-value (blue) learning. There was no difference between the social and non-social (roulette) groups. Boxes = standard error of the mean, shaded region = standard deviation, individual datapoints are displayed, MPH = methylphenidate, PLA = placebo, * indicates statistical significance.

Figure 2—figure supplement 1

Download asset Open asset

Illustration of relationship between experienced-value $β$ _{WSLS_vol-stable} and accuracy.

Experienced-value $β$ _{WSLS_vol-stable} was significantly positively correlated with accuracy under both PLA (Top: Pearson’s r = 0.453, p < 0.001) and MPH (Bottom: Pearson’s r = 0.236, p = 0.017).

Figure 3

Download asset Open asset

Drug effects on learning rates.

(A) Learning rate for experienced-value learning, in stable and volatile periods, under MPH (purple) and PLA (green). (B) Learning rate for inferred-value learning, in stable and volatile periods, under MPH and PLA. (C) There was a significant interaction between drug and volatility for experienced-value learning, but not for inferred-value learning. (D) The interaction between drug and volatility was specific to experienced-value learning even for a sub-sample of participants who were keen inferred-value learners. Boxes = standard error of the mean, shaded region = standard deviation, individual datapoints are displayed, MPH = methylphenidate, PLA = placebo, * indicates statistical significance.

Figure 4

Download asset Open asset

Experienced-value learning rates under PLA, but not MPH, were further from the optimal value in the stable compared to volatile phase.

Figure 5

Download asset Open asset

Participant data (left) juxtaposed against model simulations (right).

Top: Accuracy, in stable and volatile phases, under MPH (purple) and PLA (green). Boxes = standard error of the mean, shaded region = standard deviation, individual datapoints are displayed. Middle and Bottom: Running average, across 5 trials, of blue choices for probabilistic randomisation schedules 1 (middle) and 2 (bottom). Shaded region = standard error of the mean. MPH = methylphenidate, PLA = placebo.

Figure 6

Download asset Open asset

Appendix 4—figure 1

Download asset Open asset

$p (y | m)$ = posterior model probability, ϕ = exceedance probability, MPH = purple, PLA = green.

Author response image 1

Download asset Open asset

Tables

Table 1

Coefficients from regression model with MPH-PLA accuracy as dependent variable.

	$β$	$β$ (std. Err.)	standardised $β$	t	p
constant	−0.008	0.006		−1.3	0.197
$β$ _MPH-PLA	0.014	0.002	0.36	5.763*	<0.001
ζ_MPH-PLA	−0.241	0.024	−0.658	−9.943*	<0.001
$α$ _{stable_experienced_MPH-PLA}	0.005	0.032	0.011	0.172	0.864
$α$ _{vol_experienced_MPH-PLA}	0.126	0.029	0.285	4.376*	<0.001
$α$ _{stable_inferred_MPH-PLA}	0.014	0.041	0.023	0.345	0.731
$α$ _{volatile_inferred_MPH-PLA}	0.02	0.041	0.03	0.483	0.63

* indicates statistical significance.

Appendix 1—table 1

There were no significant differences between the social and non-social groups in age, gender or socioeconomic status as measured using the Barratt Simplified Measure of Social Status (Barratt, 2012).

	Social	Non-social
Age in years ( $\bar{x} (σ)$ )	21.8 (2.6)	21.3 (2.0)	t(100) = 1.1, p = 0.3
Socioeconomic status ( $\bar{x} (σ)$ )	60.4 (7.8)	59.2 (8.4)	t(100) = 0.5, p = 0.6
N(Male:Female)	50 (26:24)	52 (25:27)	$χ^{2}$ (1) = 0.16, p = 0.7

Appendix 3—table 1

Win-stay, lose-shift analysis.

Main effects and interactions not involving the factor drug.

Main effect/interaction	Statistics	Further information
Learning type	F(1,100) = 58.595, p < 0.001	$β$ _WSLS for experienced ( $\bar{x} (σ_{\bar{x}})$ = 0.366(0.017)) > inferred ( $\bar{x} (σ_{\bar{x}})$ = 0.188(0.012))
Volatility	F(1,100) = 70.671, p < 0.001	$β$ _WSLS for volatile ( $\bar{x} (σ_{\bar{x}})$ = 0.327(0.010)) > stable ( $\bar{x} (σ_{\bar{x}})$ = 0.227(0.011))
Group	F(1,100) = 4.001, p = 0.048	$β$ _WSLS for non-social ( $\bar{x} (σ_{\bar{x}})$ = 0.295(0.013)) > social ( $\bar{x} (σ_{\bar{x}})$ = 0.259(0.013))
Learning type x group	F(1,100) = 7.729, p = 0.006	$β$ _WSLS for social ( $\bar{x} (σ_{\bar{x}})$ = 0.138(0.017)) < non-social ( $\bar{x} (σ_{\bar{x}})$ = 0.239(0.017)) for inferred (F(1,100) = 17.743, p < 0.001). No difference between groups for experienced (social $\bar{x} (σ_{\bar{x}})$ = 0.381 (0.024), non-social $\bar{x} (σ_{\bar{x}})$ = 0.352 (0.024), F(1,100) = 0.694, p = 0.407)

Appendix 3—table 2

Learning rate analysis.

Main effects and interactions not involving the factor drug. Note that the learning type x volatility interaction is consistent with previous studies of experienced-value learning which have reported higher learning rates in volatile relative to stable environments (Behrens et al., 2007; Behrens et al., 2008; Browning et al., 2015; Krugel et al., 2009; Nassar et al., 2010).

Main effect/interaction	Statistics	Further information
Learning type	F(1,100) = 65.409, p < 0.001	$α$ _direct ( $\bar{x} (σ_{\bar{x}})$ = 0.328(0.014)) > $α$ _inferred ( $\bar{x} (σ_{\bar{x}})$ = 0.194(0.010))
Learning type x volatility	F(1,100) = 4.882, p = 0.029	$α$ _{vol_experienced} ( $\bar{x} (σ_{\bar{x}})$ = 0.341(0.017)) > $α$ _{stable_experienced} ( $\bar{x} (σ_{\bar{x}})$ = 0.315(0.015); F(1,100) = 3.916, p = 0.051). No significant difference between $α$ _{stable_inferred} ( $\bar{x} (σ_{\bar{x}})$ = 0.201(0.012)) and $α$ _{vol_inferred} ( $\bar{x} (σ_{\bar{x}})$ = 0.187(0.011); F(1,100) = 1.675, p = 0.199)
Learning type x group	F(1,100) = 4.097, p = 0.046	$α$ _social ( $\bar{x} (σ_{\bar{x}})$ = 0.161(0.014)) < $α$ _non-social ( $\bar{x} (σ_{\bar{x}})$ = 0.227(0.014)) for inferred (F(1,100) = 11.349, p = 0.001). No difference between groups for experienced (social $\bar{x} (σ_{\bar{x}})$ 0.329(0.021), non-social $\bar{x} (σ_{\bar{x}})$ 0.328(0.020); F(1,100) = 0.002, p = 0.966)

Appendix 5—table 1

Coefficients from regression models with MPH and PLA accuracy as dependent variables.

	$β$	$β$ (std. Err.)	standardised $β$	t	p
*PLA*
constant	0.67	0.02		30.15*	<0.001
$β$ _PLA	−0.25	0.02	−0.66	−12.84*	<0.001
ζ_PLA	0.03	0.00	0.43	8.53*	<0.001
$α$ _{stable_experienced_PLA}	0.05	0.04	0.09	1.35	0.18
$α$ _{vol_experienced_PLA}	0.10	0.04	0.18	2.67*	0.01
$α$ _{stable_inferred_PLA}	−0.03	0.04	−0.03	−0.56	0.58
$α$ _{volatile_inferred_PLA}	0.13	0.05	0.16	2.84*	0.01
*MPH*
constant	0.69	0.02		39.88*	<0.001
$β$ _MPH	−0.26	0.02	−0.78	−15.92*	<0.001
ζ_MPH	0.02	0.00	0.44	9.15*	<0.001
$α$ _{stable_experienced_MPH}	0.03	0.03	0.06	1.05	0.30
$α$ _{vol_experienced_MPH}	0.10	0.03	0.22	4.05*	<0.001
$α$ _{stable_inferred_MPH}	0.05	0.03	0.08	1.49	0.14
$α$ _{volatile_inferred_MPH}	0.03	0.04	0.05	0.92	0.36

* indicates statistical significance at p < 0.05.

Appendix 6—table 1

Mean(standard deviation) scores for neuropsychological tests and self-report questionnaires for the groups that received PLA and MPH on day 1.

Significance level for the between group differences are reported. Self-report questionnaires include the Barratt Impulsiveness Scale (BIS-11; Patton et al., 1995), the Behavioural Inhibition Scale/Behavioural Activation Scale (BISBAS) (Carver and White, 1994), Need for Cognition Scale (NCS) (Cacioppo et al., 1984), Multidimensional Scale of Perceived Social Support (MSPSS) (Zimet et al., 1988), Barratt Simplified Measure of Social Status (BSMSS) (Barratt, 2012), Sociable and Aggressive Dominance Questionnaire (SADQ) (Kalma et al., 1993), Beck Depression Inventory II (BDI-II) (Beck et al., 1996), Spielberger Trait Anxiety Inventory (STAI) (Spielberger et al., 1983).

		Group 1 PLA Day 1	Group 2 MPH Day 1
Neuropsychological tests	Listening span	5.0 (0.9)	4.6 (1.2)	p = 0.160
	NLV	94.4 (7.6)	92.6 (7.6)	p = 0.230
	Digit span – forward	17.2 (3.7)	16.2 (3.6)	p = 0.158
	Digit Span - backward	14.7 (3.4)	13.9 (2.7)	p = 0.219
Self-report questionnaires	Impulsivity (BIS-11)	63.5 (8.9)	60.2 (7.9)	p = 0.052*
	Behavioural inhibition (BIS)	16.4 (3.7)	16.3 (3.5)	p = 0.899
	Behavioural activation (BAS)	22.8 (3.9)	23.9 (4.0)	p = 0.166
	Need for cognition (NCS)	64.5 (10.5)	62.2 (10.5)	p = 0.264
	Social support (MDSPSS)	71.1 (10.1)	69.3 (9.6)	p = 0.349
	Social status (BSMSS)	49.8 (12.1)	45.9 (12.7)	p = 0.114
	Social dominance (SADQ)	4.1 (0.9)	4.1 (0.8)	p = 0.823
	Aggressive dominance (SADQ)	2.6 (0.6)	2.6 (0.6)	p = 0.691
	Depressive symptoms (BDI)	3.5 (3.7)	3.6 (3.9)	p = 0.969
	Anxiety symptoms (STAI)	32.4 (6.6)	32.4 (7.2)	p = 0.995

^*One participant had an outlying score on the BIS-11. Without outlier: p = 0.090.

Additional files

Transparent reporting form: https://cdn.elifesciences.org/articles/51439/elife-51439-transrepform-v2.docx
Download elife-51439-transrepform-v2.docx

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Mendeley

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

Jennifer L Cook
Jennifer C Swart
Monja I Froböse
Andreea O Diaconescu
Dirk EM Geurts
Hanneke EM den Ouden
Roshan Cools

(2019)

Catecholaminergic modulation of meta-learning

eLife 8:e51439.

https://doi.org/10.7554/eLife.51439

Share this article

Cite this article

Task design.

Probabilistic schedules.

Drug effects on win-stay, lose-shift beta values.

Illustration of relationship between experienced-value βWSLS_vol-stable and accuracy.

Drug effects on learning rates.

Experienced-value learning rates under PLA, but not MPH, were further from the optimal value in the stable compared to volatile phase.

Participant data (left) juxtaposed against model simulations (right).

Participant data (left) juxtaposed against model simulations (right).

p(y|m) = posterior model probability, ϕ = exceedance probability, MPH = purple, PLA = green.

Coefficients from regression model with MPH-PLA accuracy as dependent variable.

There were no significant differences between the social and non-social groups in age, gender or socioeconomic status as measured using the Barratt Simplified Measure of Social Status (Barratt, 2012).

Win-stay, lose-shift analysis.

Learning rate analysis.

Coefficients from regression models with MPH and PLA accuracy as dependent variables.

Mean(standard deviation) scores for neuropsychological tests and self-report questionnaires for the groups that received PLA and MPH on day 1.

Transparent reporting form

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

Illustration of relationship between experienced-value $β$ _{WSLS_vol-stable} and accuracy.

$p (y | m)$ = posterior model probability, ϕ = exceedance probability, MPH = purple, PLA = green.