1. Neuroscience
Download icon

Noradrenaline blockade specifically enhances metacognitive performance

  1. Tobias U Hauser Is a corresponding author
  2. Micah Allen
  3. Nina Purg
  4. Michael Moutoussis
  5. Geraint Rees
  6. Raymond J Dolan
  1. University College London, United Kingdom
  2. Max Planck University College London Centre for Computational Psychiatry and Ageing Research, United Kingdom
Research Advance
Cited
2
Views
5,277
Comments
0
Cite as: eLife 2017;6:e24901 doi: 10.7554/eLife.24901

Abstract

Impairments in metacognition, the ability to accurately report one’s performance, are common in patients with psychiatric disorders, where a putative neuromodulatory dysregulation provides the rationale for pharmacological interventions. Previously, we have shown how unexpected arousal modulates metacognition (Allen et al., 2016). Here, we report a double-blind, placebo-controlled, study that examined specific effects of noradrenaline and dopamine on both metacognition and perceptual decision making. Signal theoretic analysis of a global motion discrimination task with adaptive performance staircasing revealed that noradrenergic blockade (40 mg propranolol) significantly increased metacognitive performance (type-II area under the curve, AUROC2), but had no impact on perceptual decision making performance. Blockade of dopamine D2/3 receptors (400 mg amisulpride) had no effect on either metacognition or perceptual decision making. Our study is the first to show a pharmacological enhancement of metacognitive performance, in the absence of any effect on perceptual decision making. This enhancement points to a regulatory role for noradrenergic neurotransmission in perceptual metacognition.

https://doi.org/10.7554/eLife.24901.001

Main text

Making a decision is often accompanied by a conscious feeling of confidence (Flavell, 1979). Subjective confidence reports typically show a good correspondence to actual task performance, reflecting a metacognitive ability for accurate introspection (Fleming et al., 2010). Impairments in metacognition can compromise decision making and lead to misjudgements of actual performance, as found in several psychiatric dimensions, such as schizophrenia, attention-deficit/hyperactivity disorder or compulsivity (Frith, 1992; Knouse et al., 2005; Lysaker et al., 2010; Hauser et al., 2017).

The neurocognitive mechanisms from which confidence, and metacognitive ability in general, arise are ill understood. While classic accounts see confidence as a mere extension of a perceptual sampling process (Kiani and Shadlen, 2009; Pleskac and Busemeyer, 2010; Meyniel et al., 2015; Moran et al., 2015), other evidence points to a non-trivial relationship between decision making and confidence that invoke distinct decision making and metacognitive processes (Fleming et al., 2010; Fleming and Dolan, 2012; Allen et al., 2016; Allen et al., 2017). Metacognition can thus be understood as incorporating both decision-related and domain-general information (Fleming and Daw, 2017) and frontopolar and hippocampal brain structures, for example, have been shown to contribute specifically to metacognition but not to perceptual decision making (Fleming et al., 2010, 2012; Allen et al., 2017). In our previous study, we provided evidence that arousal can bias metacognition independently of decision accuracy (Allen et al., 2016), in accordance with other studies showing confidence-accuracy dissociations (Fleming et al., 2015; Spence et al., 2016) and suggested that these biases might be under neuromodulatory control via neural gain (Eldar et al., 2013; Hauser et al., 2016).

Two candidate neuromodulators likely to affect metacognition are the catecholamines noradrenaline and dopamine. These neurotransmitters have their origin in brainstem nuclei that project broadly to cortical and subcortical regions, including prefrontal cortex and hippocampus (Hauser et al., 2016). Both dopamine and noradrenaline contribute to the regulation of arousal and higher-order cognition (Usher et al., 1999; Aston-Jones and Cohen, 2005; Yu and Dayan, 2005; Pessiglione et al., 2006; De Martino et al., 2008; Chowdhury et al., 2013; Eldar et al., 2013; Rigoli et al., 2016), and their dysregulation is widely inferred to contribute to various manifestations of psychiatric illness (Yamamoto and Hornykiewicz, 2004; Laruelle, 2013; Hauser et al., 2016). However, whether and which of these neuromodulators influences metacognition is unknown.

Here we assessed whether noradrenaline or dopamine have a distinct influence on metacognition, independent of any effect on perceptual decision making. The latter consideration is important because differences in perceptual decision making can distort an assessment of metacognition (Fleming and Lau, 2014). Consequently, we used an approach where we kept perceptual decision making equivalent across subjects, using a staircase procedure. This, combined with a signal detection analysis (Green and Swets, 1966; Galvin et al., 2003), enabled us to measure drug effects on metacognition while controlling for potential effects on perceptual decision making. This overcomes a limitation of a previous study that assessed the impact of dopamine on confidence (Lou et al., 2011).

To examine the influence of noradrenaline and dopamine we employed two pharmacological manipulations. Many pharmacological agents have high affinity for both dopaminergic and noradrenergic receptors and synaptic function. On this basis, we selected drugs with selective high affinity, the β-adrenoceptor antagonist propranolol in the case of noradrenaline, and the D2/3 receptor antagonist amisulpride in the case of dopamine. In a double-blind, placebo-controlled design we demonstrate that the noradrenergic agent propranolol uniquely improves metacognition in the absence of an effect on perceptual performance, with no effect seen following administration of the dopamine antagonist amisulpride.

Results

Noradrenaline blockade modulates metacognition

To examine effects of noradrenaline and dopamine (versus placebo) on metacognition we performed a double-blind, between-subjects, placebo-controlled study. Each of the three groups consisted of 20 subjects matched for gender, age, affect (Watson et al., 1988), and intellectual abilities (Wechsler, 1999) (Table 1). Due to differences in their pharmacokinetic properties we administered active drugs orally at two different time points. The dopamine group received 400 mg of amisulpride (selective D2/3 antagonist) 110 min prior to a metacognition task and an additional placebo 30 min after the amisulpride administration (Figure 1A). The noradrenaline group received a placebo at 110 min prior to the task and then 40 mg of propranolol (non-selective β-adrenoceptor antagonist) 30 min after placebo administration. The placebo group received placebo at the both time points to match the administration schedules of the other groups. A post-experiment evaluation revealed that subjects were not aware of whether and which drug they received (χ2(4)=1.26, p=0.868; missing data from two subjects). There were no effects of drug on mood (PANAS; Watson et al., 1988) ratings (main effect of drug: F(2,57)=.16, p=0.852; time x drug: F(2,57)=.19, p=0.827; time x affect x drug: F(2,57)=2.17, p=0.124).

Experimental design and metacognition task.

(A) After filling out a baseline mood questionnaire (PANAS pre), subjects received two different drugs 110 and 80 min prior to the metacognition task. A dopamine subject group first received 400 mg amisulpride (dopamine D2/3 receptor antagonist) and subsequently placebo, whereas the noradrenaline group first received placebo and then 40 mg propranolol (β-adrenoceptor antagonist). Subjects of a placebo group received placebo at both times. Eighty minutes after the second drug administration, subjects filled out a second mood questionnaire (PANAS post) and then performed a metacognition task. (B) To assess subjects’ metacognitive abilities, we used a global motion discrimination task with subsequent confidence judgements. After a fixation period, subjects saw 1100 dots moving randomly with an average motion pointing either to the left or right. After 250 ms, subjects had to indicate the overall direction of the moving dots by using keyboard arrows. Subsequently, they indicated their confidence about their decision using a sliding visual analogue scale. Subjects were instructed to use the full width of the scale by indicating high confidence on the right and low confidence on the left side of the scale.

https://doi.org/10.7554/eLife.24901.002
Table 1

Group characteristics. The three groups did not differ in their gender, age, intellectual abilities (IQ) (Wechsler, 1999) and their positive and negative affective states before and after drug administration. PANAS: positive and negative affective schedule (Watson et al., 1988), PA: positive affect, NA: negative affect, pre: before drug administration, post: after drug administration (mean±SD).

https://doi.org/10.7554/eLife.24901.003

PlaceboPropranololAmisulpride
gender (f/m)10/1010/1010/10
age24.50 ± 4.1623.15 ± 4.3122.35 ± 2.21F(2,57)=1.74, p=0.185
IQ112.45 ± 12.22118.75 ± 8.55114.60 ± 11.77F(2,57)=1.70, p=0.191
PANAS PA pre31.15 ± 10.0827.70 ± 8.2828.90 ± 6.60F(2,57)=0.86, p=0.428
PANAS NA pre11.70 ± 2.2313.55 ± 5.4813.10 ± 3.23F(2,57)=1.23, p=0.300
PANAS PA post29.22 ± 10.4727.15 ± 7.7527.80 ± 8.12F(2,57)=0.286, p=0.752
PANAS NA post11.45 ± 2.3711.95 ± 4.8711.25 ± 1.92F(2,57)=0.236, p=0.790

Eighty minutes after the second drug administration, subjects performed a visual global motion discrimination task that included confidence judgements (Figure 1B). Subjects decided whether the overall motion of a short burst of randomly moving dots was directed to the left or right of the vertex. Subsequently, and in the absence of feedback on whether they were correct or not, they indicated confidence in their decision on that trial using a sliding visual analogue scale. To control for potential drug effects on perceptual performance, we matched the subjects’ decision accuracy by continuously adapting the global motion orientation using a staircase procedure (Cornsweet, 1962).

To examine metacognitive abilities, we analysed type-II performance as derived from signal detection theory (Green and Swets, 1966; Galvin et al., 2003). This measures subjects’ awareness into their own performance by assessing how well their confidence ratings match their true accuracy (i.e., ‘how much more confident am I if I make a correct vs incorrect decision’). We calculated type-II area under the receiver-operating-characteristics curve (AUROC2) (Fleming et al., 2010, 2012; Weil et al., 2013; Allen et al., 2017) for each subject and then compared this metric between groups using an ANOVA with drug group as a between-subjects factor. The analysis revealed a significant effect of drug on AUROC2 (Figure 2, F(2,55)=5.192, p=0.009, η²=0.16), with follow up t-tests showing the propranolol group performed significantly better than a placebo group (t(38)=4.00, p<0.001, d = 1.26). The propranolol group also performed marginally better than an amisulpride group (t(36)=2.02, p=0.051, d = 0.65) with the latter having an equal performance as the placebo group (t(36)=.74, p=0.465, d = 0.23). To evaluate evidence for this null effect in the amisulpride group, we additionally performed a Bayesian two-sample t-test comparing the placebo and amisulpride groups (Rouder et al., 2009; Dienes, 2014). This analysis revealed a Bayes Factor of 3.31, corresponding to moderate evidence for the null hypothesis. These results indicate that inhibition of noradrenergic function improves metacognitive insight, in the absence of any effect of amisulpride.

Propranolol improves metacognitive abilities.

(A) Signal detection theoretic analysis revealed a significantly increased metacognitive ability, as measured by the type-II area under the ROC curve (AUROC2). (B) A highly significant effect of propranolol compared to placebo shows that propranolol increases metacognitive abilities. The difference between propranolol and amisulpride suggests that this performance increase might be specific to an influence on noradrenaline but not dopamine function. (C) Individual AUROC2 metrics show that most subjects in the propranolol group perform above the median metacognitive performance (dotted line), while perceptual decision making performance was relatively stable across all groups. mean ±1 SEM; fat line: ANOVA; square brackets: t-tests.

https://doi.org/10.7554/eLife.24901.004

Improved metacognition mainly driven by confidence on error trials

To further understand the noradrenaline-induced metacognitive enhancement, we compared median confidence ratings for correct and incorrect trials between the propranolol and placebo group. A significant group-by-correctness interaction (F(1,38)=8.66, p=0.006, η²=0.19) in the absence of a group main effect (F(1,38)=1.83, p=0.185, η²=0.05) suggests that a confidence rating difference between correct and incorrect trials underlies the observed group differences. Subsequent t-tests demonstrated that this effect was primarily driven by error trials (error trials: t(38)=2.17, p=0.036, d = 0.69, correct trials: t(38)=-.193, p=0.848, d = 0.06), suggesting that the propranolol group exhibited lower confidence for error trials.

Metacognitive differences are not explained by perceptual performance

Metacognitive measures, as used here, can be influenced by differences in perceptual performance (Fleming and Lau, 2014). We deliberately used a staircase procedure to keep performance equivalent between groups (mean accuracy: F(2,55)=1.60, p=0.212, η²=0.05, cf. Figure 2C). A signal-detection theoretic analysis confirmed these findings, revealing the absence of any significant differences in either perceptual sensitivity d’ (Figure 3C, F(2,55)=1.69, p=0.194, η²=0.07) or response bias c (F(2,55)=2.29, p=0.112, η²=0.08) (Green and Swets, 1966). To additionally ensure that differences in AUROC2 were not influenced by any of these measures, we also compared AUROC2 using an ANCOVA with d’, response bias c, and stimulus signal strength (mean orientation) as covariates, revealing the same group difference for AUROC2 after controlling for these potential biases (F(2,51)=4.99, p=0.010, η²=0.17).

Drug effects on perceptual decision making.

No drug effects were observed on the signal strength (A, stimulus motion orientation) or the response speed (B). In line with no difference in accuracy, perceptual sensitivity d’ did not differ between groups (C). Median confidence ratings (D) showed no difference revealing that there was no bias in the average rating behaviour between groups. These findings suggest that noradrenaline blockade selectively boosts metacognitive sensitivity in the absence of any effect on perceptual decision making. mean ±1 SEM; n.s. p>0.10.

https://doi.org/10.7554/eLife.24901.006

No drug effect on perceptual decision making

To test whether perceptual decision making was affected by our drug interventions, we analysed whether stimulus strength, measured by mean stimulus motion orientation, differed between groups. There was no significant difference in stimulus strength (Figure 3A, F(2,55)=1.16, p=0.321, η²=0.04), indicating perceptual performance was not significantly affected by the drug manipulations. Likewise, there was no drug effect on reaction times (Figure 3B, F(2,55)=.87, p=0.424, η²=0.03). Lastly, to test whether there were baseline differences in how the groups were utilising the confidence scale, we examined the median confidence ratings, but found no difference (Figure 3D, F(2,55)=.38, p=0.684, η²=0.01), supporting the result that an enhanced metacognitive ability under propranolol is not due to a bias in use of the confidence rating scale.

Discussion

Confidence determines how much we trust our decisions and how strongly they influence future behaviour. A read out of confidence in a decision that fails to reflect actual performance will lead to poor decisions and long-term adverse outcomes. Impaired metacognition is reported in psychiatric disorders (Frith, 1992; Knouse et al., 2005; Lysaker et al., 2010; Wells, 2011; Hauser et al., 2017), and its pharmacological remediation could provide a target for treatment (Wells, 2011). Here we show that inhibition of central noradrenaline (by means of propranolol) function enhances perceptual metacognitive ability. A dopamine blockade (by means of amisulpride) had no impact on metacognition and neither drug manipulation had an impact on core perceptual performance.

Noradrenaline is known to impact arousal and higher-order cognition, but the precise mechanisms remain obscure. Influential accounts propose noradrenergic modulation of information processing, either through neural gain (Aston-Jones and Cohen, 2005; Eldar et al., 2013) or by signalling unexpected uncertainty (Yu and Dayan, 2005; Dayan et al., 2006). Our finding that blocking noradrenaline leads to improved metacognitive performance can be understood within both frameworks. Metacognition can be thought of as a higher-order process that follows a perceptual decision making stage and integrates perceptual and other sources of information, such as interoceptive states and general arousal (Allen et al., 2016), to form an overall confidence judgement. The neural gain hypothesis (Aston-Jones and Cohen, 2005; Eldar et al., 2013) proposes that noradrenaline amplifies strong and diminishes weak signals throughout the brain, with the effect of an increased ‘contrast’ between strong and weak signals. Due to a nonlinearity in this amplification it is likely to neglect subtle signal differences, and thus omit the breath and detail of information conveyed. This in turn means noradrenaline might render detailed stimulus properties unavailable to the metacognitive process, impairing the precision of a metacognitive judgement. The latter theory (Dayan et al., 2006) suggests phasic noradrenaline is elicited by unexpected uncertainty or arousal, such as when making an erroneous choice (Ullsperger et al., 2010). This phasic burst acts by interrupting ongoing processes and leads to a resetting and erasure of currently maintained information to enable an orienting response (Sokolov et al., 2002; Dayan et al., 2006). In the context of our paradigm, this suggests that following an incorrect response accumulated sensory information is reset and unavailable for confidence judgement, leading to poorer metacognitive performance. This is supported by our finding of a primary drug-effect on confidence in erroneous trials. On both accounts, a blockade of noradrenaline by propranolol hinders the noradrenaline-related loss of information to provide complete perceptual information to a confidence-related process. This is also in line with our previous findings showing that unexpected arousal biases metacognition (Allen et al., 2016), an effect possibly modulated by noradrenaline.

In our experiment, perceptual metacognition was solely influenced by manipulation of noradrenaline, and not by blocking dopamine D2/3 receptors. This is of interest as both neuromodulators are often ascribed similar functions including a role in exploration (Hauser et al., 2016), neural gain (Eldar et al., 2013; Fiore et al., 2016; Hauser et al., 2016), salience (Bromberg-Martin et al., 2010; Kahnt and Tobler, 2013), (prediction) error signalling (Holroyd and Coles, 2002; Hauser et al., 2014) and effort processing (Bouret et al., 2012). Likewise, neural recordings from dopaminergic and noradrenergic brainstem nuclei have revealed surprisingly similar neural response patterns (Bouret et al., 2012; Varazzani et al., 2015). Our finding of enhanced perceptual metacognition with noradrenaline blockade might reflect a rare sensitivity to the actions of one of these neuromodulators. Our findings also raise the possibility that a recent report of increased performance and confidence ratings following dopaminergic enhancement (Lou et al., 2011) may reflect a performance, but not a metacognitive, effect. This finding is akin to a previous report of specific testosterone effects, where there was an effect on perceptual performance but not metacognition (Wright et al., 2012).

An important caveat for comparing the amisulpride and propranolol groups directly is that little is known about the precise pharmacokinetics and how comparable the dosage effects are. We took great care in the design of the study to render the two drug conditions as comparable as possible. First, because of the slightly different absorption rates, we administered amisulpride 30 min before propranolol, in keeping with previous drug schedules (Peretti et al., 1997; Ramaekers et al., 1999; Strange et al., 2003; Silver et al., 2004; Strange and Dolan, 2004; Hurlemann et al., 2005; Alexander et al., 2007; Gibbs et al., 2007; De Martino et al., 2008; Kahnt et al., 2015; Kahnt and Tobler, 2017). Second, to render the cognitive effects of the drugs as similar as possible, we selected dosages that were commonly reported in previous studies of neurocognition (i.e. 40 mg propranolol, 400 mg amisulpride) (e.g., Ramaekers et al., 1999; Strange et al., 2003; Silver et al., 2004; Strange and Dolan, 2004; Hurlemann et al., 2005; Alexander et al., 2007; Gibbs et al., 2007; De Martino et al., 2008; Kahnt et al., 2015; Kahnt and Tobler, 2017). However, we know little about the magnitude of these drug effects on the brain. A previous study of sulpiride, which has a similar chemical formulation to amisulpride, but slightly different pharmacokinetics, suggests that a single-dose of 400 mg leads to a occupancy of ~28% of D2 receptors (Mehta et al., 2008). Unfortunately, there are no PET studies reporting on single-dose amisulpride, and there are no occupancy studies of propranolol, thus rendering it difficult to directly quantify and compare our dosage effects.

In this study, we show that noradrenaline specifically influences perceptual metacognition but not perceptual decision making. It is interesting to speculate whether our findings are generalizable to metacognition in non-perceptual domains. Recent studies show that a metacognitive ability is relatively stable across different perceptual decision making tasks (Song et al., 2011; McCurdy et al., 2013), even when probing different sensory modalities (de Gardelle et al., 2016; Garfinkel et al., 2016). However, it is unclear whether metacognition within different cognitive domains (e.g., perception vs memory) rely on the same processes, with evidence from neuroimaging suggesting that these functions utilise unique neural networks (Baird et al., 2013; Fleming et al., 2014). Given that noradrenaline modulates activity on a whole-brain level (Hauser et al., 2016), it is possible that noradrenergic metacognition effects can be observed in domains other than perception, an interesting area for future studies. Lastly, our previous findings of an embodied reflection of confidence by means of cardiac and pupil responses (Allen et al., 2016) raise the question as to whether the observed noradrenaline effects are purely a consequence of central changes, or whether peripheral effects of this drug influence metacognitive performance independently. A drug that exclusively targets peripheral, but not central, noradrenaline (cf. De Martino et al., 2008) could provide insight into the question of visceral contributions to metacognition as suggested in ideas on embodied cognition (Allen and Friston, 2016).

In conclusion, using a double-blind, placebo-controlled drug manipulation we show that noradrenaline has a controlling influence on metacognitive ability. Metacognition is enhanced following a blockade of noradrenergic β-adrenoceptors an observation suggesting potential remedial avenues for metacognitive insight deficits seen in psychiatric patients.

Materials and methods

Subjects

Sixty subjects participated in this double-blind, placebo-controlled, between-subjects study. Each subject was randomly allocated to one of three drug groups, controlling for an equal gender balance in all groups. Candidate subjects with a history of neurological or psychiatric disorders, current health issues, regular medications (except contraceptives), or prior allergic reactions to drugs were excluded from the study. Subjects had normal or corrected-to-normal vision. The groups were matched for age, mood (PANAS; before and after drug administration) (Watson et al., 1988) and intellectual abilities (WASI abbreviated version) (Wechsler, 1999) (Table 1). Subjects were reimbursed for their participation on an hourly basis. The study was approved by the UCL research ethics committee and all subjects gave written informed consent.

Drug manipulation and procedures

To attenuate noradrenergic function we administered 40 mg of propranolol, a non-selective β-adrenoceptor antagonist. To attenuate dopamine function we administered 400 mg of amisulpride, a selective D2/3 antagonist. These drugs were chosen because of their selective high affinity effects on either one or the other of these two neuromodulators, enabling a specific dissociation of their contribution to metacognitive ability. The dosage and timing of both propranolol and amisulpride were based on previous studies that have investigated their effects on cognition (e.g., Ramaekers et al., 1999; Strange et al., 2003; Silver et al., 2004; Strange and Dolan, 2004; Hurlemann et al., 2005; Alexander et al., 2007; Gibbs et al., 2007; De Martino et al., 2008; Kahnt et al., 2015; Kahnt and Tobler, 2017).

Prior to the task the drugs were administered at two different time points, based upon pharmacokinetic considerations (Figure 1A). The first drug was administered 110 min prior to the metacognition task. At that time, the dopamine group received amisulpride while the other groups received placebo. After 30 min, subjects consumed a second drug. This time, the noradrenaline group received propranolol, while the dopamine and placebo group consumed a placebo. A placebo group received placebo at both times. The task was performed 80 min after the second drug administration.

Experimental paradigm

To measure metacognitive ability we applied an adaptive visual global motion detection paradigm, similar to the version in our previous study (Allen et al., 2016), implemented using Psychtoolbox-3 (www.psychtoolbox.org) for MATLAB (R2010a). On every trial, subjects viewed a brief burst of motion (250 ms), followed by a forced choice to determine if the overall motion direction was to the left or right of vertical. Subjects then rated their subjective confidence using a continuous sliding scale marked at four equal intervals by horizontal lines. To prevent response preparation, the starting point of the confidence marker was jittered up to 12% to the left or right of scale midpoint on each trial. Subjects had up to 1500 ms to make their motion choice, and 2500 ms to report their confidence.

At the start of the experiment, each subject was instructed that the goal of the task was to measure their perceptual and metacognitive sensitivity. This was operationally defined as their ability to detect motion direction and how accurately their confidence ratings reflected their actual detection performance. Subjects first completed a short training session of 140 detection-only trials to establish motion thresholds. All subjects achieved staircase stabilization before continuing to the main experiment (the motion direction threshold reached in the final staircase of training was used as the starting point for main task). Subjects were encouraged to use the entire scale to report their subjective feeling of confidence, and to carefully reflect on each trial on the decision they had just made. Confidence reports were then binned into six equally sized bins for further analysis similar to previous reports (Fleming et al., 2010, 2012; Allen et al., 2017; Hauser et al., 2017).

On each trial subjects viewed a cloud of 1100 moving black dots of 0.08 degrees visual angle (DVA), presented for 250 ms within a 15.69 DVA circular array at random starting positions and advancing at a speed of 0.06 DVA per frame. Dots which moved beyond the stimulus aperture were replaced at the opposite edge to maintain constant dot density. To prevent fixation on local motion directions, all dots had a randomized limited lifetime of maximum 93% (14 frames). Each motion stimulus was defined by a global motion direction (‘orientation’) to the left or right of vertical. Following experiments investigating confidence with global motion stimuli, dot mean and variance were manipulated independently of one another (Allen et al., 2016; Spence et al., 2016). To this end, all dots were ‘signal’ dots and the standard deviation of the mean direction was adjusted across conditions. On each trial the motion signal was thus calculated using the formula:

Dotdirection=(LeftvsRight)MeanOrientation+GaussianNoiseStandardDeviation

To control task difficulty and thus ensure an unbiased estimate of metacognitive sensitivity (Fleming and Lau, 2014), the mean direction of motion was continuously adapted for each subject using a 2-up-1 down staircase, which converges at the limit on a 71% detection accuracy. To render the staircase opaque to subjects, and maximize confidence variability, we deployed two separate staircase conditions with a fixed motion variance equal to either 20 or 30 degrees standard deviation. Subjects completed a total of 144 trials (72 high variance, 72 low variance) divided evenly between four blocks.

Statistical analysis and metacognition modelling

The goal of our analyses were two-fold: First, we wanted to ensure all groups expressed equivalent perceptual decision making performance, as performance differences can influence estimates of metacognitive ability (Fleming and Lau, 2014). Second, we wanted to test for differences in metacognition using signal detection theory (SDT), examining the metacognitive detection performance using an area under the curve for a type-II receiver-operating-characteristics (ROC) (AUROC2) metric (Fleming et al., 2010, 2012).

To ensure homogeneous performance across all subjects, we excluded two outlier subjects who performed worse than the rest (as measured using boxplots; both subjects belonged to the amisulpride group). We omitted all trials from the first block to ensure staircase stabilisation, similar to previous studies (Fleming et al., 2010, 2012; Allen et al., 2017; Hauser et al., 2017). Trials with early (<100 ms), late (>1500 ms) or missing responses were excluded. We collapsed low and high variance trials, leaving a total of 108 trials per subject, per recommended procedures, for optimal estimation of SDT measures (Fleming and Lau, 2014). To compare perceptual abilities between groups, we assessed their performance in terms of accuracy and signal strength (mean stimulus orientation). We further assessed reaction times and the subjects’ average confidence ratings. We used ANOVAs with a between-subject factor group (placebo, propranolol, amisulpride) and post-hoc t-tests in SPSS (version 22, IBM). To evaluate the evidence for the null hypothesis that amisulpride had no effect on AUROC2, we performed a Bayesian two-sample t-test using version 0.9.8 of the BayesFactor package, computed using R version 3.3.2 (2016-10-31) on x86_64-w64-mingw32 (Rouder et al., 2009; Morey and Rouder, 2015). A unit-information Bayes Factor with r scale parameter = 1 was calculated for the Placebo vs Amisulpride contrast (Rouder et al., 2009). This Bayes factor expresses the continuous evidence for the null hypothesis of no drug effect, where Bayes Factors > 3 correspond to ‘moderate’ evidence (Rouder et al., 2009; Dienes, 2014).

To examine subjects’ metacognitive abilities, we assessed the type-II performance using SDT. In this framework, metacognition can be modelled as the sensitivity of subjective confidence to underlying ground truth discrimination performance. By defining metacognitive ‘hits’ (i.e., high confidence for correct detections) and ‘misses’ (high confidence for incorrect detections), metacognitive sensitivity can be expressed as the area under a type-II receiver-operating-characteristics curve (AUROC2). In contrast to classical measures of metacognition (e.g., the correlation of confidence and accuracy), AUROC2 is unbiased by a subject's overall level of confidence (or metacognitive bias/criterion) if detection performance is held constant across subjects (Fleming and Lau, 2014). Further, being nonparametric, AUROC2 is not susceptible to issues such as non-normal confidence distributions.

AUROC2 was calculated using the same metric as in Fleming et al. (2010), Kornbrot (2006):

(1.1) AUROC2=14k=112i([hk+1fk]2[hkfk+1]2)+14k=12ii([hk+1fk]2[hkfk+1]2)

where i indicates the six confidence rating bins, h depicts the relative frequency of this rating for correct choices (hi=p(confidence==i|correct)) and f describes the counterpart for incorrect responses (fi=p(confidence==i|incorrect)).

To ensure that the groups did not differ in their type-I detection performance (d‘) or response bias (c), we additionally examined these metrics (Green and Swets, 1966; Fleming et al., 2010):

(1.2) d=12(z(H)z(FA))
(1.3) c=.5(z(H)z(FA))

where z describes the inverse of a cumulative normal distribution, H is the correct hits and FA the false alarms for two-alternative forced choice tasks.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
    The neural basis of metacognitive ability
    1. SM Fleming
    2. RJ Dolan
    (2012)
    Philosophical Transactions of the Royal Society B: Biological Sciences 367:1338–1349.
    https://doi.org/10.1098/rstb.2011.0417
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
    The Cognitive Neuropsychology of Schizophrenia
    1. CD Frith
    (1992)
    Psychology Press.
  26. 26
  27. 27
  28. 28
  29. 29
    Signal Detection Theory and Psychophysics
    1. DM Green
    2. JA Swets
    (1966)
    John Wiley and Sons.
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
    BayesFactor
    1. RD Morey
    2. JN Rouder
    (2015)
    BayesFactor, Version 0.9.11-3.
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
    The Orienting Response in Information Processing
    1. EN Sokolov
    2. JA Spinks
    3. R Näätänen
    4. H Lyytinen
    (2002)
    Mahwah, NJ, US: Lawrence Erlbaum Associates Publishers.
  57. 57
  58. 58
    Computations underlying confidence in visual perception
    1. ML Spence
    2. PE Dux
    3. DH Arnold
    (2016)
    Journal of Experimental Psychology: Human Perception and Performance 42:671–682.
    https://doi.org/10.1037/xhp0000179
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
    Wechsler Abbreviated Scale of Intelligence
    1. D Wechsler
    (1999)
    New York: The Psychological Corporation: Harcourt Brace and Company.
  66. 66
  67. 67
    Metacognitive Therapy for Anxiety and Depression, Reprint Edition
    1. A Wells
    (2011)
    New York, NY: Guilford Press.
  68. 68
  69. 69
  70. 70

Decision letter

  1. Haozhe Shan
    Reviewing Editor; University of Chicago, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Noradrenaline, but not dopamine, modulates metacognitive ability". Your article has been reviewed by two peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation process was overseen by Sabine Kastner as Senior Editor. The reviewers were supportive of the manuscript, but recommended the following major revisions, before a decision on publication can be made.

In the current submission, the authors put forth a clearly designed behavioral task to probe the role of noradrenaline and dopamine on perceptual confidence. Using amisulpride and propranolol as dopamine/noradrenaline blockers, the authors reported that administration of propranolol affected perceptual confidence judgment in the subjects, while that of amisulpride did not. The reviewers' biggest concern is one that is common with studies using pharmacological manipulations: the two agents used have effects other than the ones desired in the study, and they have different pharmacokinetics and dosage effects. The authors gave some considerations to these factors, but there are still uncontrolled aspects. The timeframe of the administration, the dosage of the two drugs, and the other targets of the drugs are factors that concern the reviewers the most. The authors are advised to discuss these factors further and provide justification for the practice in the study, or tone down the thesis to a comparison between propranolol and amisulpride administration, not noradrenaline and dopamine.

Reviewer #1:

In this Research Advance article, Hauser and colleagues presented an elegant and straightforward experiment, probing roles of dopamine noradrenaline in metacognition. Administration of noradrenaline but not dopamine affects metacognitive judgment accuracy but not the cognitive performance itself. The results pave a road for further investigations into how visceral senses influence metacognition. The manuscript is overall clearly written and informative. However, there are a few issues to be addressed.

First, the timing of drug administration is different in the noradrenaline group and the dopamine group. The authors stated that the motivation is that amisulpride and propranolol have different kinetics, which is reasonable. However, is there any control method to make sure that these two different schedules are equivalent to each other? In other words, if propranolol was injected further ahead of the testing, could it have had an effect? The authors should address the potential discrepancies here. (While the authors do cite previous studies as motivation for these specific times in the methods section, note that its metacognition, not cognition, that's investigated here; the authors themselves have argued in the Introduction that these two may have different mechanisms. Do findings regarding the kinetics of these drugs still apply in a metacognition study?)

Second, pharmacological agents are never precise effectors (e.g. the high affinity issue). Instead they may have side effects that cause issues. What are some of the effects amisulpride and propranolol may have, how strong they are at the doses used, and how may they confound interpretation of the results?

Finally, could any of the findings regarding metacognition be task-specific? For instance, could non-perceptual, more slow-paced decision making have different interactions with the drugs? The authors should include a discussion.

Reviewer #2:

This paper reports the effects of single dose administration of propranolol or amisulpride on self-rated confidence following perceptual decision making. In a double blind between-subjects design, they found that administration of propranolol increased the likelihood that participants assigned low confidence to decisions that were wrong. Amisulpride had no effect on self-rated confidence. Neither drug influenced perceptual performance. These results are framed in the context of the influence of noradrenaline and dopamine on meta-cognition.

The behavioural study is well designed and the results clearly reported. My concerns relate mostly to the pharmacological inferences and the framing of the paper.

1) Amilsulpride did not significantly influence performance (perceptual or self-rated confidence). It is quite possible that this null effect is simply due to the low occupancy of dopamine receptors with this single dose. To support this dose, the authors' cited paper (Kahnt et al., 2015) in turn cites Mehta (2008) whom in turn use PET (raclopride) to show that a single dose of 400mg of amilsupride leads to an estimated 28% receptor occupancy (much lower than clinical effects) which notably caused no significant changes across a broad armoury of cognitive tests (had they properly corrected for multiple tests!). This is consistent with Ramaekers (1999) – cited by Gibbs 2007 to justify the dose – who found no cognitive or behavioural effects following a single dose unless it was administered for 5 days in a row.

We do not know what the comparable receptor occupancy of central adrenoreceptors is following a single dose of 40mg of propranolol (there are no citations in the text and I believe there is no PET assay). Therefore, the apparent head-to head competition between noradrenaline and dopamine modulation is not unequivocally demonstrated. The difference could simply be due to a marked difference in the central effective pharmacological effects of these single doses. This might be fine if the paper framed itself around the actual pharmacological manipulations given – but the paper clearly stamps itself as a dopamine versus noradrenaline head-to-head. The authors offer insufficient justification for this imputation and list no caveats or limitations.

2) As I'm sure the authors are acutely aware, propranolol has both central and peripheral β blockade – in fact prior work by some of the authors used propranolol in combination with the selective peripheral blocker, nadolol, to disambiguate the central from the peripheral action. So there are inevitably cardiovascular and likely other autonomic consequences to the β blockade. Given the same group recently published a paper that linked autonomic fluctuations to decision confidence, this seems quite an oversight.

3) There is more of a style issue, but I found the start of the Discussion a much more accessible/more descriptive account of the goals of the paper, than the title or the Abstract – that is, I would have preferred to see a more descriptive approach to the actual task up front – self-rated decision following a perceptual task which was then interpreted as "metacognition". The latter is a conceptual construct that must surely be composed of other processes, and which is only (imperfectly) addressed in the present task. The broad readership will have to read deeply into the paper until they discover that this is a simple rating of confidence following a decision.

4) Were the subjects fasting? Were there checks to see if they were able to unblind themselves (particularly for the propranolol).

5) Was there a main effect of drug on PANAS?

Citations:

Mehta, M. A., Montgomery, A. J., Kitamura, Y., & Grasby, P. M. (2008). Dopamine D2 receptor occupancy levels of acute sulpiride challenges that produce working memory and learning impairments in healthy volunteers. Psychopharmacology, 196(1), 157-165.

Ramaekers JG, Louwerens JW, Muntjewerff ND, Milius H, de Bie A, Rosenzweig P, Patat A, O'Hanlon JF: Psychomotor, cognitive, extrapyramidal, and affective functions of healthy volunteers during treatment with an atypical (amisulpride) and a classic (haloperidol) antipsychotic. J Clin Psychopharmacology 1999; 19:209-221

https://doi.org/10.7554/eLife.24901.008

Author response

Reviewer #1:

In this Research Advance article, Hauser and colleagues presented an elegant and straightforward experiment, probing roles of dopamine noradrenaline in metacognition. Administration of noradrenaline but not dopamine affects metacognitive judgment accuracy but not the cognitive performance itself. The results pave a road for further investigations into how visceral senses influence metacognition. The manuscript is overall clearly written and informative. However, there are a few issues to be addressed.

First, the timing of drug administration is different in the noradrenaline group and the dopamine group. The authors stated that the motivation is that amisulpride and propranolol have different kinetics, which is reasonable. However, is there any control method to make sure that these two different schedules are equivalent to each other? In other words, if propranolol was injected further ahead of the testing, could it have had an effect? The authors should address the potential discrepancies here. (While the authors do cite previous studies as motivation for these specific times in the methods section, note that its metacognition, not cognition, that's investigated here; the authors themselves have argued in the Introduction that these two may have different mechanisms. Do findings regarding the kinetics of these drugs still apply in a metacognition study?)

We thank the reviewer for a positive evaluation of the manuscript and for raising thoughtful questions. We agree that directly comparing two drugs comes with specific challenges, some of which cannot be completely resolved with current methods. In this study, we took great care to make the drug conditions as comparable as possible. First, we tried to match the pharmacokinetics between the drugs. For propranolol, the peak plasma concentration is reached 1-2 hours after drug administration and most experimental studies are conducted approx. 90 minutes after drug administration (usually between 75 and 120 minutes) (cf. Alexander et al., 2007; De Martino et al., 2008; Hurlemann et al., 2005; Silver et al., 2004; Strange et al., 2003; Strange and Dolan, 2004). For amisulpride, the peak plasma concentration is between 1-4 hours, and experimental tasks are usually delivered any time between 1 and 3 (up to 6) hours after administration (cf. Gibbs et al., 2007; Kahnt et al., 2015; Kahnt and Tobler, 2017; Peretti et al., 1997; Ramaekers et al., 1999). Based on these established facts and procedures, we opted to administer amisulpride ahead of administering propranolol. However, given the relatively long duration of peak plasma levels (i.e. relatively slow metabolisation), we think it is reasonable to assume we would obtain similar results with slightly different administration regiments.

We used similar reasoning for selecting dosages for the two drugs. Based on prior literature, 40mg of propranolol as well as 400mg of amisulpride can be considered standard dosages in the context of experimental studies and where these dosages are known to impact cognition and neural processing (e.g., Alexander et al., 2007; De Martino et al., 2008; Kahnt et al., 2015; Kahnt and Tobler, 2017; Ramaekers et al., 1999; Strange et al., 2003; Strange and Dolan, 2004). However, little is known about whether these dosages are comparable in terms of their effect on the brain. To our knowledge, there are no PET studies on single-dose drug administration that could allow us to quantify the extent to which these drug dosages occupy their respective receptors. And even if we had access to such knowledge it is unclear whether (and how) receptor occupancy linearly translates to effects on cognition. This means we cannot say with certainty whether an absence of an amisulpride effect is because dopamine does not play a role in metacognition, or whether it is because the dosage was too little. We have thus revised the manuscript to acknowledge this possibility, including toning down our findings with respect to amisulpride, and mention these issues explicitly in the revised discussion. Please find the revised sections below.

Lastly, it is worth mentioning that we would not expect different pharmacokinetics for cognition and metacognition. Based on the assumption that both processes are derive from brain-based computations, then there is good reason to believe that pharmacokinetics of both metacognition and cognition should be equivalent.

Title: “Noradrenaline blockade specifically enhances metacognitive performance”

Abstract: “Blockade of dopamine D2/3 receptors (400 mg amisulpride) had no effect on either metacognition or perceptual decision making.”

Introduction: “[…] we selected drugs with selective high affinity, the β-adrenoceptor antagonist propranolol in the case of noradrenaline, and the D2/3 receptor antagonist amisulpride in the case of dopamine. In a double-blind, placebo-controlled design we demonstrate that the noradrenergic agent propranolol uniquely improves metacognition in the absence of an effect on perceptual performance, with no effect seen following administration of the dopamine antagonist amisulpride.”

Results: “Noradrenaline blockade modulates metacognition […] A dopamine subject group first received 400mg amisulpride (dopamine D2/3 receptor antagonist) and subsequently placebo, whereas the noradrenaline group first received placebo and then 40mg propranolol (β-adrenoceptor antagonist). […] These results indicate that antagonism of noradrenergic function improves metacognitive insight, in the absence of any effect of amisulpride. […] A highly significant effect of propranolol compared to placebo shows that propranolol increases metacognitive abilities. The difference between propranolol and amisulpride suggests that this performance increase might be specific to an influence on noradrenaline but not dopamine function.”

Discussion: “Here we show that inhibition of central noradrenaline (by means of propranolol) function enhances perceptual metacognitive ability. A dopamine blockade (by means of amisulpride) had no impact on metacognition and neither drug manipulation had an impact on core perceptual performance. […] In our experiment, perceptual metacognition was solely influenced by manipulation of noradrenaline, and not by blocking dopamine D2/D3 receptors. […] An important caveat for comparing the amisulpride and propranolol groups directly is that little is known about the precise pharmacokinetics and how comparable the dosage effects are. We took great care in the design of the study to render the two drug conditions as comparable as possible. First, because of the slightly different absorption rates, we administered amisulpride 30 minutes before propranolol, in keeping with previous drug schedules (Peretti et al., 1997; Ramaekers et al., 1999; Strange et al., 2003; Silver et al., 2004; Strange and Dolan, 2004; Hurlemann et al., 2005; Alexander et al., 2007; Gibbs et al., 2007; De Martino et al., 2008; Kahnt et al., 2015; Kahnt and Tobler, 2017). Second, to render the cognitive effects of the drugs as similar as possible, we selected dosages that were commonly reported in previous studies of neurocognition (i.e. 40 mg propranolol, 400mg amisulpride) (e.g., Ramaekers et al., 1999; Strange et al., 2003; Silver et al., 2004; Strange and Dolan, 2004; Hurlemann et al., 2005; Alexander et al., 2007; Gibbs et al., 2007; De Martino et al., 2008; Kahnt et al., 2015; Kahnt and Tobler, 2017). However, we know little about the magnitude of these drug effects on the brain. A previous study of sulpiride, which has a similar chemical formulation to amisulpride, but slightly different pharmacokinetics, suggests that a single-dose of 400mg leads to a occupancy of ~28% of D2 receptors (Mehta et al., 2008). Unfortunately, there are no PET studies reporting on single-dose amisulpride, and there are no occupancy studies of propranolol, thus rendering it difficult to directly quantify and compare our dosage effects. […] In conclusion, using a double-blind, placebo-controlled drug manipulation we show that noradrenaline has a controlling influence on metacognitive ability.”

Second, pharmacological agents are never precise effectors (e.g. the high affinity issue). Instead they may have side effects that cause issues. What are some of the effects amisulpride and propranolol may have, how strong they are at the doses used, and how may they confound interpretation of the results?

Common side effects of amisulpride are extrapyramidal symptoms and dizziness/nausea, but these are rarely observed in single-dose administration regimes such as ours. In fact, a previous study did not find any significant effects on psychomotor or extrapyramidal functioning after 400mg amisulpride administration (Ramaekers et al., 1999). In this study, we had one participant in the amisulpride group that complained of mild transient symptoms (primarily tremor) in the aftermath of the experiment, likely to be due to the drug. For propranolol, common side effects are lowered blood pressure and pulse, nausea, and diarrhoea. This is why we examined blood pressure and pulse after the experiment. We did not observe any differences in blood pressure (systolic: F(2,57)=.53, p=.594, diastolic: F(2,57)=.825, p=.443), but observed a trend difference in pulse (F(2,57)=2.95, p=.060). However, the pulse rates in all groups were still within the normal range (placebo: 62.9 ± 8.6, propranolol: 56.3 ± 11.1, amisulpride: 62.1 ± 8.3). No subject reported issues related to low blood pressure at the end of the experiment. In addition, the effect on metacognition was much stronger than the effect on pulse, and the metacognitive effect remained even when controlling for pulse (multiple regression placebo vs propranolol: t=4.28, p<.001). Nevertheless, we think it would be interesting to investigate whether peripheral noradrenaline effects also influence metacognitive performance, and discuss possible ways of addressing this in future studies.

Discussion: “Lastly, our previous findings of an embodied reflection of confidence by means of cardiac and pupil responses (Allen et al., 2016) raises the question as to whether the observed noradrenaline effects are purely a consequence of central changes, or whether peripheral effects of this drug influence metacognitive performance independently. A drug that exclusively targets peripheral, but not central, noradrenaline (cf. De Martino et al., 2008) could provide insight into the question of visceral contributions to metacognition as suggested in ideas on embodied cognition (Allen and Friston, 2016).”

Finally, could any of the findings regarding metacognition be task-specific? For instance, could non-perceptual, more slow-paced decision making have different interactions with the drugs? The authors should include a discussion.

This is an interesting issue, which is currently under discussion in the field. Our reading of the literature is that metacognitive performance is relatively stable across different tasks within the domain of perceptual decision making (e.g., Garfinkel et al., 2016; McCurdy et al., 2013; Song et al., 2011), even between different sensory modalities (Gardelle et al., 2016). However, there also seem to be some differences between cognitive domains, e.g., between perceptual decision making and meta-memory (Baird et al., 2013; Fleming et al., 2014). Given that we have only evaluated perceptual but not other domains of metacognition, we cannot say whether the noradrenaline effect would generalise. We have thus clarified throughout the revised manuscript that we study perceptual decision making, and discuss this issue in the Discussion section.

Discussion: “Here we show that inhibition of central noradrenaline (by means of propranolol) function enhances perceptual metacognitive ability […] In this study, we show that noradrenaline specifically influences perceptual metacognition but not perceptual decision making. It is interesting to speculate whether our findings are generalizable to metacognition in non-perceptual domains. Recent studies show that a metacognitive ability is relatively stable across different perceptual decision making tasks (Song et al., 2011; McCurdy et al., 2013), even when probing different sensory modalities (Gardelle et al., 2016; Garfinkel et al., 2016). However, it is unclear whether metacognition within different cognitive domains (e.g., perception vs memory) rely on the same processes, with evidence from neuroimaging suggesting that these functions utilise unique neural networks (Baird et al., 2013; Fleming et al., 2014). Given that noradrenaline modulates activity on a whole-brain level (Hauser et al., 2016), it is possible that noradrenergic metacognition effects can be observed in domains other than perception, an interesting area for future studies.”

Reviewer #2:

[…]

1) Amilsulpride did not significantly influence performance (perceptual or self-rated confidence). It is quite possible that this null effect is simply due to the low occupancy of dopamine receptors with this single dose. To support this dose, the authors' cited paper (Kahnt et al., 2015) in turn cites Mehta (2008) whom in turn use PET (raclopride) to show that a single dose of 400mg of amilsupride leads to an estimated 28% receptor occupancy (much lower than clinical effects) which notably caused no significant changes across a broad armoury of cognitive tests (had they properly corrected for multiple tests!). This is consistent with Ramaekers (1999) – cited by Gibbs 2007 to justify the dose – who found no cognitive or behavioural effects following a single dose unless it was administered for 5 days in a row.

We do not know what the comparable receptor occupancy of central adrenoreceptors is following a single dose of 40mg of propranolol (there are no citations in the text and I believe there is no PET assay). Therefore, the apparent head-to head competition between noradrenaline and dopamine modulation is not unequivocally demonstrated. The difference could simply be due to a marked difference in the central effective pharmacological effects of these single doses. This might be fine if the paper framed itself around the actual pharmacological manipulations given – but the paper clearly stamps itself as a dopamine versus noradrenaline head-to-head. The authors offer insufficient justification for this imputation and list no caveats or limitations.

We thank the reviewer for the positive evaluation of our work and we appreciate the helpful critique related to what inferences we can draw based upon our drug manipulations. First, we 400mg of amisulpride has been used by several prior studies and we decided on this dosage because it is a fairly standard dose for non-clinical, experimental studies (e.g., Gibbs et al., 2007; Kahnt et al., 2015; Kahnt and Tobler, 2017; Ramaekers et al., 1999). Several studies have shown effects of this dose on cognitive and neural responses. However, we agree that there are also several null-results with respect to cognition in the literature, but it is difficult to reconcile whether this is due to a low dosage or because D2 receptors do not play a major role in the particular aspect of cognition under investigation. In addition, our goal was not to relate it to clinical doses as used in the treatment of conditions such as schizophrenia. We had a more fundamental question of whether amisulpride influences metacognition. The study by Mehta et al., 2008 used sulpiride instead of amisulpride. Although these two drugs are closely related, they do have different pharmacokinetic properties, so it is not straightforward to generalise from this study as the issue of the degree of receptor occupancy following a 400mg amisulpride dose. To the best of our knowledge, there is no study that investigates receptor occupancy of 400mg of amisulpride in healthy controls (single-dose), rendering it difficult to guage the extent of dopamine receptor occupancy with this dose. However, given previously mentioned effects on cognition and neural responses, an effect on perceptual decision making or metacognition is entirely possible. Nevertheless, we agree with the reviewer that a direct comparison between the drugs is difficult to interpret because of multiple factors that might confound such a comparison (for example, dosage). We revised the manuscript accordingly and toned down the dopamine aspects (e.g., removed it from the title). In addition, we cite the aforementioned papers and discuss issues of interpretation in a new Discussion section.

2) As I'm sure the authors are acutely aware, propranolol has both central and peripheral β blockade – in fact prior work by some of the authors used propranolol in combination with the selective peripheral blocker, nadolol, to disambiguate the central from the peripheral action. So there are inevitably cardiovascular and likely other autonomic consequences to the β blockade. Given the same group recently published a paper that linked autonomic fluctuations to decision confidence, this seems quite an oversight.

We agree that a dissociation between central and peripheral effects of noradrenaline is both of interest and important. However, to have a comparable control condition for both, amisulpride and propranolol, we decided to use placebo instead of a peripherally active β-blocker. Our reading of the literature on propranolol studies led us to conclude placebo is much more common as a control condition than is nadolol (where a dosage issue again arises). Note also that our previous finding (Allen et al., 2016) showed that confidence is reflected in peripheral measures. However, it is still entirely unclear whether this is a down-stream effect of a brain response (very likely for pupil dilation), or whether it is an independent peripheral effect. Importantly, as there is no prior study the effects of noradrenaline on metacognition, our prime goal was to determine whether there were any effects of noradrenaline before examining differential responses to central and peripheral noradrenaline blockade. In subsequent studies, we indeed plan to use this differentiation to investigate the specific contributions of embodied cognition. We discuss this issue now in the revised version of the manuscript.

Discussion: “[…] our previous findings of an embodied reflection of confidence by means of cardiac and pupil responses (Allen et al., 2016) raises the question as to whether the observed noradrenaline effects are purely a consequence of central changes, or whether peripheral effects of this drug influence metacognitive performance independently. A drug that exclusively targets peripheral, but not central, noradrenaline (cf. De Martino et al., 2008) could provide insight into the question of visceral contributions to metacognition as suggested in ideas on embodied cognition (Allen and Friston, 2016)”

3) There is more of a style issue, but I found the start of the Discussion a much more accessible/more descriptive account of the goals of the paper, than the title or the Abstract – that is, I would have preferred to see a more descriptive approach to the actual task up front – self-rated decision following a perceptual task which was then interpreted as "metacognition". The latter is a conceptual construct that must surely be composed of other processes, and which is only (imperfectly) addressed in the present task. The broad readership will have to read deeply into the paper until they discover that this is a simple rating of confidence following a decision.

We apologise for not being clear enough about the concept of metacognition and the analysis used in this study. We have revised the manuscript entirely to make the beginning of the paper more accessible for the broad readership of eLife and to clarify the meaning of metacognition right at the beginning of the paper. We hope that the reviewer agrees that the revised version is now more accessible.

Abstract: “Impairments in metacognition, the ability to accurately report one’s performance, are common in patients with psychiatric disorders, where a putative neuromodulatory dysregulation provides the rationale for pharmacological interventions.”

Introduction: “Making a decision is often accompanied by a conscious feeling of confidence (Flavell, 1979). Subjective confidence reports typically show a good correspondence to actual task performance, reflecting a metacognitive ability for accurate introspection (Fleming et al., 2010). Impairments in metacognition can compromise decision making and lead to misjudgements of actual performance, as found in several psychiatric dimensions […].”

4) Were the subjects fasting? Were there checks to see if they were able to unblind themselves (particularly for the propranolol).

We did not ask the subjects to fast, as in similar drug studies (e.g., Kahnt et al., 2015, 2017; Strange & Dolan, 2004), but they did not eat anything during the period of the entire study. We asked the subjects after the experiment to guess which drug they received (missing data from two subjects). Based on the reviewer’s comments, we now analysed this data and did not find any difference. We report this in the revised version of the manuscript and detail here the subjects’ guesses.

drug guessed (N)
placebopropranololamisulpride
drugs received (N)placebo864
propranolol965
amisulpride893

Results: “A post-experiment evaluation revealed that subjects were not aware of whether and which drug they received (χ2(4)=1.26, p=.868; missing data from two subjects).”

5) Was there a main effect of drug on PANAS?

No, as reported in Table 1 there was no effect on positive or negative affect. We extended these analyses in the revised manuscript by reporting a repeated-measures ANOVA with within-subjects affect (positive, negative), time (pre, post), and between-subjects factor drug. This extended analysis also showed no effects of drug on mood.

Results: “There were no effects of drug on mood (PANAS; Watson et al., 1988) ratings (main effect of drug: F(2,57)=.16, p=.852; time x drug: F(2,57)=.19, p=.827; time x affect x drug: F(2,57)=2.17, p=.124).”

https://doi.org/10.7554/eLife.24901.009

Article and author information

Author details

  1. Tobias U Hauser

    1. Wellcome Trust Centre for Neuroimaging, University College London, London, United Kingdom
    2. Max Planck University College London Centre for Computational Psychiatry and Ageing Research, London, United Kingdom
    Contribution
    TUH, Conceptualization, Data curation, Formal analysis, Supervision, Visualization, Writing—original draft, Project administration, Writing—review and editing
    Contributed equally with
    Micah Allen
    For correspondence
    t.hauser@ucl.ac.uk
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0002-7997-8137
  2. Micah Allen

    1. Wellcome Trust Centre for Neuroimaging, University College London, London, United Kingdom
    2. Institute of Cognitive Neuroscience, University College London, London, United Kingdom
    Contribution
    MA, Conceptualization, Software, Supervision, Methodology, Writing—original draft, Writing—review and editing
    Contributed equally with
    Tobias U Hauser
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0001-9399-4179
  3. Nina Purg

    Wellcome Trust Centre for Neuroimaging, University College London, London, United Kingdom
    Contribution
    NP, Investigation, Project administration, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
  4. Michael Moutoussis

    1. Wellcome Trust Centre for Neuroimaging, University College London, London, United Kingdom
    2. Max Planck University College London Centre for Computational Psychiatry and Ageing Research, London, United Kingdom
    Contribution
    MM, Conceptualization, Supervision, Investigation, Methodology, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0002-4751-0425
  5. Geraint Rees

    1. Wellcome Trust Centre for Neuroimaging, University College London, London, United Kingdom
    2. Institute of Cognitive Neuroscience, University College London, London, United Kingdom
    Contribution
    GR, Conceptualization, Supervision, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0002-9623-7007
  6. Raymond J Dolan

    1. Wellcome Trust Centre for Neuroimaging, University College London, London, United Kingdom
    2. Max Planck University College London Centre for Computational Psychiatry and Ageing Research, London, United Kingdom
    Contribution
    RJD, Conceptualization, Supervision, Funding acquisition, Writing—original draft, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0001-9356-761X

Funding

Wellcome (095844/Z/11/Z)

  • Tobias U Hauser
  • Michael Moutoussis
  • Raymond J Dolan

Wellcome (100227)

  • Micah Allen
  • Geraint Rees

UCLH Biomedical Research Council

  • Michael Moutoussis

Wellcome (091593/Z/10/Z)

  • Geraint Rees
  • Raymond J Dolan

Wellcome (098362/Z/12/Z)

  • Raymond J Dolan

Max-Planck-Gesellschaft

  • Raymond J Dolan

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

RJD holds a Wellcome Trust Senior Investigator Award (098362/Z/12/Z). A Wellcome Trust Cambridge-UCL Mental Health and Neurosciences Network grant (095844/Z/11/Z) supported RJD and TUH. GR and MA were supported by a Wellcome Trust SRF grant (100227). The UCL-Max Planck Centre is a joint initiative supported by UCL and the Max Planck Society. The Wellcome Trust Centre for Neuroimaging is supported by core funding from the Wellcome Trust (091593/Z/10/Z). We thank Gita Prabhu for the support with the ethics application. We thank Robb Rutledge, Francesco Rigoli, Geert-Jan Will, and Eran Eldar for helpful inputs on the study design. We thank Steve Fleming for helpful comments on an earlier version of the manuscript. The authors declare no conflict of interest.

Ethics

Human subjects: The study was approved by the UCL research ethics committee and all subjects gave written informed consent.

Reviewing Editor

  1. Haozhe Shan, Reviewing Editor, University of Chicago, United States

Publication history

  1. Received: January 5, 2017
  2. Accepted: April 19, 2017
  3. Version of Record published: May 10, 2017 (version 1)

Copyright

© 2017, Hauser et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 5,277
    Page views
  • 349
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.

Comments

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)

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

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

Further reading

    1. Neuroscience
    Simon M Danner et al.
    Research Article Updated