Perceptual decisions made in the face of uncertain sensory evidence are often biased by previous stimuli, choices, and choice outcomes (Gold et al., 2008; Busse et al., 2011; de Lange et al., 2013; Akaishi et al., 2014; Fischer and Whitney, 2014; Fründ et al., 2014; Abrahamyan et al., 2016; Pape and Siegel, 2016; St John-Saaltink et al., 2016; Fritsche et al., 2017; Hwang et al., 2017; Urai et al., 2017; Braun et al., 2018). In most standard laboratory tasks, the environmental state (i.e., stimulus category) is uncorrelated across trials. In that context, such history biases tend to impair performance (Abrahamyan et al., 2016). However, when the environmental state exhibits some stability across trials, as is common for natural environments (Yu and Cohen, 2009; Glaze et al., 2015), trial history biases substantially improve performance (Braun et al., 2018). Previous behavioral work shows that humans and other animals flexibly adapt their trial history bias to the correlation structure of the environment (Abrahamyan et al., 2016; Kim et al., 2017; Braun et al., 2018; Hermoso-Mendizabal et al., 2020).

How do adaptive history biases influence the formation of subsequent decisions? Prominent models conceptualize the decision formation as the accumulation of noisy sensory evidence into a decision variable (DV) that grows with time until a bound for one of the choice alternatives is crossed and the corresponding action is initiated (Bogacz et al., 2006; Gold and Shadlen, 2007; Ratcliff and McKoon, 2008; Brunton et al., 2013; Ossmy et al., 2013). In such accumulator models, history biases may shift the starting point of the DV before evidence onset and/or bias the evidence accumulation per se. Behavioral modeling indicates that individual differences in the idiosyncratic history biases occurring in random environments are better explained by biases of evidence accumulation than by starting point biases (Urai et al., 2019). Such effects have neither been assessed for adaptive biases in structured (stable or systematically alternating) environments nor have they been unraveled at the neural level. We hypothesized that adaptive history biases translate into a biased build-up rate (accumulation bias) of neural signatures of the DV, more so than an offset before decision formation (starting point), in a fashion that depends on the correlation structure of the environment.

Neural signals exhibiting functional properties of the DV have been observed in parietal and frontal cortical areas involved in action planning in both primates (Shadlen and Kiani, 2013; Peixoto et al., 2021) and rodents (Hanks et al., 2015; Brody and Hanks, 2016). Specifically, when choices are reported with hand movements, hallmark signatures of the DV are evident in motor preparatory activity in primate (human and monkey) premotor and primary motor (M1) cortex. In human motor cortex, this selective motor preparatory activity is expressed in a suppression of ongoing beta-band oscillations contralateral to the upcoming hand movement, accompanied by an enhancement of gamma-band power (Crone et al., 1998a; Crone et al., 1998b; Donner et al., 2009) and likely spiking activity. While the origin of this beta-power suppression remains under study (Sherman et al., 2016; Little et al., 2019) we here use it as a functional marker of the DV encoded in local patters of spiking activity: Like this spiking activity (Shadlen and Kiani, 2013; Peixoto et al., 2021), the beta-band suppression (i) encodes the specific choice that will later be reported, (ii) gradually builds up during decision formation, with a rate that scales with evidence strength, and (iii), in reaction time tasks, converges on a common level just before action execution (Donner et al., 2009; O’Connell et al., 2012; Wyart et al., 2012; de Lange et al., 2013; Fischer et al., 2018; Wilming et al., 2020; Murphy et al., 2021).

We combined a canonical decision-making task, discrimination of the net motion direction of dynamic random dot patterns (Gold and Shadlen, 2007), with a systematic manipulation of the environmental correlation structure. Our manipulation was motivated from an ecological perspective (Mobbs et al., 2018), specifically, the insights that (i) natural environments are commonly structured and (ii) history biases that change flexibly with this environmental structure are a hallmark of adaptive behavior. Because previous work on such correlated environments was purely behavioral (Abrahamyan et al., 2016; Kim et al., 2017; Braun et al., 2018; Hermoso-Mendizabal et al., 2020), the neural signatures of adaptive history biases have remained unknown. Although several previous studies have identified neural correlates of history biases in standard perceptual choice tasks (i.e., using unstructured environments) (Talluri et al., 2021), all but one study performed in monkeys (Mochol et al., 2021) focused on static representations of the bias in ongoing activity preceding the new decision. Therefore, it has remained unknown whether such a dynamic bias during evidence accumulation exists in the human brain.

Single-trial behavioral modeling uncovered the resulting history-dependent biases as well as their flexible adjustment to the environmental correlation structure. Relating the model-inferred time-varying history bias to magnetoencephalography (MEG) measurements of the pre-trial baseline state and subsequent build-up rate of action-selective motor cortical population activity identified a neural signature of this adaptive bias in the latter, not the former. In sum, we show that the sign and rate of the build-up of a selective neural marker of DV during evidence accumulation track a dynamic history bias that is adapted to the environmental structure.

Human participants (N=38) performed a random dot motion (up vs. down) discrimination task with varying levels of motion strength spanning psychophysical threshold (Figure 1A, Materials and methods). We alternated the task, in pseudo-random order, between three different sensory environments with distinct repetition probabilities of stimulus categories (i.e., motion directions) across trials, referred to as neutral, repetitive, and alternating, respectively (Figure 1B). These three environments were characterized by approximately equal fractions of upward and downward motion stimuli, and they were presented in blocks of 99 trials each, separated by pauses (Materials and methods). Participants were not informed about the existence of these different environments and received outcome feedback after each choice.

Figure 1 with 3 supplements see all

Download asset Open asset

Large-scale cortical dynamics of task processing

The behavioral results reported above indicate that participants adjusted their history bias to the environmental statistics, which, in turn, boosted their performance. How did these (partly) adaptive history biases affect the formation of subsequent decisions, more specifically: the dynamics of the underlying DV in the brain? Our concurrent collection of whole-brain MEG data during this task enabled us to address this question. We combined source reconstruction with established anatomical atlases and spectral analysis to characterize the cortical dynamics involved in our task across several pre-defined cortical regions known to be involved in visual processing and action planning (Wilming et al., 2020; Murphy et al., 2021).

We first identified established task-related modulations of MEG power during decision formation. Control analyses indicated minor leakage between the source estimates for neighboring cortical regions, but negligible leakage for more distant regions (Figure 2—figure supplement 1; Materials and methods). Importantly, our analysis revealed distinct functional profiles for several regions, in line with previous work (Siegel et al., 2011; Wilming et al., 2020; Murphy et al., 2021; Urai and Donner, 2022): modulations in visual cortical regions that scaled with motion coherence and encoding of the evolving action plan in frontal (motor and premotor) and parietal cortical regions (Figure 2).

Figure 2 with 2 supplements see all

Download asset Open asset

In line with previous work (Siegel et al., 2007), gamma-band power (~60–100 Hz) in visual cortex was enhanced while low-frequency power (<30 Hz) was suppressed relative to baseline during motion viewing (Figure 2A); both components of the visual responses scaled with motion coherence, predominantly in dorsal visual cortical areas V3A/B, IPS0-3, and the MT+ complex (Figure 2B). Concomitantly with these responses to visual motion, activity lateralization predicting the subsequent choice (left vs. right button) built up in downstream (anatomically more anterior) parietal and motor cortical areas (Figure 2C). Again in line with previous work (Donner et al., 2009; de Lange et al., 2013; Pape and Siegel, 2016; Wilming et al., 2020; Murphy et al., 2021), this action-selective activity build-up was a suppression of beta-band (12–36 Hz) power contra- vs. ipsilateral to the upcoming movement, and robustly expressed in the M1 hand area (Figure 2C and Figure 4A). This signal, referred to as ‘motor beta lateralization’ in the following, has been shown to exhibit hallmark signatures of the DV, specifically: (i) selectivity for choice and (ii) ramping slope that depends on evidence strength (Siegel et al., 2011; Murphy et al., 2021; O’Connell and Kelly, 2021).

This signal reached statistical significance earlier for correct than error trials and during the stimulus interval it ramped to a larger amplitude (i.e., more negative) for correct trials (Figure 2—figure supplement 2, left). But the signal was indistinguishable in amplitude between correct and error trials around the time of the motor response (Figure 2—figure supplement 2, right). We also confirmed the dependence of the ramping of the motor beta lateralization on evidence strength using a single-trial regression also including the history bias that we report in the section Adaptive history bias shapes the build-up of action-selective motor cortical activity below.

In sum, we replicated well-established signatures of visual motion processing and action-selective motor preparation in our current MEG data – for the latter signal in particular, some hallmark signatures of the DV: selectivity for choice, dependence of slope on evidence strength, and dependence of behavioral accuracy. In what follows, we systematically quantify the impact of the adaptive history bias on two fundamental aspects of this neural DV marker: its baseline level at the start of the decision process and its build-up during decision formation.

No consistent modulation of baseline state of action-selective activity by environmental context and trial history

Previous human MEG work indicates that the motor cortical baseline beta-power state is flipped relative to its state just before the previous choice, a phenomenon referred to as ‘beta rebound’ (Pfurtscheller et al., 1996; Pape and Siegel, 2016; Urai and Donner, 2022) that was also evident in our data (Figure 3A, collapsed across all three environments). Recent MEG studies of human perceptual decision-making have linked this phenomenon to either overt choice alternation (Pape and Siegel, 2016) or alternating starting points inferred from drift diffusion model fits (Urai and Donner, 2022). We, therefore, wondered if and how the baseline level of motor beta lateralization depended on the different sensory environments or on the history bias in specific trials.

Figure 3

Download asset Open asset

If this baseline lateralization state ‘inherited from’ the previous choice was involved in mediating the effect of the adaptive bias on choice, it might be expected to be reduced in repetitive vs. alternating environments, thus reducing subjects’ tendency to alternate in the former. Instead, the beta rebound effect (i.e., increased power contralateral vs. ipsilateral to previous choice) was about equally strong in all three environments (Figure 3B). We found no evidence for its modulation by sensory environment (repetitive vs. neutral: p=0.1511; repetitive vs. alternating: p=0.5667; neutral vs. alternating: p=0.0789; all two-sided permutation tests).

We then related the baseline M1 beta lateralization to the adaptive history bias. To this end, we adapted a single-trial regression procedure from recent monkey physiology work (Mochol et al., 2021) to relate the time-varying history bias to neural data in order to test if this bias modulated the baseline motor beta lateralization on a trial-by-trial basis. We used each individual’s time course of single-trial history bias estimated through the behavioral model (positive values for bias toward upward, Figure 1E) as predictors for the single-trial motor beta lateralization, whereby lateralization was assessed relative to the hand coding for up-choices in a given block (Materials and methods). This procedure took the impact of both previous stimuli and previous choices into account, estimated with an individually optimized number of lags. Thus, the single-trial bias estimates were largely independent of assumptions about the sources of the single-trial bias (stimuli, choices, lags). However, because the model was fit and applied separately to data from different environments, the resulting time course of single-trial bias estimates captured the context-dependent, adaptive bias components described in the preceding section. An involvement of the motor baseline state in the implementation of history bias predicts a stronger baseline beta-suppression contralateral to the hand favored by the bias, with a magnitude that scales with the strength of the bias. In other words, this scenario predicts significant negative beta coefficients, regardless of the environment.

We found no such effect when the analysis was run across all three environments (Figure 3C, ‘all’), again inconsistent with the notion of a generally bias-encoding neural signal. We did find an effect of the single-trial bias on the baseline beta lateralization state in the alternating environment when analyzed selectively (Figure 3C; p=0.0003, two-tailed permutation test). Such an effect was, however, not present for either of the other two environments (Figure 3C; all: p=0.4465; repetitive: p=0.1343; neutral: p=0.6571; two-tailed permutation tests). Overall, the results of our analyses of the baseline beta lateralization suggest the beta rebound from the previous trial may help promote choice alternation when performing in an alternating context, but does not generally encode adaptive stimulus history biases.

Adaptive history bias shapes the build-up of action-selective motor cortical activity

The analyses from the previous section assessed the dependence of the starting point (i.e., baseline level) of a neural DV-proxy on environmental context and adaptive history bias. Behavioral modeling has shown that idiosyncratic history biases in a variety of tasks in random environments are accounted for by history-dependent biases in the build-up (i.e., drift) of the DV, rather than in its starting point (Urai et al., 2019). We, therefore, next asked whether the adaptive history biases identified here might shape the build-up rate of our neural proxy of the DV during decision formation.

To test this idea, we used two complementary approaches both of which again exploited our model-inferred single-trial bias estimates. In one of those approaches, we grouped the single-trial bias estimates into three equally spaced bins, with two bins containing strong biases of opposite direction (up vs. down) and the middle bin containing trials with little bias (Figure 1E). We used the ‘up’ and ‘down’ bins to visualize the impact of the history bias on the ramping of the neural DV, by computing the time course of beta lateralization contra- vs. ipsilateral to the button-press for the direction of the bin-wise bias (Materials and methods). The behavioral choice was, by definition, correlated with both, the single-trial bias and the action-selective motor beta lateralization (Figure 2C and Figure 4A). This could yield a correlation between bias and motor cortical lateralization even in the absence of any direct effect of bias on motor beta lateralization. To isolate a genuine effect of the bias on our neural DV, we subsampled the data from the up and down bias bins to yield an equal number of upward and downward choices within each bin (Materials and methods). For each bin, we then computed the time course of beta lateralization relative to the button-press for the bias and collapsed the resulting time courses across bins. This procedure isolated the impact of the model-inferred history bias on the neural DV, independent of the choice. The resulting time course ramped into the direction of the single-trial bias, reaching statistical significance at about 700 ms after motion onset, before the end of the decision interval (Figure 4B). We used linear regression to estimate the slope of the ramp for an interval that exhibited clear linear ramping in the average motor beta lateralization across all trials (gray-shaded in Figure 4A, Materials and methods). As expected, the slope was smaller than zero (p=0.048; one-tailed permutation test against zero; Figure 4C, left), indicating that the time-varying history bias contributed to the build-up of action-selective motor cortical activity during decision formation. We obtained qualitatively identical results as in Figure 4B and C when first removing (using linear regression) the beta-rebound from the previous trial (Figure 3A) from the time course of the beta-band lateralization (data not shown).

Figure 4 with 1 supplement see all

Download asset Open asset

Second, we again fit a single-trial regression model, now to simultaneously quantify the impact of the history bias and current evidence on the dynamics of the neural DV in a time-variant fashion. We ran two separate regression models, one on the amplitude of motor beta lateralization for a range of time windows, the other on the slope of motor beta lateralization, assessed locally in time for the same time windows; in both cases, lateralization was again assessed relative to the hand coding for up-choices in a block (Materials and methods). An impact of the adaptive history bias on the ramping of motor cortical activity would predict a specific effect of the history bias, over and above the effect of the sensory evidence, on both read-out measures, in particular on the ramping slopes. Specifically, it predicts negative beta weights, reflecting steeper downward slope (i.e., stronger suppression) for stronger biases.

We found a clear and expected effect of current sensory evidence on motor beta lateralization, with a steeper downward slope for stronger evidence (Figure 4E). This dependence of the ramping of the motor beta lateralization on current motion strength is in line with previous work (de Lange et al., 2013) and confirms that the motor beta lateralization reflects hallmarks of a neural DV. Critically, and in line with our hypothesis, the same was true for the effect of the history bias: a stronger bias produced a stronger and steeper motor beta lateralization toward the direction of the bias (Figure 4D). The bias effect on the lateralization amplitude reached significance during the decision interval (from about 320 to about 720 ms after motion onset; Figure 4D, black line), and the corresponding impact on the ramping slope was significant even earlier, during the first half of the decision interval (starting at about 150 ms after the motion onset; Figure 4D, gray line). Combined, these two effects indicate that a strong bias on a given trial constituted an early force on the M1 ramping dynamics, pushing the signal into the direction of the bias even before the current evidence exerted its effect (compare with gray lines in Figure 4E); the M1 lateralization amplitude later during the decision interval reflected the bias more strongly on trials, for which the bias was strong than those, for which the bias was weak.

Our analysis of the ramping slopes in Figure 4C (left) estimated the slope for a longer (and later) time interval than the one, for which the single-trial regression in Figure 4D yielded significant slope effects. We, thus, repeated the above analysis also for the earlier time window derived from the single-trial regression results Figure 4C (right). Also, for this window did we find a robust effect of the bias estimate on the ramping slope (p=0.0158; one-tailed permutation test against zero). Taken together, model-independent and model-based analyses provided convergent evidence for the dependence of action-selective cortical ramping activity during decision formation on the time-varying, context-dependent history biases.

It has long been known that the history of preceding choices and stimuli biases perceptual judgments of the current stimulus (Fernberger, 1920). Recent behavioral modeling showed that at least part of such history biases reflect time-varying expectations that are flexibly adapted to the environmental structure (Abrahamyan et al., 2016; Braun et al., 2018; Hermoso-Mendizabal et al., 2020). Such dynamically varying expectations, largely ignored in standard neurophysiological studies of perceptual decision-making, may be a key driver of sensory-guided behavior in ecological settings (Mobbs et al., 2018). How adaptive expectations shape the neural dynamics underlying decision-making has remained unknown. Here, we addressed this issue by combining a standard task from the neurophysiology of decision-making (Gold and Shadlen, 2007; Siegel et al., 2011) with systematic manipulations of the environmental stability as well as single-trial, model-based MEG assessment of cortical decision dynamics. This revealed that the history-dependent, dynamic expectations boosted participants’ behavioral performance and selectively altered the build-up sign and rate, not (consistently) the pre-trial baseline level, of an established neurophysiological proxy of the DV: action-selective preparatory population activity in their motor cortex.

While participants’ history biases in a random environment (i.e., uncorrelated stimulus sequences) were largely idiosyncratic, as widely observed (Akaishi et al., 2014; Urai et al., 2019; Urai and Donner, 2022), we found that one component of these biases lawfully shifted between stable (frequent category repetitions) and systematically alternating environments and improved participants’ performance. It is instructive to compare this adjustment of history bias with the one observed in a previous study using a similar manipulation of environmental statistics (Braun et al., 2018). In that previous study, participants did not receive outcome feedback and thus remained uncertain about the category of the previous stimulus. Correspondingly, the history bias adjustment was evident in the impact of their previous choices (rather than previous stimulus categories), and most strongly of those made with high confidence (i.e., correct and fast). By contrast, in the current study, participants could deterministically infer the true category of the previous stimulus from the feedback. Correspondingly, we found that their history bias adjustment to the different environments was now governed by the previous stimulus category. Together, the findings from both studies support the notion that human subjects can use different types of internal signals to build up history-dependent expectations in an adaptive fashion.

The observation of an effective behavioral adjustment to differentially structured environments in participants’ steady-state behavior raises the question of how (and how quickly) they learned the different environmental structures. Our behavioral modeling approach required many trials, which precludes the assessment of the temporal evolution of the bias (i.e., weight) adjustment during the blocks of a given sensory environment. This issue should be addressed in future work, using models capable of learning environmental parameters such as transition probabilities (Yu and Cohen, 2009; Meyniel et al., 2016; Glaze et al., 2018; Hermoso-Mendizabal et al., 2020).

Previous work has characterized neural signals underlying idiosyncratic history biases in contexts where these biases would be maladaptive. Such signals were observed in several brain areas and in different formats. In a continuous spatial working memory task, activity-silent codes in prefrontal cortex during mnemonic periods seem to promote memory reactivations, which mediate serial memory biases (Barbosa et al., 2020). Studies of perceptual forced choice tasks have found signatures of persistent population activity reflecting the previous choice in posterior parietal cortex (Morcos and Harvey, 2016; Hwang et al., 2017; Scott et al., 2017; Urai and Donner, 2022), prefrontal cortex (Mochol et al., 2021), and motor cortex (Pape and Siegel, 2016; Urai and Donner, 2022). Specifically, human MEG work showed that history-dependent modulations of parietal cortical activity in the gamma-band spanned the intervals between trials and mediated idiosyncratic choice repetition biases (Urai and Donner, 2022). Such an effect was not observed for the motor beta-rebound that was similarly sustained into the next decision interval (Urai and Donner, 2022). Importantly, none of these studies quantified the build-up rate of action-selective motor cortical activity on the subsequent trial.

Idiosyncratic history biases are reflected in a persistent baseline state of action-selective neural population activity in monkey prefrontal cortex, during decision formation accompanied by a subtler modulation of the build-up rate (Mochol et al., 2021). Another human MEG study derived an action-independent proxy of the neural DV from sensor-level MEG data that required two successive judgments within a trial (Rollwage et al., 2020). The initial decision biased the subsequent build-up of that DV-proxy in a manner that depended on the consistency of new evidence with the initial decision and the confidence in that initial decision. Critically, no previous study has investigated the flexible and performance-increasing history biases that we have manipulated and studied here.

Our current results resemble the results from block-wise manipulations of the probability of a specific stimulus category (i.e., not of transitions across stimulus categories): this also biases the build-up of saccade-selective activity in monkey posterior parietal cortex (Hanks et al., 2011), just like what we found here for hand movement-selective motor cortical activity in humans, albeit with strong, but lawful, trial-by-trial variations in our current setting (Figure 1E). It is tempting to interpret both as downstream expressions of perceptual expectations in cortical circuitry involved in action planning. Indeed, modulating the build-up of an evolving DV by prior expectations can be useful in accumulation-to-bound models when reliability of the evidence varies from decision to decision (Hanks et al., 2011; Moran, 2015). Whether or not the neural signatures of idiosyncratic history biases studied in previous work have similar cognitive content and underlying mechanisms remains an open question.

Our results indicate that dynamic and adaptive expectations bias the dynamics of neural signatures of action planning during decision formation. How are these expectations implemented in upstream neural populations, so as to yield the selective changes in M1 ramping dynamics observed here? One possibility is that history biases the state of sensory cortex (Nienborg and Cumming, 2009; St John-Saaltink et al., 2016), for example via feedback from cortical areas involved in decision formation (Wimmer et al., 2015). Another possibility is that the expectations shape the read-out of sensory evidence by the evidence accumulator, with preferential accumulation of evidence that matches the expectation, in line with active inference (Friston, 2010). Yet another possibility is that the evidence accumulator receives non-sensory input from brain regions encoding history information in a sustained fashion (Talluri et al., 2021; Urai and Donner, 2022). In all these different schemes, dynamic expectations would need to be constructed in a highly flexible, context-dependent fashion in order to give rise to the adaptive biasing of action-selective activity observed here.

Raw MEG data are available at https://www.fdr.uni-hamburg.de/record/13475. Source reconstructed MEG data are available at https://www.fdr.uni-hamburg.de/record/13197. Behavioral data is available at https://www.fdr.uni-hamburg.de/record/13517. All data is available under Creative Commons Attribution 4.0 International License. The ethics protocol disallows sharing MRI data via a public repository. Data may be shared however within the context of a collaboration. No proposal is needed. In order to obtain the data, please email Anke Braun (anke.braun86@gmail.com) and Tobias H. Donner (t.donner@uke.de). The code and data immediately underlying all main and figure supplements are publicly available on https://github.com/DonnerLab/2023_BraunA_Adaptive_biasing_of_action-selective_cortical_build-up_activity_by_stimulus_history (copy archived at DonnerLab, 2023a) under GNU General Public LicenseVersion 2. The logistic regression model with history bias was fitted using a toolbox from Fründ et al., 2014, which is publicly available under https://bitbucket.org/mackelab/serial_decision/src/master/. Preprocessing of MEG data was done using a Fieldtrip pipeline from Urai and Donner, 2022, which is publicly available on https://github.com/DonnerLab/2022_Urai_choicehistory_MEG (copy archived at DonnerLab, 2022). Source reconstruction of MEGdata was done using pymeg (Wilming et al., 2020), which is publicly available under https://github.com/DonnerLab/pymeg (copy archived at DonnerLab, 2023b).

1. 1. Braun A
   2. Donner THD

   (2023) Forschungsdatenrepositorium Uni Hamburg

   Source Reconstructed MEG Data for Adaptive biasing of action-selective cortical build-up activity by stimulus history.

   https://doi.org/10.25592/uhhfdm.13196
2. 1. Braun A
   2. Donner THD

   (2023) Forschungsdatenrepositorium Uni Hamburg

   Raw MEG Data for Adaptive biasing of action-selective cortical build-up activity by stimulus history.

   https://doi.org/10.25592/uhhfdm.13474
3. 1. Braun A
   2. Donner THD

   (2023) Forschungsdatenrepositorium Uni Hamburg

   Behavioral data for Adaptive biasing of action-selective cortical build-up activity by stimulus history.

   https://doi.org/10.25592/uhhfdm.13516

1. 1. Abrahamyan A
   2. Silva LL
   3. Dakin SC
   4. Carandini M
   5. Gardner JL

   (2016) Adaptable history biases in human perceptual decisions

   PNAS 113:E3548–E3557.

   https://doi.org/10.1073/pnas.1518786113

   • PubMed
   • Google Scholar
2. 1. Akaishi R
   2. Umeda K
   3. Nagase A
   4. Sakai K

   (2014) Autonomous mechanism of internal choice estimate underlies decision inertia

   Neuron 81:195–206.

   https://doi.org/10.1016/j.neuron.2013.10.018

   • PubMed
   • Google Scholar
3. 1. Barbosa J
   2. Stein H
   3. Martinez RL
   4. Galan-Gadea A
   5. Li S
   6. Dalmau J
   7. Adam KCS
   8. Valls-Solé J
   9. Constantinidis C
   10. Compte A

   (2020) Interplay between persistent activity and activity-silent dynamics in the prefrontal cortex underlies serial biases in working memory

   Nature Neuroscience 23:1016–1024.

   https://doi.org/10.1038/s41593-020-0644-4

   • PubMed
   • Google Scholar
4. 1. Bogacz R
   2. Brown E
   3. Moehlis J
   4. Holmes P
   5. Cohen JD

   (2006) The physics of optimal decision making: a formal analysis of models of performance in two-alternative forced-choice tasks

   Psychological Review 113:700–765.

   https://doi.org/10.1037/0033-295X.113.4.700

   • PubMed
   • Google Scholar
5. 1. Braun A
   2. Urai AE
   3. Donner TH

   (2018) Adaptive history biases result from confidence-weighted accumulation of past choices

   The Journal of Neuroscience 38:2418–2429.

   https://doi.org/10.1523/JNEUROSCI.2189-17.2017

   • PubMed
   • Google Scholar
6. 1. Brody CD
   2. Hanks TD

   (2016) Neural underpinnings of the evidence accumulator

   Current Opinion in Neurobiology 37:149–157.

   https://doi.org/10.1016/j.conb.2016.01.003

   • PubMed
   • Google Scholar
7. 1. Brunton BW
   2. Botvinick MM
   3. Brody CD

   (2013) Rats and humans can optimally accumulate evidence for decision-making

   Science 340:95–98.

   https://doi.org/10.1126/science.1233912

   • PubMed
   • Google Scholar
8. 1. Busse L
   2. Ayaz A
   3. Dhruv NT
   4. Katzner S
   5. Saleem AB
   6. Schölvinck ML
   7. Zaharia AD
   8. Carandini M

   (2011) The detection of visual contrast in the behaving mouse

   The Journal of Neuroscience 31:11351–11361.

   https://doi.org/10.1523/JNEUROSCI.6689-10.2011

   • PubMed
   • Google Scholar
9. 1. Crone NE
   2. Miglioretti DL
   3. Gordon B
   4. Lesser RP

   (1998a) Functional mapping of human sensorimotor cortex with electrocorticographic spectral analysis. II. Event-related synchronization in the gamma band

   Braina 121:2301–2315.

   https://doi.org/10.1093/brain/121.12.2301

   • PubMed
   • Google Scholar
10. 1. Crone NE
    2. Miglioretti DL
    3. Gordon B
    4. Sieracki JM
    5. Wilson MT
    6. Uematsu S
    7. Lesser RP

    (1998b) Functional mapping of human sensorimotor cortex with electrocorticographic spectral analysis. I. Alpha and beta event-related desynchronization

    Brain 121:2271–2299.

    https://doi.org/10.1093/brain/121.12.2271

    • PubMed
    • Google Scholar
11. 1. Dale AM

    (1999) Optimal experimental design for event-related fMRI

    Human Brain Mapping 8:109–114.

    https://doi.org/10.1002/(SICI)1097-0193(1999)8:2/3<109::AID-HBM7>3.0.CO;2-W

    • PubMed
    • Google Scholar
12. 1. de Gee JW
    2. Colizoli O
    3. Kloosterman NA
    4. Knapen T
    5. Nieuwenhuis S
    6. Donner TH

    (2017) Dynamic modulation of decision biases by brainstem arousal systems

    eLife 6:e23232.

    https://doi.org/10.7554/eLife.23232

    • PubMed
    • Google Scholar
13. 1. de Lange FP
    2. Rahnev DA
    3. Donner TH
    4. Lau H

    (2013) Prestimulus oscillatory activity over motor cortex reflects perceptual expectations

    The Journal of Neuroscience 33:1400–1410.

    https://doi.org/10.1523/JNEUROSCI.1094-12.2013

    • PubMed
    • Google Scholar
14. 1. Donner TH
    2. Siegel M
    3. Fries P
    4. Engel AK

    (2009) Buildup of choice-predictive activity in human motor cortex during perceptual decision making

    Current Biology 19:1581–1585.

    https://doi.org/10.1016/j.cub.2009.07.066

    • PubMed
    • Google Scholar
15. Software

    1. DonnerLab

    (2022) 2022_Urai_Choicehistory_Meg, version swh:1:rev:ec05db4c27b67df9f318e9706f23ac28f904660a

    Software Heritage.

    https://archive.softwareheritage.org/swh:1:dir:98ca0aa690541884ce354f0f2c578091ec46ed52;origin=https://github.com/DonnerLab/2022_Urai_choicehistory_MEG;visit=swh:1:snp:7d1c0e3cc11ee731c2b32ddb4c85c79292ea7b83;anchor=swh:1:rev:ec05db4c27b67df9f318e9706f23ac28f904660a
16. Software

    1. DonnerLab

    (2023a) 2023_BraunA_Adaptive_biasing_of_action-selective_cortical_build-up_activity_by_stimulus_history, version swh:1:rev:f7cdb7fe875ec220d1458e2645ff181cca4fa9c8

    Software Heritage.

    https://archive.softwareheritage.org/swh:1:dir:405d31b1a95e3043667b35cf05c4fe2f85a4fdbf;origin=https://github.com/DonnerLab/2023_BraunA_Adaptive_biasing_of_action-selective_cortical_build-up_activity_by_stimulus_history;visit=swh:1:snp:4822372adb610e572c84a4e473b03c9073d761a4;anchor=swh:1:rev:f7cdb7fe875ec220d1458e2645ff181cca4fa9c8
17. Software

    1. DonnerLab

    (2023b) Pymeg, version swh:1:rev:a76d1050d9910803857a00feacf13a10a07c191e

    Software Heritage.

    https://archive.softwareheritage.org/swh:1:dir:593be9a0a108aa79adae7334f22f9595403c3117;origin=https://github.com/DonnerLab/pymeg;visit=swh:1:snp:04b82e33a6a4cbf97d3ccb3ab9eb6e4897a8f415;anchor=swh:1:rev:a76d1050d9910803857a00feacf13a10a07c191e
18. Book

    1. Efron B
    2. Tibshirani R

    (1998)

    An Introduction to the Bootstrap

    Boca Raton: Chapman & Hall/CRC.

    • Google Scholar
19. 1. Fernberger SW

    (1920) Interdependence of judgments within the series for the method of constant stimuli

    Journal of Experimental Psychology 3:126–150.

    https://doi.org/10.1037/h0065212

    • Google Scholar
20. 1. Fischer J
    2. Whitney D

    (2014) Serial dependence in visual perception

    Nature Neuroscience 17:738–743.

    https://doi.org/10.1038/nn.3689

    • PubMed
    • Google Scholar
21. 1. Fischer AG
    2. Nigbur R
    3. Klein TA
    4. Danielmeier C
    5. Ullsperger M

    (2018) Cortical beta power reflects decision dynamics and uncovers multiple facets of post-error adaptation

    Nature Communications 9:5038.

    https://doi.org/10.1038/s41467-018-07456-8

    • PubMed
    • Google Scholar
22. 1. Fischl B
    2. Sereno MI
    3. Dale AM

    (1999) Cortical surface-based analysis

    NeuroImage 9:195–207.

    https://doi.org/10.1006/nimg.1998.0396

    • PubMed
    • Google Scholar
23. 1. Friston K

    (2010) The free-energy principle: a unified brain theory?

    Nature Reviews. Neuroscience 11:127–138.

    https://doi.org/10.1038/nrn2787

    • PubMed
    • Google Scholar
24. 1. Fritsche M
    2. Mostert P
    3. de Lange FP

    (2017) Opposite effects of recent history on perception and decision

    Current Biology 27:590–595.

    https://doi.org/10.1016/j.cub.2017.01.006

    • PubMed
    • Google Scholar
25. 1. Fründ I
    2. Wichmann FA
    3. Macke JH

    (2014) Quantifying the effect of intertrial dependence on perceptual decisions

    Journal of Vision 14:9.

    https://doi.org/10.1167/14.7.9

    • PubMed
    • Google Scholar
26. 1. Glaze CM
    2. Kable JW
    3. Gold JI

    (2015) Normative evidence accumulation in unpredictable environments

    eLife 4:e08825.

    https://doi.org/10.7554/eLife.08825

    • PubMed
    • Google Scholar
27. 1. Glaze CM
    2. Filipowicz ALS
    3. Kable JW
    4. Balasubramanian V
    5. Gold JI

    (2018) A bias–variance trade-off governs individual differences in on-line learning in an unpredictable environment

    Nature Human Behaviour 2:213–224.

    https://doi.org/10.1038/s41562-018-0297-4

    • Google Scholar
28. 1. Gold JI
    2. Shadlen MN

    (2007) The neural basis of decision making

    Annual Review of Neuroscience 30:535–574.

    https://doi.org/10.1146/annurev.neuro.29.051605.113038

    • PubMed
    • Google Scholar
29. 1. Gold JI
    2. Law CT
    3. Connolly P
    4. Bennur S

    (2008) The relative influences of priors and sensory evidence on an oculomotor decision variable during perceptual learning

    Journal of Neurophysiology 100:2653–2668.

    https://doi.org/10.1152/jn.90629.2008

    • PubMed
    • Google Scholar
30. 1. Gramfort A
    2. Luessi M
    3. Larson E
    4. Engemann DA
    5. Strohmeier D
    6. Brodbeck C
    7. Parkkonen L
    8. Hämäläinen MS

    (2014) MNE software for processing MEG and EEG data

    NeuroImage 86:446–460.

    https://doi.org/10.1016/j.neuroimage.2013.10.027

    • PubMed
    • Google Scholar
31. 1. Hanks TD
    2. Mazurek ME
    3. Kiani R
    4. Hopp E
    5. Shadlen MN

    (2011) Elapsed decision time affects the weighting of prior probability in a perceptual decision task

    The Journal of Neuroscience 31:6339–6352.

    https://doi.org/10.1523/JNEUROSCI.5613-10.2011

    • PubMed
    • Google Scholar
32. 1. Hanks TD
    2. Kopec CD
    3. Brunton BW
    4. Duan CA
    5. Erlich JC
    6. Brody CD

    (2015) Distinct relationships of parietal and prefrontal cortices to evidence accumulation

    Nature 520:220–223.

    https://doi.org/10.1038/nature14066

    • PubMed
    • Google Scholar
33. 1. Hermoso-Mendizabal A
    2. Hyafil A
    3. Rueda-Orozco PE
    4. Jaramillo S
    5. Robbe D
    6. de la Rocha J

    (2020) Author Correction: Response outcomes gate the impact of expectations on perceptual decisions

    Nature Communications 11:3470.

    https://doi.org/10.1038/s41467-020-17071-1

    • PubMed
    • Google Scholar
34. 1. Hwang EJ
    2. Dahlen JE
    3. Mukundan M
    4. Komiyama T

    (2017) History-based action selection bias in posterior parietal cortex

    Nature Communications 8:1242.

    https://doi.org/10.1038/s41467-017-01356-z

    • PubMed
    • Google Scholar
35. 1. Kim TD
    2. Kabir M
    3. Gold JI

    (2017) Coupled decision processes update and maintain saccadic priors in a dynamic environment

    The Journal of Neuroscience 37:3632–3645.

    https://doi.org/10.1523/JNEUROSCI.3078-16.2017

    • PubMed
    • Google Scholar
36. 1. Little S
    2. Bonaiuto J
    3. Barnes G
    4. Bestmann S

    (2019) Human motor cortical beta bursts relate to movement planning and response errors

    PLOS Biology 17:e3000479.

    https://doi.org/10.1371/journal.pbio.3000479

    • PubMed
    • Google Scholar
37. 1. Meyniel F
    2. Maheu M
    3. Dehaene S

    (2016) Human inferences about sequences: A minimal transition probability model

    PLOS Computational Biology 12:e1005260.

    https://doi.org/10.1371/journal.pcbi.1005260

    • PubMed
    • Google Scholar
38. 1. Mobbs D
    2. Trimmer PC
    3. Blumstein DT
    4. Dayan P

    (2018) Foraging for foundations in decision neuroscience: insights from ethology

    Nature Reviews. Neuroscience 19:419–427.

    https://doi.org/10.1038/s41583-018-0010-7

    • PubMed
    • Google Scholar
39. 1. Mochol G
    2. Kiani R
    3. Moreno-Bote R

    (2021) Prefrontal cortex represents heuristics that shape choice bias and its integration into future behavior

    Current Biology 31:1234–1244.

    https://doi.org/10.1016/j.cub.2021.01.068

    • PubMed
    • Google Scholar
40. 1. Moran R

    (2015) Optimal decision making in heterogeneous and biased environments

    Psychonomic Bulletin & Review 22:38–53.

    https://doi.org/10.3758/s13423-014-0669-3

    • PubMed
    • Google Scholar
41. 1. Morcos AS
    2. Harvey CD

    (2016) History-dependent variability in population dynamics during evidence accumulation in cortex

    Nature Neuroscience 19:1672–1681.

    https://doi.org/10.1038/nn.4403

    • PubMed
    • Google Scholar
42. 1. Murphy PR
    2. Wilming N
    3. Hernandez-Bocanegra DC
    4. Prat-Ortega G
    5. Donner TH

    (2021) Adaptive circuit dynamics across human cortex during evidence accumulation in changing environments

    Nature Neuroscience 24:987–997.

    https://doi.org/10.1038/s41593-021-00839-z

    • PubMed
    • Google Scholar
43. 1. Nienborg H
    2. Cumming BG

    (2009) Decision-related activity in sensory neurons reflects more than a neuron’s causal effect

    Nature 459:89–92.

    https://doi.org/10.1038/nature07821

    • PubMed
    • Google Scholar
44. 1. O’Connell RG
    2. Dockree PM
    3. Kelly SP

    (2012) A supramodal accumulation-to-bound signal that determines perceptual decisions in humans

    Nature Neuroscience 15:1729–1735.

    https://doi.org/10.1038/nn.3248

    • PubMed
    • Google Scholar
45. 1. O’Connell RG
    2. Kelly SP

    (2021) Neurophysiology of human perceptual decision-making

    Annual Review of Neuroscience 44:495–516.

    https://doi.org/10.1146/annurev-neuro-092019-100200

    • PubMed
    • Google Scholar
46. 1. Oostenveld R
    2. Fries P
    3. Maris E
    4. Schoffelen JM

    (2011) FieldTrip: Open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data

    Computational Intelligence and Neuroscience 2011:156869.

    https://doi.org/10.1155/2011/156869

    • PubMed
    • Google Scholar
47. 1. Ossmy O
    2. Moran R
    3. Pfeffer T
    4. Tsetsos K
    5. Usher M
    6. Donner TH

    (2013) The timescale of perceptual evidence integration can be adapted to the environment

    Current Biology 23:981–986.

    https://doi.org/10.1016/j.cub.2013.04.039

    • PubMed
    • Google Scholar
48. 1. Pape AA
    2. Siegel M

    (2016) Motor cortex activity predicts response alternation during sensorimotor decisions

    Nature Communications 7:13098.

    https://doi.org/10.1038/ncomms13098

    • PubMed
    • Google Scholar
49. 1. Peixoto D
    2. Verhein JR
    3. Kiani R
    4. Kao JC
    5. Nuyujukian P
    6. Chandrasekaran C
    7. Brown J
    8. Fong S
    9. Ryu SI
    10. Shenoy KV
    11. Newsome WT

    (2021) Decoding and perturbing decision states in real time

    Nature 591:604–609.

    https://doi.org/10.1038/s41586-020-03181-9

    • PubMed
    • Google Scholar
50. 1. Pfurtscheller G
    2. Stancák A
    3. Neuper C

    (1996) Post-movement beta synchronization. A correlate of an idling motor area?

    Electroencephalography and Clinical Neurophysiology 98:281–293.

    https://doi.org/10.1016/0013-4694(95)00258-8

    • PubMed
    • Google Scholar
51. 1. Ratcliff R
    2. McKoon G

    (2008) The diffusion decision model: theory and data for two-choice decision tasks

    Neural Computation 20:873–922.

    https://doi.org/10.1162/neco.2008.12-06-420

    • PubMed
    • Google Scholar
52. 1. Rollwage M
    2. Loosen A
    3. Hauser TU
    4. Moran R
    5. Dolan RJ
    6. Fleming SM

    (2020) Confidence drives a neural confirmation bias

    Nature Communications 11:2634.

    https://doi.org/10.1038/s41467-020-16278-6

    • PubMed
    • Google Scholar
53. 1. Rouder JN
    2. Speckman PL
    3. Sun D
    4. Morey RD
    5. Iverson G

    (2009) Bayesian t tests for accepting and rejecting the null hypothesis

    Psychonomic Bulletin & Review 16:225–237.

    https://doi.org/10.3758/PBR.16.2.225

    • PubMed
    • Google Scholar
54. 1. Scott BB
    2. Constantinople CM
    3. Akrami A
    4. Hanks TD
    5. Brody CD
    6. Tank DW

    (2017) Fronto-parietal cortical circuits encode accumulated evidence with a diversity of timescales

    Neuron 95:385–398.

    https://doi.org/10.1016/j.neuron.2017.06.013

    • PubMed
    • Google Scholar
55. 1. Shadlen MN
    2. Kiani R

    (2013) Decision making as a window on cognition

    Neuron 80:791–806.

    https://doi.org/10.1016/j.neuron.2013.10.047

    • PubMed
    • Google Scholar
56. 1. Sherman MA
    2. Lee S
    3. Law R
    4. Haegens S
    5. Thorn CA
    6. Hämäläinen MS
    7. Moore CI
    8. Jones SR

    (2016) Neural mechanisms of transient neocortical beta rhythms: Converging evidence from humans, computational modeling, monkeys, and mice

    PNAS 113:E4885–E4894.

    https://doi.org/10.1073/pnas.1604135113

    • PubMed
    • Google Scholar
57. 1. Siegel M
    2. Donner TH
    3. Oostenveld R
    4. Fries P
    5. Engel AK

    (2007) High-frequency activity in human visual cortex is modulated by visual motion strength

    Cerebral Cortex 17:732–741.

    https://doi.org/10.1093/cercor/bhk025

    • PubMed
    • Google Scholar
58. 1. Siegel M
    2. Engel AK
    3. Donner TH

    (2011) Cortical network dynamics of perceptual decision-making in the human brain

    Frontiers in Human Neuroscience 5:21.

    https://doi.org/10.3389/fnhum.2011.00021

    • PubMed
    • Google Scholar
59. 1. St John-Saaltink E
    2. Kok P
    3. Lau HC
    4. de Lange FP

    (2016) Serial dependence in perceptual decisions is reflected in activity patterns in primary visual cortex

    The Journal of Neuroscience 36:6186–6192.

    https://doi.org/10.1523/JNEUROSCI.4390-15.2016

    • PubMed
    • Google Scholar
60. 1. Stolk A
    2. Todorovic A
    3. Schoffelen JM
    4. Oostenveld R

    (2013) Online and offline tools for head movement compensation in MEG

    NeuroImage 68:39–48.

    https://doi.org/10.1016/j.neuroimage.2012.11.047

    • PubMed
    • Google Scholar
61. 1. Talluri BC
    2. Braun A
    3. Donner TH

    (2021) Decision making: How the past guides the future in frontal cortex

    Current Biology 31:R303–R306.

    https://doi.org/10.1016/j.cub.2021.01.020

    • PubMed
    • Google Scholar
62. 1. Urai AE
    2. Braun A
    3. Donner TH

    (2017) Pupil-linked arousal is driven by decision uncertainty and alters serial choice bias

    Nature Communications 8:14637.

    https://doi.org/10.1038/ncomms14637

    • PubMed
    • Google Scholar
63. 1. Urai AE
    2. de Gee JW
    3. Tsetsos K
    4. Donner TH

    (2019) Choice history biases subsequent evidence accumulation

    eLife 8:e46331.

    https://doi.org/10.7554/eLife.46331

    • PubMed
    • Google Scholar
64. 1. Urai AE
    2. Donner TH

    (2022) Persistent activity in human parietal cortex mediates perceptual choice repetition bias

    Nature Communications 13:6015.

    https://doi.org/10.1038/s41467-022-33237-5

    • PubMed
    • Google Scholar
65. 1. van den Brink RL
    2. Wynn SC
    3. Nieuwenhuis S

    (2014) Post-error slowing as a consequence of disturbed low-frequency oscillatory phase entrainment

    The Journal of Neuroscience 34:11096–11105.

    https://doi.org/10.1523/JNEUROSCI.4991-13.2014

    • PubMed
    • Google Scholar
66. 1. Van Veen BD
    2. van Drongelen W
    3. Yuchtman M
    4. Suzuki A

    (1997) Localization of brain electrical activity via linearly constrained minimum variance spatial filtering

    IEEE Transactions on Bio-Medical Engineering 44:867–880.

    https://doi.org/10.1109/10.623056

    • PubMed
    • Google Scholar
67. 1. Wang XJ

    (2008) Decision making in recurrent neuronal circuits

    Neuron 60:215–234.

    https://doi.org/10.1016/j.neuron.2008.09.034

    • PubMed
    • Google Scholar
68. 1. Wang L
    2. Mruczek REB
    3. Arcaro MJ
    4. Kastner S

    (2015) Probabilistic maps of visual topography in human cortex

    Cerebral Cortex 25:3911–3931.

    https://doi.org/10.1093/cercor/bhu277

    • PubMed
    • Google Scholar
69. 1. Wilming N
    2. Murphy PR
    3. Meyniel F
    4. Donner TH

    (2020) Large-scale dynamics of perceptual decision information across human cortex

    Nature Communications 11:5109.

    https://doi.org/10.1038/s41467-020-18826-6

    • PubMed
    • Google Scholar
70. 1. Wimmer K
    2. Compte A
    3. Roxin A
    4. Peixoto D
    5. Renart A
    6. de la Rocha J

    (2015) Sensory integration dynamics in a hierarchical network explains choice probabilities in cortical area MT

    Nature Communications 6:6177.

    https://doi.org/10.1038/ncomms7177

    • PubMed
    • Google Scholar
71. 1. Wyart V
    2. de Gardelle V
    3. Scholl J
    4. Summerfield C

    (2012) Rhythmic fluctuations in evidence accumulation during decision making in the human brain

    Neuron 76:847–858.

    https://doi.org/10.1016/j.neuron.2012.09.015

    • PubMed
    • Google Scholar
72. 1. Yu A
    2. Cohen JD

    (2009)

    Sequential effects: Superstition or rational behavior

    Advances in Neural Information Processing Systems 21:1873–1880.

    • Google Scholar

The following is the authors’ response to the original reviews.

eLife assessment

This valuable work provides new insights into history-dependent biases in human perceptual decisionmaking. It provides compelling behavioral and MEG evidence that humans adapt their historydependent to the correlation structure of uncertain sensory environments. Further neural data analyses would strengthen some of the findings, and the studied bias would be more accurately framed as a stimulus- or outcome-history bias than a choice-history bias because tested subjects are biased not by their previous choice, but by the previous feedback (indicating the category of the previous stimulus).

Thank you for your constructive evaluation of our manuscript. We have followed your suggestion to frame the studied bias as ‘stimulus history bias’. We now use this term whenever referring to our current results. Please note that we instead use the generic term ‘history bias’ when referring to the history biases studied in the previous literature on this topic in general. This is because these biases were dependent on previous choice(s), previous stimuli, or previous outcomes, or combinations of some (or all) of these factors. We have also added several of your suggested neural data analyses so as to strengthen the support for our conclusions, and we have elaborated on the Introduction so as to clarify the gaps in the literature that our study aims to fill. Our revisions are detailed in our replies below. We also took the liberty to reply to some points in the Public Review, which we felt called for clarification of the main aims (and main contribution) of our study.

Reviewer #1 (Public Review):

This paper aims to study the effects of choice history on action-selective beta band signals in human MEG data during a sensory evidence accumulation task. It does so by placing participants in three different stochastic environments, where the outcome of each trial is either random, likely to repeat, or likely to alternate across trials. The authors provide good behavioural evidence that subjects have learnt these statistics (even though they are not explicitly told about them) and that they influence their decision-making, especially on the most difficult trials (low motion coherence). They then show that the primary effect of choice history on lateralised beta-band activity, which is well-established to be linked to evidence accumulation processes in decision-making, is on the slope of evidence accumulation rather than on the baseline level of lateralised beta.

The strengths of the paper are that it is: (i) very well analysed, with compelling evidence in support of its primary conclusions; (ii) a well-designed study, allowing the authors to investigate the effects of choice history in different stochastic environments.

Thank you for pointing out these strengths of our study.

There are no major weaknesses to the study. On the other hand, investigating the effects of choice/outcome history on evidence integration is a fairly well-established problem in the field. As such, I think that this provides a valuable contribution to the field, rather than being a landmark study that will transform our understanding of the problem.

Your evaluation of the significance of our work made us realize that we may have failed to bring across the main gaps in the literature that our current study aimed to fill. We have now unpacked this in our revised Introduction.

Indeed, many previous studies have quantified history-dependent biases in perceptual choice. However, the vast majority of those studies used tasks without any correlation structure; only a handful of studies have quantified history biases in tasks entailing structured environments, as we have done here (Abrahamyan et al., 2016; Kim et al., 2017; Braun et al., 2018; Hermoso-Mendizabal et al., 2020). The focus on correlated environments matters from an ecological perspective, because (i) natural environments are commonly structured rather than random (a likely reason for history biases being so prevalent in the first place), and (ii) history biases that change flexibly with the environmental structure are a hallmark of adaptive behavior. Critically, the few previous studies that have used correlated environments and revealed flexible/adaptive history biases were purely behavioral. Ours is the first to characterize the neural correlates of adaptive history biases.

Furthermore, although several previous studies have identified neural correlates of history biases in standard perceptual choice tasks in unstructured environments (see (Talluri et al., 2021) for a brief overview), most have focused on static representations of the bias in ongoing activity preceding the new decision; only a single monkey physiology study has tested for both a static bias in the pre-stimulus activity and a dynamic bias building up during evidence accumulation (Mochol et al., 2021). Ours is the first demonstration of a dynamic bias during evidence accumulation in the human brain.

The authors have achieved their primary aims and I think that the results support their main conclusions. One outstanding question in the analysis is the extent to which the source-reconstructed patches in Figure 2 are truly independent of one another (as often there is 'leakage' from one source location into another, and many of the different ROIs have quite similar overall patterns of synchronisation/desynchronisation.).

We do not assume (and nowhere state) that the different ROIs are “truly independent” of one another. In fact, patterns of task-related power modulations of neural activity would be expected to be correlated between many visual and action-related cortical areas even without leakage (due to neural signal correlations). So, one should not assume independence even for intracortically recorded local field potential data, fMRI data, or other data with minimal spatial leakage effects. That said, we agree that filter leakage will add a (trivial) component to the similarity of power modulations across ROIs, which can and should be quantified with the analysis you propose.

A possible way to investigate this further would be to explore the correlation structure of the LCMV beamformer weights for these different patches, to ask how similar/dissimilar the spatial filters are for the different reconstructed patches.

Thank you for suggesting this analysis, which provides a very useful context for interpreting the pattern of results shown in our Figure 2. We have now computed (Pearson) correlation coefficients of the LCMV beamformer weights across the regions of interest. The results are shown in the new Figure 2 – figure supplement 1. This analysis provided evidence for minor leakage between the source estimates for neighboring cortical regions (filter correlations <= than 0.22 on average across subjects) and negligible leakage for more distant regions. We now clearly state this when referring to Figure 2.

That said, we would also like to clarify our reasoning behind Figure 2. Our common approach to these source-reconstructed MEG data is to focus on the differences, rather than the similarities between ROIs, because the differences cannot be accounted for by leakage. Our analyses show clearly distinct, and physiologically plausible functional profiles across ROIs (motion coherence encoding in visual regions, action choice coding in motor regions), in line with other work using our general approach (Wilming et al., 2020; Murphy et al., 2021; Urai and Donner, 2022).

Most importantly, our current analyses focus on the impact of history bias on the build-up of actionselective activity in downstream, action-related areas; and we chose to focus on M1 only in order to avoid hard-to-interpret comparisons between neighboring action-related regions. Figure 2 is intended as a demonstration of the data quality (showing sensible signatures for all ROIs) and as a context for the interpretation of our main neural results from M1 shown in the subsequent figures. So, all our main conclusions are unaffected by leakage between ROIs.

We have now clarified these points in the paper.

Reviewer #2 (Public Review):

In this work, the authors use computational modeling and human neurophysiology (MEG) to uncover behavioral and neural signatures of choice history biases during sequential perceptual decision-making. In line with previous work, they see neural signatures reflecting choice planning during perceptual evidence accumulation in motor-related regions, and further show that the rate of accumulation responds to structured, predictable environments suggesting that statistical learning of environment structure in decision-making can adaptively bias the rate of perceptual evidence accumulation via neural signatures of action planning. The data and evidence show subtle but clear effects, and are consistent with a large body of work on decision-making and action planning.

Overall, the authors achieved what they set out to do in this nice study, and the results, while somewhat subtle in places, support the main conclusions. This work will have impact within the fields of decisionmaking and motor planning, linking statistical learning of structured sequential effects in sense data to evidence accumulation and action planning.

Strengths:

• The study is elegantly designed, and the methods are clear and generally state-of-the-art

• The background leading up to the study is well described, and the study itself conjoins two bodies of work - the dynamics of action-planning processes during perceptual evidence accumulation, and the statistical learning of sequential structure in incoming sense data

• Careful analyses effectively deal with potential confounds (e.g., baseline beta biases)

Thank you for pointing out these strengths of our study.

Weaknesses:

• Much of the study is primarily a verification of what was expected based on previous behavioral work, with the main difference (if I'm not mistaken) being that subjects learn actual latent structure rather than expressing sequential biases in uniform random environments.

As we have stated in our reply to the overall assessment above, we realize that we may have failed to clearly communicate the novelty of our current results, and we have revised our Introduction accordingly. It is true that most previous studies of history biases in perceptual choice have used standard tasks without across-trial correlation structure. Only a handful of studies have quantified history biases in tasks entailing structured environments that varied from one condition to the next (Abrahamyan et al., 2016; Kim et al., 2017; Braun et al., 2018; Hermoso-Mendizabal et al., 2020), and showed that history biases change flexibly with the environmental structure. Our current work adds to this emerging picture, using a specific task setting analogous to one of these previous studies done in rats (Hermoso-Mendizabal et al., 2020).

Critically, all the previous studies that have revealed flexible/adaptive history biases in correlated environments were purely behavioral. Ours is the first to characterize the neural correlates of adaptive history biases. And it is also the very first demonstration of a dynamic history-dependent bias (i.e., one that gradually builds up during evidence accumulation) in the human brain.

Whether this difference - between learning true structure or superstitiously applying it when it's not there - is significant at the behavioral or neural level is unclear. Did the authors have a hypothesis about this distinction? If the distinction is not relevant, is the main contribution here the neural effect?

We are not quite sure what exactly you mean with “is significant”, so we will reply to two possible interpretations of this statement.

The first is that you may be asking for evidence for any difference between the estimated history biases in the structured (i.e., Repetitive, Alternating) vs. the unstructured (i.e., Neutral) environments used in our experiment. We do, in fact, provide quantitative comparisons between the history biases in the structured and Neutral environments at the behavioral level. Figure 1D and Figure 1 – figure supplement 2A and accompanying text show a robust and statistically significant difference in history biases. Specifically, the previous stimulus weights differ between each of the biased environments and the Neutral environment and the weights shifted in expected and opposite directions for both structured environments, indicating a tendency to repeat the previous stimulus category in Repetitive and vice versa in Alternating (Figure1D). Going further, we also demonstrate that the adjustment of the history is behaviorally relevant in that it improves performance in the two structured environments, but not in the unstructured environment (Figure 1F and Figure 1 – figure supplement 2A and figure supplement 3).

The second is that you refer to the question of whether the history biases are generated via different computations in structured vs. random environments. Indeed, this is a very interesting and important question. We cannot answer this question based on the available results, because we here used a statistical (i.e., descriptive) model. Addressing this question would require developing and fitting a generative model of the history bias and comparing the inferred latent learning processes between environments. This is something we are doing in ongoing work.

• The key effects (Figure 4) are among the more statistically on-the-cusp effects in the paper, and the Alternating group in 4C did not reliably go in the expected direction. This is not a huge problem per se, but does make the key result seem less reliable given the clear reliability of the behavioral results

The model-free analyses in Figure 3C and 4B, C from the original version of our manuscript were never intended to demonstrate the “key effects”, but only as supplementary to the results from the modelbased analyses in Figures 3C and 4D, E in our current version of the manuscript. The latter show the “key effects” because they are a direct demonstration of the shaping of build-up of action-selective activity by history bias.

To clarify this, we now decided to focus Figures 3 and 4 on the model-based analyses only. This decision was further supported by noticing a confound in our model-independent analyses in new control analyses prompted by Reviewer #3.

Please note that the alternating bias in the Alternating environment is also less strong at the behavioral level compared to the bias in the Repetitive condition (see Figure 1D). A possible explanation is that a sequence of repetitive stimuli produces stronger prior expectations (for repetition) than an equally long sequence of alternating stimuli (Meyniel et al., 2016). This might also induce the bias to repeat the previous stimulus category in the Neutral condition (Figure 1D). Moreover, this intrinsic repetition bias might counteract the bias to alternate the previous stimulus category in Alternating.

• The treatment of "awareness" of task structure in the study (via informal interviews in only a subsample of subjects) is wanting

Agreed. We have now removed this statement from Discussion.

Reviewer #3 (Public Review):

This study examines how the correlation structure of a perceptual decision making task influences history biases in responding. By manipulating whether stimuli were more likely to be repetitive or alternating, they found evidence from both behavior and a neural signal of decision formation that history biases are flexibly adapted to the environment. On the whole, these findings are supported across an impressive range of detailed behavioral and neural analyses. The methods and data from this study will likely be of interest to cognitive neuroscience and psychology researchers. The results provide new insights into the mechanisms of perceptual decision making.

The behavioral analyses are thorough and convincing, supported by a large number of experimental trials (~600 in each of 3 environmental contexts) in 38 participants. The psychometric curves provide clear evidence of adaptive history biases. The paper then goes on to model the effect of history biases at the single trial level, using an elegant cross-validation approach to perform model selection and fitting. The results support the idea that, with trial-by-trial accuracy feedback, the participants adjusted their history biases due to the previous stimulus category, depending on the task structure in a way that contributed to performance.

Thank you for these nice words on our work.

The paper then examines MEG signatures of decision formation, to try to identify neural signatures of these adaptive biases. Looking specifically at motor beta lateralization, they found no evidence that starting-level bias due to the previous trial differed depending on the task context. This suggests that the adaptive bias unfolds in the dynamic part of the decision process, rather than reflecting a starting level bias. The paper goes on to look at lateralization relative to the chosen hand as a proxy for a decision variable (DV), whose slope is shown to be influenced by these adaptive biases.

This analysis of the buildup of action-selective motor cortical activity would be easier to interpret if its connection with the DV was more explicitly stated. The motor beta is lateralized relative to the chosen hand, as opposed to the correct response which might often be the case. It is therefore not obvious how the DV behaves in correct and error trials, which are combined together here for many of the analyses.

We have now unpacked the connection of the action-selective motor cortical activity and decision variable in the manuscript, as follows:

“This signal, referred to as ‘motor beta lateralization’ in the following, has been shown to exhibit hallmark signatures of the DV, specifically: (i) selectivity for choice and (ii) ramping slope that depends on evidence strength (Siegel et al., 2011; Murphy et al., 2021; O’Connell and Kelly, 2021).”

Furthermore, we have added a figure of the time course of the motor beta lateralization separately for correct and error trials, locked to both stimulus onset and to motor response (Figure 2 – figure supplement 2). This signal reached statistical significance earlier for correct than error trials, and during the stimulus interval it ramped to a larger (i.e., more negative) amplitude for correct trials (Figure 2 – figure supplement 2, left). But the signal was indistinguishable in amplitude between correct and error trials around the time of the motor response (Figure 2 – figure supplement 2, right). This pattern matches what would be expected for a neural signature of the DV, because errors are more frequently made on weak-evidence trials than correct choices and because even for matched evidence strength, the DV builds up more slowly before error trials in accumulator models (Ratcliff and McKoon, 2008).

--

As you will see, all three reviewers found your work to provide valuable insights into history-dependent biases during perceptual decision-making. During consultation between reviewers, there was agreement that what is referred as a choice-history bias in the current version of the manuscript should rather be framed as a stimulus- or outcome-history bias (despite the dominant use of the term 'choicehistory' bias in the existing literature), and the reviewers pointed toward further analyses of the neural data which they thought would strengthen some of the claims made in the preprint. We hope that these comments will be useful if you wish to revise your preprint.

We are pleased to hear that the reviewers think our work provides valuable insights into historydependent biases in perceptual decision-making. We thank you for your thoughtful and constructive evaluation of our manuscript.

We have followed your suggestion to frame the studied bias as ‘stimulus history bias’. We now use this term whenever referring to our current results. Please note that we instead use the generic term ‘history bias’ when referring to the history biases studied in the previous literature on this topic in general. This is because these biases were dependent on previous choice(s), previous stimuli, or previous outcomes, or combinations of some (or all) of these factors.

We have also performed several of your suggested neural data analyses so as to strengthen the support for our conclusions.

Reviewer #1 (Recommendations For The Authors):

One suggestion is to explore the correlation structure of the LCMV beam former weights for the regions of interest in the study, for the reasons outlined in my public review.

Again, thank you for suggesting this analysis, which provides a very useful context for interpreting the pattern of results shown in our Figure 2. We have now computed (Pearson) correlation coefficients of the LCMV beamformer weights across the regions of interest. The results are shown in the new Figure 2 – figure supplement 1. This analysis provided evidence for minor leakage between the source estimates for neighboring cortical regions (filter correlations <= than 0.22 on average across subjects) and negligible leakage for more distant regions. We now clearly state this when referring to Figure 2.

That said, we would also like to clarify our reasoning behind Figure 2. Our common approach to these source-reconstructed MEG data is to focus on the differences, rather than the similarities between ROIs, because the differences cannot be accounted for by leakage. Our analyses show clearly distinct, and physiologically plausible functional profiles across ROIs (motion coherence encoding in visual regions, action choice coding in motor regions), in line with other work using our general approach (Wilming et al., 2020; Murphy et al., 2021; Urai and Donner, 2022).

Most importantly, our current analyses focus on the impact of history bias on the build-up of actionselective activity in downstream, action-related areas; and we chose to focus on M1 only in order to avoid hard-to-interpret comparisons between neighboring action-related regions. Figure 2 is intended as a demonstration of the data quality (showing sensible signatures for all ROIs) and as a context for the interpretation of our main neural results from M1 shown in the subsequent figures. So, all our main conclusions are unaffected by leakage between ROIs.

We have now clarified also these points in the paper.

I also wondered if the authors had considered:

(i) the extent to which the bias changes across time, as the transition probabilities are being learnt across the experiment? given that these are not being explicitly instructed to participants, is any modelling possible of how the transition structure is itself being learnt over time, and whether this makes predictions of either behaviour or neural signals?

We refer to this point in the discussion. The learning of the transition probabilities which can and should be addressed. This requires generative models that capture the learning of the transition structure over time (Yu and Cohen, 2009; Meyniel et al., 2016; Glaze et al., 2018; Hermoso-Mendizabal et al., 2020).

The fact that our current statistical modeling approach successfully captures the bias adjustment between environments implies that the learning must be sufficiently fast. Tracking this process explicitly would be an exciting and important endeavor for the future. We think it is beyond the scope of the present study focusing on the trial-by-trial effect of history bias (however generated) on the build-up of action-selective activity.

(ii) neural responses at the time of choice outcome - given that so much of the paper is about the update of information in different statistical environments, it seems a shame that no analyses are included of feedback processing, how this differs across the different environments, and how might be linked to behavioural changes at the next trial.

We agree that the neural responses to feedback are a very interesting topic. We currently analyze these in another ongoing project on (outcome) history bias in a foraging task. We will consider re-analyzing the feedback component in the current data set, in this new study as well.

However, this is distinct from the main question that is in the focus of our current paper – which, as elaborated above, is important to answer: whether and how adaptive history biases shape the dynamics of action-selective cortical activity in the human brain. While interesting and important, neural responses to feedback were not part of this question. So, we prefer to keep the focus of our paper on our original question.

Reviewer #2 (Recommendations For The Authors):

Minor:

-pg. 7: "inconstant"

-some citations (e.g., Barbosa 2020) are missing from the bibliography

Thank you for pointing this out. We have fixed these.

-figure S2 is very useful! could probably go in main text.

We agree that this figure is important. But we decided to show it in the Supplement (now Figure 1 – figure supplement 2) after careful consideration for two reasons. First, we wanted to put the reader’s focus on the stimulus weights, because it is those weights, which are flexibly adjusted to the statistics of the environment rather than the choice weights, which seem less adaptive (i.e., stereotypical across environments) and idiosyncratic. Second, plotting the previous stimulus weights only enabled to add the individual weights in the Neutral condition, which would have been to cluttered to add to figure S2.

For these reasons, we feel that this Figure is more suitable for expert readers with a special interest in the details of the behavioral analyses and would be better placed in the Supplement. These readers will certainly be able to find and interpret that information in the Supplement.

Reviewer #3 (Recommendations For The Authors):

I would suggest that a more in depth description of the previous literature that explains exactly how the features of the lateralized beta--as it is formulated here-- reflect the decision variable would assist with the readers' understanding. A demonstration of how the lateralized beta behaves under different coherence conditions, or for corrects vs errors, for example, might be helpful for readers.

We now provide a more detailed description of how/why the motor beta lateralization is a valid proxy of DV in the revised paper.

We have demonstrated the dependence of the ramping of the motor beta lateralization on the motion coherence using a regression model with current signed motion coherence as well as single trial bias as regressors. The beta weights describing the impact of the signed motion coherence on the amplitude as well as on the slope of the motor beta lateralization are shown in Figure 4G (now 4E). As expected, stronger motion coherence induces a steeper downward slope of the motor beta lateralization.

Furthermore, we have added a figure of the time course of the motor beta lateralization separately for correct and error trials, locked to both stimulus onset and to motor response (Figure 2 – figure supplement 2). This signal reached statistical significance earlier for correct than error trials, and during the stimulus interval it ramped to a larger (i.e., more negative) amplitude for correct trials (Figure 2 – figure supplement 2, left). But the signal was indistinguishable in amplitude between correct and error trials around the time of the motor response (Figure 2 – figure supplement 2, right).This pattern matches what would be expected for a neural signature DV, because errors are more frequently made on weakevidence trials than correct choices and because even for matched evidence strength, the DV builds up more slowly before error trials in accumulator models (Ratcliff and McKoon, 2008).

Finally, please note that our previous studies have demonstrated that the time course of the beta lateralization during the trial closely tracks the time course of a normative model-derived DV (Murphy et al., 2021) and that the motor beta ramping slope is parametrically modulated by motion coherence (de Lange et al., 2013), which is perfectly in line with the current results.

Along similar lines, around figures 3c and 4B, some control analyses may be helpful to clarify whether there are differences between the groups of responses consistent and inconsistent with the previous trial (e.g. correctness, coherence) that differ between environments, and also could influence the lateralized beta.

Thank you for pointing us to this important control analysis. We have done this, and indeed, it identified accuracy and motion strength as possible confounds (Author response image 1). Specifically, proportion correct as well as motion coherence were larger for consistent vs. inconsistent conditions in Repetitive and vice versa in Alternating. Those differences in accuracy and coherence might indeed influence the slope of the motor beta lateralization that our model-free analysis had identified, rendering the resulting difference between consistent and inconsistent difficult to interpret unambiguously in terms of bias. Thus, we have decided to drop the consistency (i.e., model-independent) analysis and focus completely on the modelbased analyses.

Author response image 1

Download asset Open asset

Importantly, this decision has no implications for the conclusions of our paper: The model-independent analyses in the original versions of Figure 3 and 4 were only intended as a supplement to the most conclusive and readily interpretable results from the model-based analyses (now in Figs. 3C and 4D, E). The latter are the most direct demonstration of a shaping of build-up of action-selective activity by history bias, and they are unaffected by these confounds.

In addition, I wondered whether the bin subsampling procedure to match trial numbers for choice might result in unbalanced coherences between the up and down choices.

The subsampling itself did not cause any unbalanced coherences between the up and down choices, which we now show in Figure 4 – figure supplement 1. There was only a slight imbalance in coherences between up and down choices before the subsampling which then translated into the subsampled trials but the coherences were equally distributed before as compared to after the subsampling.

Also, please note that the purpose of this analysis was to make the neural bias directly “visible” in the beta lateralization data, rather than just regression weights. The issue does not pertain to the critical single-trial regression analysis, which yielded consistent results.

References

Abrahamyan A, Silva LL, Dakin SC, Carandini M, Gardner JL (2016) Adaptable history biases in human perceptual decisions. Proceedings of the National Academy of Sciences 113:E3548–E3557.

Braun A, Urai AE, Donner TH (2018) Adaptive History Biases Result from Confidence-weighted Accumulation of Past Choices. The Journal of Neuroscience:2189–17. de Lange FP, Rahnev DA,Donner TH, Lau H (2013) Prestimulus Oscillatory Activity over Motor Cortex Reflects Perceptual Expectations. Journal of Neuroscience 33:1400–1410.

Glaze CM, Filipowicz ALS, Kable JW, Balasubramanian V, Gold JI (2018) A bias–variance trade-off governs individual differences in on-line learning in an unpredictable environment. Nat Hum Behav 2:213–224.

Hermoso-Mendizabal A, Hyafil A, Rueda-Orozco PE, Jaramillo S, Robbe D, de la Rocha J (2020) Response outcomes gate the impact of expectations on perceptual decisions. Nat Commun 11:1057.

Kim TD, Kabir M, Gold JI (2017) Coupled Decision Processes Update and Maintain Saccadic Priors in a Dynamic Environment. The Journal of Neuroscience 37:3632–3645.

Meyniel F, Maheu M, Dehaene S (2016) Human Inferences about Sequences: A Minimal Transition Probability Model Gershman SJ, ed. PLOS Computational Biology 12:e1005260.

Mochol G, Kiani R, Moreno-Bote R (2021) Prefrontal cortex represents heuristics that shape choice bias and its integration into future behavior. Current Biology 31:1234-1244.e6.

Murphy PR, Wilming N, Hernandez-Bocanegra DC, Prat-Ortega G, Donner TH (2021) Adaptive circuit dynamics across human cortex during evidence accumulation in changing environments. Nat Neurosci 24:987–997.

O’Connell RG, Kelly SP (2021) Neurophysiology of Human Perceptual Decision-Making. Annu Rev Neurosci 44:495–516.

Ratcliff R, McKoon G (2008) The Diffusion Decision Model: Theory and Data for Two-Choice Decision Tasks. Neural Computation 20:873–922.

Siegel M, Engel AK, Donner TH (2011) Cortical Network Dynamics of Perceptual Decision-Making in the Human Brain. Frontiers in Human Neuroscience 5 Available at: http://journal.frontiersin.org/article/10.3389/fnhum.2011.00021/abstract [Accessed April 8, 2017].

Talluri BC, Braun A, Donner TH (2021) Decision making: How the past guides the future in frontal cortex. Current Biology 31:R303–R306.

Urai AE, Donner TH (2022) Persistent activity in human parietal cortex mediates perceptual choice repetition bias. Nat Commun 13:6015.

Wilming N, Murphy PR, Meyniel F, Donner TH (2020) Large-scale dynamics of perceptual decision information across human cortex. Nat Commun 11:5109.

Yu A, Cohen JD (2009) Sequential effects: Superstition or rational behavior. Advances in neural information processing systems 21:1873–1880.

Author details

1. Anke Braun

   1. Section Computational Cognitive Neuroscience, Department of Neurophysiology and Pathophysiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
   2. Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Psychiatry and Neurosciences, Berlin, Germany
   3. Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Child and Adolescent Psychiatry, Berlin, Germany

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing

   For correspondence

   anke.braun86@gmail.com

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1946-7765

2. Tobias H Donner

   1. Section Computational Cognitive Neuroscience, Department of Neurophysiology and Pathophysiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
   2. Bernstein Center for Computational Neuroscience, Charité – Universitätsmedizin Berlin, Berlin, Germany

   Contribution

   Conceptualization, Supervision, Funding acquisition, Writing – original draft, Project administration, Writing – review and editing

   For correspondence

   t.donner@uke.de

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-7559-6019

• 35

  Page views

• 4

  Downloads

• 0

  Citations

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