Precursor Processes of Human Schedule Responding: Goal-Directedness Emerges After Burst Initiation

Precursor Processes of Human Schedule Responding: Goal-Directedness Emerges After Burst Initiation | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL This is a preprint and has not been peer reviewed. Data may be preliminary. 31 December 2025 V1 Latest version Share on Precursor Processes of Human Schedule Responding: Goal-Directedness Emerges After Burst Initiation Authors : Xiaosheng Chen , Dan Zhang [email protected] , and Phil Reed Authors Info & Affiliations https://doi.org/10.22541/au.176718077.78729703/v1 Published Psychophysiology Version of record Peer review timeline 171 views 77 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Goal-directedness is a core feature of human volitional behaviour. The present study (N = 40) integrates microstructural behavioural analysis with psychophysiological measures to examine how goal-directed actions unfold over time. Using a reinforcement-learning paradigm with matched reward rates, we elicited both time-driven and event-driven behaviour and examined neural activity preceding two co-occurring microstructural response types within these contexts: burst-initiation responses and subsequent within-burst responses. Neural variability decreased more markedly prior to within-burst responses than prior to burst-initiation responses, suggesting stronger engagement of neural processes associated with goal-directed control following burst initiation, that is, after responding has already commenced. In addition, trial-to-trial readiness potential (RP) variability decreased more strongly prior to event-driven behaviour, which was characterised by a higher proportion of within-burst responses than time-driven behaviour. Results further revealed that, across behavioural contexts, a higher proportion of within-burst responses was associated with stronger neural signatures of goal-directed action. More broadly, these findings highlight the dynamic nature of goal-directed control and demonstrate how psychophysiological measures can capture moment-to-moment fluctuations in volitional engagement during ongoing behaviour. Precursor Processes of Human Schedule Responding: Goal-Directedness Emerges After Burst Initiation Xiaosheng Chen 1,2 , Dan Zhang 1 , and Phil Reed 3 1 Department of Psychology, Tsinghua University, Beijing 100084, China; 2 Department of Neurobiology, Stanford University, CA 94305, U.S.A.; 3 School of Psychology, Swansea University, SA2 8PP, UK. Corresponding author: Dan Zhang, Department of Psychological and Cognitive Science, Tsinghua University, Beijing, China. E-mail: [email protected] Short title: Developing goal-directedness of actions Submitted to: Psychophysiology September (2025) Abstract Goal-directedness is a core feature of human volitional behaviour. The present study (N = 40) integrates microstructural behavioural analysis with psychophysiological measures to examine how goal-directed actions unfold over time. Using a reinforcement-learning paradigm with matched reward rates, we elicited both time-driven and event-driven behaviour and examined neural activity preceding two co-occurring microstructural response types within these contexts: burst-initiation responses and subsequent within-burst responses. Neural variability decreased more markedly prior to within-burst responses than prior to burst-initiation responses, suggesting stronger engagement of neural processes associated with goal-directed control following burst initiation, that is, after responding has already commenced. In addition, trial-to-trial readiness potential (RP) variability decreased more strongly prior to event-driven behaviour, which was characterised by a higher proportion of within-burst responses than time-driven behaviour. Results further revealed that, across behavioural contexts, a higher proportion of within-burst responses was associated with stronger neural signatures of goal-directed action. More broadly, these findings highlight the dynamic nature of goal-directed control and demonstrate how psychophysiological measures can capture moment-to-moment fluctuations in volitional engagement during ongoing behaviour. Keywords: goal-directedness, RP variability, reinforcement schedule, burst-initiation, within-burst Research Transparency Statement Acknowledgments: We thank Dr. Aaron Schurger for critical review and helpful suggestions. We are also grateful to Mr. Zhongyu Li and Dr. Jiaoting Li for assistance during data analysis. Funding: This research was supported in full by the China Postdoctoral Science Foundation (Grant No. 2021M701951). Author Contributions: All authors conceptualized the project. XC performed the experiments and wrote the methods and results sections. XC wrote the introduction and discussion, prepared figures, and drafted the manuscript. DZ and PR edited and revised the manuscript. All authors approved the final manuscript. DZ supervised the project. Competing Interest Statement: The authors declare no competing interests. Artificial Intelligence: No artificial intelligence–assisted technologies were used in the conduct of this research or in the writing of this manuscript. Ethics Approval: The study was approved by the Institutional Review Board at Tsinghua University, and all participants provided informed consent in accordance with institutional guidelines. Data Availability: All data and custom code supporting the findings of this study will be made publicly available upon publication. Prior to that, materials are available from the corresponding author upon reasonable request. Goal-directed behaviour, in contrast to automatic and unintentional ‘habitual’ behaviour, refers to a class of actions guided by past experience, current cues, and anticipated future goals. It is an important feature of human volitional behaviour and serves as a key behavioural entry point for the study of consciousness/volition (e.g., Dickinson, 1985; Haggard, 2008). Understanding how goal-directed behaviour unfolds is therefore critical for elucidating the nature of human agency and cognitive control, and potentially the mechanisms underlying psychiatric disorders driven by impaired goal regulation. The use of such procedures, including free-operant schedules of reinforcement (Ferster & Skinner, 1957; Killeen, 2023), in combination with neural measurements, would allow a more ecologically valid and fine-grained analysis of fluctuations in voluntary, goal-directed action. Our current approach integrates recent advances in behavioural and neural analysis and challenges traditional dichotomies between ‘goal-directed’ and automatic, unintentional ‘habitual’ modes of control by demonstrating that volitional, goal-related engagement varies within the microstructure of freely emitted behaviour, providing new insight into how intentional goal-directedness unfolds over time. Reinforcement schedule paradigms offer powerful frameworks for investigating freely-emitted behaviours (Killeen, 2023; Zeiler, 1984). Their response components can be delineated into goal-directed ‘actions’ and stimulus-driven ‘habits’ (Chen & Reed, 2024; Dickinson, 1985; Perez & Dickinson, 2020; Thrailkill, Trask, Vidal, Alcalá, & Burston, 2018). Random ratio (RR) and random interval (RI) schedules are widely used to dissociate goal-directed from habitual behaviour (Dickinson, 1985; Garr & Delamater, 2019; Perez et al., 2019; Thrailkill & Burston, 2015). RI schedules deliver reinforcement for a response after variable time intervals and simulate time-driven behaviour, such as animal hibernation. Such schedules do not produce a strong relationship between variations in response rate and variations in reinforcement rate, and therefore tend to promote automatic or habitual responding. In contrast, RR schedules deliver reinforcement probabilistically based on response counts and simulate event-driven behaviour, such as exam performance. They exhibit strong correlations between response rate and reinforcement rate and are more conducive to action–outcome learning (Dickinson, 1985; Perez & Dickinson, 2020). Behavioural studies have highlighted the importance of analysing the microstructure of operant performance (Falligant, Hagopian, & Newland, 2024; Reed et al., 2018; Shull, 2011) – particularly the distinction between ‘burst-initiation’ responses following longer pauses, and ‘within-burst’ responses reflecting subsequent sustained responding (Shull, 2011). Burst-initiation responses are controlled by factors such as reinforcement rates in the conditioning context, motivation, and the contextual stimuli (Falligant et al., 2024; Reed et al., 2018; Shull, 2011). In contrast, within-burst responding appears controlled by the action of reinforcement (Reed, 2015; Shull, 2011). Within-burst responses are more strongly affected by reinforcer devaluation procedures, implying a greater degree of goal-directedness (Reed, 2025). They are also more sensitive to verbal instructions (Chen & Reed, 2024) and mindfulness procedures (Chen & Reed, 2024). Together, these findings indicate that within-burst responses exhibit a higher level of goal-directedness than burst-initiation responses. This is because within-burst responses are significantly more sensitive to reinforcer devaluation, verbal instructions, and mindfulness manipulations, that is, they adjust response patterns more efficiently to achieve the task goal and are associated with more consciously mediated control, compared to burst-initiation responses. Importantly, responses under RR schedules tend to produce a greater proportion of within-burst responses than under RI schedules, even when reinforcement rate is matched (Reed, 2020). This suggests that the observed schedule differences may be driven less by differences in the overall schedule, and more by changes and shifts in the distribution of response types during a session. The above behavioural analysis suggests that initial responding in a burst is evoked by contextual stimuli, and may be more habitual, and only once responding has commenced (i.e. within-burst responding) do these responses increasingly engage intentional, goal-directed control actions (Reed, 2025). Advances in neural data analysis have enabled increasingly fine-grained investigation of volitional processes, yet such approaches have rarely been applied to schedule-controlled performance. This has limited efforts to integrate behavioural analyses of ongoing action (e.g., Shull, 2011) with neural measures of volition (e.g., Libet, 1982; 1983). Kornhuber and Deecke (1965) demonstrated that a slow, negative-going ramp of converging neural activity precedes self-initiated voluntary action by several hundred milliseconds. This signal, termed the readiness potential (RP), has been widely interpreted as the electrophysiological signature of action planning, preparation, and initiation. Building on this work, Khalighinejad et al. (2018, 2019) showed that trial-to-trial variability in the RP decreases prior to self-initiated actions, suggesting that reductions in neural noise may reflect the stabilization of internal dynamics supporting volitional control. In this framework, greater reductions in RP variability are taken to index higher levels of behavioural engagement toward achieving an action goal. Such reductions likely reflect an integrated set of psychological processes, including decision formation, action initiation, and anticipatory aspects of action, that substantially overlap with the core components of goal-directed behaviour (Haggard, 2008; Balleine & O’Doherty, 2010; Schurger et al., 2012). Consistent with this view, prior work has identified RP-related neural signals as markers for distinguishing goal-directed from non-goal-directed behaviour and for indexing sensitivity to action outcomes (Dehaene & Naccache, 2011; Martinez & Colom, 2008; Schurger et al., 2012). In the present study, RP variability is adopted as a central index of the degree to which behaviour is goal-directed within an ongoing behavioural context. Notably, however, existing studies of the neural correlates of goal-directed responding have largely relied on discrete-trial paradigms (e.g., Khalighinejad et al., 2018; Schurger et al., 2012), rather than examining neuro-behavioural associations during continuous, dynamically structured behaviour. Building on these insights, we integrate RP analysis with computational modelling to examine how goal-directedness emerges dynamically across behavioural sequences maintained on free-operant schedules. We hypothesised that within-burst responses would be associated with greater reductions in trial-to-trial RP variability, reflecting more consistent, intentional goal-directed control, compared to that observed for burst-initiation responses. We further predicted that event-driven RR schedules would amplify this pattern by increasing the proportion of within-burst responses. Such a pattern of results would not only integrate a range of behavioural, neural, and computational analyses, but would also suggest that conscious awareness of behavioural goal-directed intentions fluctuates in real time as behaviour unfolds. Participants Forty volunteers were recruited. All were right-handed, had normal or corrected-to-normal vision, and had no self-reported history of seizure, epilepsy, or other neurologic or psychiatric disorders. Three participants failed to learn the behavioural task, leaving 37 participants (mean age = 22, SD ± 2). G*Power calculations conducted prior to data collection indicated that, assuming an effect size of f = .25, 80% power, and α = .05, the current sample size was sufficient (Allman et al., 2010). The study was approved by the local Ethics Committee of Tsinghua University, and written informed consent was obtained from each participant. Experimental Task —-Figure 1—- The current paradigm was adapted from previously established reinforcement schedule procedures for humans (Bradshaw & Reed., 2012; Chen et al., 2020; 2021, Figure 1a). Three boxes were presented on a computer screen (8cm wide × 3cm high) with a white background. The top box (schedule signal) was placed in the center upper portion of the screen, and displayed a single colour (blue or pink) to indicate the operative schedule (RR or RI) throughout each trial. The colour associated with each schedule was constant for each participant, but was counter-balanced across participants. The middle box was presented in the center middle of the screen, and presented the running points total (which was set to 10 at the commencement of each trial). The bottom box was in the center lower portion of the screen, and was the response area. Participants were requested to place the mouse cursor in the bottom box, and then press the space bar to earn points. The procedure employed a multiple RR-10, yoked RI reinforcement schedule. During the RR schedule, reinforcement (40 points added to the total) was presented for responses with a 1/10 probability. On the following RI schedule, reinforcement was delivered for the first response made after a specified amount of time had elapsed; the time was yoked to the time of the corresponding reinforcer on the preceding RR schedule. Each response made on each schedule subtracted 1 point from the total. This manipulation was introduced as it has previously been noted that, in the absence of cost for a keypress, there is no penalty for over-responding and, therefore, no reason to regulate performance in line with the contingency of the schedule (Bradshaw & Reed, 2012; Reed, 2001). Two forms of micro-structural response were analysed: within-burst and burst-initiation responses (Killeen, Hall, Reilly, & Kettle, 2002; Reed, 2020; Shull, 2011; Yamada & Kanemura, 2020). The 1000ms ‘cut-off’ method was employed to reveal the micro-structure of responding (Mellgren & Elsmore, 1991). This method designates inter-response times (IRTs) shorter than 1000ms as within-burst responses, and IRTs longer than 1000ms as burst-initiation responses. In the context of human schedule performance studies, an IRT of 1000ms has proved a valid index of this distinction (Chen & Reed, 2023; Reed, 2015; Reed et al., 2018). Behavioural Task Participants were tested individually in an electrically shielded chamber, with a 50cm computer screen (DELL, USA), situated approximately 60cm in front of the participant. Participants read the study information and paper instructions for the task, and the experimental procedure was explained as the EEG cap was being prepared. Participants were then presented with the behavioural task on the computer. Participants were told only that different schedules would require different response strategies, but they were not explicitly informed the type of schedule or response pattern, and they learned the reinforcement contingency through trial and error. They were given instructions, on the computer screen: “When the task begins, position the cursor over the rectangle which states “click here”, and use the space bar to score as many points as possible. There are eight games in total. The first game is identified with a large blue [pink] rectangle at the top of the screen. When the first game is over, the rectangle will change to pink [blue] to indicate the start of the next game. The rectangles alternate between blue and pink [pink and blue] to indicate the changing games for the remainder of the task. Your goal in each game is to reach the highest score possible. You will see that the points reduce according to the way in which you play but will rise again every so often, according to the pattern of space bar hits that you use. All you need to do is to find the best pattern of space bar hits to score as highly as possible in each game. It may be a good idea to respond quickly sometimes, and slowly at other times, but you need to discover this for yourself!” . Participants completed one practice trial consisting of responding on an RR schedule followed by a yoked RI schedule, to familiarise themselves with the task. As illustrated in Figure 1 (right panel), the upper panel depicts the complete task structure for a single RR–yoked RI–externally triggered section pair, whereas the lower panel shows the overall task structure comprising four RR–yoked RI–externally triggered section pairs (i.e., with the first pair serving as the training section). They then completed three RR, three yoked RI, schedules. The RR and RI trials always came in pairs, with the RR-10 schedule preceding the yoked RI schedule. Each schedule presentation was 4min long. Throughout the experiment, participants were required to keep their hands close to the space bar to avoid increasing the reaction time, and try to press without hesitation. The response hands (left or right) of the participants were counter-balanced across all participants. Once the task was completed, participants were debriefed and paid. EEG Recording EEG signals were recorded from 64 electrodes (FPZ, FP1/2, AFz, AF3/4, AF7/8, Fz, F1/2, F3/4, F5/6/, F7/8, FC1/2, FC3/4, FC5/6, FC7/8, Cz, C1/2, C3/4, C5/6, T7/8, CPz, CP1/2, CP3/4, CP5/6, TP7/8, Pz, P1/2, P3/4, P5/6, P7/8, POz, PO3/4, PO5/6, PO7/8, Oz, O1/2). The reference and ground were set to CPz and AFz. A portable wireless EEG amplifier (NeuSen. W64, Neuracle, China) was used for data recording, at a sampling rate of 1000 Hz. Electrode impedances were kept below 10 kOhm. EEG Pre-processing The recorded EEG data were down sampled to 250 Hz, and band-pass filtered at 0.1-30 Hz. Then EEG data were re-referenced to the average of the two electrodes near mastoids (T7/T8), following previous studies (Khalighinejad, 2018; Travers & Haggard, 2021). The EEG amplitudes trials were defined as the EEG segment starting 2s before each keypress to 1s afterward for voluntary (RR and RI) and externally triggered response trials. A [-0.05 s, 0.05 s] baseline was chosen (Khalighinejad et al., 2018; 2019), and the mean amplitudes within the baseline period were subtracted for all the EEG signals per trial. All responses made during the last three RR and RI blocks were analysed. As participants might make keypress responses within a relatively short time interval, any keypresses made within 300 ms were rejected to avoid ERP contamination caused by motor actions and to prevent overlap with the waveform from the previous response (cf. Woldorff, 1993; Luck, 2014). Overall, 15\(\pm 0.35\%\ \)of responses were removed for this reason. This exclusion procedure effectively balanced the number of remaining trials across the RR and RI schedules, as well as between within-burst and burst-initiation responses, because more short inter-response times (IRTs) were emitted under the RR schedule than the RI schedule, and more short IRTs occurred within bursts than at burst initiation. Following this exclusion, mean overall responses per min during blocks 2 to 4, was: 101.55±18.60 for RR, and 99.64±17.72, for RI schedule, t <1, indicating that no statistically significant difference in response rates was observed between the RR and RI schedules during blocks 2–4. Independent component analysis (ICA) was conducted for artifact rejection. The independent components (ICs) with large weights over frontal or temporal areas, together with a corresponding temporal course showing eye-movement or muscle-movement activities, were removed. The remaining ICs were then back-projected onto the scalp EEG channels, reconstructing the artifact-free EEG signals. On average, 3-4 ICs were rejected per participant. The segmented responses were subject to an additional artifact rejection procedure: responses with an absolute value exceeding 100 μV in any of the channels of interest were rejected. Here, the channels of interest (Cz/1/2/3/4) were chosen, covering the centrally-distributed RP-related electrodes (see Goldberg, Busch, & Meer, 2017; Wittmann et al., 2014; and our results). Finally, 165 ± 50.3 artifact-free responses were obtained. EEG data Analysis Readiness potential (RP) activity was analyzed from five central electrodes (Cz, C1, C2, C3, and C4), where RP amplitudes are typically maximal (Kornhuber & Deecke, 1965; Shibasaki & Hallett, 2006). The central cluster was selected to capture activity associated with the motor preparation process while minimizing the influence of non-motor components from frontal or parietal regions. Averaging across these electrodes also provided a more stable estimate of the RP signal. The FieldTrip toolbox, based on the MATLAB platform, was adopted to conduct the RP analysis. Two variables were calculated as indicators for each participant: (1) mean value and (2) standard deviation (SD) of RP amplitudes across trials, that is, trial to trial ERP variability. To compare RP fluctuation between different conditions, a 60ms window was used to divide data epochs, starting 2 s prior to a response and ending 1s after the response. To control for multiple comparisons, false discovery rate (FDR) correction was applied across all time points and electrodes (Cz, C1, C2, C3, and C4) for each condition contrast (within-burst vs. burst initiation; RR vs. RI). Modelling and Simulations In order to simulate the RP fluctuations before responses, the modified version of the Leaky Stochastic Accumulator Model was adopted, which has been validated in previous RP-modelling studies (Nima et al., 2018). While the drift-diffusion model (DDM) was originally developed for paradigms involving discrete decisions with clear evidence accumulation (i.e., one decision per trial), its application here is still conceptually justified. In the current task, participants must discover the optimal response pattern through trial and error. Each subsequent response is guided by feedback from the previous one, and participants gradually accumulate internal evidence toward the final goal. Thus, the DDM framework remains suitable for modeling this process of dynamic evidence accumulation and goal-directed adjustment over trials. The core formula of the leaky accumulator model is shown in Equation 1, illustrating that the activity of the accumulators changes stochastically but is constrained by several key parameters, including the drift rate (I), the leak parameter (k), and the noise scaling factor (cₜ). The model assumes that the accumulation process is not purely linear but “leaky,” allowing the accumulated evidence to decay over time, thereby providing a more realistic representation of dynamic decision processes in continuous responding tasks. The noise scaling factor (cₜ) was modeled as a linearly changing parameter, increasing from c₁ to c₂ to capture gradual changes in internal noise during learning. Other parameters, such as the non-decision time and boundary separation, were fixed across conditions to reduce model complexity and ensure parameter identifiability. This approach allowed us to focus on the parameters most relevant to our hypotheses (e.g., drift rate and leak), which directly reflect the efficiency and stability of goal-directed evidence accumulation in this task. \begin{equation} \delta x=\left(I-kx\right)\Delta t+c_{t}\xi\sqrt{\text{Δt}}\ (1)\nonumber \\ \end{equation} Parameter estimation procedure was conducted as described by Nima et al. (2018). Firstly, the leaky model generated 1000 long trials, based on default parameters, each trial with 50,000 timesteps. Then, the first cross-threshold event in each long trial was located, and computed as a simulated action. The simulation loss was defined as the root mean squared deviation (RMSD) of the amplitude between simulated action and real RP data. Finally, after iterating the least square algorithm, the local optimal parameters were output as an estimated result. Before calculation, all five parameters ( I, k, c 1 , c 2 , threshold) were set as free parameters in RI condition, and two of them ( c 1 , threshold) were fixed in RR condition for better comparison. Similarly, the simulation for within-burst responses also started with three free parameters, with c 1 and threshold fixed. For each participant, the estimation process produced one set of optimal parameters. Using these parameters, the generating process above was conducted, and produced the final simulated results for each participant (37 participants \(\times\) 4 conditions). The similarity of SD between the simulated data and observed RP was then demonstrated by computing the RP convergence, which equals the area under one (RI/burst initiation) curve minus the area under another (RR /within burst) curve from -2 s to 0s. Results Behavioural data Figure 1B shows that the mean overall response rate for blocks 2 to 4 was higher on the RR than the RI schedule. A paired t-test revealed a significant difference, t (39)=14.18, p <.001, d =4.092[ 95% CI: 2.307:5.861]. Mean rates of burst-initiation (Figure 1C), using a 1000ms cut-off to distinguish burst-initiation and within-burst responses (Chen & Reed, 2024; Mellgren & Elsmore, 1991), were higher for the RR than RI schedule, t (39)=9.35, p <.001, d =2.698[1.439:3.933]. Within-burst responses (Figure 1D) were emitted at a significantly higher rate on the RR compared to the RI schedule, t (11)=12.37, p 300ms to ensure signal clarity and minimize contamination from prior responses. RP signals for responses <300ms were likely contaminated by muscular activity, and not representative of cognitively driven behaviour (Foster et al., 2012; Leifer et al., 2011). This exclusion balanced the number of responses between RR (86.35 ± 16.80) and RI (99.64 ± 17.72) schedules, with no significant difference between conditions (t(36) < 1, n.s.). RR vs RI RP amplitudes and variability: RR responses showed classical RP negative-going ramps indicative of goal-directed control. RP amplitude was lower prior to RR than RI responses (-0.91s to –0.01s, t (36)=2.73, p =.010, d =.45; Figure 2A, 2C). RP variability (SD of RP amplitude) was also lower for RR, suggesting reduced neural noise and greater goal-directedness (-2.0s to –0.01s; t (36)=5.45, p <.001, d =.90; Figure 2B, 2D). Within-burst vs Burst-initiation: Within-burst responses showed lower RP amplitude fluctuation than burst-initiation responses (-2.0s to –1.15s; t (36)=3.49, p =.001, d =.57), especially at Cz, C2, C3, and C4 (Figure 2E, 2G). RP variability was also lower prior to within-burst responses in average -2s to -0.5s; t (36)=2.58, p =.035, d =.25, especially at C3 (-2s to -0.01s; t (36)=2.03, p =.049, d =.33; Figure 2F, 2H). Modelling RP convergence —-Figure 3—- Using the modified leaky stochastic accumulator model (Khalighinejad et al., 2018), we simulated RP data and matched them to observed signals. The comparison between RR and RI schedules reveals significantly higher: drift, t (36)=7.12, p <.001, d =1.17; leak, t (36)=3.41, p =.002, d =.56); and noise reduction, t (36)=26.53, p <.001, d =4.36, parameters. The simulated RP closely matched the raw RP convergence ( r =.987, p <.001; Figure 3A–C). RR-RI task responses: Comparative results showed RR task responses reduced neural noise more than externally triggered behaviour, while RI did not. For within-burst versus burst-initiation, optimized parameters again showed higher drift (t(36)=9.01, p<.001, d=1.48), leak (t(36)=18.76, p<.001, d=3.08), and noise reduction (t(36)=27.02, p<.001, d=4.43) for within-burst. Simulated vs observed RP variability was highly correlated (r=.994, p<.001; Figure 2D–F). Externally triggered responses —-Figure 4—- Further validation analyses compared RP signals preceding externally-triggered responses with short (300ms) latencies. RP amplitudes were significantly lower before longer-latency responses (−2.0s to −1.09s, t (36)=2.32, p =.023, d =0.37; Figure 4A). This finding is consistent with previous studies showing that very short response intervals often capture noise caused by muscular activity, rather than genuine cognitive processing, thereby supporting our decision to use 300ms as the cutoff for filtering. Moreover, RP amplitude and variability were both lower prior to externally-triggered responses exceeding 1s, compared to those between 300ms and 1s (−1.52s to −1.45s, −1.21 s to −0.72s, −0.61s to −0.37s, −0.31s to 0.17s; t (36)=2.26, p =.006, d =0.55; Figures 4C, 4D). Although the waveform in Figure 3C showed brief polarity changes at certain time points, these fluctuations likely reflected transient neural or peripheral activity rather than meaningful cognitive effects. In particular, the direction changes may be attributed to neural instability induced by muscle-related tension or non-advanced motor preparation in trials with excessively short response latencies. This further supports that the observed neural differences under reinforced conditions were not merely due to the arbitrary 1-second time division, but instead reflect systematic influences of distinct micro-response types and cognitive processes. Notably, this pattern is the reverse of what we observed in the RR and yoked RI conditions, suggesting that the effects found in those conditions are not artifacts of the 1000ms classification threshold. Rather, they reflect genuine differences in the neural activity preceding responses, likely arising from distinct cognitive processes engaged under schedule-controlled behaviour. Discussion This study investigated how volitional goal-directedness develops within schedule-controlled behaviour, using RP variability as a neural marker (Khalighinejad et al., 2018; Schurger et al., 2012). Consistent with prior research, we replicated that event-driven RR behavioural contexts, elicited higher overall response rates than yoked-reward RI time-driven behavioural contexts (Ferster & Skinner, 1957; Bradshaw & Reed, 2012; Reed et al., 2018), and that RR schedules promoted more within-burst responses (Shull, 2011; Reed, 2020; Chen & Reed, 2024). Critically, we found that trial-to-trial RP variability decreased more strongly under event-driven RR schedules than under time-driven RI schedules, suggesting an overall enhancement of engagement in goal-directed processes in environments that support action-outcome learning, as is the case for RR schedule learning. In parallel, within-burst responses consistently showed greater reductions in neural variability than burst-initiation responses across both schedules. This dissociation supports our central claim: goal-directedness is more tightly linked to the structure of behaviour, specifically, the transition from burst initiation to within-burst responding, than to the overall reinforcement schedule (i.e. the behavioural context) per se. These findings refine classic dichotomies in instrumental learning. Rather than assuming that RR schedules inherently drive goal-directed behaviour, we demonstrate that RR schedules promote response structures (within-burst responding) that are themselves more tightly coupled to cognitive goal-directed control. This microstructural account aligns with theoretical perspectives suggesting that goal-directedness arises not from schedule-level abstraction, but from moment-to-moment fluctuations in internal control dynamics. To validate our behavioural classifications, we analysed response latency. Responses faster than 300ms likely reflected motor artifacts or low cognitive involvement (Foster et al., 2012), while those between 300ms and 1s also lacked reliable markers of volitional preparation in externally triggered behavioural contexts. This supports our latency-based segmentation of meaningful cognitive engagement. At the neural level, these results advance the framework for interpreting trial-to-trial RP variability as an index of intentional goal-related control. Reduced variability reflects a shift toward more stable and selective neural processing (Hancock et al., 2017; Schurger et al., 2015), allowing outcome-relevant information to dominate action selection (Dehaene & Naccache, 2001). Our findings build on the work of Khalighinejad et al. (2018), extending it from isolated voluntary actions to structured sequences of behaviour. However, RP variability alone cannot establish a causal relationship with goal-directedness; rather, it provides a promising window into the neural signatures of volitional, goal-directed control. It remains possible that other factors, such as attentional state or motor preparation demands, contribute to the observed differences. Future research could manipulate RP variability directly, through neurofeedback or brain stimulation to test its causal role in adaptive control. More broadly, our data speak to enduring questions about the nature of goal-directedness. Rather than being a binary feature of behaviours or schedules, goal-directedness appears to be a fluctuating cognitive state, emerging dynamically over time and measurable at the neural level. This aligns with Libet’s (1981) view that intentionality follows rather than precedes movement preparation, and with philosophical models of volition as probabilistic and temporally extended processes resolved only at the point of commitment. In sum, we provide converging behavioural and neural evidence that goal-directedness emerges dynamically after behaviour is already underway, particularly during within-burst responding. By combining operant microstructure analysis with RP modelling, this study offers a refined, temporally sensitive view of intentional goal-directed action, while also providing a principled integrative perspective for linking classical behavioural theory with cutting-edge neuroscientific methods and discoveries. References Chen, X., & Reed, P. (2020). Factors controlling the micro-structure of human free-operant behaviour: Burst-initiation and within-burst responses are effected by different aspects of the schedule. Behavioural Processes , 175 , 104106. Chen, X., & Reed, P. (2023). The effect of brief mindfulness training on the micro-structure of human free-operant responding: Mindfulness affects stimulus-driven responding. Journal of Behaviour Therapy and Experimental Psychiatry , 79 , 101821. Chen, X., Osborne, L. A., & Reed, P. (2020). Role of psychopathic personality traits on the micro-structure of free-operant responding: impacts on goal-directed but not stimulus drive responses in extinction. Personality and Individual Differences , 163 , 110055. Chen, X., & Reed, P. (2023). The effect of brief mindfulness training on the micro-structure of human free-operant responding: Mindfulness affects stimulus-driven responding. Journal of Behavior Therapy and Experimental Psychiatry , 79 , 101821. Chen, X., Zhang, D., & Reed, P. (2025). Microstructural response patterns on schedules of reinforcement predict neural signature of goal-directedness. Under review Dehaene, S., & Changeux, J.P. (2011). Experimental and theoretical approaches to conscious processing. Neuron , 70(2) , 200-227. Dehaene, S., & Naccache, L. (2001). Towards a cognitive neuroscience of consciousness: basic evidence and a workspace framework. Cognition , 79(1-2) , 1-37. Dickinson, A. (1985). Actions and habits: the development of behavioural autonomy. Philosophical Transactions of the Royal Society of London. B, Biological Sciences , 308(1135) , 67-78. Dickinson, A., & Balleine, B. (1994). Motivational control of goal-directed action. Animal Learning and Behaviour , 22(1) , 1-1. Falligant, J.M., Hagopian, L.P., & Newland, M.C. (2024). Bursts, pauses, and units of operant performance: A primer. Perspectives on Behavior Science , 47 (3), 643-674. Ferster, C.B., & Skinner, B.F. (1957). Schedules of reinforcement . Foster, J.D., Nuyujukian, P., Freifeld, O., Ryu, S.I., Black, M.J., & Shenoy, K.V. (2012). A framework for relating neural activity to freely moving behaviour. In 2012 Annual International Conference of the IEEE Engineering in Medicine and Biology Society (pp. 2736-2739). IEEE. Haggard, P. (2008). Human volition: Towards a neuroscience of will. Nature Reviews Neuroscience, 9(12), 934–946. Hancock, R., Pugh, K.R., & Hoeft, F. (2017). Neural noise hypothesis of developmental dyslexia. Trends in Cognitive Sciences , 21(6) , 434-448. Garr, E., & Delamater, A.R. (2019). Exploring the relationship between actions, habits, and automaticity in an action sequence task. Learning & Memory , 26 (4), 128-132. Khalighinejad, N., Schurger, A., Desantis, A., Zmigrod, L., & Haggard, P. (2018). Precursor processes of human self-initiated actions. Neuroimage , 165 , 35-47. Killeen, P.R. (2023). Theory of reinforcement schedules. Journal of the Experimental Analysis of Behavior , 120 (3), 289-319. Libet, B. (1982). Brain stimulation in the study of neuronal functions for conscious sensory experiences. Human Neurobiology , 1 , 235-42. Libet, B., Wright Jr, E. W., & Gleason, C.A. (1983). Preparation-or intention-to-act, in relation to pre-event potentials recorded at the vertex. Electroencephalography and clinical, Neurophysiology , 56(4) , 367-372. Perez, O. D., & Dickinson, A. (2020). A theory of actions and habits: The interaction of rate correlation and contiguity systems in free-operant behavior. Psychological review , 127 (6), 945. Pérez, O.D., Aitken, M.R., Zhukovsky, P., Soto, F.A., Urcelay, G.P., & Dickinson, A. (2019). Human instrumental performance in ratio and interval contingencies: A challenge for associative theory. Quarterly Journal of Experimental Psychology , 72 (2), 311-321. Reed, P. (2015). The structure of random ratio responding in humans. Journal of Experimental Psychology: Animal Learning and Cognition , 41 (4), 419. Reed, P. (2025). Reinforcer devaluation effects within-burst responses to a greater degree than burst-initiation responses on human free-operant responding. Under review. Reed, P., Smale, D., Owens, D., & FrRPard, G. (2018). Human performance on random interval schedules. Journal of Experimental Psychology: Animal Learning and Cognition , 44 (3), 309. Schurger, A., Sitt, J.D., & Dehaene, S. (2012). An accumulator model for spontaneous neural activity prior to self-initiated movements. Proceedings of the National Academy of Sciences , 109(42) , E2904 E2913. Schurger, A., Sarigiannidis, I., Naccache, L., Sitt, J.D., & Dehaene, S. (2015). Cortical activity is more stable when sensory stimuli are consciously perceived. Proceedings of the National Academy of Sciences , 112(16) , E2083-E2092. Shull, R L. (2011). Bursts, changeovers, and units of operant behaviour. European Journal of Behaviour Analysis , 12(1) , 49-72. Thrailkill, E.A., & Burston, M.E. (2015). Contextual control of instrumental actions and habits. Journal of Experimental Psychology: Animal Learning and Cognition , 41 (1), 69. Thrailkill, E.A., Trask, S., Vidal, P., Alcalá, J. A., & Burston, M. E. (2018). Stimulus control of actions and habits: A role for reinforcer predictability and attention in the development of habitual behavior. Journal of Experimental Psychology: Animal Learning and Cognition , 44 (4), 370. Zeiler, M.D. (1984). The sleeping giant: Reinforcement schedules. Journal of the Experimental Analysis of Behavior , 42(3) , 485-493. Figure 1. (A) Left: Positions of all 64 electrodes, with the locations of the five selected electrodes highlighted by a black square. Right, upper panel: Task structure for a single RR–yoked RI–externally triggered section pair. Right, lower panel: Timeline of the overall experimental procedure, comprising four RR–yoked RI–externally triggered section pairs, with the first pair serving as the training section. (B) Group-mean overall response rates for RR and RI schedules. (C) Group-mean response rates for burst-initiation responses. (D) Group-mean response rates for within-burst responses. Error bars indicate 95% confidence intervals. Figure 2. RP activity prior to responses. (A) Grand-average RP amplitude for RR and RI schedules. (B) Trial-to-trial RP variability (SD) for RR and RI schedules. (C) RP amplitude for each channel for RR and RI schedules. (D) Trial-to-trial RP variability (SD) across participants for each channel for RR and RI schedules. (E) Grand-average RP amplitude for each type of micro-response. (F) Trial-to-trial RP variability (SD) across participants for each type of response. (G) RP amplitude for each channel for each type of response. (H) Trial-to-trial RP variability (SD) across participants for each channel for each type of response. The x-axis represents time (s) for all figures. Graded error boundaries indicate 95% confidence intervals. The pink line represents burst-initiation responses, and the cyan line represents within-burst responses. Shaded grey areas denote significant differences (p < .01, FDR-corrected). Figure 3. (A) Empirical trial-to-trial RP variability (SD) across trials for RR and RI responses. (B) Simulated trial-to-trial RP variability (SD) across trials. (C) Optimal parameters of the leaky model for RI and RR conditions. (D) Mean observed trial-to-trial RP variability (SD) across trials for within-burst and burst-initiation responses. (E) Simulated trial-to-trial RP variability (SD) across trials. (F) Optimal parameters of the leaky model for burst-initiation and within-burst responses. The x-axis represents time (s) for all figures. Graded error boundaries indicate 95% confidence intervals. Figure 4. (A) Grand-average RP amplitude for responses with inter-response intervals greater than 0.3 s but less than 1 s, and for responses with inter-response intervals greater than 1 s, in the externally triggered condition. (B) Trial-to-trial RP variability (SD) across participants for each type of response. (C) Grand-average RP amplitude for responses with inter-response intervals greater than 0.3 s but less than 1 s, and for responses with inter-response intervals greater than 1 s, in the externally triggered condition. (D) Trial-to-trial RP variability (SD) across participants for each type of response. The deep green line represents responses with longer intervals, and the light green line represents responses with shorter intervals. The x-axis represents time (s) for all figures. Shaded grey areas denote significant differences (p < .01, FDR-corrected). Information & Authors Information Version history V1 Version 1 31 December 2025 Peer review timeline Published Psychophysiology Version of Record 25 May 2026 Published Copyright This work is licensed under a Non Exclusive No Reuse License. Authors Affiliations Xiaosheng Chen Tsinghua University Department of Psychology View all articles by this author Dan Zhang [email protected] Tsinghua University Department of Psychology View all articles by this author Phil Reed Swansea University School of Psychology View all articles by this author Metrics & Citations Metrics Article Usage 171 views 77 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Xiaosheng Chen, Dan Zhang, Phil Reed. Precursor Processes of Human Schedule Responding: Goal-Directedness Emerges After Burst Initiation. Authorea . 31 December 2025. DOI: https://doi.org/10.22541/au.176718077.78729703/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.176718077.78729703/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a026d89ccd8f52ad',t:'MTc3OTkwNDA2MA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();