Under the joint supervision of Assistant Professor Zhang Jianfeng and Professor Chen Qi of the School of Psychology, postdoctoral researcher Long Zhengkun has published a paper online in Communications Biology entitled “Temporal signatures of thought—Neurodynamics distinguish on- and off-task thoughts.” The study examined differences in neural dynamics between on-task thoughts and off-task thoughts, the latter referring to mind-wandering or task-unrelated thought. By focusing on the temporal dynamic properties of brain activity, the research identified temporal signatures associated with different thought states, thereby deepening understanding of the neural mechanisms underlying spontaneous human thought.
1.Research Background
Human thought is highly dynamic. Our minds often drift spontaneously away from the task at hand toward task-unrelated content, a common phenomenon referred to as mind wandering or off-task thought. Research has shown that people spend roughly 30% of their waking lives in a state of mind wandering. More than a century ago, psychologist William James compared thought to the life of a bird—alternating between flight and perch—capturing with remarkable insight the dynamic nature of human thinking. Yet the neural dynamic features that drive continuous fluctuations between task-related thought and mind wandering remain a key unresolved question.
In recent years, researchers have developed a range of neurodynamic indices to characterize the temporal patterns of brain signals. These include the autocorrelation window (ACW), which reflects temporal continuity; Lempel–Ziv complexity (LZC), which indexes informational complexity; the power-law exponent (PLE), which captures the nesting relationship between fast and slow frequencies; and median frequency (MF), which reflects the balance of frequency components. These measures have been widely applied in research on cognition related to consciousness, the self, and attention. However, their relationship to thought states—particularly on-task thought and mind wandering—has remained largely unexplored. Against this background, the present study aimed to use multiple electroencephalography (EEG)-based neurodynamic indices to systematically examine differences in the temporal dynamics of the brain between on-task thought and mind wandering, identify the temporal signatures of distinct thought states, and further explore the hierarchical organization of neurodynamic indices across different timescales.
2.Methods
The research team recruited 29 right-handed university students and simultaneously recorded 64-channel electroencephalography (EEG) data and behavioral responses while they performed a sustained motor task. The experiment used a block design combined with thought sampling. At the end of each block, participants responded to a thought probe indicating whether their attentional state during the preceding period had been characterized by on-task thought or mind wandering.
The analyses were organized at two temporal levels: the block level, based on 17 seconds of continuous EEG data preceding each thought probe, and the trial level, based on 3 seconds of EEG data spanning the pre- and post-stimulus period. These two levels were taken to correspond to the “background” and “foreground” layers of thought, respectively. At the block level, the researchers calculated the power-law exponent (PLE), reflecting temporal nesting between fast and slow frequencies; median frequency (MF), reflecting the temporal speed of neural activity; autocorrelation window (ACW), reflecting the temporal continuity of the signal; and Lempel–Ziv complexity (LZC), reflecting the temporal compressibility of the signal. At the trial level, ACW and LZC were computed.
Linear mixed-effects models were used to test whether these neurodynamic indices could distinguish between on-task thought and mind wandering, while structural equation modeling was used to examine the unidirectional influence from longer timescales at the background level to shorter timescales at the foreground level. The central hypothesis was that mind wandering operates on longer timescales and therefore should be associated with higher PLE, lower MF, longer ACW, and lower LZC. The researchers further hypothesized that neurodynamic indices across different timescales would exhibit a hierarchical organizational structure extending from background to foreground. In addition, the team conducted a full replication in an independent sample of 30 participants.
Figure 1. (a) Experimental task; (b) data analysis.
3.Results
(1) Compared with on-task thought, mind wandering was associated with a higher power-law exponent (PLE) and a lower median frequency (MF), indicating a greater dominance of low-frequency components in neural activity. PLE reflects the nesting relationship between fast and slow frequencies, whereas MF indexes the overall temporal speed of neural activity. Taken together, these findings suggest that mind wandering represents a temporally “slower” form of thought with a higher degree of temporal nesting.
Figure 2. Differences between mind wandering and on-task thought in frequency-domain neurodynamic indices (PLE and MF).
(2)Mind wandering also showed a longer autocorrelation window (ACW) than on-task thought. ACW reflects the duration over which brain signals remain temporally autocorrelated, with a longer ACW indicating that neural activity integrates information over a more extended time window. This finding suggests that mind wandering is accompanied by a more sustained process of temporal integration, consistent with an internally oriented cognitive state marked by greater temporal continuity.
Figure 3. Differences between mind wandering and on-task thought in the autocorrelation window (ACW).
(3)Mind wandering exhibited lower Lempel–Ziv complexity (LZC) than on-task thought. LZC reflects the informational complexity and compressibility of neural signals. This result indicates that mind wandering is associated with more repetitive and compressible patterns of neural activity, whereas on-task thought is characterized by greater informational diversity.
Figure 4. Differences between mind wandering and on-task thought in Lempel–Ziv complexity (LZC).
(4)The study also examined the relationship between neurodynamic indices and behavioral performance. Mind wandering was associated with slower and more variable reaction times, while higher PLE, longer ACW, and higher cvLZC all predicted slower behavioral responses. Mediation analyses further revealed a hierarchical relationship among these indices: ACW influenced behavioral performance indirectly through its effects on LZC and PLE, suggesting that ACW operates at a deeper “background” level, whereas LZC and PLE are more directly linked to “foreground” behavioral execution.
Figure 5. Prediction of behavioral reaction time by neurodynamic indices and their mediation effects.
(5)Differences in the hierarchical background–foreground organization of neurodynamics differentially shaped on-task thought and mind wandering. Structural equation modeling showed that neurodynamic indices at longer timescales (the block level) exerted a unidirectional influence on indices at shorter timescales (the trial level), whereas the reverse relationship was not supported, revealing a hierarchical organization from “background” to “foreground.” Multi-group comparisons further showed that the coupling between background and foreground levels was tighter during on-task thought, but significantly weaker during mind wandering. This suggests that variation in the degree of coupling across the background–foreground hierarchy may represent a key mechanism distinguishing these two thought states.
This finding may also help explain why mind wandering exhibits distinctive neurodynamic features such as higher PLE, longer ACW, and lower LZC. Compared with on-task thought, mind wandering may be less constrained by the short-timescale demands of external tasks and may rely more heavily on longer-timescale “background” dynamics. As a result, it may be more strongly shaped by slow-frequency-dominated background activity, giving rise to temporal patterns that are slower, more sustained, and more compressible in informational terms.
Figure 6. The temporal background–foreground hierarchical structure based on structural equation modeling and its differences between the two thought states.
4.Conclusion
This study systematically revealed marked differences between on-task thought and mind wandering in the temporal characteristics of EEG neurodynamics. Mind wandering was characterized by greater temporal nesting, reflected in higher PLE and lower MF, stronger temporal continuity, reflected in longer ACW, and greater temporal compressibility, reflected in lower LZC. These findings indicate that the two thought states are associated with fundamentally distinct neurodynamic profiles: mind wandering operates on longer and slower timescales, whereas on-task thought unfolds on shorter and faster timescales.
More importantly, the study found that neurodynamic indices across different timescales form a hierarchical organization extending from “background” to “foreground.” Neural dynamics at longer timescales (the block level, 17 seconds) exerted a unidirectional influence on neural dynamics at shorter timescales (the trial level, 3 seconds), rather than the reverse. This background–foreground structure manifested differently across the two thought states. During on-task thought, coupling between long and short timescales was tighter, whereas during mind wandering, this coupling was looser. Such reduced coupling may allow internally oriented mental activity to become more dominant, thereby facilitating the occurrence of mind wandering.
These findings provide empirical support for the recently proposed Dynamic Layer Model of Brain (DLB) and suggest that timescale differences at the neurodynamic level may be reflected at the psychological level in the form of a “common currency.” In other words, the temporal characteristics of thought may serve as a bridge linking neural activity with subjective mental experience.
5.Author Contributions
Postdoctoral researcher Long Zhengkun of the School of Psychology at Shenzhen University is the first author of the paper. Professor Fu Xiaolan of Shanghai Jiao Tong University is a co-author. Professor Chen Qi and Assistant Professor Zhang Jianfeng of the School of Psychology at Shenzhen University, together with Professor Georg Northoff of the University of Ottawa, are the corresponding authors. This research was supported by the National Key Research and Development Program of China (2024YFC3606800), the STI2030—Major Projects initiative (Neural Circuit Mechanisms and Neurocomputational Models of Attention, 2021ZD0203800), and the National Natural Science Foundation of China (32571283, 32201129), among other funding sources.
6.References
Long, Z., Fu, X., Chen, Q., Zhang, J., & Northoff, G. (2026). Temporal signatures of thought—Neurodynamics distinguish on- and off-task thoughts. Communications Biology, 9(1), 437. https://doi.org/10.1038/s42003-026-09715-7
