Exploring weight initialization, diversity of solutions, and degradation in recurrent neural networks trained for temporal and decision-making tasks.

Recurrent Neural Networks (RNNs) are frequently used to model aspects of brain function and structure. In this work, we trained small fully-connected RNNs to perform temporal and flow control tasks with time-varying stimuli. Our results show that different RNNs can solve the same task by converging to different underlying dynamics and also how the performance gracefully degrades as either network size is decreased, interval duration is increased, or connectivity damage is induced. For the considered tasks, we explored how robust the network obtained after training can be according to task parameterization. In the process, we developed a framework that can be useful to parameterize other tasks of interest in computational neuroscience. Our results are useful to quantify different aspects of the models, which are normally used as black boxes and need to be understood in order to model the biological response of cerebral cortex areas.

Longitudinal analysis of social isolation effects on finger tapping in the Blursday database.

The Blursday database is a collection of data obtained online from a longitudinal study where participants were asked to participate in several behavioral tasks and questionnaires during the COVID-19 pandemic from their homes. In this study, we analyzed the published data to explore (1) the longitudinal changes in temporal cognition observed from the data collected in the home-based setting (2), the effects of the voluntary quarantine measures implemented in Japan on temporal cognition, (3) whether the participant's temporal cognition is altered by the change in their psychological state or their cognitive abilities, and (4) whether the effects of the quarantine measures depend on the age of the individual. Results show that confinement measures were good predictors for the performance in both spontaneous finger-tapping task and paced finger-tapping task, though these were dependent on the age of the participant. In addition, cognitive scores were good predictors of the performance in the paced finger-tapping task but not the spontaneous finger-tapping task. Overall, this study provides evidence suggesting changes in both psychological, cognitive, and temporal cognition during the pandemic on the Japanese population despite its voluntary measures to deal with the new situation.

Paradoxical self-sustained dynamics emerge from orchestrated excitatory and inhibitory homeostatic plasticity rules.

Self-sustained neural activity maintained through local recurrent connections is of fundamental importance to cortical function. Converging theoretical and experimental evidence indicates that cortical circuits generating self-sustained dynamics operate in an inhibition-stabilized regime. Theoretical work has established that four sets of weights (WE←E, WE←I, WI←E, and WI←I) must obey specific relationships to produce inhibition-stabilized dynamics, but it is not known how the brain can appropriately set the values of all four weight classes in an unsupervised manner to be in the inhibition-stabilized regime. We prove that standard homeostatic plasticity rules are generally unable to generate inhibition-stabilized dynamics and that their instability is caused by a signature property of inhibition-stabilized networks: the paradoxical effect. In contrast, we show that a family of "cross-homeostatic" rules overcome the paradoxical effect and robustly lead to the emergence of stable dynamics. This work provides a model of how-beginning from a silent network-self-sustained inhibition-stabilized dynamics can emerge from learning rules governing all four synaptic weight classes in an orchestrated manner.

The Blursday database as a resource to study subjective temporalities during COVID-19.

The COVID-19 pandemic and associated lockdowns triggered worldwide changes in the daily routines of human experience. The Blursday database provides repeated measures of subjective time and related processes from participants in nine countries tested on 14 questionnaires and 15 behavioural tasks during the COVID-19 pandemic. A total of 2,840 participants completed at least one task, and 439 participants completed all tasks in the first session. The database and all data collection tools are accessible to researchers for studying the effects of social isolation on temporal information processing, time perspective, decision-making, sleep, metacognition, attention, memory, self-perception and mindfulness. Blursday includes quantitative statistics such as sleep patterns, personality traits, psychological well-being and lockdown indices. The database provides quantitative insights on the effects of lockdown (stringency and mobility) and subjective confinement on time perception (duration, passage of time and temporal distances). Perceived isolation affects time perception, and we report an inter-individual central tendency effect in retrospective duration estimation.

Time-oriented attention improves accuracy in a paced finger-tapping task.

Finger tapping is a task widely used in a variety of experimental paradigms, in particular to understand sensorimotor synchronization and time processing in the range of hundreds of milliseconds (millisecond timing). Normally, subjects do not receive any instruction about what to attend to and the results are seldom interpreted taking into account the possible effects of attention. In this work we show that attention can be oriented to the purely temporal aspects of a paced finger-tapping task and that it affects performance. Specifically, time-oriented attention improves the accuracy in paced finger tapping and it also increases the resynchronization efficiency after a period perturbation. We use two markers of the attention level: auditory ERPs and subjective report of the mental workload. In addition, we propose a novel algorithm to separate the auditory, stimulus-related components from the somatosensory, response-related ones, which are naturally overlapped in the recorded EEG.

Spatiotemporal perturbations in paced finger tapping suggest a common mechanism for the processing of time errors.

Paced finger tapping is a sensorimotor synchronization task where a subject has to keep pace with a metronome while the time differences (asynchronies) between each stimulus and its response are recorded. A usual way to study the underlying error correction mechanism is to perform unexpected temporal perturbations to the stimuli sequence. An overlooked issue is that at the moment of a temporal perturbation two things change: the stimuli period (a parameter) and the asynchrony (a variable). In terms of experimental manipulation, it would be desirable to have separate, independent control of parameter and variable values. In this work we perform paced finger tapping experiments combining simple temporal perturbations (tempo step change) and spatial perturbations with temporal effect (raised or lowered point of contact). In this way we decouple the parameter-and-variable confounding, performing novel perturbations where either the parameter or the variable changes. Our results show nonlinear features like asymmetry and are compatible with a common error correction mechanism for all types of asynchronies. We suggest taking this confounding into account when analyzing perturbations of any kind in finger tapping tasks but also in other areas of sensorimotor synchronization, like music performance experiments and paced walking in gait coordination studies.

Response to perturbations as a built-in feature in a mathematical model for paced finger tapping.

Paced finger tapping is one of the simplest tasks to study sensorimotor synchronization. The subject is instructed to tap in synchrony with a periodic sequence of brief tones, and the time difference (called asynchrony) between each response and the corresponding stimulus is recorded. Despite its simplicity, this task helps to unveil interesting features of the underlying neural system and the error-correction mechanism responsible for synchronization. Perturbation experiments are usually performed to probe the subject's response, for example, in the form of a "step change," i.e., an unexpected change in tempo. The asynchrony is the usual observable in such experiments and it is chosen as the main variable in many mathematical models that attempt to describe the phenomenon. In this work we show that although asynchrony can be perfectly described in operational terms, it is not well defined as a model variable when tempo perturbations are considered. We introduce an alternative variable and a mathematical model that intrinsically takes into account the perturbation and make theoretical predictions about the response to novel perturbations based on the geometrical organization of the trajectories in phase space. Our proposal is relevant to understand interpersonal synchronization and the synchronization to nonperiodic stimuli.

The Times of Our Lives: Interaction Among Different Biological Periodicities.

Environmental cycles on Earth display different periodicities, including daily, tidal or annual time scales. Virtually all living organisms have developed temporal mechanisms to adapt to such changes in environmental conditions. These biological timing structures-ranging from microsecond to seasonal timing-may have intrinsic properties and even different clock machinery. However, interaction among these temporal systems may present evolutionary advantages, for example, when species are exposed to changing climatic conditions or different geographic locations. Here, we present and discuss a model that accounts for the circadian regulation of both ultradian (less than 24-h) and infradian (more than 24-h) cycles and for the interaction among the three time scales. We show two clear examples of such interaction: (i) between the circadian clock and the seasonal regulation of the Hypothalamic-Pituitary-Thyroid (HPT) axis; and (ii) between the circadian clock and the hypothalamic-nigrostriatal (HNS) ultradian modulation. This remarkable interplay among the otherwise considered isolated rhythms has been demonstrated to exist in diverse organisms, suggesting an adaptive advantage of multiple scales of biological timing.

Sensorimotor synchronization: neurophysiological markers of the asynchrony in a finger-tapping task.

Sensorimotor synchronization (SMS) is a form of referential behavior in which an action is coordinated with a predictable external stimulus. The neural bases of the synchronization ability remain unknown, even in the simpler, paradigmatic task of finger tapping to a metronome. In this task the subject is instructed to tap in synchrony with a periodic sequence of brief tones, and the time difference between each response and the corresponding stimulus tone (asynchrony) is recorded. We make a step towards the identification of the neurophysiological markers of SMS by recording high-density EEG event-related potentials and the concurrent behavioral response-stimulus asynchronies during an isochronous paced finger-tapping task. Using principal component analysis, we found an asymmetry between the traces for advanced and delayed responses to the stimulus, in accordance with previous behavioral observations from perturbation studies. We also found that the amplitude of the second component encodes the higher-level percept of asynchrony 100 ms after the current stimulus. Furthermore, its amplitude predicts the asynchrony of the next step, past 300 ms from the previous stimulus, independently of the period length. Moreover, the neurophysiological processing of synchronization errors is performed within a fixed-duration interval after the stimulus. Our results suggest that the correction of a large asynchrony in a periodic task and the recovery of synchrony after a perturbation could be driven by similar neural processes.

Robust timing and motor patterns by taming chaos in recurrent neural networks.

The brain's ability to tell time and produce complex spatiotemporal motor patterns is critical for anticipating the next ring of a telephone or playing a musical instrument. One class of models proposes that these abilities emerge from dynamically changing patterns of neural activity generated in recurrent neural networks. However, the relevant dynamic regimes of recurrent networks are highly sensitive to noise; that is, chaotic. We developed a firing rate model that tells time on the order of seconds and generates complex spatiotemporal patterns in the presence of high levels of noise. This is achieved through the tuning of the recurrent connections. The network operates in a dynamic regime that exhibits coexisting chaotic and locally stable trajectories. These stable patterns function as 'dynamic attractors' and provide a feature that is characteristic of biological systems: the ability to 'return' to the pattern being generated in the face of perturbations.

Small perturbations in a finger-tapping task reveal inherent nonlinearities of the underlying error correction mechanism.

Time processing in the few hundred milliseconds range is involved in the human skill of sensorimotor synchronization, like playing music in an ensemble or finger tapping to an external beat. In finger tapping, a mechanistic explanation in biologically plausible terms of how the brain achieves synchronization is still missing despite considerable research. In this work we show that nonlinear effects are important for the recovery of synchronization following a perturbation (a step change in stimulus period), even for perturbation magnitudes smaller than 10% of the period, which is well below the amount of perturbation needed to evoke other nonlinear effects like saturation. We build a nonlinear mathematical model for the error correction mechanism and test its predictions, and further propose a framework that allows us to unify the description of the three common types of perturbations. While previous authors have used two different model mechanisms for fitting different perturbation types, or have fitted different parameter value sets for different perturbation magnitudes, we propose the first unified description of the behavior following all perturbation types and magnitudes as the dynamical response of a compound model with fixed terms and a single set of parameter values.

Learning of temporal motor patterns: an analysis of continuous versus reset timing.

Our ability to generate well-timed sequences of movements is critical to an array of behaviors, including the ability to play a musical instrument or a video game. Here we address two questions relating to timing with the goal of better understanding the neural mechanisms underlying temporal processing. First, how does accuracy and variance change over the course of learning of complex spatiotemporal patterns? Second, is the timing of sequential responses most consistent with starting and stopping an internal timer at each interval or with continuous timing? To address these questions we used a psychophysical task in which subjects learned to reproduce a sequence of finger taps in the correct order and at the correct times - much like playing a melody at the piano. This task allowed us to calculate the variance of the responses at different time points using data from the same trials. Our results show that while "standard" Weber's law is clearly violated, variance does increase as a function of time squared, as expected according to the generalized form of Weber's law - which separates the source of variance into time-dependent and time-independent components. Over the course of learning, both the time-independent variance and the coefficient of the time-dependent term decrease. Our analyses also suggest that timing of sequential events does not rely on the resetting of an internal timer at each event. We describe and interpret our results in the context of computer simulations that capture some of our psychophysical findings. Specifically, we show that continuous timing, as opposed to "reset" timing, is consistent with "population clock" models in which timing emerges from the internal dynamics of recurrent neural networks.

Smooth operator: avoidance of subharmonic bifurcations through mechanical mechanisms simplifies song motor control in adult zebra finches.

Like human infants, songbirds acquire their song by imitation and eventually generate sounds that result from complicated neural networks and intrinsically nonlinear physical processes. Signatures of low-dimensional chaos such as subharmonic bifurcations have been reported in adult and developing zebra finch song. Here, we use methods from nonlinear dynamics to test whether adult male zebra finches (Taenopygia guttata) use the intrinsic nonlinear properties of their vocal organ, the syrinx, to insert subharmonic transitions in their song. In contrast to previous data on the basis of spectrographic evidence, we show that subharmonic transitions do not occur in adult song. Subharmonic transitions also do not arise in artificially induced sound in the intact syrinx, but are commonly generated in the excised syrinx. These findings suggest that subharmonic transitions are not used to increase song complexity, and that the brain controls song in a surprisingly smooth control regimen. Fast, smooth changes in acoustic elements can be produced by direct motor control in a stereotyped fashion, which is a more reliable indicator of male fitness than abrupt acoustic changes that do not require similarly precise control. Consistent with this view is the presence of high fidelity at every level of motor control, from telencephalic premotor areas to superfast syringeal muscles.

Population clocks: motor timing with neural dynamics.

An understanding of sensory and motor processing will require elucidation of the mechanisms by which the brain tells time. Open questions relate to whether timing relies on dedicated or intrinsic mechanisms and whether distinct mechanisms underlie timing across scales and modalities. Although experimental and theoretical studies support the notion that neural circuits are intrinsically capable of sensory timing on short scales, few general models of motor timing have been proposed. For one class of models, population clocks, it is proposed that time is encoded in the time-varying patterns of activity of a population of neurons. We argue that population clocks emerge from the internal dynamics of recurrently connected networks, are biologically realistic and account for many aspects of motor timing.

Bilateral source acoustic interaction in a syrinx model of an oscine bird.

The bilateral simultaneous generation of sound in some oscine songbirds leads to complex sounds that cannot be described in terms of a superposition of the isolated sources alone. In this work, we study the appearance of complex solutions in a model for the acoustic interaction between the two sound sources in birdsong. The origin of these complex oscillations can be traced to the nonlinear mode-mode interaction arising when both sources are active. As an example, we analyze a remarkable sound produced by an oscine songbird and show that the proposed dynamical scenario is compatible with the observed behavior.

Modeling source-source and source-filter acoustic interaction in birdsong.

We present a simple model for birdsong production in Oscine songbirds that allows us to study the acoustic interaction between their two sound sources, as well as the acoustic coupling between sources and vocal tract. This model allows us to study complex phenomena in which the traditionally assumed source-filter separation hypothesis does not hold. We make testable hypotheses about the source of complexity in the song of some birds.

Highly structured duets in the song of the South American Hornero.

The South American Hornero (Furnarius rufus) is a suboscine bird widely known for its mud-made, oven-looking nest. Beyond their architectural skills, the male and female Horneros sing in highly structured duets. The analysis of field recordings reported in this work reveals that as the male increases the note production rate the female responds by switching to different locking states: the ones predicted by the theory of nonlinear forced oscillators. This gives the duet a most appealing rhythm, and unveils the nonlinear nature of the underlying brain activity needed to generate the song.

Neuromuscular control of vocalizations in birdsong: a model.

We present a dynamical model of the processes involved in birdsong production, relating qualitatively its parameters with biological ones. In this way, we intend to unify the activity patterns of the muscles controlling the vocal organ with the resulting vocalization. With relatively simple paths in the parameter space of our model, we reproduce experimental recordings of the Chingolo sparrow (Zonotrichia capensis).

Diversity within a birdsong.

We present a model for the activities of neural circuits in a nucleus found in the brains of songbirds: the robust nucleus of the archistriatum (RA). This is a fore brain song control nucleus responsible for the phasic and precise neural signals driving vocal and respiratory motor neurons during singing. Driving a physical model of the avian vocal organ with the signals generated by the neural model, we produce synthetic songs. This allows us to show that certain connectivity architectures in the RA give rise to a wide range of different vocalizations under simple excitatory instructions.