Behavior Research Methods

Engagement is widely recognised as central to learning and academic achievement. Electrodermal activity (EDA) has emerged as an objective physiological indicator of engagement, as it measures sympathetic nervous system activation. However, the high cost of wearable EDA sensors has limited its widespread application. This study answers the call for affordable, high-temporal-resolution engagement measures by validating a video-based quantitative assessment method. Researchers collected 75 minutes of synchronised EDA and video data from 12 upper secondary students (aged 17-18) during regular instruction. Novel software was developed to analyse student movement and sound level for academically relevant content. The OpenPose AI model for pose estimation was also applied. This approach produced six distinct movement variables: two AI-based and four non-AI-based. Six linear models using varying movement variables and sound level were tested to predict tonic EDA levels. All models effectively predicted EDA levels, with non-AI-based movement metrics outperforming AI-based alternatives. The four non-AI-based movement models showed similar performance, indicating that compressed versions reduced computational time without sacrificing predictive power. These findings validate a novel, objective method for comparing engagement across learning activities on short timescales. This method is particularly useful for collaborative learning environments and enables controlling for movement and sound in quantitative classroom analyses.

Singing is an innate human behaviour present across cultures and the lifespan. Despite lacking direct biological advantages, its ubiquity suggests that it is intrinsically rewarding. This research aimed to investigate the underlying factors that explain variability in sensitivity to deriving reward and enjoyment from natural singing in the general population. In Study 1 (n = 606), an initial pool of items describing daily, non-professional singing behaviours were administered to an international adult sample. Exploratory factor analysis revealed a unidimensional structure of 20 items with acceptable model fit, organized into five facets representing distinct domains of singing-related rewards: 1) pleasure and emotional evocation, 2) social singing reward, 3) singing frequency, 4) mood regulation through singing, and 5) inattentional singing during routine tasks. In Study 2 (n = 430), confirmatory factor analysis in a new sample supported this structure. When both samples were combined (n = 1036), the unidimensional model defined by these five facets showed acceptable to excellent goodness-of-fit indices, supporting the conceptualization of singing reward as a multidimensional construct with differentiated facets. This led to the Barcelona-Aarhus Natural Singing Engagement Questionnaire (BANSEQ), which demonstrated excellent reliability ( = .94) and population-level stability. Study 3 (n = 1036) tested the convergent validity of BANSEQ with measures of music reward and engagement and identified sociodemographic and psychological correlates across the five facets of singing reward. Overall, these findings characterize the sources of individual differences in the hedonic experience of natural singing and propose BANSEQ as a robust psychometric tool for its assessment in the general population.

Mind-wandering (MW) is a frequent and pervasive phenomenon, yet it is commonly assessed using self-reports or probe-based methods that offer limited temporal precision regarding its onset. In this study, we introduce a novel paradigm, ReMind, that estimates the onset and duration of MW episodes during natural reading by combining retrospective self-reports with eye-tracking. Participants indicated the words where they believed their mind started and stopped wandering, and these reports were aligned with gaze timestamps to estimate MW onset. Using data from 44 participants, we examined whether knowledge of MW onset improves the detection of MW from eye-tracking signals. To evaluate relevance for both self-report and thought-probe paradigms, we additionally simulated thought probes by randomly sampling time points during reading. Logistic regression classifiers trained on eye-tracking features extracted from time windows anchored to MW onset achieved AUROC scores of 0.659 and 0.621 under the self-report and simulated thought-probe paradigms, respectively, using leave-one-subject-out cross-validation. In both cases, onset-aligned windows outperformed classifiers trained using arbitrary MW windows. Sliding-window analyses further revealed systematic temporal changes around MW onset, with classification performance peaking at approximately 3 seconds after onset. Feature-level analyses showed reduced fixation rate and fixation dispersion, along with increased pupil size following MW onset. Together, these findings characterize the temporal progression from on-task reading to MW. Overall, ReMind provides a useful framework for studying the temporal dynamics of MW during naturalistic reading.

Dance is a core form of human-environment interaction and a powerful medium for emotional expression, yet dancers are routinely exposed to environmental affective cues that may shape their movement. We tested whether a negative emotional context induced immediately before improvisation alters dance biomechanics. Twenty professional dancers performed two 3-min improvised dances. Between dances, they viewed either Neutral or Negatively valenced pictures from the International Affective Picture System (IAPS; 2 min 40 s, 5 s per image). Eye tracking verified attention to the visual stream. Mood was assessed at four time points (PT1-PT4) using the Brazilian Mood Scale (BRAMS), and full-body, three-dimensional kinematics were captured at 300 Hz using a 9-camera optoelectronic system (Qualisys) and processed to measure global movement amplitude and expansion. Negative IAPS exposure increased tension, depression, fatigue, and decreased vigor from PT2 to PT3. Biomechanically, the Negative Stimulus dancers showed a significant reduction in global movement amplitude after negative IAPS exposure, with reduced movement amplitude of the body extremities. In contrast, global movement expansion remained unchanged; that is, the extremities were not positioned closer or farther from the pelvis. Neutral images produced no mood change and no measurable modulation of movement amplitude or expansion. Together, these results support the hypothesis that improvised dance carries biomechanical signatures of the dancers current affective state, beyond the intended expressive content, and provide an automated motion-capture workflow for studying emotion-movement coupling in spontaneous dance. HighlightsNegative visual context shifted dancers mood toward negative affect Negative images reduced movement amplitude in improvised dance Movement expansion remained stable despite mood induction Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=113 SRC="FIGDIR/small/711707v1_ufig1.gif" ALT="Figure 1"> View larger version (19K): org.highwire.dtl.DTLVardef@aeaacdorg.highwire.dtl.DTLVardef@14f9bf5org.highwire.dtl.DTLVardef@18805fcorg.highwire.dtl.DTLVardef@1411256_HPS_FORMAT_FIGEXP M_FIG C_FIG

Computational bioacoustics has seen significant advances in recent decades. However, the rate of insights from automated analysis of bioacoustic audio lags behind our rate of collecting the data - due to key capacity constraints in data annotation and bioacoustic algorithm development. Gaps in analysis methodology persist: not because they are intractable, but because of resource limitations in the bioacoustics community. To bridge these gaps, we advocate the open science method of data challenges, structured as public contests. We conducted a bioacoustics data challenge named BioDCASE, within the format of an existing event (DCASE). In this work we report on the procedures needed to select and then conduct useful bioacoustics data challenges. We consider aspects of task design such as dataset curation, annotation, and evaluation metrics. We report the three tasks included in BioDCASE 2025 and the resulting progress made. Based on this we make recommendations for open community initiatives in computational bioacoustics.

Empathy, a key element of social interaction, involves both cognitive and affective processes and is commonly investigated through measures such as empathic accuracy and affective physiological synchrony. While physiological synchrony offers a continuous measure of affective processes, empathic accuracy typically relies on discrete self-reports, leaving their temporal relationship largely unexplored. Advancing this line of research requires datasets that integrate time-continuous self-reports with physiological signals, yet such datasets--particularly those focusing on the empathizee--remain limited. To fill this gap, we present EMPAC (Empathy Measurement: Physiological, Affective, and Cognitive), a multimodal dataset constructed. To create empathy-eliciting stimuli, professional actors performed emotionally intense, pseudo-autobiographical narratives while their physiological signals (e.g., ECG, EDA) and continuous self-reported emotional states were recorded. We then conducted two observer experiments using these video recordings. In Experiment 1, to validate the stimuli as empathy-eliciting materials, observers continuously rated emotional intensity without being informed of the specific emotion portrayed, following the protocol of previous studies on time-series empathic accuracy. Yet this approach sometimes revealed a gap between the emotion category portrayed by the target and that perceived by the observers. In Experiment 2, we introduced a revised procedure in which the target emotion category was disclosed prior to viewing, revealing that specifying the target emotion led to a different relationship between individual empathy traits and empathic accuracy than observed in Experiment 1. EMPAC thus provides a rich, temporally aligned resource for investigating empathy dynamics in naturalistic settings and for evaluating methodological variations in empathic accuracy paradigms.

PurposeWe present a large-scale (N=120) comparative study of gel-based and dry electroencephalography systems for cognitive load analysis in tasks involving information visualization stimuli. Although dry systems are increasingly adopted owing to their portability and fast setup, their sensitivity to cognitive-related measurements (as compared to gel-based systems) remains debated. This limits the understanding of whether dry systems provide sufficient sensitivity for cognitive load assessment under controlled task conditions. MethodsWe analyzed a diverse set of signal quality metrics, such as signal-to-noise ratio and channel retention, combined with spectral features across frequency bands to evaluate the ability for each device to capture workload-related neural markers during information visualization tasks. ResultsAlthough the gel-based device showed consistently better quality results than the dry one, the effect sizes suggest a small practical significance of the differences between systems. These results demonstrate that dry systems can provide adequate physiological sensitivity for cognitive load assessments. ConclusionOur findings highlight the trade-off between usability (setup, calibration, etc.) and data fidelity, providing practical guidance for choosing electroencephalography systems for cognitive workload monitoring and applied neuroengineering research. Overall, the results suggest that dry systems can support coarse-grained cognitive load assessment, while gel-based systems remain advantageous when greater sensitivity is required.

Animal welfare is a central aspect in animal-based research where mice are most commonly used. Their facial expression can be analyzed to assess their well-being status using the Mouse Grimace Scale. However, its manual application becomes increasingly impractical when used on a large number of animals. This lead to the ongoing integration of computer vision methods to automate the analysis. While such methods have proven effective qualitatively, a systematic assessment to verify their reliability largely remains an open research gap. In this work, we attempted to close this gap as we evaluated three dominant paradigms (i.e., classification from supervised learning features, self-supervised learning features, or landmark locations) for the binary (i.e., well-being un-/impaired) classification of facial mouse images. Our quantitative results showed that such methods can be employed successfully with as low as 16% type II error rates. For qualitative assessment, we visualized the decision-making process and demonstrated that mainly pixels associated with the mouse rather than its environment are used. We further discovered that visual characteristics of the mice beyond those described by the Mouse Grimace Scale contributed to the classification. Our work showed that the automated well-being status assessment in mice is trustworthy and urges towards widespread adoption.

Lucid dreaming (LD), during which the dreamer is aware that they are dreaming, is frequently induced in laboratory settings by delivering sensory cues during rapid eye movement (REM) sleep. These cues should be incorporated into ongoing dreams and can trigger reflective awareness. This approach relies on the continuity between waking experiences and dream content. In sleep laboratories, participants often dream of the experimental setting itself (lab dreaming), providing a predictable context in which lucidity may emerge. The present studies leveraged this phenomenon by explicitly training participants to associate the sleep laboratory with reflective awareness prior to sleep. Across three studies (total N = 101), participants completed a morning nap following verbal LD instructions and presleep audio designed to prime recognition of the laboratory context in dreams. In addition, conditions included immersive virtual reality (VR) rehearsal of the laboratory environment, VR combined with haptic stimulation (HS) during REM sleep, or VR containing subtle fake system errors intended to prompt reflective checking. LD frequency was assessed through external ratings of signal-verified LD (SVLD) dream reports. Lucidity rates were high across all conditions, with approximately 40-45% of dreams externally rated as lucid and 11%-32% SVLDs occurring in every group. However, neither VR rehearsal, haptic stimulation, nor implicit VR errors increased lucidity relative to the baseline laboratory induction procedure. Exploratory analyses investigated the overlap between laboratory dreaming, false awakenings (FAs), and lucidity. These findings suggest that explicit training focused on the predictable context of the sleep laboratory may already provide a powerful pathway to lucidity, with additional technological manipulations offering limited benefit under a single-nap protocol. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=140 SRC="FIGDIR/small/711049v1_ufig1.gif" ALT="Figure 1"> View larger version (47K): org.highwire.dtl.DTLVardef@191373corg.highwire.dtl.DTLVardef@c1490corg.highwire.dtl.DTLVardef@1a2c193org.highwire.dtl.DTLVardef@52c5d1_HPS_FORMAT_FIGEXP M_FIG C_FIG

Zebrafish (Danio rerio) are an important vertebrate model for vision and neuroscience research. In the larval stages, the aquatic species begins to elicit the optomotor response (OMR) to stabilize themselves in water -- a behaviour that may be exploited in the laboratory to measure visual acuity. However, up to now, the measurement of the OMR in juvenile and adult zebrafish has been limited due to their behavioural complexity. Here, we optimize a protocol to assay zebrafish aged between 4 and 9 weeks-post-fertilization, by displaying sinusoidal gratings parallel to the zebrafish eye to elicit a robust OMR. We assessed the visual spatial-frequency tuning function of an environmentally induced myopia model to confirm the sensitivity and robustness of the protocol. Additionally, we show the OMR is sensitive to the contrast and temporal resolution of the sinusoidal gratings. Furthermore, we found that the time between stimulus presentations impact the spatial-frequency tuning function likely as time is required for zebrafish to return to baseline swimming after eliciting the OMR. Finally, we found that the OMR after ten versus twenty seconds of stimulus onset appears comparable; indicating that robust OMR responses in zebrafish can be elicited through relatively short stimulus presentations. Through the experiments conducted, we present an optimized protocol specific to zebrafish. The protocol may be used to follow the progression or treatment efficacy of progressive neurological disorders including specific visual disorders and higher brain functions with visual endophenotypes. Ultimately, this protocol allows for high-throughput robust measures of visual and neural function in zebrafish.

The replicability of experimental results is considered a cornerstone of empirical research. However, the reproducibility of results from groups that have not undergone experimental manipulation -- the standard against which comparisons in an experiment are made -- has been almost entirely neglected in animal research. Our aim is to address this gap by exemplarily determining within-laboratory replicability using research in pigs, an increasingly popular large animal model species. Drawing on data from twelve independent porcine holeboard studies conducted in our laboratory, we examine the replicability of groups that were not subjected to experimental manipulation (typically the control group), eventually including groups on which the experimental treatments had no effect. These analyses were also performed on all eight studies involving the Terra x Finnish Landrace x Duroc pig breed, with all other breeds excluded to increase genetic uniformity. The holeboard is a complex spatial discrimination task in which an animal must learn to find food at four of sixteen possible locations (holes) arranged in a 4 x 4 matrix. The main variables measured are spatial working memory (WM), reference memory (RM) and the inter-visit interval (IVI), which serves as an index of motivation. All studies showed predominantly linear improvements in WM and learning rates across successive trial blocks. IVI showed greater variation across trialblocks, but this did not affect WM and RM learning, which are robust and replicable indices of spatial learning in pigs. Assessing replicability provides relevant information, such as whether behavioural and physiological traits in a model species are stably expressed and robust across studies. Including replicability research and publishing its results can stimulate the development and use of more replicable methods and workflows, thereby increasing scientific rigour. Provided the data are available and accessible, the next step should be to expand replicability studies to include those conducted in different laboratories.

Virtual Reality (VR) applications depend on eliciting spatial presence, the subjective experience of being physically located within a virtual environment. Although individual differences have long been theorised to contribute to this experience, their role in highly immersive VR systems remains contested. The present study investigated whether trait absorption predicts spatial presence and whether this relationship is mediated by attention allocation. Seventy participants (44 female, 26 male; M age = 22.90, SD = 4.88) completed a 6-minute VR session using a Meta Quest 3 Head-Mounted Display and validated self-report measures of trait absorption (Tellegen Absorption Scale), attention allocation, and spatial presence (MEC-Spatial Presence Questionnaire). Path analysis confirmed a significant, complete mediation pathway: trait absorption positively predicted attention allocation ({beta} = 0.27, p = .013), which in turn strongly predicted spatial presence ({beta} = 0.54, p < .001). The direct path from absorption to spatial presence was non-significant ({beta} = 0.11, p = .325), indicating complete mediation. The indirect effect was significant ({beta} = 0.15; 95% BCa CI [0.025, 0.291]). The model explained a sizeable 33.8% of the variance in spatial presence (Cohens f{superscript 2} = 0.51). Post-hoc dose-response analysis revealed that trait absorption acts as a cognitive amplifier: the strength of the attention-presence relationship tripled from low-absorption ({beta} = 0.33, R{superscript 2} = .15) to high-absorption individuals ({beta} = 1.00, R{superscript 2} = .56). These findings demonstrate that individual differences remain important in highly immersive VR by modulating the effectiveness of attentional focus, offering promising directions for tailoring VR interventions.

Summary paragraphA hallmark of learning is the need for sensory stimuli (Ginns, 2015; McGraw et al., 2009; Reinwein, 2012; Spence, 1950) so that learning is fundamentally based on sensory input signals affecting behaviour, physiology, and neurology. If behavioural measures of learning can be causally linked to physiological and neurological variables, a broader understanding of the mechanisms related to learning in schools, learning disabilities, and learning and health issues may emerge (McGraw et al., 2009). Despite decades of research on the physiological/neurological variable of sympathetic activation, learning, and achievement (Horvers et al., 2021), any causal relation remains unclear (Cowley et al., 2014; Mason et al., 2020; Pijeira-Diaz et al., 2016; Sung et al., 2023; Yu et al., 2024) and issues with instrument validation remain (Costantini et al., 2023; Hu et al., 2024; Milstein & Gordon, 2020; Van Der Mee et al., 2021). Here we investigate the effect of sensory input on sympathetic activation by using validated instruments for skin conductance measurement (Batista et al., 2019) and whether sympathetic activation is connected to learning in a cognitive laboratory context and an ecologically valid classroom context. In both contexts, we found a physiological variable which correlated with learning and that sensory input affected this variable while student movement did not. These sensory inputs varied depending on the different instructional activities the students participated in. Together, these findings bring us one step closer to a model linking sensory input to behavioural, physiological, and neurological variables.

One of our sensory systems key functions is to detect threats in the environment. Sensory information eliciting negative emotions, such as fear or disgust, triggers instinctive avoidance reactions. This core survival mechanism is believed to be expressed as subtle non-conscious postural reactions, even when participants are instructed to stand still. Such avoidance behavior has mainly been studied using indirect measures that make participants aware of their posture (e.g. force-plate based methods) or measures that depend on explicit cognitive tasks, like moving a joystick to indicate an urge to approach or avoid the stimulus; experimental tasks with limited ecological validity and generalizability. Therefore, despite the importance of this basic survival strategy, its underlying mechanisms are still poorly understood. Here, we used a novel 3D-camera-based method allowing direct but implicit measures of postural reactions with high precision. Participants are aware that they are being filmed but, crucially, are not informed that distance measures are obtained. We assessed this ecologically valid measure of approach/avoidance responses in two different sensory modalities: olfaction and vision. Participants were standing upright while exposed to either olfactory or visual stimuli and verbally rating their perceived valence in each trial. In response to subjectively unpleasant odors and images, participants moved away from the stimulus source, as compared to pleasant stimuli. These results demonstrate a putative modality-independent early proxy for avoidance behavior in response to perceived negative valence. Considering its face validity and general applicability, this novel experimental method presents new possibilities for assessing non-conscious approach-avoidance responses in humans.

Virtual reality (VR) offers immersive and interactive learning environments that can improve student engagement and 3D visualization. However, its application in medical education is mostly limited to clinical settings and its potential for better understanding complex concepts, or empathy with the patients, remains underexplored. Here, we describe SIGHT (Simulated Immersive Guidance for Human Training), an immersive VR application, designed to teach core concepts in the physics and functional neuroanatomy, or neurophysiology of human vision. Its two integrated learning modules allow first-person experience of normal and pathological conditions: the optics module enables users to manipulate lenses, experience refractive errors such as myopia, presbyopia, and astigmatism and correct them through appropriate lens selection; the neurophysiology module allows learners to navigate the visual pathways from the retina to the visual cortex and to simulate lesions, experiencing the corresponding visual field deficits. User authentication and interactive evaluation steps provide analytical feedback of the experience and learning process. A pilot group of medical students reported high usability, engagement and deeper understanding of the vision-related concepts, showing how the approach of SIGHT can support experiential learning in medical education.

Despite questionable accuracy, subjective methods to categorize skin color are heavily relied upon in research and medicine. Objective skin color determination is expensive requiring specialized instrumentation and interpretation. We compare three subjective approaches, i) Fitzpatrick Skin Type Scale (FST), ii) Pantone SkinTone Guide (PST) and, iii) Monk Skin Tone Scale (MST), with objectively measured skin color from a spectrophotometer in 87 volunteers to understand the limitations of each method. In agreement with others, we show that the popular FST questionnaire correlates poorly with the objective approach. However, PST color swatches provide good correlation with spectrophotometer measurements. PST consists of 110+ swatches that are inexpensive and easy to use, however, similar to other reports, the volunteers found the number of swatches overwhelming and/or excessive. We found that the recently introduced MST is not representative of reality with only 3 of the 10 color groups representing our volunteers and published populations of volunteers. In future, we propose using 9 color swatches to split the spectrum of human skin color into 10 groupings (Nottingham Skin Categories - NSC) that are representative of the global population. This new approach would be easy to implement and inexpensive in research, healthcare and cosmetics settings, and maps directly to objective, quantitative, measures taken with a spectrophotometer. For the testing and development of new optical devices, NSC would provide increased comparability between studies and ensure studies are representative of local/global populations. In the clinic NSC would be useful for dermatology, photodynamic therapy and dosage assessment for topical medicine, for example.

Object-based tasks are widely used in rodent behavioral research, yet object selection remains largely unsystematic. We present a paired-object validation protocol in which objects differ along one researcher-defined feature, allowing assessment of whether that feature is salient to the animal. Using six object pairs varying in height, color, shape, or aperture presence, we tested two wild-caught mice species with contrasting ecologies. Wood mice (Sylvaemus uralensis) and striped field mice (Apodemus agrarius) showed equal preference for both objects in most pairs, indicating that color, apertures, and apex shape differences are not salient under the tested conditions and can be used interchangeably in object recognition tasks. Height, however, produced ecology-predicted responses: arboreal wood mice avoided the shortest object while open-habitat striped field mice did not. These results demonstrate that the protocol successfully detects feature salience when present and that ecological background predicts which features matter. Summary StatementA systematic paired-object protocol reveals that most researcher-defined features (color, holes, shape) do not affect rodent exploration, but height preferences emerge in ecology-predicted patterns, demonstrating that feature salience is species-specific.

Biodesign is an interdisciplinary research domain that incorporates principles from design and the life sciences to develop new systems, processes, and objects. Collegiate biodesign educators face unique pedagogical challenges, including an absence of relevant scholarship on curriculum design and instructional best practices for cultivating student scientific literacy. These difficulties may be overcome with newly available technologies, like generative AI systems, that enable personalized learning through domain-specific semantic spaces. This article examines the instructional value of one such domain-specific LLM, Biodesign Buddy, through a mixed-methods analysis of an eight-week study involving 64 students participating in an international biodesign competition. Results indicate strong support for integrating AI into biodesign coursework. Surveys captured attitudes toward AI, scientific literature, and learning experiences to assess AIs impact on learning outcomes. Findings suggest that integrating AI into biodesign pedagogy can meaningfully redress conceptual issues in biodesign while informing broader debates on AIs role in higher education. Impact StatementThis article introduces Biodesign Buddy, a domain-specific generative AI system for collegiate biodesign education, and reports on its exploratory deployment, offering design principles and preliminary findings to inform the development of AI-supported pedagogies for interdisciplinary biodesign instruction.

This paper presents a six-stage methodological framework for Convolutional Neural Net-work (CNN)-based cetacean vocalization detection and classification in Passive Acoustic Monitoring (PAM), implemented as the open-source toolkit ai-pam-pipeline. The frame-work is generalizable across species and fully parameterised through a single configuration file, guaranteeing exact experimental reproducibility. Two experiments are reported. Experiment A examines the effect of FFT window length Nfft [isin] {256, 512, 1024} on binary Bottlenose dolphin (Tursiops truncatus) whistle detection using stratified 10-fold cross-validation on an in-domain dataset (Oltremare, 192 kHz) and a cross-domain benchmark (DCLDE 2022). In-domain performance is uniformly high (macro F1{approx} 0.98; Wilcoxon, all p > 0.05). Cross-domain results diverge substantially: Nfft = 256 is significantly superior (p = 0.006, rank-biserial r = 0.89). The mechanism is an upsampling amplification effect: coarser spectral bins produce wider, higher-contrast FM traces after bilinear resampling to fixed image dimensions. This superiority is threshold-invariant: precision equals 1.000 across all configurations and thresholds{theta} [isin] [0.1, 0.9], confirming that the advantage is not an artifact of threshold choice. These findings demonstrate that preprocessing choices -- often treated as secondary implementation details -- can significantly affect cross-domain generalisation. While Nfft serves here as a controlled case study, the framework is designed to enable systematic, reproducible evaluation of arbitrary preprocessing parameters within a unified experimental protocol. Experiment B demonstrates multiclass capability on five T. truncatus vocalization cate-gories (macro F1 = 0.843); inter-class confusion between click trains and burst-pulse sounds reflects biological signal overlap rather than classifier failure.

Selective auditory attention decoding (AAD) enables tracking which of multiple concurrent speakers a listener attends to and is a key building block for neuro-steered hearing devices. While AAD integrated in a closed-loop system with real-time neurofeedback (NFB) is hypothesized to improve decoding through neural adaptation and error-correction behaviour, the short-term behavioral and algorithmic impact of such a bilateral human-machine interaction remains poorly understood. Here we evaluated the effects of NFB on AAD accuracy and user experience in a single-session AAD paradigm with online NFB involving nineteen participants. They performed a selective listening task with enforced attention switches across four conditions: open-loop (OL), closed-loop with auditory gain feedback (CLA), closed-loop with visual feedback (CLV), and a condition with pseudo-auditory gain control (psCLA) decoupled from the participants individual neural activity. AAD was performed online using both subject-specific and subject-independent linear decoders on 5 s sliding windows, followed by Hidden Markov Model post-processing. Online analysis showed comparable decoding performance across all conditions. However, offline posthoc analysis using subject-independent decoders revealed that AAD accuracy in the CLA condition was significantly lower than in the OL baseline. Subjectively, participants reported that CLA was significantly more distracting and required higher switching effort. Crucially, a causal analysis of the psCLA condition found no robust evidence that higher audio gains inherently improve decoding accuracy. Our results demonstrate that within a single-session paradigm with rapidly varying feedback cues, auditory neurofeedback may degrade AAD performance by increasing cognitive load and distraction. These findings suggest that suboptimal feedback can impede rather than facilitate learning. We conclude that more accurate and stable decoders and longitudinal, multi-session training protocols are likely essential prerequisites for achieving beneficial neurofeedback effects in closed-loop auditory attention systems.