Behavior Research Methods

Purpose. The aim of this study was to evaluate how eye-trackers are currently used in clinical settings and to explore whether guidelines are perceived to be needed to support their use in a clinical context. Methods. A survey was developed and distributed to clinical eye-tracker users, manufacturers, professional bodies and other key stakeholders. The survey explored current practices, perceived barriers to clinical eye-tracking, guideline availability, and the perceived need for guidelines to support clinical practice. Results. Eighty responses to at least one of the survey questions were received. Of these, three were excluded from some analyses due to having no clinical involvement in eye-tracking (their involvement was in research only). Eye-tracking was performed by ophthalmologists, orthoptists, neurologists, optometrists, allied scientists and audiologists. Whilst most respondents were assessing eye movements daily, eye-tracking was mostly performed either rarely (less than 10%) or moderately often (10-60%) in their practice. Most clinical respondents (76%) reported they both performed eye-tracking and interpreted the results. The most commonly identified barriers to clinical eye-tracking were eye-trackers being considered primarily research tools, considered unsuitable for performing the assessments, and lack of access to an eye-tracker. Most respondents (71%) were unaware of any existing guidelines for clinical eye-tracking. Where guidelines were reported to exist, they were described as somewhat sufficient to completely insufficient. Most respondents (93%) supported the development of guidelines for clinical eye-tracking. A small number of respondents (7%) reported concerns about introducing clinical eye-tracking guidelines. Conclusion. A broad range of clinicians, professionals and other stakeholders involved in clinical eye-tracking responded to this survey. Their responses suggest a general consensus of a need for clinical eye-tracking guidelines. Whilst guidelines may address some of the reported barriers to eye-tracking clinically, and support benchmarking and standardisation across a broad range of clinical disciplines, concerns around a perceived threat to autonomous clinical practice should also be considered.

In recent years, rodent touchscreen-based paradigms have gained popularity for their flexible task design, automated data collection and improved standardization. These devices minimize experimenter intervention and contribute to more replicable and reliable research. However, the high costs and proprietary hardware / software associated with commercially available systems present a financial barrier for smaller labs or researchers with limited resources. To address this issue, we developed "NC4Touch", an open-source scalable, modular rodent testing apparatus. It features three independent touchscreens and a feeding port for easy use in a wide array of tests comparable to those available with standard operant chambers, with the added benefit that visual stimuli can be highly customized. The system includes a user-friendly graphical interface that offers real-time control, task customization, video recording, and data management. We demonstrate its effectiveness in a visual discrimination task using both rats and mice. Several devices can be operated in parallel from the same computer and user interface, allowing high- throughput data collection. We provide detailed assembly instructions for the hardware and well- documented and easily-configurable software. We hope that the affordable open-source nature of NC4Touch will expand the scope of behavioral testing, allowing researchers to overcome traditional financial barriers and create a more collaborative community. Significance StatementTouchscreen-based testing is a powerful tool for assessing cognition in rodent models, however commercial systems remain cost-prohibitive for many researchers. NC4Touch is a rodent touchscreen apparatus that provides an open-source, customizable, and affordable alternative that enables high-throughput, cognitive testing in both mice and rats. By lowering financial barriers and promoting hackability, this system enables greater accessibility and collaboration in research.

Introduction/Aims: The e-norms method derives reference values from a laboratory's own mixed patient data, without recruiting healthy volunteers. A decade of use exposed three things the original description did not address: a dependence on how finely measurements are recorded, a minimum sample size below which derived limits are systematically too permissive, and an extension of the method to engineered composite features. This work characterizes the first two and describes a corrected plateau rule. Methods: Twenty-six sensory amplitude datasets (18 to 3,301 measurements), three sensory conduction velocity datasets, nine median-nerve datasets and a 2,000-measurement peroneal file were analyzed. Amplitudes were progressively rounded; velocities were made finer by adding noise of 0.05 m/s. Subsamples of 50 to 750 were drawn from thirteen large datasets on raw, logarithmic and square-root scales. The corrected rule was validated against expert visual plateau selection and against the plateau proportions reported in 2015, and limit stability was estimated by resampling. Results: At n = 50, amplitudes as recorded returned 0.28 of the full-sample limit with 73.1% failing; rounded to whole numbers they returned 1.03 with none failing. Integer velocities returned 1.03 with no failures, but failed in all 300 draws once noise was added. Derived limits were systematically too low below 250 to 300 measurements. The corrected rule returned 1.00 with 0.1% failures and left integer velocities essentially unchanged. Discussion: The dependence is intrinsic to locating the plateau from gaps between sorted values. Obtaining 250 normal studies means accumulating 530 to 1,250, depending on a laboratory's normal rate.

Cognitive tasks are moving from the laboratory onto participants' own devices. Browser runtimes make that move easy. Their stimulus timing, however, can vary by tens of milliseconds across devices. We asked how much timing error cognitive and neural measures can tolerate. We injected per-trial stimulus-timing jitter into a public EEG dataset (ERP CORE, N = 40) and re-measured three event-related potential components. Under a prespecified criterion of half the between-participant SD, the N170 tolerated jitter up to 15.8 ms, limited by peak-latency bias. The slower MMN and P3 tolerated more than 20 ms. The full jitter distribution predicted the loss. Heavy-tailed jitter cost less than Gaussian jitter of the same RMS. In simulated reaction-time data, the same tail shape determined whether platform error raised or lowered the ex-Gaussian {tau}, a candidate attentional biomarker. A single RMS timing figure therefore cannot tell whether a platform is adequate for these measures. We built the Mobile Experimental Research Investigative Toolkit (MERIT) to meet these requirements without code. Researchers design tasks in a visual builder that supports reaction-time, continuous-control, drawing, and sequential task families. A Unity engine embedded natively in the Curious iOS and Android apps runs the tasks. Measured with an optical sensor, inter-stimulus-interval SD was 0.71 to 0.96 ms on three current phones and tablets. It was 3.73 ms on a 2019 budget Android phone. Every device measured at least four times below the N170 ceiling. In the laboratory, the same task definition streams event markers to Lab Streaming Layer (LSL) for synchronized EEG and physiological recording. In the field, it runs on ecological momentary assessment schedules. Native mobile players, no-code builders, and EMA platforms each exist as separate systems. MERIT combines them, so the task characterized against brain and body recordings in the laboratory is the same task participants complete at home.

The current study aims to identify materials that most people experience as being unpleasant to touch and to uncover the perceptual attributes that link to unpleasantness. We sampled 79 everyday materials covering a wide range of sensory properties such as fluidity, granularity, roughness, and softness, e.g., cooking oil, sand, coconut, and silk. We asked participants to freely explore these materials with their dominant hand and subsequently rate their unpleasantness and 23 sensory attributes (e.g., elasticity, stickiness) in the absence of visual and auditory feedback. We collected data on material unpleasantness from two populations: Germany and the UK. Our wide array of materials showed different unpleasantness levels that correlated with specific perceptual material attributes. In general, viscous and deformable materials (e.g., hair gel, play-doh) felt more unpleasant to touch than surface-soft and smooth materials. Results were comparable for the German and UK samples. Overall, our results revealed a relationship between specific sensory attributes of materials and their unpleasantness, which seems to be consistent across German and UK samples. Our findings contribute to the understanding of how various sensory attributes are related to haptic unpleasantness. These insights advance our understanding of haptic perception with various implications, such as product design and affective computing systems, as both seek to integrate touch for detecting and modulating human affective states.

Museum visits typically follow curator-defined routes that constrain how visitors shape their own experience, yet choice is widely held to heighten engagement, autonomy, and enjoyment. Virtual reality (VR) offers a setting in which to study these processes because it combines ecological immersion with precise, continuous behavioral measurement. We investigated (i) whether VR- derived behavioral signals are associated with self-reported enjoyment during a virtual museum tour, and (ii) whether the level of agency afforded to visitors influences enjoyment. Forty-eight adults completed a room-scale, life-size VR tour (8 * 4 m) of seven paintings from the Tel Aviv Museum of Art, each accompanied by a synchronized audio guide. Synchronized gaze and head- position streams were logged continuously (50 Hz) and segmented into painting-level viewing episodes using a trial-and-tile pipeline that intersects each painting's trial interval with an empirically defined spatial window in front of the canvas. Participants were randomly assigned to one of three agency conditions, Active (choice before every artwork), Semi-Active (choice for the first three), or Passive (fixed route),while the artwork sequence was held identical. Self- reported enjoyment at the tour and painting levels did not differ reliably across agency conditions. Among VR-derived measures, gaze engagement during the audio guide showed the clearest (though modest) association with painting-level liking, whereas locomotion and pacing measures were weak and inconsistent predictors. Agency nonetheless reliably modulated several gaze- and time-based viewing measures. The findings reveal a dissociation between subjective enjoyment and the micro-structure of viewing, and establish a reusable framework for full-tour, painting-level behavioral analysis in immersive settings.

Recognizing others emotions is central to social interaction. Traditional biological psychology infers emotional responding via laboratory measures, whereas contemporary computer vision algorithms claim to identify emotions unobtrusively from facial video. However, the validity of such algorithms for classifying spontaneous emotional responses occurring without explicit communicative intent remains debated. We compared established psychophysiological measures (EEG, facial EMG, EDA activity) with the open-source facial behavior toolkit OpenFace for classifying participants spontaneous responses during free viewing of happiness-inducing, disgust-inducing, and neutral pictures. Participants provided valence and arousal ratings and later selected the basic emotion that best matched their reaction which served as the classification criterion. Using within-participants single-trial support vector machine (SVM) classification, EEG achieved the highest accuracy (40%), followed by facial EMG (37%); OpenFace reached 36%. All methods except EDA exceeded chance performance (33.3%) and were lower compared to human raters (48%). Predictions declined slightly for across-participants SVMs, being at chance for OpenFace and EDA. The results indicate that in principle both, psychophysiological measures and video-derived facial action units, can capture diagnostically relevant aspects of emotional responding during picture viewing, but that their performance is limited when expressions are spontaneous and not produced for communicative purposes. Inter-individual variability in expressivity and physiological responding likely contributes to these limitations and should be considered when deploying automatic emotion recognition in research or applied settings.

BackgroundAutomated behavioral tracking is increasingly used in biological and biomedical research; however, robustness across heterogeneous imaging conditions remains a major challenge. Domain shifts caused by changes in illumination, contrast, or acquisition setup can substantially degrade the performance of computer vision and deep learning models, limiting their practical applicability. This issue is particularly relevant for small-scale biological datasets, where extensive annotation and retraining are often impractical. MethodsWe developed a behavioral tracking framework for Galleria mellonella larvae combining a classical computer vision (CV) pipeline, a YOLOv8s-seg + ByteTrack deep learning pipeline, and generative domain adaptation methods. Behavioral recordings were collected in two independent experimental sessions under distinct illumination conditions (top and bottom lighting), creating a natural cross-session domain shift. The classical pipeline was based on contour detection, adaptive preprocessing, temporal smoothing, and trajectory reconstruction. Deep learning models were trained on 320 manually annotated frames and evaluated in both within-session and cross-session settings. To improve generalization, we investigated generative augmentation using StyleGAN2-ADA and unpaired image-to-image translation using CycleGAN. Tracking outputs were further analyzed through behavioral descriptors, including trajectories, traveled distance, velocity, spatial occupancy heatmaps, and directional movement patterns. ResultsThe classical CV pipeline achieved high detection performance in both recording sessions, with mean detection rates of 99.05% and 99.71%, respectively. A YOLOv8s-seg model demonstrated strong within-session performance but exhibited severe degradation under cross-session evaluation, with larval mask segmentation performance dropping to mAP@0.5 = 9.1%, confirming the presence of a substantial domain shift. Despite differences in detection methodology, behavioral metrics derived from YOLO and CV pipelines showed strong agreement at the group level (Pearson correlation r = 0.89; median distance ratio = 0.99). Generative augmentation with StyleGAN2-ADA did not yield meaningful gains in tracking robustness -- likely because baseline performance was already near ceiling -- whereas CycleGAN-based domain adaptation substantially reduced the domain gap and improved cross-session detection performance while preserving biologically relevant trajectory structures and spatial behavioral patterns. ConclusionsCross-session variability represents a critical challenge for automated behavioral tracking in biological experiments. Our results demonstrate that carefully designed classical computer vision approaches can achieve highly reliable tracking in small-data settings, while deep learning models require explicit strategies to address domain shift. Generative domain adaptation, particularly CycleGAN-based image translation, offers an effective solution for improving cross-session generalization without additional manual annotation. The proposed framework provides a robust foundation for scalable behavioral phenotyping of Galleria mellonella and other small biological model organisms.

Natural-sound research requires stimulus sets that combine acoustic standardization with detailed behavioural characterization. We present 1,377 two-second sounds representing 240 expert-defined source--action classes. We call this database "MaMa Sounds", as it resulted from the collaborative effort of two academic teams in Maastricht and Marseille. The sounds were manually curated, segmented, sampled at 16 kHz, and labelled with a noun identifying the source and a verb identifying the action. We release deidentified trial-level identification and familiarity data together with multiple per-sound norms (e.g., identification accuracy, confidence and agreement; familiarity), along with overall norms derived with principal component analysis. Noun, verb, and joint noun--verb norms are provided as direct means and medians with the number of contributing observations. This battery preserves process-specific information, while two principal-component scores provide compact overall behavioural-identifiability measures derived from response ease, semantic correspondence, agreement, and familiarity. The repository also contains deterministic response-cleaning code, participant and reference Word2Vec representations, and code reproducing the public sound-level tables. The resource supports stimulus selection, matching, and continuous modelling in auditory cognition and neuroscience.

PurposePupil diameter in daily life depends on both the light reaching the eye and the observers age, but established prediction formulas require laboratory quantities that are rarely measured in natural environments. We developed a compact age-corrected model that predicts pupil diameter from melanopic equivalent daylight illuminance (mEDI). MethodsWe used an existing field dataset in which binocular pupil diameter and near-corneal spectral irradiance were recorded while 83 adults aged 18-87 years moved through indoor and outdoor environments. The analysis included 10,082 valid paired observations. We fitted a bounded sigmoid relating pupil diameter to mEDI and age, with each participant given equal influence, and assessed prediction in participants excluded from model fitting. Performance was compared with simpler models, a flexible generalised additive model (GAM), and Watson-Yellott predictions based on assumed field geometry. ResultsPupil diameter decreased smoothly as mEDI increased. Age primarily reduced the difference between pupils in dim and bright conditions, by 0.768 mm per decade, while the predicted bright-light diameter changed little with age. In held-out participants, the bounded model had a participant-balanced root mean squared error (RMSE) of 0.630 mm and mean absolute error of 0.537 mm. The GAM had a slightly lower point-estimate RMSE of 0.610 mm, but the difference was small and uncertain. The bounded model outperformed the tested log-linear, reduced, age-only, and Watson-Yellott alternatives. ConclusionAge and mEDI are sufficient to provide useful population-average pupil predictions across the observed adult age and real-world light range. The model is transparent, physiologically bounded, and nearly as accurate as a flexible GAM, but predictions approaching darkness remain uncertain because valid mEDI measurements were not available in that range. Key pointsO_LIA compact equation predicts population-average pupil diameter from age and mEDI alone. C_LIO_LIAge mainly compresses the pupils response range by reducing pupil diameter under dimmer conditions. C_LIO_LIPrediction error in unseen participants was close to that of a flexible GAM, without requiring a fitted smooth object. C_LIO_LIThe model is intended for the observed adult age and field-light range, not for extrapolation into darkness. C_LI

Understanding how our different senses interact to shape our perception is essential to design realistic and immersive virtual and augmented reality (VR/AR) experiences. The present study investigated how roughness perception can be modulated through haptic, visual, and auditory cues in AR using a vibrotactile wristband. Participants compared virtual textures varying in vibration frequency/amplitude, visual grain size, and friction sound. Results revealed strong linear relationships between stimulus parameters and perceived roughness, with haptic frequency and visual cues driving the highest discrimination performance. Adding non-informative sensory feedback reduced perceptual sensitivity, acting as noise. Individual differences emerged: participants who rated haptic as the easiest modality showed greater sensitivity to haptic variations, while visual-reliant participants performed better with visual cues. We conclude that roughness in AR can be systematically manipulated, but is vulnerable to perceptual interference from irrelevant inputs, where our work provides actionable insights for implementing optimized and adaptive AR/VR interfaces.

Automated home-cage systems measure individual behavior in social groups for days to months. Among these systems, the IntelliCage has become a widely used platform for longitudinal and socially embedded behavioral phenotyping. Yet the analysis layer often remains less standardized than the experiment itself: raw exports, phase definitions, exclusion rules, time alignment, and derived behavioral metrics are transformed by lab-specific scripts that are difficult to audit, compare, or reuse. We present ic-analysis, an open-source Python toolkit for standardized, scriptable, and shareable IntelliCage workflows. The toolkit separates user-defined experiment metadata and workflow scripts from a reusable analysis core with modular analysis and plotting functions, allowing users to flexibly assemble experiment-specific pipelines without editing package internals. Due to its modular design, the analysis core can be applied to a broad range of experimental paradigms, including general activity, exploratory, motivational, cognitive, and social readouts, rather than being limited to a single fixed protocol. It aligns biological phase windows across staggered cage runs and exports plots together with quantitative result tables, applied settings, and audit files that support reproducible and FAIR reporting. Here, we demonstrate this flexible design in a realistic synthetic place-learning/place-reversal experiment with two mouse groups and deliberately offset cage starts. The workflow recovered the implanted behavioral differences while preserving the required experimental-time alignment. Group A showed stronger endpoint saccharin preference (81.7 +/- 1.7% vs. 27.4 +/- 3.4%, p = 1.9e-), higher liquid uptake, faster place-learning onset (56.6 +/- 8.1 vs. 278.3 +/- 33.2 visits), and better reversal performance (64.1 +/- 1.3% vs. 25.6 +/- 1.7% rewarded correct-corner visits) compared to Group B. This demonstration shows how standardized, explicitly defined readouts can turn complex IntelliCage exports into interpretable behavioral profiles while preserving the analysis history needed for inspection and reuse. ic-analysis therefore provides both a working analysis scaffold and an extensible, community-friendly route toward IntelliCage workflows that are easier to reproduce, compare, extend, and share.

Wayfinding in hospitals is often hindered by ineffective signage; however, the cognitive mechanisms of healthcare wayfinding symbols comprehension remain under-researched. This study utilized eye-tracking and spatial gaze mapping to examine how visual complexity, abstraction, and human figuration modulate perception in 40 healthy adults viewing 24 hospital-related healthcare wayfinding symbols. Results indicate that pupil size is a sensitive physiological marker of cognitive load, significantly influenced by visual complexity ({chi}2 = 11.32, p = .022) and abstraction ({chi}2 = 7.49, p = .027). Human figuration reduced fixation duration and increased saccade amplitude, facilitating efficient semantic integration. Furthermore, human-centric healthcare wayfinding symbols elicited streamlined gaze trajectories, whereas abstract/complex designs induced chaotic scanpaths. These findings suggest that human figuration acts as a cognitive scaffold, reducing mental effort. We provide evidence-based guidelines for optimizing healthcare wayfinding symbols by prioritizing human body representations and balancing abstraction levels. HighlightO_LIPupil size indexes cognitive load during symbol comprehension. C_LIO_LIHuman figuration cuts fixation duration, boosting wayfinding efficiency. C_LIO_LIAbstract symbols increase pupil dilation, raising cognitive load. C_LI

Tinnitus is one of the most common hearing disorders affecting one-third of Americans and is defined as the buzzing or ringing sound one perceives in one or both ears in the absence of an acoustic stimulus. One objective method for tinnitus screening in rodents is gap prepulse inhibition of the acoustic startle reflex (GPIAS), a reduction in the abrupt motor response elicited by an intense auditory stimulus following a silent gap in noise. Reduced inhibition by gaps embedded in narrow-band noise is hypothesized to reflect the primary tinnitus pitch, as the tinnitus "fills-in" the gap. However, Lobarinas et al. (2013) identified a critical limitation: after acoustic trauma or hearing loss, rodents often exhibit markedly diminished acoustic startle reflexes, creating a "floor effect" where further suppression becomes undetectable even if gap perception remains intact, leading to false-positive tinnitus screening results. To address this limitation, we utilized the CBA/CaJ mouse model to assess tactile airpuff modification efficacy and employed a machine learning algorithm to classify startle responses, achieving 98% accuracy in differentiating startles from non-startles. We induced unilateral conductive hearing loss via ear plugging and found enhanced gap detection ability, contrasting with false-positive tinnitus indicators observed in rats by Lobarinas and colleagues. We also pharmacologically induced tinnitus via sodium salicylate, revealing frequency-specific alterations in gap detection patterns. Our findings suggest that differences in data analysis methodology, specifically using a machine learning algorithm to filter non-startle responses, may explain species-specific differences between mouse and rat models and significantly improve GPIAS validity as a tinnitus screening assessment tool.

BackgroundPersonal light dosimeters enable exposure assessment under free- living conditions, but sensors are rarely positioned near the eyes--the relevant site for visual and non-visual responses. Chest and wrist placement may improve adherence and scalability, yet placement-dependent error and its consequences for derived outcomes remain insufficiently characterised. ObjectiveTo quantify how chest and wrist placement affects time-resolved estimates of eye-level light exposure across naturalistic contexts and derived exposure metrics. MethodsWe analysed concurrent 10-s melanopic equivalent daylight illuminance measurements from identical dosimeter models at the glasses, chest, and wrist across eight sites in seven countries. Melanopic equivalent daylight illuminance was compared at two analytical scales. Generalised additive mixed models quantified time-resolved placement error across contexts (N=787 participant-days), and mixed- effects models compared 54 daily metrics (N=604 participant-days). Hierarchical bootstrap resampling quantified the precision of population-average metric bias across participant numbers and monitoring durations. ResultsBody-worn dosimeters generally underestimated eye-level exposure. Across categorical contexts, estimated mean errors ranged from -23.1% to -4.2% at the chest and from -54.7% to -30.3% at the wrist, with larger errors generally observed at night. Nineteen of 54 chest-derived and 27 of 54 wrist-derived metrics differed nominally from glasses. Among these, median absolute bias was 5% at both placements, but maxima reached 50% and 84%, respectively. Timing outcomes were comparatively insensitive, whereas level and temporal-dynamics outcomes were more sensitive. With seven days per participant, the class-median bias standard deviation reached the 5% precision tolerance with 2-16 participants for all classes except temporal dynamics, which required 33 at the chest and 59 at the wrist. SignificanceDosimeter-placement validity depends on analytical scale and exposure construct. Chest placement can support scalable studies focused on aggregated timing or duration outcomes, whereas wrist placement introduces greater and less predictable bias. Near-eye measurement remains preferable for small- sample studies, time-resolved exposure levels and temporal-dynamics analyses.

Background: Auditory Brainstem Responses (ABRs) are objective and highly standardized evoked potentials, but their clinical interpretation still largely depends on expert visual assessment. Most automated ABR approaches focus on signal detection, threshold estimation, or waveform recognition, whereas the diagnostic reasoning linking ABR features to clinically meaningful categories remains insufficiently formalized. Methods: We developed a theoretical, rule-based framework for ABR classification. Nine clinically relevant descriptors were defined: amplitudes of waves I, III, and V; latencies of waves I, III, and V; the I/V amplitude ratio; and the I-V and III-V interpeak intervals. The complete theoretical combinatorial space was generated and progressively reduced using rules that accounted for non-applicable descriptors, algebraic consistency, and neurobiological plausibility. Two expert raters independently labeled valid configurations and compared with language-model-assisted and logic-based rule systems. Agreement was assessed using percentage concordance and Cohen's k;. Descriptor importance was explored using Random Forest analysis. Results: The initial space of 5,832 possible ABR configurations was reduced to 343 valid configurations. Expert raters showed high agreement, with 84.55% concordance and Cohen's k = 0.620. Agreement between expert classifications and language-model-assisted rules was lower, whereas logic-based rules showed higher agreement with both raters and were retained as the final proposed rule set. The diagnostic space was highly asymmetric, with NORM, PPSHA, and PPC occupying narrow regions, and PB and MIXED representing broader diagnostic domains. Wave I amplitude was the most influential descriptor across classification agents. Conclusions: This work provides an interpretable framework for ABR classification by transforming a broad theoretical combinatorial space into a constrained diagnostic domain. This proof-of-concept study supports the feasibility of modelling ABR interpretation as a structured diagnostic space grounded in auditory neurobiology, providing a reproducible foundation for future clinical validation and semi-automated decision-support tools.

This tutorial seeks to facilitate the use of mobile electroencephalography (EEG) in young children. Mobile EEG allows researchers to investigate neural correlates of behavioral and cognitive processes in ecologically valid settings. While mobile EEG has been widely adopted in adult research, studies applying it to freely moving children remain scarce. However, investigating neural processes during active behavior and in interaction with movement is indispensable for advancing our understanding of neural correlates of cognitive development in early childhood. Here, we provide a practical tutorial on the collection and preprocessing of mobile EEG data from children. Drawing on experience and data from a large-scale study with toddlers, we summarize key methodological considerations, discuss common challenges and practical recommendations, and present a preprocessing pipeline developed for developmental mobile EEG data. In addition, we provide example EEG datasets from naturally moving 18-20-month-old toddlers. We argue that incorporating mobile EEG into developmental cognitive neuroscience allows researchers to (i) investigate cognitive development in naturalistic environments, (ii) examine associations among bodily movement, neural activity, and behavior, and (iii) study samples that are difficult to reach with stationary research. HighlightsO_LIMobile EEG enables investigation of neural processes during active behavior C_LIO_LIMobile EEG in moving freely children comes with unique methodological challenges C_LIO_LIWe provide practical recommendations for developmental mobile EEG studies C_LIO_LIA dataset and preprocessing pipeline from a toddler mobile EEG study are provided C_LI

The measurement of light received at the cornea of the eye is a paramount consideration for the understanding of the relation between environmental illumination and the non-image-forming effects of light. The field of view (FOV) at the cornea is less than a full hemisphere, because it is partially occluded by human facial morphology. The International Commission on Illumination (CIE) has defined a standard model of human FOV. A suitably designed physical occluder attached to the sensor (of a light meter) has been proposed as a means of incorporating the effect of human FOV when taking measurements. Similarly, when using simulation to predict light received at the cornea, a geometrical description of the occluder at the eye point(s) can be added to the 3D model of the scene. The first occluder model proposed to represent CIE human FOV was enumerated in terms of: the CIE definition; the radius of the occluder; and, the radius of the light sensor disc. We present a simpler model based only on the CIE definition and the occluder radius. Both models were tested using a virtual goniophotometer. Various sensor response functions describing the spatial sensitivity across the sensor disc, including several we characterized through laboratory measurements, were included in the test. For all functions considered, the performance of the simpler occluder model was equivalent to or better than the model first proposed.

Large language models (LLMs) are increasingly evaluated for clinical tasks, but safe use requires reliable uncertainty estimates as well as high accuracy. We tested eight general-purpose LLMs on 100 validated neurology board-style multiple-choice questions and compared three confidence signals: self-reported confidence, critique-implied confidence from an independent model instance, and consistency-based confidence from repeated sampling. Accuracy varied from 45% (Llama 3.1 8B) to 92% (Gemini 3 Flash/Pro), but confidence was generally overestimated, particularly in lower-accuracy models. Consistency-based confidence provided the best discrimination between correct and incorrect answers (pooled AUROC 0.74), outperforming self-report, although calibration differences between methods were not significant. Calibration should therefore be reported alongside accuracy when clinical LLMs are evaluated.

Background The nonfluent/agrammatic (nfv) and logopenic (lv) variants of primary progressive aphasia (PPA) disrupt fluency through distinct underlying neurocognitive mechanisms. Differential diagnosis currently requires hours of cognitive-linguistic testing, with additional barriers for bilingual patients due to a shortage of bilingual service providers and a lack of well-established assessment methods. In English speakers, a promising automated approach for differentiating nfvPPA and lvPPA is to derive speech-timing measures and linguistic features from connected speech as input to machine learning (ML) classification algorithms. To our knowledge, this approach has not been evaluated in the context of bilingualism. Methods Thirty four Catalan Spanish simultaneous bilingual patients (lv = 24, nfv = 10) were asked to describe a picture (Western Aphasia Battery Picnic Scene) in both their dominant and non-dominant language. From the participant's recorded response, we derived four feature sets: speech-timing measures, derived with PRAAT; word-level parameters, derived from corpora; linguistic features, derived with the natural language processing tools SpaCy and CLAN; image-text congruence scores, derived with the vision-language encoder Multilingual-CLIP. Each feature set was fed into classification algorithms for differentiating nfv from lv in participants' non-dominant and dominant samples. Then, we combined each feature set's classifier into an ensemble model. We used the McNemar test to determine the statistical significance of differences in classification performance between responses in the non-dominant and dominant language. Results The best-performing classifier achieved F1 macro scores of 93% (word-level parameters) and 92% (ensemble) in the non-dominant and dominant language, respectively. For all feature sets and ensemble models, classification performance did not significantly differ between the non-dominant and dominant language. Ensemble modeling did not significantly improve classification performance in either language. Conclusions Taking advantage of recent advances in multilingual multimodal machine learning, we accurately differentiate Catalan-Spanish bilingual individuals with nfvPPA and lvPPA using a largely automated, time-efficient (1-2 minutes), and ecologically valid connected-speech-based approach. Future directions include evaluating this approach on larger datasets balanced by PPA subtype, using automated transcriptions of connected speech. Our study represents a step towards addressing current inequities in PPA differential diagnosis for non-English-speaking bilingual speakers. Trial registration Data from the clinical trial NCT05741853 was retrospectively analyzed