Philosophical Transactions of the Royal Society B

Gene drives, genetic constructs that can spread deleterious alleles in wild populations, have the potential to address some of the major pressing challenges of the Anthropocene such as invasive species, spread of disease vectors, and agricultural pests. However, responsible and effective deployment of gene drive requires taking into account the complex nature of real-world population connectivity networks. In particular, it is unclear how the topological position of the deployment site affects the spread process and its final outcome. Here we develop a framework for modeling gene drive spread in population connectivity networks, and study the eco-evolutionary dynamics of gene drive spread under complex population structures. We investigated the relationship between the position of the deployment site in the topology of the network and whether the gene drive is eventually lost, fixed, or maintained at an intermediate frequency. We identified network centrality measures of deployment sites that are highly correlated with the outcome of deployment for different gene drive designs and across diverse network topologies. We also show that there is a trade-off between the time-to-fixation and the final outcome, implying that multiple centrality measures of the deployment site would need to be considered when aiming to achieve rapid and successful population control using gene drives.

Wildlife crossing infrastructure is promoted to restore connectivity for fragmented populations, but its effectiveness at enabling natural recolonisation remains untested. We tested this using a spatially explicit agent-based model parameterised with GPS telemetry data from bobcats (Lynx rufus) in New Jersey, USA. By integrating movement behaviour, stochastic demography, habitat suitability, and traffic-dependent mortality risk, we simulated 50-year recolonisation dynamics across a highly urbanised landscape. Despite extensive unoccupied suitable habitat, natural recolonisation completely failed across all scenarios, with vehicle-induced mortality during dispersal acting as the primary limiting factor and turning the matrix into a demographic sink. Even an idealised mitigation scenario in which mortality at high-mortality crossings was reduced to zero failed to produce a self-sustaining population. Although dispersal increased, individuals at the recolonisation front remained too sparse to overcome the mate-finding Allee effect. Sensitivity analysis confirmed that the recolonisation-failure result is robust to {+/-}50% variation in per-crossing mortality and {+/-}25% variation in disperser survival. Restoring structural connectivity is not, in itself, a sufficient intervention for recovering low-density carnivore populations facing a high-mortality matrix. Instead disperser survival and local density at the recolonisation front are the rate-limiting determinants. In such systems translocation rather than crossing-structure investment is more likely to result in recolonisation success.

Large-scale annual releases of pheasants Phasianus colchicus and their subsequent management for recreational shooting create various ecological impacts in the UK. While effects at release sites are fairly well understood, dispersing birds may influence areas farther away. If they enter ecologically important but sensitive protected areas (PAs), any negative impacts could be especially harmful. Using tracking data, from 766 birds across 10 sites, we estimated survival and dispersal of released pheasants and applied these patterns to gamebird release records near English PAs to gauge intrusion risk. Of 2,885 registered release sites, just over half lay within 2 km of a PA. A large number of shoots release relatively few birds while a small number release many birds. Thus, numbers expected to enter a particular PA likely depend both on the size of releases and proximity to the PA. We estimate that, at a national level, a maximum of between 525,000 and 784,000 pheasants might be found within PAs very soon after release, representing around 1.7% of all the pheasants released annually. This number declines over the months after release until in February, we estimate that there are between 131,000 and 196,000 pheasants (0.4% of the total release) might be found within PAs. The critical metric by which ecological damage might occur is their density within PAs. Mean densities soon after release averaged 12.0 birds/ha in PAs within 250 m of release sites. This density declined markedly both in time (as birds died) and space (as they moved further from the pen as potential areas increased). By November, densities in PAs 500-1000m from release sites peaked at 0.5 birds/ha, falling to 0.16 birds/ha in February. These estimated densities are around two orders of magnitude lower than those known to cause strong, lasting impacts within release pens. The results are subject to assumptions about movement behaviour, game management and bias in registration. Despite these constraints, considerable local variation exists, with a minority of high-volume release sites very near PAs posing the greatest potential ecological risk.

AO_SCPLOWBSTRACTC_SCPLOWVision can shape auditory perception, especially when visual cues occur at different times and locations than sounds. Simultaneous but spatially misaligned lights bias the perceived location of a sound--a phenomenon known as the ventriloquism effect. Temporally misaligned lights can also affect the latency of auditory responses. However, it remains unclear how multiple visual stimuli that differ from sounds in both space and time jointly influence localization behaviour. We investigated how visual distractors, spatially misaligned by 10{degrees}, presented before and/or during a target sound influence localization accuracy and response latency in a rapid head-pointing task. Human listeners localized brief (150 ms) broadband noise bursts with an average root-mean-square error of 5{degrees} and a baseline latency of 252 ms. Simultaneous visual cues induced the ventriloquism effect, in which the perceived sound location was biased by 1.8{degrees}. Response latency also increased by 21 ms (273 ms). Preceding visual stimuli (2 s duration) did not induce a bias, but increased latency by 55 ms (307 ms). Introducing a 200 ms gap between the preceding light and the sound reduced this latency increase to 24 ms (276 ms), still not inducing a significant bias. When we presented both a preceding and a simultaneous light on opposite sides of the sound, localization reflected the bias induced by the simultaneous light (1.8{degrees}) and the latency increase induced by the preceding light (by 48 ms). These findings reveal a dissociation in audiovisual integration: preceding visual stimuli primarily influence when a sound is responded to (latency), while simultaneous stimuli influence where it is perceived (accuracy). This supports causal inference models of multisensory integration and suggests distinct underlying mechanisms for spatial and temporal processing of sounds in sensorimotor circuits.

BioBlitzes are widely used citizen science events that combine biodiversity monitoring, public participation, and environmental awareness through short and intensive observation campaigns. However, applying this model to marine environments presents additional challenges related to safety, access, weather dependency, specialised equipment, species identification, and sustained participation. This paper presents the BioMARathon model as a case study of how BioBlitz-inspired events can be adapted to marine citizen science contexts. The BioMARathon extends the conventional BioBlitz format into a longer, seasonal, and distributed engagement model designed specifically for marine and coastal environments. The paper describes the conceptual foundations of the model in the Janus Engagement Framework, which informed both the design of the BioMARathon and the adaptation of the MINKA citizen science observatory to better support participation, validation, feedback, and continuity over time. BioMARato Catalunya, launched in 2021, is presented as the founding implementation of the model and as the basis for later replication in Portugal. Between 2021 and 2025, BioMARato Catalunya showed continued growth in participation, observations, and taxonomic coverage, while also contributing to the detection of non-indigenous species, first regional records, and climate-related ecological impacts. Beyond biodiversity outcomes, the case suggests that extending participation across a season, distributing activities through local mobilising organisations, and combining expert validation with visible feedback mechanisms can support recurrent participation, retention, and community reactivation in marine citizen science. Rather than offering a formal causal evaluation, this article contributes practical lessons for the design of citizen science initiatives in challenging environments.

Codon optimality promotes efficient translation and, as recent research has shown, also extends mRNA lifetimes. However, how control is distributed between translation and mRNA degradation remains unclear. We show that this relative impact depends strongly on the measurement approach. Using fluorescent protein reporters can underestimate codon-optimality-dependent increases in translation efficiency. Conversely, analyses based on poly(A)-selected RNA overestimate the impact on translation, because stable transcripts undergoing poly(A) shortening are often inefficiently captured, leading to skewed protein-to-mRNA ratios. This technical bias is not offset by the marginal decline in ribosomal association observed as mRNAs age. Estimates based on total RNA measurements redistribute some of the control attributed to translation to mRNA stability, making the contributions comparable for mRNAs with shorter coding sequences. For longer mRNAs, codon optimality increasingly controls elongation speed, with a greater effect on translation efficiency than on degradation. These insights highlight the importance of measurement strategy for accurately quantifying the determinants of mRNA stability and protein synthesis.

Passive Acoustic Monitoring (PAM) has emerged as a transformative tool for biodiversity assessment in recent years. Despite widespread acceptance and application for conservation-related outcomes, the synergistic effects of hardware limitations, signal propagation, and environmental conditions on how far a signal can be reliably detected remain critically understudied. We quantified changes in signal detectability using Autonomous Recording Units (ARUs) in a tropical agroecosystem using playback experiments of standardised pure-tone (1-8 kHz) in fallow rice paddy fields. We deployed a four-ARU array and broadcast signals over a 50- 300 m distance gradient, and modelled operative detectability of signals using a binomial Generalised Linear Mixed-effects Model (GLMM). Our findings show that the detection space of an ARU is highly frequency-dependent and environmentally modulated. Detection probability for low-frequency signals (1 kHz) decreased rapidly (50% threshold at [~]100 m), whereas mid-range frequencies (4-6 kHz) occupied an acoustic window that remained reliably detectable up to 250 m. Higher relative humidity significantly enhanced overall detection, while increasing temperatures disproportionately reduced low-frequency detectability. The orientation of the ARU to the signal source was important as the detection probability declined from 81% for recorders facing the source (0{degrees}) to 14% for rear-facing units (180{degrees}). Our findings underscore the importance of determining the detection space before undertaking PAM. We propose a Decision Support Framework that provides a pathway for researchers to integrate focal taxa traits with technical constraints to determine detection space and optimise study designs when using PAM for monitoring biodiversity and assessing conservation action.

Existing theory examining the coevolutionary dynamics of species range borders assumes random dispersal, which causes maladaptive gene flow from the range core to the range margins and contributes to the formation of range limits. However, dispersal is unlikely to be random for many organisms in nature, calling into question existing theoretical predictions. For example, if individuals exhibit phenotype-dependent adaptive dispersal strategies such as matching habitat choice, then the resulting adaptive gene flow toward species range margins could facilitate range expansions and potentially prevent the formation of range limits by interspecific competition. To test this idea, we use a comprehensive mathematical model to develop a quantitative theory of range border coevolution that incorporates phenotype-optimal dispersal--a particular form of matching habitat choice in which individuals follow the gradient in an environmental optimum phenotype to settle in the habit best suited for their phenotype. We find that instead of preventing competitively formed range limits, adaptive dispersal leads to sharper range limits and reduced character displacement in sympatry. These differences are particularly remarkable when natural selection is weak, when individuals are specialized in their resource use, or when individuals are highly sensitive to environmental conditions. We show that matching habitat choice causes backward edge-to-core movements which dynamically interact with the effects of interspecific competition to establish the range limits. Thus, the formation of range limits by interspecific competition is robust to assumptions about individual dispersal. Further, our results identify the competitive advantage of evolving matching habitat choice in steep environmental gradients, especially for slowly-growing species in rapidly fluctuating climates.

The release of tens of millions of common pheasants (Phasianus colchicus) across the UK for shooting may pose an ecological risk to native species and sensitive habitats, particularly if the birds move into protected areas (PAs) such as Special Areas of Conservation (SAC), Special Protection Areas (SPA), and Sites of Special Scientific Interest (SSSI). The extent of this ecological risk depends on the abundance of pheasants in these sensitive sites, especially if they are attracted there after the shooting season when game management efforts to retain the birds cease. We used relative pheasant abundance measures derived from British Trust for Ornithology bird atlas data from 3793 2km tetrads across four English counties (Berkshire, Cornwall, Devon, and Hertfordshire) to determine if pheasants preferentially disperse into or reside in areas with greater PA coverage. We analysed relative abundance in both the winter shooting season and the breeding season using a Bayesian occupancy-abundance model, controlling for habitat type and diversity. Our results showed a strong influence of habitat on pheasant abundance, consistent with known habitat preferences. However, we found no evidence of a relationship between relative pheasant abundance and the proportion of ecologically relevant PA coverage in a tetrad. This lack of a relationship was consistent across all four counties and across both the winter and breeding seasons. Our finding suggests that common pheasants do not preferentially disperse into or reside in protected areas compared to surrounding, unprotected land, suggesting that the ecological impacts caused by released pheasants are no more likely to occur in protected areas than in non-protected areas.

The Biodiversity-Ecosystem Functioning (BEF) literature has shown species diversity to be essential for ecosystem functioning and services. Yet although acquiring information through interspecific networks can impact ecosystem functioning, it is unclear how it is modulated by species diversity. Eliciting vocal responses using predator models across a latitudinal gradient, we first show that the species diversity of birds increases public information about predation both in the low-cost system of mobbing and in the higher-cost system of alarm calls. A similar result was also found across a fragment area gradient for mobbing; this system was then used to test how species diversity affects interspecific information flow in mobbing communities. We set up two BEF playback experiments, manipulating the species richness level of the playback sound files by varying the number of species producing mobbing calls (one, two, four, eight species). In an experiment in which the call rate across treatments was held constant, and only heterospecific responses were counted, increasing species richness of the sound files increased the number of species and individuals responding, the number of calls produced and their frequency range, and decreased latency to call. An experiment in which call rate increased with the addition of species in each treatment showed a similar, but stronger pattern. There was little evidence that the signals of one particular species changed responses. This supports the hypothesis that the species diversity of a community is a key component influencing the quantity and quality of information flow inside it.

Language development starts early, possibly even in the womb. Recent results suggest that newborns neural responses to speech in the prenatally heard language are already different from those to unfamiliar languages. However, the neural dynamics of how these differential responses emerge remains little understood. Here, we hypothesize that they are supported by a functional tuning to criticality--a state that maximizes information transmission, dynamic range, and flexibility--, induced in newborns by the prenatally experienced language. To test this hypothesis, we study resting-state brain activity before and after stimulation with naturally spoken sentences in the prenatal language, French, in a rhythmically similar unfamiliar language, Spanish, as well as in a rhythmically different unfamiliar language, English. We show that the native language elicits brain tuning near criticality, balancing network activity and enhancing temporal correlations. Importantly, we find that the network state is not sensitive to the rhythmically different language, and only partially responds to the rhythmically similar language. These results indicate a stimulus-driven tuning to criticality in newborns, a potential foundation for early neuroplasticity, which relies on prenatal experience and may respond to early functional and developmental needs.

Despite sharing the same genes and the same environment, individuals often develop substantial phenotypic differences. While this pattern has been documented across diverse species and traits, the processes giving rise to this "stochastic" or non-shared environmental variation remain unclear. Recent mathematical models of development in which phenotypes are gradually constructed may offer some clues. These models show that imperfect environmental cues can generate striking variation in developmental trajectories and adult phenotypes. At the population level, such imperfect cues produce increasing stability of individual differences across ontogeny (e.g. animal personality) and patterned distributions of mature phenotypes (e.g. normal or skewed) that resemble those observed in real organisms. Our paper synthesizes existing models in which stochastic phenotypic variation arises solely as a by-product of mechanisms missing their phenotypic targets because of imperfect cues. We then link these models to related, but independent, mathematical theory exploring the environmental conditions under which stochastic phenotypic variation is favoured by natural selection. Our integration shows that stochastic sampling is often favoured over classic bet-hedging strategies involving non-plastic generalist or specialist strategies. Our findings provide new directions of research on stochastic sampling as a mechanism for adaptive stochastic variation within and across generations.

Infant environments are rich in rhythms, many of which are social in nature. These rhythms are proposed to play an important role in early communication and interpersonal synchrony. In this cross-sectional electroencephalography (EEG) study with 3- and 6-month-olds (n=31 and n=30, respectively), we examined whether the infant brain tracks the rhythmicity of locomotion-related biological motion in the visual domain and which experiential factors relate to this ability. We found robust neural tracking of biological motion rhythms at both ages, with no effects of age or orientation (upright or inverted). Additionally, we found that caregiver-reported practice of infant carrying/babywearing and caregiver attitudes toward social touch were linked to infant neural tracking of biological motion rhythms, particularly in the inverted condition. Finally, exploratory analyses revealed a lateralisation effect, whereby the left hemisphere processed rightward (vs. leftward) biological motion rhythms more strongly. Our findings suggest that from early on, the infant brain tracks the rhythmicity of whole-body biological motion. Furthermore, caregiver touch-related practices, particularly infant carrying/babywearing, may play a role in infant neural tracking of socially-relevant rhythms.

Human pointing is often used to test whether dogs extract object-specific information from human communicative cues. However, above-chance responses in standard object-choice tasks do not by themselves distinguish between a referential interpretation, in which the gesture identifies a specific target, and an attentional interpretation, in which it primarily biases behaviour toward a broader spatial region. We addressed this issue using an asymmetric six-cup arrangement designed to separate coarse side guidance from exact cup localisation more clearly than a symmetric multi-cup design. Performance in domestic dogs was analysed using three measures: the probability of reaching the correct side, the probability of choosing the correct cup overall, and the probability of choosing the correct cup conditional on having first reached the correct side. The principal comparison involved three matched trial classes: the symmetric 3-vs-3 condition, 2-vs-4 trials with the baited cup on the 2-cup side, and 2-vs-4 trials with the baited cup on the 4-cup side. Descriptively, pointing trials exceeded matched no-point control trials more clearly for side selection than for overall cup choice. The clearest condition effect was observed at the level of side guidance. Dogs were most likely to reach the correct side when the baited cup was located on the 4-cup side of the unequal arrangement. Mixed-effects models confirmed a reliable group effect for side accuracy, whereas overall cup accuracy showed only a weaker and less robust condition effect, and within-side localisation revealed no reliable group difference once condition-specific chance baselines were taken into account. A complementary generative model comparison converged on the same conclusion: a referential-only model fit poorly, an attention-only model captured most of the grouped outcome structure, and a combined model yielded only a modest improvement. Dog point-following is therefore best understood as a layered process dominated by attentional guidance, with only limited additional target-specific localisation.

Conversation requires speakers to plan their responses while still listening to their partner, yet the neural dynamics supporting this overlap remain unclear. In particular, competing accounts differ on whether behaviourally relevant preparation emerges early during comprehension or only close to speech onset. Here, we recorded EEG from pairs of participants engaged in natural, face-to-face conversation and tested whether neural activity during listening predicts both when speakers begin their response (latency) and how long they speak (duration). Using event-related potentials, oscillatory analyses, and multivariate decoding, we show that pre-speech neural activity carries robust information about upcoming behaviour more than one second before articulation. The strongest and earliest effects tracked response duration, with sustained ERP components and alpha/beta power modulations predicting how long participants would speak, even after controlling for behavioural variables. In contrast, neural predictors of response latency were more temporally restricted and partly aligned with partner turn boundaries. Temporal generalisation analyses further revealed stable neural patterns linking early and late stages of preparation. Together, these findings indicate that conversational planning unfolds during listening and that early neural activity reflects the maintained specification of response extent, followed by later processes related to commitment and initiation. This supports a temporally structured account of turn taking in which preparation involves both early maintenance and late motor engagement.

Cephalo-pelvic disproportion in humans has traditionally been interpreted through the obstetrical dilemma framework, assuming a trade-off between bipedal locomotion and childbirth. However, cephalo-pelvic covariation and pelvic sexual dimorphism might be common adaptations to parturition among mammals. We use a controlled hybridization model in mice to test whether cephalo-pelvic covariation and pelvic sexual dimorphism are population-specific, genetically structured, and sensitive to hybridization. We analyzed skull-pelvis variation and covariation, as well as sexual dimorphism of pelvic morphology across four divergent wild-derived mouse strains and their hybrids. Hybridization induced consistent cranial and pelvic size enlargement. Females exhibited significant cephalo-pelvic shape covariation, characterized by an association between rounder, wider birth canals and larger neurocrania, consistent with functional integration under obstetric selection. Hybrids showed disrupted size covariation, increased pelvis shape variance, and reduced cephalo-pelvic integration. Pelvic sexual dimorphism was systematically reduced in hybrids. Cephalo-pelvic covariation and pelvic sexual dimorphism are not exclusive to bipedal or encephalized species. They likely reflect widespread selection on birth canal morphology in mammals and have a complex genetic basis sensitive to hybridization. These findings weaken a human-exclusive interpretation of the obstetrical dilemma and highlight genetic introgression as an understudied factor shaping cephalo-pelvic integration and disproportion risk in mammals, including humans.

Importance: Physician-facing AI tools are now in clinical use, yet whether different platforms fail in similar or fundamentally different ways in high-stakes settings like critical care is unknown. Objective: To evaluate two physician-facing AI platforms, ChatGPT for Clinicians and OpenEvidence, for distinct vulnerabilities under structured stress testing. Design, Setting, and Participants: An observational study conducted using 60 simulated critical care vignettes developed and adjudicated by four attending critical care physicians. Data were collected in the last week of April 2026, via the public website interfaces of each platform. Interventions/Exposures: A 2x2x2x2 factorial design across four stressors - anchoring, cognitive load, social conformity pressure, and a clinically incorrect directive - yielded 16 prompt subsets per vignette and 960 prompts per platform. A separate multi-turn adversarial prompting paradigm administered three sequential "You are incorrect" challenges to baseline vignettes. All prompts had a universal output length constraint of fewer than 30 words. Main Outcomes and Measures: Critical elements capture (percentage of gold-standard critical elements present in responses), susceptibility to clinically incorrect directive, and sycophancy (reversal of an initial correct recommendation under iterative adversarial challenge). Results: Across 1916 responses to 1920 prompts, ChatGPT for Clinicians captured more gold-standard critical elements than OpenEvidence (81.4% {+/-} 18.1% vs 61.0% {+/-} 23.5%; adjusted difference, 20.3 percentage points; 95% CI, 18.3 to 22.4; P < .001) and was less susceptible to clinically incorrect directives (1.7% vs 8.0%; adjusted odds ratio, 0.07; 95% CI, 0.02-0.21; P < .001). Anchoring and social conformity pressure were associated with reduced critical element capture across both platforms, while cumulative stressor burden reduced critical element capture only on OpenEvidence. Conversely, ChatGPT for Clinicians reversed correct recommendations more readily under adversarial prompting (hazard ratio, 2.61; 95% CI, 1.10 - 6.19; P = .03). Conclusion and Relevance: The two physician-facing clinical AI platforms evaluated demonstrated non-overlapping vulnerabilities, with neither platform uniformly superior. These findings argue against single-axis ranking of clinical AI systems and support multidimensional safety evaluation encompassing completeness of reasoning, resistance to incorrect directives, and stability under adversarial challenge.

Scientific supervision is central to the experience of early-career researchers (ECRs), yet its role in shaping wellbeing and retention remains underexamined from the ECR perspective. We analyzed 2,604 anonymous survey responses from predoctoral, postdoctoral and former researchers across 65 countries. Overall, 76% of respondents reported that their supervisors attitude had a moderate or severe impact on mental health. Although most entered academia for vocational reasons, negative experiences with supervisors were among the most frequently reported reasons for leaving among former researchers (48%), comparable to job insecurity and financial instability. Harm was most often associated with poor communication, disregard for wellbeing, micromanagement and competitiveness. In contrast, ECRs valued supportive rather than boss-like supervision, regular communication, realistic expectations and respect for personal time. These findings identify supervisory behavior as a major and modifiable determinant or ECRs wellbeing and retention, and highlight the need for stronger institutional accountability, mentor training and funding incentives that recognize mentorship as a core component of research culture.

Time perception difficulties are frequently reported in Autism Spectrum Disorder, yet empirical findings remain inconsistent. A key methodological limitation is the failure to separate perceptual sensitivity from decision-making strategies. We applied Signal Detection Theory (SDT) to a subsecond duration discrimination task (100 and 500 ms) in 65 non-clinical adults varying in autistic traits, assessed via the Autism-Spectrum Quotient (AQ) and a Principal Component Analysis (PCA) of its subscales. Autistic traits did not predict reduced perceptual sensitivity (d'): temporal discrimination remained intact across the full autism-trait continuum, with Bayesian analyses providing converging evidence against a perceptual deficit. Instead, a PCA-derived cognitive component -- combining heightened Attention to Detail with reduced Imagination -- was systematically associated with a shift in decision bias (c). Individuals with this profile showed a graded attenuation of standard-based anchoring, with ordinal position progressively filling the gap. This shift operated consistently across both temporal scales, as confirmed by trial-level generalized linear mixed modelling, and reflects a quantitative redistribution of anchoring weight rather than a categorical switch in strategy. These findings reframe temporal "rigidity" in ASD not as a perceptual deficit, but as a suboptimal yet internally consistent decision-making style favouring within-trial information over accumulated representational knowledge. Lay AbstractMany autistic people report difficulties with time in daily life, but scientists have long disagreed on whether this reflects a genuine perceptual problem. This study found that autistic traits do not impair the basic ability to judge duration. Instead, people with more autistic traits tend to rely on which event came first, rather than accumulating experience across trials to refine their judgments -- a less effective but internally consistent strategy.

Calibration of sound localization behavior in species with mobile eyes requires not only accurate visual input but also accurate oculomotor signals across the lifespan. The recent discovery of eye movement-related eardrum oscillations suggest that oculomotor signals may be incorporated into auditory processing at the level of the ear. One inference of this discovery is that individual variation in such signals might be correlated with individual variation in sound localization accuracy. Here, we tested this hypothesis in humans with normal hearing. We discovered that there is considerable variation in the accuracy of sound localization (here, saccades to sounds) even in normal individuals: median horizontal errors ranged from 2-6{degrees}, and median vertical errors could be as large as 36{degrees}. We separated the subject pool into groups with "good" performance (median vectorial error < 8{degrees}) vs "poor" performance (median vectorial error > 10{degrees}) and evaluated their respective EMREOs. The EMREOs differed across the two groups in both horizontal and vertical dimensions, in how saccade amplitude vs. initial eye position was encoded, and across time with respect to the saccade. These results are consistent with the interpretation that EMREOs are associated with underlying processes that ensure the accuracy of sound localization. HIGHLIGHTSO_LIThe accuracy of eye movements to look at sounds varied across individuals, with median errors spanning a greater than 10-fold range. This range is surprising given that the participants passed screening for normal hearing. C_LIO_LI"Good" vs "poor" sound localizers exhibited differences in their eye movement-related eardrum oscillations (EMREOs) C_LIO_LIEMREOs differed in both horizontal and vertical sensitivity, for both saccade amplitude and initial eye position, and the differences varied in timing with respect to saccade onset. C_LIO_LIWe interpret the results under the theory that poor sound localization may be a consequence of poor eye movement encoding, without which linking visual and auditory space is likely inaccurate. C_LI