While the visual world is rich and complex, importantly, it nevertheless contains many statistical regularities. For example, environmental feature distributions tend to remain relatively stable from one moment to the next. Recent findings have shown how observers can learn surprising details of environmental color distributions, even when the colors belong to actively ignored stimuli such as distractors in visual search. Our aim was to determine whether such learning influences orienting in the visual environment, measured with saccadic eye movements. In two visual search experiments, observers had to find an odd-one-out target. Firstly, we tested cases where observers selected targets by fixating them. Secondly, we measured saccadic eye movements when observers made judgments on the target and responded manually. Trials were structured in blocks, containing learning trials where distractors came from the same color distribution (uniform or Gaussian) while on subsequent test trials, the target was at different distances from the mean of the learning distractor distribution. For both manual and saccadic measures, performance improved throughout the learning trials and was better when the distractor colors came from a Gaussian distribution. Moreover, saccade latencies during test trials depended on the distance between the color of the current target and the distractors on learning trials, replicating results obtained with manual responses. Latencies were slowed when the target color was within the learning distractor color distribution and also revealed that observers learned the difference between uniform and Gaussian distributions. The importance of several variables in predicting saccadic and manual reaction times was studied using random forests, revealing similar rankings for both modalities, although previous distractor color had a higher impact on free eye movements. Overall, our results demonstrate learning of detailed characteristics of environmental color distributions that affects early attentional selection rather than later decisional processes.

Time has an immense influence on our memory. Truncated encoding leads to memory for only the 'gist' of an image, and long delays before recall result in generalized memories with few details. Here, we used crowdsourced scoring of hundreds of drawings made from memory after variable encoding (Experiment 1) and retentions of that memory (Experiment 2) to quantify what features of memory content change across time. We found that whereas some features of memory are highly dependent on time, such as the proportion of objects recalled from a scene and false recall for objects not in the original image, spatial memory was highly accurate and relatively independent of time. We also found that we could predict which objects were recalled across time based on the location, meaning, and saliency of the objects. The differential impact of time on object and spatial memory supports a separation of these memory systems.

Recent findings suggest that it is possible for people to proactively avoid attentional capture by salient distractors during visual search. The results have important implications for understanding the competing influences of top-down and bottom-up factors in visual attention. Nevertheless, questions remain regarding the extent to which apparently ignored distractors are processed. To assess distractor processing, previous experiments have used a probe method in which stimuli are occasionally superimposed on the search display-requiring participants to abort the search and identify the probe stimuli. It has been recently shown that such probe tasks may be vulnerable to decision-level biases, such as a participant's willingness to report stimuli on to-be-ignored items. We report here results from a new method that is not subject to this limitation. In the new method, the non-target search elements, including the salient distractors, contained features that were either congruent or incongruent with the target. Processing of the non-target elements is inferred from the effects of the compatibility of the shared features on judgments about the target. In four experiments using the technique we show that ignored salient distractors are indeed processed less fully than non-target elements that are not salient, replicating the results of earlier studies using the probe methods. Additionally, the processing of the distractors was found to be reduced at least in part at early perceptual or attentional stages, as assumed by models of attentional suppression. The study confirms the proactive avoidance of capture by salient distractors measured without decision-level biases and provides a new technique for assessing the magnitude of distractor processing.

Yousif et al. (2024) have raised a number of pertinent objections to the idea that number adaptation is a straightforward account of the readily-observable aftereffects that affect perceived numerosity. Their criticisms appear well-motivated, but their particular version of the old-news proposal, involving specific dots, may be insufficiently abstract given that adaptation accumulates. Two new experiments are presented that are meant to buttress their critique by (1) confirming their predictions concerning neutral adaptation, and (2) re-evaluating related claims concerning number vs. density comparisons that have been widely accepted. Present behavioral evidence dissociating effects of adapter size, adapter number and adapter density, supports the idea that density adaptation is implicated as a primary source of most phenomenologically-compelling aftereffects of perceived numerosity. Experiment 2 was preregistered on AsPredicted.org. The pre-registration is available at the following link: https://aspredicted.org/PC7_2ZB The full raw data sets for the two experiments reported her are available on OSF at the following links: Experiment 1: https://osf.io/b9qwy/?view_only=73beb62d9c2046c3aa08cdeb96cd5cca Experiment 2: https://osf.io/6ax5j/?view_only=723ceb0b44da47dba99e56db12db02a9.

Adaptation is a ubiquitous property of perceptual systems, increasing sensitivity to change and allowing them to operate over a large dynamic range. The number sense, like most other perceptual systems, is adaptable. Yousif et al. (2024) challenge the existence of number adaptation, offering an alternate explanation that they term the "old news hypothesis". Here we consider the major evidence advanced for their theory and show that, while their predicted effects may reach statistical significance, they are far too small to begin to explain the robust phenomenon of adaptation. We also highlight a series of studies using fMRI, EEG, pupillometry and psychophysical techniques that support the existence of adaption, and are inconsistent with "old news". We conclude that number adaptation, while not fully understood, does indeed exist, and remains an invaluable concept for understanding the number sense.

Motor imagery is accompanied by a subjective multisensory experience. This sensory experience is thought to result from the deployment of internal models developed for the execution and monitoring of overt actions. If so, how is it that motor imagery does not lead to overt execution? It has been proposed that inhibitory mechanisms may prevent execution during imagined actions such as imagined typing. To test this hypothesis, we combined an experimental with a modelling approach. We conducted an experiment in which participants (N = 49) were asked to alternate between overt (executed) and covert (imagined) typing. We predicted that motor inhibition should lead to longer reaction and movement times when the current trial is preceded by an imagined vs. an executed trial. This prediction was borne out by movement times, but not by reaction times. We introduced and fitted an algorithmic model of motor imagery to disentangle potentially distinct inhibitory mechanisms underlying these effects. Results from this analysis suggest that motor inhibition may affect different aspects of the latent activation function (e.g., the shape of the activation function or the motor execution threshold) with distinct consequences on reaction times and movement times. Overall, these results suggest that typing imagery involves the inhibition of motor commands related to typing acts. Preregistration, complete source code, and reproducible analyses are available at https://osf.io/y9a3k/.

Religious beliefs are among the most ubiquitous ideological beliefs in the world and often critical to people's worldview. Nonetheless, there is a great deal of variability in the strength and persistence of such beliefs, both across and within cultures. Here, we are interested in what underlying cognitive processes are associated with the phenomena of religious belief change. Although previous research has linked the tendency to engage in analytic thinking with religious dis-belief, this work has missed the potentially larger relationship between analytic thinking and belief change more broadly - that is change in any/either direction over time. Using a cross-sectional correlational study across two large datasets, including 16 countries and 50,827 individuals, we found that roughly 25 % of individuals indicated having substantively changing their beliefs at least once. Further, the relationship between analytic thinking and belief change appears independent from the association between analytic thinking and reported level of belief. Therefore, although analytic thinking is generally associated with a decrease in religious belief, we find some evidence that it may also support an increase in belief among those indicating past change. In total, this work provides evidence for a robust link between analytic thinking and religious belief change over time.

Prior work has suggested that visual working memory as measured in change detection tasks can be based on recollection, whereby participants consciously identify a specific feature of a stimulus that has changed, or on familiarity, whereby participants sense that a change has occurred but are unable to consciously access what has changed. Whether recollection and familiarity also contribute to auditory working memory is unclear. The present study aims to address that gap in knowledge by having participants make confidence judgments in change detection tests for speech sounds and pure tones. The results indicated that both recollection and familiarity contribute to auditory working memory across a variety of conditions, and showed that these two processes are functionally dissociable. With speech sounds, subjects were better able to detect syllable changes compared to tone or location changes, and this benefit reflected a selective increase in recollection rather than familiarity. Moreover, for pure tones, both recollection and familiarity also contributed to performance, but recollection was found to be selectively eliminated under stimulus-limited test conditions (i.e., noise-masked, brief dichotic presentations). The results indicate that recollection and familiarity contribute to auditory working memory in a manner that is functionally similar to that observed in visual working memory.

Sequential decision-making, where choices are made one after the other, is an important aspect of our daily lives. For example, when searching for a job, an apartment, or deciding when to buy or sell a stock, people often have to make decisions without knowing what future opportunities might arise. These situations, which are known as optimal stopping problems, involve a risk associated with the decision to either stop or continue searching. However, previous research has not consistently found a clear connection between individuals' search behavior in these tasks and their risk preferences as measured in controlled experimental settings. In this paper, we explore how particular characteristics of optimal stopping tasks affect people's choices, extending beyond their stable risk preferences. We find that (1) the way the underlying sampling distribution is presented (whether it is based on experience or description), (2) the sequential presentation of options, and (3) the unequal frequencies of choices to reject versus to accept significantly bias people choices. These results shed light on the complex nature of decisions that unfold sequentially and emphasize the importance of incorporating context factors when studying human decision behavior.

The prior distribution of values for a specific feature can be categorized as long- or short-term priors based on their respective learning durations. Studies have demonstrated that the visual system can integrate both priors through weighted averaging and then utilize the integrated prior to efficiently encode stimuli. It is unclear what determines the two priors' relative weights. To address this question, we arranged the orientations according to three distributions: natural, anti-natural, and natural with increased-amplitude distributions. The natural distribution mirrors the distribution of orientations in the natural world, so it does not conflict with the long-term prior; according to the Kullback-Leibler divergence analysis, the natural distribution had a higher conflict with the anti-natural distribution than with the natural distribution with increased amplitude. It was found that the cardinal bias - the orientation estimates are biased away from the cardinal orientations - was strongest in the natural distribution with increased amplitude but weakest in the anti-natural distribution. These results were accurately predicted by an efficient Bayesian observer model in which the prior is the weighted integration of the long- and short-term priors. Importantly, the weight of the short-term prior in the new prior decreased as the level of the conflict between the long- and short-term priors increased. Therefore, this study reveals that the visual system integrates the long- and short-term priors through weighted averaging, with the conflicting level between the two priors determining their relative weights in the integration prior. The integrated prior was used by visual systems to efficiently encode stimuli.

Emerging evidence suggests that children may think of robots-and artificial intelligence, more generally-as having moral standing. In this paper, we trace the developmental trajectory of this belief. Over three developmental studies (combined N = 415) and one adult study (N = 156), we compared participants' judgments (Experiments 1-3) and donation choices (Experiment 4) towards a human boy, a humanoid robot, and control targets. We observed that, on the whole, children endorsed robots as having moral standing and mental life. With age, however, they tended to deny experiential mental life to robots, which aligned with diminished ascription of moral standing. Older children's judgments more closely mirrored those of adult participants, who overwhelmingly denied these attributes to robots. This sheds new light on children's moral cognitive development and their relationship to emerging technologies.

Cognitive and social factors can deteriorate eyewitness identification performance in children and older adults. An identification procedure that mitigates the effect of such factors could be beneficial for child and older adult witnesses. In a field experiment, we mapped identification performance in a large community sample (N = 1239) across the lifespan (ages 6-79 years) for two different identification procedures: classic lineups and reaction time-based Concealed Information Test (RT-CIT). Visitors of a science museum or science fair witnessed a recorded mock theft and then took either a classic lineup, or the RT-CIT. Young adults (18-35-year-olds) outperformed younger and older age groups in lineup performance. The RT-CIT showed a moderate capacity to diagnose face recognition and absence of recognition in the target-absent condition. Age did not affect identification with the RT-CIT. However, children were often not able to follow the RT-CIT instructions, leading to a large number of exclusions. A direct comparison of lineup vs. RT-CIT performance showed that children and adolescents showed better identification performance in RT-CIT than lineups. For young adults, there was no difference between the two procedures. The trend turned around at mid-adult age who showed better identification performance when they were given a lineup compared to an RT-CIT. These findings suggest that the RT-CIT might be considered an alternative identification procedure for children and adolescents, offering protection for innocent suspects.

How are people able to understand everyday physical events with such ease? One hypothesis suggests people use an approximate probabilistic simulation of the world. A contrasting hypothesis is that people use a collection of abstractions or features. While it has been noted that the two hypotheses explain complementary aspects of physical reasoning, there has yet to be a model of how these two modes of reasoning can be used together. We develop a "blended model" that synthesizes the two hypotheses: under certain conditions, simulation is replaced by a visuo-spatial abstraction (linear path projection). This abstraction purchases efficiency at the cost of fidelity, and the blended model predicts that people will make systematic errors whenever the conditions for applying the abstraction are met. We tested this prediction in two experiments where participants made judgments about whether a falling ball will contact a target. First, we show that response times are longer when straight-line paths are unavailable, even when simulation time is held fixed, arguing against a pure-simulation model (Experiment 1). Second, we show that people incorrectly judge the trajectory of the ball in a manner consistent with linear path projection (Experiment 2). We conclude that people have access to a flexible mental physics engine, but adaptively invoke more efficient abstractions when they are useful.

Wilson et al. (2025) report a failed attempt to replicate the reductive allure effect: Unlike prior work, they do not find that participants judged explanations of scientific phenomena to be higher quality when they contained irrelevant reductive language. The current commentary considers three possible reasons for this failure to replicate: (1) a change in the nature of online study participants, (2) a change in the background knowledge that people bring to judgments of scientific explanations, and (3) a change in the kinds of explanations that people find satisfying.

Understanding how early scene viewing is guided can reveal fundamental brain mechanisms for quickly making sense of our surroundings. Viewing is often initiated from the left side. Across two experiments, we focused on search initiation for lateralised targets within real-world scenes, investigating the role of the cerebral hemispheres in guiding the first saccade. We aimed to disentangle hemispheric contribution from the effects of reading habits and distinguish between an overall dominance of the right hemisphere for visuospatial processing and finer hemispheric specialisation for the type of target template representation (from pictorial versus verbal cues), spatial scale (global versus local), and timescale (short versus longer). We replicated the tendency to initiate search leftward in both experiments. However, we found no evidence supporting a significant impact of left-to-right reading habits, either as a purely motor or attentional bias to the left. A general visuospatial dominance of the right hemisphere could not account for the results either. In Experiment 1, we found a greater probability of directing the first saccade toward targets in the left visual field but only after a verbal target cue, with no lateral differences after a pictorial cue. This suggested a contribution of the right hemisphere specialisation in perceptually simulating words' referents. Lengthening the Inter-Stimulus Interval between the cue and the scene (from 100 to 900 ms) resulted in reduced first saccade gain in the left visual field, suggesting a decreased ability of the the right hemisphere to use the target template to guide gaze close to the target object, which primarily depends on local information processing. Experiment 2, using visual versus auditory verbal cues, replicated and extended the findings for both first saccade direction and gain. Overall, our study shows that the multidetermined functional specialisation of the cerebral hemispheres is a key driver of early scene search and must be incorporated into theories and models to advance understanding of the mechanisms that guide viewing behaviour.

Machines powered by artificial intelligence have the potential to replace or collaborate with human decision-makers in moral settings. In these roles, machines would face moral tradeoffs, such as automated vehicles (AVs) distributing inevitable risks among road users. Do people believe that machines should make moral decisions differently from humans? If so, why? To address these questions, we conducted six studies (N = 6805) to examine how people, as observers, believe human drivers and AVs should act in similar moral dilemmas and how they judge their moral decisions. In pedestrian-only dilemmas where the two agents had to sacrifice one pedestrian to save more pedestrians, participants held them to similar utilitarian norms (Study 1). In occupant dilemmas where the agents needed to weigh the in-vehicle occupant against more pedestrians, participants were less accepting of AVs sacrificing their passenger compared to human drivers sacrificing themselves (Studies 1-3) or another passenger (Studies 5-6). The difference was not driven by reduced occupant agency in AVs (Study 4) or by non-voluntary occupant sacrifice in AVs (Study 5), but rather by the perceived social relationship between AVs and their users (Study 6). Thus, even when people adopt an impartial stance as observers, they are more likely to believe that AVs should prioritize serving their users in moral dilemmas. We discuss the theoretical and practical implications for AV morality.

Face recognition is crucial for social interactions. Traditional approaches primarily rely on subjective judgment, utilizing a pre-selected set of facial features based on literature or intuition to identify critical facial features for face recognition. In this study, we adopted a reverse-correlation approach, aligning responses of a deep convolutional neural network (DCNN) with its internal representations to objectively identify facial features pivotal for face recognition. Specifically, we trained a DCNN, namely VGG-FD, to possess human-like capability in discriminating facial identities. A representational similarity analysis (RSA) was employed to characterize VGG-FD's performance metrics, which was subsequently reverse-correlated with its representations in layers capable of discriminating facial identities. Our analysis revealed a higher likelihood of face pairs being perceived as different identities when their representations significantly differed in areas such as the eyes, eyebrows, or central facial region, suggesting the significance of the eyes as facial parts and the central facial region as an integral of face configuration in face recognition. In summary, our study leveraged DCNNs to identify critical facial features for face discrimination in a hypothesis-neutral, data-driven manner, hereby advocating for the adoption of this new paradigm to explore critical facial features across various face recognition tasks.

The Continued Influence Effect (CIE) is the phenomenon that retracted information often continues to influence judgments and inferences. The CIE is rational when the source that retracts the information (the retractor) is less credible than the source that originally presented the information (the informant; Connor Desai et al., 2020). Conversely, a CIE is not rational when the retractor is at least as credible as the informant. Thus, a rational account predicts that the CIE depends on the relative credibility of informant and retractor. In two experiments (N = 151, N = 146), informant credibility and retractor credibility were independently manipulated. Participants read a fictitious news report in which original information and a retraction were each presented by either a source with high credibility or a source with low credibility. In both experiments, when the informant was more credible than the retractor, participants showed a CIE compared to control participants who saw neither the information nor the retraction (ds > 0.82). When the informant was less credible than the retractor, participants showed no CIE, in line with a rational account. However, in Experiment 2, participants also showed a CIE when informant and retractor were equally credible (ds > 0.51). This cannot be explained by a rational account, but is consistent with error-based accounts of the CIE. Thus, a rational account alone cannot fully account for the complete pattern of results, but needs to be complemented with accounts that view the CIE as a memory-based error.

Interactions between the context in which a sensorimotor skill is learned and the recall of that memory have been primarily studied in limb movements, but speech production requires movement, and many aspects of speech processing are influenced by task-relevant contextual information. Here, in ecologically valid speech (read sentences), we test whether English-French bilinguals can use the language of production to acquire and recall distinct motor plans for similar speech sounds spanning the production workspace. Participants experienced real-time alterations of auditory feedback while producing interleaved English and French sentences. The alterations were equal in magnitude but opposite in direction between languages. Over three experiments (n = 15 in each), we observed language-specific sensorimotor learning in speech that countered the alterations and persisted after the alterations were removed. The effects were not observed in a fourth experiment (n = 15) when the feedback alterations were tied to a non-linguistic cue. In a fifth experiment (n = 15), we provide further confirmation that the observed language-specific changes in speech production were confined to sentence production, the linguistic level at which they were learned. The results contrast with recent work and theories of second language learning that predict broad interference between L1 and L2 phonetic representations. When faced with contrasting sensorimotor demands between languages, bilinguals readily acquire and recall highly specific motor representations for speech.

We examine the impact of partisan language (i.e., language that describes events in a manner that supports a political agenda), both with regard to peoples' perceptions of the speakers who use it and their evaluations of the events it is used to describe. In two experiments, we recruited 1121 Democrats and Republicans from the United States. Using a set of liberal-biased (e.g., expand voting rights) and conservative-biased (e.g., reduce election security) terms, we find that partisans judge speakers describing polarizing events using ideologically-congruent language as more trustworthy than those describing events in a non-partisan way (e.g., expand mail-in voting). However, when presented to rival partisans, ideologically-biased language promoted negative evaluations of opposing partisans, with speakers attributed out-group language being viewed as far less trustworthy than non-partisan speakers. Furthermore, presenting Democrats and Republicans with ideologically-congruent descriptions of political events polarized their attitudes towards the events described. Overall, the present investigation reveals how partisan language, while praised by co-partisans, can damage trust and amplify disagreement across political divides.

Wayfinding, a large-scale spatial ability, involves the navigation of one's environment and can be classified into three types of knowledge: route, landmark, and survey. Newcombe and Shipley (2015) proposed a double-dimension spatial framework that classifies spatial abilities into four categories: intrinsic-static, intrinsic-dynamic, extrinsic-static, and extrinsic-dynamic. These abilities are usually assessed in small-scale environments, such as in desktop settings. Examining the relationship between wayfinding knowledge and small-scale spatial abilities has important theoretical implications for understanding spatial cognition at different environmental scales. It also has practical implications for designing more effective training programs to improve wayfinding skills. However, the existing literature linking the two is limited and mixed, especially in children. The current study utilized the double-dimension framework to examine the relationship between small-scale and large-scale spatial abilities in children. We hypothesized that intrinsic abilities should be associated with landmark knowledge while dynamic abilities should be related to route knowledge. Eight small-scale spatial tasks measuring four spatial categories and one wayfinding task measuring route and landmark knowledge were administered to 171 typically developing children between the ages of four and nine. Dynamic spatial abilities significantly predicted route knowledge and mediated the effects of age and sex on route knowledge. In addition, dynamic and intrinsic-static abilities predicted landmark knowledge and also mediated the effects of age on landmark knowledge. These results showed the associations between small-scale spatial abilities and large-scale spatial abilities were selective and specific yet strong, providing insights into further theoretical advancements in spatial cognition. Practical implications were also discussed.

Prior work has established that laypeople do not consistently treat moral questions as being objectively true or as merely true relative to different perspectives. Rather, these metaethical judgments vary dramatically across moral issues and in response to different social influences. We offer a potential explanation by examining how objectivists and relativists are evaluated in different contexts. We provide evidence for a novel account of metaethical judgments as signaling tolerance or intolerance of disagreement. The social implications of signaling tolerance or intolerance in different contexts may motivate different metaethical judgments. Study 1 finds that relativists are perceived as more tolerant, empathic, having superior moral character, and as more desirable as social partners than objectivists. Study 2 replicates these findings with a within-participants design and also shows that objectivists are perceived as more morally serious than relativists. Study 3 examines evaluations of objectivists and relativists regarding concrete moral issues, finding these results vary across situations of moral agreement and disagreement. Study 4 finds that participants' metaethical stances likewise vary when responding in the way they think would make a person who agrees or disagrees with them evaluate them more positively. However, in Study 5, we find no effect on metaethical judgment of telling participants they will be evaluated by a person who agrees or disagrees with them, which suggests either a failure to induce reputational concerns or a more limited influence of reputational considerations on metaethical judgments, despite strong effects on social evaluation.

A growing body of literature suggests a powerful role of predictions on memory through prediction violation and prediction confirmation. Violation appears to enhance memory for the event violating the prediction, meanwhile, confirmation boosts memory for the predicted event instead. Crucially, however, the effect of prediction by itself has not been identified as it has typically been studied with its violation or confirmation. Here, we demonstrate the power of explicit predictions on memory by isolating it from its direct violation and confirmation. In a series of experiments, participants were presented with a real-world object along with three characters and they predicted which character the object belonged to. Upon prediction, participants received either visual confirmation (predicted character showing the item), visual rebuttal (another character showing the item) or no feedback (none of the characters showing the item) with regard to their prediction. When their memory was tested, participants were more likely to falsely remember that their predicted character showed them the item than the other characters did, even when no feedback was provided. This false memory was not eliminated by visual rebuttal and it was not weakened when participants had a strong item memory. Experiments 2-4 eliminated action (selecting a predicted character) as an alternative explanation and demonstrated that this prediction-based false memory could be modulated through indirect prediction confirmation and rebuttal. Taken together, our findings show that explicit predictions can be sufficient to induce false memory of predicted events that are robust enough to withstand its direct rebuttal.

Are people who agree on something more likely to be right and competent? Evidence suggests that people tend to make this inference. However, standard wisdom of crowds approaches only provide limited normative grounds. Using simulations and analytical arguments, we argue that when individuals make independent and unbiased estimates, under a wide range of parameters, individuals whose answers converge with each other tend to have more accurate answers and to be more competent. In 6 experiments (UK participants, total N = 1197), we show that participants infer that informants who agree have more accurate answers and are more competent, even when they have no priors, and that these inferences are weakened when the informants were systematically biased. In conclusion, we speculate that inferences from convergence to accuracy and competence might help explain why people deem scientists competent, even if they have little understanding of science.

Generic statements (e.g., "Ducks lay eggs") provide generalizations about kinds that can be judged as true, even in the face of exceptions. Although past research has focused on the positive evidence that justifies a generic, little work has explored the role of evidence that does not match the generic claim (e.g., ducks that do not lay eggs). The current studies aim to understand how different types of exceptions may differentially undermine generic claims. In Studies 1 and 2, adults (n = 560) and children ages 5-11 (n = 141) were asked to judge the truth of generic statements about fictitious animal kinds (e.g., Wugs have blue horns). Accompanying each statement was a set of 6 kind members, some of which displayed the target property (e.g., blue horns), and others of which displayed either an alternative property (e.g., red horns), or an absence of the property (e.g., no horns). Study 1 found that adults were less likely to endorse generic statements when non-matching examples displayed an alternative property than when they displayed an absence of the property. Study 2 indicated that children as well as adults were less likely to endorse generic statements when presented with alternative evidence, regardless of the salience of the alternative. Study 3 replicated these findings with a more sensitive task in which adults (n = 120) and children (n = 97) were asked to choose between sets with either alternative or absence evidence. These studies provide the first evidence that children and adults attend to non-matching evidence when making judgments about generic statements, interpret alternative evidence to be stronger counterevidence than absence evidence, and do not use the salience of alternative properties to determine the strength of alternative evidence. We discuss the implications of this work for problematic generic claims in language and thought.

In reply to Magnuson, Crinnion, Luthra, Gaston, and Grubb (2023), we challenge their conclusion that on-line activation feedback improves word recognition. This type of feedback is instantiated in the TRACE model (McClelland & Elman, 1986) as top-down spread of activation from lexical to phoneme nodes. We give two main reasons why Magnuson et al.'s demonstration that activation feedback speeds up word recognition in TRACE is not informative about whether activation feedback helps humans recognize words. First, the same speed-up could be achieved by changing other parameters in TRACE. Second, more fundamentally, there is room for improvement in TRACE's performance only because the model, unlike Bayesian models, is suboptimal. We also challenge Magnuson et al.'s claim that the available empirical data support activation feedback. The data they base this claim on are open to alternative explanations and there are data against activation feedback that they do not discuss. We argue, therefore, that there are no computational or empirical grounds to conclude that activation feedback benefits human spoken-word recognition and indeed no theoretical grounds why activation feedback would exist. Other types of feedback, for example feedback to support perceptual learning, likely do exist, precisely because they can help listeners recognize words.

In recent years, several influential computational models and metrics have been proposed to predict how humans comprehend and process sentence. One particularly promising approach is contextual semantic similarity. Inspired by the attention algorithm in Transformer and human memory mechanisms, this study proposes an "attention-aware" approach for computing contextual semantic relevance. This new approach takes into account the different contributions of contextual parts and the expectation effect, allowing it to incorporate contextual information fully. The attention-aware approach also facilitates the simulation of existing reading models and their evaluation. The resulting "attention-aware" metrics of semantic relevance can more accurately predict fixation durations in Chinese reading tasks recorded in an eye-tracking corpus than those calculated by existing approaches. The study's findings further provide strong support for the presence of semantic preview benefits in Chinese naturalistic reading. Furthermore, the attention-aware metrics of semantic relevance, being memory-based, possess high interpretability from both linguistic and cognitive standpoints, making them a valuable computational tool for modeling eye-movements in reading and further gaining insight into the process of language comprehension. Our approach emphasizes the potential of these metrics to advance our understanding of how humans comprehend and process language.

Human behavior is often assumed to be hierarchically structured, made up of abstract actions that can be decomposed into concrete actions. However, behavior is typically measured as a sequence of actions, which makes it difficult to infer its hierarchical structure. In this paper, we explore how people form hierarchically structured plans, using an experimental paradigm with observable hierarchical representations: participants create programs that produce sequences of actions in a language with explicit hierarchical structure. This task lets us test two well-established principles of human behavior: utility maximization (i.e. using fewer actions) and minimum description length (MDL; i.e. having a shorter program). We find that humans are sensitive to both metrics, but that both accounts fail to predict a qualitative feature of human-created programs, namely that people prefer programs with reuse over and above the predictions of MDL. We formalize this preference for reuse by extending the MDL account into a generative model over programs, modeling hierarchy choice as the induction of a grammar over actions. Our account can explain the preference for reuse and provides better predictions of human behavior, going beyond simple accounts of compressibility to highlight a principle that guides hierarchical planning.

Episodic events are typically retrieved and forgotten holistically. If you recall one element (e.g., a person), you are more likely to recall other elements from the same event (e.g., the location), a pattern that is retained over time in the presence of forgetting. In contrast, representations of individual items, such as objects, may be less coherently bound, such that object features are forgotten at different rates and retrieval dependency decreases across delay. To test the theoretical prediction that forgetting qualitatively differs across levels in a representational hierarchy, we investigated the potential dissociation between event and item memory across five experiments. Participants encoded three-element events comprising images of famous people, locations, and objects. We measured retrieval accuracy and the dependency between the retrieval of event associations and object features, immediately after encoding and after various delays (5 h to 3 days). Across experiments, retrieval accuracy decreased for both events and objects over time, revealing forgetting. Retrieval dependency for event elements (i.e., people, locations, and objects) did not change over time, suggesting the holistic forgetting of events. Retrieval dependency for object features (i.e., state and colour) was more variable. Depending on encoding and delay conditions across the experiments, we observed both fragmentation and holistic forgetting of object features. Our results suggest that event representations remain coherent over time, whereas object representations can, but do not always, fragment. This provides support for our representational hierarchy framework of forgetting, however there are (still to be determined) boundary conditions in relation to the fragmentation of object representations.

How do people hold others responsible? Responsibility judgments are affected not only by what actually happened, but also by what could have happened if things had turned out differently. Here, we look at how replaceability - the ease with which a person could have been replaced by someone else - affects responsibility. We develop the counterfactual replacement model, which runs simulations of alternative scenarios to determine the probability that the outcome would have differed if the person of interest had been replaced. The model predicts that a person is held more responsible, the more difficult it would have been to replace them. To test the model's predictions, we design a paradigm that quantitatively varies replaceability by manipulating the number of replacements and the probability with which each replacement would have been available. Across three experiments featuring increasingly complex scenarios, we show that the model explains participants' responsibility judgments well in both social and physical settings, and better than alternative models that rely only on features of what actually happened.