AbstractThe relation between the fundamental frequency of the voice (f0) and vertical space has been shown in previous studies; however, the underlying mechanisms are less clear. This study investigates the relationship between head angle and f0 in iconic prosody, along with the influence of object size on lip opening and formant frequencies. In the experiment, participants pointed to objects of two different sizes and in various vertical positions while saying the words “piff” or “paff,” which induced vertical head position change. Head angle emerged as a reliable predictor of f0, with a larger angle increasing the f0. This effect was consistent despite individual variations in head movement. While the vertical position of the object also showed a reliable effect on f0, head angle substantially outperformed it as a predictor, suggesting that head angle represents the primary physiological mechanism predicting f0 changes. Conversely, object size did not predict either lip opening or formant dispersion. Lip opening and formant dispersion were purely indexical, tracking vowel-specific articulatory configurations rather than external object properties. These findings underscore the role of head position in modulating f0 through direct physiological coupling, potentially underpinning iconic prosody, while revealing the limits of size-related iconicity in parameters constrained by phonemic requirements.

AbstractThe relation between the fundamental frequency of the voice (f0) and vertical space has been shown in previous studies; however, the underlying mechanisms are less clear. This study investigates the relationship between head angle and f0 in iconic prosody, along with the influence of object size on lip opening and formant frequencies. In the experiment, participants pointed to objects of two different sizes and in various vertical positions while saying the words “piff” or “paff,” which induced vertical head position change. Head angle emerged as a reliable predictor of f0, with a larger angle increasing the f0. This effect was consistent despite individual variations in head movement. While the vertical position of the object also showed a reliable effect on f0, head angle substantially outperformed it as a predictor, suggesting that head angle represents the primary physiological mechanism predicting f0 changes. Conversely, object size did not predict either lip opening or formant dispersion. Lip opening and formant dispersion were purely indexical, tracking vowel-specific articulatory configurations rather than external object properties. These findings underscore the role of head position in modulating f0 through direct physiological coupling, potentially underpinning iconic prosody, while revealing the limits of size-related iconicity in parameters constrained by phonemic requirements.

AbstractConsensus-based social learning strategies often outcompete other strategies in evolutionary models. But while formal proofs suggest that consensus’ reliability is compromised when individual judgments are not independent, this makes for a notoriously implausible assumption in the biological world: the people we learn from are constantly learning from each other as well. How do we avoid being misled by consensus? We present three experiments and a computational model examining commonsense reasoning about how people’s public and private judgments are influenced by the consensus and social status of those around them. Results suggest that while people realize that these two factors can cause others’ public and private judgments to diverge, their own trust in public consensus depends on how accurately they believe it reflects their informants’ true beliefs.

AbstractWithin minutes of birth a newborn gnu or giraffe works to stand and walk, asserting postural balance and organised animate behaviour in an apparently goal-directed manner. In contrast, robots learning to stand and walk from scratch begin with random flailing, the behaviour cohering over time as the robot internalises some reward/value signal. How does the newborn gnu ‘innately know’ what goal to aim for, and decide to work towards it? How could similar goal-directed balance learning be implemented in robots? Currently, animate balance inherits its axiomatic definition from the Newtonian formulation for inanimate balance; static mechanical equilibrium. This is arguably inappropriate for animate balance, because animals need to move and are never in static mechanical equilibrium, giving rise to the ‘posture-movement paradox’. The present Perspective proposes a more fluid, dynamical axiomatic task definition and goal which (a) isolates resisting gravity, (b) admits and enables movement, and (c) subsumes static mechanical equilibrium as a special case. This novel definition is founded upon inevitable biophysical requirements and observable developmental process. The article explains how animals apprehend and embed this goal through prenatal development suspended in equidense amniotic fluid, and then are challenged to self-maintain it by the perinatal transition. The account entails a paradigmatic shift in putative physiological organisation and associated conceptual framework for balance; from a subsidiary sensorimotor control task to a vital mechano-regulation task, organisationally akin to thermo-regulation. This vital mechano-regulation model of balance has practical implications and implies a range of predictions.

AbstractConsensus-based social learning strategies often outcompete other strategies in evolutionary models. But while formal proofs suggest that consensus’ reliability is compromised when individual judgments are not independent, this makes for a notoriously implausible assumption in the biological world: the people we learn from are constantly learning from each other as well. How do we avoid being misled by consensus? We present three experiments and a computational model examining commonsense reasoning about how people’s public and private judgments are influenced by the consensus and social status of those around them. Results suggest that while people realize that these two factors can cause others’ public and private judgments to diverge, their own trust in public consensus depends on how accurately they believe it reflects their informants’ true beliefs.

AbstractWithin minutes of birth a newborn gnu or giraffe works to stand and walk, asserting postural balance and organised animate behaviour in an apparently goal-directed manner. In contrast, robots learning to stand and walk from scratch begin with random flailing, the behaviour cohering over time as the robot internalises some reward/value signal. How does the newborn gnu ‘innately know’ what goal to aim for, and decide to work towards it? How could similar goal-directed balance learning be implemented in robots? Currently, animate balance inherits its axiomatic definition from the Newtonian formulation for inanimate balance; static mechanical equilibrium. This is arguably inappropriate for animate balance, because animals need to move and are never in static mechanical equilibrium, giving rise to the ‘posture-movement paradox’. The present Perspective proposes a more fluid, dynamical axiomatic task definition and goal which (a) isolates resisting gravity, (b) admits and enables movement, and (c) subsumes static mechanical equilibrium as a special case. This novel definition is founded upon inevitable biophysical requirements and observable developmental process. The article explains how animals apprehend and embed this goal through prenatal development suspended in equidense amniotic fluid, and then are challenged to self-maintain it by the perinatal transition. The account entails a paradigmatic shift in putative physiological organisation and associated conceptual framework for balance; from a subsidiary sensorimotor control task to a vital mechano-regulation task, organisationally akin to thermo-regulation. This vital mechano-regulation model of balance has practical implications and implies a range of predictions.

AbstractPart of language’s great expressivity comes from its users creating new forms by applying familiar rules to novel items. But linguistic rules aren’t all created equal—some are more readily generalisable than others. In this paper, we focus on how rule generalisation is affected by certain properties of frequency distributions. In an artificial language learning experiment that asks adult learners to generalise using one of two suffixes, we find that they probability-match their input but slightly prefer whichever suffix they encountered with more low-frequency stems. Then with an urn model of learning, we show that previous explanations of generalisation that focus only on a distribution’s type count or its skew fail to capture participants’ behaviour—only the low-frequency preference yields convergent results. We model learners’ behaviour in terms of rational Bayesian inference about how likely a rule is to apply to more word types than somebody has already encountered. Overall, we suggest that linguistic rule generalisation is a self-sustaining process: by creating novel and therefore low-frequency items, rule generalisation produces the very same distributional structure that feeds it.

AbstractA Ganzfeld is a homogeneous visual field, devoid of any focal points. Such a stimulus has been used by researchers to study perceptual phenomena in the absence of changes in sensory structure. Others have used it to study altered states of consciousness (ASCs). Until now, these different facets have been studied separately with little attention for the emotional subjective experience. This study aimed to elucidate the perceptual, phenomenal, and emotional experience of the multifaceted Ganzfeld using a multi-method approach combining behavioral (eye-tracking) and neural (electroencephalography; EEG) measures, with qualitative (interviews) and quantitative (questionnaires) assessments. We show that Ganzfeld spaces induce ASCs and offer immersive, full-body experiences, including bodily effects. Our results pertaining to bodily sensations further prompted us to identify a perceptually grounded cognitive processing type with either an inward-directed or externally-directed focus. We also identified the presence of an abstract cognitive processing type characterized by an introspective focus and meditative experiences. At the behavioral level, decays were characterized by decreased eye movements. The lag in reporting decays and the subjective experience of decays point to the notion of mind blanking. At the neural level, we found increased theta activity preceding decays, further hinting at a potential interrelation between perceptual decays and mind blanking. Finally, decays were characterized by more alpha activity, a pattern often associated with attenuated sensory processing and states of reduced external engagement (Jensen & Mazaheri, 2010), such as relaxation. Our findings contribute to a more in-depth understanding of all the components contributing to the rich Ganzfeld experiences.

AbstractCognitive scientists have discovered normative and heuristic principles that capture human subgoal preferences when partitioning problems into smaller ones. However, it remains unclear where such preferences come from and why they tend to be both effective and efficient. In this work, we study the processes through which these preferences may be implicitly encoded over learning as learners improve towards optimal traversals. We build on the graph-based environments from prior work and use neural networks as model learners to test if learning shortest-path traversal can lead to human-like path decomposition. We find that simple transformer models develop a preference for paths containing nodes that occur frequently on the shortest paths, consistent with human subgoal preferences found in prior work. This preference is observed when models solve shortest path traversals for unseen problems in both known graphs and new graphs, demonstrating that human-like subgoal preferences can arise without requiring explicit preference computation or exhaustively searching over all possible paths. The same preference does not emerge when models learn to perform random or Hamiltonian traversals. Our findings are robust across several transformer variants as well as recurrent neural networks, suggesting they depend more on the data distribution than the network architecture.

AbstractPart of language’s great expressivity comes from its users creating new forms by applying familiar rules to novel items. But linguistic rules aren’t all created equal—some are more readily generalisable than others. In this paper, we focus on how rule generalisation is affected by certain properties of frequency distributions. In an artificial language learning experiment that asks adult learners to generalise using one of two suffixes, we find that they probability-match their input but slightly prefer whichever suffix they encountered with more low-frequency stems. Then with an urn model of learning, we show that previous explanations of generalisation that focus only on a distribution’s type count or its skew fail to capture participants’ behaviour—only the low-frequency preference yields convergent results. We model learners’ behaviour in terms of rational Bayesian inference about how likely a rule is to apply to more word types than somebody has already encountered. Overall, we suggest that linguistic rule generalisation is a self-sustaining process: by creating novel and therefore low-frequency items, rule generalisation produces the very same distributional structure that feeds it.

AbstractA Ganzfeld is a homogeneous visual field, devoid of any focal points. Such a stimulus has been used by researchers to study perceptual phenomena in the absence of changes in sensory structure. Others have used it to study altered states of consciousness (ASCs). Until now, these different facets have been studied separately with little attention for the emotional subjective experience. This study aimed to elucidate the perceptual, phenomenal, and emotional experience of the multifaceted Ganzfeld using a multi-method approach combining behavioral (eye-tracking) and neural (electroencephalography; EEG) measures, with qualitative (interviews) and quantitative (questionnaires) assessments. We show that Ganzfeld spaces induce ASCs and offer immersive, full-body experiences, including bodily effects. Our results pertaining to bodily sensations further prompted us to identify a perceptually grounded cognitive processing type with either an inward-directed or externally-directed focus. We also identified the presence of an abstract cognitive processing type characterized by an introspective focus and meditative experiences. At the behavioral level, decays were characterized by decreased eye movements. The lag in reporting decays and the subjective experience of decays point to the notion of mind blanking. At the neural level, we found increased theta activity preceding decays, further hinting at a potential interrelation between perceptual decays and mind blanking. Finally, decays were characterized by more alpha activity, a pattern often associated with attenuated sensory processing and states of reduced external engagement (Jensen & Mazaheri, 2010), such as relaxation. Our findings contribute to a more in-depth understanding of all the components contributing to the rich Ganzfeld experiences.

AbstractCognitive scientists have discovered normative and heuristic principles that capture human subgoal preferences when partitioning problems into smaller ones. However, it remains unclear where such preferences come from and why they tend to be both effective and efficient. In this work, we study the processes through which these preferences may be implicitly encoded over learning as learners improve towards optimal traversals. We build on the graph-based environments from prior work and use neural networks as model learners to test if learning shortest-path traversal can lead to human-like path decomposition. We find that simple transformer models develop a preference for paths containing nodes that occur frequently on the shortest paths, consistent with human subgoal preferences found in prior work. This preference is observed when models solve shortest path traversals for unseen problems in both known graphs and new graphs, demonstrating that human-like subgoal preferences can arise without requiring explicit preference computation or exhaustively searching over all possible paths. The same preference does not emerge when models learn to perform random or Hamiltonian traversals. Our findings are robust across several transformer variants as well as recurrent neural networks, suggesting they depend more on the data distribution than the network architecture.

AbstractThe sperm whale communication system, consisting of groups of clicks called codas, has been primarily analyzed in terms of the number of clicks and their inter-click timing. This paper reports spectral properties in sperm whale vocalizations and demonstrates that spectral properties are highly structured, discretely distributed across codas, and uttered in dialogues, rather than being a physical artefact of whale movement. We report formant structure in whale codas and uncover previously unobserved spectral patterns. We argue that these spectral properties freely combine with the traditionally analyzed properties. We present a visualization technique that allows the description of several previously unobserved patterns. Codas are on many levels analogous to human vowels and diphthongs and can be conceptualized in terms of the source-filter theory: vowel duration and pitch correspond to the number of clicks and their timing (traditional coda types), while spectral properties of clicks correspond to formants in human vowels. We identify two recurrent and discrete coda-level spectral patterns that appear across individual sperm whales and across traditional coda types: the a- and i-coda vowels. We also report that sperm whales have diphthongal patterns on individual codas: with rising, falling, rising-falling and falling-rising formant patterns observed. These uncovered patterns suggest that spectral properties have the potential to add to the communicative complexity of codas independent of the traditionally analyzed properties and add a new dimension to the study of a cetacean communication system.

AbstractMost deaf and hard-of-hearing children are born to hearing parents, often delaying exposure to their first language. This negatively influences development of not only language, but also many other aspects of cognition, including exact representations of large quantities. The core knowledge view of numeracy predicts that delays in language exposure should not affect nonverbal representations of small quantities (1–3). This study is the first to investigate effects of language modality (spoken vs. signed) and timing of language experience (early, from birth vs. later) on the representation of small quantities of objects. We adapted the “Mr. Elephant” task (Shusterman et al., 2017) and examined whether children (age 3 to 7 years) succeeded on trials involving quantities 2 and 3. A logistic regression found that Timing and Socioeconomic Status significantly predicted Mr. Elephant performance, while Modality and Age did not. Early-exposed children were more likely to succeed on the task than Later-exposed children. For an exploratory follow-up, two measures of language were added into the analysis: Highest Count, which records children’s recitation of the count list, and Give-a-Number (‘Give-N’), which assesses children’s understanding of the cardinal principle (CP). This logistic regression found that Timing and Give-N performance significantly and independently predicted Mr. Elephant performance, but Socioeconomic Status and Highest Count did not. Children who were CP-knowers were more likely to succeed on Mr. Elephant than non-CP-knowers. These results suggest that the representation of small quantity representations is associated with the timing of children’s language exposure and their knowledge of the cardinal principle.

AbstractMetacognition comprises the ability to differentiate the accuracy of predictions about the world. This is often called Type 2 performance (with Type 1 performance being the overall accuracy). Typical measures of metacognition are based on signal detection theory and require the strong assumption of truncated normal noise underlying confidence ratings. To minimize distributional assumptions, measures based on classical information theory have been proposed. We further this approach by providing bounds on its key quantity, the transmitted information. We show that classifiers making predictions with a certain accuracy can transmit information only within a limited range, depending on the underlying noise distribution: The lowest transmitted information indicates the worst Type 2 performance and corresponds to binary noise; the highest transmitted information indicates the best Type 2 performance and corresponds to uniform noise. Because normal noise is only an intermediate case, traditional measures based on this assumption can bias interpretations of Type 2 performance. Based on these bounds, we suggest a new measure: Relative metainformation (RMI). RMI scales from 0 (lower bound) to 1 (upper bound) and therefore advances towards the much-needed decoupling of Type 2 from Type 1 performance measures. To demonstrate the strengths of RMI, we apply it to groups: In a setting where multiple independent group members with fixed accuracies combine their predictions in an optimal way, we show that the group performance depends directly on RMI: Group accuracy is best vs. worst if the group members have highest vs. lowest RMI values. Overall, our theoretical bounds allow to better evaluate measures of Type 2 and group performance.

AbstractMany empirical studies have found a curvilinear (inverted-U) relationship between knowledge and curiosity, such that curiosity is induced when stimuli are neither unknown nor too familiar. While various theoretical accounts have been proposed to explain this phenomenon, no clear link between them have been delineated. In this Perspective, we review seven psychological accounts of the inverted-U relationship between knowledge and curiosity (“the U”) and provide a coherent framework integrating them. According to this framework, the U emerges as a consequence of the imperative to pursue learning progress and thus maximize knowledge. We show that some theories of curiosity address this issue by explicitly stipulating knowledge maximization as the computational objective, and learning-progress maximization as an optimal means of achieving it (i.e., normative theories). Other theories focus on psychological mechanisms or factors that drive curiosity (i.e., process theories). We propose that these process-theoretic mechanisms could also work in a manner that maximizes learning by signaling situations in which some relevant prior knowledge exists, but is incomplete. The implications of this framework for future theoretical work on curiosity and its connections to related phenomena are discussed.

AbstractThe sperm whale communication system, consisting of groups of clicks called codas, has been primarily analyzed in terms of the number of clicks and their inter-click timing. This paper reports spectral properties in sperm whale vocalizations and demonstrates that spectral properties are highly structured, discretely distributed across codas, and uttered in dialogues, rather than being a physical artefact of whale movement. We report formant structure in whale codas and uncover previously unobserved spectral patterns. We argue that these spectral properties freely combine with the traditionally analyzed properties. We present a visualization technique that allows the description of several previously unobserved patterns. Codas are on many levels analogous to human vowels and diphthongs and can be conceptualized in terms of the source-filter theory: vowel duration and pitch correspond to the number of clicks and their timing (traditional coda types), while spectral properties of clicks correspond to formants in human vowels. We identify two recurrent and discrete coda-level spectral patterns that appear across individual sperm whales and across traditional coda types: the a- and i-coda vowels. We also report that sperm whales have diphthongal patterns on individual codas: with rising, falling, rising-falling and falling-rising formant patterns observed. These uncovered patterns suggest that spectral properties have the potential to add to the communicative complexity of codas independent of the traditionally analyzed properties and add a new dimension to the study of a cetacean communication system.

AbstractRobots express a great deal of diverse human-like capacities, ranging from communicating in natural languages to displaying emotions to responding to physical touch. Here we examined the role of different kinds of mental capacities on children’s evaluations of, and trust in, robots. We presented 6- to 9-year-olds with identical-looking humanoid robots described as having one (or none) of the following capacities: cognitive-perceptual, social-emotional, or physiological. Across three studies (N = 287), we found that children differentially evaluated (Studies 1A and 1B) and selectively trusted (Study 2) robots with different types of minds. The diverging evaluations (i.e., of benevolence, intelligence, affinity, and epistemic appeal) of robots with different minds emerged between ages 7 and 8 and became stronger with age. Moreover, these differences translated into selective trust choices: children trusted robots with cognitive-perceptual capacities over robots with social-emotional capacities in a factual, but not a social, context, and over robots with bodily capacities across both contexts. Altogether, these findings open avenues for future interdisciplinary research on children’s reasoning about emerging technologies.

AbstractRobots express a great deal of diverse human-like capacities, ranging from communicating in natural languages to displaying emotions to responding to physical touch. Here we examined the role of different kinds of mental capacities on children’s evaluations of, and trust in, robots. We presented 6- to 9-year-olds with identical-looking humanoid robots described as having one (or none) of the following capacities: cognitive-perceptual, social-emotional, or physiological. Across three studies (N = 287), we found that children differentially evaluated (Studies 1A and 1B) and selectively trusted (Study 2) robots with different types of minds. The diverging evaluations (i.e., of benevolence, intelligence, affinity, and epistemic appeal) of robots with different minds emerged between ages 7 and 8 and became stronger with age. Moreover, these differences translated into selective trust choices: children trusted robots with cognitive-perceptual capacities over robots with social-emotional capacities in a factual, but not a social, context, and over robots with bodily capacities across both contexts. Altogether, these findings open avenues for future interdisciplinary research on children’s reasoning about emerging technologies.

AbstractHuman self-beliefs hinge on social feedback, but their formation and revision are not solely based on new information. Biases during learning, such as confirming initial expectations, can lead to inaccurate beliefs. This study uses computational modeling to explore how initial expectations about one’s own and others’ abilities and confidence in these beliefs affect processes of belief formation and belief revision in novel behavioral domains. In the first session, participants formed performance beliefs through trial-by-trial feedback. In the second session, feedback contingencies were reversed to promote a revision of beliefs. Results showed that people form and revise beliefs in a confirmatory manner, with lower initial expectations being linked to more negatively biased belief formation and revision, while growing confidence strengthened these beliefs over time. Once formed, these beliefs proved resistant to change even when faced with contradictory feedback. The findings suggest that newly formed beliefs become entrenched and resistant to new, contradictory information in a short period of time. Understanding how self-beliefs are formed, the role that confidence plays in this process, and why established beliefs are difficult to revise can inform the development of interventions aimed at promoting more adaptive learning in educational, clinical, and social contexts.

AbstractBoth human children and animals seek information following a violation-of-expectation event, but little research suggests the latter do so for the sake of it. In this preregistered experiment, we compared epistemic curiosity—the pursuit of information for its own sake—in kea parrots (Nestor notabilis) and three-year-old human children (Homo sapiens) following a violation-of-expectation event. Subjects were trained to push a tool into an apparatus that produced a reward before the apparatus was surreptitiously made non-functional in following trials. In both functional and non-functional trials, after solving the task, subjects were rewarded and allowed to explore the apparatus for thirty seconds with the opportunity to peek into the side of the apparatus. We found that relatively more kea peeked than children, but the children and not the kea were significantly more likely to peek in the non-functional versus functional trials, particularly when the researcher was absent. While both species showed markers of curiosity in the experiment, we found expectancy-violation-induced epistemic curiosity only in the children and not the kea in this context.

AbstractDevelopmental language models (DLMs) aim to replicate the efficiency of child language acquisition but often focus solely on the estimation of exogenous linguistic input. We argue that a child’s linguistic growth is also critically shaped by endogenous processes, including (1) co-opting language in non-linguistic perception and cognition, (2) engaging in private and inner speech, and (3) benefiting from neural replay of linguistic information during sleep. These endogenous processes amplify and refine exogenous linguistic input in ways that current DLMs do not replicate. To align DLMs with child language acquisition, we propose redefining “linguistic exposure“ to encompass both exogenous and endogenous linguistic input. By integrating label feedback, self-generated speech, and sleep-like consolidation, researchers can narrow the gap between artificial and human learning. Collaborations across machine learning, psychology, and linguistics will be essential to ground models in empirical data on child behavior and build DLMs that truly reflect the marvel of language acquisition.

AbstractHumans are remarkably efficient at learning new tasks, in large part by relying on the integration of previously learned knowledge. However, research on task learning typically focuses on the learning of abstract task rules on minimalist stimuli, to study behavior independent of the learning history that humans come equipped with (i.e., semantic knowledge). In contrast, several theories suggest that the use of semantic knowledge and labels may help the learning of new task information. Here, we tested whether providing existing, semantically rich task embeddings and response labels allowed for more robust task rule encoding and less (catastrophic) forgetting and interference. Our results show that providing semantically rich task settings and response labels resulted in less task forgetting (Experiment 1), both when using pictorial symbols or words as labels (Experiment 2), or when contrasted with visually matched shape labels without inherent meaning (Experiment 4). Using a subsequent value-based decision-making task and reinforcement learning modeling (Experiment 3), we demonstrate how the learned embedding of novel stimuli in semantically rich, representations, further allowed for a more efficient, feature-specific processing when learning new task information. Finally, using artificial recurrent neural networks fitted to our participants’ task performance, we found that task separation during learning was more predictive of learning and task performance in the semantically rich conditions. Together, our findings show the benefit of using semantically rich task rules and response labels during novel task learning, thereby offering important insights into why humans excel in continual learning and are less susceptible to catastrophic forgetting compared to most artificial agents.

AbstractDespite cross-linguistic diversity in how kin relations map to terminology, there are constraints on which kin may be categorised together. But what are the constraints on kin term variation, and where do they come from? One proposed constraint is internal co-selection—an evolutionary process where terminological changes in one generation of kin co-occur with parallel changes in other generations. This results in kin categories which are predictable on the basis of other kin categories, a property we call predictive structure. To determine the strength of this constraint, we measured the predictive structure of kinship terminology systems from 731 languages. We found that kinship terminologies exhibit a significant degree of predictive structure, and we argue that its prevalence reflects a cognitive pressure for simplicity imposed during the generalisation of known kin categories to new referent types. We tested this claim using an artificial kin term generalisation task. Our results suggest that people do favour predictive structure when generalising from known kin categories to new referents, but that this preference faces interference from other pressures to distinguish kin by features like gender.

AbstractNavigating the social world is guided by remembering which people know each other. Yet, different factors might influence how social relationships are remembered, where people’s shared attributes could distort a social network’s mnemonic representation. Here, we study whether dyadically shared contexts and personality traits impact how people remember relationships in social networks. Through varying levels of network topological complexity, we find the contexts where people know each other are most memorable and that better contextual retrieval predicts relationship recall. In contrast, shared personality traits affect relationship recall differently depending on social network complexity, where shared negatively valenced traits relate to worse relationship recall in the simple network. Subsequent modeling revealed that as networks become more complex, relationships between more centrally positioned individuals that share negatively valenced traits are better recalled compared to less well-connected individuals. These results suggest contextual memory can serve as a scaffold for remembering relationships in a social network, while affective traits’ impact on social network retrievability depends on emotional valence and the individuals involved. More generally, our findings give insight into how the same social network can be represented differently based on one’s past experience.