Recent Submissions

  • Infants’ interpretation of information-seeking actions

    Varga, Bálint; Csibra, Gergely; Kovács, Ágnes Melinda; Department of Cognitive Science (Cognitive Science Society, 2021)
    Although infants can frequently observe others gathering information, it is an open question whether and how they make sense of such activities since the mental causes and intended effects of these are hidden and underdetermined by the available evidence. We tested the hypothesis that infants possess a naive theory that leads them to grasp the purpose of information-gathering actions when they serve as sub-goals of higher-order instrumental goals. We presented 14-month-old infants with actions that were inefficient with respect to the agent’s instrumental goal but could or could not be justified as information-seeking behavior via this theory. We expected longer looks in the condition where the detour could not be justified and the results were in line with our predictions. While this evidence is compatible with our hypothesis, further studies are in progress to rule out alternative interpretations of our findings.
  • Facilitation of object encoding in infants by the observation of giving

    Tatone, Denis; Hernik, Mikołaj; Csibra, Gergely (Springer Science and Business Media LLC, 2021)
    We propose that humans are prepared to interpret giving as a diagnostic cue of reciprocal–exchange relations from infancy. A prediction following from this hypothesis is that infants will represent the identity of an object they see being given, because this information is critical for evaluating potential future reciprocation. Across three looking-time experiments we tested whether the observation of a transfer action induces 12-month-olds to encode the identity of a single object handled by an agent. We found that infants encoded the object identity when the agent gave the object (Experiment 1), but not when she took it (Experiment 2), despite being able to represent the goal of both actions (Experiments 1 and 3). Consistent with our hypothesis, these results suggest that the infants’ representation of giving comprises information necessary for comparing the value of transferred goods across sharing episodes.
  • Infants’ representation of asymmetric social influence

    Bas, Jesús; Sebastian-Galles, Nuria; Csibra, Gergely; Mascaro, Olivier; Department of Cognitive Science (Elsevier BV, 2023)
    In social groups, some individuals have more influence than others, for example, because they are learned from or because they coordinate collective actions. Identifying these influential individuals is crucial to learn about one’s social environment. Here, we tested whether infants represent asymmetric social influence among individuals from observing the imitation of movements in the absence of any observable coercion or order. We defined social influence in terms of Granger causality; that is, if A influences B, then past behaviors of A contain information that predicts the behaviors and mental states of B above and beyond the information contained in the past behaviors and mental states of B alone. Infants (12-, 15-, and 18-month-olds) were familiarized with agents (imitators) influenced by the actions of another one (target). During the test, the infants observed either an imitator who was no longer influenced by the target (incongruent test) or the target who was not influenced by an imitator (neutral test). The participants looked significantly longer at the incongruent test than at the neutral test. This result shows that infants represent and generalize individuals’ potential to influence others’ actions and that they are sensitive to the asymmetric nature of social influence; upon learning that A influences B, they expect that the influence of A over B will remain stronger than the influence of B over A in a novel context. Because of the pervasiveness of social influence in many social interactions and relationships, its representation during infancy is fundamental to understand and predict others’ behaviors.
  • Infants expect agents to minimize the collective cost of collaborative actions

    Mascaro, Olivier; Csibra, Gergely; Department of Cognitive Science (Springer Science and Business Media LLC, 2022)
    This paper argues that human infants address the challenges of optimizing, recognizing, and interpreting collaborative behaviors by assessing their collective efficiency. This hypothesis was tested by using a looking-time study. Fourteen-month-olds (N = 32) were familiarized with agents performing a collaborative action in computer animations. During the test phase, the looking times were measured while the agents acted with various efficiency parameters. In the critical condition, the agents’ actions were individually efficient, but their combination was either collectively efficient or inefficient. Infants looked longer at test events that violated expectations of collective efficiency (p = .006, d = 0.79). Thus, preverbal infants apply expectations of collective efficiency to actions involving multiple agents.
  • Structural asymmetries in the representation of giving and taking events

    Yin, Jun; Csibra, Gergely; Tatone, Denis; Department of Cognitive Science (Elsevier BV, 2022)
    Across languages, GIVE and TAKE verbs have different syntactic requirements: GIVE mandates a patient argument to be made explicit in the clause structure, whereas TAKE does not. Experimental evidence suggests that this asymmetry is rooted in prelinguistic assumptions about the minimal number of event participants that each action entails. The present study provides corroborating evidence for this proposal by investigating whether the observation of giving and taking actions modulates the inclusion of patients in the represented event. Participants were shown events featuring an agent (A) transferring an object to, or collecting it from, an animate target (B) or an inanimate target (a rock), and their sensitivity to changes in pair composition (AB vs. AC) and action role (AB vs. BA) was measured. Change sensitivity was affected by the type of target approached when the agent transferred the object (Experiment 1), but not when she collected it (Experiment 2), or when an outside force carried out the transfer (Experiment 3). Although these object-displacing actions could be equally interpreted as interactive (i.e., directed towards B), this construal was adopted only when B could be perceived as putative patient of a giving action. This evidence buttresses the proposal that structural asymmetries in giving and taking, as reflected in their syntactic requirements, may originate from prelinguistic assumptions about the minimal event participants required for each action to be teleologically well-formed.
  • Nonverbal Action Interpretation Guides Novel Word Disambiguation in 12-Month-Olds

    Pomiechowska, Barbara; Csibra, Gergely; Department of Cognitive Science (MIT Press - Journals, 2022)
    Whether young infants can exploit sociopragmatic information to interpret new words is a matter of debate. Based on findings and theories from the action interpretation literature, we hypothesized that 12-month-olds should distinguish communicative object-directed actions expressing reference from instrumental object-directed actions indicative of one’s goals, and selectively use the former to identify referents of novel linguistic expressions. This hypothesis was tested across four eye-tracking experiments. Infants watched pairs of unfamiliar objects, one of which was first targeted by either a communicative action (e.g., pointing) or an instrumental action (e.g., grasping) and then labeled with a novel word. As predicted, infants fast-mapped the novel words onto the targeted objects after pointing (Experiments 1 and 4) but not after grasping (Experiment 2) unless the grasping action was preceded by an ostensive signal (Experiment 3). Moreover, whenever infants mapped a novel word onto the object indicated by a communicative action, they tended to map a different novel word onto the distractor object, displaying a mutual exclusivity effect. This reliance on nonverbal action interpretation in the disambiguation of novel words indicates that sociopragmatic inferences about reference likely supplement associative and statistical learning mechanisms from the outset of word learning.
  • Young domestic chicks spontaneously represent the absence of objects

    Szabó, Eszter; Chiandetti, Cinzia; Téglás, Ernő; Versace, Elisabetta; Csibra, Gergely; Kovács, Ágnes Melinda; Vallortigara, Giorgio; Department of Cognitive Science (eLife Sciences Publications, Ltd, 2022)
    Absence is a notion that is usually captured by language-related concepts like zero or negation. Whether nonlinguistic creatures encode similar thoughts is an open question, as everyday behavior marked by absence (of food, of social partners) can be explained solely by expecting presence somewhere else. We investigated 8-day-old chicks’ looking behavior in response to events violating expectations about the presence or absence of an object. We found different behavioral responses to violations of presence and absence, suggesting distinct underlying mechanisms. Importantly, chicks displayed an avian signature of novelty detection to violations of absence, namely a sex-dependent left-eye-bias. Follow-up experiments excluded accounts that would explain this bias by perceptual mismatch or by representing the object at different locations. These results suggest that the ability to spontaneously form representations about the absence of objects likely belongs to the initial cognitive repertoire of vertebrate species.
  • Representations of uncertainty: Where art thou?

    Koblinger, Ádám; Fiser, József; Lengyel, Máté (Elsevier BV, 2021)
    Perception is often described as probabilistic inference requiring an internal representation of uncertainty. However, it is unknown whether uncertainty is represented in a task-dependent manner, solely at the level of decisions, or in a fully Bayesian manner, across the entire perceptual pathway. To address this question, we first codify and evaluate the possible strategies the brain might use to represent uncertainty, and highlight the normative advantages of fully Bayesian representations. In such representations, uncertainty information is explicitly represented at all stages of processing, including early sensory areas, allowing for flexible and efficient computations in a wide variety of situations. Next, we critically review neural and behavioral evidence about the representation of uncertainty in the brain agreeing with fully Bayesian representations. We argue that sufficient behavioral evidence for fully Bayesian representations is lacking and suggest experimental approaches for demonstrating the existence of multivariate posterior distributions along the perceptual pathway.
  • A universal probabilistic spike count model reveals ongoing modulation of neural variability

    Liu, David; Lengyel, Máté (Neural Information Processing Systems Foundation, Inc. (NeurIPS), 2022)
    Neural responses are variable: even under identical experimental conditions, single neuron and population responses typically differ from trial to trial and across time. Recent work has demonstrated that this variability has predictable structure, can be modulated by sensory input and behaviour, and bears critical signatures of the underlying network dynamics and computations. However, current methods for characterising neural variability are primarily geared towards sensory coding in the laboratory: they require trials with repeatable experimental stimuli and behavioural covariates. In addition, they make strong assumptions about the parametric form of variability, rely on assumption-free but data-inefficient histogram-based approaches, or are altogether ill-suited for capturing variability modulation by covariates. Here we present a universal probabilistic spike count model that eliminates these shortcomings. Our method builds on sparse Gaussian processes and can model arbitrary spike count distributions (SCDs) with flexible dependence on observed as well as latent covariates, using scalable variational inference to jointly infer the covariate-to-SCD mappings and latent trajectories in a data efficient way. Without requiring repeatable trials, it can flexibly capture covariate-dependent joint SCDs, and provide interpretable latent causes underlying the statistical dependencies between neurons. We apply the model to recordings from a canonical non-sensory neural population: head direction cells in the mouse. We find that variability in these cells defies a simple parametric relationship with mean spike count as assumed in standard models, its modulation by external covariates can be comparably strong to that of the mean firing rate, and slow low-dimensional latent factors explain away neural correlations. Our approach paves the way to understanding the mechanisms and computations underlying neural variability under naturalistic conditions, beyond the realm of sensory coding with repeatable stimuli.
  • Planning in the brain

    Mattar, Marcelo G.; Lengyel, Máté (Elsevier BV, 2022)
    Recent breakthroughs in artificial intelligence (AI) have enabled machines to plan in tasks previously thought to be uniquely human. Meanwhile, the planning algorithms implemented by the brain itself remain largely unknown. Here, we review neural and behavioral data in sequential decision-making tasks that elucidate the ways in which the brain does—and does not—plan. To systematically review available biological data, we create a taxonomy of planning algorithms by summarizing the relevant design choices for such algorithms in AI. Across species, recording techniques, and task paradigms, we find converging evidence that the brain represents future states consistent with a class of planning algorithms within our taxonomy—focused, depth-limited, and serial. However, we argue that current data are insufficient for addressing more detailed algorithmic questions. We propose a new approach leveraging AI advances to drive experiments that can adjudicate between competing candidate algorithms.
  • Adaptive erasure of spurious sequences in sensory cortical circuits

    Bernacchia, Alberto; Fiser, József; Hennequin, Guillaume; Lengyel, Máté (Elsevier BV, 2022)
    Sequential activity reflecting previously experienced temporal sequences is considered a hallmark of learning across cortical areas. However, it is unknown how cortical circuits avoid the converse problem: producing spurious sequences that are not reflecting sequences in their inputs. We develop methods to quantify and study sequentiality in neural responses. We show that recurrent circuit responses generally include spurious sequences, which are specifically prevented in circuits that obey two widely known features of cortical microcircuit organization: Dale’s law and Hebbian connectivity. In particular, spike-timing-dependent plasticity in excitation-inhibition networks leads to an adaptive erasure of spurious sequences. We tested our theory in multielectrode recordings from the visual cortex of awake ferrets. Although responses to natural stimuli were largely non-sequential, responses to artificial stimuli initially included spurious sequences, which diminished over extended exposure. These results reveal an unexpected role for Hebbian experience-dependent plasticity and Dale’s law in sensory cortical circuits.
  • Contextual inference in learning and memory

    Heald, James B.; Lengyel, Máté; Wolpert, Daniel M. (Elsevier BV, 2023)
    Context is widely regarded as a major determinant of learning and memory across numerous domains, including classical and instrumental conditioning, episodic memory, economic decision-making, and motor learning. However, studies across these domains remain disconnected due to the lack of a unifying framework formalizing the concept of context and its role in learning. Here, we develop a unified vernacular allowing direct comparisons between different domains of contextual learning. This leads to a Bayesian model positing that context is unobserved and needs to be inferred. Contextual inference then controls the creation, expression, and updating of memories. This theoretical approach reveals two distinct components that underlie adaptation, proper and apparent learning, respectively referring to the creation and updating of memories versus time-varying adjustments in their expression. We review a number of extensions of the basic Bayesian model that allow it to account for increasingly complex forms of contextual learning.
  • Training stochastic stabilized supralinear networks by dynamics-neutral growth

    Soo, Wayne W.M.; Lengyel, Máté (Neural Information Processing Systems Foundation, Inc. (NeurIPS), 2022)
    There continues to be a trade-off between the biological realism and performance of neural networks. Contemporary deep learning techniques allow neural networks to be trained to perform challenging computations at (near) human-level, but these networks typically violate key biological constraints. More detailed models of biological neural networks can incorporate many of these constraints but typically suffer from subpar performance and trainability. Here, we narrow this gap by developing an effective method for training a canonical model of cortical neural circuits, the stabilized supralinear network (SSN), that in previous work had to be constructed manually or trained with undue constraints. SSNs are particularly challenging to train for the same reasons that make them biologically realistic: they are characterized by strongly-connected excitatory cells and expansive firing rate non-linearities that together make them prone to dynamical instabilities unless stabilized by appropriately tuned recurrent inhibition. Our method avoids such instabilities by initializing a small network and gradually increasing network size via the dynamics-neutral addition of neurons during training. We first show how SSNs can be trained to perform typical machine learning tasks by training an SSN on MNIST classification. We then demonstrate the effectiveness of our method by training an SSN on the challenging task of performing amortized Markov chain Monte Carlo-based inference under a Gaussian scale mixture generative model of natural image patches with a rich and diverse set of basis functions -- something that was not possible with previous methods. These results open the way to training realistic cortical-like neural networks on challenging tasks at scale.
  • The Computational and Neural Bases of Context-Dependent Learning

    Heald, James B.; Wolpert, Daniel M.; Lengyel, Máté (Annual Reviews, 2023)
    Flexible behavior requires the creation, updating, and expression of memories to depend on context. While the neural underpinnings of each of these processes have been intensively studied, recent advances in computational modeling revealed a key challenge in context-dependent learning that had been largely ignored previously: Under naturalistic conditions, context is typically uncertain, necessitating contextual inference. We review a theoretical approach to formalizing context-dependent learning in the face of contextual uncertainty and the core computations it requires. We show how this approach begins to organize a large body of disparate experimental observations, from multiple levels of brain organization (including circuits, systems, and behavior) and multiple brain regions (most prominently the prefrontal cortex, the hippocampus, and motor cortices), into a coherent framework. We argue that contextual inference may also be key to understanding continual learning in the brain. This theory-driven perspective places contextual inference as a core component of learning.
  • Perceiving others’ cognitive effort through movement: Path length, speed, and time

    Székely, Marcell; Michael, John (SAGE Publications, 2023-07-03)
    Effort perception is a crucial capacity underpinning characteristically human forms of sociality, allowing us to learn about others’ mental states and about the value of opportunities afforded by our environment, and supporting our ability to cooperate efficiently and fairly. Despite the crucial importance and prevalence of effort perception, little is known about the mechanisms underpinning it. Across two online experiments (N = 462), we tested whether adults estimate others’ cognitive effort costs by tracking perceptible properties of movement such as path length, time, and speed. The results showed that only time had a consistent effect on effort perception, that is, participants rated longer time as more effortful. Taken together, our results suggest that within the context of our task—observing an agent deciphering a captcha—people rely on the time of others’ actions to estimate their cognitive effort costs.
  • An initial but receding altercentric bias in preverbal infants' memory

    Manea, Velisar; Kampis, Dora; Grosse Wiesmann, Charlotte; Revencu, Barbu; Southgate, Victoria (The Royal Society, 2023-06-07)
    Young learners would seem to face a daunting challenge in selecting to what they should attend, a problem that may have been exacerbated in human infants by changes in carrying practices during human evolution. A novel theory proposes that human infant cognition has an altercentric bias whereby early in life, infants prioritize encoding events that are the targets of others’ attention. We tested for this bias by asking whether, when the infant and an observing agent have a conflicting perspective on an object's location, the co-witnessed location is better remembered. We found that 8- but not 12-month-olds expected the object to be at the location where the agent had seen it. These findings suggest that in the first year of life, infants may prioritize the encoding of events to which others attend, even though it may sometimes result in memory errors. However, the disappearance of this bias by 12 months suggests that altercentricism is a feature of very early cognition. We propose that it facilitates learning at a unique stage in the life history when motoric immaturity limits infants' interaction with the environment; at this stage, observing others could maximally leverage the information selection process.
  • Semantic incorporation and discourse prominence: Experimental evidence from English pronoun resolution

    Wittenberg, Eva; Trotzke, Andreas; Department of Cognitive Science (Elsevier, 2021)
    The semantic incorporation of nouns into predicates, like give a hug, is not morphologically marked in English, and how syntactic incorporation strategies like light verb constructions influence the discourse-prominence structure of an utterance has not yet been studied systematically. One hypothesis is that since semantically incorporated nouns are not morphosyntactically incorporated in English, they can function like any other noun as prominent and accessible referents for anaphora. Another hypothesis is that their semantic status and their predicative meaning influence their discourse prominence, and hence their accessibility by anaphoric means. We tested these two hypotheses in two experimental studies on different anaphoric preferences of English pronouns. Our studies demonstrate that the felicity patterns for the two different pronominal reference strategies are determined at different linguistic levels: For it, we found an impact of morphosyntactic form; for that, the semantic type of the referent (object vs. event) seems to play a role. Crucially, the degree of semantic incorporation does not affect discourse prominence and pronoun choice to the extent that we had expected.
  • A Psycholinguistic Investigation into Diminutive Strategies in the East Franconian NP: Little Schnitzels Stay Big, but Little Crooks Become Nicer

    Wittenberg, Eva; Trotzke, Andreas; Department of Cognitive Science (Cambridge University Press, 2021)
    Upper German dialects make heavy use of diminutive strategies, but little is known about the actual conceptual effects of those devices. This paper is the first to present two large-scale psycholinguistic experiments that investigate this issue in East Franconian, a dialect spoken in Bavaria. Franconian uses both the diminutive suffix -la and the quantifying construction a weng a lit. ‘a little bit a’ to modify noun phrases. Our first experiment shows that diminutization has no effect on conceptualization of magnitude: People do not think of a smaller/weaker/shorter etc. referent when the NP is modified by the morphological diminutive, the quantifying construction, or their combination. The second experiment involves gradable NPs and shows that, again, the morphological diminutive has no effect on how people conceptualize the degree to which a gradable nominal predicate holds; in contrast, a weng a reduces it significantly. These experiments suggest that diminutization does not have uniform effects across semantic domains, and our results act as a successful example of extending the avenue of cognitive psychology into dialectology with the active participation of a speaker community.
  • Chimpanzee Coordination and Potential Communication in a Two-touchscreen Turn-taking Game

    Voinov, Pavel V.; Call, Josep; Knoblich, Günther; Oshkina, Marina; Allritz, Matthias; Other (Springer Nature, 2020)
    Recent years have seen a growing interest in the question of whether and how groups of nonhuman primates coordinate their behaviors for mutual benefit. On the one hand, it has been shown that chimpanzees in the wild and in captivity can solve various coordination problems. On the other hand, evidence of communication in the context of coordination problems is scarce. Here, we investigated how pairs of chimpanzees ( Pan troglodytes ) solved a problem of dynamically coordinating their actions for achieving a joint goal. We presented five pairs of chimpanzees with a turn-taking coordination game, where the task was to send a virtual target from one computer display to another using two touch-screens. During the joint practice of the game some subjects exhibited spontaneous gesturing. To address the question whether these gestures were produced to sustain coordination, we introduced a joint test condition in which we simulated a coordination break-down scenario: subjects appeared either unwilling or unable to return the target to their partner. The frequency of gesturing was significantly higher in these test trials than in the regular trials. Our results suggest that at least in some contexts chimpanzees can exhibit communicative behaviors to sustain coordination in joint action.
  • Infants do not use payoff information to infer individual goals in joint-action events

    Tatone, Denis; Schlingloff-Nemecz, Laura; Pomiechowska, Barbara; Department of Cognitive Science (Elsevier, 2023)
    For observers the occurrence of a joint action (JA) is inherently ambiguous with respect to the goals of the individuals involved. People may work together, for instance, because they are motivated to obtain material rewards or to help others. We hypothesized that to solve this interpretive ambiguity observers leverage information about the JA’s payoff structure. Specifically, when a JA yields material rewards for a participating agent (as well as their partner), their behavior can be straightforwardly explained as instrumental to the obtainment of these rewards. Conversely, when a JA does not yield material rewards for the agent (but does so for her partner), the unrewarded agent’s contribution needs to be accounted for by positing other types of goals, such as assisting her partner in obtaining her rewards. We examined this hypothesis across three looking-time experiments with 12-month-olds: specifically, we tested whether the absence of material rewards for an agent participating in a JA would prompt infants to interpret her participation as prosocially motivated. Consistent with this hypothesis, Experiment 1 showed that, after having been familiarized to two dyadic JA events resulting in one or both agents being rewarded, infants selectively expected the unrewarded agent to act altruistically towards her former JA partner by giving her a resource. Experiments 2–3 examined whether this expectation was driven by the prosocial interpretation of the unrewarded agent’s behavior or by changes in the number or distribution of resources between familiarization and test. Contradicting the hypothesis that infants interpreted the agent’s behavior as prosocially motivated, in Experiment 2 we observed similar looking times when the unrewarded agent performed a prosocial action (giving) or an antisocial action (taking) towards her partner at test. Further, in a close replication of the original experiment in which the change in the number of familiarized objects occurred in the test event featuring the rewarded agent rather than the unrewarded one (Experiment 3), infants produced a looking-time pattern opposite to the one first obtained. Taken together, these findings suggest that infants encoded the payoff structure of JA events (i.e., the number/distribution of resources that the interaction brought about) but did not leverage this information to infer the individual goals of participating agents. The present evidence calls for a critical re-evaluation of our original hypothesis and for further research into the mechanisms by which infants disambiguate the motives of agents involved in joint actions.

View more