Interactivity and Individual Differences in Mental Arithmetic

Symposium: Advances in Thinking, Problem Solving and Creativity: A Festschrift for Kenneth J. Gilhooly Organiser: Linden Ball Rationale: Since the 1970s Ken Gilhooly has made major and lasting contributions to the fields of thinking, problem solving and creativity through his research on topics such as the role of working memory and individual differences in reasoning and planning, the processes involved in insight and non-insight problem solving, the strategies underpinning creative idea production and the characteristics of real-world domain expertise. This Festschrift will present a survey of emerging issues in the psychology of thinking, problem solving and creativity in tribute to Ken’s own substantial research contributions in these areas. The symposium speakers represent a subset of the many colleagues and collaborators who Ken has influenced over the years and who are all currently working at the cutting-edge of thinking, problem solving and creativity research. Within the proposed symposium these contributors will report their latest findings and conceptual advances in relation to topics such as planning and skilled problem solving, mental imagery and creative thinking, insight problem solving, thought suppression and incubation effects, and individual differences in problem solving and learning. The symposium will offer an integrative perspective on foundational issues as well as on new developments in thinking, problem solving and creativity research. The event will be rounded off by Ken himself, who will draw on his own recent work to present a reflective review of the current state-of-the-art in research and theorising in the area of insight problem solving. Insight into Problem Solving? Symposium Keynote speaker: Kenneth Gilhooly (University of Hertfordshire, Hatfield), [email protected] Have we advanced much in understanding problem solving since Poincaré, Wallas and Duncker? This “Trinity” still loom large both in popular science accounts of the field and in introductory textbooks. While we now have a good understanding of problem solving in small-scale, well-defined problems, as search within a problem space, and can relate solving processes to cognitive resources, such as working memory, the areas of insight and creativity still strike many as mysterious. Insight tasks require a change in how they are represented (re-structuring) in contrast to tasks that can be solved by search within a problem space. Theories involving representational change and progress monitoring offer explanations which remove some of the mystery of insight. Creative problem solving in which novel solutions are required does seem to involve unconscious processes (as do many areas of psychology). Studies of “incubation” effects are beginning to remove some of the mystery but clear models are still lacking. Recent studies based on priming and embodied cognition have produced eye-catching results which although suggestive are lacking clear process explanations. Finally, the possible value of brain imaging in problem solving research will be discussed – with some caution urged in view of: (1) the differential effects of being supine (as in a scanner) as against upright on problem solving; and (2) the argument that cognitive theory illuminates imaging data rather than vice versa. Incubation and Suppression Processes in Problem Solving George Georgiou (University of Hertfordshire, Hatfield), [email protected] Experimental research has shown that setting a problem aside for a period of time can facilitate solutions when the problem is resumed. This break away from a problem is referred to as an ‘incubation’ period. A number of possible mechanisms have been proposed to account for this phenomenon, including: intermittent conscious work; selective forgetting; attention withdrawal; relief from fatigue; and unconscious work. However, a crucial aspect of the incubation procedure is for the problem solver not to think about the problem during the break, by possibly suppressing thoughts of the problem whilst engaged in interpolated activities. Attempts to suppress thoughts have been shown to result in a subsequent increase in the accessibility and frequency of the suppressed thought, known as a rebound effect. The research reported here explores the extent to which the facilitating effects of an incubation period may be due to suppression effects, which paradoxically activate the avoided thoughts, thus aiding problem solving. The results of studies using delayed and immediate paradigms will be discussed. Interactivity and Individual Differences in Mental Arithmetic Frédéric Vallée-Tourangeau (Kingston University, London), [email protected] This paper reports data from a mental arithmetic task where participants were invited to complete short and long additions in two different contexts. In a first, they solved the additions without using their hands or artefacts and in a second the same problems were presented as a set of manipulable tokens. In a static context, working memory resources may be stretched, particularly by the long additions. In contrast, the cognitive system created by the coupling of internal and external resources in an interactive context may enhance arithmetic performance. The addition problems were interpolated among a series of other tasks that measured numeracy and arithmetic skills, working memory capacity, visuo-spatial processing speed and attention switching. Mental arithmetic performance was measured in terms of accuracy, latency and efficiency – indexed as the ratio of accuracy over time invested in completing the problems. A significant interaction between condition and problem type was observed: Performance was slightly better in the static condition for the short sums but declined substantially relative to performance in the interactive condition for the long sums. Individual differences in terms of working memory capacity and arithmetic skills explained a fifth of the variance in performance in the static condition, while arithmetic skills, working memory capacity and attention switching skills explained 45% of the performance variance in the interactive condition. Implications for process models of mental arithmetic in static and interactive contexts of reasoning are discussed. Process Dissociations in Problem Solving with Verbal and Visual Insight Tasks Linden Ball (Lancaster University), [email protected] Two accounts of insight problem solving dominate the literature: the ‘special-process theory’, which attributes insight to implicit processes of spreading activation, and the ‘progress monitoring theory’, which proposes that both insight and non-insight problems draw on the same analytic planning mechanisms. We report two experiments that manipulated concurrent tasks demands during insight problem solving in terms requirements to engage in either thinking aloud (TA) or articulatory suppression (AS). Experiment 1 used verbal insight problems whereas Experiment 2 used visual insight problems. Experiment 1 revealed that insight into verbal problems was hindered by AS but facilitated by TA, whereas Experiment 2 showed the exact opposite pattern, with insight into visual problems being impaired by TA and enhanced by AS. Explanations of these process dissociations in insight problem solving will be examined, including the possibility that a rapprochement between the special process and progress monitoring accounts may be necessary to capture the full pattern of data across both kinds of insight tasks. Planning: The Force Behind Skilled Problem Solving, Insight, and Creativity Tom Ormerod (Lancaster University), [email protected] The contribution made by Ken Gilhooly in bringing to light important contributions made by working memory and individual differences to a full understanding of human reasoning, problem-solving, skill and expertise, cannot be underestimated. Yet, working memory and individual differences are often ignored. This is problematic, because working memory is fundamental to planning, and planning, it will be argued here, is fundamental to most successful problem-solving, and its absence can often explain much failure to solve. Here, the role of planning will be assessed across insight and creative problem-solving domains, from puzzles to real-world design tasks. It will be argued that planning, while essentially a process that guides conscious search, is both constrained and encouraged by individual differences among solvers and differences in working memory requirements generated by the problem to be solved. Mental Imagery and Creative Thinking David Pearson (University of Aberdeen), [email protected] Mental imagery has a long history of association with imagination and creative thought. Experimental studies will be discussed which have examined in particular the process of ‘mental synthesis’, in which visual imagery is used to manipulate and combine separate components into new configurations. The experience of mental synthesis has been linked to successful performance across a wide range of different creative tasks, including the visualisation and development of scientific models, the conceptual stage of architectural design, and many aspects of general everyday problem-solving. Presented data from experimental studies will explore the contribution made by working memory, expertise, and mood to successful synthesis performance. Adult Age Effects on Belief Reasoning Louise Phillips (University of Aberdeen), [email protected] Older adults often perform poorly on Theory of Mind (ToM) tests that require understanding of others’ beliefs and intentions. There are opposing theories about the course and specificity of age changes in belief reasoning across the adult lifespan. In a recent study, we found that there are non-linear effects of aging on belief reasoning, with specific impairments in the ability to process false beliefs in older adults aged 65-88. Difficulties in updating information in working memory and the ability to decode social cues partially mediated the age differences in false belief reasoning. These results indicate that age differences in decoding social cues and updating information in memory may be important influences on the specific problems encountered when reasoning about false beliefs in old age. Symposium: Analogical Reasoning Organiser: Robert G. Morrison Executive Function and Knowledge in Analogical Reasoning: An Integrated Developmental, Neurocognitive and Computational Approach Symposium keynote speaker: Robert G. Morrison (Loyola University Chicago), [email protected] Executive functions, such as inhibitory control, are critical to our ability to process analogies, and may as well be critical for learning the relational knowledge necessary for analogical reasoning. I will highlight developmental, neuropsychological, neuroimaging, and computational findings showing the importance of inhibitory control during relational learning and reasoning, and suggest the importance of moving cognitive neuroscience approaches to analogy beyond just reasoning to include how we learn relational knowledge necessary for reasoning. From your eyes only: Tracking children's and adults' strategies in analogy making Jean-Pierre Thibaut & Robert French (Université de Bourgogne), [email protected] We will present eye-tracking data in which we compare children and adults in analogy making tasks. These data suggest that children do not allocate the same amount of time to the different stimuli that compose the task as the adults or that correct trials are not organized in the same way as error trials. We will present data that capitalize on these results showing that the way the stimuli are introduced in the task influence performance. Development of relational reasoning during adolescence Iroise Dumontheil (University College London), [email protected] Relational integration refers to the ability to jointly consider several structured mental representations, or relations. As such, it is involved in analogical reasoning. I will present data on the development of relational reasoning performance from late childhood to adulthood, and on the neural correlates of relational reasoning during adolescence and adulthood. Growing analogy from recycled parts Denis Mareschal (Birkbeck College, University of London), [email protected] In this talk I will discuss the need for models of complex cognition to be developmentally tractable. This will be illustrated using a connectionism model of early analogical reasoning. The strengths and limitations of this approach will be discussed. Toward a neurocomputational model of relational reasoning Keith Holyoak (University of California, Los Angeles), [email protected] I will argue that recent developments in cognitive neuroscience suggest possible mechanisms by which a computational model of relational reasoning might be realized in the brain. Relational reasoning may involve (1) use of neural synchrony and other temporal information to bind information coded in different brain areas, (2) neurons in prefrontal cortex capable of rapid learning, (3) cross-frequency coupling as a modulator of learning, and (4) different types of inhibitory connections to control sequential processing of relational information in working memory. Analogical processes support learning relational categories Dedre Gentner (Northwestern University), [email protected] Several lines of evidence converge to suggest an important role for common relational structure in category learning and use. The role of common relational structure in categories is most clearly seen in relational categories such as bisector and carnivore. Because such categories lack perceptual support, they pose a challenge to learners. I’ll describe studies that indicate that analogical comparison is a major contributor to learning relational categories, in both children and adults. Further, this work suggests that the use of consistent relational language amplifies the effects of analogical comparison. Learning structured representations from unstructured inputs Leonidas A Doumas (University of Hawaii), [email protected] Relational thinking requires explicitly relational representations. To be explicitly relational, a representation must specify relations (or relational roles) independently of their arguments and bind them to their arguments dynamically. Despite the centrality of relational thinking in human cognition, a clear description of how children and adults acquire the kinds of representations that can support relational thinking has been elusive. I will present a theory, instantiated in a computer model called DORA, of how structured explicitly relational representations are learned from unstructured examples. The resulting model provides a powerful and comprehensive account of many aspects of cognitive development and adult cognition. Why we’re (not) so smart with numbers John E. Opfer, Vyacheslav Y. Nikitin, & Frank J. Kanayet (Ohio State University), [email protected] The human capacity for quantitative thinking expands profoundly during the first decade of life. What accounts for these changes? And why do adults remain innumerate across so many contexts? In this talk, I will present evidence that symbolic coding of quantities plays a key role in determining when we are (and are not) proficient at processing quantitative relations. Analogy and education: Encouraging analogy through helping learns to manage process demands Lindsey Richland (University of Chicago), [email protected] Analogy is a powerful tool for generalization and transfer, though experimental participants often fail to notice or use relevant analogs. This talk explores the relationship between the cognitive demands of learning and problem solving by analogy and instructional practices that can mitigate the likelihood that those demands overwhelm learners. Specifically, working memory and inhibitory control demands are high when learners solve problems or reason analogically about novel stimuli, but these can be reduced by strategic yet practicable use of visual information (the chalkboard, gesture, visual representations). Making culture: Roles for analogy Jeffrey Loewenstein (University of Illinois, Urbana-Champaign), [email protected] Some knowledge is remembered, gets talked about, becomes selected, generates new products, guides companies, shapes institutions, and otherwise makes culture. Analogy plays important roles in many of these processes. Data on cultural narratives show advantages of analogy for a wide range of outcomes, including social selection, engagement and attitudes. Symposium: Argumentation Organiser: Ulrike Hahn Rationale: Argument is central to our everyday thought and much of the reasoning we perform takes place in the service of argument, broadly construed, that is, in the service of attempts to convince ourselves or others of a particular conclusion. The symposium brings together recent research on argumentation, focusing primarily on argument quality and the extent to which people are sensitive to argument quality in both argument evaluation and generation. Looking for arguments: The argumentative function of reasoning Symposium keynote speaker: Hugo Mercier (University of Neuchâtel), [email protected] Since it allows finding and evaluating arguments, reasoning is essential to argumentation. Dan Sperber and I have suggested that in fact argumentation is the main function of reasoning: reasoning would have evolved so we can produce arguments to convince others and evaluate arguments so as to be convinced when appropriate. Here I will support this hypothesis by showing that reasoning is designed and performs in the way expected of an argumentative device. When reasoning produces arguments it displays a confirmation bias and it starts out with lax criteria of argument quality, both traits that make sense in a dialogic context. While these traits of reasoning are detrimental for the lone reasoner, they allow for good performance in argumentative settings, as demonstrated by results in the psychology of reasoning, social psychology, developmental psychology, education and political science. Is high-quality evidence more persuasive than low-quality evidence in the presence of counterevidence? A cross-cultural study Jos Hornikx (Radboud University Nijmegen), [email protected] Different types of data (evidence) can be used to support claims, such as anecdotal and statistical evidence. Hornikx and Hoeken (2007) demonstrated that high-quality evidence in support of claims is more persuasive than low-quality evidence for Dutch participants, but not for French participants. The French participants may not have been centrally processing the claims and evidence, which is necessary to distinguish between strong and weak evidence (cf. Petty & Cacioppo, 1986). Heckler and Childers (1992) demonstrated that people process information more carefully when the information contains incongruent elements. Therefore, an experiment was conducted in which participants received incongruent evidence: evidence supporting the claim, and counterevidence. The central question was whether normatively strong evidence was more persuasive than normatively weak evidence for French participants under conditions of incongruent evidence. Dutch and French students (N = 544) indicated how probable they found claims followed by statistical evidence and normatively strong/weak expert evidence; each piece of evidence served in some cases as evidence, and in other cases as counterevidence. Analyses showed that normatively strong expert evidence was more persuasive than normatively weak expert evidence for Dutch participants, but that the two types of evidence were equally persuasive for French participants. This means that, even in a condition that promotes central processing, French participants were found to be insensitive to evidence quality. Arguing about probability: Lay people’s criteria to assess argument quality Hans Hoeken, Ester Sorm, & Peter Jan Schellens (Radboud University Nijmegen), [email protected] Argument quality can have strong and lasting persuasive effects. This raises the question of what criteria people use to distinguish high quality from low quality arguments. In an experiment, 196 participants without any training in argument theory rated their acceptance of 30 probability claims supported by either an argument from authority, from cause to effect, or from example. Arguments were systematically manipulated to violate nine specific criteria. For seven criteria, violation decreased acceptance of the claim supported. These findings provide insights into people's argumentative competence and enable a more precise description of the persuasion process. Finally, the results have implications for the use of argument quality as a methodological tool in persuasion research. Drawing inferences from absent evidence: A Bayesian network approach Adam J. L. Harris, David A. Lagnado, & Victoria Cullen (University College London), [email protected] The Bayesian formalisation of the argument from ignorance (Hahn & Oaksford, 2007) demonstrated that inferences from missing evidence can be valid for a range of probabilistic parameters. We go beyond this pioneering research and investigate the causal networks underlying these parameters. The research is situated within a legal setting, and concerns the absence of eyewitness testimony. We show that the normatively appropriate inference from absent evidence differs according to its likely causes, and report the results of a behavioural study demonstrating that the juror-eligible participants in our study are sensitive to this. Thus, explanations for the absence of key trial evidence are critical pieces of information for jurors. The effect of this information, both normatively and descriptively, is well understood within a Bayesian network approach to evidential reasoning. Utility templates for consequential argumentation Jean-François Bonnefon, Matthew Haigh & Andrew Stewart (Universite de Toulouse II), [email protected] People use conditional sentences when describing actions, their preconditions, and their consequences. Oftentimes, these actions and consequences matter, in the sense that they have value or utility to various agents: they then constitute consequential arguments. For example: If she praises him, he will support her (therefore, she will praise him); If you leave me, I will be crushed (therefore, don't leave me). Many arrangements of agents, targets and values are possible beyond these two examples, but our contention is that some of these arrangements have the special status of /utility templates/ that guide and constrain interpretation. We argue that whenever it is possible, people will interpret or re-interpret a conditional sentence in order to make it coincide with one of their utility templates. We identify four potential templates through a sentence completion survey, and demonstrate their properties in two experimental studies. Symposium: Bayesian Approaches Organiser: Nick Chater, Peter Dayan and Adam Sanborn A taxonomy of inductive problems Symposium keynote speaker: Charles Kemp (Cannergie Mellon University, Pittsburgh), [email protected] Inductive inferences about objects, features, and relations have been studied for many years but there are few attempts to chart the full range of inductive problems that humans are able to solve. I will present a taxonomy of inductive problems that helps to clarify the relationships between familiar inductive problems such as generalization, categorization, and identification, and that introduces new inductive problems for psychological investigation. I will discuss several computational accounts of inductive reasoning and will argue that the Bayesian approach provides a promising way to explain how people solve all of the problems in the taxonomy. To support this claim, I will present Bayesian models that help to explain how people simultaneously generalize across objects and features, how people identify categories and features, and how people imagine novel categories and category exemplars. Learning time-varying categories: The role of absolute and relative judgment Dan Navarro (University of Adelaide), Jeff Beck, Ingmar Kanitscheider and Alex Pouget [email protected] Most theories of category learning assume that the structure of categories is static, but many categories change over time. To the extent that this problem has been studied previously, it has been assumed that memory based strategies can account for human sensitivity to change over time (i.e, weighting recent observations more heavily). We present the results of a category learning experiment involving categories that change over time in a systematic fashion, and find evidence that a simple recency weighting strategy is not sufficient to explain human performance, and that people can learn to anticipate changes before they occur. Additionally, we introduce a probabilistic modelling framework able to explore the representational origins of this effect, in terms of the extent to which people rely on both absolute stimulus magnitudes and relative stimulus magnitudes when forming a category representation. A normative theory of Weber’s law Jeff Beck, Ingmar Kanitscheider and Alex Pouget (Universities of Geneva and Rochester), [email protected] A diverse array of studies have shown that discrimination thresholds for sensory variables are often proportional the stimulus magnitude, a phenomenon known as Weber’s law. Typical explanations invoke a finely tuned combination of a nonlinear neural representation and noise in neural responses. For instance, one such explanation assumes that neural responses are sensitive to the logarithm of the sensory variable (or the ratio of sensory variables) corrupted by a noise with fixed variance. Here we propose a purely computational explanation for Weber’s Law which does not require an appeal to internal representation or noise. Rather, we suggest that it arises purely from the statistical nature of the problem faced by the brain. For example, imagine having to estimate the number of items in a scene. If we treat the items as blobs of activity in feature maps, the total sum of the activity provides a numerosity estimate. If the variability within the feature map is independent, the variance of the estimate scales with the mean numerosity, in contrast to Webers’ law which predicts that the variance scale with the square of the mean. However, the independence assumption is problematic because activity in such maps is likely to be scaled by global parameters which vary from trial to trial such as the overall luminosity in the image or the level of attention. It is easy to show that the presence of such global scaling parameters correlates neural activity in such a way that the variance of the total sum scales with the square of the mean, thus yielding Weber’s law. This simple intuition can be generalized to more complex models, by considering inference in scale mixture models such as the Gaussian Scale or Gamma-Poisson mixture models which precisely replicate Weber’s Law when the variance of the scale parameter is large enough. Reconciling intuitive physics and Newtonian mechanics for colliding objects Adam Sanborn, Vikash Mansinghka and Thomas Griffiths (University of Warwick, Coventry), [email protected] People have strong intuitions about the influence objects exert upon one another when they collide. Because people's judgments appear to deviate from Newtonian mechanics, psychologists have suggested that people depend on a variety of task-specific heuristics. This leaves open the question of how these heuristics could be chosen, and how to integrate them into a unified model that can explain human judgments across a wide range of physical reasoning tasks. We propose an alternate framework, where people's judgments are based on optimal statistical inference over a Newtonian physical model that incorporates sensory noise and intrinsic uncertainty about the physical properties of the objects being viewed. This ``noisy Newton" framework can be applied to a multitude of judgments, with people's answers determined by the uncertainty they have for physical variables and the constraints of Newtonian mechanics. We investigate a range of effects in mass judgments that have previously been taken as strong evidence for heuristic use and show that they are well explained by the interplay between Newtonian constraints and people's uncertainty. We also consider an extended model that handles causality judgments, and obtain good quantitative agreement with human judgments across tasks that involve different judgment types, with a single consistent set of parameters. The Bayesian Boom: Good thing or bad? Ulrike Hahn (Cardiff University), [email protected] Two separate proposals have recently critiqued the role of normative frameworks, and, in particular, Bayesian probability in the study of cognition (Elqayam & Evans, 2012; Jones & Love, 2011), encouraging a greater focus on process theories and implementation. In light of these, the talk seeks to characterise more precisely the role of Bayesian probability in cognitive theories, trying to distinguish different ways in which the framework is brought to bear, and highlighting both its unique contributions and limitations in its present use. Heuristics, optimality, and children’s sequential information search Jonathan Nelson, Bojana Divjak, Gudny Gudmundsdottir, Laura Martignon, and Björn Meder (Max Planck Institute for the History of Science, Berlin), [email protected] Consider a game of guessing which person has been chosen at random from among several people. The task is to identify the person with the smallest number of yes-or-no questions, about specific features that some people have (e.g. "Is the person wearing earrings?"). It is impractical or impossible to check which of all possible sequences of questions is most efficient. Are any heuristic or stepwise-optimal strategies effective? Does it depend on what environment the people are from? We addressed this in a Representative Environment with similar numbers of male and female people, and in a predominantly male Nonrepresentative Environment. Exhaustive search revealed that in the Nonrepresentative Environment, beard is the best first question. In the Representative Environment, gender is the best first question. Remarkably, a simple heuristic strategy-asking about the feature possessed by closest to half of the possible individuals-identifies the optimal decision tree in both environments. We conducted an experiment to explore 4th-grade children's strategies in this game, using cards with cartoon faces to represent the possible people. The children adapted their searches to each environment and preferentially asked the best first question in each environment. In the Nonrepresentative Environment, the best first question (beard) initially tied with gender for most popular. In the Representative Environment, a strong majority of children asked the most useful question (gender) first. This could suggest that people's searches are especially efficient in real-world environments. Symposium: Causal Learning and Reasoning Organiser: David Lagnado Causality and cognition Symposium keynote speaker: Noah Goodman (Stanford University), [email protected] I will explore the ways that formal theories of causal understanding can be integrated with theories of conceptual representation and inference. This will include grounding of causal concepts into perception, learning abstract knowledge about the "causes" relation, and rich causal knowledge representations formalized via probabilistic programming. The causal structure of choice leads to self-deception: i. Data Steven Sloman, York Hagmayer, & Philip Fernbach (Brown University), [email protected] What can people legitimately learn about themselves by observing their own behavior? The answer hinges on what people believe about the causes of action: Is a given action a product of free will and agency, a talent, a skill, or something else? We argue that uncertainty in the causes of actions enables self-deception. We characterize different types of self-deception in terms of the distinction between intervention and observation in causal reasoning. Diagnostic self-deception arises when people deceive themselves about the diagnostic value of their own actions, specifically when people intervene but choose to view their actions as observations in order to find support for a self-serving diagnosis. In earlier work we have shown that such self-deception depends on imprecision in the environment that gives freedom to represent one’s own actions as either observations or interventions depending on which offers a more favorable inference. More recently, we have obtained more direct evidence for a motivational shift in the causal construal of choice. We report studies showing that the conditions promoting diagnostic self-deception lead to a negative correlation between the amount of effort participants report putting into a task and the actual effort they expend. The causal structure of choice leads to self-deception: ii. Model York Hagmayer, Steven Sloman, & Philip Fernbach (Kings College London), [email protected] A causal theory of choice will be presented which entails diagnostic self-deception under certain conditions. The basic idea is that we use causal models of choice to decide on actions and update our beliefs about the causes of our behavior based in part on which beliefs are deemed favorable. The theory states that people are uncertain about which of three models of choice govern a specific action: (i) an Intervention Model according to which actions are only dependent on intentions, (ii) an Observation Model according to which behaviors are due to an internal factor (e.g., ability, personality, sensitivity), and (iii) a Conjunctive Model according to which both intentions and other factors are necessary for action. People also assign values to (i) actions because of their consequences and (ii) internal properties that cause behavior. When choosing, people pick the action that has the highest expected value. When updating their beliefs about the choice models, they consider the available evidence and the values entailed by the updated beliefs and pick the most favorable set. This theory predicts self-deceptive behavior when belief in the observation model is high, a potential internal cause has a high value, and the action has some negative consequences. Results from various simulation studies will be presented. Time and causality: Causal binding leads to shifts in event perception. Marc J. Buehner (Cardiff University), [email protected] Temporal binding of action to consequence refers to a subjective shortening of elapsed time between the former and the latter. Originally it was thought that temporal binding is specific to motor learning and arises as a consequence of either sensory adaptation (Stetson, Cui, Montague, & Eagleman, 2006) or the associative nature of the forward model of motor command (Haggard, Clark, & Kalogeras, 2002). Both of these interpretations rely on the intentional quality of the causative action (c.f. Wohlschläger, Haggard, Gesierich, & Prinz, 2003). I will present evidence of temporal binding resulting from not only intentional actions but also from mechanical causation, suggesting that intentional action is not necessary for temporal binding, and that binding results from the causal relation linking both events. Consequently, ‘intentional binding’ is a special case of more general ‘causal binding’, which can be explained by a theory of Bayesian ambiguity reduction (Buehner & Humphreys, 2009, 2010; Eagleman & Holcombe, 2002). What are the limits of propositionally mediated associative learning? Jan De Houwer (Ghent University), [email protected] Associative learning refers to a change in behavior that results from the relation between events (e.g., the presence of two stimuli or a behavior and a stimulus). It is now generally accepted that associative learning is often mediated by the formation of propositions about the related events. However, this conclusion is based almost exclusively on studies in which participants had full control over the behavior that changed (e.g., causality ratings). I review the results of a number of studies on causal learning and evaluative conditioning in which associatively induced changes in more automatic reactions were examined. Whereas some studies suggest that even these instances of associative learning can be mediated by the formation of propositions, others reveal learning that does not seem to be due to propositional processes. We review the evidence and evaluate whether it is necessary to abandon the idea that all instances of associative learning are propositionally mediated. Symposium: Cognitive models of decision making Organiser: Nick Chater Rationale: Thinking is only of practical consequence if it leads to decisions. Over the last fifty years, a vast empirical literature has documented a vast range of decision making phenomena. But what the principles that govern how people make decisions? This symposium focusses on models aiming to explain the mechanistic and functional basis of human decision making; and to provide an integrated account of how people turn convert thoughts into decisions. How we think when we decide: What evidence, exactly, do we accumulate when we choose? Neil Stewart (University of Warwick, Coventry), [email protected] When a choice is repeated, we do not always make the same decision. Choices that are more finely balanced take longer to make. Both these pieces of evidence strongly implicate a stochastic accumulation process, where evidence is repeatedly sampled until a decision is made. Findings from single cell recording and ERP support this account. But what, exactly, is the evidence that we accumulate? There are several competing mathematical models which make strong claims about what is accumulated. I will explore these models and use eye-tracking process data to constrain an account of information accumulation in choice---the essence of how we think when