(When sufficient titles, and abstracts for talks and posters arrive, I will begin posting them in this section)
| B |
| Speaker's Name: | Bruno Bocanegra |
| First Author's Name: | Bruno Bocanegra |
| First Author's Affiliation: | Erasmus University, Rotterdam |
| Title: | Emotion impairs high frequency spatial vision |
| Abstract: | In order to respond adaptively to threat in the environment our brains are equipped with specialized mechanisms that enhance the visual processing of emotional events. Although a previous study indicated that emotion enhances vision, the generality of this finding remains unknown. Do the benefits of emotion extend to all basic dimensions of vision or are they limited in scope? Here, we provide the first demonstration that emotion not only improves but also impairs low-level vision. Our results indicate that the brief presentation of a fearful face enhances orientation sensitivity for low spatial frequency gabors, but diminishes sensitivity for high spatial frequency gabors. We show this counterintuitive pattern of benefits and deficits is due to a sensitivity shift across the spatial frequency spectrum, triggered by the global facial configuration in the fearful expressions. Consistent with previous neuroimaging data, the selective low-frequency benefits suggest that emotion enhances magnocellular visual processing. Additionally, we propose that the high-frequency deficits might be due to cross-inhibition between magno- and parvocellular visual pathways. Our results reveal an emotional mechanism that improves the detection of coarse features at the expense of fine-grained visual details, presumably in order to facilitate responses to motivationally significant stimuli. |
|
| Speaker's Name: | Gordon D.A. Brown |
| First Author's Name: | Gordon D.A. Brown |
| First Author's Affiliation: | University of Warwick |
| Second Author's Name: | Caroline Morin |
| Second Author's Affiliation: | University of Warwick |
| Title: | Time-based models of memory |
| Abstract: | Time-based models of memory differ in a number of important ways. Models that represent the locations of items as point sources in time appear to fail on data that require consideration of the temporal extension of items. We describe a new relative-time based model that builds on the positive features of existing temporal models but can yet accommodate data that prove problematic for previous models. We also report the results of a number of experiments designed to test the predictions of the new model. |
|
| Speaker's Name: | Tom Busey |
| First Author's Name: | Tom Busey |
| First Author's Affiliation: | Indiana University, Bloomington |
| Second Author's Name: | Larry Humes |
| Second Author's Affiliation: | Indiana University, Bloomington |
| Third Author's Name: | James Craig |
| Third Author's Affiliation: | Indiana University, Bloomington |
| Fourth Author's Name: | Diane Kewley-Port |
| Fourth Author's Affiliation: | Indiana University, Bloomington |
| Title: | Temporal order judgments measured in auditory, visual and tactile
modalities: Effects of Aging. |
| Abstract: | Age-related declines in temporal processing have been noted in all three primary sensory
modalities and are consistent with a general slowing of temporal processing. To date,
however, no study has addressed commonalities among these modalities. In a project
with over 90 elders and 20 younger participants we measured sensitivity in several
different temporal order tasks that vary in their informational complexity and spatial
location of the stimuli. Despite a literature that suggests a common cause for age-related
declines that span sensory and cognitive tasks, we find a pattern of results that is more
consistent with a relative independence of the three modalities. This result obtains
despite the fact that we carefully match the tasks across the modalities. We discuss
the implications of these results, as well as the data that suggest commonalities among
the modalities. |
|
| D |
| Speaker's Name: | Eddy Davelaar |
| First Author's Name: | Eddy Davelaar |
| First Author's Affiliation: | Birkbeck, University of London |
| Title: | Primary memory: contributions to free recall performance |
| Abstract: | On the same page on which William James introduced the terms primary and secondary memory, he presented his view on how they are implemented in the brain, which is not different from Hebb's view. Here I start with the view that primary memory is the activated part of long-term memory, or better yet that primary memory is the process (see also, Norman, 1968) by which temporarily activated long-term knowledge is maintained in active state beyond its expected unaided life-time. The current debate in the literature is whether a short-term buffer needs to be postulated in order to account for performance on free recall tasks (Brown, et al, 2007; Davelaar, et al, 2005; Howard, et al, 2007). I will briefly touch on dissociations in recency effects over the short- and long-term, on dissociations between immediate and longer-term free recall tasks, and some points of confusions. One source of confusion is the unclarity of the concept of activation-based short-term store. I will show using simulations why an activation-buffer is different than a fixed-box buffer and its association with working memory models that assume that the content of working memory is the activated part of long-term memory (Cowan, 1999; McElree, 2006; Oberauer, 2002). |
|
| Speaker's Name: | Simon Dennis |
| First Author's Name: | Simon Dennis |
| First Author's Affiliation: | Ohio State University |
| Second Author's Name: | Dennis Mehay |
| Second Author's Affiliation: | Ohio State University |
| Title: | What do cows drink? |
| Abstract: | In this talk, we will present the case for a return to an associative framework for understanding sentence processing. Firstly, we will briefly review the literature on the impact of association on sentence processing in behavioral and event related potential studies. Then we will propose an associative model, based on multinomial logistic
regression (a single layered softmax network), and show how it is able to account for phenomena which have been argued to preclude simple associative models. These include the ability to process center embedded recursive structures and the ability to form hierarchical phrase structure representations. By training this model on 1M
sentences from the gigaword corpus, we show that the approach is scalable and can predict several results concerning when sentences will be difficult for people to process. Finally, we present new results that question the psychological reality of propositional
representations and demonstrate how an extended version of the associative model is able to fulfill some functions that have previously been attributed to propositions. We then use the extended model to answer the central question "What do cows drink?". |
|
| F |
| Speaker's Name: | Simon Farrell |
| First Author's Name: | Simon Farrell |
| First Author's Affiliation: | University of Bristol |
| Title: | How relative are positional representations in serial recall? |
| Abstract: | The evidence from serial recall of temporally grouped lists suggests that items are associated with some external positional or timing signal, and that this signal is used to hierarchically organize sequences of items. Some models (e.g., the Start-End Model of Henson, 1998), hold that all items in a group are anchored both to the start and the end of that group: a relative representation. Using a competitive comparison of computational models, we show that the existing evidence, focussing on confusions of terminal group items between groups of different sizes, does not mandate a model in which all items are anchored to the end of a group (i.e, an extensive end marker), but is instead consistent with a simpler alternative model in which terminal items are discretely tagged as end items (a restricted end marker). We then present some new data using lists grouped in a 4-4 fashion, which allows inspection of confusions of internal items between groups. These more discriminating data, and model fits, suggest that an extensive marker does play a role, though limited, in representing order in short-term memory.
Henson, R. N. A. (1998). Short-term memory for serial order: The Start-End Model. Cognitive Psychology, 36, 73-137. |
|
| Speaker's Name: | Stefan Frank |
| First Author's Name: | Stefan Frank |
| First Author's Affiliation: | Radboud Univ. Nijmegen |
| Title: | Testing the Surprisal Theory of Word-reading Time |
| Abstract: | According to so-called "surprisal theory" (Hale, 2001; Levy, 2008), a word's probability
of occurrence given its sentence context is inversely logarithmically related to the time
required to read that word. Tests of this theory have implicitly assumed that the
(subjective) probabilities a reader assigns to words correspond to the (objective)
probabilities as extracted from text corpora. If surprisal theory is correct and subjective
probabilities indeed correspond to objective probabilities, an objectively more accurate
probability model should also provide more accurate predictions of word-reading times.
To investigate whether this relation holds, we compared two models that can generate
word probabilities and that have been suggested as a basis for psycholinguistic models
of sentence processing. However, they make very different assumptions as they originate
from disparate fields: One is a Simple Recurrent Network (SRN; the quintessential
connectionist model), the other a Probabilistic Context-Free Grammar (PCFG; the
standard model in computational linguistics). Both models were trained on a part of the
Wall Street Journal corpus and tested on the Dundee corpus, which contains both
newspaper texts and corresponding eye-movement data. Preliminary results show that
the SRN generates more accurate word probabilities, whereas the PCFG provides
better predictions of reading times. This suggests that subjective probabilities cannot
be estimated by objective probabilities or that surprisal theory is simply incorrect.
If we do hold on to the theory, the results indicate that a system based on tree-structures
forms a better psycholinguistic model than does a connectionist system. |
|
| Speaker's Name: | Robert M. French |
| First Author's Name: | Rosemary A. Cowell |
| First Author's Affiliation: | Univ. de Bourgogne |
| Second Author's Name: | Robert M. French |
| Second Author's Affiliation: | Univ. de Bourgogne |
| Title: | The Emergence of Rules in Category Learning:
A Semi-supervised Neural Network Model |
| Abstract: | We present a neural network model of category learning that addresses the question of how rules for category membership are acquired. The architecture of the model comprises a statistical-learning (Kohonen) network and a rule network whose weights, crucially, emerge from the statistical network. The statistical-learning network is implemented with a neurobiologically plausible Hebbian learning mechanism and forms category representations on the basis of perceptual similarity. The rule network gradually extracts rules from the statistical-learning network to discover which of the stimulus features are sufficient to determine category membership. These rules are weightings of individual features; weights are stronger for features that convey more information about category membership. We demonstrate that this model predicts a cognitive advantage in classifying perceptually ambiguous stimuli over a system that relies only on perceptual similarity. In addition, it produces reaction time data that reflect the level of agreement between the statistical- and rule- learning components. Finally, the model demonstrates that occasional feedback greatly enhances the categorization performance of the system, which has implications for the “poverty of the stimulus” debate. |
|
| H |
| Speaker's Name: | Pernille Hemmer [6-10 for the talk] |
| First Author's Name: | Pernille Hemmer |
| First Author's Affiliation: | University of California Irvine |
| Second Author's Name: | Mark Steyvers |
| Second Author's Affiliation: | University of California Irvine |
| Title: | A Bayesian Account of Reconstructive Memory |
| Abstract: | It is well established that prior knowledge influences reconstruction from memory, but the specific interactions of memory and knowledge are unclear. Extending work by Huttenlocher et al. (1991, 2000) we propose a hierarchical Bayesian model of reconstructive memory in which prior knowledge interacts with episodic memory at multiple levels of abstraction. The combination of prior knowledge and noisy memory
representations is dependent on familiarity. We present empirical evidence of the hierarchical influences of prior knowledge, showing that the reconstruction of familiar objects is influenced toward the specific prior for that object, while unfamiliar objects are influenced toward the overall category. |
|
| Speaker's Name: | Jared Hotaling |
| First Author's Name: | Jared Hotaling |
| First Author's Affiliation: | Indiana University |
| Second Author's Name: | Andrew Cohen |
| Second Author's Affiliation: | UMass, Amherst |
| Third Author's Name: | Jerry Busemeyer |
| Third Author's Affiliation: | Indiana University |
| Fourth Author's Name: | Richard Shiffrin |
| Fourth Author's Affiliation: | Indiana University |
| Title: | Information Integration in Perceptual Decision Making |
| Abstract: | One set of researchers studying judgment and decision making have repeatedly shown that people employ simple and often less than optimal strategies when integrating information from multiple sources. Another set of researchers has had great success using Bayesian optimal models to explain information integration in fields such as categorization, memory, and perception. In an attempt to understand the differences in a common paradigm, we studied the dilution effect in a perceptual decision setting. The dilution effect has been studied in many verbal problems typically stated in terms of probabilities: Subjects often judge combined weak and strong evidence favoring a single outcome to lie between the two: in effect the weak evidence has diluted the strong evidence. In our perceptual studies, information from the upper (U) and lower (L) halves of faces had to be combined to make a decision whether the test face T matched target face A or B. Sometimes the face halves were split apart horizontally, forcing observers to combine matching information cognitively. Suppose there is strong evidence U (say) that produces odds for A over B: O(A/B|U) >> 1.0. Suppose there is weak but still positive evidence L giving odds for A over B: O(A/B|L) > 1.0. Bayesian analysis would say the odds given both U and L should be the product of the two separate odds, thus increasing the resultant odds. We found that splitting faces produced a tendency to dilute evidence combination compared with non-split faces. I will present a preliminary model of the mechanisms underlying these results. |
|
| Speaker's Name: | David E. Huber |
| First Author's Name: | Yoonhee Jang |
| First Author's Affiliation: | University of California, San Diego |
| Title: | Measuring the statistical dependence between familiarity and recall |
| Abstract: | Following study of paired words, participants were given a sequence of test trials that alternated between 1) forced choice recognition of single words and 2) use of the target from step 1 as a cue for recall of the studied associate. This recognition followed by cued
recall procedure produces a 2X2 table for memories that do or do not support recognition combined with memories that do or do not support cued recall. In addition, recognition performance was manipulated through immediate repetition priming (i.e., the Jacoby-Whitehouse paradigm) and also through speeded versus non-speeded responding.
Multinomial Processing Tree (MPT) models of these data were used to examine the influence of priming on recognition and the influence of speeded responding on recognition. Finally, an appropriate MPT was used to measure the statistical dependence between familiarity based recognition versus recall based recognition. |
|
| K |
| Speaker's Name: | George Kachergis |
| First Author's Name: | George Kachergis |
| First Author's Affiliation: | Indiana University |
| Second Author's Name: | Richard Shiffrin |
| Second Author's Affiliation: | Indiana University |
| Title: | Context Effects in Recognition Memory: A Bayesian Analysis |
| Abstract: | Many people have had the experience of knowing what song will play next on an album (even one heard only a few times). Conversely, many people fail to recognize an acquaintance encountered in an unfamiliar context. Our studies explore the automatic or incidental storage of associations between successively encountered words on a list (storage of such associations is a key assumption found in the models of Howard and Kahana, and the REM-II model of Mueller, Nelson, and Shiffrin). We find evidence in episodic recognition memory for such storage. We explore the effects of the previous item’s familiarity, semantics, and study-list position on the accuracy and response time for a target word preceded at test by same, similar, or different context words. We are in the process of using Hierarchical Bayesian data analyses to analyze the data and formulate models. |
|
| L |
| Speaker's Name: | Daniel Lafond |
| First Author's Name: | Daniel Lafond |
| First Author's Affiliation: | Université Laval |
| Second Author's Name: | Yves Lacouture |
| Second Author's Affiliation: | Université Laval |
| Title: | Testing the Predictive Accuracy of Decision Tree Models of Categorization |
| Abstract: | This work examines the descriptive and predictive accuracy of three decision tree models of categorization adapted from Trabasso, Rollins and Shaughnessy (1971). These models aim to provide a quantitative account of categorization response times, choice proportions and typicality judgments at the individual-participant level. Study I modeled results from Cohen and Nosofsky's (2003) experiment. Overall, the decision tree models achieved
comparable fits to that of two exemplar models, the EGCM-RT (Lamberts, 2000) and the EBRW-PE (Cohen & Nosofsky, 2003). In Study II, we replicated and extended Cohen and Nosofsky's experiment by asking participants to give subjective typicality ratings for each stimulus. A post-test phase called the "four-questions game" (Sayeki, 1969) provided the constraints required to systematically identify a unique decision tree for each participant. Model I, II and III showed increasingly good fits to the data, which follows from their respective complexity and flexibility. However, there is a risk that more flexible models provide better fits simply by adjusting to noise in the data (overfitting). Cross-validation tests showed that the decision tree models had a good predictive accuracy-though overfitting was observed. The key result, however, is that without any mathematical modeling, the process tracing method used to identify decision trees successfully predicts how participants classify new stimuli. |
|
| Speaker's Name: | Stephan Lewandowsky |
| First Author's Name: | Stephan Lewandowsky |
| First Author's Affiliation: | University of Western Australia |
| Title: | Short term-memory: New data and a model |
| Abstract: | I review the extensive benchmark data on short-term memory for order, and discuss the prominent computational theories accounting for serial order memory. On the basis of recent diagnostic results, I identify four explanatory constructs that must be instantiated in a model in order to provide an adequate account of those data. I present one such model (C-SOB), developed by Simon Farrell and myself and show that it handles existing benchmark data as well as recent diagnostic results. Some of the model’s novel predictions are explored, for example the relationhip between response suppression and recency, and the interplay between the nature of interfering material and the extent of observed forgetting. |
|
| Speaker's Name: | Casimir J.H. Ludwig |
| First Author's Name: | Casimir J.H. Ludwig |
| First Author's Affiliation: | University of Bristol |
| Second Author's Name: | Lucy A. Ellis |
| Second Author's Affiliation: | University of Bristol |
| Third Author's Name: | Iain D. Gilchrist |
| Third Author's Affiliation: | University of Bristol |
| Fourth Author's Name: | Roger H.S. Carpenter |
| Fourth Author's Affiliation: | Cambridge |
| Fifth Author's Name: | Simon Farrell |
| Fifth Author's Affiliation: | University of Bristol |
| Title: | The mechanism underlying inhibition of saccadic return |
| Abstract: | Inhibition of Return (IOR) refers to a bias against orienting to a location that was inspected recently. Inhibition of Saccadic Return (ISR) appears a closely related phenomenon in which human observers take longer to re-fixate an already visited location. We present data from a novel, gaze-contingent saccade sequences paradigm in which 3 successive target locations are cued. When observers are directed to re-fixate the immediately preceding target location a reliable ISR was found for low contrast, peripheral cues and for symbolic, central cues. To identify the underlying mechanism of ISR saccade latency distributions from individual observers were fit with a set of
sequential sampling models. These models share the assumption that activity is accumulated over time to a response threshold. ISR could be explained as a reduction in the accumulation rate for return saccades, as an increase in the separation between baseline and threshold activity levels, or as a combination of the two. Competing model selection showed that ISR is best accounted for as a change in the accumulation rate. We argue this parameter represents the overall desirability of a particular course of action, which may be derived from a variety of sensory and non-sensory sources. The diminished desirability of return saccades may reflect learned assumptions about the temporal structure of the natural world. |
|
| M |
| Speaker's Name: | Gail McKoon |
| First Author's Name: | Gail McKoon |
| First Author's Affiliation: | Ohio State University |
| Second Author's Name: | Jessica Love |
| Second Author's Affiliation: | Ohio State University |
| Title: | Pronouns and Rules of Engagement |
| Abstract: | In 1992, Greene, McKoon, and Ratcliff suggested that readers might not resolve pronouns in sentences like "Judy saw Thomas steal a car but she didn't tell the police." Readers might not know, at the end of the sentence, whether it was Judy or Thomas who didn't tell the police. Greene et al.'s data supported this suggestion with on-line single-word recognition experiments: From before a pronoun (before "she" in the Judy-Thomas sentence) to the end of the sentence, there was no increase in accessability for the referent (Judy) relative to the nonreferent (Thomas). In new experiments, we show how simply readers can be engaged into a story world where pronouns are completely understood. |
|
| Speaker's Name: | Ben Murdock |
| First Author's Name: | Ben Murdock |
| First Author's Affiliation: | University of Toronto |
| Title: | The TODAM working memory model for serial-order effects in short-term memory. |
| Abstract: | The TODAM working memory model (TWM) is a revision of two previous TODAM
serial-order models. Unlike its predecessors (but like many other contemporary models) an item combines content and context where context drifts slowly but continuously through study, distraction, and test. It is an augmented chaining model where each item is associated with its predecessor and is stored in a common memory vector. Retrieval is the
correlation of either the retrieved item or the retrieved information and deblurring is assumed to be successful if the best match exceeds the next-best match by a criterion. TWM can explain two basic serial-order effects (the serial position effect and the list-length effect) with no free parameters, can explain various serial recall, probe recognition, and serial learning effects with varying degrees of success, and is not
subject to the standard objections to a chaining model. |
|
| N |
| Speaker's Name: | Angela Nelson |
| First Author's Name: | Angela Nelson |
| First Author's Affiliation: | Indiana University |
| Second Author's Name: | Rich Shiffrin |
| Second Author's Affiliation: | Indiana University |
| Title: | The influence of context and recency on the effects of trained frequency |
| Abstract: | A previous experiment by Nelson & Shiffrin showed that training subjects on novel Chinese characters to varying degrees produced frequency effects in post-training memory and perception tasks similar to those found for word frequency. The current experiments expand on this research by examining the effects of contextual diversity
(experiment 1) and recency of presentation (experiment 2) on various memory and perception tasks. In experiment 1, 16 subjects were trained on novel Chinese characters using the same visual search task used by Nelson & Shiffrin. The characters varied in their frequency of occurrence, as well as in their contextual diversity: half the subjects
were trained in a variable contextual diversity condition, and half in a controlled contextual diversity condition. Experiment 2 also trained subjects on novel Chinese characters in visual search with differing frequency of occurrence. In addition to frequency, the recency of occurrence of each character was varied: half the subjects viewed the high frequency characters more recently (in relation to post-training
tests) than the low frequency, and the other half viewed the low frequency characters more recently than high frequency. The results of these two studies will be discussed, as well as their relation to our current model of semantic and episodic memory development. |
|
| O |
| Speaker's Name: | Klaus Oberauer |
| First Author's Name: | Klaus Oberauer |
| First Author's Affiliation: | University of Bristol |
| Second Author's Name: | Stephan Lewandowsky |
| Second Author's Affiliation: | University of Western Australia |
| Title: | Implementing the Time-Based Resource Sharing Model |
| Abstract: | The complex span task is the paradigm most frequently used to measure working memory capacity. It consists of an immediate serial recall task with a distracting processing task in between presentation of the items. Barrouillet and colleagues have presented the Time-Based Resource-Sharing (TBRS) model to account for performance in complex span tasks (Barrouillet, Bernardin, & Camos, 2004). Memory traces are assumed to decay unless refreshed by an attentional bottleneck. The bottleneck time must be shared between refreshing and the distractor task. The model accounts for three key findings with the complex span task: (1) Performance in complex span is much worse than in comparable simple span tasks; (2) Performance depends on the proportion of time between items that is required for the distractor task, and (3) The length of the distractor activity has little effect on memory performance. So far, the TBRS has been presented only informally. We present several alternative computational implementations and show that the intuitive predictions of effects 1-3 arise from the model only with specific assumptions about the details of decay and rehearsal. We also investigate whether variants of TBRS in which interference replaces decay can generate equivalent predictions. |
|
| P |
| Speaker's Name: | Diane Pecher |
| First Author's Name: | Diane Pecher |
| First Author's Affiliation: | Erasmus University, Rotterdam |
| Title: | The Testing Effect: Lowering the rate of forgetting. |
| Abstract: | Long-term retention can be enhanced by intermediate testing compared to additional study. The most interesting finding is that at short retention intervals performance is better for materials that received additional study rather than for materials that were tested, but at long retention intervals (e.g., one week after study) performance is better for materials that were tested rather than for materials that received additional study. Thus, the forgetting rate is lower in the test condition than in the additional study condition. We studied this testing effect for various materials (unrelated word pairs, foreign languages, geographical knowledge). We found slower forgetting even when the final test was different from the intermediate test. We also found that the more difficult the intermediate test was, the lower the forgetting rate. These findings clearly have implications for education. |
|
| R |
| Speaker's Name: | Jeroen Raaijmakers |
| First Author's Name: | Jeroen Raaijmakers |
| First Author's Affiliation: | University of Amsterdam |
| Title: | A critical discussion of the inhibition theory for forgetting |
| Abstract: | Recent years have seen a revival of the idea that memory traces may be actively suppressed and that such suppression may lead to a decrease in performance on a later memory test. Although the inhibition account appears to be relatively successful there are a number of problems with the theoretical accounts that are given. Some of these problems are quite similar to the problems that plagued the "unlearning" notion in the classical Two-Factor Theory of interference and forgetting. I will discuss these developments and highlight some of the problems with a strict application of the inhibition assumption. |
|
| Speaker's Name: | Roger Ratcliff |
| First Author's Name: | Roger Ratcliff |
| First Author's Affiliation: | Ohio State University |
| Title: | Modeling decision processes |
| Abstract: | I will talk about 2-3 topics in simple decision making. First is the use of EEG measures that demonstrate trial by trial variability in drift rate in the diffusion model. The second is application of the model to research in sleep deprivation showing the effects on model parameters of 57 hours of deprivation. The third is the application of the model to
data from subjects with reduced blood sugar. |
|
| Speaker's Name: | Cowell Rosie |
| First Author's Name: | Cowell Rosie |
| First Author's Affiliation: | University of Kent |
| Title: | The Effect of Brain Lesions in the Ventral Visual Object Processing Pathway: A Connectionist Model and a Reinterpretation of Old Data |
| Abstract: | A previous model of the effects of perirhinal cortex (PRh) removals on visual discrimination learning in monkeys suggested that behavioural deficits are due to compromised representations of visual stimuli (Bussey & Saksida, 2002). This model is based on a hierarchical organisation of visual representations in the ventral visual stream
(VVS), with simple feature representations stored in a "caudal" layer and representations of the conjunctions of those features stored in a more rostral, "PRh" layer. The model claims that the nature of the stimuli used is critical to whether a deficit will be observed following PRh lesions. Animals with PRh damage are impaired in discriminating complex stimuli that possess "high feature ambiguity", i.e. in which rewarded and non-rewarded stimuli possess many common features.
Here we extend the previous model, enabling simulation of discrimination learning with a lesion of the caudal layer. Following lesions to caudal VVS, the model predicts impaired discrimination of simple stimuli, since representations of simple features - critical to the discrimination - are compromised. In contrast, networks can still acquire discriminations
between complex stimuli. The model accounts for a double dissociation in visual discrimination learning found in studies employing lesions at opposite ends of the ventral visual stream, which have hitherto been explained by positing dissociable functions of memory and perception for rostral and caudal areas of the VVS, respectively (e.g., Iwai and Mishkin, 1968). The model provides an alternative framework for interpreting visual discrimination learning data, placing emphasis not on what brain function is tested by different discrimination tasks, but rather on which brain region contains the representations necessary for the discrimination. |
|
| Speaker's Name: | Adina Roskies |
| First Author's Name: | Adina Roskies |
| First Author's Affiliation: | Dartmouth College |
| Title: | What can neuroscience tell us about free will |
| Abstract: | I will discuss the problem of free will in the context of new information from neuroscience. Specifically, I will consider whether the picture of decision-making that is emerging from neuroscience provides reason to think our actions are determined or not under our control. I provide several reasons to think that our philosophical positions about the correct relation between freedom and determinism is immune to input from the biological sciences, but I also argue that dependency on determinism is misconceived. Instead, I outline another view, and show how our philosophical views about free action can be consistent with a mechanistic model of brain function, and illustrate this by reference to current neuroscientific work. |
|
| S |
| Speaker's Name: | Adam Sanborn |
| First Author's Name: | Adam Sanborn |
| First Author's Affiliation: | Gatsby Institute |
| Second Author's Name: | Thomas Griffiths |
| Second Author's Affiliation: | UC Berkeley |
| Third Author's Name: | Daniel Navarro |
| Third Author's Affiliation: | U. Adelaide |
| Title: | Rational Approximations to Rational Models of Categorization |
| Abstract: | Rational models have been successfully used to explain behavior as the optimal solution to a computational problem in many areas of cognition, including memory, reasoning, generalization, and causal induction. While these models can be used to explore the assumptions people make in a particular task, the computation required to produce the optimal solution is often intractable and thus not a reasonable model of the computations performed by people. To make working with rational models practical, computer scientists have developed approximation algorithms with asymptotic convergence guarantees, such as Gibbs sampling and particle filtering. We propose to use these same algorithms to generate rational process models from rational models of cognition – making the assumption that cognition utilizes these statistical algorithms to approximate intractable rational models. In particular, we show that a particle filter approximation to the Rational Model of Categorization (RMC; Anderson, 1990) can reproduce human data, including more human-like order effects than are produced by the RMC. |
|
| Speaker's Name: | Frederick Schauer |
| First Author's Name: | Frederick Schauer |
| First Author's Affiliation: | John F. Kennedy School of Government and Harvard Law School, Harvard University |
| Title: | Why Precedent in Law (and elsewhere) Is Not Totally
(or even substantially) about Analogy |
| Abstract: | Cognitive scientists and others who do research on analogical reasoning often claim that the use of precedent in law is an application of reasoning by analogy. In fact, however, law’s principle of precedent is quite different. The typical use of analogy, including the use of analogies to earlier decisions in legal argument, involves the retrieval and selection of an analog from multiple candidates in order to help make the best decision now. But the legal principle of precedent requires that a prior decision be treated as binding, even if the current decision-maker disagrees with that decision. When the identity between a prior decision and the current question is obvious and inescapable, precedent thus imposes a constraint quite different from the effect of a typical argument by analogy. The importance of drawing this distinction is not so much in demonstrating that a common claim in the psychological and cognitive science literature is mistaken, but that the possibility of making decisions under the constraints of binding precedent is itself an important form of decision-making that deserves to be researched in its own right. |
|
| Speaker's Name: | Richard M. Shiffrin |
| First Author's Name: | Richard M. Shiffrin |
| First Author's Affiliation: | Indiana University |
| Second Author's Name: | Angela Nelson |
| Second Author's Affiliation: | Indiana University |
| Third Author's Name: | Shane Mueller |
| Third Author's Affiliation: | Klein Associates |
| Title: | The co-evolution of event memory and knowledge |
| Abstract: | We lay out a general approach to the way event encoding, storage and retrieval co-evolve with knowledge storage and retrieval, indicating how it explains episodic memory effects, knowledge retrieval tasks, and priming. The theory is instantiated in a model applied to a study in which Chinese characters are trained in visual search for several weeks. Different characters receive different amounts of training. Then three transfer tests are given: 1) episodic recognition memory; 2) pseudo-lexical decision; 3) forced choice perceptual identification. Frequency effects are found throughout, and we show how the model accounts for these, in part through a context mechanism: An item’s episodic and knowledge traces both accumulate features from items studied in nearby temporal and spatial proximity. |
|
| Speaker's Name: | Vladimir Sloutsky |
| First Author's Name: | Vladimir Sloutsky |
| First Author's Affiliation: | Ohio State University |
| Title: | Statistical Density and Category Learning |
| Abstract: | This research focuses on the role of category structure in category
learning and category representation across points of development. I
introduce a measure of structure – statistical density – or the
proportion of category-relevant variance to the total variance and
consider the role of statistical density in category learning and
category representation. These considerations are tested in multiple
experiments. The results demonstrate a clear dissociation between dense
and sparse categories: Whereas dense categories were readily learned
without supervision, learning of sparse categories required supervision.
There were also developmental differences in how statistical density
affected category representation. Although children represented both
dense and sparse categories on the basis of the overall similarity,
adults represented dense categories on the basis of similarity and
represented sparse categories on the basis of the inclusion rule. The
results support the notion that statistical structure interacts with the
learning regime in their effects on category learning. In addition,
these results elucidate important developmental differences in how
categories are represented, which presents interesting challenges for
theories of categorization. |
|
| Speaker's Name: | Bobbie Spellman |
| First Author's Name: | Bobbie Spellman |
| Title: | Relying on other's statements of confidence |
| Abstract: | How do people decide whether to believe the statements of others? Psychology research shows that people often over-use informants’ statements of confidence as a proxy for their likely accuracy. We call this error the “Presumption of Calibration” and demonstrate that when people are given information about the informants’ confidence and accuracy together, they override the initial presumption and use the
confidence-accuracy relation to assess the witness’s actual calibration. That assessment of calibration is then used to judge credibility. We believe that people use either or both absolute measures of the confidence-accuracy relation (technically “calibration”) and relative measures of that relation (“resolution”). In experiments using trial scenarios, we show that mock-jurors prefer the testimony of well-calibrated to poorly-calibrated witnesses; that they prefer well-calibrated witnesses to more “modest” witnesses; and
that they recognize that justified errors, and understandable lies, do not detract from an informants’ calibration. In analogous studies with children, we show that 5-6 year olds do not evaluate calibration even though they recognize confidence and accuracy alone. We suspect that assessing calibration is a highly attention-demanding process
because it involves acknowledging and encoding the covariation between two types of information. |
|
| V |
| Speaker's Name: | Ingmar Visser |
| First Author's Name: | Ingmar Visser |
| Second Author's Name: | Maartje Raijmakers |
| Third Author's Name: | Emmanuel Pothos |
| Title: | Individual Strategies in Artificial Grammar Learning |
| Abstract: | Artificial Grammar Learning (AGL) has been used extensively to study theories of learning. However, we argue that compelling conclusions cannot be forthcoming without an analysis of individual strategies. We describe a new statistical method for doing so, based on the increasingly popular framework of latent variable models, which are especially suited to capture heterogeneity in participant responses. In the current study, we apply the method of latent class regression models, in which the intercept and regression coefficients can have different values in different latent groups of participants; each latent group represents different reliance on the (potentially) available sources of knowledge in AGL, such as grammaticality and fragment overlap. The results show that grammaticality and fragment overlap can indeed be understood as distinct aspects of learning performance, as evidenced by different groups of participants adopting predominantly one or the other strategy in a series of comparable datasets from AGL studies. |
|
| Z |
| Speaker's Name: | Rene Zeelenberg |
| First Author's Name: | Rene Zeelenberg |
| First Author's Affiliation: | Erasmus University, Rotterdam |
| Second Author's Name: | Gino Camp |
| Second Author's Affiliation: | Erasmus University, Rotterdam |
| Third Author's Name: | Diane Pecher |
| Third Author's Affiliation: | Erasmus University, Rotterdam |
| Title: | Are Independent Cues Truly Independent? |
| Abstract: | The independent cue technique has been developed to test traditional interference theories against inhibition theories of forgetting. In a series of experiments we tested the critical criterion for the independence of retrieval cues: studied cues that are not presented during test (and unrelated to test cues) should not contribute to the retrieval process. During the study phase of the experiment subjects studied weakly related word pairs (e.g., rope-sailing, sunflower-yellow). A subset of the words (e.g., rope) received additional study. In the test phase a category cue (e.g., sports, color) was presented and subjects were instructed to recall an item from the study list that was a member of the category (e.g., sailing, yellow). In three experiments we showed that the additional study of the paired-associate word (e.g., rope) enhanced category cued recall even though this word was not presented at test. I will argue that this result complicates the use of the independent cue technique to differentiate between interference and inhibition theories of forgetting. |
|
| Speaker's Name: | Matt Zeigenfuse |
| First Author's Name: | Matt Zeigenfuse |
| First Author's Affiliation: | University of California, Irvine |
| Title: | Finding Feature Representations of Stimuli: Combining Feature Generation
and Similarity Judgment Tasks |
| Abstract: | A widely-used assumption cognitive modeling is that stimuli are represented in terms of features.
Two experimental approaches to finding appropriate features, and characterizing stimuli in
terms of these features, involve feature generation and similarity judgment tasks. In feature
generation, people list a set of candidate features, and then decide whether or not each
stimulus has each feature. In similarity judgment tasks, people rate the similarity between
pairs of stimuli, and models like additive clustering are used to infer features, and their
patterns of belonging to stimuli. In this paper, we show how relating feature generation
and similarity judgment can provide a powerful method for finding feature representations.
We describe a model that constrains a potentially large set of generated features to only
those that are needed to explain similarity judgments. Using modern computational Bayesian
methods, we apply our model to part of the Leuven natural language database,
considering a set of 30 mammals and 764 candidate representational features. We show
that the inferred feature representation is interpretable, is able to describe the existing
similarities, and provides good generalization performance to withheld similarities. |
|
-- Send in your