9 February 2011
UB Department of Psychology
and Center for Cognitive Science
Mapping Individual Variations in Learning Capacity
ABSTRACT:
Individual differences in learning capacity are evident in humans and most other animals. Traditionally, such differences are described in terms of variations along a relatively small number of psychological dimensions corresponding to behavioral traits. Here, an alternative approach is considered in which individual differences in learning capacity are characterized by spatially sorting behavioral patterns. To illustrate this approach, a two-dimensional, self-organizing, feature map was used to analyze patterns in the performances of intact and cortically-lesioned rats engaged in multiple learning tasks. After training, the spatial structure of the map revealed systematic variations in learning across rats that were related to the degree of brain damage. Individual nodes within the map described prototypical performance profiles that corresponded closely to patterns of learning seen in individual rats, including individuals with idiosyncratic profiles. Techniques that automatically identify modal patterns of performance during learning may provide new insights into the processes that determine what an individual organism can learn.
RECOMMENDED READING:
Mercado, Eduardo, III (2011), "Mapping Individual Variations in Learning Capacity", International Journal of Comparative Psychology 24: 4–35.
16 February 2011
UB Department of Psychiatry
Ontology Research Group
New York State Center of Excellence in Bioinformatics and Life Sciences
Spatiotemporal Reasoning in Referent Tracking Systems:
Why Is It So Hard?
ABSTRACT:
Referent Tracking is a paradigm for representing portions of reality that requires strict adherence to the principles of Ontological Realism and the representational elements offered in the Basic Formal Ontology (BFO) and the Relation Ontology. Whereas BFO provides representational units for repeatable entities, Referent Tracking offers a syntax for describing particulars and their relationships. One principle is that representations should mimic the structure of reality, and this in turn requires keeping track at all times of whether representations are about (1) first-order entities, (2) beliefs, or (3) representations themselves. While building domain ontologies and using these ontologies for data annotation under the BFO constraints is already hard, developing reasoning schemas that can be used to infer on the basis of observations or descriptions what other (non-observed or non-described) entities exist is extremely challenging. We here report on our efforts to develop reasoning schemas for the annotation of situations captured on video, thereby paying in the first place attention to the many pitfalls that come with naïve solutions.
RECOMMENDED READING:
Ceusters, W.; Corso, J.; Fu, Y.; Petropoulos, M.; & Krovi, V. (2010), "Introducing Ontological Realism for Semi-Supervised Detection and Annotation of Operationally Significant Activity in Surveillance Videos", Proceedings of the 5th International Conference on Semantic Technologies for Intelligence, Defense, and Security (STIDS 2010), Fairfax, VA, October 27–28.
2 March 2011
Department of Psychology and Department of Linguistics
University of Michigan
Bounded Optimality in Language, Thought, and Action:
Adaptation under Cognitive Constraint
and Its Implications for Understanding Individual Differences
ABSTRACT:
In this presentation, we explore a theoretical framework that construes cognitive and linguistic processing as boundedly optimal control problems—as rational processes constrained by both the structure of the external environment and the structure and limitations of the cognitive architecture. Underlying the approach are computational methods for evaluating large spaces of possible behavioral strategies in terms of their expected utility given these constraints, rather than their fit to observed data. We demonstrate the generality of the approach through its application to elementary dual-tasking, "fast-and-frugal" decision making, verbal short-term memory, and eye-movement control in reading. A key theoretical payoff is an understanding of individual differences in performance as the empirical signatures of strategies that are adaptations to individually varying processing constraints. We discuss how the framework builds on and complements related approaches, including rational analysis, bounded rationality, Bayesian modeling, architectures, reinforcement learning, and signal-detection theory. The key feature of bounded optimality is the theoretical role assigned to processing constraints: They are used to help define the optimization problem, rather than used to explain departures from optimality.
RECOMMENDED READING:
Howes, Andrew; Lewis, Richard L.; & Vera, Alonso (2009), "Rational Adaptation under Task and Procesing Constraints: Implications for Testing Theories of Cognition and Action", Psychological Review 116(4): 717–751.
9 March 2011
Communication Science and Disorders
School of Health and Rehabilitation Sciences
University of Pittsburgh
Automatic Processing and Recovery of Complex Sentences in Aphasia
ABSTRACT:
The production and the comprehension of syntactically complex sentences is impaired in aphasia. For example, both Wh-movement sentences (such as object-extracted, relative clauses) and NP-movement sentences (such as passives) elicit chance performance by adults with aphasia in off-line comprehension tasks like sentence-picture matching. However, it remains unclear how exactly impaired adults try (and often fail) to comprehend such sentences in real time. This talk reviews evidence from a series of studies examining the on-line comprehension of complex sentences by adults with aphasia (Dickey & Thompson 2004, 2009; Dickey, Choy, & Thompson 2007). The evidence suggests that significant residual capacity for syntactic processing remains following brain damage, capacity that may go undetected using traditional, off-line methods. Furthermore, this intact capacity may be what underlies successful response to language treatment targeting complex sentences (Dickey & Thompson 2007). Language treatment that directly stimulates aphasic adults' capacity to use this residual ability has significant evidence of efficacy, and on-line tasks that tap this ability may be useful in predicting treatment outcomes in aphasia (Dickey & Yoo 2010).
RECOMMENDED READING:
Dickey, Michael Walsh; & Thompson, Cynthia K. (2009), "Automatic Processing of Wh- and NP-Movement in Agrammatic Aphasia: Evidence from Eyetracking", Journal of Neurolinguistics 22: 563–583.
6 April 2011
Department of Linguistics
Institute for Advanced Computer Studies (UMIACS)
and Department of Computer Science
University of Maryland
The Linguistics of Spin
ABSTRACT:
Most work on computational analysis of people's attitudes relies on words that express overt opinions, e.g., using the word "awesome" in a movie review as a clue to the fact that it views the movie favorably. However, underlying perspective can also reside in less obvious linguistic choices. The chairman and president of BP America, describing BP's Deepwater Horizon oil spill, testified to Congress that "eleven people were lost in an explosion and fire" (May 11, 2010). In contrast, the progressive media outlet Democracy Now! described the "explosion that killed eleven workers" (May 13, 2010). The two statements describe the same event, but they differ in their underlying view of what happened. The difference illustrates how language can be used "to select some aspects of a perceived reality and make them more salient in a communicating text, in such a way as to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation". Entman (1993) calls this framing, and deliberately framing in a way that manipulates or deceives is sometimes called spin.
In the first part of this talk, I introduce the idea of grammatical framing, i.e., framing accomplished via choice of grammatical structure. I demonstrate that grammatical framing is fundamentally connected to underlying properties of events that are well known to lexical semanticists, and show how observable syntactic reflexes of those properties can be used, fully automatically, to accurately label a text with respect to its perspective on a topic, even in the absence of overtly opinionated language. In the second part of the talk, I discuss my group's recent work on lexical framing, which includes computational models to detect underlying divergences of viewpoint and the words and phrases associated with those divergences.
This research includes joint work with Stephan Greene, Jordan Boyd-Graber, Eric Hardisty, and Viet An Nguyen.
RECOMMENDED READING:
Greene, Stephan, & Resnik, Philip (2009), "More than Words: Syntactic Packaging and Implicit Sentiment", HLT-NAACL 2009 (Association for Computational Linguistics): 503–511.
13 April 2011
Department of Psychology
Barnard College
I Would Know that Voice Anywhere!
The Role of Phonetic Sensitivity in the Perceptual Identification of Individual Talkers
ABSTRACT:
A listener's ability to identify a familiar talker is often ascribed to sensory samples of the acoustic attributes of vocal quality. In idealizations of this aspect of speech perception, unique, long-term characteristics of the vocal source of acquaintances are represented in a gallery in long-term memory, and such characteristics function as standards for evaluating an unknown signal that challenges the auditory system. The ability to identify a linguistic message inheres in a different set of acoustic properties, those of finer grain that underlie the perception of consonant and vowel sequences used to identify spoken words. Neuropsychological findings of a dissociation between aphasia and phonagnosia suggest a system architecture in which the perception of a linguistic message is independent of the perception of the identity of the talker who produced it. The plausibility of this conceptualization can be assessed in light of our studies of individual identification without recourse to auditory impressions of familiar vocal quality. This evidence shows that phonetic attributes can contribute to the perception and identification of individual talkers.
RECOMMENDED READING:
20 April 2011
Department of Psychology
University of Texas at Austin
Retrieving Personal Names
ABSTRACT:
Personal names differ from object names in being more prone to tip-of-the-tongue states, harder to learn, and particularly vulnerable to deficits with brain damage (for review, see Valentine, Brennen, & Brédart 1996). I will discuss results of a study examining the types of substitution errors parents make in addressing their children, which has been the impetus for further research on the use and processing of personal names.
RECOMMENDED READING:
Griffin, Zenzi M. (2010), "Retrieving Personal Names, Referring Expressions, and Terms of Address", in B. Ross (ed.), The Psychology of Learning and Motivation (San Diego, CA: Elsevier), Vol. 53, pp. 345–387.
27 April 2011
UB Department of Philosophy
and Center for Cognitive Science
Moral Objectivism across the Lifespan
ABSTRACT:
The received wisdom among philosophers has been that practically everyone is an objectivist about moral claims; i.e., they treat the truth or falsity of these claims as being as factual as scientific statements about the physical world. I report results from a series of studies that investigate the degree of objectivity that ordinary people attribute to moral claims, showing that moral objectivism is more likely to be endorsed during some stages of life than others and that some of the factors affecting the degree of objectivity attributed to moral claims include the degree of perceived societal disagreement about an issue and the perceived cultural distance between oneself and the person one is making a moral judgment about.
RECOMMENDED READING:
Goodwin, Geoffrey P.; & Darley, John M. (2008), "The Psychology of Meta-Ethics: Exploring Objectivism", Cognition 106: 1340-1366.
14 September 2011
UB Department of Biochemistry
What's So Memorable about NMDA Receptors?
ABSTRACT:
At the vast majority of synapses in brain, two types of glutamate-activated channels, AMPA and NMDA receptors, generate the excitatory postsynaptic current (epsc). These receptors are highly homologous in primary sequence and 3-D structure, yet each generates characteristic electrical signals and mediates distinct functions in synaptic transmission and plasticity. AMPA receptors serve to faithfully transmit electrical signals, whereas NMDA receptors serve to detect memorable signals and to integrate and relay these as post-synaptic calcium fluxes, in effect initiating synaptic plasticity or apoptosis. The lecture will touch upon recent evidence that links biophysical properties of NMDA receptors to these physiologic functions. In particular, the kinetics of glutamate binding/dissociation, activation mechanism, and gating behavior will be discussed as they relate to synaptic plasticity.
RECOMMENDED READING:
28 September 2011
Co-sponsored by the UB Department of Philosophy
Silver Professor of Philosophy, Psychology, and Neural Sciences
Department of Philosophy
New York University
The Overflow Wars
ABSTRACT:
The overflow argument appeals to variants of iconic memory to argue that a conscious phenomenology system "overflows"—that is, has a higher capacity than—the cognitive access system. Recently, these arguments have come under pressure both empirically and conceptually. At the same time, research on a newly recognized variant of iconic memory has generated exciting findings. This talk assesses the overflow argument in light of these recent contributions.
Brief presentations of stimuli give the perceiver the impression that one can grasp and report only a few of the rich array of objects and their properties that one sees. Sperling showed participants an array of letters (for example 3 rows of 4 letters) for a brief period, finding that subjects could report only 3–4 items from the whole matrix but could also report 3–4 items from any row cued after stimulus offset, suggesting their introspective judgment was correct.
This introspective judgment is part of the basis of an argument for overflow. Many have argued that this introspective judgment stems from a kind of cognitive illusion, what has been called the Refrigerator Light Illusion: that the perceiver confuses easy availability upon attending with presence in the visual field. However, it has recently been claimed that perceivers in this type of experiment are victims of a perceptual rather than cognitive illusion, in which they confuse fragments or features of letters with letters, or confuse generic "gists" perceived with diffuse attention with specific letters with specific attended shapes, or are victims of a "postdictive" illusion, or are victims of "inattentional inflation of subjective perception".
This talk reviews the controversy, arguing that the critics of overflow are committed to unconscious image-like representations, a postulation for which there is no independent evidence.
RECOMMENDED READING:
Block, Ned (2008), "Consciousness and Cognitive Access", Proceedings of the Aristotelian Society 108(Part 3):289–317.
5 October 2011
Department of Brain & Cognitive Sciences
University of Rochester
Syntactic Adaptation When Encountering the Unexpected
ABSTRACT:
Many artificial language learning experiments have implicated implicit statistical learning as a feasible mechanism of grammar induction, typically demonstrating that children and adults are able to learn about the underlying structure of sequential input with surprisingly short amounts of exposure to it. In this talk, I will detail the results of a series of experiments highlighting the role that these types of implicit learning mechanisms are likely to play during natural-language processing in adulthood. In all of the experiments to be discussed, subjects encountered distributions of probabilistic cues to syntactic structure that deviated in various ways from the distributions normally encountered in natural language. These deviations produced conditions under which various structural expectations were violated. In each case, subjects exhibited protracted adaptation to the atypical distributions of cue-based information over the course of an experiment, as indexed by cumulative changes in patterns of reaction times elicited in response to expectation violations. This type of short-term learning, which we refer to as syntactic adaptation (Fine, Farmer, & Jaeger (in progress); see also Farmer, Monaghan, Misyak, & Christiansen (in press)) is consistent with accounts of language processing and acquisition that provide for continuous updating of distributional information over time, all the while highlighting the strong interrelationship that exists between processing and learning. The talk will conclude with a discussion of the degree to which the type of learning observed in these experiments is likely to be context-specific versus generalizable, under what conditions it is most likely to occur, and what these types of learning effects mean for current theories of real-time syntactic processing.
RECOMMENDED READING:
12 October 2011
Department of Brain & Cognitive Sciences,
Center for Language Sciences
University of Rochester
Word Order Universals Reflect Biased Learning
ABSTRACT:
In this talk, I provide evidence for a key assumption of generative linguistics—namely, that biases in the language-learning system constrain the space of possible human languages. Recent claims that typological universals do not exist, or are the result of factors outside the cognitive system, have highlighted the need for experimental evidence connecting learning biases to typological preferences (e.g., Evans & Levinson 2009). I focus on a word-order universal, first formulated by Greenberg (1963), which bans a language from using both pre-nominal adjectives and post-nominal numerals. I report the results of several artificial-language learning experiments showing that learning biases can provide an explanation for this universal.
RECOMMENDED READING:
19 October 2011
Department of Brain & Cognitive Sciences
University of Rochester
Neural Basis of Perceptual Learning:
Improved Inference in Early Sensory Areas
ABSTRACT:
Extensive training on simple tasks such as fine orientation discrimination results in large improvements in performance, a form of learning known as Perceptual Learning (PL). Psychophysical experiments manipulating external noise have been used to argue that PL is due to improved decision making. In contrast, single-cell recordings have demonstrated that response properties of neurons in early sensory areas are modified by training, suggesting a change in early sensory representations, not decision making. No model has successfully reconciled these contradictory conclusions. In this talk, considering the example of orientation discrimination, I will argue that specific changes to the feed-forward connectivity between the Lateral Geniculate Nucleus (LGN) and the primary visual cortex (V1) can improve performance in a manner consistent with both psychophysical and neurophsyiological findings. Importantly, the system behaves as if the decision stage has been improved, even though the changes took place early in the model. Further, I will show that the improvement in performance during PL need not be represented by a steepening of tuning curves in early sensory areas—as suggested by previous work. Rather, when learning-induced changes in correlations are taken into account, performance depends crucially on exactly how sensory representations are modified—modifications that improve inference result in improved performance, irrespective of tuning-curve changes.
RECOMMENDED READING:
Bejjanki, V.R.; Beck, J.M.; Lu, Z.L.; Pouget, A. (2011), "Perceptual Learning as Improved Probabilistic Inference in Early Sensory Areas", Nature Neuroscience 14: 642–648.
26 October 2011
Communication Sciences Program
Hunter College (CUNY)
The Resilience of Structure in Talk:
Evidence from Language Acquisition and Language Loss
ABSTRACT:
Within a larger debate in cognitive science, questions about the existence of abstract, structural representations and processes operating independently of specific content have dominated much research in the psycholinguistics of language comprehension, production, and acquisition. In this talk, I approach this question with data from two "limiting cases": language acquisition by young three-year-olds and language loss by older speakers with Alzheimer's Disease (AD). In the first limiting case, I will present data from a syntactic priming experiment with young children, indicating that young, monolingual-English children have more abstract, sentence-level representations than suggested by lexicalist accounts of language acquisition. In the second limiting case, I will present data from a sentence-production task with Italian and English speakers with AD, showing that speakers' knowledge of the fundamental structural properties of their language remains intact even when much else is lost. I will discuss these data within current, theoretical debates in the cognitive neuroscience of language, as well as their implications for language interventions for both children and adults.
RECOMMENDED READING:
9 November 2011
UB Department of Computer Science & Engineering
Affiliated Faculty, Department of Philosophy
Affiliated Faculty, Department of Linguistics
and Center for Cognitive Science
How Cognition Could Be Computing
ABSTRACT:
Computationalism should be the view that cognition is computable; therefore, computationalism can be true even if (human) cognition is not the result of computations in the brain. Semiotic systems should be understood as systems that interpret signs; therefore, both humans and computers are semiotic systems. Minds can be considered as virtual machines implemented in certain semiotic systems, primarily the brain, but also AI computers. I take issue with James H. Fetzer's arguments to the contrary.
RECOMMENDED READING:
Rapaport, William J. (2011), "Semiotic Systems, Computers, and the Mind: How Cognition Could Be Computing" (ms.) [ link]
16 November 2011
UB Department of Linguistics
and Center for Cognitive Science
Personification and Grammatical Gender:
New Ideas and Data from Experimental Research and from German Advertising
ABSTRACT:
Renewed interest in linguistics and cognitive psychology in the effects of language on cognition (neo-Whorfian theory) has led to a flurry of research on whether and how speakers of a gender language tend to personify inanimate concepts in accordance with lexicalized grammatical gender. German, e.g., has masc-gender for ‘key’ (der Schlüssel) and fem-gender for ‘bridge’ (die Brücke), while Spanish does the opposite (la llave and el puente). Do German speakers think of keys as more masculine and bridges as more feminine, while Spanish speakers do the opposite? The presentation will (a) briefly review research that both shows and fails to show this connection between language and cognition, (b) present evidence from German poetry and art traditionally used to argue for the connection, (c) present new evidence from German advertising (print and TV) showing that gender-based personification is alive and well in current German cultural discourse, and (d) using this evidence, examine a set of hypotheses limiting the cognitive reach of personification, including:
RECOMMENDED READING:
30 November 2011
Co-sponsored by the UB Department of Philosophy
Co-Director, Center for Cognitive Studies
University Professor
Austin B. Fletcher Professor of Philosophy
Tufts University
PLEASE NOTE: This talk will take place in the Student Union Theater
Failures of Imagination
and the "Mystery" of Consciousness
ABSTRACT:
It is difficult to imagine a theory of consciousness, but that does not mean it is impossible, as some philosophical pessimists like to insist. With some help from earlier revolutionary thinkers, in particular Darwin and Turing, we can imagine a family of paths to pursue that can explain consciousness—but we have to abandon a few, deeply held intuitions. Those who refuse to entertain the prospect of abandoning those intuitions will just have to sit on their hands and hope I'm wrong, since, by their own accounts, these intuitions declare the game over: We'll never, ever understand consciousness! Time will tell.
RECOMMENDED READING:
Dennett, Daniel (2009), "Darwin's ‘Strange Inversion of Reasoning’", Proceedings of the National Academy of Sciences Vol. 106, Suppl. 1 (16 June): 10061–10065.
7 December 2011
UB Department of Linguistics and Center for Cognitive Science
and Department of Linguistics and Information Science,
Heinrich Heine University Düsseldorf
and
Max Planck Institute for Psycholinguistics
The Argument from the Poverty of the Stimulus Redux
ABSTRACT:
Berwick et al. (2011) revisit some of the classic poverty of the stimulus [POS] arguments advanced by Chomsky in earlier work (e.g., Chomsky 1975, 1986) and maintain that they remain unrebutted. They argue that there are four interacting factors that are involved in the development of human capacities like language (2011:1209):
The first three are most relevant to language acquisition, and the crucial issue revolves around the existence of (1)—innate, domain-specific factors—which have been taken by generative linguists to constitute the Language Acquisition Device/Universal Grammar. POS arguments are the primary source for evidence in favor of, and concerning the nature of, (1). This talk explores a classic POS argument, one that is widely considered to be one of the most compelling ones for Chomsky's position: the constraints on the formation of WH-questions, relative clauses, and related constructions known as "island constraints". Data from two languages are examined, English and Lakhota (a Siouan language of North America), and it will be argued that the Chomskyan analysis of these phenomena strongly supports the conclusion that there are principles of grammar that cannot be learned. [NOTE: No prior detailed knowledge of Chomskyan syntactic theory presupposed or required.] An alternative account of these restrictions is then proposed and applied to the two languages. An account of how these restrictions could be learned is presented, with special emphasis on what the relevant evidence available to the language learner could be. The implications of this competing account for the argument from POS are discussed, and it is suggested that there is evidence for innate, domain-general factors—i.e., (2)—but not for innate, domain-specific factors—i.e., (1).
RECOMMENDED READING: