William Bialek
April 20, 2007

Searching for simple models

The world is a complicated place. In trying to understand it, physics often has made progress by focusing on simple model problems, places where we can build our intuition before tackling the full complexity of nature. Can we do this is cognitive science, or does putting "simple" and "cognitive" in the same sentence already mean that we are talking nonsense?

In this lecture I'll look at how a small corner of the fly's brain makes sense out of the visual world, and at how networks of neurons in the salamander retina cooperate to generate surprising collective behavior. These systems certainly are simpler than the human brain, and this provides us with an opportunity to push our understanding much further, to the point where we really have a mathematical theory for what is going on rather than just parameterized models. Within this theoretical framework, we can see, at least tentatively, how to bridge the gap from the simpler systems to the more challenging problems of cognition and the dynamics of large networks in the cortex. I'll emphasize the concrete predictions that we can make, and perhaps most importantly I'll argue that the lessons from simpler systems point to new kinds of experiments that we should be doing in order to give a deeper characterization of cognitive function.

Elissa L. Newport
March 17, 2006

How children shape languages: Language acquisition and the emergence of signed and spoken languages

As human children and adults learn their native languages, two apparently distinct phenomena occur. First, children (and, to some degree, adults) are remarkably adept at learning the details of the particular language to which they are exposed. Children exposed to English learn English, while those exposed to Japanese learn Japanese. To account for how they do this, Richard Aslin and I have been developing an approach to language acquisition known as ‘statistical learning.’ Our basic idea is that human language acquisition involves naturally and unconsciously computing, over a stream of speech, such things as how frequently sounds co-occur, how frequently words occur in similar contexts, and the like. Learners use these computations to determine regular versus accidental properties of the language and to learn its rules. Our studies show that adults, infants, and even nonhuman primates perform such computations online and with remarkable speed, on both speech and nonspeech materials.

At the same time, children do not always acquire what they are exposed to: under certain circumstances, they reliably change languages. Studies of the emergence of Nicaraguan Sign Language, as well as other signed and spoken languages, suggests that children are a prime force in developing and expanding languages. Our research shows that this phenomenon can also be incorporated into an understanding of statistical learning. Even in the lab, given certain types of input, learners reliably change the patterns of the language; and children do this strikingly more often than adults.

Taken together, our studies of language acquisition under natural and laboratory circumstances are beginning to help us understand how children learn and also create languages.

Colin Camerer
February 25, 2005

Behavioral Game Theory

Colin Camerer's specialty in the last few years has been "behavioral game theory", a subfield (or "franchise") of behavioral economics which uses experimental evidence to establish how psychological limits on the ability to make calculations and plan ahead, the way in which people react to fairness, and learning from experience, influence behavior in situations described by "game theory". Game theory is a mathematical analysis of any social situation in which one player-- typically a person, but possibly a firm or nation-- tries to figure out what other players will do, and choose the best strategy given those guesses about others. Most game theory describes the fictional behavior of an ideal, hypercalculating, emotionless player (like Dr. Spock from Star Trek) and, as a result, is not always a good guide to how normal people who don't plan too far ahead will actually behave. My 2003 book Behavioral Game Theory describes hundreds of different experimental studies which show where game theory predicts well and predicts poorly, and suggests some new kinds of theory. Behavioral game theory gives precise predictions about how people who think only a couple of steps ahead, have both guilt and envy toward others, and learn from experience, are likely to behave.

Ray Jackendoff
February 20, 2004

Toward a Cognitive Science of Culture and Society

It is a commonplace nowadays to speak of social categories and cultural institutions as being "socially constructed." This lecture will explore the question of what it takes for an individual to participate in such "socially constructed" entities -- what kind of mind one has to have in order to live in a human society. The question has strong parallels with contemporary linguistics, which investigates the cognitive structure required to be a speaker of a human language (another "socially constructed" entity). As in the case of language, the study of social cognition is consistent both with acknowledging cultural diversity and with a substantive universal cognitive framework that underlies the ability to learn and function in one's culture. The inquiry leads quite directly to connections with issues in anthropology, primatology, evolutionary psychology, legal and moral philosophy, economics, and religion, as well as more traditional cognitive sciences such as linguistics and developmental psychology.

Handout: Toward a Cognitive Science of Culture and Society

Geoffrey Hinton
March 7, 2003

Learning Representations by Unlearning Beliefs

Neural networks need to learn good ways of representing the data they receive from sensors. When there is no teacher to specify how each sensory input ought to be represented, a network can learn useful features and constraints by discovering combinations of sensor values that occur much more often or much less often than might be expected. Using its features and constraints, the network can associate a "surprise level" with any pattern of sensory values. A good set of features and constraints is one that associates a low surprise level with real sensory data and a high surprise level with all other possible combinations of sensor values.

It is very hard to evaluate a proposed set of features and constraints because all possible combinations of sensor values must be considered. Nevertheless, there is a simple way of improving the network's current features and constraints. Starting with a real pattern of sensor values, the network makes many small adjustments to the data to make it less surprising. Having corrupted the data to fit in with its current beliefs, the network then modifies its features and constraints to make the real data less surprising and the corrupted data more surprising. I shall show some examples of this learning procedure in action.

Stanislas Dehaene
March 7, 2002

Cerebral Bases of the Number Sense in the Parietal Lobe

What are the origins of the human sense of numbers and arithmetic? Parietal cortex is systematically activated whenever we calculate, and its lesion can cause severe deficits of number manipulation. The human parietal lobe may therefore contain a category-specific representation of numerical quantity. An alternative possibility, however, is that it is merely involved in generic visuo-spatial and/or linguistic processes that are not specific to the number domain. To address this issue, several fMRI experiments will be presented. Those experiments reveal a systematic map of visuo-spatial, language, and calculation activations in the parietal lobe, amongst which an area in the middle intraparietal sulcus responds solely during number processing, not to other linguistic or spatial tasks. This region shows notation-independent subliminal quantity priming, suggesting that it unconscious encodes the quantity meaning of numbers. A new model is proposed according to which a quantity-specific area of the intraparietal sulcus interacts with other regions of the parietal lobe involved in phonological and spatial attention processes. The model suggests an evolutionary expansion of the inferior parietal lobule in humans, and helps understand the patterns of adult and developmental dyscalculia.

Martin Nowak
February 9, 2001

Evolution of Language: From Animal Communication to Universal Grammar

Abstract Not Available

Elizabeth Spelke
January 21, 2000

Core Knowledge and Cognitive Development

This talk explores three old-fashioned ideas about human thinking: that it differs qualitatively from that of all other animals, that it increases in scope and power as children learn, and that it depends on language. Although decades of research and argument have cast doubt on these ideas, I will defend them all, drawing first on studies of spatial memory and navigation and then on studies of number and arithmetic in animals, children, and human adults. These studies suggest that a critical ingredient in human intelligence is our promiscuous capacity to combine old concepts in new ways, and that natural languages provide the medium for these combinations.

Daniel Dennett
October 2, 1998

Things About Things

Every science thrives on oversimplifications, and cognitive science is no exception. What makes cognitive science cognitive is its assumption of states or processes or events that exhibit intentionality or aboutness, but some of the most popular idealizations that attempt to capture aboutness turn out to be false friends. There are alternative things about things that offer somewhat better prospects.

---

The Benjamin and Anne A. Pinkel Endowed Lecture Fund was established though a generous gift from Sheila Pinkel on behalf of the estate of her parents, Benjamin and Anne A. Pinkel, and serves as a memorial tribute to the lives of her parents. Benjamin Pinkel, who received a BSE in Electrical Engineering from the University of Pennsylvania in 1930, was actively interested in the philosophy of the mind and published a monograph on the subject, Consciousness, Matter, and Energy: The Emergence of Mind in Nature, in 1992, the objective of which is a "re-examination of the mind-body problem in the light of...new scientific information." The lecture series is intended to advance the discussion and rigorous study of the deep questions which engaged Dr. Pinkel's investigations.

Current year's lecture