Additional Reading

The following reading is of relevance to the course. Explore yourself, and feel free to bring any issues, observations, etc to class!

Bullock, T. H., Bennett M. V. L.., Johnston D., Josephson R., Marder E. and Fields R.D. (2005) The Neuron Doctrine, Redux, Science 310." Intro: After a century, neuroscientists are re- thinking the Neuron Doctrine, the fundamental principle of neuro- science. This proposition, developed pri- marily by the great Spanish anatomist and Nobel laureate Santiago Ramón y Cajal, holds that a neuron is an anatomically and functionally distinct cellular unit that arises through differentiation of a pre- cursor neuroblast cell. In principle, part of this tenet has held up, but technology and research have extended our knowl- edge far beyond this simple descrip- tion. What has evolved is a modern view of the neuron that allows a more broad and intricate perspective of how information is processed in the nervous system. One hundred years since its inception, an examination of the Doctrine indicates that it no longer encompasses important aspects of neuron function. If we are to under- stand complex, higher level neuronal processes, such as brain function, we need to explore beyond the limits of the Neuron Doctrine.
Kay, Lily E. From logical neurons to poetic embodiments of mind: Warren S. McCulloch's project in neuroscience. Science in Context 14.4 (2001): 591-614. This article provides some societal context to the emergence of cognitivism, the role of logic in the origins of neural networks, the links between neural networks, information theory, and the military during and after the Second World War, and the driving passions of Warren McCulloch, described as "psychiatrist, experimental epistemologist, poet, militarist, and theological engineer". A good read.
Brandon C. Roy, Michael C. Frank and Deb Roy. (2009) Exploring Word Learning in a High-Density Longitudinal Corpus, Proceedings of the 31st Annual Meeting of the Cognitive Science Society. Amsterdam, Netherlands. Abstract: What is the role of the linguistic environment in children’s early word learning? Here we provide a preliminary analysis of one child’s linguistic development, using a portion of the high-density longitudinal data collected for the Human Speechome Project. We focus particularly on the development of the child’s productive vocabulary from the age of 9 to 24 months and the relationship between the child’s language development and the caregivers’ speech. We find significant correlations between input frequencies and age of acquisition for individual words. In addition, caregivers’ utterance length, type-token ratio, and proportion of single-word utterances all show significant temporal relationships with the child’s development, suggesting that caregivers "tune" their utterances to the linguistic ability of the child.
Bechtel, W. and A. Abrahamsen. (2006) Phenomena and mechanisms: Putting the symbolic, connectionist, and dynamical systems debate in broader perspective. In R. Stainton (Ed.), Contemporary debates in cognitive science. Oxford: Basil Blackwell.

Conclusions: Our strategy through this paper has been to show that the range of phenomena for which mechanistic models are sought is extremely varied and to illustrate briefly some of the kinds of models of mechanisms that have been advanced to account for different phenomena. The focus in the philosophical literature on systematicity and other general properties of cognitive architectures presents a distorted view of the actual debates in the psychological literature over the types of mechanisms required to account for cognitive behavior. Even in the domain of language, where systematicity is best exemplified, many additional phenomena claim the attention of cognitive scientists. We discussed two that are quantitative in nature: the U-shaped acquisition of irregular past tense forms and the exponential acquisition of early vocabulary. Beyond language we have alluded to work targeting the rich domains of perceptual and motor behavior, memory, and problem solving. Phenomena in all of these domains are part of the explanandum of mechanistic explanation in cognitive science. Such explanatory attempts, which like the phenomena themselves often are quantitative, go back as far as Weber’s psychophysics and currently are moving forward in dynamical approaches to perception, cognition and development.

We have also emphasized that cognitive science, despite its many disputes, has progressed by continually combining and recombining a variety of influences. The use of equations both in characterizing and explaining phenomena are among these. When combined with other influences and commitments, the outcomes discussed here have ranged from information processing models with quantified operations to connectionist networks to both global and mechanistic dynamical accounts. Each of these approaches has provided a different answer to the question of whether the mind processes language-like representations according to formal rules, and we have argued that the overall answer need not be limited to just one of these. Cognitive science takes multiple shapes at a given time, and is protean across time.

"I am large, I contain multitudes." (Walt Whitman, 1855, Leaves of Grass. Brooklyn: Rome Brothers)

McClelland, J. L. and Jenkins, E. (1991) Nature, nurture, and connections: Implications of connectionist models for cognitive development. In Architectures for intelligence: The Twenty-second Carnegie Mellon Symposium on Cognition. (pp. 41-73). Hillsdale, NJ, England: Lawrence Erlbaum Associates, Inc. x, 436 pp. Abstract: When it comes to selecting an architecture for modeling cognition, we have a choice, We can start with a symbolic architecture, in which the putative constitu- ents of abstract cognition (symbols) are taken as modeling primatives; or we may adopt an alternative view, that symbolic behavior emerges from the operation of a system of simple, sub-symbolic processing units, Connectionist models take this latter tack, In these models, processing occurs through the propagation of activation among a number of simple processing units, The knowledge that governs processing is stored in the strengths of the connections among the units, And learning occurs through the gradual adjustment of the strengths of these connections, At fust glance it may seem that such mechanisms are far removed from symbolic thought, Yet we will argue in this chapter that they may form the basis of the acquisition of a number of cognitive abilities, and that they may help us answer basic questions about the process of cognitive development. Several different kinds of answers have been given to these questions, We will see how the connectionist framework opens them anew and suggests what may be differ- ent answers in many cases.
Anna C. Schapiro, James L. McClelland (2009) A connectionist model of a continuous developmental transition in the balance scale task, Cognition, Volume 110 Abstract: A connectionist model of the balance scale task is presented which exhibits developmental transitions between `Rule I' and `Rule II' behavior [Siegler, R. S. (1976). Three aspects of cognitive development. Cognitive Psychology, 8, 481-520.] as well as the `catastrophe flags' seen in data from Jansen and van der Maas [Jansen, B. R. J., & van der Maas, H. L. J. (2001). Evidence for the phase transition from Rule I to Rule II on the balance scale task. Developmental Review, 21, 450-494]. The model extends a connectionist model of this task [McClelland, J. L. (1989). Parallel distributed processing: Implications for cognition and development. In R. G. M. Morris (Ed.), Parallel distributed processing: Implications for psychology and neurobiology (pp. 8-45). Oxford: Clarendon Press] by introducing intrinsic variability into processing and by allowing the network to adapt during testing in response to its own outputs. The simulations direct attention to several aspects of the experimental data indicating that children generally show gradual change in sensitivity to the distance dimension on the balance scale. While a few children show larger changes than are characteristic of the model, its ability to account for nearly all of the data using continuous processes is consistent with the view that the transition from Rule I to Rule II behavior is typically continuous rather than discrete in nature.
Schafer, G. and Mareschal, D. (2001) Modeling infant speech sound discrimination using simple associative networks, Infancy, 2(1) Abstract: Infants' responses in speech sound discrimination tasks can be non-monotonic over time. Stager and Werker (1997) reported such data in a bimodal habituation task. In this task, 8-month-old infants were capable of discriminations that involved minimal contrast pairs, while 14-month-old infants were not. It was argued that the older infants' attenuated performance was linked to their processing of the stimuli for meaning. The authors suggested that these data are diagnostic of a qualitative shift in infant cognition. We describe an associative connectionist model showing a similar decrement in discrimination without any qualitative shift in processing. The model suggests that responses to phonemic contrasts may be a highly non-monotonic function of experience with language. The implications of this idea are discussed. The model also provides a formal framework for studying habituation-dishabituation behaviors in infancy.
Van Gelder, T. and R. F. Port (1995) It's About Time: An Overview of the Dynamical Approach to Cognition Chapter 1 of Van Gelder, T. and Port, R. F. (1995) Mind as Motion, MIT Press Chapter 1 of a book from 1995, which sought to introduce Dynamical Systems Theory as a new framework for modeling cognition. The book itself features a wide variety of approaches, not all mutually compatible.
Eck, D. and Jaeger, H. (2007) Can't get you out of my head: A connectionist model of cyclic rehearsal, in: Modeling Communications with Robots and Virtual Humans, Springer-Verlag, 2007 Somewhat clunky model of the earworm phenomenon, where after brief exposure to a repeating auditory stimulus, the tune drives itself, as it were, repeating itself in your poor aflicted head endlesly. Illustrates one potential use of the dynamic capabilities of the Echo State Network, though the implementation is, to my jaundiced eye, somewhat baroque and inelegant.
Dreyfus, Hubert L. (2007) Why Heideggerian AI Failed and How Fixing it Would Require Making it More Heideggerianl, Philosophical Psychologyi>, 20:2, 247-268 Representation is a tricky business. This rather excellent discussion of the shortcomings of representational approaches to artificial intelligence may provide a route in for some of you to the manner in which phenomenology has provided important an important critique of representational approaches even in more modern forms of cognitive science.