eng TQMP Tutorials in Quantitative Methods for Psychology 1913-4126 2006-09-01 2 2 38 42 article Guy L. Lacroix gll@carleton.ca 1 Denis Cousineau denis.cousineau@umontreal.ca 2 Carleton University Université de Montréal This special issue of Tutorials in Quantitative Methods for Psychology presents four papers on the Power Law of Learning to celebrate the 25th anniversary of A. Newell and P. Rosenbloom’s (1981) seminal paper “Mechanism of Skill Acquisition and the Law of Practice”. This introduction highlights the main points of Newell and Rosenbloom’s work and then presents the contributors’ articles. http://www.tqmp.org/Content/vol02-2/p038/p038.pdf Editorial power law eng TQMP Tutorials in Quantitative Methods for Psychology 1913-4126 2006-09-01 2 2 43 51 article Paul S. Rosenbloom Rosenbloom@usc.edu 1 University of Southern California This article traces a line of research that began with the establishment of a pervasive regularity in human performance – the Power Law of Practice – and proceeded through several decades' worth of investigations that this opened up into learning and cognitive architecture. The results touch on both cognitive psychology and artificial intelligence, and more specifically on the possibility of building general learning mechanisms/systems. It is a story whose final chapter is still to be written. http://www.tqmp.org/Content/vol02-2/p043/p043.pdf Power law SOAR eng TQMP Tutorials in Quantitative Methods for Psychology 1913-4126 2006-09-01 2 2 52 65 article Graig P. Speelman c.speelman@ecu.edu.au 1 Kim Kirsner pkirsmer@bigpond.net.au 2 Edith Cowan University University of Western Australia The Newell and Rosenbloom (1981) depiction of the power law of learning implies that improvements in task performance that result from practice can be described by a power function with one variable for amount of practice. We suggest that performance on all but the simplest of tasks relies on component skills that differ in their practice history. As a result, power functions with one term for practice could not be expected to provide accurate descriptions of learning curves. In particular, transfer situations that involve a mixture of old and new skills are likely to lead to perturbations in learning curves that require more than the simple version of the power law to describe. We explore the types of functions that are necessary in these situations and note the impact of transfer factors on learning rates. http://www.tqmp.org/Content/vol02-2/p052/p052.pdf Power law Transfer of learning eng TQMP Tutorials in Quantitative Methods for Psychology 1913-4126 2006-09-01 2 2 66 76 article Karl M. Newell kmn1@psu.edu 1 Gottfried Mayer-Kress gmk@santafe.edu 1 Yeou-The Liu ytl@where.edu 2 The Pennsylvania State University Taiwan Normal University The central proposal of A. Newell and Rosenbloom (1981) was that the power law is the ubiquitous law of learning. This proposition is discussed in the context of the key factors that led to the acceptance of the power law as the function of learning. We then outline the principles of an epigenetic landscape framework for considering the role of the characteristic time scales of learning and an approach to system identification of the processes of performance dynamics. In this view, the change of performance over time is the product of a superposition of characteristic exponential time scales that reflect the influence of different processes. This theoretical approach can reproduce the traditional power law of practice – within the experimental resolution of performance data sets - but we hypothesize that this function may prove to be a special and perhaps idealized case of learning. http://www.tqmp.org/Content/vol02-2/p066/p066.pdf Power law Superposition of exponential curves eng TQMP Tutorials in Quantitative Methods for Psychology 1913-4126 2006-09-01 2 2 77 83 article Denis Cousineau denis.cousineau@umontreal.ca 1 Guy L. Lacroix guy_lacroix@carleton.ca 2 Université de Montréal Carleton University Response time data in learning experiments show a typical trend. They start out slow, quickly improve, before finally tending toward optimal performance. This trend provides critical information that can be used to test various theories of learning. One convenient way to characterize the data is the use of a learning curve; an idealized curve that passes through the observed data points as a function of training. This idealized curve has free parameters that must be estimated using optimization techniques. In this tutorial, we show how to estimate learning curve parameters using three softwares (Excel, SPSS, and Mathematica) assuming that the idealized curve is a power function. The techniques can easily be adapted to other functions. Finally, details are provided on related topics (maximizing block sizes, testing curvatures, etc.). http://www.tqmp.org/Content/vol02-2/p077/p077.pdf Power law How-to guide eng TQMP Tutorials in Quantitative Methods for Psychology 1913-4126 2006-09-01 2 2 84 84 article The CMU Soar research group cmu@cmu.edu 1 Carnegie Mellon University Courtesy of the Carnegie Mellon University Archives/Allen Newell Collection. http://www.tqmp.org/Content/vol02-2/p084/p084.pdf Power law Obituary