Please download to get full document.

View again

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
  AGL StimSelect 1 AGL StimSelect: Software for automated selection of stimuli for Artificial Grammar Learning Todd M. Bailey School of Psychology Cardiff University Emmanuel M. Pothos Department of Psychology Swansea University Please address correspondence to Todd Bailey, School of Psychology, Cardiff University, Cardiff CF10 3AT, UK, or Emmanuel Pothos, Department of Psychology, Swansea University, Swansea SA2 8PP, UK. Electronic mail may be sent at   or   Running head : AGL StimSelect; Word count : 7,865 (including abstract)  AGL StimSelect 2 Abstract Artificial Grammar Learning (AGL) is an experimental paradigm that has been used extensively in cognitive research for many years to study implicit learning, associative learning, and generalization based either on similarity or rules. Without computer assistance it is virtually impossible to generate appropriate grammatical training stimuli along with grammatical or non-grammatical test stimuli that control relevant psychological variables. We present the first flexible, fully automated software for selecting AGL stimuli. The software allows users to specify a grammar of interest, and to manipulate characteristics of training and test sequences, and their relationship to each other. The user thus has direct control over stimulus features that may influence learning and generalization in AGL tasks. The software enables researchers to develop AGL designs that would not be feasible without automatic stimulus selection. It is implemented in Matlab.  AGL StimSelect 3 Artificial Grammar Learning (AGL) experiments test people’s sensitivity to sequential dependencies. In its most common form, an AGL study involves letter strings (e.g.  XVSX  ) which do or do not conform to some simple set of rules (a finite state grammar). Strings that conform to the rules are  grammatical   (G strings), and strings that do not conform are ungrammatical   (NG strings). Without knowing anything about rules, participants study a sample of strings that conform to the rules. After being told that the studied stimuli all conformed to an unspecified set of rules, participants are then asked to observe a set of novel strings and decide which are G and which are NG. The first AGL study was presented in 1967 (A. Reber, 1967) and since then there have been more than 125 studies, spanning research themes as diverse as implicit cognition (Berry & Dienes, 1993; Pothos, 2007; A. Reber, 1993; Shanks, 2005), associative learning (Boucher & Dienes, 2003; Perruchet et al., 2002; Servan-Schreiber & Anderson, 1990), rules vs. similarity (Ashby et al., 1988; Pothos, 2005), and cognitive neuropsychology (including amnesia, Alzheimer’s disease, and Parkinson’s disease; Know lton & Squire, 1996; Poldrack et al., 2001; P. Reber & Squire, 1999; Witt, Nuehsman, & Deuschl, 2002). The AGL paradigm has also been used recently to study the psychopathology associated with dyslexia (Pothos & Kirk, 2004), and to examine cognitive processes supporting the maladaptive behavior of alcohol abuse (Pothos & Cox, 2002). In principle, the structural properties of AGL stimuli can be controlled to a high degree of specificity to allow rigorous examination of different theories of learning. For example, Vokey and Brooks (1992; Brooks & Vokey, 1991) manipulated the similarity of test strings to training strings, across both G and NG test items. Knowlton and Squire (1996) factorially combined test string grammaticality with “chunk strength” to test e ffects of the relative familiarity of subsequences within the test strings (pairs or triplets of letters, called chunks or fragments). Others have combined grammaticality with both similarity and chunk  AGL StimSelect 4 strength (e.g. Meulemans & van der Linden, 2003). Another possibility is to experimentally control some properties (particularly grammaticality), and then analyze the effects of other  properties through mathematical modeling (Johnstone & Shanks, 1999; Pothos & Bailey, 2000). The research above took great care to control important stimulus properties. It is far from trivial to construct AGL stimulus sets that simultaneously control multiple  psychologically relevant factors. At present, AGL researchers employ trial-and-error methods to identify appropriate stimuli, with more or less assistance from whatever software tools they develop themselves on an ad hoc basis. These informal methods are perhaps sufficient to control two or three key stimulus properties across small stimulus sets with fewer than about 50 test items. However, these methods do not scale up well to more complex designs. Also, although AGL studies typically aim to draw general conclusions, they usually rely on a single set of stimuli on which all participants are tested. Although it would often be preferable to test each participant on a different set of stimuli (Dienes & Altmann, 2003; R. Reber & Perruchet, 2003; Redington & Chater, 1996), it is not usually feasible to identify more than one suitable set of stimuli for a given study using the present trial-and-error methods. Thus, the lack of an automated procedure to generate AGL stimuli is a major limitation in AGL research. In this paper we present AGL StimSelect, a software package that automatically generates training and test strings that embody structural properties that can be selected by the user from a flexible and extendable set of parameterized constraints. With StimSelect, a user can quickly generate AGL stimuli, while controlling multiple variables that are likely to influence performance on an AGL task. Program Design
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!