Annual Scientific Meeting 2011 abstracts and links
Annual Scientific Meeting 2011 abstracts and links
Music and amusis - an experience sampling study
Diana Omigie, Goldsmith’s University (London)
Congenital amusia (CA) is a developmental disorder characterized by deficits in melody perception and production. Empirical research into this condition has the potential to throw light on questions like how the brain processes music and why music listening can bring such pleasure to its listeners. However recent work has focused mainly on the perceptual abilities of people with C.A. and there has been little research into whether and how the observed perceptual deficits affect their appreciation of music in every day life. Assessing the degree with which amusics willingly engage with music in everyday life is a useful way of inferring how they feel about it and consequently evaluating the relationship between music perception and appreciation. We used hierarchical cluster analysis to evaluate the experiences ampling data collected from a total of 34 participants (17 amusics and 17 matched controls) and then observed how amusics were distributed over the resulting analysis solution. We found that at least 60 percent of amusics demonstrated a listening profile that was clearly distinct from that of controls. These amusics showed little evidence of wanting to engage with music in their everyday life. However the remaining amusics who fell into the cluster that mainly contained controls showed evidence of normal music appreciation; choosing to listen regularly and reporting obtaining pleasure from it. This research is important as it reveals contrasting attitudes towards music within the condition known as Congenital amusia. The reasons why appreciation may arise in the absence of normal perception are explored and further analysis is carried out to try to explain why the two groups of amusics show such different attitudes to music.
Negative attitudes towards learning research methods (RM) are associated with poor grades and dropout. The present cross-sectional internet survey explored preferences for learning RM and factors associated with RM grades. Psychology students (N = 134) from high school to postgraduate level reported that more interaction with their teacher would improve grades. Those with most RM difficulty also wanted: practical work, visual teaching aids, more interesting textbooks, humour in teaching, smaller seminar groups, and more seminars. Using ordinal regression, the significant predictors of better RM grades relative to grades for other modules were: lower stress (p < .001), more advantaged social background (p < .005), and internal locus of control (p < .013). The effect of motivation was mediated by stress. Dyscalculia was not associated with RM grade. These findings have implications for ways to improve the teaching of research methods to psychology students.
Multiple hypothesis testing when hypotheses are related logically
using Shaffer’s R test: A hierarchical step down procedure with a step up test at each step
Andrew Rutherford, Keele University
Hochberg’s (1988) presented a powerful test based on Simes’ (1986) inequality. Rom (1990) later improved this test by defining and calculating exact p-values. Later, Hochberg and Rom (1995) described how to apply their tests when hypotheses were related logically, as described by Shaffer (1986). However, as Hochberg and Rom’s (1995) account is not easily understood by sophisticated statistical mathematicians, it presents real problems for most other scientists. This may explain the lack of application of this important work. In response to my request for worked examples, Juliet Shaffer provided valuable insights into multiple hypothesis testing and Dror Rom delivered a new short-cut method to test logically related hypotheses, which he named Shaffer’s R test. The background and application of Shaffer’s R test will be described.
Latent class analysis is a powerful technique that enables researchers to glean insights into ‘hidden’ psychological experiences. It has been used in a variety of domains, such as with attempts: to understand psychosis as measured along a continuum of symptom expression; to identify features of computer games that are integral to the gaming experience; and to assess the characteristics of a range of trauma and suicidal behaviour typologies. The technique is grounded in the psychometric approach and item response theory and is a versatile method to dealing with nominal data in a deep and psychologically grounded way. This presentation will involve discussion of the key principles and practices when undertaking a latent class analysis. To illustrate the art and science of latent class analysis, a case example will draw on the nuances of data obtained from a general population survey of over 4,300 respondents and will model a set of various class solutions to unearth the barriers to engagement in the arts that could be present within a community.
Towards a rational use of mathematics in the psychology of reasoning
Andy Fugard, University of Salzburg
Logic is the mathematics of reasoning. Traditionally, logic in the psychology of reasoning was taken to mean classical logic and most non-mathematicians received a diagnosis of illogical. However, there are many logics, including probability logics. For so-called "basic" conditionals such as "If the card shows a square, then it's red", most people's degree of belief is given by a conditional probability, P(red|square), a justifiably rational interpretation. However, a significant minority treat an "if" as an "and" (conjunction), which is less easy to justify. Previously we found that, given a long series of trials, many of those initially showing a conjunction interpretation spontaneously shift to a conditional probability. In this talk I will present evidence that automatic stimulus-oriented processes are responsible for the conjunctions and inhibitory function is required for a shift. In another experiment we tested the effect of different expressions of identical (from the perspective of probability theory) conditionals, e.g., for conditionals concerning four cards numbered 1 to 4: "If the card shows a 2, then it shows a 2 or a 4" versus "If the card shows a 2, then it shows an even number". For the former type of conditional, most participants' degree of belief was 0, versus 1 for the latter type of conditional. A theory of relevant deduction, originally developed for classical logic, explains these two interpretations. These results illustrate the utility, and limitations, of logic for guiding theorizing and designing experiments about how non-mathematicians reason.
Adaptive design for model discrimination
Maarten Speekenbrink, University College London
Psychology is rich in formal models of learning, categorization and decision-making, to name but a few areas. While competing models differ in their substantive assumptions, they often make highly similar predictions. For this reason, model comparison based on empirical data is often inconclusive. Optimizing the design of an experiment for model discrimination is difficult, especially when individual participants differ widely in terms of model parameters. To resolve this problem, we present a method to design experiments adaptively whilst running them, at each trial choosing the stimulus which is expected to minimize the entropy of the posterior probability distribution over a set of competing models. We show the advantages of adaptive design in simulation study. We then present data from an experiment in which the method was applied to discriminate between competing models of category learning, including an exemplar model (the Generalized Context Model) and a decision bound model.