Behavioural Science Blog

The Science of Human Behaviour

Archive for the ‘Research Methodology’ Category

The Probabilistic Nature of Behavioural Phenotypes

leave a comment »


The probabilistic nature of behavioural phenotypes proclaims that people with a specific syndrome will have a heightened probability to exhibit specific behavior or developmental pathways that are different to others, who do not have the specific syndrome. The advantage of probabilistic nature of the construct is that it can account for exceptions and it is probably more ecological valid that a simple causal explanation. (Short excursion: I believe that statistical models that only test direct effects will be outdated in a few years, as we come to appreciate the complexity of behavioral science. Developmental psychopathology is leading the way with extensive use of structural equation modelling and a flexible theoretical framework that works with risk and protective factors. Such models have to potential to raise explanatory value beyond the simple causal models that are employed in most research today.) The behavioural phenotypes are less important for individual therapy where extensive diagnose should be performed, but they have stronger implications for intervention. If risk factors can be identified that apply for a certain population, specifically tailored interventions might be given to the whole group instead of individual therapy later on. That way specific skills and behaviors might help mentally retarded children to cross into a developmental stage, that could not have been reached without the intervention. The crucial thing is that intervention might be necessary at an early age, when children are most sensitive to the intervention, even thou the problematic behaviour might arise only in a later developmental stage.

Written by Martin Metzmacher

July 13, 2008 at 10:28 am

Unaided Categorical Psychodiagnosis is not Coherent with Normative Models

with one comment


Normative models require the data to be linear. Each case should be weighted linearly and combined additively. That is certainly not the case in clinical settings. Not all therapists have clear guidelines for themselves on what to base a categorical decision. When comparing the data at hand they rely on statistical models that are fallible. They are subject to the same biases that “normal people” are subject to: They think that the information presented matches the diagnose even if the behaviour could be interpreted as ambiguous and that other people would consent with their diagnosis.

Written by Martin Metzmacher

April 2, 2008 at 1:07 am

Experimental Design for the Study of Explanatory Diagnosis

leave a comment »


The study of explanatory diagnosis can best be conducted with an experiment, manipulating the data in a way so that an ambiguous, specific cause for mental illness is rendered more active in the mind of the clinician. This could be a for example a case about psychotic behaviour that could be attributed to a depression or dementia. One would prime either the concept of depression or dementia beforehand and see if that influences the clinicians explanation of the possible causes of the illness. In fact this poses a 2 (information data) x 2 (priming) design, that allows to test some assumptions about the process of evaluation. For example it would be interesting to see if the clinician is more likely to assimilate data if he or she has been primed with the congruent concept beforehand. The priming manipulation could be done in several ways, supra and subliminally. One ecological valid manipulation would be to have the clinician read other patients cases beforehand that have a distinct diagnosis (either depression or dementia) and see if the still active construct is carried on to the evaluation of the next patient.

Written by Martin Metzmacher

April 1, 2008 at 1:12 am