Autism and Inhibition of Return - A note about sample size
Posted Oct 03 2008 11:31am
A review of: Rinehart, N.J., Bradshaw, J.L., Moss, S.A., Brereton, A.V., Tonge, B.J. (2008). Brief report: Inhibition of return in young people with autism and Asperger's disorder. Autism, 12 (3), 249-260. DOI: 10.1177/1362361307088754
The authors designed an interesting experiment based on Minshew’s Complex Information Processing theory of autism, which seeks to understand the neuropsychological patterns of strengths and weaknesses in Autism as the foundations for the specific deficits in social cognition observed in Autism.
Side Note: Please note that the ‘Neuropsychological’ is usually misinterpreted, even by trained researchers and clinicians, as referring to physiological, as in ‘biological psychology’. Instead, neuropsychological refers in general to neurocognitive functioning – that is, cognitive and motor domains linked to brain processes. Thus, a test of neuropsychological functioning would include assessment of memory, attention, motor control, visual perception, auditory skills, general intelligence, language, etc.
Neuropsychological researchers have noted intact or superior abilities to detect unique patterns or items in visual search tasks in children with autism. This is inconsistent with the finding that these children also show difficulty with visual orientation and attentional set-shift (controlling the shifting of attention when needed from one set of items/tasks to others). Thus, how could children with autism have excellent item detection skills in light of their difficulty with attention and visual orientation? One possible explanation explored by the authors is that children with autism have a pronounced Inhibition of Return (IOR). IOR is a cognitive process that facilitates visual search by inhibiting searching on areas that have already been searched. For example, when looking for a letter ‘p’ on a poster full of letters ‘b’ and ‘d’, IOR allows you to search more effectively and those with a more pronounce IOR would be faster.
In this study the authors compared 12 kids with high functioning autism, 12 kids with Asperger’s, and 12 typically developing kids, all matched for age (mean 10), sex, and IQ. Diagnoses were confirmed via ADI. The participants completed a series of tasks to measure IOR. The authors found no significant differences in IOR between those with ASD and typically developing kids. However, there was a trend at the (p = .052) level suggesting a more pronounced IOR among children with Asperger’s than the other two groups. This is worth noting because the authors ability to find statistically significant results is directly affected by the number of participants in each group. Thus, if these same results had been obtained with more participants, it is very likely that the difference observed would be statistically significant. This sample size related effect is more of a concern when the N of participants is very small. Thus, I will be more likely to ‘trust’ non-significant findings from a study using 1,000 participants than those from a study using only 20. One rule of thumb you can use when examining studies with small sample size (number of participants) is that statistically significant results are more "informative" than results not showing statistical significance. Why? Because the small sample size makes it more difficult to find statistically significant differences, so they require greater differences between the groups you are comparing. Thus, when you see a study using a small sample size that found a statistically significant difference, you can assume that the difference between the groups is there (assuming no other methodological problems). However, when the results show no statistical difference, this could mean that the difference between the groups does not exist, or that they needed a larger sample size.