The study has a number of limitations. We considered various exclusion criteria including diagnosis of ADHD. This was because ADHD has been associated with abnormal reward processing previously as well as various psychiatric symptoms such as depressive symptoms and psychotic symptoms; furthermore ADHD is commonly treated with stimulant medication that may alter brain activation during reward anticipation, making results challenging to interpret. Thus, excluding ADHD is helpful in reducing the chances of finding spurious associations between various domains of psychopathology and altered brain reward anticipation signals; however, it reduces the sample size, thus reducing statistical power, and it makes the sample less representative of the general population. Additional children had to be excluded for other reasons, such as artefacts, movement in the scanner or non-engagement with the task. The study data were collected on multiple MRI scanners, which is likely an additional source of noise and variance. We try to compensate for this by taking scanner site into account in statistical analyses. Acquiring data at a single site using a single scanner could be preferable, in terms of reducing sources of variability; however, a major advantage of a multi-site study is the potential for much larger sample size. There are many measures of potential interest that can be generated from the MID task. We calculated reward anticipation by contrasting anticipation of a large reward versus no reward was it is an effective and reliable contrast in eliciting ventral striatal and medial prefrontal reward anticipation associated activation. We used this same contrast in our previous MID study (Arrondo et al 2015) and as did Jia et al from IMAGEN in the largest MID study previously. Many other related contrasts could have been used simply to probe reward anticipation, and the task data could also or alternatively be used to examine anticipation of punishment, brain responses to reward or punishment receipt or omission of expected rewards and/or punishments; numerous connectivity analyses are also possible. Whilst these alternative analyses would be of considerable interest, each additional analysis increases the liklihood of type I error unless rigorous correction for multiple comparison is made, in which case the likelihood of type II error is increased. It is possible that a voxel-wise or vertex-wise analysis could be more sensitive than the approach we used, of regionally averaged scores. However, we note that the regionally averaged scores revealed highly significant task activation in the expected areas of striatal subdivisions, thalamus and medial frontal cortex, supporting their validity.
In summary, in spite of the MID task eliciting robust subcortical and cortical activation in a large sample of 9-10 year olds, there were no strong associations with severity of psychopathology (psychotic-like experiences or depressive or anxiety symptoms) in this age group. It will be of considerable interest to examine whether baseline activation in this age is predictive of the development of future psychopathology, or whether abnormal patterns of brain activation emerge during the maturation process, perhaps associated with concomitant development of specific of general measures of psychopathology.