Consistent with the hypothesis that better vitamin D status, indicated by higher concentrations of 25(OH)D, protects against schizophrenia we found an inverse association of serum concentrations of 25(OH)D3 and total 25(OH)D with definite psychotic experiences in childhood, though the association was weaker when suspected cases were also included in the outcome. Contrary to expectations from previous studies of total 25(OH)D or those of dietary or supplement intake of vitamin D with schizophrenia, we found that higher concentrations of 25(OH)D2 were associated with psychotic experiences. However, it should be noted that the results are presented per doubling of exposure and thus the associations of 25(OH)D3 and 25(OH)D2 (15% risk decrease and 26%risk increase in the adjusted model, respectively) were modest.
The associations were robust to adjustment for a range of potential confounding factors that we selected based on appropriate methods. 
Decisions regarding such selection can be difficult, e.g. with 25(OH)D3
adjustment for seasonality and UVB exposure might be considered as overadjustment, given their role in determining 25(OH)D3
. However, the associations were similar when we used 25(OH)D3
without adjustment for seasonality (Table S2
) and also when we removed indicators of UVB exposure (data not presented). Adjustment for seasonality in studies assessing the association of 25(OH)D with outcomes can improve statistical efficiency and reduce confounding when the outcome is also likely to be seasonally patterned and does not lead to overadjustment. 
There has been debate about whether BMI or adiposity influences 25(OH)D concentrations or the other way around (i.e. 25(OH)D influences adiposity. A recent Mendelian randomization study supports the direction being form greater adiposity causing lower 25(OH)D, 
than the other way round, which supports inclusion of BMI as a potential confounder.
To our knowledge one previous study has examined the association of vitamin D with psychotic experiences. 
In that large cross-sectional study of women, dietary vitamin D intake, estimated with food frequency questionnaires, was inversely associated with psychotic experiences. It is difficult to directly compare these results with ours as we find the inverse association only with 25(OH)D3
, which is largely determined by UVB, rather than diet. Another prospective study found that maternal report of vitamin D supplementation in the first year of life was protective against schizophrenia in adulthood in men but not in women. 
Since the exposure in that cohort study was maternal report of vitamin D3
supplement use with no detailed data on which supplement the children received, 
it is impossible to know whether that represented predominantly an association of vitamin D3
with outcome. Notably, in these previous studies the outcome was assessed in adulthood, 
while we studied the association with psychotic experiences in childhood/early adolescence.
We had the opportunity to explore the relative associations of both 25(OH)D3
with psychotic experiences. This is important because on the basis of associations with bone outcomes 
and affinity for vitamin D binding protein 
it has been suggested that vitamin D3
is more potent than vitamin D2
but nutritional supplements are available in both forms and for non-bone outcomes any differences between the associations of two forms of 25(OH)D are, as yet, unknown. We have previously shown that although 25(OH)D3
was the major contributor to total 25(OH)D in this cohort, 25(OH)D2
makes an important contribution to total 25(OH)D concentrations in some children, especially those with low total 25(OH)D concentrations. 
Our data linking outdoor behaviour with 25(OH)D3 is consistent with the hypothesis that the children in this cohort largely received their vitamin D3 via sunlight exposure on the skin, not dietary intake. While over a third of the cohort had no detectable 25(OH)D2, it was of interest to note that those with higher concentrations of this form of circulating 25(OH)D (which is exclusively derived from dietary sources such as fortified margarine and cereals as well as supplements) were more likely to report psychotic experiences. This finding may be a chance occurrence, given the lack of any previous studies with which to compare the results. It could reflect residual confounding by one or more characteristics that are related to increased dietary intake of vitamin D2 and also increased risk of psychotic experiences. It could be causal, but as yet we are unaware of any biological mechanism that would explain it or different associations of 25(OH)D3 and 25(OH)D2. We emphasise that this finding should be treated with caution unless replicated in other studies.
Experimental studies suggest that vitamin D is important for neuronal function and brain development 
. The majority of the animal studies that have examined effects on neuronal function have been based on vitamin D receptor knock-out models or developmental vitamin D deficiency. The relevance of these studies to our findings depends on the extent to which the consequences of early life vitamin D deficiency in animal models, including lateral ventricle size/structure, increased cellular proliferation, reduced apoptosis, and altered neurogenesis 
continue to operate in children and adolescents. There is some emerging evidence to suggest that low vitamin D status may impact on brain outcomes differently in the post-natal brain (e.g. vitamin D may be ‘neuroprotective’ and help the post-natal brain cope with stressors). 
In interpreting our findings it is important to consider what psychotic experiences reflect. If they represent an early expression of pathological neurodevelopmental processes that later lead to schizophrenia, our results would have potential implications for predication and for understanding aetiology of schizophrenia. This possibility is supported by suggestions that people with non-clinical psychotic experiences represent a valuable and valid group for studying the aetiology of clinical psychosis, and the similarity of risk factors for non-clinical and clinical psychotic symptoms. 
However, it is also true that in many individuals who report psychotic experiences, these phenomena are transient, and despite the clear association between earlier psychotic experiences and later clinical psychotic disorders, the majority of those with psychotic experiences do not progress to clinical disorders. 
Even if such experiences are not a useful indicator of future clinical outcomes, they are associated with distress and functional impairment, 
and may represent an important public health concern, in the same way that symptoms of depression and anxiety that do not reach thresholds used to formally diagnose depression do. 
Thus, understanding more about the aetiology of such experiences is likely to be important.
The study has several important limitations. We lacked detailed information on supplement intake for the cohort members and thus were not able to infer how much of the circulatory 25(OH)D3
was derived from dietary sources. Although we were able to adjust for confounders, including measures of socioeconomic position and outdoor exposure, it is possible that these dimensions were not fully captured and there is some residual confounding. Reverse causality is possible, for example if those with psychotic experiences were less likely to go outdoors and be exposed to UVB as a result of these symptoms. However, the prospective nature of our study reduces this possibility and prospectively we actually found a weak positive association of reported time spent outdoors with psychotic experiences, making this possibility unlikely. Our findings may be due to chance and thus further replication in large prospective cohorts would be valuable, as would comparing the associations of both forms of 25(OH)D with hard endpoint of clinically diagnosed schizophrenia, in order to establish the true association with this disorder. Mendelian randomization studies, in which genetic variants that are robustly associated with circulating concentrations of 25(OH)D3
are used as instrumental variables to assess their causal effects on outcomes 
would also be valuable, but these would require very large sample sizes. Ultimately, large randomised trials would be required to determine whether supplementation with vitamin D was an effective means of preventing psychosis-related outcomes. 
Our findings do suggest that vitamin D3
would be the more appropriate supplement to be assessed in trials examining the effect of this vitamin on psychotic experiences if these further studies were to suggest causal protective effects of this form of vitamin D.
We only used data from a single measurement of exposures which may be inadequate, 
although a single measurement may be a useful biomarker of season-specific vitamin D status over a longer time. 
. Our sample included mainly white children (the number of non-white children inthe main analyses was 84 (2.6%)), which may limit the generalisability of the results, but does mean that residual confounding due to ethnicity is unlikely. Similar to other prospective cohort studies there was loss to follow-up and those who have attended follow-up clinics tend to be from higher socioeconomic groups. 
However, the prevalence of psychotic experiences was similar in complete cases and those who were excluded due to missing data.
In conclusion, our findings provide some support for an inverse association of total 25(OH)D and 25(OH)D3 concentrations with psychotic experiences, which if psychotic experiences are related to development of schizophrenia, also support a possible protective association of higher 25(OH)D3 concentrations with schizophrenia. The positive association of 25(OH)D2 concentrations with psychotic experiences was unexpected and is currently unexplained. Further replication of our findings in large prospective studies would be valuable, as would the use of studies, such as Mendelian randomized controlled trials and randomised controlled trials, that are more able to establish whether associations are likely to be causal or not.