This paper is a review of more than 35 years of research on artificial food colors and how children behave. It also draws on testimony given to an FDA advisory committee. It does not run its own study. Instead it pulls together earlier trials and meta-analyses, including large tests of non-clinical children in Southampton (277, 153, and 144 kids, some preschoolers and some aged 8 to 9).
The authors conclude that food colors have a small but real harmful effect on children's behavior. This effect is not limited to children diagnosed with ADHD. The authors frame the dyes as more of a general public-health issue than a specific cause of ADHD.
The effects measured across the cited studies were small. One review found an effect judged too small to matter (0.11). Others found effects in parent ratings closer to 0.18 to 0.28. One analysis shrank from 0.18 to 0.12 once a bias toward positive results was accounted for.
The authors stress that the underlying studies were all flawed, with problems such as weak diagnosis methods, imperfect or missing blinding, and dyes tested in mixtures rather than one at a time. They call the evidence inconclusive but too large to ignore, and suggest limiting children's exposure to dyes until safety is clearer.
Source: Arnold L, Lofthouse N, Hurt E. Artificial Food Colors and Attention-Deficit/Hyperactivity Symptoms: Conclusions to Dye for. Neurotherapeutics (2012). doi:10.1007/s13311-012-0133-x