Literary reading and eating disorders: survey evidence of therapeutic help and harm.

The highly significant trends revealed in the statistical analysis may in part reflect the fact that respondents report what they think the effects of reading have been, rather than direct effects of exposure to certain text types.

also in 132,142 other papers

These reflections need qualifying by mention of some obvious limitations of the present research, the first being that it is wholly observational: it offers no way of separating correlation from causation, so there may be significant confounding factors (such as severity of illness, or motivation for recovery) at play in the results.

also in 13,097 other papers