In my view, this blog entry does not accurately represent the method and findings of the study to which it refers, which is newly published in Archives of General Psychiatry (and which I just read). For example, it is quite inaccurate to say that the study “cannot say a single word about causation.” When it comes to drawing causal conclusions, there is a big difference between correlations in a cross-sectional study (data collected at a single time point) and correlations in a longitudinal study that follows the same people over time (often called a panel design). With cross-sectional data, causal order generally cannot be determined (e.g., does media use lead to depression, or do depressed people use more media?). This study used a longitudinal design. The authors carefully reviewed the different explanations for an association between depression and media use, and note that most past studies have been cross-sectional. They analyzed data collected at two time points (7 years apart) from a nationally representative sample of participants who were in grades 7 to 12 at time 1. In addition, they excluded from their analysis adolescents who were depressed at time 1. They also included numerous controls to rule out other factors. The findings showed that television and overall media use in adolescence was linked with the *later* development of depression. Of course, no single study can “prove” any hypothesis, and many other factors undoubtedly contribute to the development of depression. And as this blog entry noted, no study can ever measure “all relevant variables.” In my judgment, rather than glibly dismissing the findings, a more complete discussion of the study, including its strengths (e.g., the longitudinal design and inclusion of controls) and limitations, would have helped to counteract misinformation about mental health issues that is often presented in the news media. I should note that I have no connection to this study or this specific research topic, but I do study and teach media uses and effects.

Well, to say it’s a “good longitudinal study” and then dismiss it anyway does not seem to be a productive response. The *point* of a longitudinal panel study is to examine relationships in a sample over time, which can meet at least the time-order requirement for causality (that is, the cause must precede the effect). Of course it cannot PROVE causality, but it adds evidence that is consistent with that possibility. The authors themselves were clear in acknowledging this.

But let’s consider the example of teenager A, and the question that was posed: If teenager A “becomes more depressed by age 22, would we not reasonably expect their life activities to reflect their state of mind?” Of course their activities at age 22 should reflect their state of mind at age 22. But this is NOT what the study found! Instead, to extrapolate the findings to this example, the study found that what teenagers were doing AT AGE 17 was related to how depressed they were at age 22. Again, it is the associations OVER TIME that are unique to this study!

As for the opposite possibility that was mentioned: If the study had found that watching more TV in high school was associated with LESS depression 7 years later (instead of more), I would be asking why. For example, I might consider whether there is something about watching television that helps people develop ways for coping with negative emotions or life stressors. Although the study found the opposite, I don’t think this possibility is “patently absurd.”

And, again, of course no study can control for all possible variables. This study was based on data from the National Longitudinal Study of Adolescent Health; the authors of this study did not design the questions themselves (and undoubtedly would not have had the resources to collect data from such a large representative sample). Scientific research is cumulative. No study is perfect, but each study should build on and add to previous work.

Given the findings, I think we need to be asking: Why was TV and overall media use in high school related to depression 7 years later? Does TV viewing directly impact mood states in a negative way? If so, through what mechanism? Does it matter what types of television programs young people are watching (the data set did not include this information)? Or, do people who spend more time watching television communicate less with family and friends, which reduces the availability of interpersonal support? Or could it be that young people who were watching LESS television were engaged in other activities that acted as a buffer for stressful life events (e.g., more involved with extra-curricular activities or volunteer work?). These are questions that could grow out of this study, and which might yield valuable answers.