MWAK asked how to distinguish between major depression and bereavement. The good news is that you don't have to. If the patient meets criteria for major depression, they have major depression and you should refer the patient. The exclusionary bereavement criterion is misleading. The Archives article shows that the syndrome is independent of cause: major depression that develops in the context of bereavement or in the context of medical hospitalization or the context of incarceration or in the context of job loss is clinically indistinguishable.
The good thing about using a clinical syndrome definition is that it separates diagnosis from etiology. This frees the clinician from zeroing in on one etiology at the cost of another. This is important in psychiatry because an attribution error can be physically risky. The bereaved clinically depressed 75 year old woman may also be suffering from hypothyroidism. The bereaved 35 year old gay man who lost his companion to AIDS may also have an HIV-associated affective syndrome. A diagnosis of a clinical syndrome is only the first step. The importance of psychiatry (as opposed to non-medically trained mental health professionals) is that we know the next step is a good diagnostic workup.
The need to attribute causation really only seems to come into play with when we're talking about major depression even though there are other psychiatric syndromes that can be precipitated by life stress: the first episode of psychosis in schizophrenia for example, or mania. Yet we don't attribute mania to life circumstances. If someone is impulsively spending thousands of dollars, talking a mile-a-minute and staying up all night cleaning the house we don't say "Of course you're happy, you have a great husband and a great job and great kids (or a great dog)." We say, "Please please go back on your lithium." (My favorite causation story is of a psychologist who once asked me, "Is this inmate demented because of the stress of getting an 80 year sentence?" I explained that it was from the stress of having an 80 year old brain.)
The other thing the Times story didn't mention is that all the groups in the study had recurring episodes. Both the loss- and bereavement-triggered groups had a mean of two to four episodes. The trigger was identified only for the index (first) episode. You could make an argument that someone might get misdiagnosed once---but four times? And at least 5% of subjects were hospitalized. I think Roy might have something to say about this. How many times do you see people on your inpatient unit for "intense normal sadness" (the authors' classification) as opposed to clinical depression?
The authors were concerned about the stigma of a posssibly inaccurate diagnosis. Given the information they presented about their own subjects, I'd rather err on the side of caution.