Before thinking about the thorny question of how shamans heal, it is worth posing a logically prior question: do they heal? There are remarkably few data on this question. In particular, even moderately long-term follow-up is lacking. As anthropologist and medical doctor Gilbert Lewis puts it, “It is rare to find examples of anthropologists who record the frequency of therapeutic failures, do follow ups, or find out how many people do not bother to come back next time to the shaman.” Robert Desjarlais, a psychological anthropologist, points out that most research on ritual healing attempts to explain how it works, without demonstrating whether — and in what ways — patients actually feel better.

So: how well do shamans actually cure sickness? The answer is that no one knows. To a great extent the body heals itself without intervention; most diseases are self-limiting. Another answer is a question: cured compared to what? It is difficult to devise a metric: return to work? return to premorbid functioning? return for follow-up? consumer satisfaction? Indeed, we cannot even assume that people or cultures have unitary or unequivocal resolutions of suffering, or that we can recognize a culturally relevant resolution of suffering when it occurs. We do not know how long a follow-up is useful, even if we knew what we were following.

Still, a study might go something like this.

Select a health problem with outcomes that can be clearly ranked, and of the sort might be brought to a biomedical facility for treatment — diabetes, for example, or a chronic inflammatory disease such as lupus, rheumatoid arthritis, ulcerative colitis, or Crohn’s disease. Then, at a regional hospital, such as that in Iquitos, ask all patients with that diagnosis if they would accept treatment from a curandero in addition to their biomedical treatment. The curandero treatment would consist of the foundational triad of mestizo shamanism — shacapar, rattling; chupar, sucking; and soplar, blowing tobacco smoke. Among those who would be willing to accept the additional treatment, randomly assign some to a treatment and some to a non-treatment group.

Thus the study would have three groups of participants — those who are uninterested in shamanic treatment, those who are interested and get it, and those who are interested and do not. Then compare two sorts of outcomes — objective findings and laboratory test results, on the one hand, and quality of life measures on the other, such as the World Health Organization cross-cultural quality of life test instruments.

Now there are lots of uncontrolled variables in a study like this — patients on different medication regimens, patients with different levels of compliance, issues of social support, and effects of other concurrent problems such as alcoholism. Still, the presence or absence of statistically significant differences in any outcome measure between the shaman-treated and the two non-shaman-treated groups would certainly raise questions worth thinking about.

A study protocol might even use two different curanderos, in order to see if one is effective and one is not. And then try to figure out why.

Any reason why a study like this shouldn’t be done?