The article and summary are misleading in focusing on overall counts of infection rather than rates of infection, which take into account both the number of cases and the relative rates of vaccination.
The better statistics to summarize from the article are the 'attack rates' (number of those who get the disease per 100,000 person-years). The article breaks these down by age and by vaccination status (Table 2, p. 23).
For ages 2-7, attack rates were 359 for vaccinated and 606 for under and non-vaccinated. This difference in attack rates, though fairly large, was not large enough to reach statistical significance.
For ages 8-12, attack rates were 2453 and 3211 respectively, again a non-significant difference.
For ages 13-18, attack rates were 452 and 2189 -- this difference does yield a statistically significant advantage for those vaccinated
Looking over all ages, attack rates were 1011 and 2073, respectively, again a statistically significant advantage for those vaccinated.
The two big points are a) the large overall advantage of being vaccinated compared to being under- or non-vaccinated, and b) the age-related increase in the attack rate for those vaccinated that occurs during gaps between boosters (between 8-12).
Overall, both points are strong evidence for the efficacy of the vaccine over non-vaccination.
The author's articles claim in the introduction that "our unvaccinated and under-vaccinated population did not appear to contribute significantly to the increased rate of clinical pertussis" (p. 5). This seems contradicted by their data, but it is not revised or re-considered in the discussion. They may perhaps mean that given the low(ish) rate of under/non vaccination that this group had a relatively small contribution to the overall number of cases. It is clear from the attack rate analysis, however, those in this group were 4.8X more likely to develop whooping cough than those fully vaccinated (overall attack rate of 2189/100,000ppy for under/non divided by 452 for those vaccinated).
The summary and headline should probably be revised. By tallying raw counts it's like saying "Honda's are most involved in accidents" which may be true in terms of raw counts due to their popularity but not true in terms of accident rates.
Conversation on this story is long over, but it's never too late to post some actual facts about the study.
I. Rates of obedience.
Milgram did several replications of his original experiment. His obedience rates hovered around 65%. In the one experiment he did comparing men to women, similar obedience rates were found (Experiment 8, Milgram, 1974). In a survey of 14 replications by other researchers, Blass (2000) reported that obedience rates varied from a low of 28% to a high of 91%. It is interesting that posters familiar with these studies tended to overestimate obedience rates (claiming in the high 90s). Research has shown that invididuals *unfamiliar* with the studies tend to *underestimate* obedience rates, often by an order of magnitude. For example, Milgram polled a group of 40 Yale psychologists unfamiliar with his studies; the predicted rate of obedience was 0.125% (Milgram, 1963).
II. History of the research.
Both the summary and the source article claim that research using Milgram's paradigm has ceased for over 40 years. This is incorrect. The review of research by Blass (2000) reports on replications as late as 1985 (Schurz, 1985). In a brief search I wasn't able to find more recent replications, but they could well be out there. Blass found no significant changes in obedience rates over time in the years from Milgram's original research through 1985. It's strange that this paper made it through review with such a glaring factual error. Moreover, their command of the literature on obedience seems quite superficial and beneath what I was able to glene simply from Blass's review (2000) (which, obviously, I'm making heavy use of here).
III. Ethicality of the research.
Both the summary and the source blandly claim that this research is no longer possible on ethical grounds. What these ethical grounds are supposed to be, however, is not clear. I think the strangest part of the lore surrounding this experiment is our reflexive acceptance that this research is wrong. With a bit of thought, I think it becomes clear that this research is actually quite acceptable.
1) Harm? The primary objection to this research is that obedient subjects may be harmed in some way by learning that they are willing to murder on command. There is no empirical evidence for this claim. Milgram conducted follow-up interviews with many subjects--most were grateful to have participated in the study. He wrote that the tension produced was below that produced by watching a good horror movie. As far as I know, none of Milgram's many subjects (several hundred, I believe) cam forward then or now with claims of adverse effects. Moreover, it seems likely that participating in this experiment would produce more benefit than harm. After all, this type of insight into your behavior could well help you avoid destructive obedience in the future.
2) Potential gains. In determining the ethicality of a research proposal, the potential harm to subjects is always weighted, to some extent, against the potential gains from the research. For example, we often let terminally ill patients try experimental and risky drugs/procedures because the potential gains outweigh the risks. Of course, potential gains can only bear a moderate weight (shouldn't create genius killer sharks to cure Alzheimers!). But we routinely allow volunteers to risk taking experimental drugs in clinical trails, even for non-threatening medical conditions (viagra. Thus, it seems we are willing to tolerate small levels of risk even for relatively small gains. Now consider that Milgram's research was specifically designed to understand the roots of genocidal behavior. His experiments were conducted after WWII; he was interested to know if mere obedience was sufficient to induce murder or if some mental or social disorder was also required (as was commonly hypothesized about the perpetrators of the Holocaust). This is not just an academic point. Over 100 million+ died in genocide in the 20th century (see the work of RJ Rummel,
Slashdot Mirror
The article and summary are misleading in focusing on overall counts of infection rather than rates of infection, which take into account both the number of cases and the relative rates of vaccination. The better statistics to summarize from the article are the 'attack rates' (number of those who get the disease per 100,000 person-years). The article breaks these down by age and by vaccination status (Table 2, p. 23). For ages 2-7, attack rates were 359 for vaccinated and 606 for under and non-vaccinated. This difference in attack rates, though fairly large, was not large enough to reach statistical significance. For ages 8-12, attack rates were 2453 and 3211 respectively, again a non-significant difference. For ages 13-18, attack rates were 452 and 2189 -- this difference does yield a statistically significant advantage for those vaccinated Looking over all ages, attack rates were 1011 and 2073, respectively, again a statistically significant advantage for those vaccinated. The two big points are a) the large overall advantage of being vaccinated compared to being under- or non-vaccinated, and b) the age-related increase in the attack rate for those vaccinated that occurs during gaps between boosters (between 8-12). Overall, both points are strong evidence for the efficacy of the vaccine over non-vaccination. The author's articles claim in the introduction that "our unvaccinated and under-vaccinated population did not appear to contribute significantly to the increased rate of clinical pertussis" (p. 5). This seems contradicted by their data, but it is not revised or re-considered in the discussion. They may perhaps mean that given the low(ish) rate of under/non vaccination that this group had a relatively small contribution to the overall number of cases. It is clear from the attack rate analysis, however, those in this group were 4.8X more likely to develop whooping cough than those fully vaccinated (overall attack rate of 2189/100,000ppy for under/non divided by 452 for those vaccinated). The summary and headline should probably be revised. By tallying raw counts it's like saying "Honda's are most involved in accidents" which may be true in terms of raw counts due to their popularity but not true in terms of accident rates.
Conversation on this story is long over, but it's never too late to post some actual facts about the study. I. Rates of obedience. Milgram did several replications of his original experiment. His obedience rates hovered around 65%. In the one experiment he did comparing men to women, similar obedience rates were found (Experiment 8, Milgram, 1974). In a survey of 14 replications by other researchers, Blass (2000) reported that obedience rates varied from a low of 28% to a high of 91%. It is interesting that posters familiar with these studies tended to overestimate obedience rates (claiming in the high 90s). Research has shown that invididuals *unfamiliar* with the studies tend to *underestimate* obedience rates, often by an order of magnitude. For example, Milgram polled a group of 40 Yale psychologists unfamiliar with his studies; the predicted rate of obedience was 0.125% (Milgram, 1963). II. History of the research. Both the summary and the source article claim that research using Milgram's paradigm has ceased for over 40 years. This is incorrect. The review of research by Blass (2000) reports on replications as late as 1985 (Schurz, 1985). In a brief search I wasn't able to find more recent replications, but they could well be out there. Blass found no significant changes in obedience rates over time in the years from Milgram's original research through 1985. It's strange that this paper made it through review with such a glaring factual error. Moreover, their command of the literature on obedience seems quite superficial and beneath what I was able to glene simply from Blass's review (2000) (which, obviously, I'm making heavy use of here). III. Ethicality of the research. Both the summary and the source blandly claim that this research is no longer possible on ethical grounds. What these ethical grounds are supposed to be, however, is not clear. I think the strangest part of the lore surrounding this experiment is our reflexive acceptance that this research is wrong. With a bit of thought, I think it becomes clear that this research is actually quite acceptable. 1) Harm? The primary objection to this research is that obedient subjects may be harmed in some way by learning that they are willing to murder on command. There is no empirical evidence for this claim. Milgram conducted follow-up interviews with many subjects--most were grateful to have participated in the study. He wrote that the tension produced was below that produced by watching a good horror movie. As far as I know, none of Milgram's many subjects (several hundred, I believe) cam forward then or now with claims of adverse effects. Moreover, it seems likely that participating in this experiment would produce more benefit than harm. After all, this type of insight into your behavior could well help you avoid destructive obedience in the future. 2) Potential gains. In determining the ethicality of a research proposal, the potential harm to subjects is always weighted, to some extent, against the potential gains from the research. For example, we often let terminally ill patients try experimental and risky drugs/procedures because the potential gains outweigh the risks. Of course, potential gains can only bear a moderate weight (shouldn't create genius killer sharks to cure Alzheimers!). But we routinely allow volunteers to risk taking experimental drugs in clinical trails, even for non-threatening medical conditions (viagra. Thus, it seems we are willing to tolerate small levels of risk even for relatively small gains. Now consider that Milgram's research was specifically designed to understand the roots of genocidal behavior. His experiments were conducted after WWII; he was interested to know if mere obedience was sufficient to induce murder or if some mental or social disorder was also required (as was commonly hypothesized about the perpetrators of the Holocaust). This is not just an academic point. Over 100 million+ died in genocide in the 20th century (see the work of RJ Rummel,