Smoking and Tobacco Control Monograph No. 7 
Doesn't this mean that a numerical scale — say, from 1 to 27 — is 
necessarily misleading and that a categorical rating system would be a more 
valid way to report such information? You had four categories, 1 believe, and 
you showed variance. What I am saying is, I was following that, and it 
looked to me like you were about to recommend a four-category system. 
DR. TOWNSEND: The whole idea of a categorical system, or so-called banded 
system, has been put on the table by a number of people in the past. 
Conceivably, that accomplishes the same endpoint with one exception, 
1 believe. The same endpoint, of course, is that it provides a comparative 
ranking for consumers. The flaw in that approach, if that is the only 
ranking, and discrete numbers are not also included, is that, of course, 
you would expect products to come up against the ceiling of each band. 
DR. COHEN: I thought you established, with your own analysis, that 
individual numbers were inherently misleading? 
DR. TOWNSEND: I did not say that individual numbers were misleading. 
In fact, 1 believe individual numbers — a numerical rating system — is, in fact, 
the best and that is what the FTC test method is. 
With those data, then, manufacturers have advertised their products as 
light or ultralight, to fit some range of tar numbers. 
DR. COHEN: I think you missed my point. 
DR. TOWNSEND: The FTC method is the method that provides the useful 
comparative information for the smoker. 
DR. COHEN: 1 thought you established with your charts that there was 
variance due to what is technically an interaction between product design 
characteristics and smokers' adaptations to them. Is that correct? 
DR. TOWNSEND: Okay, you are confusing me. Let me put the chart back 
up. Standard deviation for replicate measures of that particular product. 
Different particular puff frequencies. Puff frequency of 60 seconds happens 
to be blue bars, 45 seconds is the red bars, 30 seconds is the yellow bars. 
So, the ultralight product that was smoked at 30-second frequency — in fact, 
this is the variability we saw in 10 replicate measures of that one cigarette. 
DR. COHEN: Okay, then 1 did misunderstand that, but a lot of the 
presentations today have essentially suggested that smokers respond to 
product design characteristics by modifying behavior. And 1 am not talking 
about idiosyncratic behavior, but standard ways that you put up — puff 
frequencies, puff duration, number of puffs. 
Ihat creates variance around a point estimate. And would it not be 
more valid to acknowledge that those variances exist when you provide this 
information to consumers? Ranking is the least informative scale. 
DR. rOWNSEND: You are talking about something like the EPA gas mileage 
ranking, where you have highway and city. 
186 
