Does Caffeine Enhance the Taste of Cola?
Soft drink manufacturers claim that caffeine enhances the flavor of cola drinks and that this is why they make caffeinated colas. Researchers from Johns Hopkins University disputed this claim in a study reported in the Archives of Family Medicine (Griffiths and Vernotica, 2000). The researchers conducted a taste test to find out if people can tell by tasting whether a cola contains caffeine or not. Each participant tried 20 times to identify which of two cola drinks was "sample A" and which was "sample B." Unknown to the participants, one sample (A or B) contained caffeine and one did not.

The particular sample (A or B) with caffeine was the same for all 20 trials. On each trial, a participant tasted drinks from two unlabeled cups and guessed which was sample A.

After every trial, the participant was told whether he or she was correct. To learn about the tastes of the two samples, participants did five practice trials before the 20 trials that counted in the results. Suppose that a participant cannot ever detect the difference between the cola samples and randomly guesses on every trial. In that case, X = number of correct guesses is a binomial random variable with n = 20 trials and p = .5 (the probability of correctly guessing just by luck).

The investigators called a participant's flavor detection performance "significant" when the individual made 15 or more correct identifications in 20 trials. The probability is only .0207 (about 1 in 50) that somebody who guesses every time could do this well. The calculation of this probability is

P(X ≥ 15) = 1 - P(X ≤ 14) = 1- .9793= .0207.

The binomial probability of P(X ≤ 14) could be found using Minitab, Excel, or a calculator such as a TI-84. Twenty-five individuals participated, and only two made 15 or more correct identifications. This isn't much better than what would be expected if everybody randomly guessed for all trials. As we already determined, the probability of 15 or more correct when randomly guessing is .0207, which is about 1 in 50. If all 25 participants guessed, we would expect 25(.0207) = .5175, or about 0 or 1 participants to fall into the "significant" category.

The researchers interpreted their results to mean that people generally can't taste whether a cola contains caffeine. Soft drink manufacturers objected to several aspects of the study, including the sample sizes for trials and participants and the way in which the researchers prepared the caffeinated cola samples.