"Walnuts add a happy crunch to life, and they may even help prolong it" 
NYTimes 28 Sept 1994

"Effects of walnuts on serum lipid levels and blood pressure in normal 
men" The New England Journal of Medicine, Vol.328 No.9, pp.603-607

1a.  The newspaper article contains recipes for walnuts.  The walnut 
seadon in California is Sept to Nov, and the study was included to 
presumable increase interest in the article. (article, col 2)

b. The study was conducted at Loma Linda University.  It was funded by 
the California Walnut Commission, and Dennis Balint is head of the 
commission.  Dr. Sabate is head of the research team at Loma Linda. 
(article, col 1)

2. The subjects of the experiment are 24 men who volunteered for the 
study.  The volunteers do not: eat nuts frequently, have known food 
allergy, smoke, take medication, have a history of hypertension, have 
extrememly high (>80th percentile) of low (<20th percentile) serum 
cholesterol levels. (p.603, col 2)
        The treatments are 2 diets:
        A cholesterol-lowering reference diet, and a diet that replaces 
some of the fatty foods in the reference diet with walnuts. All 
subjects were on each diet for 4 weeks.  Half of them received the 
reference diet for the first 4 weeks, and the walnut diet for the 
second 4 weeks.(p.604, col1)

3. controlled - the investigators assigned subjects to treatments
        single-blind - the subjects knew which diet they were on.  Those 
that measured serum cholesterol did not know which subject was on which diet.
        randomized - subjects were assigned to treatments at random
        cross-over - each subject received both treatments.  They 
"crossed-over" to the other treatment half way through the study.

4. The subjects had to eat breakfast and dinner in a common dining area, 
and they were provided lunch and snacks at breakfast time.  They were 
also paid a small monetary sum for participating in the experiment.

5.  The SD for the reference diet is 23 mg/dl.  The walnut diet has the 
same SD.
        If we were to create an SD for the difference in the two 
averages, ignoring the pairing it would be:

        Square Root [(23^2)/9 + (23^2)/9] = 11

So, a confidence interval for the difference 160-182 = -22 would cover 0.
However, with pairing we look at the differences in serum cholesterol for 
the 2 diets individually, and find the SD of these differences.  This 
reduces the overall variability and the confidence interval is -28 to 
-17, much smaller.

6.The carry over effect is dismissed on page 605, col 1, paragraph 2.  
It's the effect of the 1st diet on the serum cholesterol level during the 
period of the 2nd diet.

7. Figure 1 indicates that there is no carry over effect, because the 
mean serum levels for the walnut diet are close (both about 25 mg/dl 
lower than their respective levels for the reference diet).

8. ?