One of the most frequent criticisms of the recall approach is the fallibility of human memory. Where total recall of all
food items consumed is sought, respondents tend to omit some items, even in short-term recall situations. However,
a review of food intake research literature revealed that subjects' recall accuracy was best for foods that were major
components of a meal and eaten in relatively large quantities. Flesh foods, meats, meat substitutes and casseroles had
the greatest recall accuracy in several studies (Mullen, et al., 1984; Eck, et al., 1989; Dubois and Boivin, 1990). In other
studies where fish intake was sought, recall for fish was among the most accurate of any foods (Guthrie, 1984; Yuhas,
et al., 1989; Jensen, et al., 1984).

The accuracy of recall can also be enhanced by the use of specific and meaningful memory cues (Krall and Dwyer,
1987; Dubois and Boivin, 1990). This aided recall approach was used in the questionnaires. Six commonly eaten types
of finfish and five major types of shellfish were used as memory cues to screen for finfish and shellfish consumers.
Respondents that had eaten finfish were then read a list of 27 species to determine which ones they had eaten during
the previous seven days. Subjects that had eaten some type of shellfish were read a list of 13 shellfish species to
determine consumption. The lists of finfish and shellfish included those most commonly consumed in Florida (Teehan,
1991).

Another potential source of error with food intake recall stems from respondents' estimates of portion sizes. Many
studies, particularly those using the food frequency approach, do not solicit portion size estimates from respondents.
Instead, a "standard" portion, as defined by professional dietitians or others, is assumed. Some researchers defend the
use of a "standard" portion by arguing that intraindividual variability is greater than interindividual variability (Hunter,
et al., 1988). Others counter that standard portions reflect "measurement convenience and approximation rather than
any behavioral truth about the portions people actually consume" (Block, et al., 1986). One alternative to using one
standard portion size is to use a range of sizes such as "small," "medium" and "large"; this method has been shown to
improve portion size estimates by introducing personal variability in respondents' intake estimates (Block, et al., 1986).
This technique was used in a modified manner in our study for estimating consumption of some shellfish.

In order to obtain an accurate estimate of portion size, a very common unit of measure, which the respondents could
easily visualize, needed to be found. Because visual aids, such as photographs for food replicas, could not be used in
a telephone survey the respondents were asked to "imagine one slice of sandwich bread" for the finfish items. They
were then asked to provide an estimate of portion size in fractions or multiples of the bread slice. The rationale for using
a slice of bread was that it was a familiar, common item that a large proportion of the population would be able to
visualize, particularly in the context of eating. Respondents' portion estimates in terms of fractions or multiples of the
bread slice were converted to a finfish weight consumption figure using a volume estimate for a bread slice and the
average specific gravity for finfish flesh.

Portion sizes for shellfish items were based upon counts and known industry size standards whenever possible. For
example, respondents that had eaten shrimp were simply asked for the number eaten and the approximate size. Size
options were "jumbo", "large", "medium" and "small", all commonly used size classifications for which edible yield
data are available.


 Procedure

The surveyed households were selected via a random digit dialing procedure to assure inclusion of unlisted telephone
numbers. Within each household, the "next birthday" method of respondent selection was used to randomly select an
adult (age 18 or older) for interviewing (Salmon and Nichols, 1983).

The randomly selected adult in the telephone surveys was asked for detailed away-from-home seafood consumption
data for the previous seven days. After this information had been obtained, the interviewer asked to speak to the primary
meal preparer. In many cases, the randomly selected adult was also the primary meal preparer. If the primary meal
preparer was available, he or she was asked to provide at-home seafood consumption data for every member of the
household, including the randomly selected adult. The rationale for asking the primary meal preparer for data on every
member of the household was simple: The meal preparer was likely to be the one that was most aware of individual