George (1966) with geese, Carleton (1966) 
with sciurids, and Tietjen et al. (1967) with 
pocket gophers. 
Techniques using enclosures (Crouch 1966), 
exclosures (Webb 1959), and photography 
(Greenwalt and Jones 1955) have also been of 
value in appraising diets. Den (Errington 
1937) and roost or nest debris (Errington 
1932), rodent feeding sites (Takos 1947), and 
cheek pouches of some rodents (Johnson 1961; 
Bradley 1968) have been used as additional 
sources of food habits data. 
METHODS OF HANDLING MATERIAL 
Methodology in the study of animal diets has 
been treated in some detail by McAtee (1912), 
Cottam (1936), Glading et al. (1943), Hartley 
(1948), Martin (1949), Martin and Korschgen 
(1963), and Dzieciolowski (1966). 
Collection 
Observation of the contents of stomachs, in- 
testines, crops, gizzards, gullets, and proventri- 
culi have all provided dietary information. 
Frequently several of these contents are ob- 
served to improve estimates. Sometimes the en- 
tire digestive tract of small mammals and 
many birds is collected and preserved. Crops of 
gallinaceous birds and doves are preferable to 
gizzards because such contents will not have 
been exposed to the more destructive digestive 
processes. In many studies, both crops and giz- 
zards have been analyzed. Gullets of waterfowl 
sometimes contain food particles in suitable 
condition for appraisal. 
It is often impractical to save the entire 
stomach contents of larger herbivorous ani- 
mals, and a sample, commonly about a quart, is 
removed. Among ruminants, samples are gen- 
erally taken from the rumen or rumen-reticu- 
lum. Contents are either mixed prior to sam- 
pling or samplings are taken from several 
parts of the food mass. 
Dietary composition in domestic livestock 
has been appraised by use of fistulated animals 
(Heady and Torell 1959; Van Dyne and Heady 
1965; Cook et al. 1967; and others). A similar 
approach in ruminant game animals, although 
admittedly much more difficult, may offer com- 
pensating rewards. The difficulties inherent in 
fistulating and maintaining an animal with 
physiological and psychological characteristics 
far different from those of more docile domes- 
tic animals can be appreciated. Successful 
rumen fistulation has been reported for deer by 
Short (1962) and Dziuk et al. (1963). Both 
rumen and esophageal fistulation techniques 
are currently being investigated for use on 
tamed or penned animals.’ 
*Personal correspondence with A. Lorin Ward, 
Apr. 30, 1968. 
134 
Flushing tubes (Vogtman 1945; MacGregor 
1958) induced regurgitation (Errington 1932), 
and stomach pumps? are also used to obtain 
food materials for analysis. 
Establishment of Sampling Limits 
The number of samples gathered for analy- 
sis should be a major consideration in animal 
food studies if diets are to be reliably esti- 
mated. When sample sizes are too limited, re- 
sults may be of questionable validity. When 
data are summarized by season or other units 
of time, sampling adequacy is further lessened. 
The contents of a single stomach or crop may 
appreciably distort findings. How sampling 
limited in time can alter conclusions was 
clearly demonstrated by Korschgen (1958), 
who found widely divergent food habits in 
mourning doves for a given month in different 
years. 
Because knowledge of dietary habits of some 
of the less abundant animal species is lacking, 
and because sample material for analysis is 
difficult to acquire, results based on only a few 
specimens have been reported. Errington 
(19382) did not regard stomach analyses as 
providing the best quantitative estimate of 
raptor food habits, at least partly because the 
number of stomachs examined that were repre- 
sentative of a given set of conditions was often 
too small to have much mathematical signifi- 
cance. The limitations of small samples should 
be recognized and interpreted accordingly. 
Apparently, only a few workers have been 
much concerned about the number of samples 
required for reliable estimates. Davison (1940) 
compared results obtained from examination of 
4, 9, 18, 95, and 471 bobwhite crops. Data ob- 
tained from 95 crops were essentially the same 
as those from 471 crops. Korschgen (1948) re- 
ported similar results and concluded that a 
minimum of 100 bobwhite crops should be ob- 
tained from a large area of similar plant 
growth to provide a representative sample. 
Anderson et al. (1965) used statistical methods 
recommended by Hanson and Graybill (1956) 
to estimate sampling adequacy in mule deer 
stomach content analyses. A sample size of 93 
provided an adequate estimate of major food 
items within 15 percent of the true mean at the 
95 percent confidence level. Van Dyne and 
Heady (1965) calculated the number of fistu- 
lated sheep and cattle required to sample die- 
tary constituents within a desired level of pre- 
cision by the procedure of Stein (1945). The 
number of animals required to estimate dietary 
composition within 10 percent of the mean 
with 90-percent confidence was variable, de- 
pending on animal class, grazing period, and 
dietary component. Generally, many animals 
