VALIDITY OF THE SURVEY METHOD OF RESEARCH. H 
possible to get fairly reliable results by securing large numbers of 
estimates and using only averages of them. This principle is taken 
advantage of in the study of farm practice, and there is reason to 
believe that, within the proper limits of use of the results obtained, 
studies of this kind are entitled to at least as much consideration 
from the standpoint of accuracy as are those involving experi- 
mental work conducted under the most favorable field conditions. 
Indeed it is believed that when carefully conducted by those 
properly trained both in the collection of data and in the in- 
terpretation of these data, the results of such studies approach in 
accuracy those obtained in laboratory investigations. 
The so-called law of averages is merely one manifestation of the 
laws of probability, or chance. It is not feasible here to discuss these 
laws in detail. They are fully treated in standard texts, with which 
every experimentalist should be familiar. In fact, the interpretation 
of experimental results which does not take into account the law of 
error is nearly as apt to be wrong as it is to be right. A little con- 
sideration will show that in a highly variable quantity, such as the 
yield of a given plot treated in a given way, six duplicate plots is 
far too small a number to insure with any degree of certainty 
that the action of the law of averages will eliminate the departures 
from the true average. In general, the average of six such yields, 
no matter how accurately each yield is measured, is far less reliable 
than would be the average of 60 estimates of farmers based on years 
of experience with a given field. Sixty such estimates give a chance 
for the law of averages to eliminate a large proportion of the errors 
in the individual estimates, and these errors are in general no larger 
than those in plot yields, no matter how accurately these yields are 
measured. 
While we may not here consider the laws of chance in detail, a 
few illustrations of them may serve to show that such laws actually 
exist. 
In flipping a penny it is an even chance whether heads or tails 
turn up at any particular throw. Now, it has been proven by abun- 
dant experiment that as the number of times the penny is thrown in- 
creases, the tendency for the total number of heads to equal the total 
number of tails increases. In other words, where the chance is even 
the event will, on the average, turn out in one of two possible ways as 
often as it does in the other. 
In throwing a single die there are six possible results, all equally 
likely to occur. There is thus a tendency, when a die is thrown many 
times, for any one of the six faces to turn up one time in six on the 
average. 
An excellent illustration of the workings of the laws of chance was 
recently found in tabulating the replies to a circular letter sent out 
