For this, as for the other problems, extreme differences in method and
in time and degree of success appear for the different settings. Thus,
while settings 1, 2, 3, 6, 7, and 10 proved to be easy, settings 4, 5,
8, and 9 were evidently more difficult.

[Illustration: FIGURE 20.--Error curve of learning for the solution of
problem 3 (alternately the first box at the left end and the first at
the right end) by Sobke.]

From the first this problem promised to be much easier for Sobke than
problem 2, and although the actual number of trials necessary for the
solution is greater by sixty for problem 3 than for problem 2,
comparison of the data of the tables justifies the statement that the
third problem was both easier and more nearly adequately solved than the
second. This is not surprising when the nature of the two problems is
considered, for whereas problem 2 requires choice by perception of the
relationship of secondness from the right end of the group, problem 3
requires, instead, the choice of the end member of the group each time,
with the additional variation of alternation of ends. Now as it happens,
the end member is easily selected by the monkey, and it appears further
that alternation was relatively easy for Sobke to acquire. Consequently,
the combination of end and alternation proved easier than the choice of
the second from the right end of the group.

The above statements are supported by comparison of the curves of
learning. The curve for problem 2, figure 19, is extremely irregular;
that for problem 3, figure 20, much more regular. Similarly, the daily
ratios of right to wrong choices as exhibited in tables 5 and 6 indicate
smaller variations for the third problem than for the second.