888 
the result of selecting the student body, and 
would be vertical at the right side if it were 
possible to perfect the selection so as to ex- 
clude all below a certain standard and admit 
all above that standard. 
Reed College uses ten grades. Ten, rather 
than any other number, are used, because tests 
by approved statistical and psychological 
methods show that fewer grades are inade- 
quate to designate readily discernible differ- 
ences, and many more grades can not be used 
with intelligent discrimination. The defi- 
nitions of these grades have a scientific basis. 
Grades 1-5 indicate that a student stands in 
the upper half of an average class; grades 
6-10 indicate that he is in the lower half. 
For example, 2 designates the work which will 
be done (in the long run) by the best 5 per 
cent. of all students, and 6 the work done by 
that quarter of an average class standing just 
below the middle. These particular percent- 
ages were chosen somewhat arbitrarily for the 
sake of convenient round numbers, but they 
correspond fairly well to the distribution 
shown in the probability curve as skewed from 
normal by the selection of students. 
Grade 1 is rarely given, representing a de- 
gree of excellence attainable by not more than 
one student in four or five hundred; similarly, 
grade 10 records correspondingly bad failures. 
The lowest grade called passable is 8, which 
covers all cases where credit is granted con- 
ditionally or upon the satisfaction of some 
special requirement; 9 is for ordinary cases of 
failure. 
It will be observed that the symbol 1 of the 
Reed College scale is a grade of real distinc- 
tion, as the A of the usual scale is not; and 
that even the 3 of the Reed College scale rep- 
resents quite as much distinction as the A if 
the latter grade were obtained by 15 per cent. 
of a class. 
The grades, however, can not be interpreted 
in qualitative terms, as good, poor, A, C, 90 
per cent.; nor do they designate rank in the 
particular class. They show the group in 
which the student would appear if the classes 
of several years were subdivided as indicated 
above. 
SCIENCE 
[N.S. Vou. XXXV. No. 910 
Such a definition will not make a “2” = 
“9” as regards the actual quality of work 
done by the student, since this is a matter not 
only of the student’s relative standing in the 
subject, but also of the actual standards set 
for that subject. But despite this, for courses 
of the same general sort (as introductory 
courses without prerequisites, whether coming 
in freshman or senior year), the relative 
rank in an average class seems to be the best 
available criterion of the student’s merits. 
In adopting the Reed College System, we 
attempted to divide the base line of the nor- 
mal probability curve into equal parts; but we 
found this awkward, as grade 2 would then 
fall to but 14 per cent. of the students. We 
then tried (each side of the median) a distri- 
bution of 20 per cent., 15 per cent., 10 per 
cent., 5 per cent., which corresponds to equal 
bases for grades 4, 5, 6, 7 with somewhat 
longer bases for 3 and 8, and much longer 
ones for 2 and 9. This distribution seemed 
more convenient. We changed the distribu- 
tion on the lower side, however, to 25 per 
cent., 15 per cent., 6 per cent., 4 per cent., to 
allow for skewing of the curve due to the influ- 
ence of selection. What sort of divisions of 
the base line this requires no one can say 
without knowing the exact shape of the 
skewed curve (which depends on how effec- 
tive the selective cut is), and knowing also 
how far the new median has been moved above 
the median in the normal curve (which de- 
pends on what area was cut off, according to 
the standard of selection). In using unequal 
divisions of the base line, therefore, we make 
a choice of percentages which is somewhat 
arbitrary, except that it follows roughly the 
sort of distribution in the skewed curve with 
equal base divisions. 
With such a basis for grading students in 
their college courses, it is possible to give a 
definite and just reward for high scholarship 
by allowing a course completed with high 
credit to count more toward a degree than a 
course completed with lower credit. Such a 
plan for counting quality has the great ad- 
vantage of enabling the students who do the 
best work to graduate in less than the normal 
