Four carrion crows were trained to choose the greater array from pairs
of numerousness, discriminanda in the range of 1-12. In the process o
f training all irrelevant attributes of the arrays (geometric form, si
ze and colour of elements, patterns of distribution) were varied. Seve
ral control procedures were employed to make extraneous cues unlikely.
In particular, in order to preclude the use of cumulative area or oth
er magnitude cues the ratio of cumulative area of elements to their nu
mber was varied (in a half of presentations the greater array consiste
d of bigger elements while the lesser array consisted of smaller eleme
nts and in the other half the greater array consisted of smaller eleme
nts while the lesser array consisted of bigger elements). All the crow
s demonstrated high accuracy of comparisons (75,3 +/- 2,4%). Correct p
erformance dominated also under conditions with minimal difference bet
ween the compared arrays. It was concluded that these crows were able
to compare the arrays in the range of 1-12 by numerosity itself. When
arrays in the new range of 10-20 were presented all the crows demonstr
ated successful transfer of acquired reaction without any additional t
raining (71,5 +/- 2,3%). However, when the ratio of cumulative area of
elements to their number was varied only two of the four crows perfor
med successfully (71,4 +/- 3,6). Taken together these data suggest tha
t crows are capable of forming the concept <<more than>> based on nume
rical discrimination up to 20.