The Evolution
40
be interesting to see if the above experiment yielded a larger order
effect when restaged without such ambiguous intersections.
The Optimal Word Order for a Language
From all of the above it follows that if memory wishes to carry
Iout the Type II intersection of, say, Africa and striped, the most
effective order (i.e. the order that will most easily yield an inter-
sected concept such as "zebra") would be Africa striped, as it is
natural for memory to carry out Type II intersections with whole-to-
whole links activated first, and part-whole links second. In carrying
out this Type II intersection memory should find the word order Africa
striped easier, not because of any rule of the English language (actually
the phrase Africa striped is not grammatical and can hardly be regarded
as a standard way of denoting "zebra"), but because by placing the part-whole
links of the concept "striped" second, activation interference is avoided.
It is important to note that not all phrases are as simple for memory to
deal with as is Africa striped. With such a phrase ther is no doubt that
for a Type II intersection to take place the whole-to-whole links must be
provided by the concept "Africa" (or the word Africa), and the part-whole
links by the concept "striped." Whatever the word order this must be so,
as the concept "striped" can hardly be expected to provide the whole-to-
whole links.
But some Type II phrases (i.e. phrases that give rise to Type II
intersections) are Congnitively ambiguous in the sense that memory has
a choice not merely of word order, but also a choice as to which word will
provide the whole-to-whole links and which the part-whole links.