that is abstracted from the lexicogrammar, by the interaction of the lexicogrammar and
the interpreter.
So, given the implicit problems that occur when we employ semantic terminology, surely
we would be better served by the use of the morpheme as our main unit of quantification.
Such units would be much more adaptable to a statistical analysis, given that they are
quantitatively represented in the lexicogrammar and do not require any subjective inter-
pretation to delineate them in a text. Individual morphemes have a formal unity, in that
each allomorph is represented consistently in the lexicogrammar. They possess functional
unity, as each morpheme fulfills a definable function in the syntax of a language. Every
morpheme also has its own semantic unity, possessing a fundamental meaning which ex-
ists in all occurrences.^103
However, a system of quantification based on morphological delimitation raises problems
for our study due to the various writing systems we encounter. That is, not all of our texts
permit themselves to a consistent morphological breakdown while maintaining an equal
representation in the statistical analysis. For example, documents written in cuneiform
script can represent entire words with either a single logogram, or through a more elon-
gated string of syllabic signs. Now, suppose we were to encounter, as we do, a cuneiform
text that is represented by multiple parallel copies. Suppose also that some copies of this
text represent particular words with logograms, while the same words are represented in
(^103) See D.G. Lockwood, "The Problem of Inflected Morphemes," (^) Readings in Stratificational Linguistics
(eds A. Makkai and D.G. Lockwood; Alabama: The University of Alabama Press, 1973) 190, for this de-
scription of the qualities of morphemes.