An important question in language evolution is whether segmentation as a linguistic process is able to yield compositionality. Segmentation is hypothesized to be a process to bridge the gap between holistic and compositional lexicons. However, to date no thorough analytical method has been provided to test the feasibility of segmentation. In this paper, an analytical model is presented that can predict the probability of encountering various kinds of overlaps by observing utterance pairs, and the probability of finding confirmation in the (...) language for newly extracted segments. Language users start by using a previously evolved holistic lexicon to communicate about simple environments. They segment these holistic utterances to smaller pieces, which can be used as elements of a compositional lexicon. The model reveals that the feasibility of segmentation depends on the definition of counterexamples, i.e. those associations (pairs), which either cause ambiguous extraction of segments, or do not allow segmentation at all. On one hand if inexact overlaps are considered to be contradictory (i.e. causing confusion) to a perfect exact overlap, then the probability is so minuscule that it renders the role of segmentation marginal during language evolution. On the other hand, if such inexact counterexamples are able to be segmented unambiguously due to extra cognitive capacities, segmentation may have a much higher feasibility. Keywords: segmentation; fractionation; analysis; holistic; protolanguage; compositionality. (shrink)