In the subsection 4.3.4.2 The ‘odd number paradox’ of Cognitive Linguistics by W. Croft & D. A. Cruse
We read:
The ‘odd number paradox’ has also been put forward as a problem for prototype theory. Armstrong et al. (1983) found that people will grade ODD NUMBERS for centrality, even though the category ODD NUMBER has a clear definition in terms of necessary and sufficient features. Their proposed solution, the so-called ‘dual representation’ hypothesis, combines the prototype approach and the classical approach (Smith et al. 1974). The idea is that concepts have two representations, which have different functions. There is a ‘core’ representation, which has basically the form of a classical definition. This representation will govern the logical properties of the concept. The other representation is some sort of prototype system which prioritizes the most typical features, and whose function is to allow rapid categorization of instances encountered. With this set-up, the odd-number effect ceases to be a puzzle. However, this conjunction of two theories inherits most of the problems of both of them: in particular, it reinstates a major problem of the classical theory that prototype theory was intended to solve, namely, the fact that for a great many everyday concepts there is no available core definition.
It is clear for me that they speak about conjecture of the two representations of a concept. But I is unclear for me what do they mean in this sentence 'people will grade ODD NUMBERS for centrality, even though the category ODD NUMBER has a clear definition in terms of necessary and sufficient features.'