Let's make a dumbbell theory of some people's personalities.
Uniframers disregard discrepancies in favor of imagined regularities. They tend to be perfectionists but also tend to think in terms of stereotypes. This sometimes leads to recklessness because they have to reject some evidence in order to make their uniframes. Accumulators are less extreme. They keep collecting evidence and hence are much less prone to make mistakes. But then they're also slower to make discoveries.
Of course these imaginary personalities are only caricatures, and everyone blends both extremes. Most people find some reasonable compromise, though a few of us lean more in one direction than the other. I'm sure we all use mixtures of different learning strategies
— accumulations of descriptions, K-lines, uniframes, or whatever. On the surface, it might seem easier to make accumulations than to make uniframes — but choosing what to accumulate may require deeper insight. In any case, whenever an accumulation becomes too large and clumsy, we try to replace some groups of its members with a uniframe. But even when we succeed in finding a suitably compact uniframe, we can expect it, too, to accumulate exceptions eventually, since first descriptions rarely work for all our later purposes.
For example, when a child first encounters dogs, an attempt might be made to create a uniframe that catalogs those animals' parts — eyes, ears, teeth, head, body, tail, legs, and so on. But the child will eventually have to learn that even here there are exceptions.
Furthermore, that uniframe won't help answer the child's most urgent questions about any one dog in particular: Is it friendly? Does it have a loud bark? Is it the kind that tends to bite? Each such concern could require building a different kind of hierarchy-tree.
This leads to an inescapable difficulty. Our various motives and concerns are likely to require incompatible ways to classify things. You can't predict a dog's bite from its bark. Each of the classifications we build must embody different kinds of knowledge, and we can rarely use more than a few of them at once. When we have a goal that is simple and clear, we may be able to select one particular kind of description that makes the problem easy to solve. But when goals of several types conflict, it is harder to know just what to do.