Canonical features and their interaction
with Part of Speech categories
Greville G. Corbett
Often features are presented as clean, neat, simple. This is understandable, since it is the
contrast with the disorderliness of lexical entries which gives the intuitive justification for
features. But real life and real languages are more complex. There are many instances
where we need to ask whether we genuinely have a feature, or feature value. We must
recognize that feature systems vary according to how well founded they are and in how
they are distributed across the lexicon, sometimes in principled ways, sometimes almost
randomly. In order to talk about this difficult area, the penumbra of feature systems, a
canonical approach proves helpful. Having justified this approach, we then consider a set
of converging criteria for canonical features and values. This gives us a point in the
theoretical space from which to calibrate the difficult instances which abound in feature
systems. We ask whether the problems we find are feature-specific or whether they recur
in the different morphosyntactic features. This leads to the question of whether they can
co-exist. Finally, we address the issue of differentiating canonical morphosyntactic
features; since they are idealizations they might appear indistinguishable. We show how
at an abstract level, minimal differences allow us to differentiate them. It is here that the
interaction with parts of speech proves particularly significant.