ABSTRACT
There is no necessary connection between the constraints demanded for easy learnability and those needed for easy parsability. After all, learnability is a property of a family of grammars, while efficient parsability is a property of a single grammar. For example, if there were only one grammar, or language, it could be easily selected, yet it could be of arbitrary recognition complexity. Likewise, it is well known that there are families of languages that are not identifiable from positive evidence, even though each member of the family is easily parsable (e.g., a collection of all the finite languages plus an infinite, finite-state language that covers them all). This leaves us with the question of how learnability and parsability constraints are connected. Do these dual functional demands conflict or reinforce one another?
