There was discussion a short time ago about the non-attributive use of certain adjectives that beqin with the prefix a-.  I couldn’t think at the time what article I had recently read about this phenomenon, but came across it this morning: Julia Schlüter, “Constraints on the attributive use of ‘predicative-only’ adjectives” in Graeme Trousdale, Nikolas Gisborne, ed., Topics in English Linguistics: Constructional Approaches to English Grammar (Mouton de Gruyter, Berlin, 2008) pp. 145-179.  Some of her conclusions may be of some interest. 

 

1)      aghast, agog, aloof, askew appear occasionally in attr. position

2)      adrift, alive ashamed, averse, awake, aware, awry appear more often in attr.position when premodified

3)      afloat, afraid, akin, asleep appear in attr. position only when premodified

 

Premodification includes use with a prefix (like un-), compounding, and modification by an adverb, which complexity in the attr. construction has been available only since the nineteenth century.  The linguistic explanations available are based on 1) semantics, and 2) phonetic (stress clashes).  Both forces were shown to be at work in this construction, sometimes one, sometimes the other.   This seems to support a lexically based theory of explanation.