Many linguistics and teaching publications mention the A-adjectives, adjectives like asleep, afraid, afloat, alive, ablaze. The property that sets them apart and that is noted by all of these publications is that they do not like to appear prenominally, which is a position where adjectives generally do appear:
(1) ??an afraid child, ??an afloat ship, ??an ablaze building
(2) a scared child, a floating ship, a blazing building
This is as far as most publications go, although many point out that some of these are derived from prepositional phrases historically. Most assume that this distributional irregularity of the A-adjectives is synchronically idiosyncratic, and just has to be learned as such. This quote from Boyd and Goldberg, to appear, is typical: ``Since there is no general semantic or phonological reason to treat the members of the class as anything other than adjectives, their unusual distribution poses a clear learnability challenge.'' (Boyd and Goldberg to appear, page 10.)
I argue here that, in fact, there is strong distributional evidence that A-adjectives are synchronically PPs. Basically, I will argue that an A-adjective like ablaze has the same structure as the PP on fire. This explains why they do not appear prenominally: PPs do not appear prenominally, either:
(3) *an on fire building
But both can appear predicatively:
(4) That building is ablaze/on fire.
First, there is abundant evidence that the A-adjectives are not adjectives. They cannot do any of the things that typical adjectives can do. They cannot be used with the definite determiner the to form kind-denoting NPs, for instance:
(5) the scared, the living, the fat, the sleepy, the aloof, the lonely (will be expelled from the earth)
(6) *the alive, *the afraid, *the ablaze, *the afloat, *the asleep, *the alone (will be expelled from the earth)
A-adjectives cannot have comparatives with -er, even though most of them meet the phonological requirements for -er suffixation (maximally two syllables), and there is no ban on morphologically complex stems for -er:
(7) sleepier, scareder (I hear this used), floatier, lonelier
(8) *aliver, *afraider, *ablazer, *afloater, *aloner
They cannot be turned into adverbs with -ly:
(9) sleepily, burningly, fearfully, floatingly, aloofly, astutely
(10) *afraidly, *ablazely, *afloatly, *alonely
So, A-adjectives do not act like other adjectives. As I will now show, not only do they not pattern with adjectives, there are ways in which they pattern with PPs.
First, A-adjectives permit right modification, just like PPs and unlike adjectives:
(11) he fell right asleep, they went right ashore, it went clear askew/awry, it came right alive
(12) it caught right on fire, it went clear off the mark, they stepped right on shore, it came right to life
(13) *he became right sleepy, *he got clear lonely, *it came right lively
Second, it has often been noted that A-adjectives become better in prenominal position if they are modified:
(14) She flashed me an aware, amused glance. (attested example cited in Huddleston and Pullum 2002, page 559)
However, PPs also become better in prenominal position under the same conditions:
(15) She flashed me an at-ease, amused glance.
(16) He gave an on-topic but hastily prepared speech.
(17) a rapidly burning, totally ablaze building; a rapidly burning, totally on fire building
Third, certain small clauses permit PPs but not adjectives. They also permit A-adjectives:
(18) I have fond memories of him at work/ on shore/ with his friends.
(19) *I have fond memories of him crazy/ proud of his son/ sleepy.
(20) I have fond memories of him asleep/ alone in his office/ ashore.
So, there is quite a bit of distributional evidence indicating that A-adjectives are actually PPs. Presumably this evidence is available to language learners, who then construct a mental grammar where the A-adjectives are actually PPs. Given that they are PPs, they do not normally occur prenominally. There is nothing idiosyncratic about the A-adjectives at all.
Finally, this analysis has some consequences for particle shift in English. Some A-adjectives participate in particle shift, a fact that has been remarked upon but is generally thought to be mysterious or archaic (e.g., Bolinger 1971):
(21) They brought aboard the passengers. / They brought the passengers aboard.
(22) They took ashore the passengers. / They took the passengers ashore.
However, I personally feel that I have strong intuitions about these, even though I rarely hear or read them, which is odd if particle shift with the A-adjectives is archaic. If A-adjectives are actually PPs, though, this intuition about particle shift makes sense, because we can have a very simple generalization about particle shift: PPs that consist of a single morphological word can undergo particle shift. This includes the common particles out, up, in, off, and so on, but also the A-adjectives, which are PPs that are a single morphological word.
Bolinger, Dwight. (1971). The Phrasal Verb in English. Cambridge, MA: Harvard University Press.
Boyd, Jeremy K. and Adele E. Goldberg (to appear). Learning What Not to Say: The Role of Statistical Preemption and Categorization in A-Adjective Production. To appear in Language. Draft dated 10/10/2010 consulted here.
Huddleston, Rodney and Geoffrey K. Pullum (2002). The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press.