Variation in the English Indefinite Article
The problem of variation in the English indefinite article between the forms a and an has long vexed linguists. In his 1933 classic, Language, Bloomfield cited this case as an example of free variation at the morphological level, saying, “There seems to be no principled basis for predicting which form occurs in which contexts.” This solution was accepted by the neo-Bloomfieldians in general.
It was Jespersen who first questioned the Bloomfieldian solution. In 1941, he proposed that the syntactic class of the following word determined the form of the indefinite article; specifically, an occurred before adjectives, and a before nouns. He cited the following examples to support his theory.
1. an old man
2. a board
3. an elegant woman
4. a desktop
5. an orange monkey
6. a photograph
This solution was widely hailed as a breakthrough in a previously misunderstood area. Over the next few years, it came to be generally accepted. However a small minority of linguists tried to poke holes in the theory. The most significant early attack was based on example 7, below,
7. an orange
where a noun is preceded by an. The usual explanation for this anomaly was that orange, although a noun, was derived from an adjective and still retained adjectival force. This explanation, of course, assumed a synchronic rather than a historical derivation, since in fact the adjective historically derives from the noun. A slightly different explanation was that an orange is a truncated form of the phrase an orange fruit.
In 1954, the second major attack, which was actually an extension of the first, was issued, based on example 8, below.
8. a green (as in golf)
The argument ran as follows: if #7 is to be explained by any of the arguments given above, these arguments should also apply to a green. According to the theory, the form should be an green—this form, though, is in fact ungrammatical. Some adherents of the Jespersenian theory argued that the truncated solution to account for an orange likewise accounted for a green, since a green is not an obvious truncation of any longer phrase; or, alternately, it is a truncation of the phrase a golf green, where golf, though used as a modifier of sorts, retains its nominal force. However, most linguists considered these arguments weak and admitted that the Jespersenian solution was flawed. Nonetheless, since there was no paradigm which could replace it, the Jespersenian solution was still generally accepted.
So matters stood when, in 1959, Chomsky published his landmark article “The Inadequacy of Immediate Constituent Structure Grammars: Further Evidence from the English Noun Phrase.” Chomsky criticized earlier linguists for thinking that the morphological form of the indefinite article was determined by the syntactic class of whatever word happened to follow it. In fact, Chomsky pointed out, the article functions as a determiner of the noun which is its head, and one should therefore expect its morphology to be governed by the noun. After examining the standard data, i.e. those given as examples above, he proposed the revolutionary theory that an is the indefinite article used with animate nouns and a that used with inanimates. All of the data used so far will be found to comply with this solution. As to the pesky an orange/a green, Chomsky proposed that oranges, single living things, are animate in a way that greens, collections of independently living entities not possessed of a collective life, are not. As support for this explanation he presented the following further examples.
9. an elm
10. an apple
11. an elephant
12. a field
13. a sea of people
14. a flat surface
15. a long fence
#9-11 were meant to show that individual living things, including plants, took an, while #12-13 were evidence that groups of living things, whether plants or more traditionally “animate”, took a. #14-15 showed that inanimates took a. Significantly, #9-11 and 14-15 were inexplicable by Jespersen’s theory.
Chomsky’s theory, which seemed to solve the problem so well, was universally accepted. Soon one of its new adherents noted that Chomsky’s analysis paralleled the traditional analysis of variation in the Spanish indefinite article, where the form is dictated by the noun’s gender. Because the theory seemed to work so well, because its creator was so illustrious, and because it had typological parallels, it was not challenged for many years.
When the challenge came, it originated in typology. Greenberg’s work in African languages had convinced him that most linguists had a very limited view of what possible noun classes could be. They classed nouns usually according to gender (masculine, feminine, neuter) or animacy (animate or inanimate). Greenberg believed that further investigation might show that the a/an dichotomy reflected a sort of noun class less familiar to European languages, but no less typologically possible. Reviewing the data presented so far, he determined that it was possible to divide up the nouns into classes of “items having significant size in only two dimensions,” which took a, and “others,” which took an. For the data already given, Greenberg’s theory worked quite as well as Chomsky’s. Greenberg next sought out and found the crucial examples below,
16. an orb
17. a manta ray
which contradict Chomsky’s theory but support Greenberg’s.
The year was 1971; linguistics was a volatile field; and people who had staked their reputations on the Chomskyan theory were not going to take this attack lying down. The first counterattacks concentrated on example #13, a sea of people; it was argued that a sea of people had a significant depth as well as length and width. Most reasonable scholars, however, agreed that in a sea of people the only significant dimensions are lateral. After all, there are no truly two-dimensional items in the world, only those we treat as such in thought and language. It was not until 1972 that Postal issued the truly damning attack on Greenberg’s theory, based on the two examples below.
18. an album
19. a basketball
Plainly, Greenberg’s theory could not account for #18-19. On the other hand, neither could Chomsky’s, and though attempts were made to rework it, all of them failed. With both Greenberg and Chomsky’s analyses invalidated, there began a mad scramble to find some other syntactic basis for what was still considered a morphological problem. Theories were proposed and discarded, more and more examples were considered, but to this day there has been no success. Indeed, many linguists have despairingly returned to the solution offered by Bloomfield and now claim that this is a case of morphological free variation.
Is this the solution? Do linguists delude themselves when they think that there might still be a solution to this seemingly intractable problem? The answer is a resounding no. Indeed, linguists are misguided on their views on this issue, but the neo-neo-Bloomfieldian free variationists are no less so than their colleagues. Their mistake, the mistake of all who have considered this problem so far, lies in their belief that the variation is morphological in nature. It is an acknowledged fact that the neo-Bloomfieldians concentrated on syntax and semantics to the virtual exclusion of phonology. This sin has persisted in linguistics since the neo-Bloomfieldian days, although lately generativists have been less guilty of it than others. Because of this, no one ever considered the simple, obvious solution which I now present: an occurs before words beginning with vowels and a occurs before words beginning with consonants. This solution is simple, economical, and accounts for all of the data in English, barring performance errors. Note that when an occurs before words beginning with orthographic h, this h is not pronounced; e.g., an herb. This non-pronunciation of initial h is especially common among French and British people, who can’t pronounce English very well.
In conclusion, I should like to urge that this article be taken to heart by those linguists who still ignore phonology in their analyses. As demonstrated here, many seemingly intractable problems can be solved with a little bit of effort in the right direction.
| Tim Pulju |
Michigan State University |
I would like to thank Jim Copeland for valuable comments and suggestions on an earlier draft of this paper. However, I can’t, because he made no valuable comments or suggestions on an earlier draft of this paper. As a matter of fact, there was no earlier draft.
 |
![[redacted]](/images/b.gif "[redacted]")
—J. Fraser Bennett |
 |
English is the Original Language—Hans Melkor |
 |
Ps. Q. Vol XVI, No 4 Contents |