What prevents certain grammatical forms to be analysed as one word?

Question

When analysing a language, when do we analyse certain morphemes as one word as opposed to multiple, or is this arbitrary?

For instance, I could make the claim that (in certain cases) 'a/an' is a bound morpheme, and I could analyse it as a prefix on nouns. So, I would write 'anapple' instead of 'an apple'. We already see a considerable number of cases of this in English (eg: 'a lot' becoming 'alot')

(Since it's been pointed out that the 'a/an' example is in fact incorrect, a better example would be analysing 'of' as a head-marked posessive suffix.)

The converse is also possible. Why should I see Turkish 'evimiz' (house-our, "our house") as one word, and not 'ev imiz'? Turkish has bound words that display vowel harmony too, so why exactly must it be analysed as such?

Answer 1

It is not arbitrary, but it is very theory-dependent. One popular criterion for affix-hood is that affixes tend to affix to a particular word-class, thus if you treat "an" as an affix, you would expect that it only attaches to nouns (for example); but in fact it attaches to anything that can be on the left edge of an NP ("an old apple; an enormously expensive apple", etc.). However, there are "edge-inflections", where a certain feature is realized on the left or right edge of some constituent. From what I can tell, word / affix distinction is nonexistent in contemporary Minimalist approaches. Arnold Zwicky has written a fair amount about the problem of "words" and affixes.

Answer 2

It's generally assumed that you can't stick material of arbitrary length and near-abitrary content into the middle of a word. As mentioned in Yellow Sky's comment, we can say, in addition to "an apple", "a yellow apple", "a lovely, delicious yellow apple", and so on.

"Of" is not analyzed as a head-marked possessive suffix for a various reasons. There are many contexts where "of" phrases are used where "of" is not directly after the head of a possessive construction. "Of..." can be used predicatively or used in coordination ("the well-being of the people and of the government"). Also, you can put a conversational aside before "of", like "the perception--which I think is false--of X as Y...".

There are some real cases where it's unclear whether to analyze something as one word, but I don't think any of the examples you gave are of this kind.

Some examples I know of that have actually been considered to cause some difficulty for analysis:

• French verbs and the prefixed pronouns (especially the non-subject pronouns, but to an extent even the subject pronouns)

• French prepositions before proper nouns (like country names). This is discussed in "The Principle of Phonology-Free Syntax: four apparent counterexamples in French", by Miller, Pullum, & Zwicky (1997) (it's on the page that user6726 linked to; the analysis is as "phrasal inflection" of the first word of the NP).

• English compound nouns of the form adjective + noun that are written closed (like "blackbird") vs English adjective + noun sequences that are written open, but stressed on the first element (like "solar system" and "social worker"). This is discussed in "Adjectives, Compounds, and Words", by Laurie Bauer, which is also the source of these examples.

Answer 3

Words are the minimal forms that speakers memorize. (There might be exceptions to that general principle.)

There are several peculiarities of words not shared by other expressions, but so far as I know, there is no good discovery procedure for words -- that is, there is no set of observable facts about an expression that give a definite answer about whether or not it is a word. In that respect, words are like phonemes. Giving the words of a language is like giving the phoneme inventory; it is part of a theory of the language. There is no way of knowing for sure that it is entirely correct. Theories can be wrong.

And it is probably no accident that not only do words and phonemes share this abstract property, but also that words are always made of phonemes, with the exception of words made up of a single morpheme with some non-native sounds in it (I'm thinking here of "Bach" [bax], but I'm not sure whether the [x] always reverts to /k/ when an affix is added.)