Mind, New Series, Vol. 73, No. 290 (Apr., 1964), 204-214.

Is the sentence ' He had a green thought ' ungrammatical ? Some say " Of course not " and at once, without hesitation. I say :

1. Grammatical theory is a complex subject, abounding in technicalities. I mean to avoid technicalities here in so far as I can. The points I want to make are, to begin with anyway, fundamentally simple. They can be understood without a detailed investigation of their divers ramifications. So I shall for the most part speak in a relatively vague and intuitive way, using plain words even in cases where more technical locutions are available and would be vastly more precise. I shall also take various things for granted.

2. I know and I shall take for granted that ' He had a green thought ' is an English sentence. Thus one can sensibly ask whether or not it is a grammatical English sentence. (How I know that that is an English sentence is a hard question. The identification of an expression as a particular English sentence involves phonetic, phonemic, morphological, syntactic, and non-syntactic semantic considerations. I don't want to go into the question here but, as will be seen, I can't altogether avoid it.)

3. Suppose we have a grammar of English. Then, speaking somewhat vaguely, we could ask whether or not the sentence ' He had a green thought ' is in accordance with the grammar. (What the phrase ' in accordance with the grammar ' is supposed to mean is far from clear, how it is understood will depend on what kind of grammar is being considered. ) Suppose the sentence is in accordance with the grammar. Then if the grammar is a correct grammar of English, we should have to say that the sentence ' He had a green thought ' is a grammatical English sentence.

So the important question is : is the grammar a correct grammar of English ?

4. I think the point I am making here is obvious. But since some seem inclined to deny it, I mean to labour it a bit.

Some seem to think that since they weren't taught at school not to say ' He had a green thought ', there's nothing ungrammatical about the sentence. The sentence is grammatical according to the grammar they learned at school, so the sentence is grammatical.

The obvious reply to this is the point I'm labouring, viz. that if the grammar they learned at school is a correct grammar of English then what they say is so. But is the grammar they learned at school a correct grammar of English ?

" The way I speak is in accordance with the grammar I learned at school " : but is it ?

5. To know whether a grammar is a correct grammar of English we have to have some idea whether or not various expressions are grammatical English sentences and we have to determine whether or not these sentences are in accordance with the grammar.

This can be said in another way. Before we can hope to assess the correctness of a grammar, we have to have some intuitive idea of grammaticalness. An adequate grammar will be one that captures this intuition (and perhaps does other things as well). (A choice between alternative grammars might be based on considerations of simplicity, utility in teaching the language to foreigners, and so forth.) So let's look at the matter in an intuitive way.

6. Intuitively and somewhat vaguely speaking then, a native speaker balks when an ungrammatical sentence is uttered (other than by way of example, quotation, and the like and apart from rather special and specifiable contexts and discourses). (A more precise technical term is available here. Instead of speaking of speakers balking when a sentence is uttered, we could speak of semantically deviant utterances and semantically deviant sentences.[1] Thus a sophisticated reader of poetry does not balk at the phrase ' a green thought ' in Marvell's The Garden ; even so, it is semantically deviant.) Thus if children speak ungrammatically, some of us are apt to correct them ; some of us take steps to get them to cease talking in that way.

This suggests one necessary condition of ungrammaticalness, viz. that if a sentence is ungrammatical then native speakers balk when the sentence is uttered. (The qualification " unless the sentence is uttered by way of quotation, example, and the like and apart from rather special and specifiable contexts and discourses " is of course required, but I shall not bother to say all this each time it may be necessary. My aim in this paper is to convey some intuitive understanding of ungrammaticalness, that and nothing more.)

7. Consider the sentence ' He don't believe it ' : is this sentence ungrammatical ?

I say yes and no, for it depends on which dialect of English the sentence is supposed to occur in. In certain dialects of American English, such a sentence is grammatical : native speakers do not balk when it is uttered. (Indeed, in some cases such speakers might look askance at the sentence ' He doesn't believe it '.) In other dialects of English, speakers do balk when the sentence ' He don't believe it ' is uttered. In such cases the sentence can perhaps be correctly classed as ungrammatical. (I don't deny that some people believe that ' He don't believe it ' is ungrammatical even when native speakers don't and aren't apt to balk at the sentence. There's no accounting for what some people believe.)

8. Suppose I say ' It's raining ' when it isn't raining. Then what I've said isn't so. A native speaker of English aware of the state of the weather might balk when I utter the sentence ' It's raining ' : after all, if children go around saying things that aren't so, some of us take steps to get them to cease. So why not class the sentences uttered in saying what isn't so as ungrammatical ?

One reason is this : I might say ' It's raining ' when it is raining. There need be nothing to balk at in such a case.

9. If I say ' It's raining ' when it isn't raining, a native speaker might balk at what I say but he wouldn't be apt to balk at the sentence. Even if he did balk at the sentence, he would be balking at the sentence token, not the sentence type.

It's not necessary to confuse sentence tokens and sentence types. ' That's a cow ' and ' That's a cow ' are two different sentence tokens of one sentence type.

A sentence is ungrammatical if and only if every token of the sentence type is ungrammatical. Ungrammaticalness pertains primarily to sentence types.

Thus if I utter the sentence ' It's raining ' when it isn't raining and we class this sentence token as ungrammatical then we must also class the token ' It's raining ', uttered when it is raining, as ungrammatical. And this means that we would be balking at tokens that we have no reason to balk at.

10. But this is also something that wants explaining : why does ungrammaticalness pertain primarily to sentence types and not sentence tokens ?

Suppose we were to class only certain sentence tokens of a certain sentence type as ungrammatical. It would follow that it would not be possible to tell whether or not a given sentence token of the type in question was ungrammatical simply by examining the token. For since it is only by virtue of what they have in common that two sentence tokens are tokens of the same type, there could be no relevant difference between them on the basis of which one but not the other could be correctly classed as ungrammatical. (But this is not to deny that there are homonymous sentence types. For example, in the spoken language, ' I saw an Alaskan bare ' and ' I saw an Alaskan bear ' are homonymous sentence types. Again, the ungrammatical sentence ' I were going to the store ' need not be identified with the homonymous expression that occurs in the sentence ' If I were going to the store, I would get it for you '. The latter expression may well be a morphophonemic variant of the grammatical sentence type ' I was going to the store '.)

11. So, still intuitively and somewhat vaguely speaking, I suppose we can say this : if a sentence is ungrammatical then native speakers balk when an arbitrary token of the type is uttered. But there's more to be said.

If a sentence is ungrammatical, not only do native speakers balk when the sentence is uttered but they balk at it in that they balk because of certain structural features of an arbitrary token of the type.

For example, in certain dialects of English, the sentence ' He don't believe it ' is ungrammatical. Not only do the speakers of these dialects balk when such a sentence is uttered but their basis for doing so is quite clear. Thus grammatically sophisticated and articulate speakers of such dialects would point out that the subject pronoun ' he ' calls for the third person singular form ' does ' instead of ' do ' (or ' doesn't ' instead of ' don't ').

12. It's time to be a bit more precise, and unfortunately, a bit more technical.

A grammar of a language can be thought of as a system of some sort (though it can also and perhaps profitably be likened to a machine, or a device, or a procedure, etc.) for attributing structural descriptions to certain expressions of the language. Thus if we had a correct grammar of English, it would be possible correctly to attribute a structural description to any English sentence or expression whatsoever, whether grammatical or not. (If we couldn't do that, we could have no reason for classing a given expression either as an English expression or as an English sentence.)

Furthermore, the set of structural descriptions attributable to certain expressions of the language will, in a correct grammar of the language, be divided into at least two exclusive proper subsets. I shall speak of the set of accepted structural descriptions and of the set of nonaccepted structural descriptions. Supposing we have a correct grammar of the language, I shall say that a sentence of the language is grammatical if and only if its associated structural description is a member of the set of accepted structural descriptions. I shall say that a sentence of the language is ungrammatical if and only if its associated structural description is a member of the set of nonaccepted structural descriptions.

13. If a sentence is ungrammatical then, in a correct grammar of the language, that sentence will have a nonaccepted structural description. And this means that the sentence will have a structural description that is classed as nonaccepted on theoretic grounds. (If it were not so classed on theoretic grounds, it could not be so classed in the grammar.)

So suppose we say this : a sentence of a language is ungrammatical if and only if first, native speakers of the language balk when an arbitrary token of the type is uttered, and secondly the sentence type has an associated structural description that is classed as nonaccepted on theoretic grounds.

And the question now is : what must a structural description be like if it is to be classed as nonaccepted on theoretic grounds ?

14. Given a correct grammar of a language, an ordered set of categories can be associated with a given sentence. The categories in question may be word categories, e.g. Noun, Adjective, Verb, etc., or word-group categories, e.g. Noun-phrase, Verb-phrase, Subject, Predicate, Sentence, Passive Sentence, etc. (In more sophisticated grammars, the categories may be construction types, or constituent types, or transformational types, and so forth. To avoid unnecessary complications and technicalities, I shall here be concerned simply with word and word group categories.)

To give a structural description of a sentence is to associate an ordered set of categories with the sentence. (Thus in a familiar phrase-grammar, the ordered set {Sentence, Noun-Phrase, Verb-Phrase, Definite Article, Noun, Verb, Adjective] constitutes a structural description of the sentence ' The cat is hungry '.)

15. For a structural description to be classed as nonaccepted on theoretic grounds, it must be possible to designate an ordered set of categories as nonaccepted. But if this were all there was to it, then any structural description whatever could be classed as nonaccepted on theoretic grounds. Thus any given sentence type that occasions a balk by native speakers could then be classed as ungrammatical.

For example, knowledgeable native speakers might conceivably balk when the sentence ' An oyster isn't a bivalve ' is uttered, though presumably they wouldn't balk when the sentence ' An oyster is a bivalve ' is uttered. Let us introduce and invoke the categories �_� and C₂ such that ' an oyster ' falls under �_� and ' isn't a bivalve ' falls under (7₂. We can then designate the ordered set of categories {�_� C₂] as nonaccepted.

The obvious objection here is that the categories that have been introduced and invoked are completely ad hoc. (That we would also have introduced an asymmetry here by classing only the negative form of a sentence as ungrammatical is a relevant but not a conclusive point. At a relatively casual level of analysis, such asymmetry is already to be found in the language ; e.g. ' I haven't any money ' is grammatical but ' I have any money ' is ungrammatical, and so is ' It's not the case that I have any money '. However, there is good reason to say that ' I haven't any money ' is the negative form not of ' I have any money ' but rather of ' I have some money '. Given such a transformational analysis one could perhaps deny that such asymmetry is in fact already to be found in the language.)

16. When are categories ad hoc and when not ? That's an enormously difficult problem that I must and mean to side step for the time being. (But see below.) Let us suppose that somehow we can assign a degree of utility to a category or to an ordered set of categories such that the degree of utility will be the inverse of the degree of ad hocness.

Presumably even in a correct grammar of a language the grammatical categories will differ in degree of utility. Thus intuitively speaking, I suppose that the category Noun-Phrase is of greater utility than the category Mass Noun. (Any measure of utility would, I believe, have to accord with such an intuition to be acceptable.) But presumably in anything that we are prepared to call a correct grammar of a language, all the categories will be of a sufficient degree of utility to avoid the charge of ad hocness. (For if the categories were not of such a degree of utility, the systematic character of the grammar would be called into question.)

17. Let n be the minimum degree of utility sufficient to avoid the charge of ad hocness. Then, supposing again that we have a correct grammar of the language, we can say that a sentence of the language has a degree of ungrammaticalness greater than zero if and only if its associated structural description is in terms of an ordered set of categories such that the ordered set is of a degree of utility equal to or greater than � and where the structural description is a member of the class of nonaccepted structural descriptions.

Intuitively speaking, I am inclined to suppose that the degree to which a native speaker is apt to balk when an ungrammatical sentence is uttered could (in part anyway) be accounted for in terms of the degree of ungrammaticalness of the sentence. For example, a native speaker of my dialect is likely to balk when the sentence ' I have a elephant ' is uttered, but such a speaker would balk more strongly when the sentence ' I has an elephant ' is uttered. This difference probably reflects the difference in the degree of ungrammaticalness. The sentence ' I have a elephant ' exemplifies a deviation with respect to a morpho-phonemic category : before a vowel, the sandhi form ' an ' is required. Presumably such a category is of only minor utility. But the sentence ' I has an elephant ' exemplifies a deviation with respect to a proper subcategory of the category Verb, the subcategory in question being of considerable utility, or so I am inclined to suppose.

(But it must be understood that these remarks are of a speculative and intuitive character. In default of a measure of utility, one cannot speak otherwise.)

18. Is the sentence ' He had a green thought ' ungrammatical ? Do native speakers of the relevant dialects balk when the sentence is uttered (other than by way of example, quotation, and the like and apart from rather special and specifiable contexts and discourses, e.g. in poetry)? I think they do. (Whether a speaker is apt to balk of course depends not only on the utterance but on the speaker. But such complications are easily avoided if one proceeds in a more precise but more technical way, viz. by considering whether or not the utterance in question is semantically deviant. An intuitive account of complex matters can be achieved only at the cost of vagueness and imprecision.)

At least one distinguished philosopher maintains that ' He had a green thought ' is grammatical but nonsensical. To say that what is said in saying ' He had a green thought ' is nonsensical is to balk when the sentence is uttered, at least in the loose sense of ' to balk when the sentence is uttered ' that I intend here.

Of course, in so far as the philosopher insists that the sentence is grammatical, it might seem that he cannot be said to balk at the sentence. He doesn't think he's balking at the sentence uttered, he thinks he's balking at what is said in uttering the sentence. But maybe he's confused about what he's balking at. And anyway, it is clear enough that he balks when the sentence is uttered, and that's all that's required here. (I shall not discuss whether or not ' He had a green thought ' is nonsensical. However, I can see not the slightest reason to suppose that it is nonsensical. When Andrew Marvell wrote " Annihilating all that's made to a green thought in a green shade " he was not making a mistake, neither did he write anything nonsensical.)

19. The question then is : does the sentence ' He had a green thought ' have a structural description that can be classed as nonaccepted on theoretic grounds ?

Can we attribute to the sentence a structural description belonging to the class of nonaccepted structural descriptions, where the structural description is in terms of categories of a sufficiently high degree of utility to avoid the charge of �� Aocwess? To determine this, we have to determine which categories are relevant, and then we must in some way assess the degree of utility of the ordered set of categories associated with the sentence.

It is, I believe, quite clear that if we restricted our attention to categories presumably of a relatively high degree, e.g. to Noun, Verb, Adjective, etc., the sentence 'He had a green thought' would have to be classed as grammatical. More simply, in so far as we are concerned with categories of such a degree, there is nothing exceptional to be noted in connection with the phrase ' a green thought ' : it is simply an example of a noun phrase, not differing in that respect from ' a green house ' or ' a pink flower ', etc.

Hence if the question is whether ' He had a green thought ' is ungrammatical at some extremely high level of ungrammatical-ness, the answer would seem to be no. It does not follow that it is not ungrammatical to some lesser degree.

20. Consider such sentences as ' The fool is utter ', ' The fool is out and out ', ' She hurt himself '. Are these sentences ungrammatical ? It would be hard to deny it.

But it is, I believe, quite clear that if we restricted our attention to categories presumably of a relatively high degree of utility, e.g. to Noun, Verb, Adjective, etc., these sentences would have to be classed as grammatical. More simply, in so far as we are concerned with such categories, there is nothing exceptional to be noted in connection with the expression ' is utter ' : it is simply an example of a verb phrase, not differing in that respect from ' is tall ' or ' is morose ', etc.

A somewhat subtler analysis is evidently called for.

21. Consider the sentence ' It's nice to have tree ' : is this sentence ungrammatical ? I am inclined to suppose so. But if it is, what's ungrammatical about it ?

The word ' chicken ' in English falls under the category Count Noun and also under the category Mass Noun, and thus under the disjunctive category Count/Mass Noun, all of which are proper subcategories of the category Noun. The relevant difference between mass nouns and count nouns is that the former requires neither an article nor a plural affix. Thus the sentence ' It's nice to have chicken ' and the sentences ' It's nice to have a chicken ' and ' It's nice to have chickens ' are all grammatical. But if ' tree ' is simply a count noun and not a mass noun, and if these categories have a degree of utility equal to or greater than n, the sentence ' It's nice to have tree ' is ungrammatical.

To avoid sentences that, intuitively speaking, are clearly ungrammatical, it is necessary to attend to categories presumably of a lesser degree of utility than those of Noun, Verb, Adjective, etc. In particular, we must introduce and invoke proper sub-categories of these categories.

22. We must introduce and invoke proper subcategories of the familiar categories cited. But how many?

I suggest : a great many, about 7,023. (Why that number ? It's a nice big number and a big number is what is wanted here.) If a grammar is, as it were, to capture the intuition of even ordinary speakers and hearers, a great variety of categories is called for.

It will help here to look at the matter in a general way.

23. One might suppose that ideally every sentence type that occasions a balk should be classed as ungrammatical. Such an ideal (whether desirable or not) is unattainable. Alternatively, one might suppose that every sentence type that is itself balked at (and not merely one that occasions a balk) should be classed as ungrammatical. But even this ideal is in fact unattainable.

There are limits to what a grammar can enable us to do while still retaining its character as a grammar.

24. Is ' She hurt himself ' ungrammatical ? I am inclined to say so. Just so 'The lady hurt himself, 'The woman hurt himself ', ' His wife hurt himself ', etc. are all ungrammatical in my dialect of English.

Suppose we introduce a category of feminine and a category of masculine expressions : thus ' wife ', ' woman ', ' herself ', etc. are feminine expressions ; any endocentric noun construction with a feminine expression as head is also a feminine expression : thus ' the lady in the hallway ' is a feminine expression. The category of masculine expressions could be specified in the same way. We could then designate the structural description associated with ' She hurt himself ' as nonaccepted owing to the fact that the environment ' . . . hurt-------(-self' is open only to expressions of the same category with respect to masculine or feminine. (The details of such an analysis are somewhat complex, but the point is simple enough.) So we could say that ' She hurt himself ' and ' His wife hurt himself ' are ungrammatical.

25. But now consider the sentences ' His wife hurt himself ' and ' The adult female person he is married to hurt himself. These sentences may not be exactly synonymous but the difference in significance is surely slight. Even so, we can hardly maintain that ' The adult female person he is married to hurt himself ' is ungrammatical, and this despite the fact that native speakers would be apt not only to balk when the sentence is uttered but to balk at the sentence. To class such a sentence as ungrammatical, very special categories would have to be introduced and invoked. Presumably such categories would be of little utility. And perhaps more significantly, the same sort of difficulty would only crop up again.

For consider the sentence ' The person of the opposite sex he was married to before he became a man hurt herself ' : this sentence poses the same problem. To avoid it, we should have to maintain that the expression ' the person of the opposite sex he was married to before he became a man ' falls under the category Masculine. But on what grounds, or more precisely, on what syntactic grounds ? Or to make matters even worse, consider the environment ' The person of the sex opposite to his at that time since when he has changed sex seven times, that person hurt . . .-(-self ' : what fills the blank, ' him ' or ' her '? Assuming that elementary arithmetic is not a part of grammar, this is not a grammatical question.

26. There are limits to what a grammar can enable us to do while still retaining its character as a grammar. But the exploration of these limits and the determination of precisely what does and what does not fall within these limits is a subtle and difficult task for grammarians. This is not an area in which the idle pronouncements of philosophers are apt to be illuminating.

Are the following sentences ungrammatical ? He wants my own hat. A perfect fool is looked by him. I have any books A filth is to be avoided. Puppies look barking. The fork eats. He was reading bore books. The men grief the women. He has a red good apple. He had a green thought. In each case, to answer the question calls for work. To say anything sensible here, one has to leave the prim path of rosy speculation and muck about with the data. I shall illustrate this (if only briefly) in connection with ' He had a green thought '.

27. Let us introduce and invoke the categories N_a and A_s such that ' thought ', ' fact ', ' idea ', ' statement ', ' proposition ', etc. fall under the word category N_s, and such that ' interesting ', ' strange ', ' surprising ', ' true ', ' possible ', etc. but not ' green ' fall under the word category A_s. (Thus the categories in question can be effectively characterized by a complete list of the words that fall under them.) Let �A_s be the category of words that do not fall under the category A_s. We can then designate the ordered set of categories {Article, � A,, N_s} as a member of the set of nonaccepted structural descriptions. Thus ' He had a green thought ' is ungrammatical.

28. The question now is : are the categories in question, viz. N_s, A_s, and �A_s of a sufficient degree of utility to avoid the charge of ad hocness ? In default of any measure of utility, no precise answer can be given to this question. Even so, we can look at the matter in an intuitive way.

29. Consider the following pairs of sentences :

la He had a green thought.

Ib He had a strange thought.

2a The thought that George was alive was green.

2b The thought that George was alive was strange.

�� That George was alive was green.

3b That George was alive was strange.

4a It was green that George was alive.

4b It was strange that George was alive.

Notice that when ' green ' is replaced by ' strange ' there is nothing strange. Further notice that the members of N, are all intimately associated with sentences ; technically speaking they appear to be sentence nominalizations, they replace sentences. The members of A_s are also intimately associated with sentences ; technically speaking they appear to be sentence adjuncts, they occur as modifiers of sentences.

I suggest that this is very strong evidence and thus a very strong indication that the categories introduced and invoked are not wholly ad hoc. (Technically speaking, sentences lb-4b indicate that the categories in question will be of utility in a transformational analysis of English.)

Is ' He had a green thought ' ungrammatical ? That depends on what the or a correct grammar of English proves to be. But for the time being I am certainly inclined to suppose that it is. The University of Pennsylvania

[1] See my Semantic Analysis, Cornell University Press, 1960.