GRAMMAR. According to the definition of the late Dr. Henry Sweet a grammar gives the general facts of language, while a dictionary deals with the special facts of language. But the two domains frequently overlap, so that one and the same fact finds its place in the grammar as well as in the dictionary; this is because in order to state a rule correctly we must also state its limitations, i.e., the special cases in which it does not hold good. If we give the rule that English substantives form their plurals in -s, we must add that besides this regular forma tion we have the irregular plurals men, women, oxen, etc. And as languages are not constructed after ideal patterns, such excep tions to the rules must necessarily take up much space in all books on grammar.
To the ordinary man, grammar means a set of more or less arbitrary rules which he has to observe if he wants to speak or write correctly. This is especially the case if he is engaged in the study of a foreign language, but he is often led to the same point of view by the grammar of his own native language, as taught in schools. Grammar treated in this way may be called normative or prescriptive grammar.
But to the scientific grammarian the subject has a different aspect : to him the rules are not what he has to observe but what he observes (in a different sense) when he examines the way in which speakers and writers belonging to a particular community or nation actually use their mother-tongue. His attitude to wards linguistic phenomena is therefore much more that of a naturalist observing the facts of nature ; he stands more objec tively outside the language he is studying, and perhaps never has to form one single sentence in it for himself. This we term descriptive grammar.
The grammatical observer, like the observer in other fields, seeks, wherever possible, to go beyond the mere facts in order to find their explanation. This is the function of comparative historical grammar, a creation of the 1 gth century (see PHILOLOGY). Many things which seem strange from the point of view of merely descriptive grammar find their natural explana tion when viewed in the light of earlier periods of the same lan guage or of related languages. Take such an abnormal plural as feet from foot: the historian finds that its long vowel goes back through a regular phonetic development to an earlier oe which, wherever it was found, was treated in the same way (thus in feed, green, sweet) and like other oe's was a mutated form of a still earlier o—the vowel that is better preserved in the singular foot, where, however, it has now been shortened and raised. The mutation was here, as elsewhere, due to the existence of an earlier i in the final syllable, which was dropped in all analogous cases. Now we know that the ending in the plural in the earliest Ger manic was very often -iz, which corresponds to a still earlier preserved in Latin and Greek; the form feet, which from the one-sided Modern English point of view was an isolated fact, is thus seen to correspond to the Greek plural vroSes and to be con nected with that form through a long series of perfectly normal historical changes, which do not only affect such plural forms but find parallels in other words as well. The historical and compara tive method of explaining grammatical facts has been carried to a rare degree of perfection, but it is clear that it can only be em ployed to the full where we have early linguistic documents of the same language or of nearly related languages to refer to. The great majority of languages are only known to us in quite recent stages ; here, however, a similar method of explanation may be used if there are other now existing languages that are akin to that we are examining, and the comparative method then sometimes allows us within certain limits to reconstruct a common basis from which the several languages have started, as with the numerous African languages known as Bantu.
Grammatical reconstructions should always be made with great caution, for the ways in which languages develop are not always easy to calculate. We may take the Romanic languages (Italian, French, Spanish, etc.) as a test case: all these languages have been known to us for several centuries ; now in some cases it would be possible from existing forms in them to infer what the common basis must have been, and the forms thus reconstructed would agree pretty closely with the forms of what we know to have been the basis, namely Latin; but the method fails utterly, as has been well remarked, with regard to many other forms: no one would be able, for instance, to conclude from the forms of Romanic substantives that Latin had ever had an accusative in -m, for the only remnant is French rien from Latin rem "a thing" —and that now means "nothing" and can no longer be called an accusative.
The method of comparative grammar was especially developed in the study of our own family of languages, the Aryan or Indo European family (q.v.), and at a certain stage of its development scholars were naturally tempted to dwell on and to a certain extent exaggerate those features that were common to these lan guages, and to take less account of features which were peculiar to one or a few of them. There was always a tendency to think that these were survivals of primitive common phenomena which were lost in the other languages of the group. This may be true in some cases, but more often we see that something found in one language only is a recent development that has really noth ing to do with the rest of the family and may constitute a new grammatical type or phenomenon. Comparative grammar should therefore always be supplemented by separative grammar which does full justice to what is peculiar to each separate language and treats each on its own merits.
Thus not only separate grammatical forms, but whole gram matical categories may be dropped in course of time. Generally this does not take place all at once but gradually, those forms which are in constant use being sometimes preserved for a long time after the others have been given up. The old Aryan (Indo European) languages had separate forms for the dual number, distinct from the plural, but that distinction has been nearly universally lost. In Greek the dual was an archaism in Homer, though it lived on as a colloquialism in Attic till finally it dis appeared there too. In the oldest English a few pronominal forms such as unc us two, inc you two, are the only survivals of a separate dual, and from about 1250 they go completely out of use. In Russian the dual, which ended in -a, has left some curi ous traces which are no longer felt as a separate number : some words denoting parts of the body which are found in pairs, form their plural in -a: glaza eyes, yoga horns, etc. ; after the numerals 3. 4 a form that looks like a genitive singular is used instead of the usual plural form : it is the old dual in -a, which is extended to three and four.
In dealing with any definite period of a language it is important to state exactly which categories are found and which not. Old English had, but Modern English has not, a dative case. When the old forms were given up in the Middle English period, traces of them were still preserved in some survivals, e.g., in Chaucer of towne (with e pronounced as a separate syllable), yeer by yere, by weste; a few isolated remnants exist still, though no longer felt as separate case-forms: alive="on life" (dat.), Atterbury= "at the (dat.) borough." In a sentence like "he gave his children food" or a phrase like "from his children," Old English used the form cildrum, while the form was cildru in the nominative and accusative plural. Now the distinction has disappeared. To say that English still uses a dative case in these combinations is just as unhistorical as to say that Normandy and Massachusetts still form parts of the British Empire. This does not, of course, amount to denying that children in the sentence above is an in direct object, to be distinguished from the direct object (food). If we were to speak of a dative case here we might just as well say that in "Tom and Mary are children" the last word is gram matically in the dual number, but who would say this? A word or form belonging to one grammatical category may in course of time be shifted insensibly into another one. Thus near at first was the comparative of nigh, with a superlative next; but in such a sentence as "Come near ! " the meaning might be equally well taken as "closer" or "close," and thus the word passed into its modern use as a "positive" and it became possi ble to form a new comparative and superlative nearer, nearest, while the old superlative next was specialized in its use, and nigh became obsolete.
New grammatical categories may develop; examples are the English "expanded" or "progressive" tenses : he is running, was running, has been running, etc., as distinct from he runs, ran, has run. The distinction between "absolute" (primary) and "conjoint" (adjunctive) possessive pronouns, e.g., mine, yours as distinct from my, your, is another case in point. There is in some languages a tendency in regard to personal pronouns to merge the distinction of nominative and objective in that of conjoint and absolute, the old nominative being used only when it stands in immediate connection with a verb as subject, and the old objec tive in all other positions. This has become the rule in French, where je is used only in combinations like je dis, dis-je (I say, say I) and moi, which is the stressed form of the accusative me, is found in c'est moi and Qui l'a dit? Moi (Who said it? I) . In Italian we see similar tendencies, and in modern colloquial Eng lish me tends to supersede the literary I in It is me and Who's there? Me.
As languages are thus seen to be in constant flux, and as gram matical categories may to a certain extent change from language to language and even from one period to another—and as exotic and "savage" languages possess many categories unknown to our European languages, it will easily be understood how injurious it is to a scientific conception of grammar to measure it always and everywhere by the same standard. But that is what grammarians of former centuries and even recent writers have been and are in the habit of doing ; for Latin was for centuries the only lan guage studied grammatically, and its privileged position made people think it a pattern by which to measure all other languages. Not only those languages that were similar in structure to Latin, but even the most heterogeneous languages were indiscriminately saddled with the elaborate Latin system of tenses and moods, and by means of such Procrustean methods the actual facts of many languages were distorted and misrepresented. Discrimina tions which had no foundation in reality were nevertheless insisted on, while discriminations that happened to be non-existent in Latin were apt to be overlooked. As A. H. Sayce writes in the 9th edition of the Encyclopedia Britannica, "The endeavour to find the distinctions of Latin grammar in that of English has only resulted in grotesque errors, and a total misapprehension of the usage of the English language." Happily things are improving very considerably in this respect.