5. Conclusion: Some General Principles*
No attempt is made here to account for all of the universals described in the preceding sections and repeated in Appendix III. Some general principles, however, are proposed which seem to underlie a number of different universals and from which they may be deduced. Attention is first directed to those universals which are most closely connected with the basic order typology and the closely associated genitive construction. Two basic notions, that of the dominance of a particular order over its alternative and that of harmonic and disharmonic relations among distinct rules of order, are introduced. This latter concept is very obviously connected with the psychological concept of generalization.
We may illustrate the reasoning involved by reference to Universal 25, according to which, if the pronominal object follows the verb, the nominal object does so likewise. In other words, in the tetrachoric table resulting from the alternative for each of the combinations there is a single blank. Since the nominal object may follow the verb whether the pronoun object precedes or follows, while the nominal object may precede the verb only if the pronoun precedes, we will say that VO is dominant over OV since OV only occurs under specified conditions, namely when the pronominal object likewise precedes, while VO is not subject to such limitations. Further, the order noun object-verb is harmonic with pronoun object-verb but is disharmonic with verb-pronoun object since it does not occur with it. Likewise verb-noun object order is harmonic with verb-pronoun object and disharmonic with pronoun object-verb. We may restate our rule, then, in terms of these concepts as follows:
A dominant order may always occur, but its opposite, the recessive, occurs only when a harmonic construction is likewise present. Note that the notion of dominance is not based on its more frequent occurrence but on the logical factor of a zero in the tetrachoric table. It is not difficult to construct an example in which one of the recessive alternatives is more frequent than the dominant. Dominance and harmonic relations can be derived quite mechanically from such a table with a single zero. The entry with zero is always the recessive one for each construction, and the two constructions involved are disharmonic with each other.Harmonic and disharmonic relations, as noted earlier, are examples of generalization. In similar constructions, the corresponding members tend to be in the same order. The basis for the correspondence in the present instance is obvious, in that pronoun and noun are both objects of the verb, and the other pair verb-verb is identical. In regard to harmonic and disharmonic relations, a fair amount of freedom will be exercised based on transformational and other relations among constructions, not merely the occurrence of a zero in a tetrachoric table.
Proceeding on this basis, we now consider Universal 3. It will be noted that this universal amounts to an assertion of the nonexistence of postpositional languages of type I. Since in all of the types, I, II and III, S precedes O. this is irrelevant for the present context. This leads to the following conclusions:
Prepositions are dominant over postpositions, and SV order is dominant over VS order. Further, prepositions arc harmonic with VS and disharmonic with SV, while postpositions are harmonic with SV and disharmonic with VS. What distinguishes type II from type III is that in type II the object follows the verb, a characteristic shared with type I. On the other hand, type III has the object before the verb. From Universal 4, which states that with overwhelmingly more than chance frequency SOV is associated with postpositions, the conclusion is drawn that OV is harmonic with postpositions while VO is harmonic with prepositions. The constructional analogies which support this are discussed later with reference to the closely associated genitive constructions. For the moment it may be noted that the relations between types I, II, and III and Pr/Po may now be recapitulated in these terms: Type I has VS which is harmonic with prepositions, and SO which is likewise harmonic with prepositions. Further, prepositions are dominant. All languages of type I, in fact, are prepositional. Type II has SV which is harmonic with postpositions and VO which is harmonic with prepositions, and prepositions are dominant. In fact, a definite majority of languages of type II have prepositions. Type III has SV and OV, both of which are harmonic with postpositions. However, prepositions are dominant. In fact, the preponderant majority of languages which have type III have postpositions, with but a handful of exceptions.From the overwhelming association of prepositions with governing noun-genitive order and of postpositions with genitive-governing noun order but with a small number of exceptions of both types, the conclusion is drawn that prepositions are harmonic with NG and postpositions with GN.
The close connection between genitive order and Pr/Po is a simple instance of generalization. The relation of possession is assimilated to other relational notions, for example, spatial relations. In English, 'of' which marks possession is a preposition with the same order properties as 'under', 'above', etc. Further, such spatial and temporal relations are often expressed by nouns or nounlike words, for example, English 'in back of'. In many languages 'behind' = 'the back + genitive'; hence: 'X's back' = 'in back of X' parallels 'X's house'; and 'back of X' = 'in back of X' parallels 'house of X'.
The connection between these genitives and the analogous prepositional or postpositional phrases on the one hand, and subject-verb and object-verb constructions on the other, is via the so-called subjective and objective genitive. Note that in English 'Brutus' killing of Caesar started a civil war' has the same truth value as 'The fact that Brutus killed Caesar started a civil war'. The order of elements is likewise similar. In other words, in such transformations, the noun subject or object corresponds to the genitive, and the verb to the governing noun. In fact, there are languages in which the subject or the object of the verb is in the genitive. For example, in Berber argaz 'man' is the general form of the noun, and urgaz is either the dependent genitive or the subject of the verb, provided it follows immediately. Thus iffer urgaz, 'went out the man', exactly parallels axam urgaz, 'the house of the man'. Berber, it will be noted, is a language of type I, and the genitive follows the noun. It likewise has prepositions rather than postpositions.
A further relationship among the variables of the basic order typology may be posited, that between genitive order and adjective order. Both the genitive and qualifying adjectives limit the meaning of the noun. There are further facts to support this. There are languages like Persian, in which both adjective and genitive dependence are marked by exactly the same formal means. Where pronominal possession is involved, some languages use a derived adjective, while others use a genitive of the pronoun. There are even instances where adjectives are used in the first and second person, while a genitive is used in the third person (e.g., Norwegian).
We may summarize these results by stating that all of the following are directly or indirectly harmonic with each other: prepositions, NG, VS, VO, NA. We have here a general tendency to put modified before modifier, and the most highly "polarized" languages in this direction are those of type I with NG and NA, a considerable group of languages. The opposite type is based on harmonic relations among postpositions, ON, SV, OV, and AN. This is also a very widespread type, as exemplified by Turkish and others in the present sample. On the other hand, the general dominance of NA order tends to make languages of the Basque type (i.e., III/Po/NA with GN order) very nearly as common as the Turkish type. It should also be pointed out that languages being highly complex structures, there are other factors at work in individual cases not included among the five factors cited at this point. One of them, demonstrative-noun order, has already been mentioned.
It is more difficult to account for the dominances than for the harmonic relations, to explain, for example, why the adjective tends to follow the noun. It may be suggested, however, that noun-adjective predominance arises from the same factor as that which makes subject-verb the dominant order. In Hockett's terminology, there is a general tendency for comment to follow topic. There is some evidence that noun-adjective does parallel subject-verb in this way. In many languages all adjectival notions are treated as intransitive verbs. The qualifying adjective is then a relative or participle of the verb. The tendency of relative clauses, it has been seen, is even stronger than that of adjectives to follow the noun. In some languages such as Arapesh in New Guinea, 'The good man came' would be literally translated 'The man is-good that-one he came'. Adjective-noun order, then, is somewhat ambivalent since analogies with other constructions involving modifiers make it indirectly harmonic with VS while the factor of topic-comment order makes it analogous with SV.
All this is far from a complete theory. Nevertheless, it does suggest that one should examine instances in which, contrary to the prevailing rules, the genitive construction is disharmonic with Pr/Po. One would reason that in such cases the genitive construction is, as it were, being attracted by the adjective-noun construction which, as has been seen, has sources of determination that are to some extent outside of the general framework of harmonic relations connected with the order of modifier and modified. For example, if, in spite of the general rule, we find genitive-governing noun order with prepositions, the reason might be the opposing pull of order adjective-noun which is harmonic with genitive-governing noun. Otherwise stated, the genitive construction should only be disharmonic with Pr/Po when Pr/Po is disharmonic with the adjective-noun order. One may include here cases in which a language has two genitive orders, indicating a probable change of type since one must, in all likelihood, be older than the other. One may further conjecture that if there are exceptions, they will be in type II, which, having both SV and VO which are disharmonic, can provide an anchor in either case for deviant genitive order.
It will be noted that Universal 5, insofar as it refers to postpositional languages of type III (the vast majority), gives a particular instance of this hypothesis; for this statement asserts that a language of type III if it has NG will also have NA. If such a language is postpositional, then NG will be disharmonic with postpositions but harmonic with NA. If we include languages with both genitive orders, then there are at least six cases, all favorable (i.e., with NA rather than AN). These are Somali and Maba with both genitive orders, and Kanuri, Galla, Teda, and Sumerian which have SOV, postpositions, NG, and NA .
This hypothesis will, however, produce some further predictions. For prepositional languages of type III, the hypothesis will be that with varying genitive order or with GN, which is disharmonic with prepositions, the adjective-noun order will be AN. I know of only two cases, Tigrinya with both genitive orders, and Amharic with GN. Both have AN in accordance with our hypothesis. For languages of type II which are prepositional and which have GN, and should therefore have AN, we have Danish, Norwegian and Swedish (possibly a single case), and English with two genitive orders. Both fulfill the hypothesis in that they have AN. Among postpositional languages of type II, we have the Moru-Madi group in the Sudan and the fairly distantly related Mangbetu, both of which, with alternative genitive orders, have the predicted NA. We now encounter the only exceptions of which I am aware, Araucanian in Chile, with both genitive orders; and a group of Daghestan languages in the Caucasus, including some like Rutulian with NG, and others like Tabassaran with both genitive orders. Apparently all those languages of the Daghestan group which are of type III have only GN harmonising with both postpositions and AN. If so, this is an important indication of the general validity of our hypothesis. Finally, since all languages of type I are prepositional, we have only a single case to consider, prepositional languages with GN. I know of only one example, the Milpa Alta dialect of Nahuatl described by Whorf. It has AN as expected.
Another type of relation than those that have just been considered is illustrated by Universals 20 and 29. These may be called proximity hierarchies. What we have is a rule that certain elements must be closer to some central element than some other satellite. The central element may be the root morpheme or base of a word or the head-word of an endocentric construction. Such a proximity hierarchy is likely to be related to an implicational hierarchy in the instance of inflectional categories. Just as the category of number is almost always closer to the base than expressions of case, so there are many languages with the category of number but without the category of case, and very few with case but without number. Since, by the proximity hierarchy, number is closer, it is more likely to become amalgamated with the base and so become an inflection. These hierarchies are presumably related to degrees of logical and psychological remoteness from the center, but no analysis is attempted here.
These phenomena are likewise related to those of neutralization. The more proximate category, or the implied category, tends to be more elaborate, and it is the less proximate or the implying categories which tend to be neutralized in its presence. Universals 36 and 37 are related in this manner. Number is the implied category. Gender categories are often neutralized in the marked number (i.e. nonsingular). It is much rarer for number to be neutralized in some particular gender (e.g., the neuter in Dravidian languages). With regard to number and case, number is, as has been seen, more proximate and generally present when case is present, while the opposite relation holds far more rarely. It is likewise common for certain case distinctions to be neutralized in number, while the opposite phenomenon perhaps never occurs.
Another principle is evident from Universal 34. We do not have such systems as the following: a particular grammatical category for the trial, while another embraces the dual and all numbers larger than three. In other words, disjunctiveness or lack of continuity in this respect is never tolerated.
Universals 14 and 15 possibly illustrate the same principle. The order of elements in language parallels that in physical experience or the order of knowledge. In the instance of conditionals, although the truth relations involved are timeless, logicians have always symbolized in the order implying, implied exactly as in spoken language. If modus ponens used in proof, then we have a pragmatic example which follows the order of reasoning. No one thinks to write a proof backwards.
Universals 7, 8, and 40, although superficially very different, seem to be examples of the same general tendency to mark the end of units rather than the beginning. For example, in rigid subtype III, the verb marks the end of the sentence. When the infections occur only with the final member of the noun phrase, this marks the end of the phrase. This is probably related to the fact that we always know when someone has just begun speaking, but it is our sad experience that without some marker we don't know when the speaker will finish.
The existence of a rigid subtype III, whereas there are no examples of a rigid subtype of I, is probably related to still another factor. In general the initial position is the emphatic one, and while there are other methods of emphasis (e.g., stress), the initial position always seems to be left free so that an element to which attention is directed may come first. Here Universal 12 is an example. It seems probable that in all languages expressions of time and place may appear in the initial positions in the sentence.
The discontinuity of the predicate, which commonly appears in such instances (e.g., German, 'Gestern ist mein Vater nach Berlin gefahren'), illustrates a further principle. On the whole, the higher the construction in an immediate constituent hierarchy, the freer the order of the constituent elements. It has been seen that practically all languages have some freedom of order regarding subject and predicate as a whole; whereas only a small minority have variant order in genitive constructions, and then almost always along with other differences, not merely a difference of order. Within morphological constructions, order is the most fixed of all. On the whole, then, discontinuous constituents are far less frequent than continuous ones.
As indicated in the initial section of this paper, the principles described in this section are to be viewed as no more than suggestive. It is hoped that some of them at least will prove useful for further investigation.
*Joseph H. Greenberg, "Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements", In: Joseph H. Greenberg (ed.). Universals of Language. London: MIT Press, pp. 96-104.