Anthropology

by A. L. Kroeber

48. Criteria of Relationship

The question that first arises in regard to linguistic families is how the relationship of their constituent idioms is determined. In brief, the method is one of comparison. If a considerable proportion of the words and grammatical forms of two languages are reasonably similar, similar enough to indicate that the resemblances cannot be due to mere accident, these similar words and forms must go back to a common source, and if this source is not borrowing, the two tongues are related. If comparison fails to bring out any such degree of resemblance, the languages are classed in distinct families.

Of course it is possible that the reason two languages seem unrelated is not that they are really so, but that they have in the lapse of ages become so much differentiated that one cannot any longer find resemblance between their forms. In that event true relationship would be obscured by its remoteness. Theoretically there is high probability that many families of languages, customarily regarded as totally distinct, do go back in the far past to a common origin, and that ignorance of their history, or inability to analyze them deeply, prevents recognition of their relationship. From time to time it happens that groups of languages which at first seemed unrelated are shown by more intensive study to possess elements enough in common to compel the recognition of their original unity. In that case what were supposed to be several “families” become merged in one. The scope of a particular family may be thus enlarged; but the scope of the generic concept of “family” is not altered.

Whether there is any hope that comparative philology may ultimately be prosecuted with sufficient success to lead all the varied forms of human speech back to a single origin, is an interesting speculation. A fair statement is that such a possibility cannot be denied, but that the science is still far from such a realization, and that progress toward it is necessarily slow. Of more immediate concern is an ordering and summarizing of the knowledge in hand with a view to such positive inferences as can be drawn.

In an estimate of the similarity of languages, words that count as evidence must meet two requirements: they must be alike or traceably similar in sound; and they must be alike or similar in meaning. This double requirement holds, whether full words or separable parts of words, roots or grammatical forms, are compared. The English word eel and the French île, meaning island, are pronounced almost exactly alike, yet their meaning is so different that no sane person would regard them as sprung from the same origin. As a matter of fact île is derived from Latin insula, whereas eel has a cognate in German aal. These prototypes insula and aal being as different in sound as they are in meaning, any possibility that eel and île might be related is easily disposed of. Yet if the Latin and German equivalents were lost, if nothing were known of the history of the English and French languages, and if île meant not island but, say, fish or watersnake, then it might be reasonable to think of a connection.

Such doubtful cases, of which a certain proportion are likely to be adjudged wrongly, are bound to come up in regard to the less investigated languages, particularly those of nations without writing, the earlier stages of whose speech have perished without trace. In proportion as more is known of a language, or as careful analysis can reconstruct more of its past stages, the number of such borderline cases obviously becomes fewer.

Before genetic connection between two languages can be thought of, the number of their words similar in sound and sense must be reasonably large. An isolated handful of resemblances obviously are either importations—loan words—or the result of coincidence. Thus in the native Californian language known as Yuki, ko means go, and kom means come. Yet examination of Yuki reveals no further instances of the same kind. It would therefore be absurd to dream of a connection: one swallow does not make a summer. This lone pair of resemblances means nothing except that the mathematical law of probability has operated. Among the thousands of words in one language, a number are likely to be similar in sound to words of another language; and of this number again a small fraction, perhaps one or two or five in all, will happen to bear some resemblance in meaning also. In short, the similarities upon which a verdict of genetic relationship is based must be sufficiently numerous to fall well beyond possibility of mere coincidence; and it must also be possible to prove with reasonable certainty that they are not the result of one language borrowing words from another, as, for instance, English borrowed from French and Latin.

At the same time it is not necessary that the similarities extend to the point of identity. In fact, too close a resemblance between part of the stock of two languages immediately raises a presumption of borrowing. For every language is continually changing, and once a mother tongue has split into several branches, each of these goes on modifying its sounds, and gradually shifting the meaning of its words, generation after generation. In short, where connection is real, it must be veiled by a certain degree of distortion.

Take the English word foot and the Latin word of the same meaning, pes. To offhand inspection the sounds or forms of the two words do not seem similar. The resemblance becomes more definite in other forms of pes, for instance the genitive case ped-is or the accusative ped-em. Obviously the stem or elementary portion of the Latin word is not pes but ped-; and the d is closer to the English t of foot than is the s of pes. The probability of relationship is increased by the Greek word for foot, pous, whose stem proves to be pod-, with vowel closer to that of English. Meanwhile, it would be recognized that there are English words beginning with ped-, such as pedal, pedestrian, pedestal, all of which have a clear association with the idea of foot. All these words however possess almost exact equivalents in Latin. One would therefore be justified in concluding from these facts what indeed the history of the languages proves, namely, that pedal, pedestrian, and pedestal are Latin words taken over into English; whereas foot and pes and pous, and for that matter German fuss, are derivatives from a common form which once existed in the now extinct mother tongue from which Greek and Latin and English and German are derived.