¹ Where it is necessary to distinguish between morphemes recognized without applying the method here proposed, and the morphemes which result from the application of this method, the former will be called continuous morphemes, and the latter new.

² Leonard Bloomfield, Language.

³ Read: minus /æ/ plus /e/. In the pair man: men, it is of course arbitrary to consider men rather than man as containing two morphemes. We might have said that man consists of men plus a morpheme indicating ‘singular’. We might also have said that man and men are two unrelated single morphemes. The choice among these ways of analyzing man and men depends on the relation of these morphemes to other morphemes in the language, and to the utterances in which they occur.

⁴ Braces { } will be used to indicate morphemic units.

⁵ See Z. S. Harris, Morpheme Alternants in Linguistic Analysis, Lang. 18.169–80 (1942), where it is shown that such groupings of morphemes into one morphemic unit can be performed, without arbitrariness or resort to meaning, on the basis of distributional criteria.

⁶ More rigorous criteria, with less reliance upon meaning, can be stated, but are not necessary for the present purpose.

⁷ Stanley Newman, Yokuts Language of California 120 (1944).

⁸ Leonard Bloomfield, Language 180. The late Manuel J. Andrade showed me in 1940 a similar case in his Guatemalan material.

⁹ These morphemes can of course be broken down into case, number, and (usually) gender morphemes, but that is not relevant here.

¹⁰ E.g. it occurs also in τ — άδ∊λφ—, as in των άδ∊λφων ‘of the brothers’.

¹¹ Or /et/ after the present tense morpheme.

¹² I am indebted to Nathan Glazer for obtaining these utterances from our informant, and for his valuable collaboration in the Swahili research of the Intensive Language Program of the American Council of Learned Societies.

¹³ The phonetic value of a space is to indicate loud stress on the second vowel before it. The letters may be pronounced with their usual values for an approximation to Swahili sounds.

¹⁴ When an object of the verb is stated, its class prefix may also be repeated before the verb, after the subject class prefix.

¹⁵ In the case of ulaa, it is u (‘class 6‘). The noun laa occurs with the class prefix u. The sequence a-ulaa occurs with the class prefix of the noun which precedes it.

¹⁶ When a noun is followed by an adjective plus an adjectival noun phrase (i.e. a noun phrase preceded by a-), the adjectival noun phrase comes last. Hence there is no confusion as to whether an adjective refers directly to the head noun or to the noun that is preceded by a- (even if both are of the same class): if the adjective follows the a- noun, it refers to it.

¹⁷ Allowing for variations which depend on the particular demonstrative, etc.

¹⁸ These few discontinuous morphemes may indeed be considered to be the ultimate noun class of the language, since the nouns may be considered to modify these class prefixes in somewhat the same way that the adjectives modify the nouns. The formal relation, however, is not the same, since nouns are limited to particular class prefixes whereas adjectives are not limited to particular nouns.

¹⁹ I am indebted for this suggestion to Freeman Twaddell, and have also profited froma discussion of the question with Bernard Bloch.

²⁰ The ‘person class’ plural wa would have to be broken into two morphemes, one a plural having the same domain as the class prefixes, the other a variant of the' person class' prefix which occurs only in the environment of the new plural morpheme.

²¹ Since the domain of the class prefix now becomes the domain of the noun, or rather of its first phoneme or two.

²² X and Y would traditionally be described as being in agreement. The present method replaces this agreement relation by a single morpheme Z. The method can clearly be extended so as to replace other types of grammatical agreement by single morphemes, but the results would not be as simple as in this limited type.

²³ The fact that we have to state the environment in order to know the form of our discontinuous morpheme in each particular occurrence, means that the environment functions as an independent factor in determining the variant of our new morpheme. Therefore it is possible to generalize the present method, and to take any two morphemes, and treat any factor which determines their coexistence as part of their environment. However, it is descriptively advantageous to do so only if the environment or the variation is of a general character, or correlates with other features of the language.

²⁴ With a by-product of such forlorn morphemes as -ide, -ick, etc.