¹ I am deeply indebted to Dr. and Mrs. Chaim Shatan, Dr. Dorothy Kipnis, and Mrs. Harold L. Williams, parents of the children, for their careful record keeping and continuous cooperation over a long period. The investigation was supported in part by USPHS small research grant M-5116(A) from the National Institute of Mental Health, Public Health Service.

² D. McCarthy, ‘Language development in children’, in Manual of child psychology (ed. L. Carmichael; New York, 1954), reviews a number of studies which employ traditional grammatical categories instead of treating the child's speech as sui generis. Of greater interest are a number of studies of children's knowledge of English inflexional rules, reviewed by J. Berko and R. W. Brown, ‘Psycholinguistic research methods’, in Handbook of research methods in child development (ed. P. H. Mussen; New York, 1960); but these studies use subjects at least three or four years old, who have already mastered much of English phrase structure. Some corpora of utterances by children who seem somewhat more advanced than those reported here have been recently discussed by R. W. Brown and C. Fraser, ‘The acquisition of syntax’, in Verbal learning and verbal behavior (ed. C. N. Cofer; New York, 1962), and also by W. Miller and S. Ervin, ‘The development of grammar in child language’, in The acquisition of language, Child development monographs (ed. U. Bel-lugi and R. W. Brown; in press). Miller and Ervin's paper came to my attention after this article was written; they note some of the same phenomena discussed here. Their term ‘operator’ appears to correspond to my term ‘pivot’.

³ No tape recordings were made of Gregory and Andrew in the early months because of the uneconomically large number of hours spent in recording and listening that would have been required to obtain the small number of word combinations at the child's command. A few tape recordings were made in the fifth month, and after the sixth month a high proportion of each corpus was tape-recorded, the written record being eventually abandoned as no longer practicable. This later material is not discussed here. Comparison of the written record with tape recordings made at the same age revealed that constructions present in the one were always present in the other, and with about equal frequency, a fact which is evidence of the reliability of the written records.

⁴ The procedure described leads to a lexical and not a phonemic representation of morphemes. With very young children there are reasons for not making a phonemic analysis preparatory to an investigation at the morphological and syntactic levels. For example, it is not easy to use the 21-month-old as an informant who gives same-different judgments to questions. Moreover, the contrasts that exist are not constant over time. The strong likelihood that a child may hear contrasts which he cannot produce not only means that partially separate analyses of receptive and productive functions would be necessary, but may also lead to a special situation in which phonemic strings would frequently have to be inferred from lexical ones.

⁵ Some of the pivots (e.g. more) occur as single-word utterances, others (e.g. all) do not. Those that occur alone seem also to occur as X-words in word combinations (e.g. no more, more down there); they can therefore properly be regarded as belonging to both classes.

⁶ In deciding whether utterances containing infrequently occurring words were pivotal constructions, the writer was sometimes guided by the child's subsequent development. For example, although mail occurs three times in first position (mail man, mail car, mail come), it is not classified as a pivot because there are no further cases of mail in first position in subsequent weeks; on the other hand, although come only recurs twice in final position, the subsequent uses of it in final position suggest that it is a pivot. Similarly, in Gregory's corpus (Table 1), although more and pretty only occur twice, they are regarded as pivots because they recur frequently in initial position before X-words in the ne: t month of Gregory's development.

⁷ A. A. Hill, ‘Grammaticality’, Word 17.1–10 (1961); H. Maclay and M. Sleator, ‘Responses to language: Judgments of grammaticalness’, HAL 26.275–82 (1960).

⁸ By the end of the second month the recorded vocabulary was sufficiently large that the parents had difficulty in distinguishing which single-word utterances were new, and they recorded many fewer single-word utterances in the third and fourth months than in the second. However, during the first four months 150–200 words were recorded as having been uttered five times or more. There were probably many more less frequently occurring words which were not recorded. Even if the usual rank-frequency relation (Zipf's law, rf = K, in which the constant K approximates the total vocabulary, since it is the rank of the least frequently occurring word) fits the single-word utterances poorly, the total vocabulary must be very substantially greater than that recorded.

⁹ The next phase of development in these children will be fully discussed in a subsequent article. See also the longer discussions of children they have followed by Brown and Fraser, op.cit., and Miller and Ervin, op.cit.

¹⁰ They may form constructional homonyms with pivotal constructions, e.g. bàby chair (‘little chair’: pivotal construction), usually without an intonation break, and báby #cháir (‘the baby is in the chair‘). The two constructions, however, certainly cannot always be distinguished from their intonation alone (# is used here and in the text as a general juncture symbol).

¹¹ Actually there are two parts to this assertion : (a) that the children learn the positions of certain words, and (b) that a sentence typically has just two positions—first and last. But (b) needs no discussion when utterances are just two words long. In experiments in children's learning of simple artificial languages, reported elsewhere (Braine, On learning the grammatical order of words', Psychological review, in press) learning positions is explored when ‘position’ in a sentence is defined by successive fractionations, i.e. when a sentence has just two positions, each of which may be occupied by a word or a phrase, and when each constituent phrase in turn may be divided into two positions.

¹² J. Berko, ‘The child's learning of English morphology’, Word 14.150–77 (1958).

¹³ For example, in the context (—) boat, Gregory places byebye, pretty, nightnight, and big. These are all pivots. Since the evidence indicates that not any word can precede boat, but only a small set, there is no basis for assuming that the position of boat has been learned. The occurrence of the specific utterances byebye boat, pretty boat, etc., can better be explained by assuming that the positions of byebye, pretty, etc. have been learned.

¹⁴ For example, in the context (—) it, Gregory places do, push, close, buzz, move, and in the context P₁(—) he places such words as boy, dirty, plane, hurt, mommy, etc. If the occurrence of only verbs before it is not accidental, Gregory has not learned merely the position of it, but something more which limits its environment to English verbs.