§ 1. The purpose of the following account is to show that the morphological rule for pluralization of nouns in Brazilian Portuguese can be simply expressed given (a) a basic representation for morphs which might be said to be ‘independently justified’ by alternations elsewhere in the morphology of the language, and (b) a set of phonological rules operating on and modifying the basic representation, the majority of which again seem to be widely applicable in Portuguese morphophonology.

One of the perennial problems in diachronic linguistics is how to reconcile, on the one hand, the Neogrammarian postulate of sound laws operating without exception, and, on the other hand, the embarrassingly numerous irregularities we observe in many languages. On most occasions linguists have attempted to solve the problem by positing interdialectal borrowing or analogical levelling and have largely overlooked the possibility of the gradual diffusion of phonological changes across the lexicon. As a result of the lexical gradualness of sound changes, exceptions may be created either through the incompletion of a sound change, or owing to the conflict of two sound changes overlapping along the time dimension. It is the latter concept that we will attempt to elaborate and illustrate with two sets of data, both from Peking dialect. We have chosen Chinese as a case study for an obvious reason: it is possible in the case of Chinese, like few other cases, to follow sound changes step by step through the phonological dictionaries, rhyme charts and other records compiled at various stages of history. The columns on Tables 1 and 2 contain information taken from the various datable phonological dictionaries.

I. The facts about the behaviour of fricatives in Old English in relation to surrounding segments are well known; but these facts (and some others) illustrate some interesting general principles that seem not to have been thoroughly discussed. A classic formulation of the data is that of Sweet (1953: 3):

… f and s, in addition to their modern values, could represent respectively the sounds of v and z, letters which were not normally used in O.E. These three letters, f, s, p, had the sounds of f, s, and th in thin (‘breathed’ or ‘voiceless’) initially and finally in accented words; next to ‘voiceless’ consonants (such as p, t); and when double; … They had the sounds of v, z, and th in then (‘voiced’) when single between vowels, or between a vowel and another ‘voiced’ sound (such as l, r, m, n)….

To generalize this, we might say that in Old English fricatives were voiced between sonorants, and voiceless if geminate or contiguous to a major boundary, or in clusters with voiceless obstruents (of which as we will see geminates are a special case). One way of capturing these facts would be a rule which states simply that fricatives are voiced between sonorants, and voiceless elsewhere.

During the past few years linguists have begun to express doubts about the validity of segmental analysis for the study of child language and have come to feel that it is probable that a child perceives spoken language differently from an adult, e.g. Bellugi & Brown (1964: 113), Ingram (1966: 218), Ladefoged (1967: 148–9), Lenneberg (1967: 279–281), Weir (1962: 30). They are also beginning to realize the importance of treating a child's language as having its own independent system (Carroll, 1961: 332; Fry, 1966: 194). To date, however, investigators of child speech have made their analyses on a segmental and distributional basis and have expressed the child's phonological system in terms of the adult's phonemic system; one may cite for example Cohen (1969), Grégoire (1933, 1947), Jakobson (1941), Jakobson & Halle (1961), Leopold (1939, 1947, 1961), Lewis (1968), Ohnesorg (1959), Velten (1943) and Weir (1962). Such studies, valuable though they be, cannot be said to have succeeded in explaining the relationship between the forms and structures of child and adult and have left many questions unanswered such as: (1) What governs the choice of sound that the child will use as a ‘substitute’ for an adult sound? (2) Why does the child ‘drop’ certain sounds of the adult form or ‘substitute’ for them when he is already capable of making such sounds and is in factusing them in some other contexts? (3) Why does the child use homonyms for adult forms which appear to be quite unlike each other and which have been proved to be semantically clearly differentiated for the child? (4) What governs the form that reduplicated words take?

In this paper I present a stylistic analysis of a short discourse by Bertrand Russell. My purpose is twofold: first, to suggest an approach to syntactically based stylistic analysis that goes beyond mere frequency counts, and, second, to draw out some linguistic ramifications of the approach.

1. It may be claimed that current views regarding the nature of sound change fall into two broad categories: the more traditional attitude would treat an individual sound change as a complex trend or process taking perhaps several generations to establish itself, and then retaining its activity over a long period of time; whence the characteristic concern of classical historical linguistics with the establishment of absolute and relative termini post and ante quern, i More recently adherents of the generative–transformational school have interpreted sound changes as readjustments in the system of phonological rules; thus Postal (1968: 270) claims:

‘What really changes is not sounds but grammars. And grammars are abstract objects – sets of rules represented in human organisms.’

In this paper some data from Russian are presented and it is suggested, contrary to the proposals put forward by Lakoff & Peters (1966), that in the deep structure of Russian two types of coordination must be recognized, an ‘and’ type and a ‘with’ type. It is also suggested that the underlying ‘with’ coordinating element may sometimes be realized as i (‘and’) or as s (‘with’), and that the s may derive from an underlying ‘comitative’ case (cf. Fillmore, 1968: 81–83).

What has been described as ‘one of the most remarkable (facts) in quantitative linguistics’ is the constancy of the logarithmic type/token ratio. If V denotes vocabulary and N text length, then log V/log N ‘remains sensibly constant for samples of different size from a given literary text’.1

This note is concerned with historical sound change expressed in rules added to a previous set of rules that develop a language from an earlier to a later form. I am not prepared at this time to explore in full the relationship between rules of sound change and rules of generative phonology, although it is obvious that this relationship exists in Maltese, as in many other languages. Because my considerations are primarily historical ones, perhaps ‘law’ would be a better term than ‘rule’, in spite of its being hopelessly out of date. However, since the term ‘rule’ and the format in which the information is presented are familiar and useful, I will use them here.