^{page 556 note 1} ‘ Some Prosodic Aspects of Retroflexion and Aspiration in Sanskrit ’, BS0AS., 13, 1951, pp. 939 ff.Google Scholar

^{page 556 note 2} ‘ Structure ’ and ‘ system ’ are here distinguished, on the lines suggested by Prof. J. R. Firth, as referring to two complementary dimensions of a framework of linguistic description. At the phonological level of statement structure may be stated in general terms of the elements C/V and π; and values are provided for these elements by the establishment of systems of phonematic and prosodic units (also called prosodies) respectively. Such systems are justified by their applicability to the elements of structure and are exhaustive of the data selected for notice. The terms of a system, i.e. (at the phonological level) the phonematic or prosodic units, are said to commute with one another within that system.

Prosodic units may further be described as properties of the structures or sub-structures to which the systems within which they commute are relevant.

Robins, R. H. cf, TPS., 1953, p. 109Google Scholar, n. 2 and further refs.

^{page 556 note 3} cf.Arch. Ling., 3, 1951, pp. 135 f.Google Scholar; TPS., 1953, pp. 84 f., 97 ff.

^{page 556 note 4} From the foregoing note it will be understood that prosodic systems form only part of a descriptive framework in which phonematic systems are also established; ‘ prosodic statement ’ is here used as an abbreviation denoting a phonological statement which, unlike a phonemic statement, is characterized by the distinction of prosodic and phonematic systems. On this distinction see further Firth, , ‘ Sounds and Prosodies ’, TPS., 1948, pp. 127 ff.Google Scholar (esp. p. 150), and Lingua, 1, 1948, p. 403Google Scholar (‘ The Semantics of Linguistic Science ’).

^{page 556 note 5} cf. Bazell, C. E., ‘The Correspondence Fallacy in Structural Linguistics’ (Studies by Members of the English Dept., Istanbul Univ., vol. 3, 1952), pp. 1 f.Google Scholar

^{page 556 note 6} Whilst the group to which the writer belongs is characterized by a theoretical as opposed to a procedural approach, it adopts the principle that all theory must be based upon experience, and that connexion must be renewable in experience, cf. the remarks of Hjelmslev, , Prolegomena to a Theory of Language, pp. 8 ff.Google Scholar, and AL., 4, p. 145; Benveniste, , Conf. de l'Inst. de Ling, de l'Univ. de Paris, 11, 1952 3, p. 47.Google Scholar

^{page 557 note 1} cf. ‘ Prosodic Aspects … ’, p. 943, n. 1.

^{page 557 note 2} The following typographical conventions are employed:—

(a) Phonological statements: i. Prosodic units: Italic.

ii. Phonematic units: Clarendon.

(b) Phonemic and orthographic transcriptions: Italic.

(c) Phonetic transcriptions (in I.P.A. terms): Clarendon, in brackets.

^{page 557 note 3} Traditional phonetic classification does not always provide categories appropriate to the requirements of phonological statement (Firth, cf, Lingua, 1, p. 401Google Scholar, and BSOAS., 12, 1948, p. 864)Google Scholar. Until a more flexible system is established, some degree of arbitrary labelling is inevitable, and the terms here employed are not to be considered as necessarily related to or as prejudicing any such system. The interrelationships of certain categories of the present statement may be shown as follows:—

The scale of ‘ obstruence ’ provides a measure of the distinction between ‘ occlusion ’ and ‘ constriction ’ (viz. maximal: non-maximal oral obstruence) and between the sub-categories within each of these.

In case it should subsequently prove desirable to discuss as a class all articulations other than non-nasal occlusion, the term ‘ continuance ’ is available.

The traditional category of ‘ liquid ’ is here provisionally accepted without phonetic criticism or rejustification, and simply on the basis of its phonological usefulness in the present statement (Jakobson, cf, Preliminaries to Speech Analysis, p. 19Google Scholar; de Saussure, Cours, p. 74).

‘ Apicality ’ refers to tongue-tip (‘ apex ’) articulation; for the system within which this term commutes in the statement see table on p. 561 below (cf. also Firth, , BSOAS., 12, pp. 861, 863).Google Scholar

^{page 557 note 4} See preceding note.

^{page 558 note 1} See e.g. Harris, Z., Methods in Structural Linguistics, pp. 133, 143 ff.Google Scholar; Jakobson, op. cit.

^{page 558 note 2} The phonic data are not to be considered as ‘ brute facts ’ of utterance. In that they can enter into a linguistic description only by being named, they are abstractions in terms of a theory of naming; and the theory appropriate to the naming of such data is a theory of phonetic description, Firth, cf, Lingua, 1, p. 403Google Scholar; and on the general principle Popper, K., Economica, 12, 1945 (‘ The Poverty of Historicism ’), pp. 75, 79CrossRefGoogle Scholar: ‘ Science … cannot start with observations, or with the “ collection of data ” … at no stage of scientific development do we begin without something in the nature of a theory … which in some way guides our observations.’

The relevant phonic data may be said to be allotted to the abstracted phonological categories, i.e. to the terms of prosodic or phonematic systems; and the categories may be said to have such data as their exponents.

^{page 558 note 3} Sinclair, W. A. cf, The Conditions of Knowing, p. 35.Google Scholar

^{page 558 note 4} See e.g. Trubetzkoy, , Grundzüge der Phonologie, p. 67.Google Scholar

^{page 558 note 5} The traditional Roman orthographic forms are to be considered simply as typographically convenient transliterations of the Devanagari spelling, and so as substituting for the latter in its function of labelling and identifying utterances.

^{page 558 note 6} See ‘ Prosodic Aspects … ’, p. 940, with n. 7.

^{page 559 note 1} ibid., p. 943.

^{page 559 note 2} In the latter case a ‘ renewal of connexion ’ would probably require to notice retroflexion as a coarticulation only with the exponents of certain pre-focal units in addition to the focal unit.

^{page 559 note 3} Based on the figures in Whitney, ‘ On the Comparative Frequency of Occurrence of the Alphabetic Elements in Sanskrit ’, JAOS., 10, pp. 151 fGoogle Scholar; cf. Skt. Gr., § 75.

On the phonetic terminology see p. 557, n. 3 above.

^{page 559 note 4} For details see Whitney, loe. cit.

^{page 559 note 5} Bloch, J. cf, L'Indo-Aryen, pp. 53 ff.Google Scholar

^{page 559 note 6} See Allen, , Phonetics in Ancient India, pp. 53 ff.Google Scholar

^{page 560 note 1} Skt. Gr., § 46. For the principles involved cf. Aasta Stene, English Loanwords in Modern Norwegian; Henderson, E. J. A., ‘ The Phonology of Loanwords in some South-east Asian Languages ’, TPS., 1951, pp. 131 ff.Google Scholar

^{page 560 note 2} pp. 945 f.

^{page 560 note 3} op. cit.

^{page 560 note 4} op. cit., pp. 143 ff.

^{page 561 note 1} cf. n. 7 below.

^{page 561 note 2} p. 941, n. 3.

^{page 561 note 3} The question of retroflex articulation after k is excluded from the present discussion: cf. Phonetics in Ancient India, p. 79.

^{page 561 note 4} The most notable exception is provided by the sequence -aṣṭ-, which, however, is almost entirely restricted to the exponents of certain verbal categories, where it is in paradigmatic relationship with -aśV- or -ajV-. In all such cases a grammatical structure Root + Suffix is involved, and the retroflexion is to be stated not as an exponent of a property of structure of the exponents of one or the other grammatical category, but as an exponent of their junction. There are also cases of -iṣṭ-, -uṣṭ-, -ṛṣṭ- in paradigmatic relationship with -iśV-, -uśV-, -ͬśV- as well as with -iṣV-, -uṣV-, -ͬṣV-; R would be establishable in the latter cases but not in the former.

Apart from verbals, an exception is provided by the numeral aṣṭau; the category of numerals also provides a rare example of initial ṣ in ṣaṭ.

Other exceptions are sufficiently infrequent to require exclusion on the grounds stated above (item 3).

^{page 561 note 5} Firth, cf, TPS., 1948, p. 145.Google Scholar

^{page 561 note 6} cf. Phonetics in Ancient India, pp. 12 f., 62 ff.

^{page 561 note 7} Corresponding to the ‘ extension ’ of R in the previous article (pp. 941 ff.). It is possible that, given more precise phonic data, it might prove possible to equate this extension with the ‘ word ’ (cf. remarks at end of item 4 above), i.e. to establish R as a word-prosody.

^{page 562 note 1} For R only a single diacritic (a) would be establishable, and as forming no system is not to be stated.

^{page 562 note 2} See Firth, , TPS., 1948, p. 151Google Scholar: ' The suggested approach will not make phonological problems appear easier or oversimplify them. It may make the highly complex patterns of language clearer both in descriptive and historical linguistics ’. cf. Bazell, op. cit., pp. 33, 41 and n.: ‘ … the syntax of a language is far more complex than the syntax of algebra.… There are two principles of description, simplicity and adequacy … one cannot have the most of both at the same time.’

^{page 562 note 3} My italics (e.g. Pike, Phonemics: a Technique for Reducing Languages to Writing).

^{page 562 note 4} Introduction ⁸, p. 96.

^{page 563 note 1} In Indo-Iranian reflex-1 also appears after i < IE *ә.

^{page 563 note 2} See TPS., 1953, pp. 87 ff.

^{page 563 note 3} See Entwistle, , TPS., 1944, pp. 33 f.Google Scholar; Martinet, , Word, 7, 1951, pp. 91 ff.CrossRefGoogle Scholar

^{page 563 note 4} ‘ The Language of the Ashkun Kafirs ’, NTS., 2, 1929, pp. 192 ff.Google Scholar

^{page 563 note 5} e.g. an affricate reflex of the IE voiceless ‘ palatal ’ series (see Morgenstierne, , ‘ IE ḱ in Kafiri ’, NTS., 13, 1945, pp. 225 ff.Google Scholar; Bloch, op. cit., p. 51).

^{page 563 note 6} ‘ Ashkun Kafirs ’, pp. 199 f.

^{page 563 note 7} See e.g. Malmberg, B., Études sur la phonétique de l'espagnol parlé en Argentine, p. 160Google Scholar: ‘ … l'impression acoustique est très différent entre les s de visto, de hasta et de busco … une palatale après i, e, une postpalatale après a, et une vélaire après u, o.’ Essen, O. van cf, ‘ Zur Phonetik des ostwallonischen h ’, Arch. f. vgl. Phon., 7, pp. 58 ff.Google Scholar; Broch, , Slavische Phonetik, pp. 222 f.Google Scholar

^{page 563 note 8} ‘ Ashkun Kafirs ’, p. 200.

^{page 564 note 1} cf. Arch. Ling., 3, pp. 132 ff.Google Scholar

^{page 564 note 2} TPS., 1944, p. 33.

^{page 564 note 3} p. 941, n. 3.

^{page 564 note 4} ‘ Concerning some Slavic and Aryan Reflexes of IE ṡ ’ Word, 7, pp. 91 ff.CrossRefGoogle Scholar

^{page 564 note 5} cf. Av./Slav. is, us, rs < *iḱ, uḱ, rḱ; and the Skt. junction replacing Ved.

^{page 564 note 6} As a result of -ast- < *-ast- beside -ašt- < *-aḱt-, whereby s and š are established in identical environments, and š is identified with a ‘ high variant of s ’. In a phonemic statement, as Martinet says (p. 93),‘ this limited merger was sufficient to secure the phonemicization of the high variant ’. See, however, p. 561, n. 4, above on the Skt. reflexes of *-aḱt-.

^{page 564 note 7} cf. for a Norwegian parallel Vogt, H., NTS., 12, 1942, pp. 22 ff.Google Scholar

^{page 564 note 8} Morgenstierne, cf, NTS., 12, pp. 44, 57 (‘ Orthography and Sound-system of the Avesta ’).Google Scholar

^{page 565 note 1} Jakobson, cf, Remarques sur l'évolution phonologique du Russe (TCLP., 2), pp. 19 ff., 60 ff.Google Scholar

^{page 565 note 2} cf. Jakobson, op. cit., p. 62.

^{page 565 note 3} See TPS., 1953, pp. 90 ff. Gardiner, cf, The Nature of Historical Explanation, p. 3Google Scholar (‘ … the notions of past and future … do not enter into functional explanation, as it occurs in the advanced sciences, at all; for there considerations of time-order are superseded by considerations relating solely to structural order ’).