|Home||Feature Inventory||Related Issues||Search||Contact|
Tracking the origins of the concept 'feature'
The development of the concept of 'feature' as the ultimate component of speech can be tracked back into antiquity, and the principle of distinctive functions of sounds has underlied most attempts to design and reform alphabets. At the end of the 19th century, the only vocabulary available for the discussion of the sounds of speech were the words used also to talk about writing, but soon the concept of the 'phoneme' was developed, and discussions of the phoneme led to the emergence of features. Following the definition of the phoneme in terms of distinctive function, the concept of the distinctive feature as a core phonological notion and the theory of distinctive features were pioneered by the Prague School in the first half of the 20th century. Since it was taken for granted by the majority of phonologists that any proposed feature system should have universal rather than language specific applicability, the goal of distinctive feature theory has been 'the attempt to develop a set of features adequate to the task of distinguishing and relating the elements of phonological systems across the range of possible human languages' (Brasington 1993:1042). Decades of research have demonstrated the usefulness of distinctive features in phonological analysis, as well as led to the discussion of properties of features and feature systems in general (such as binarity/n-arity, markedness, underspecification, the notion of default, the structuring of the feature systems). Despite this, current research in phonology calls into question the assumption that distinctive features are innate and describe natural classes (e.g. Mielke 2004).
The sections below give a critical account of the rise of features. For up-to-date discussion of features in phonology, please check the proceedings of the recently held international conference entitled 'Where do features come from? Phonological primitives in the brain, the mouth, and the ear' (University of Paris 3, 4-5 October 2007). For a critical analysis of the current situation in morphological and syntactic description and theorising, where features are ubiquitous and yet poorly understood, see Corbett (2006; 2010). The critique of the current situation, with indications of wide use and regular misunderstandings of features, forms a justification for several papers which may be found in the outputs of the 'Features Project' recently completed in the Surrey Morphology Group, University of Surrey.
The development of the concept of 'feature' as a unit of linguistic analysis can be tracked far back into antiquity. In their analyses of Sanskrit, ancient Indian grammarians used the concept of sphoṭa. This term referred to a linguistic unit (a sentence, a word group, a word, a combination of morphemes, or a sound) with respect to the semantic value which 'flows forth' from the form of the unit (Jakobson 1958/601:394-3952). Already by the time of Patañjali's treatise from 2nd century BC (see also Bhartṛhari's Vākyapadīya from around 5th century AD), the lowest level of sphoṭa had been identified as a 'letter-sound' varṇa sphoṭa, an invariable unit of distinctive sound capable of semantic differentiation. The varṇa sphoṭa itself was devoid of meaning, but if one was replaced by another, this led to the production of a different word. The constant entities of sphoṭa were strictly distinguished from dhvani, 'speech sounds' (or higher level events) actually realising the sphoṭa.3 This opposition is often considered to have foretold the dichotomies of modern structuralism, and in particular the modern concept of the phoneme (see e.g. Jakobson 1958/60:394-395; Robins 1967:140; Fischer-Jørgensen 1975:4).
Independently of the Indians ancient Greek grammarians developed the concept of a 'primary element', the stoicheion, used to designate both physical and linguistic elemental units. In language, stoicheia were the ultimate components of speech, the indivisible sound units capable of forming larger units (syllables) which, in turn, formed meaningful strings (words, then sentences). In Jakobson's view, Plato's discussion of language can be understood to mean that 'language imposes upon the infinitely divisible physical continuum of the gross sound matter a coherent system that contains a limited number of discrete formal units with definite interconnections' (Jakobson 1958/60:395). Similarly, medieval philosophers were also aware of the distinction between sounds and 'sign-vehicles'. 'Thomas Aquinas treated speech sounds as "primarily designed to convey meaning" (principialiter data ad significandum), but having no meaning in themselves. He regarded this use of sounds as a human artifact (significantia artificialiter)' (Jakobson 1958/60:395-396).
1 In references to Jakobson's works, the date before
the forward slash refers to the year of writing while the date after the
forward slash refers to the year of publication. This procedure has been
adopted after Ivić (1965).
The development of alphabetic writing is naturally linked to the principle of distinctive function due to which separate letters are proposed only for those sounds which have a distinctive function and not for sound variants which are determined by their environment. The principle of distinctiveness also plays a role whenever an alphabet is transferred from one language to another. The earliest existing record of a proposal to reform the orthography of a language (Icelandic) to make it suit better the phonemic inventory of that language is the so-called First Grammatical Treatise written in Iceland by an anonymous author in the middle of the 12th century (edited several times, e.g. Haugen 1972). The Treatise is arguably 'the best phonological description of any language before the early 1930s' (Fischer-Jørgensen 1975:5-6). There is no doubt that the author's reasoning is based on distinctive function. He uses the commutation test ('minimal pairs' of words) to demonstrate which distinctions of sound are accompanied by distinctions of meaning and argues that sound variants conditioned by the environment can be written with the same letter.
The distinctive function of sounds is also used in phonological descriptions of standard languages by 'classical' phoneticians such as Jost Winteler [1846-1929], Henry Sweet [1845-1912], Paul Passy [1859-1940] or Otto Jespersen [1860-1943]. Winteler (1876) also uses a commutation test and makes a distinction between 'essential' and 'accessory' features of sounds. Sweet (1877) argues that in phonetic transcription separate symbols are needed only for the distinctive differences in each language, i.e. 'those broader distinctions of sound which actually correspond to distinctions of meaning in a given language'. Passy (1888) was influenced by Sweet and gives very similar formulations, e.g. '[t]here should be a separate letter for each distinctive sound; that is for each sound which being used instead of another can change the meaning of a word'. Jespersen (1897-1899:509, 611ff) holds a similar, functional view of phonetics and, furthermore, foretells the distinction between phoneme and allophone.
Fischer-Jørgensen (1975:7-8) points out that the pupils of 'classical' phoneticians, when confronted with the work of the Prague phonologists, maintained that these theories were not as new as their authors considered them. However,
'[t]he recognition of the functional point of view as an important one did not in classical phonetics result in the construction of any new theory or in any clearly formulated principles of description, and it was mainly in the description of the standard languages that the classical phoneticians, partly bound by orthography, restricted themselves to sound differences with distinctive functions. In the dialect descriptions of the same period most authors enumerated a large number of sound nuances without distinguishing the functions of these nuances.'
At the beginning of the 20th century, the only vocabulary available for the discussion of the sounds of speech were the words used also to talk about writing. In German, Sprachlaut, 'a speech sound', had come into use only in the 1840s. In English, a speech sound had sometimes been used from around the same time, but in order to avoid any misunderstanding the American linguist William Dwight Whitney still resorted to the awkward phrase letter of the spoken alphabet in the 1860s (Hockett 1995:6). The French equivalent was son du langage.
The term phonème was coined in 1873, by a French phonetician A. Dufriche-Desgenettes who used it in a paper on nasal consonants presented at meeting of the Société de linguistique de Paris. The new word first appeared in print in the same year in a brief anonymous report on the meeting, and was taken up in 1874 by the eminent Romance philologist Louis Havet, with the following comment (Mugdan 1985:140, also his translation): 'Phoneme, a term I borrow from Mr Dufriche-Desgenettes (...), designates any articulated sound [son articulé], whether vowel or consonant' (1874:321n).
In 1878 Ferdinand de Saussure [1857-1913] used the term phonème in his Mémoire sur le système primitif des voyelles dans les langues indo-européennes, but without giving its definition, and apparently 'only to render a strictly historical concept' (Jakobson 1958/60:397), namely 'to designate a common prototype in a parent language which is reflected by different sounds in the languages derived from this parent language' (Fischer-Jørgensen 1975:8-9). For this reason, some linguists interpret Saussure's phonème as no longer a 'speech-sound', but '[a]n element of a phonological system where, regardless of its exact articulation, it is recognized as being different from any other element' (Godel 1957:272; also Jakobson 1958/60:397). However, Saussure appears to be inconsistent and the terms phonème and, for example, voyelle, are used more or less interchangeably in the Mémoire (Mugdan 1985:139).
In 1880 Mikołaj Kruszewski [1850-1887], a student of Jan Baudouin de Courtenay [1845-1929] at the University of Kazan', for the first time used the new term to contrast it with sound: 'This word can be used to advantage as a term to designate the phonetic unit, while the word "sound" could designate the unit in the so-called physiology of sound' (1880:36; transl. Mugdan 1985:138). Although Kruszewski did not mention the source of the term, Baudouin pointed out that phoneme was 'a term borrowed from de Saussure, who, however, used it in a different sense' (Baudouin 1881:339; transl. Mugdan 1985:139). In 1881, in a German version of the introduction to his master's thesis, Kruszewski added a slight clarification to his definition: 'I suggest calling the phonetic unit (i.e. that which is phonetically indivisible) by the name phoneme, as opposed to sound, the anthropophonic unit' (Kruszewski 1881:14; transl. Mugdan 1985:137). Despite this definition, however, Kruszewski's examples show that at that time he used the term mostly to designate units that enter into (mostly diachronic or genetic) correspondences and correlations, regardless of whether the unit is a single sound or not, which was, in fact, not radically different from Saussure (Mugdan 1985:139).
Baudouin quickly adopted Kruszewski's suggestions regarding the phoneme and used the term for 'that which is indivisible from the point of view of the comparability of the phonetic parts of a word' (Baudouin 1881:333; transl. Mugdan 1985:140), and later coined the term morpheme (a unit indivisible from the morphological point of view) modelling it on phoneme. After various worthwhile attempts to broaden the definition of the phoneme to cover both its diachronic and synchronic applications, in his later linguistic work Baudouin apparently reverted to the earlier equation 'phonemes, or sounds', and finally decided to abandon the criterion of interlingual correspondences and intralingual correlations by defining the phoneme, in 1895, as a psychological unit: a 'psychological [mental] equivalent of the speech sound' (Jakobson 1958/60:420; Fischer-Jørgensen 1975:9; Mugdan 1985:143).
Baudouin's attempts to define phonemes as 'abstractions' or 'the results of generalization, devoid of positively given properties of actual implementation or existence' (1881:71, transl. Jakobson 1958/60:411) arose from his understanding of a dichotomy which he had already talked about in his inaugural lecture at the University of St. Petersburg in 1870. Namely, according to Jakobson,
'[Baudouin] (...) paid particular attention to the importance of distinguishing the two aspects of language which are interrelated and which imply one another. The first of them, or "language as a definite complex of certain constituents and categories, which exists only in potentia", he called simply 'language' (jazyk). The second aspect, "language as a continually repeatable", received the name of 'speech' (reč')' (Jakobson 1958/60:411).Furthermore, at the same lecture, Baudouin set up three separate tasks for phonetics:
'(1) the description of sounds from a physiological point of view, (2) the description of the role played by sounds in the mechanism of language (also described as their significance for the linguistic intuition of the speech community), and (3) the description of sound change. By the "role in the mechanism of language" he meant, above all, the role of sounds in morphological alternations' (Fischer-Jørgensen 1975:8).
Saussure, who lectured in Paris in 1881-91, met Baudouin in Paris on several occasions in 1881-82, corresponded with both Baudouin and Kruszewski, possessed a number of Kruszewski's works, and believed that the two Polish scholars should be on the list of those few names 'that should be cited'4 when discussing cardinal contributions to the theory of language. By the time of his lectures in Geneva (1907-1911) he had already adopted the distinction between langue and parole, and argued for a number of other dichotomies between different aspects of language. The dichotomies can be traced back to various other scholars besides Baudouin and Kruszewski (e.g. Wilhelm von Humboldt [1767-1835]; Georg von der Gabelentz [1840-93]; the neogrammarian Hermann Paul [1846-1921]; the American linguist William Dwight Whitney [1827-1894]), but the original way in which Saussure combined them into a coherent and new theory started a new era in European linguistics. Saussure's lectures were reconstructed (from his students' notes) and published under the title Cours de linguistique générale in 1916 (Fischer-Jørgensen 1975:11-13).
As for Saussure's use of the phonème at this time, it was, apparently, not very consistent and played a negligible role in his lectures (Mugdan 1985:140). He confined himself only to scattered examples and did not go into details concerning either phonemes or phonemic systems. However, it is generally considered that his language theory underlies the phonemic theories of the subsequent period (Fischer-Jørgensen 1975:16). It is, therefore, not surprising that Ščerba, a student of Baudouin, wrote with regret that 'some of our linguists are ready to ascribe to Saussure, in some degree or other, even the doctrine of the phoneme' instead of giving due credit to Baudouin and his school (1945).
4 Jakobson 1958/60:420-421, based on 'Notes inédites de F. Saussure', Cahiers Ferdinand de Saussure XII (1954), p. 66; and also Godel (1957:51): 'Baudouin de Courtenay and Kruszewski were closer than anyone else to a theoretical view of language, without digressing from purely linguistic considerations; yet they are unknown to most Western scholars'.
Baudouin's later, psychological definition of the phoneme as a sound image was problematic and ultimately led to an entirely different analysis along his previous line of thought, in which a given sound always belongs to the same 'phoneme' irrespective of morphological considerations.
According to Jakobson:
'Baudouin, who from his first note "on the use of Latin alphabet in the area of the Slavic languages" (1865) had constantly studied linguistic problems of writing and spelling, worked during the years of his Petersburg professorship on the relationship between the Russian language and its orthography, to which problem he later devoted his book, Ob otnošenii [russkogo pis'ma k russkomu jazyku (St. Petersburg, 1912)]' (1958/60:425, ft. 123).In 1900 he began an 'extensive correspondence' with Sweet. In 1908-1909 his student Lev V. Ščerba [1880-1944] visited Paris, came in contact with the leaders of the International Phonetic Association, and
'made it clear to them that the concept of a "distinctive sound" [see §2 above] required a deeper methodological substantiation, as we find it outlined in Baudouin's quest for the essence of the phoneme. Subsequently, under the influence of Passy, whose lectures Ščerba attended at the Sorbonne, he introduced the distinctive function of the phoneme into the very definition of this concept (...)' (Jakobson 1958/60:425).
In 1911, in a short description of Russian sounds (Court exposé de la prononciation russe) in a supplement to Le maître phonétique published by the International Phonetic Association, Ščerba explained that in his table of Russian sounds he presented 'sounds that have significative value (i.e. phonemes in the terminology of Baudouin de Courtenay)5 in bold type and nuances that have no significative value in ordinary type' (1911:2, transl. Jakobson 1958/60:425 ft. 124). In 1912, in the introduction to his book Russkie glasnye v kačestvennom i količestvennom otnošenii,
'Ščerba refers to the Exposé de principes de l'Association Phonétique Internationale (1908), written by Passy, "one of the few phoneticians who have completely understood the simple idea about the necessity for distinguishing 'les elements significatifs d'une langue' from sounds that 'n'ont aucune valeur distinctive'". He quotes Passy's view that if two sounds "do not play the role of semantic constituents", then "there is no difference between them from the linguistic point of view" (p.10)' (Jakobson 1958/60:426, ft. 124).
Ščerba's 'conclusive definition of the phoneme' (which is how he himself referred to it in 1912:14) is: 'the shortest general phonetic representation in a given language that is capable of associating with representations of meaning and of differentiating words and that can be emphasized in speech without distorting the phonetic structure of the word'6 (transl. Mugdan 1985:146). Thus, for the first time, the phoneme was defined as a distinctive unit, and Ščerba also pointed at the relevance of the phoneme to the phonetic constitution of the word.7
According to Jakobson, Ščerba's conversations with Antoine Meillet [1866-1936] (Saussure's student) in Paris 'strengthened the latter's conviction that the phoneme was necessary for linguistic operations' (Jakobson 1958/60:425-426). It is also noteworthy that Ščerba's ideas exerted profound influence on Daniel Jones [1881-1967]. Jones had studied French phonetics with Passy in 1905-1906, in 1911 read Ščerba's paper and immediately realised the full scope and far reaching importance of the phoneme concept (Fischer-Jørgensen 1975:50; Jakobson 1958/60:426). In 1913 Baudouin's theory was explained to him in more detail by Titus Benni, another student of Baudouin. Soon afterwards Jones incorporated these ideas into teaching and, in 1957, in his book on The History and Meaning of the Term "Phoneme" he wrote: 'The immense importance of the theory then became very clear to me, especially its relation to the construction of phonetic transcriptions, to the devising of alphabets for languages hitherto unwritten or unsuitably written, and in general to the practical teaching of foreign spoken languages' (p. 6).
5 Jakobson (1958/60:420) believes that even in
Baudouin's later writings there were to be found 'individual amazing flashes
of insight', and that Baudouin's best students (notably Ščerba)
'inspired by the fresh cultural currents of the new century, managed to
extract this core from the superfluous chaff and to find an empirical
application for their teacher's phonological inklings.' 'L.
Ščerba declared: "Baudouin himself emphasized that he had recourse
to psychological labels because of the impossibility of using any other
terminology, given the contemporary state of the science. (...) It seems
to me that Baudouin's psychologism can easily be removed from his
linguistic theories and everything will remain in place", (Izvestija
po russkomu jazyku i slovesnosti AN SSSR, III, p. 315)'
(1958/60:420, ft. 98).
Due to Baudouin and Ščerba there was an unbroken phonological tradition in Russia from the beginning of the twentieth century, which was one of the conditions for the development of a structural linguistic school in Prague. In Prague, the native Czech tradition (the philosopher T.G. Masaryk; Vilem Mathesius [1882-1945]) was complemented by the influences from the east (Baudouin and Ščerba; the Fortunatov School in Moscow) and west (Baudouin had met Saussure in Paris, and from about 1900 he corresponded with Sweet;8 Ščerba studied with Passy in Paris in 1908-1909), and its scholarship was 'decisively intensified' by the arrival of three young Russian linguists: Roman Jakobson [1896-1982], who came to Czechoslovakia in 1920, Nikolai S. Trubetzkoy9 [1890-1938], who after fleeing the revolution in Russia was appointed professor in Vienna in 1922, and Sergei I. Karcevskij [1887-1955], who had studied in Geneva, lived in Prague in the mid twenties, and went back to Geneva in 1928 as a university teacher.
In 1926 the Prague Linguistic Circle was founded and quickly started gaining influence all over Europe (Fischer-Jørgensen 1975:19, 102, 322). In October 1927, Jakobson formulated a set of theses to be presented at the First International Congress of Linguists in the Hague in April 1928. According to these theses, the aim of a phonological theory should be: (1) to set up phonological systems, (2) to account for the significant differences ('les différences significatives entre les images acoustico-motrices'), (3) to find correlations (e.g. contrasts like p/b, t/d, k/g), (4) to formulate general laws concerning the structure of phonological systems, and (5) to account for historical change in terms of a teleological development of the system (Fischer-Jørgensen 1975:20). The theses were countersigned by Karcevskij and Trubetzkoy and were the first explicit statement of the objectives of the Prague School. They also contain the first definition of the term 'phonological correlation' in terms of a proportional bilateral opposition: 'a phonological correlation consists of a series of binary oppositions defined by a common principle which can be thought of independently of each couple of opposite terms',10 even though after 1930 Jakobson did not use the term 'correlation' any more (while retaining the idea of binarity), and Trubetzkoy, who retained it, used it in a slightly different sense (Ivić 1965:41).
Neither Jakobson nor Trubetzkoy had published anything pertaining specifically to phonological theory before the Proposition au Premier Congrès International de Linguistes of 1928. 'When Trubetzkoy, whose outlook had till then been mainly historical, was stirred by these new ideas, he systematically began to elaborate them in a long series of articles which appeared in quick succession (Fischer-Jørgensen 1975:20). Starting from the conception of the phonological system 'as a lawful structural whole' (Jakobson 1932), the members of the Prague School emphasised the distinctive function of the phoneme, undertook the analysis and classification of phonological oppositions and applied phonology also to diachronic problems. Trubetzkoy's Grundzüge der Phonologie, published in 1939 after his death, offers a systematization of sounds according to distinctive oppositions and can be seen as 'the chief statement of Prague phonology' (Fischer-Jørgensen 1975:21) in its application of the 'structural method' to the science of languages.11
As for the connection between European and American linguistics, on the other hand, it was initially very loose. It appears that most American linguists 'preferred to start from scratch, unhampered by the European tradition' (Fischer-Jørgensen 1975:65). At the time he was writing Language (1921), Edward Sapir [1884-1939] had probably not been well acquainted with Saussure's theory, though by 1925 his own phonological theory was already much more elaborate than Saussure's (e.g. he defined the phoneme as 'a functionally significant unit in the rigidly defined pattern or configuration of sounds peculiar to a language', or as a 'functionally significant point in a complex system of relatedness').
Leonard Bloomfield [1887-1949], on the other hand, had studied in Leipzig. Apart from being familiar with the works of Passy and Sweet (and, according to Hockett 1995:7, having been influenced greatly by Sweet), he was acquainted with Saussure's Cours as well as with the Travaux du Cercle Linguistique de Prague (published from 1929), but he did not draw a clear distinction between 'language' and 'speech', was not clear about 'meaning' as opposed to 'form' ('expression'), and considered language as a social system to be the subject of linguistics. His account of phonology was very brief and, although from 1925/26 he gave definitions of the phoneme similar to those of the Prague phonologists (i.e. based on distinctive phonetic features),12 he offered no detailed procedure by which the phonemes of a language may be found. In his Language (1933) he defines phonology (or, 'practical phonetics') as 'the study of significant speech sounds' (p. 78) and phonemes as 'the smallest units which make a difference in meaning' (p. 136). On p. 77ff he explains the difference between distinctive and non-distinctive phonetic features and states that '[t]hese distinctive features occur in lumps or bundles, each one of which we call a phoneme. The speaker has been trained to make sound-producing movements in such a way that the phoneme-features will be present in the sound-waves, and he has been trained to respond only to these features and to ignore the rest of the gross acoustic mass that reaches his ears'13 (p. 79). For Bloomfield, then, the phoneme was a physical unit, that part of a sound which recurs in all its occurrences and which distinguishes it from all others (Fischer-Jørgensen 1975:65-68, 75).
On closer reading of Bloomfield it is clear that he had not anticipated Jakobson's proposition that phonemes are made up of even smaller components (§7). For Bloomfield, a phoneme is a distinctive feature of a sound:
'Further experiment fails to reveal any more replaceable parts in the word pin: we conclude that the distinctive features of this word are three indivisible units. Each of these units occurs also in other combinations, but cannot be further analyzed by partial resemblances: each of the three is a minimum unit of distinctive sound-feature, a phoneme. Thus we say that the word pin consists of three phonemes (...)' (1933:79).
'Among the gross acoustic features of any utterance, then, certain ones are distinctive, recurring in recognizable and relatively constant shape in successive utterances' (1933:79).
'The phonemes of a language are not sounds, but merely features of sound which the speakers have been trained to produce and recognize in the current of actual speech-sound' (1933:80).
'The important thing about language (...) is not the way it sounds (...) [but it] is its service in connecting the speaker's stimulus (...) with the hearer's response (...). This connection depends (...) upon only a relatively few features of the acoustic form, upon the features which we call phonemes' (1933:128).Even in the often quoted sentence '[t]hese distinctive features occur in lumps or bundles, each one of which we call a phoneme' (1933:79), in the light of Bloomfield's other explanations, the word 'phoneme' must be understood to refer to 'these distinctive features', and not to 'lumps or bundles'.
Other 'American structuralists', or (loosely) 'post-Bloomfieldians', who contributed to the further development of phonological theory included Morris Swadesh [1909-1967], William Freeman Twaddell [1906-1982], Zellig S. Harris [1909-2002], and Charles F. Hockett [1916-2000]. Swadesh (1934, 1935, 1936) discussed the criteria for establishing phoneme inventories and the importance of pattern congruity; Twaddell's monograph On Defining the Phoneme (1935) contains a discussion of the definitions of the phoneme given by the Prague School, Jones, and Bloomfield, as well as his own, new definition of phonemes as 'abstractional, fictitious units': microphonemes (characterised by minimal phonological differences) and macrophonemes (sums of all microphonemes of similar minimum phonological differences among forms); Harris's conception of phonemes is similar to Twaddell's in that he regards them as 'purely logical symbols, upon which various operations of mathematical logic can be performed' (1951:18) (Fischer-Jørgensen 1975:71-77).
Except for Harris and Hockett, Bloomfield and most of his successors considered the phoneme the minimal unit and were not interested in the analysis of distinctive features, or in the arrangements of systems according to distinctive features. However, Harris (1944, 1945, and 1951:125ff) analyses strings of phonemes into so-called long components (i.e. certain features such as voicing or nasality), but does not regard them as belonging to a new analytical level, but as characterising phonemes (Fischer-Jørgensen 1975:71-102). Furthermore, according to Fischer-Jørgensen (1975:101), Hockett, who was clearly influenced by the Prague School, entertained the idea of decomposing phonemes into distinctive features as early as in (1942):14
'he mentions that it is possible to regard unvoicing as a suprasegmental feature in a cluster like pt in English. In his description of Sierra Populuca (1947) he attempts a feature decomposition of the phonemes and writes these features in columns. Labial closure is marked with p, oral articulation with O, unvoicing with H, voicing with V, and nasality with N. (...) Hockett compares this transcription to musical notation.'
8 See Jakobson (1958/60:425). According to
Fischer-Jørgensen, Baudouin corresponded with Passy, she does not
mention Sweet - but she may have simply made a mistake here substituting
one name for the other.
Although the idea of features as components, rather than properties (dimensions), of phonemes may not yet have been clear in Jakobson's mind, his intuition about features occurring simultaneously in 'bundles' and, thus, making up phonemes, occurred in print as early as 1932, in an entry in a Czech encyclopedia: '[P]honeme is (...) a set of those concurrent sound properties which are used in a given language to distinguish words of unlike meaning'. This appears to be the first definition of the phoneme as a bundle of distinctive features. 'It is only a very short step from this formulation to the conception of the phoneme as consisting of components such as voicing, labiality, etc., whereby these components, and not the phonemes, become the smallest phonological units' (Fischer-Jørgensen 1975:145).
Jakobson's subsequent publications on phonology until his escape to America:15 Observations sur le classement phonologique des consonnes (1938/39), the review of van Wijk's Phonologie (1939), and Kindersprache, Aphasie und allgemeine Lautgesetze (1939-1941/41), provide the first analyses of distinctive properties of phonemes and thus begin 'the second epoch in the development of Jakobson's views on phonology, characterized by a consistent effort to define each phoneme in any linguistic pattern by an exhaustive set of distinctive features' (Ivič 1965:53).
Although his early concept of 'correlation' implied the concept of distinctive features, Jakobson did not use the term 'correlation' (or the concept of the 'archiphoneme', tied to the concept of 'correlation' by its definition) after 1930, but instead replaced it with the universalised conception of distinctive feature (Ivič 1965:41, 43). With Observations, he marked 'a departure from the common Prague patrimony' which was not shared by Trubetzkoy whose tragic death in 1938 prevented him from formulating his position.16 'In establishing his theory Jakobson [had] disproved de Saussure's tenet of the caractère linéare du signifiant: the distinctive features belong to the SIGNIFIANT, but they occur simultaneously' (Ivič 1965:61-62; also Fischer-Jørgensen 1975:146). Furthermore, Jakobson's insistence on binarity transformed the amorphous gradual relationships on a linear axis of possible articulation places of consonants into three dimensions: three binary contrasts. He subsequently worked on the distinctive characteristics of vowels, then glides, and finally, in 1949, he performed the first systematic componential analysis of all phonemes in a language (Standard Serbocroatian) (see §8 below).
15 After Kindersprache (1941) he did not
publish any work in theoretical phonology until 1949; cf. Jakobson (1951).
In 1941 Jakobson fled from Sweden to the United States. His first two papers on phonology, both written and published in 1949 (one co-authored with John Lotz), 'mark (...) a transition to the third epoch in Jakobson's development as a phonologist' (Ivić 1965:67), as they launch the first distinctive feature matrices (Jakobson 1949 for Serbocroatian, and Jakobson & Lotz 1949 for French). The paper entitled On the identification of phonemic entities, which was the first to introduce the tabulation of the distinctive features, did not yet distinguish between cells with minuses and those where the feature is redundant (Ivić 1965:36, ft. 1). Some features are represented as ternary, with the sign +/- as the third possibility, which pointed to the conflict between the binary character of distinctive features and the economy of their number. The diagrams are not free of infelicities, but they expressed the assertion by Jakobson that '[distinctive feature] analysis is the only way to distinguish between phonemes on strictly linguistic criteria, since it selects, from the mass of phonetic facts, those which are linguistically relevant' (Ivić 1965:66-67).
Both 1949 papers contain paragraphs on theoretical aspects of the problem. On the identification refers once again to Saussure: (p. 420 in Selected Writings I) 'This arbitrary thesis [of postulating a linear character of the signifier] prevented both its author and the phonemicists following in his tracks from solving the pivotal problem which he himself astutely foresaw: the task of "determining the distinctive features (éléments différentiels) of the phonemes". The group relations were still examined only in terms of sequences and not at all in terms of bundles'; and in another place (p. 421 in Selected Writings I): 'Only when brought up to the level of distinctive features, does the linguistic analysis enable us to verify Saussure's cardinal statement on phonemic units as first and foremost "entités oppositives". The phoneme by itself is not a term of opposition.' Furthermore, there are references to Twaddell's microphonemes and macrophonemes and, crucially, a clearly negative reference to Bloomfield (p. 425 in Selected Writings I):
'Only in resolving the phonemes into their constituents and in identifying the ultimate entities obtained does phonemics arrive at its basic concept (which insures the consistent use of linguistic criteria sought by Hjelmslev) and thereby definitely breaks with the extrinsic picture of speech vividly summarized by L. Bloomfield: a continuum which can be viewed as consisting of any desired, and, through still finer analysis, an infinitely increasable number of successive parts [Language (New York, 1933), p. 76.]. Linguistic analysis, with its concept of ultimate phonemic entities, signally converges with modern physics, which has revealed the granular structure of matter as composed of elementary particles.'
Jakobson's 1951 paper For the correct presentation of phonemic problems contains 'a new account of the basic postulates of the binaristic doctrine, with a hint that they could be confirmed also by spectrograms - which was done in the Preliminaries' (Ivić 1965:68), as well as a claim that Jakobson had viewed the phoneme as a 'bundle of distinctive features' since 1932.
In 1951/52 Jakobson, Gunnar Fant and Morris Halle produce Preliminaries to Speech Analysis: The Distinctive Features and their Correlates, the first full-scale account of the acoustic properties of the distinctive features which also contains some novel solutions of particular problems in componential analysis (Ivić 1965:68, 71). According to the authors (1952:1), it was an attempt 'to resolve speech into ultimate units' with the use of distinctive features. A further paper by Jakobson, Colin Cherry and Morris Halle from 1952/53 (Toward the logical description of languages in their phonemic aspect) dispenses with the +/- sign.17 The ground was prepared for the discussion that followed of the big issues regarding the properties of features and feature systems in general, such as binarity/n-arity, markedness, underspecification, the notion of default, and the structuring of the feature systems.
17 Fischer-Jørgensen (1975:101) notes that in his Manual of Phonology (1955:126ff) Hockett 'gives examples of a feature analysis which bears a strong resemblance to that of Roman Jakobson (...) and which is arranged in a matrix with pluses and minuses. However, the features which Hockett sets up are more traditionally phonetic.'
Primary sources and historical references
Secondary sources and contemporary references
How to cite this entry: