The best known theory of neutralization and the archiphoneme is the one progressively developed by the Russian N. S. Trubetzkoy during the inter-war period of the Prague Linguistic Circle. Although he lived in Vienna rather than Prague itself where Prague School flourished, his contribution to phonology was immense and his colossal correspondence with R. Jakobson eloquently shows us how the two Russian giants brought forward in a spirit of co-operation various aspects of this theory based on the principles of functionalism and structuralism.

In the domain of phonology, among other concepts whose genesis is attributed to the interwar Prague School (1926-1939),¹ the concept of ‘neutralization’ is unequivocally congruous with the basic principles of functional and structural linguistics. So are also the concept of ‘opposition’ (which logically precedes that of ‘neutralization’) and that of ‘phoneme’ (which itself is logically preceded by that of ‘opposition’).

The concept of ‘neutralization’ was adumbrated in 1929, in the following lines by Mathesius (1929: 81).

Les deux langues [Czech and German] concordent en ce que la différence dans la sonorité des consonnes n’a pas dans ces langues de valeur phonologique à la fin des mots. Ceci signifie qu’à la fin du mot dans l’une et l’autre langues, on trouve confondus en un phonème unique les couples de consonnes b/p, d/t, v/f, z/s, h/x, et en outre en tchèque d’/t’, ž/š, en allemand g/k.

Mathesius’s expression ‘un phonème unique’ foreshadows what will at a later date be known as ‘an archiphoneme’ which is intimately linked to neutralization.

The term ‘archiphonème’ (Archiphonem. Archifonéma. Архифонема)² is defined in ‘Projet de terminologie phonologique standardisée’ [hereafter ‘Projet’] (1931: 315) as follows.

Élément commun de deux ou plusieurs phonèmes corrélatifs, qu’on peut concevoir abstraction faite des propriétés de corrélation. [Ex: L’a latin abstraction faite de la longueur et de la brièveté (ɑ̄/ɑ).]

The term ‘propriété de corrélation’ (Korrelationseigenschaft. Korelační vlastnost) is defined as follows (‘Projet’: 313).

Opposition de la présence et de l’absence d’un certain caractère phonique qui différencie plusieurs couples d’unités phonologiques et qui, dans le système phonologique donné, peut être conçue abstraction faite de couples particuliers en opposition.

It may be noted that ‘Projet’ was largely drafted by Jakobson (cf Jakobson (1985) = Jakobson (2006)): Lettre 64, Note 3, where we read ‘... « Projet de terminologie phonologique standardisée » élaboré dans ses grandes lignes par RJ’.

The definition of ‘archiphoneme’ seen above is inadequate because it recognizes only such archiphonemes associated with ‘correlative oppositions’ (e.g. /p/ vs /b/ in English, /t/ vs /t’/, /d/ vs /d’/ and /t/ vs /t’/ vs /d/ vs /d’/ in Russian). These correlative oppositions can be associated with the archiphonemes /p-b/, /t-t’/, .d-d’. and /t-t’-d-d’/.³ The above definition of the term ‘archiphonème’ does not apply to what may be called ‘disjunct opposition’⁴ formed by ‘unités disjointes’ or ‘unités phonologiques appartenant à un système, sans former entre eux un couple de corrélation’.⁵ Examples of disjunct oppositions in English would be e.g. /m/ vs /n/, or /n/ vs /ŋ/, or /m/ vs /n/ vs /ŋ/ (my examples), none of which would be considered correlative pairs. Consequently, the archiphonemes /m-n/, /n-ŋ/ and /m-n-ŋ/ could not be recognized as archiphonemes⁶ as defined in ‘Projet’.

Whereas Trubetzkoy approved of ‘correlative opposition’, his reservations about ‘disjunct opposition’ and his disagreement with Jakobson on this subject ‘persisted even during their very last discussions that took place on 12 and 13 February 1938, just a few months before Trubetzkoy’s decease on June 25, aged 48. Trubetzkoy also remained unconvinced of Jakobson’s total binarism till his last days’ (Akamatsu 2013: 147-148).

The definition, if not the term, of ‘propriété de corrélation’ is interesting in that ‘un certain caractère phonique’ would correspond to ‘mark’, and opposition between ‘la présence du caractère phonique’ and ‘l’absence du caractère phonique’ ― which opposition is none other than ‘propriété de corrélation’ ― would give rise, when phonologically evaluated, to two opposing phonological values, i.e. two opposing relevant features. According to this interpretation of mine, ‘mark’ which corresponds to ‘caractère phonique’ is literally a phonic entity, not a phonological one, and therefore ‘mark’ should not be confused with ‘presence of the mark’ (‘mark’ ≠ ‘presence of the mark’). In this connection it would be useful to quote further the term and definition of ‘marque de corrélation’ found in ‘Projet’ (313) which runs as follows.

Marque de corrélation (Korrelationsmerkmal. Korelační přiznak. Корреляционноый признак) ‘Caractère phonique qui, opposé à l’absence de ce caractère, forme une propriété de corrélation.’ Exemple: La longueur des voyelles en latin ...

Vachek (1966a: 48) copies the definition from ‘Projet’ (133), and adds Trubetzkoy’s (1939: 77) definition of ‘Korrelationsmerkmal’⁷ via the French translation in Trubetzkoy (1949: 89) which runs as follows:

Une marque de corrélation est une particularité phonologique [my emphasis] par l’existence ou non-existence de laquelle une série de paires corrélatives est caractérisée (par ex. la nasalité vocalique qui, en français, différencie les paires corrélatives ɑn ― ɑ, on ― o, in ― e, un ― eu).

The example ‘nasalité (vocalique)’ given above appropriately explains what is ‘marque de corrélation’. Nasality in French vowels is a phonic feature (not a phonological feature). Presence of nasality in [ɑ̃] and its absence in [ɑ] gives rise to two phonological values, i.e. two relevant features, “nasal” and “non-nasal” attributable to /ɑ̃/ and /ɑ/, respectively. The same applies mutatis mutandis to other pertinent examples. What is said here is what I consider to be an important point to which I will return further below for discussion.

Trubetzkoy plentifully employs the concepts and terms ‘marked’ (‘merkmalhaltige’) and ‘unmarked’ (‘merkmallose’) in connection with the two phonemes of a correlative opposition and, of course, those of a neutralizable correlative opposition in particular. Such two terms are not defined in ‘Projet’ but they appear in part of the terms ‘série corrélative marquée’ (‘merkmalhaltige korrelative Reihe’) and ‘série corrélative non-marquée’ (‘merkmallose korrelative Reihe’) in ‘Projet’ (314). The distinction between ‘marked (phoneme)’ and ‘unmarked (phoneme)’ is important in Trubetzkoy’s characterization of the two correlative phonemes of a neutralizable opposition, particularly in connection with the archiphoneme of a correlative opposition (see infra Fig 3 and Fig 4).

To judge from the definition of ‘archiphonème’ in ‘Projet’ (315) cited above, it is apparent that the archiphoneme is conceived of as a global phonological entity which is not analyzable into smaller phonological entities. This view is supported when we additionally confirm that the phoneme is defined as follows in ‘Projet’ (311): ‘Unité phonologique non susceptible d’être dissociée en unités phonologiques plus petites et plus simples.’ For the concept of ‘archiphoneme’ as a sum of relevant features, one has first to look for e.g. Trubetzkoy (1939: 35) (= 1949: 40 = 1969: 36) who writes: “Man darf sagen, daß das Phonem die Gesamtheit der phonologisch relevanten Eigenschaften eines Lautgebildes ist [Trubetzkoy’s emphasis].”⁸ Only if the phoneme is conceived of as a distincitve unit that consists of a sum of relevant features can the archiphoneme be likewise conceived.

As for the term ‘archiphoneme’, this occurs in Trubetzkoy’s five letters sent to Jakobson in 1930, and another five letters subsequent to 1931. For details see infra fn 9.

The total absence, anywhere in 'Projet', of the term ‘neutralisation’ or any term that would correspond to it, let alone any related terms like ‘neutralisable’ and ‘position de neutralisation’, is both surprising and incomprehensible. It is not easy to pinpoint the date, either prior to or subsequent to 1931, at which the term ‘neutralisation’ may have been first launched by whichever member(s) of the Prague Linguistic Circle).⁹ Predictably, it attracted criticisms from various quarters. One well-known criticism came from Martinet (1936: 46, 47) who pointed out that Trubetzkoy (1932: 19) did mention the terms ‘Neutralisierung’ / ‘Aufhebung’ and also define the concept of ‘neutralisation’.¹⁰ Martinet (1936: 47) does not fail to point out that Trubetzkoy (1932: 19) furthermore admits of neutralization of disjunct oppositions (as well as that of correlative oppositions).

Trubezkoy’s liberal use of both the terms ‘neutralisation’ and ‘archiphonème’ in his correspondence with Jakobson, especially in 1933 onward, is not surprising in view of the publication of Trubetzkoy’s three articles, viz Trubetzkoy (1935¹¹, 1936a, 1936b) as well as Trubetzkoy (1933) in which the subject of neutralization and the archiphoneme receives extensive discussions.

Trubetzkoy (1933: 233) provides the following definition of ‘phoneme’.

un phonème est un élément différentiel, une valeur linguistique au sens de F. de Saussure [Trubetzkoy’s italics], il ne peut être défini que par ses rapports avec les autres phonèmes du même système¹².

This is a functionalist stance which I for one completely agree with.

We note that, shortly afterwards, in Trubetzkoy (1935: § 1) ¹³, the same functionalist definition of the phoneme is given but, surprisingly, not in Trubetzkoy (1936a: 6‑7)¹⁴.

In Trubetzkoy (1935: §§ 20-22), there appear the terms ‘aufgehoben’, ‘neutralisiert’ and ‘aufhebbare Gegensätze’. Strangely, the term ‘Archiphonem’ does not appear.¹⁵ As to what happens when a neutralizable opposition is actually neutralized in the position of neutralization, Trubetzkoy (1935: § 22) has the following to say.

Phonologische Gegensätze die in gewissen Stellungen aufgehoben werden, (“aufbebbare Gegensätze”) bekommen in den übrigen Stellungen (d. i. dort, wo sie nicht aufgehoben sind) einen spezifischen Charakter ... Und wenn der aufhebbare Gegensatz sich als Gegensatz zwichen “Vorhandensein” und “Nichtvorhandensein” eines Merkmals denken läßt, so werden die betreffenden Phoneme als das “merkmaltragende” und das “merkmallose” gewertet. Merkmallos ist dabei immer dasjenige Phonem, welche bei der phonologischen Aufhebung des betreffenden Gegensatzes als einziger Vertreter des entsprechenden Phomempaares auftritt ...

What is characteristic throughout in all his writings on ‘neutralization’, Trubetzkoy’s concept of ‘neutralizable opposition’ (nearly always between two, not more, phonemes) is that the two phonemes of the neutralizable opposition are the marked and the unmarked, respectively. Furthermore, the unmarked phoneme is that phoneme of the two which occurs in the position of neutralization in its capacity as what Trubetzkoy calls ‘Archiphonemvertreter’ (archiphoneme representative). According to Trubetzkoy, the archiphoneme itself does NOT occur in the position of neutralization. This is, as I have always argued in my relevant writings, very controversial.

One may wonder what prompts Trubetzkoy to conceive of the concept of ‘archiphoneme representative’. Is it fundamentally due to his preoccupation with the distinction between ‘marked phoneme’ and ‘unmarked phoneme’ and, further, because the phonological content of the unmarked phoneme is equivalent to that of the archiphoneme? I shall have further occasion to mention this point later and suggest some other factor which has more to do with the status that Trubetzkoy accords to the archiphoneme itself.

In Trubetzkoy (1936a) we find the term ‘opposition supprimable’ (‘opposition neutralisable’) and related terms. In Trubetzkoy (1936a: 13), the term ‘archi-phonème’ (‘archiphoneme’) and ‘opposition supprimable’ itself occur where there is a difference between, on the one hand, one of the phonemes of a neutralizable opposition in the position of relevance, and on the other, the unmarked phoneme (not the archiphoneme) which is said to occur in the position of neutralization. What is interesting is that Trubetzkoy refers implicitly to the concept of ‘phoneme’ which differs from the one we already saw in ‘Projet’ (311). The keyword here is ‘traits’. Trubetzkoy (loc cit), refers to the archiphoneme by the expression ‘un phonème’ and says ‘... un phonème dont le contenu phonologique se réduit aux traits communs aux deux termes de l’opposition donnée.’ In other words, the phoneme, and consequently the archiphoneme as well, are presented as NOT being global phonological entities which are unanalyzable into smaller phonological elements. The definition of ‘phoneme’ which is implicit in that of ‘archiphoneme’ here departs from that earlier given in ‘Projet’ (311).

Trubetzkoy (1936a: 13) explains first what the phonological content of each of the two terms (i.e. the two phonemes) of a neutralizable opposition in the position of relevance are like and then what the phonological content of that phoneme which occurs in the position of neutralization is like.¹⁶ He continues to say: ‘les termes d’une opposition supprimable se décomposent phonologiquement en « archiphonème + qualité spécifique ». As he refers to “les termes”, not ‘un des termes’, I understand that ‘qualité spécifique’ is possessed by each of the two phonemes, consequently two mutually different ‘qualités spécifiques’ are involved. The two phonemes (/A/ and /B/) of a neutralizable opposition (/A/ vs /B/) in the position of relevance are then understood to be as shown below.

Fig 1

/A/ = archiphonème + qualité spécifique 1 /B/ = archiphonème + qualité spécifique 2

Trubetzkoy (1936a: 15) explains what happens in the position of neutralization to the two phonemes of a neutralizable opposition in terms of their phonological contents.

... dans les oppositions [rechte positions] où le terme d’une opposition supprimable est le seul admis, il est identifié avec l’archiphonème de cette opposition ... ses qualités spécifiques perdent leur valeur phonologique ... l’autre terme de la même opposition apparaît comme le seul porteur d’une qualité qui manque à son « partenaire » ...

The main difference between the passage quoted further above from Trubetzkoy (1935: 20) and the passage just quoted from Trubetzkoy (1936a: 15) is that the former includes clear references to the marked phoneme and the unmarked phoneme whereas this is not the case in the latter and also that the former does not mention ‘archiphoneme’ while the latter does. The terms ‘marqué and ‘non-marqué’ appeared earlier in ‘Projet’ (314) as part of the terms ‘Série corrélative marquée’ and ‘Série corrélative non-marquée’, respectively. The German terms ‘Merkmaltragend’ and ‘Merkmallos’ (which correspond to ‘marqué’ and ‘non-marqué’ in French, and ‘marked’ and ‘unmarked’ in English, for example) which Trubetzkoy (1936b: 34 et passim) in his account of ‘neutralization’ employs are very important since they clearly show Trubetzkoy’s concepts of ‘neutralization’ and ‘archiphoneme’ (which I do not share, as will be shown further below).

What happens in the position of relevance is only implicit in Trubetzkoy’s words but need be clarified. Using the wording found in Trubetzkoy (1936a: 13), we can show what happens to /A/ and /B/ in the position of relevance as follows.

In the position of relevance

Fig 1

/A/ = archiphonème + qualité spécifique 1 /B/ = archiphonème + qualité spécifique 2

In the position of neutralization

Fig 2

/A/ = archiphonème + zero (i.e. loss of qualité spécifique 1) /B/ = archiphonème + qualité spécifique 2

Trubetzkoy (1936b: 33-34) writes about the two phonemes of a neutralizable correlative opposition as neutralization takes place.

... erscheint in der Aufhebungsstellung eines den Oppositionsgliedern ... eines von den Oppositionsgliedern in dieser Stellung [i.e. in the position of neutralization] als Stellvertreter des entsprechenden Archiphonems auftritt ... werden seine spezifischen Züge irrelevant, während die spezifischen Züge seines Partners volle phonologische Relevanz bekommen: somit wird das erste Oppositionsglied als »Archiphonem + Null«, das zweite dagegen als »Archiphonem + ein bestimmes Merkmal« gewertet.

The passage quoted just above is the one that subsequently recurs in Trubetzkoy (1939: 73) except that the word ‘bekommen’ is replaced by ‘erhalten’.¹⁷ Thus, Trubetzkoy (1939: 73) reads as follows.

... eines von den Oppositionsgliedern in dieser Stellung als Stellvertreter des entsprechenden Archiphonems auftritt, werden seine spezifischen Züge irrelevant, während die spezifischen Züge seines Partners volle phonologische Relevanz erhalthalten: somit wird das erste Oppositionsglied als “Archiphonem + Null”, das zweite dagegen als “Archiphonem + ein bestimmtes Merkmal” gewertet.

I believe we can understand what has just been quoted as the definitive statement by Trubetzkoy about what happens to the two terms of a neutralizable opposition in the position of neutralization (and, paradoxically, as we shall see later, what the two terms should be in the position of relevance).

It can safely be assumed that, in the passages quoted above, Trubetzkoy has pairs of correlative phonemes in mind. In other words, /A/ vs /B/ is a neutralizable phonological opposition whose two terms are a pair of correlative phonemes.

Trubetzkoy says that, in the position of neutralization, one of the two phonemes of a neutralizable opposition occurs. I do not understand this, and anyway I disagree. Trubetzkoy does not explain why and why only one of the two phonemes ― but which one? (determined by what criterion?) ― of the phonemes of a neutralizable correlative opposition occurs in the position of neutralization. I personally find it incomprehensible that one of the two phonemes in question occurs in the position of neutralization at all. Trubetzkoy does not explain why this phoneme, once it occurs in the position of neutralization, loses the validity of its mark. It is my view that when a neutralizable opposition is neutralized (dissolved) in the position of neutralization, neither one nor the other phoneme of the opposition occurs in that position. I will say that, on the other hand, both the marked phoneme and the unmarked phoneme occur in the position of relevance. Trubetzkoy says that one of the phonemes (the one that proves to be the unmarked phoneme) of a neutralizable opposition occurs to the exclusion of the other phoneme (the marked phoneme) in the position of neutralization.

I suspect that, for example, in the neutralization of /t/ vs /d/ in German which takes place moneme-finally and elsewhere, Trubetzkoy would choose /t/, not /d/, as the phoneme that occurs in the position of neutralization. I surmise that the basis for this choice is the phonetic fact that, in the position of relevance, /t/ and /d/ are realized voiceless and voiced, respectively. Trubetzkoy’s intention is to seek the mark which is present in one of the phoneme, i.e. /d/, but absent in the other phoneme, i.e. /t/. Consequently voice is considered as the mark.

Trubetzkoy (1936b: 33-34; 1939: 73) shows /A/ and /B/ in the position of neutralization as follows.

Fig 3

/A/ = « Archiphonem + Null » /B/ = « Archiphonem + ein bestimmtes Merkmal »

‘Merkmal’ is short for ‘Korrelationsmerkmal’, i.e. ‘mark’. ‘Null’ means literally ‘absolutely nothing’, hence « Archiphonem + Null » = « Archiphonem ». The two terms of a neutralizable opposition may therefore be tentatively conceived of as follows, though I will argue later that this is highly problematic.

Fig 4

/A/ = archiphoneme + zero /B/ = archiphoneme + mark

There are major problems with what Trubetzkoy presents as ‘Merkmal’ and ‘Null’ (mark and zero) in the well-known formulas shown in Fig 3 and Fig 4. We need to scrutinize this formulas.

There are two serious issues I wish to critically discuss which arise from the formulas shown in Fig 3 and Fig 4. One of the issues concerns the notion of what Trubetzkoy calls ‘Archiphonemvertreter’ (archiphoneme representative). This problem will be discussed in V. THE ARCHIPHONEME REPRESENTATIVE. The other issue arises from the fact that the formulas in Fig 3 and Fig 4 which are supposed to pertain to the neutralization of /A/ vs /B/ actually pertains to the position of relevance of /A/ vs /B/. This problem will be discussed in IV which immediately follows. I will discuss them in turn.

1) The readers may be wondering why I have not properly defined the archiphoneme before. It is true that ‘archiphoneme’ was mentioned a number of times, but the main point in the previous sections concerned the question of the distinction between the marked and the unmarked, leaving the archiphoneme unaffected in one way or another, that is, irrespective of whether the archiphoneme is properly defined or not. However, in what follows, the proper definition of the archiphoneme is absolutely necessary.

Trubetzkoy (1936b: 32) defines the archiphoneme in the following words: ‘... wir unter Archiphonem die Gesamtheit der Züge verstehen, die zwei Phonemen gemein sind.’¹⁸ This definition of the archiphoneme recurs almost verbatim in Trubetzkoy (1939: 71) except that the word ‘Züge’ is amelioratively replaced by the wording ‘distinktiven Eigenschaften’ by the additional word distinktiven.¹⁹

2) Trubetzkoy (1936b: 33-34; 1939: 73) famously shows the two phonemes of a neutralizable correlative opposition in the position of neutralization as follows, as earlier shown.

Fig 3

/A/ = « Archiphonem + Null » /B/ = « Archiphonem + ein bestimmtes Merkmal »

In this formula, /B/ as a whole is a distinctive unit whose phonological content consists in a sum of relevant features. Consequently, ‘[ein bestimmtes] Merkmal’ should not be ‘marque de corrélation’ (‘Korrelationsmerkmal’, ‘mark of correlation’) because this would be a phonic feature. Instead, ‘ein bestimmtes Merkmal’ should be replaced by ‘presence of the mark’ (‘mark’ ≠ ‘presence of the mark) and ‘Null’ by ‘absence of the mark’, which may conveniently be symbolized as ‘+m’ and ‘-m’, respectively, where the two ‘m’s are one and the same mark, so that ‘presence of the mark’ and ‘absence of the mark’ are opposed to each other (cf ‘propriété de corrélation’, ‘Korrelationseigenschaft’, ‘quality of correlation’) and neither exists without the other. For example, if ‘mark’ is ‘voice’, ‘presence of the mark’ (i.e. ‘+m’) is the relevant feature “voiced” while ‘absence of the mark’ (i.e. ‘-mark’) is the other relevant feature, “voiceless”, so that “voiced” and “voiceless” are opposed to each other and neither is conceivable without the other. The idea comparable to ‘presence of the mark’ and ‘absence of the mark’, rather than just the mark, already appears in Trubetzkoy (1931: 97) where he writes:

Die zwei Glieder eines korrelativen Gegensatzes sind nicht gleichberechtigt: das eine Glied besitzt das betreffende Merkmal (oder besitzt es in seiner positiven Form), das andere besitzt es nicht (oder besitzt es in seiner negativen Form). [Emphasis for the whole quotation is Trubetzkoy’s, and the additional underlines are mine.]

3) Trubetzkoy attaches great importance to the type of opposition which he calls ‘privative opposition’ in Trubetzkoy (1939: 67 = 1949: 77 = 1969: 75) in the passage quoted below.

Das Oppositionsglied, das durch das Vorhandensein des Merkmals gekennzeichnet ist, heißt “merkmaltragend”, das durch das Fehlen des Merkmals gekennzeichnete Oppositionsglied “merkmallos” [Trubetzkoy’s emphasis]. Diese Art von Oppositionen ist für die Phonologie außerordentlich wichtig.

My understanding is that what Trubetzkoy calls ‘privative opposition’ is a phonic difference in which the mark is a phonic entity, where ‘the presence of the mark’ and ‘the absence of the mark’ when viewed in opposition to each other (cf ‘propriété de corrélation’) are endowed with phonological values, i.e. relevant features. What are presented as “stimmhaft” – “stimmlos”, “nasaliert” – “unnasaliert”, “gerundet” – “ungerundet” (Trubetzkoy 1939: 67), or “sonore” – “sourd”, “nasalisé” – “non nasalisé”, “arrondi” – “non arrondi” (Trubetzkoy 1949: 77), or “voiced”/“voiceless”, “nasalized”/“nonnasalized”, “rounded”/“unrounded” (Trubetzkoy 1969: 75), are in each case a pair of relevant features. What Trubetzkoy presents as a ‘privative opposition’ assumes its counterpart phonological opposition which is in fact a phonological opposition whose function is just like any other distinctive opposition, that is, to borrow Trubetzkoy’s word, an ‘equipollent’ phonological opposition. Thus, e.g. [p] vs [b] (privative (phonic) opposition) → /p/ vs /b/ (phonological opposition); there is no ‘privative’ relationship between /p/ vs /b/ because the phonological content of /p/ and that of /b/ consist of the same number of relevant features. The relevant feature “voiced” (in /b/) is not equal to the relevant feature “voiceless” + “?”.

4) Consequently, it is impossible for /A/ to possess ‘Null’. Instead of ‘Null’, /A/ should possess ‘absence of the ‘mark’ (i.e. ‘-m’). Therefore,

Fig 3

/A/ = « Archiphonem + Null » /B/ = « Archiphonem + ein bestimmtes Merkmal »

need to be replaced by

Fig 5

/A/ = ‘archiphoneme + absence of the mark’ /B/ = ‘archiphoneme + presence of the mark’

or

Fig 6

/A/ = ‘archiphoneme + –m’ /B/ = ‘archiphoneme + +m’

or

Fig 7

/A/ = ‘archiphoneme + relevant feature 1’ /B/ = ‘archiphoneme + relevant feature 2’

or

Fig 8

/A/ = ‘archiphoneme + “voiceless”’ /B/ = ‘archiphoneme + “voiced”’

What is variously indicated as ‘absence of the mark’ vs ‘presence of the mark (Fig 5), ‘‑m’ vs ‘+m’ (Fig 6), relevant feature 1 vs relevant feature 2 (Fig 7) and “voiceless” vs “voiced” (Fig 8) are different manifestations of ‘propriété de corrélation’. For the neutralization of /A/ vs /B/ to take place, ‘propriété de corrélation’ is annulled, leaving the archiphoneme alone valid which thus occurs in the position of neutralization.

5) In connection with the impossibility of ‘Null’ appearing in the formula shown in Fig 3, we need to recall that Trubetzkoy (1933: 238) himself correctly and emphatically says that ‘une qualité phonologique n’existe que comme terme d’une opposition phonologique’ and that Trubetzkoy (1939: 85) repeats this important functional principle by saying: ‘Es darf nicht vergessen werden, daß eine distinktive Eigenschaft nur als Glied einer distinktiven Opposition bestehen.²⁰ This same point is what I subsequently emphasized in Akamatsu (1988: 90-91) in terms of ‘the relevant feature and its inherent opposability (the functional principle of the relevant feature)’.

6) Since the archiphoneme itself and alone occurs in the position of neutralization, it makes no sense to say, as does Trubetzkoy, that the archiphoneme is represented by one of the two phonemes of the neutralizable opposition. All this means that the well-known and time-honoured formula proposed by Trubetzkoy (1936b: 33-34, 1939: 73), as shown in Fig 3, is flawed.

7) The most serious flaw is that neutralization of the opposition /A/ vs /B/ does not in reality take place in the formula shown in Fig 3. Trubetzkoy says that the archiphoneme is identified, but he does not say that the archiphoneme itself occurs in the position of neutralization.

8) In denying ‘mark (of correlation)’, I am equally denying the status of ‘phonological feature’ to ‘mark’. What, if it is not ‘phonological feature’, is the status of ‘mark’? In my view, ‘mark’ is simply a ‘phonic feature’. Tcheu (1969: 240, 241) writes as follows.

La marque est donc un caractère phonique qui oppose, par sa présence et son absence, deux phonèmes dont les traits pertinents sont par ailleurs identiques ... mais elle-même n’est qu’un caractère phonique particulier.

I have always been thoroughly in agreement with Tcheu on this point; see Akamatsu (1988: 408).

9) We need to pay particular attention to the definition of ‘propriété de corrélation’, which is ‘opposition de la présence et de l’absence d’un certain caractère phonique’ and opposition between ‘la présence du caractère phonique’ and ‘l’absence du caractère phonique’, which ultimately give rise to two opposing relevant features. We have already accepted that the mark (= correlative mark) is a phonic entity, not a phonological one.

10) Tcheu’s understanding as well as mine of the concept of ‘mark’ may be illustrated as follows.

Fig 9

	‘presence of mark’ (= relevant feature 1)
‘mark’ (phonic feature)
	‘absence of mark’ (= relevant feature 2)

If we take the example of ‘voice’ as a mark, its relation to the two relevant features, “voiced” and “voiceless”, can be shown below.

Fig 10

	presence of ‘voice’ = “voiced” (relevant feature 1) in e.g. /b/
‘mark’ (phonic feature)
	absence of ‘voice’ = “voiceless” (relevant feature 2) in e.g. /p/

11) It seems apposite to note here what Martinet (1964²: 4.15) = Martinet (2005³: 4.15) writes, which I will quote in full below.

Ce qu’on appelle absence de marque doit être ici compté comme une caractéristique positive puisque correspondant à un type articulatoire distinct: pour réaliser l’absence de voix, par exemple, il faut prendre garde à ne pas laisser le contexte imposer des vibrations glottales; l’absence de voix correspond à une organisation bien définie de la glotte.

What Martinet says so rightly here is directly relevant to Figs 7, 8, 9 & 10 (supra) and Figs 11, 12 & 13 (infra) while it is indirectly relevant to Figs 3, 4, 5 & 6 (supra). All this points to the functional principle whereby no relevant feature is conceivable without there being in the same phonological system another or other relevant features to which it is opposed; in other words, no relevant feature is opposed to zero. Inadmissibility of the formula shown in Fig 3 is obvious.

Likewise, if the mark is ‘nasality’, ‘presence of nasality’ will be phonologically conceived of as the relevant feature “nasal” while ‘absence of nasality’ will be phonologically conceived of as the relevant feature “non-nasal”. And “nasal” and “non-nasal” are opposed to each other, and neither relevant feature is conceivable without the other. In terms of articulation, ‘absence of nasality’ requires total velic closure (i.e. closure between the velum and the back wall of the pharynx) while ‘presence of nasality’ requires velic opening (i.e. the velum is held free from the back wall of the pharynx).

12) One should be cautious in correctly understanding a statement like the one quoted below made in Martinet (1960: III-15).

Deux séries comme /p f t s š k/ et /b v d z ž g/... forment ce qu’on appelle une corrélation ... Le trait pertinent [my emphasis] qui distingue les deux séries s’appelle la marque. Ici la marque est la « sonorité ».

In the above quoted statement, the ‘mark’ is said to be a relevant feature (‘trait pertinent’), which is then identified as « sonorité ». It is obvious that the ‘mark’ is not presented here as a phonic feature ‘voice’. However, the presence of the mark should be “voiced” for /b v d z ž g/ and the absence of the mark should be “voiceless” for /p f t s š k/. The term ‘sonorité’ Martinet chooses is perfectly appropriate to designate the mark (a phonic feature ‘voice’).²¹ However, what Martinet presents as « sonorité »²² is evidently a relevant feature (un trait pertinent, as Martinet himself says), not a phonic feature. It would be unproblematic if ‘voice’ (i.e. ‘sonorité’) is understood as the mark, and « voiceless » (i.e. absence of ‘voice’) or « sourd » characterizes /p f t s š k/ and « voiced » (i.e. presence of ‘voice’) or « sonore » characterizes /b v d z ž g/. Retaining the term « sonorité » that Martinet uses, but understanding this term to mean the ‘mark’, I would then reformulate Fig 10 as follows.

Fig 11

	presence of ‘sonorité’ = « sonore » (relevant feature 1) in e.g. /b/
‘sonorité’ (phonic feature)
	absence of ‘sonorité’ = « sourd » (relevant feature 2) in e.g. /p/

or, for German

Fig 12

	presence of ‘Stimm’ = “stimmhaltig” (relevant feature 1) in e.g. /b/
‘Stimm’ (phonic feature)
	absence of ‘Stimm’ = “stimmlos” (relevant feature 2) in e.g. /p/

or, for Japanese

Fig 13

	presence of ‘声’ = “有声” = “voiced” (relevant feature 1) in e.g. /b/
‘声’24
	absence of ‘声’ = “無声” = “voiceless” (relevant feature 2) in e.g. /p/

The task of choosing a nomenclature whereby to designate a ‘mark’ in different languages is not always an easy one, since such a nomenclature should be one which is, so to speak, a term of hypernym, e.g. ‘voice’, with two cohyponyms, ‘voiced’ and ‘voiceless’. Another example of a designation of a ‘mark’ is ‘nasality’ which is a hypernym and ‘nasal’ and ‘non-nasal’ which are two cohyponyms. However, there is some difficulty in finding an appropriate French term to designate for a mark associated with ‘sonore’ and ‘sourd’. We have seen ‘sonorité’ above but this term can also be comprehended as being opposed to ‘sourdité’ (of course, not ‘surdité’ = deafness).

13) I have so far deliberately refrained to designate the two phonemes of a neutralizable correlative opposition as ‘marked phoneme’ and ‘unmarked phoneme’ because I foresee some possible problem to crop up as to how to understand ‘marked phoneme’ and ‘unmarked phoneme’. A phoneme characterized by the presence of a mark may be called ‘marked phoneme’, e.g. b, and a phoneme characterized by the absence of a mark may be called ‘unmarked phoneme’, e.g. p. As I understand, ‘marked’ and ‘unmarked’ mean, respectively, ‘characterized by the presence of a mark’ and ‘characterized by the absence of a mark’, the mark here being one and the same mark. But ‘mark’ is not a relevant feature, so that the marked phoneme and the unmarked phoneme should not be conceived of as follows.

Fig 14

A = archiphoneme + no mark B = archiphoneme + mark

14) Trubetzkoy’s statements about the archiphoneme, the archiphoneme representative, the mark, the marked and the unmarked as I have related above created in some quarters an explicatory account as the following which I find bordering on gobbledegook.

Marque. C’est ce dernier type de neutralisation qui a donné naissance à la notion de marque. Comme c’est toujours le même élément qui apparaît dans les positions où un seul des deux peut apparaître, on l’appelle non marqué, ou encore extensif (l’autre, d’usage plus limité, étant intensif ou marqué). Dans les contextes où seul l’élément non marqué est possible, on dit qu’il représente l’opposition toute entière, ou encore qu’il représente l’archiphonème, c’est-à-dire ce qui est commun aux deux phonèmes de l’opposition. On peut aller plus loin et postuler que le non-marqué représente toujours l’archiphonème ― même dans les contextes où il s’oppose au marqué. Leur opposition peut alors, selon le terme de Troubetzkoy, être appelée privative, en ce sens que l’un des deux termes, le marqué, possède des traits distinctifs [394]²⁵ dont l’autre est privé. Ducrot & Schaeffer (1995: 276-277).

It is difficult to know which individual writer(s) for this collaborative work to attribute the above quoted view to ― I am not interested to do so ― but the view expressed above does not seem correct to me, nor does it seem to elucidate the phenomenon of neutralization and the archiphoneme.

Given a number of problems arising from the notion of ‘mark’, I expressed my view held in the late 1970’s in Akamatsu (1978) which contained my total agreement with Tcheu (1969) about the status of ‘mark’ as a phonic entity and about the phonological value which the presence and absence of a mark are endowed with, i.e. relevant features. My paper brought forth a sympathetic reaction from Vion (1978) from the floor who said in agreement with me (and also indirectly with Tcheu) that “... dès qu’un phonème est défini par l’absence de marque. Cette absence lui confère un trait distinctif ...”. Vion went farther by saying “Comme la tentation est grande, même en dehors de la phonologie, de n’attribuer de valeur linguistique qu’au seul terme marqué, il conviendrait peut-être ‘d’abandonner cette notion.”

There exists an excellent huge work on the concept of ‘mark’ and related concepts in Prague school phonology written from a historical point of view, and from a neutral standpoint without being polemic, namely, Viel (1984). In Akamatsu (1986), a review of Viel (1984), among other not a few points I discussed, I mention critically the view of the mark held by Viel that a mark is a phonological, not a phonic, entity, which probably leads to the idea that a mark is a relevant feature. I take exception to the idea of the two terms of a neutralizable opposition /A/ and /B/ being as

/A/ “a b c” /B/ “a b c d”

and instead I believe in

Fig 15

/A/ “a b c e” /B/ “a b c d”

This rounds up my critical discussions of one of the two major issues arising from what is shown in Fig 3 and Fig 4. I will continue below to discuss other major issues in Trubetzkoy’s concepts of neutralization, the archiphoneme, the archiphoneme representative, the mark, the marked phoneme and the unmarked phoneme.

1) Trubetzkoy (1936b: 31-32, 1939: 70-71) writes: “In der Aufhebungsstellung wird somit ein Oppositionsglied zum Stellvertreter des »Archiphonems [Trubetzkoy’s emphasis]« des betreffenden Gegensatzes ...” Any statement made along this line does not seem to occur in Trubetzkoy (1936a) or other writings by Trubetzkoy of earlier dates. Since /A/ (»Archiphonem + Null«) which Trubetzkoy says occurs in the position of neutralization with the phonological content that is identical with that of »Archiphonem« which itself does not occur in that position, Trubetzkoy is led to believe that /A/ represents »Archiphonem«, hence the concept of ‘Archiphonemvertreter’ (‘archiphoneme representative’). I have always and totally opposed to the concept and intervention of ‘archiphoneme representative’ and discussed deleterious aspects of ‘archiphoneme representative on a few occasions (see e.g. Akamatsu (1972, 1975, 1976)). Incidentally, it is important to stress here that ‘-vertreter’ in ‘Archiphonemvertreter’ or ‘representative’ in ‘archiphoneme representative’ does not mean ‘realization’.²⁶ In this connection, it is also important to heed a caution sounded by Mounin (1976). This is recorded in Akamatsu (1976: 101) where it reads: ‘... il convient surtout de ne pas comprendre «représentation» comme «réalisation».’ In other words, ‘archiphoneme representative’ should be understood as ‘the phoneme which represents (not realizes) the archiphoneme in the position of neutralization.’ All functionalists are agreed that a phoneme is realized (by sounds); so is an archiphoneme. They are not represented.

2) Allowing for the concept of ‘archiphoneme representative’ nullifies effectively the concept of the archiphoneme itself. The entity ‘archiphoneme representative’ is inadmissible as it vitiates the very concept of ‘archiphoneme’.²⁷ I have long argued against the notion of ‘archiphoneme representative’ in the past. Among other writers who completely agree with me by rejecting ‘archiphoneme representative’ is Davidsen-Nielsen (1978: 52-53). He too emphasizes that accepting ‘archiphoneme representative’ results in negating neutralization and, he adds, pave the way to misrepresenting neutralization as if it were ‘defective distribution’.

3) According to Trubetzkoy, even if the phoneme that represents the archiphoneme is objectively (i.e. phonetically?) identified as marked, this phoneme should be identified as the unmarked. This shows Trubetzkoy’s insistent view that the archiphoneme is represented by the unmarked, not the marked, phoneme. In this connection Vachek (1966a: 18), who abandons the term (and the concept) ‘archiphoneme representative’, quotes from Trubetzkoy (1931: 98) as follows.

In den Stellungen wo die korrelative Eigenschaft eines Phonems seine phonologische Gültigkeit verliert, wird dieses Phonem mit dem merkmallosen Korrelationsglied identifiert, selbst wenn es objektiv mit dem merkmalhaltigen Korrelationsglied identisch ist. [The entire passage quoted is italicized by Trubetzkoy. The additional emphasis by a single underline is mine.]

4) One particular ambiguity that persists about ‘archiphoneme representative’ concerns the question of whether the archiphoneme representative is a phonic entity (sound) or a phonological entity (phoneme). This question may surprise the majority of scholars who take it granted that the archiphoneme representative is a phonic entity. Nevertheless, this is a serious question that need to be answered. My own answer is that it is a phoneme, a phonological entity. On this question of the archiphoneme representative being either a phonic entity or a phonological entity, see Akamatsu (1988: 368-376).

5) Trubetzkoy’s introduction of ‘archiphoneme representative’ (both the concept and the term) and his employing it throughout his descriptions of neutralization is very unfortunate. In my considered view, recourse to ‘archiphoneme’ is necessary and sufficient in the description of every case of neutralization. According to Trubetzkoy, the archiphoneme itself, and as such, does not seem to occur in the position of neutralization. This is, in my view, unacceptable because both the archiphoneme and the phoneme are phonematic units which inherently occur sequentially as the distinctive units of the second articulation.²⁸

6) It is significant that Trubetzkoy never speaks of realizations of archiphonemes, whereas he always speaks of realizations of phonemes. Both the phoneme and the archiphoneme are phonematic units and consequently both are realized by sounds. One wonders if, for Trubetzkoy the status of what he calls the archiphoneme is different from that of the phoneme and this is why the archiphoneme needs the phoneme which ‘represents’ it since the archiphoneme is not, according to Trubetzkoy, realized. It seems that ‘representative’ (‘represent’) and ‘realization’ (‘realize’) are two distinct notions and too heterogeneous ever to be interchangeable.

7) In connection with the fact that Trubetzkoy needs the notion and term ‘archiphoneme representative’, I wish to quote two statements I made in Akamatsu (1988: 378). They are as follows.

... he [Trubetzkoy] rarely presents the archiphoneme as a distinct unit which the phoneme doubtlessly is, and seldom, if ever, presents the archiphoneme itself as actually occurring in the position of neutralization. [The italics are in the original.]

Should Trubetzkoy be ... of the view that the archiphoneme is a distinctive unit which itself occurs in the position of neutralization, there would be no need for him to speak of the archiphoneme being ‘represented’ by a distinctive unit, the phoneme.

8) For an extensive discussion by me on various aspects of ‘archiphoneme representative’, see Akamatsu (1988: 367-398).

It would not be surprising if not a few of the readers who are interested in the activities of the Prague Linguistic Circle believe that the concept and term of ‘archiphoneme’ continue to be indisputably one of the prominent features of Prague School phonology. The readers could be forgiven to assume that the theory of neutralization and the archiphoneme as seen in Trubetzkoy’s writings on phonology culminating in his magnum opus, Grundzüge der Phonologie, is here to stay in our days. The reality is as follows. During the so-called ‘interwar period’, i.e. 1926 to 1939, during which Prague School phonology enjoyed steady and fruitful development, the concepts of neutralization and the archiphoneme were the joint and pivotal notions in the formulation of the theory of neutralization and the archiphoneme. However, immediately following the publication of Grundzüge der Phonologiein 1939, the term (with the concept) ‘archiphoneme’ was abandoned by the Prague School, though they most probably continue to subscribe to the term (and the concept?) of ‘neutralization’. The predominant majority of non-Pragueans who have read up on Prague School phonology up to and including Grundzüge der Phonologie (or its translations) but not also Prague School’s writings in the post-1939 period may well not be aware of the official abandonment of the term (and the concept) ‘archiphoneme’ and the inevitable consequence that this event has had in the post-1939 Prague phonology. I hasten to add that functional phonology being practised by what I personally call ‘the Paris School’ led by and associated with Martinet continues to adhere to the archiphoneme as well as neutralization, both of which are alive and well.

We read the following in Vachek (1966a: 18).

†archiphonème (Archiphoneme. Archiphonem. Archifoném/Archifonéma) Élément commun de deux ou plusieurs phonèmes corrélatifs, qu’on peut concevoir abstraction faite des propriétés de corrélation.’ ... ‘Par l’archiphonème nous entendons l’ensemble des particularités distinctives qui sont communs aux deux’ [= aux deux termes d’une opposition phononlogique neutralisable]. Tb39c.71 = TC49.81²⁹ [Terme abandonné après 1939]

The term ‘archiphoneme’ is said to be abandoned but the concept itself is assumed to be also abandoned. Logically, any concept precedes the corresponding term.

The abandonment of the term and concept of ‘archiphoneme’ is unambiguously mentioned by Vachek (1959: 110) who says that “... the idea of ‘archiphoneme’ which has by now been wholly abandoned by the Prague phonemicists ...”

The direct consequence of the demise of ‘archiphoneme’ is the demise of ‘archiphoneme representative’ (‘Archiphonemvertreter’). It stands to reason that if ‘archiphoneme’ is rejected, an entity which is supposed to represent the archiphoneme becomes unnecessary. It is no surprise that, having abandoned the term (and the concept) of ‘archiphoneme’, the post-1939 Prague School also abandoned the term (and the concept) of ‘archiphoneme representative’. That this is so is proven when we see Vachek (1966a: 18) writing as follows.

†archiphoneme représenté par le phonème non-marqué (Archiphonem durch merkmalloses Phonem symbolisiert. Archiphoném zastoupen bezpřiznakoým fonémen)

This is explained in German and French, and is exemplified by p (for the correlation p/b ou p/p’) and s’ in s’vèčka (for the correlation s/s’) in Russian. References are made to Tb31a,98 (= Trubetzkoy, ‘Die phonologischen Systeme’ in TCLP 4.96-116 (1931)).

We have seen that the post-1939 Prague School has abandoned ‘archiphoneme’ and ‘archiphoneme representative’. One is naturally curious to know why ‘archiphoneme’ has had to be abandoned. Vachek (1966b: 62) has this to say:

One cannot therefore posit the presence in this [spoken] chain of a unit placed on a subphonemic [my italics] level, if it should be placed on the level of the phoneme. To put the matter still differently, any simultaneous bundle of distinctive features must be evaluated, by definition, as a phoneme, not as a unit subordinated to the phoneme [my italics]. It appears thus, that in concrete phonological structures, implemented by spoken utterances, the concept of archiphoneme has no justification. And it is hardly chance that since Trubetzkoy’s Grundzüge this term has been virtually abandoned in phonological books and papers by the Prague group―this has obviously been due to its unfruitfulness.

I disagree with Vachek in that I do not believe that the level at which the archiphoneme is posited is at a subphonemic level. I have in the past, in Akamatsu (1972: 1069, 1079), briefly succumbed to positing the archiphoneme at a hyperphonemic (or a supraphonemic) level.³⁰ However, I quickly dismissed this error when I realized that my reasoning was purely logical, not phonological. Since then, my view has been that the archiphoneme is posited at the same level as is the phoneme. I believe that the archiphoneme and the phoneme are both minimum distinctive units of the second articulation and consequently they are at an equivalent level. In other words, both the phoneme and the archiphoneme are the minimum phonematic units³¹ which are at the level of the second articulation. This can be clearly seen in syntagmatics. For example, in the case of the neutralization of /s/ vs /z/ in English, the archiphoneme associated therewith is /s-z/, as in cats /kat s-z/ (or /katS/ as many writers would notate). /k/, /a/, /t/ and /s-z/ (or /S/) are all posited at the same level, the archiphoneme being at neither a higher nor a lower level than the phoneme. Another reason for which I disagree with Vachek’s rejection of ‘archiphoneme’ is that the list of the phonematic units of a language should consist of both phonemes and archiphonemes. Martinet (1949: 7) correctly says that ‘... the phoneme list should include archiphonemes ...’, though I would personally say ‘the phonematic list’ instead of ‘the phoneme list’ as ‘phonematic entities’ in my parlance are both phonemes and archiphonemes. It seems to me that Vachek’s objection to ‘archiphoneme’ is neither convincing nor well founded.

Vachek might object to ‘archiphoneme’ for another reason, though he does not mention this. Trubetzkoy’s notions of ‘archiphoneme’ and ‘neutralization’ are such that ‘archiphoneme’ itself does not occur in the position of neutralization.

Fig 16

/A/ = archiphoneme + zero (unmarked phoneme) /B/ = archiphoneme + mark (marked phoneme)

Fig 16 shows the two terms, /A/ and /B/, are in opposition to each other and form a neutralizable correlative opposition, /A/ vs /B/.³² According to Trubetzkoy, the phonological entity that occurs in the position of neutralization is the unmarked phoneme (which is said to represent the archiphoneme.) In Fig 16 above, /A/ should not be taken as the ‘archiphoneme’ but as the unmarked phoneme. In this sense ‘archiphoneme’ enjoys no status as a phonological (phonematic, to be precise) entity. Such an entity will have no raison d’être. It is effectively a non-entity. Even in /B/, the archiphoneme is contained in the marked phoneme and has no independent status as a phonological entity. Vachek could object to ‘archiphoneme’ for this reason.

Had Trubetzkoy survived beyond 1938, he would have continued to preserve ‘archiphoneme’ and ‘archiphoneme representative’. He would have continued to hold that the archiphoneme representative is always the unmarked phoneme. However, in the phonological practice of all those like me who operate with ‘archiphoneme’ as well as ‘neutralization’ but not with ‘archiphoneme representative’, the concept of ‘archiphoneme representative’ is a deleterious one in the theory of neutralization and the archiphoneme, and the demise of ‘archiphoneme representative’, whilst ‘neutralization’ is maintained, is welcome.

We have seen that Prague School has abandoned both ‘archiphoneme’ and ‘archiphoneme representative’ since 1939. How about neutralization’? Vachek (1966b: 62) mentions the terms ‘neutralisation’, Aufhebung/Neutralisation, and Neutralisace/rušení), and regards ‘neutralisation’ as ‘Terme fondamental de la phonologie’. Furthermore, Vachek (1966b: 61) positively retains the term ‘neutralisation’. We presume that the concept as well as the term of ‘neutralization’ is retained. Vachek (1966b: 55) refers us to ‘opposition neutralisable’ where we are shown the example which is earlier given by Trubetzkoy (1939: 70).³³ The example runs as follows: [e] and [ɛ] in French function phonologically as /e/ vs /ɛ/ in final open syllable but in all other positions the occurrence of [e] and that of [ɛ] are mechanically controlled³⁴ in that [e] occurs in non-final open syllable and [ɛ] in closed syllable so that [e] and [ɛ] are combinatory variants of a single phoneme in all positions except in final open syllable, and /e/ vs /ɛ/ is neutralized in all positions except in final open syllable. Trubetzkoy’s (1939: 70) words ‘ein[es] einzig[en] Phonem[s]’, and Trubetzkoy’s (1949: 80) words ‘un même phonème’ or Trubetzkoy’s (1968: 78) words ‘a single phoneme’ refer of course to ‘an archiphoneme, i.e. /e-ɛ/. My disagreement with Trubetzkoy’s description of the occurrences of [e] and [ɛ] is that, unlike him, I would say that, in non-final open syllable, not only [e] but also [ɛ] occurs (as in e.g. maison [me-] ~ [mɛ -]), in which case [e] and [ɛ] are in free variation. To be precise, I should add that, in non-final open syllable, not only [e] and [ɛ] but any vowels positable between them also occur in free variation, so that /e/ vs /ɛ/ is neutralized. My conclusion is that /e/ vs /ɛ/ is valid in final open syllable but is neutralized in the other positions, i.e. in closed syllable and in non-final open syllable. For these reasons I would not talk about there being free variation of just [e] and [ɛ] (as Trubetzkoy means) in positions other than final open syllable.

It is significant that the date of the abandonment of the term (as well as that of the concept of) ‘archiphoneme’ as noted in Vachek (1959: 110, 1966a: 18) is presumably just after the year 1939. The year 1939 is the year in which Grundzüge der Phonologie posthumously saw the light of day. The year 1939 postdates 25 June 1938 when Trubetzkoy passed away. It can be presumed that the formal abandonment of ‘archiphoneme’ is ascribable to the Prague Linguistic Circle not long after Trubetzkoy’s decease. It is anybody’s guess how Trubetzkoy might have reacted to the abandonment of ‘archiphoneme’ by the Prague Linguistic Circle. All the same I am tempted to imagine that Trubetzkoy would not have been happy and most probably objected to the abandonment of the concept and term of ‘archiphoneme’.

In Akamatsu (1992b), I pointed out in detail detrimental consequences that the abandonment of (the concept and term of) ‘archiphoneme’ entails in the theory of neutralization and the archiphoneme.³⁵

We now turn our attention to Trubetzkoy’s analysis of neutralization of phonological oppositions other than what he would call bilateral privative correlative oppositions which each consists inevitably of two phonemes, in connection with which the notion of the mark and those of the marked and the unmarked are resorted to. However, in reality, neutralizable disjunct oppositions can each consists of more than two phonemes as well, as we shall see.

1) Trubetzkoy’s theory of neutralization and the archiphoneme is inadequate to account for neutralization of disjunct oppositions such as /e/ vs /ɛ/ in French, /m/ vs /n/ in English, /m/ vs /n/ vs /ɲ/ in Spanish, /t/ vs /d/ vs /t’/ vs /d’/ in Russian, and so on. The case of /e/ vs /ɛ/ in French is mentioned by Trubetzkoy (1936a: 12-13, 1936b: 30, 1939: 70), who notes that /e/ vs /ɛ/ is valid in final open syllable but is neutralized elsewhere, that is, in non-final open syllable and in closed syllable where a single phoneme occurs which represents the archiphoneme /e-ɛ/.

The case of e.g. /e/ vs /ɛ/ in French is mentioned by Trubetzkoy (1936a: 12-13, 1936b: 30, 1939: 70). As /e/ vs /ɛ/ in French is a disjunct opposition, what might be known as the mark is clearly inconceivable and unidentifiable. Trubetzkoy does not mention the mark, the marked phoneme, and the unmarked phoneme in connection with a neutralizable disjunct opposition such as /e/ vs /ɛ/. Instead, the following is what Trubetzkoy says about the neutralization of /e/ vs /ɛ/ in French. [ɛ] occurs in closed syllable (as in fer), while [e] occurs in open syllable (as in maison). (The examples, fer and maison, are mine). Trubetzkoy considers [e] and [ɛ] as combinatory variants of a ‘single phoneme’ which he calls an ‘archiphoneme’ in closed syllable and in open (correctly, non-final open) syllable.

2) In the case of the neutralization of /m/ vs /n/ in English before /f/ or /v/ (as in comfort, canvas), it would be impossible for Trubetzkoy to declare which, /m/? or /n/?, is the unmarked phoneme.

3) What about the neutralization of /m/ vs /n/ vs /ɲ/ in Spanish? The very fact that this neutralizable opposition consists of three phonemes makes it impossible to identify the mark(s), the marked and the unmarked. Anyhow Trubetzkoy does not show us in his writings how one could go about analyzing neutralization of disjunct oppositions, be they oppositions between two phonemes (e.g. /e/ vs /ɛ/) or more than two phonemes (e.g. /m/ vs /n/ vs /ŋ/, /m/ vs /n/ vs /ɲ/ vs /ŋ/).

4) Another interesting case is the neutralization of /r/ vs /ɾ/ in Spanish. This neutralizable opposition consists of two phonemes, but is it amenable to the distinction between the marked phoneme and the unmarked phoneme? If it is, what is the mark? I have in the past discussed the neutralization of /r/ vs /ɾ/ in Spanish in Akamatsu (2019: § 16.8 = 408-409) so will be brief here. /r/ will be defined as “multiple tap” and /ɾ/ “single tap”. /r/ and /ɾ/ are distinguished from each other in word-medial position through the opposition “multiple” vs “single” (as in carro vs caro) which is cancelled when the neutralization of /r/ vs /ɾ/ takes place, and the archiphoneme /r-ɾ/ “tap” occurs in the position of neutralization, i.e. in word-initial position (as in rabo) and word-final position (as in bar). The important point is that there is no question of talking about a mark, the marked phoneme, the unmarked phoneme, and certainly not a correlative opposition. /r/ vs /ɾ/ in Spanish is evidently a disjunct opposition and what Trubetzkoy classifies as an ‘isolated opposition’.

5) In general, Trubetzkoy describes and discusses cases of neutralization of correlative oppositions³⁶ where correlative pairs of phonemes are characterized by the marked vs the unmarked distinction. Interestingly, however, Trubetzkoy also admits of neutralization of disjunct oppositions at an early stage of his discussions about neutralization without, however, naming individual languages in question. At a later date, Trubetzkoy (1939: 163-164 = 1949: 193-194 = 1969: 168) refers to neutralization of the opposition between all five nasal phonemes before vowels in Tamil, viz /m/ vs /n/ vs /n̪/ vs /ɳ/ vs /ɲ/,³⁷ and the neutralization of /m/ vs /n/ vs /ŋ/ vs /ɲ/ in some central Chinese dialects³⁸ by saying that, in all such cases, there is neutralization of ‘multilateral’(?)³⁹ oppositions between all (the emphasis is Trubetzkoy’s) nasal phonemes and that there occur the archiphonemes, which I would notate as /m-n̪-n-ɳ/ and /m-n-ŋ-ɲ/, respectively, or even better as /~/ for both. Trubetzkoy (op cit loc cit) appropriately refers to each of these archiphonemes as ‘Nasal-Archiphonem’ (‘archiphonème nasal’, ‘nasal archiphoneme’) which I will also call archiphonemes definable by the single relevant feature “nasal”. Recourse to the distinction between ‘marked’ and ‘unmarked’ would be totally impossible in analyzing neutralization of the above-mentioned disjunct oppositions even if Trubetzkoy resorted to a circuitous analysis where recourse is had to cumulative neutralizations.⁴⁰

In the foregoing parts of this paper, I have submitted to a brief critical survey the principal notions on which Trubetzkoy’s theory of neutralization and the archiphoneme is founded. In so doing, I critically discussed, among other notions, ‘neutralization’, ‘archiphoneme’, ‘archiphoneme representative’, ‘mark (of correlation)’, ‘quality of correlation’, ‘marked’ and ‘unmarked’, etc. I have not touched on all the pertinent notions. I could not have done so in this short survey, but I hope that what I have discussed will afford at least an adequate bird’s-eye view of Trubetzkoy’s theory of neutralization and the archiphoneme.

It is my intention to propose in the following lines the principal points of my own theory of neutralization and the archiphoneme. The readers will find that there exists substantial discrepancy between Trubetzkoy’s theory and mine. Interested readers may perhaps wish to consult Akamatsu (1988) which is devoted entirely to the subject of various aspects of the theory of neutralization and the archiphoneme as I conceive of. In spite of a fair measure of discordance between us on certain major issues, Trubetzkoy and I share the principles of functionalism and structuralism on which our theory of neutralization and the archiphoneme is fundamentally based.

In the foregoing pages, I have first of all attempted to give a short critical survey of essential concepts underlying Trubetzkoy’s theory of neutralization and the archiphoneme which culminated in Grundzüge der Phonologie. A series of key concepts on which Trubetzkoy’s theory in question is founded include ‘phonological opposition’, ‘phoneme’, ‘relevant feature’, ‘basis for comparison’, ‘privative opposition’, ‘bilateral opposition’, ‘mark’, ‘marked/unmarked’, ‘correlation’, ‘neutralization’, ‘archiphoneme’, ‘archiphoneme representative’, etc. I readily accept and operate with such concepts except ‘privative opposition’, ‘bilateral opposition’ (both being two types of ‘phonic difference’ to me), ‘mark’, ‘marked/unmarked’ and ‘archiphoneme representative’. My concept of ‘archiphoneme’ differs from Trubetzkoy’s in that his is not a phonematic unit that actually occur in the position of neutralization because the archiphoneme is said to be represented by the unmarked phoneme. Still constrained by ‘Jakobsonian binarism’, the terms of neutralizable oppositions must be, for Trubetzkoy, two, not more. Both ‘bilateral opposition’ and ‘multilateral opposition’ prove to be two-term phonological oppositions. Trubetzkoy gives only a partial solution to neutralization of more-than-two-term phonological oppositions because of constraint of binarism; he can only points to the common base of the terms of such phonological oppositions. The tight relation between the archiphoneme and the unmarked member of a neutralizable opposition is absolute to Trubetzkoy. To my mind, the fundamental cause that is reflected in various ways in Trubetzkoy’s theory of neutralization seems to boil down to binarism.

As shown in VIII (= VIII-1 through VIII-5), in my theory of neutralization and the archiphoneme, binarism has no place. To me, phonological oppositions are simple oppositions (each with two terms) or multiple oppositions (each with three or more terms), as the case may be. Trubetzkoy’s notions of ‘bilateral opposition’, ‘multilateral opposition’, ‘privative opposition’ and ‘gradual opposition’ are types of phonic differences,⁶⁶ not phonological oppositions (see Akamatsu 1977), and I do not operate with them. At any rate, [p] vs [b] in English (said to be a privative opposition) proves to be /p/ vs /b/, [i] vs [e] in French (said to be a gradual opposition) proves to be /i/ vs /e/, and [f] vs [k] in German (said to be an equipollent opposition) proves to be /f/ vs /k/. Functionally speaking, that is, phonologically, /p/ vs /b/, /i/ vs /e/, and /f/ vs /k/ all play the same function, that is, distinctive function (pun /pʌn/ vs bun /bʌn/; prix /pri/ vs pré /pre/; Fabel /fa:bəl/ vs Kabel /ka:bəl/). The difference between ‘privative’, ‘gradual’ and ‘equipollent’ is irrelevant in /p/ vs /b/, /i/ vs /e/, and /f/ vs /k/.

What is crucially important in my theory of neutralization and the archiphoneme is the notion of ‘exclusive opposition’. The only difference that matters between types of phonological oppositions, in the matter of neutralization, is the difference between ‘exclusive opposition’ and ‘non-exclusive opposition’. Only an exclusive opposition can be a neutralizable opposition, though an exclusive opposition can be a non-neutralizable opposition. The common base of the two or more phonemes of a neutralizable exclusive opposition is the archiphoneme which occurs in the position of neutralization, but, in the position of relevance, the common base in question is certainly not the archiphoneme, though Trubetzkoy suggests to the contrary. I do not believe that a phonematic unit can be conceived to be included in another phonematic unit.