LINGUIST List 11.156

Mon Jan 24 2000

Disc: Phonemic Analysis

Editor for this issue: James Yuells <>


  1. Richard D. Janda, Phonemic analysis

Message 1: Phonemic analysis

Date: Mon, 3 Jan 2000 23:22:04 -0500 (EST)
From: Richard D. Janda <>
Subject: Phonemic analysis

 In order not to contradict the recent discussion-contribution labeled 
"10.1976...: What Exactly Are Allophones? / *Last Posting*" (empha- 
sis added), this message concludes by focusing on a distinct issue raised 
in those earlier exchanges: "Are All Allophones Assimilatory?". Still, 
the lead-in to this topic occurs via the parts (0.1-0.4) of a prefaced post- 
script to the (allo)phon(em)e issue which I had not quite finished when 
the end-of-year break arrived. Apologies in advance for what may seem 
like beating a dead horse into the ground. 

 _Cyclopedia of Medicine_ (1933): 6.149f.: "A _phoneme_ is 
 an auditory hallucination consisting of spoken words". 
 [Quoted disapprovingly in Twadell 1935: n.9.] 
 We linguists are a strange bunch. Inconsistent, and sometimes even 
verging on hypocritical, I mean. Let a layperson say that, e.g., _din- 
ner_ really should have only its earlier sense of 'breakfast' (its etymon 
is dis-jejun-are '(to) break fast'), and most linguists would burst out 
laughing. Ah, but suggest an alternative interpretation for _allophone 
(of)_, and some linguists would apparently consider hanging, drawing, 
and quartering too good a fate for you. And this in a field that can't 
even agree on a single standard system of transcription for academic 
books and journals.... 
 Seriously, if we're consistent in our belief that meaning follows usage, 
especially the usage of large groups, then we might want to say "Pass!" 
to individual discussions of technical terms ("What _Allophones_ Mean 
to Me") and instead report either on broad trends in contemporary usage 
by today'9s linguists or on the consensus of earlier usage. The latter is 
especially intriguing in light of some recent positings suggesting that 
"... MAYBE the graduate students of the last 20-30 years have not been 
required, in the full-blown Chomskyan era which disparages all foun- 
dational structuralist work done before, to read the books and articles 
which shaped linguistics...". So, in what follows, there are presented 
some direct quotations from post-Bloomfieldian American structuralists, 
with a bit of accompanying editorializing. 
 First, though -- not forgetting those who want to prosecute any lin- 
guists perpetrating, encouraging, or advocating excessively deviant in- 
terpretations of _(allo)phone_ and similar labels -- I propose that a body 
with worldwide jurisdiction be set up for this purpose. Being an inter- 
national tribunal, it will he headquartered in Geneva, with French being 
one of its official languages, and so this court will henceforth be known 
as the Cour de linguistique ge'ne'rale.... 
 Still, to be honest, I find the discussion of whether allophones (or at 
least non-"elsewhere" [-default] allophones) are always assimilatory in 
nature to be a much more rewarding topic, so I will close by posing 
some questions in response to earlier postings addressing that issue. 

 From a recent children'9s book on dinosaurs: "_Allosaurus_ 
 comes from the Greek words allo- 'leaping' and saur- 'liz- 
 ard', so _allosaurus_ means 'leaping lizard'!". [Close, 
sense that is approximately as follows: all(o)- 'other, different, atypi- 
cal' (still, the next sense is more allophone-like: 'isisomeric in form 
[for a specified chemical compound]'). After all, Greek allo- is cognate 
with Latin alien-us, and you can hardly make something sound more 
'exclusionarily different' to students' ears than the latter term does. And 
the examples that follow again include some exclusionary-sound scien- 
tific terms (rather than collusionary, as our allophones are): _allogamy_ 
involves *cross*-fertilization; _allogenicity_ involves individuals of the 
same species that are sufficiently *unlike* genetically to interact antige- 
nically; _*allopatric* speciation_ occurs in a *different* geographical 
area (from another instance of speciation) or in isolation; _allostery_ re- 
lates to *changed* shape or activity of a protein (as an enzyme) resulting 
from combination with a substance at a point *other than* the chemical- 
ly active site; _allopurinol_ is a drug (C5H4N4O) used to promote ex- 
cretion of uric acid [emphases added]. Well, at least the last of these 
leads us back to allophones, sort of. 
 My own experience in teaching phonology is that Gleason's 1965 term
_co-allophones_ (allophones of the same phoneme) -- which can be op- 
posed to _anti-allophones_ (allophones of different phonemes) [or your 
nomination for a better opposite] -- is much more effective in helping 
students gain a handle on the relational notions in question (i.e., internal 
vs. external variant) than harping on the etymology of allo-. But, if oth- 
er terms work better, I'd be glad to use them. 

 This entire recent discussion of phonemes, allophones, and the like 
really got going back in November when Martin Salzmann [in Linguist 
List Vol-10-1778] asked (more or less) whether there aren't really two 
senses of allophone -- one being a class or type, and the other involving 
individual tokens. Among the ensuing responses, some postings reject- 
ed such suggested terms as allophoneme vs. allo-allophone (or smallo- 
phone) and argued that allophone (as type) and phone (as token) already 
covered this ground. 
 This may be common in current usage (though I was unaware of this), 
and it does have a lot going for it. If this is the wave of the future, I'd 
be happy join the crowd. However, it's important to be clear that this 
was certainly *not* the dominant post-Bloomfieldian sense of _phone_. 
At least one tradition (as my own earlier posting mentioned) treats allo- 
phones as phones that have been classified via assignment to phonemes, 
so that a phone is a sound that has been segmented but not yet classi- 
fied. The only place where I can document this at the moment is in the 
following German source (but maybe someone else can corroborate it 
with other citations): 
Fritz[, Gerd] 1973a ["Phonologie I: Die Analyse der Phonemebene"] in 
/Baumgaertner et al. 1973a [_Funk-Kolleg Sprache..., Bd. 1_]: 131, & 
 440 ["Glossar"]: "_Phon_: kleinstes Segment der Rede, das durch 
 Segmentierung gewonnen, aber dem System noch nicht durch Klas- 
 sifizierung zugewiesen ist". ['_Phone_: smallest segment of speech 
 which... [has been] gained through segmentation but not yet as- 
 signed to the system via classification'.] 
 Crucial in the above approach is that a phone is *not* here being treat- 
ed as necessarily an individual, even unique, physical sound (with a set 
of phones then being grouped together as one allophone). Rather (and 
this is confirmed by other citations from much earlier), _phone_ here 
itself stands for a class. 
 There is a reason for this, and it mostly has to do with Leonard Bloom- 
field's extreme distrust of phonetic transcriptions, and hence of phones 
as potentially important, reliable linguistic units (as opposed to the re- 
sults of laboratory-phonetic work). This conclusion is by no means a 
new one; it has already been drawn in such accounts of the history of 
(especially 20th-century) phonology as Fischer-Jo/rgensen 1975, 
Hymes & Fought 1975/1981, Anderson 1985 (see also Anderson 
1974). Even Chomsky 1962/1964 took pains to refer to (and criticize) 
"Bloomfield's apparent rejection of the level of structural phonetics". 
 Here's what L.B. himself said: 
Bloomfield 1933: 84-85: "Practical phoneticians sometimes acquire 
 great virtuosity in discriminating and reproducing all manner of 
 strange sounds. In this, to be sure, there lies some danger for lin- 
 guistic work.... The chief objection to this procedure is its inconsis- 
 tency. The phonetician's equipment is personal and accidental; he 
 hears those acoustic features which are discriminated in the langua- 
 ges he has observed. Even his most 'exact' record is bound to ig- 
 nore innumerable non-distinctive features of sounds; the ones that 
 appear in it are selected by accidental and person factors.... He 
 should remember that his hearing of non-distinctive features depends 
 upon the accident of his personal equipment, and that his most elabo- 
 rate account cannot remotely approach the value of a mechanical rec- 
 ord.... Only two kinds of linguistic records are scientifically rele- 
 vant. One is a mechanical record of the gross acoustic features, such 
 as is produced in the phonetics laboratory. The other is a record in 
 terms of phonemes, ignoring all features that are not distinctive in the 
 language. Until our knowledge of acoustics has progressed far be- 
 yond its present state, only the latter kind of record can be used for 
 any study that takes into consideration the meaning of what is spoken 
 .... In fact, the laboratory phonetician... usually formulates his prob- 
 lems not in purely acoustic terms, but rather in terms which he has 
 borrowed from practical phonetics.... In order to make a record of 
 our *observations*, we need a system of written symbols which 
 provides one sign for each *phoneme* [sic] of the language we are 
 recording. Such a set of symbols is a _*phonetic* [sic] alphabet_, 
 and a record of speech in the shape of these symbols is a _*phonet- 
 ic* [sic] transcription_..." [emphases added]. 
 For those who wish to maintain the traditions of our predecessors (and 
I'm not in any sense saying that this is always a good thing), it is crucial 
to go back and see what those scholars really said. Certainly Bloom- 
field's above use of _phonetic_ would surprise many people today. 
 Of course, Bloomfield also had to pay for his idiosyncrasy, in a way. 
His own students and colleagues sometimes took him to task for incon- 
sistency and even for disregarding his own principles (again, see the 
various works on the history of phonology mentioned above). 

 If we turn to the writings of post-Bloomfieldian American structural- 
ists and their usage of terms like _phone_ and _allophone_, we can find 
at least one instance where even _allophone_ apparently *could* poten- 
tially refer to a sound treated as a physical entity (though this quotation 
is itself not unambiguous) [citations are to books unless they specifically 
mention journals]: 
Bloch & Trager 1942: 38-40: "A phonetically trained ... student of 
 English...[,] if his [or her] ear is sharp enough..., ... may record 
 *half a dozen different vowel shades* in successive utterances of 
 _dog_.... Th[e]... examination of the phonetic material with a view 
 to sorting out the distinctive differences...[,] we call _phonemic anal- 
 ysis_...[. A] _phoneme_ is a class of phonetically similar sounds... 
 [; t]he individual sounds which compose a phoneme are its _allo- 
 phones_" [emphasis added]. 
 In any case, several other passages in the earlier American structuralist 
literature make it clear that _allophone_ was used to refer to a type, a 
class of sounds: 
Trager & Bloch 1941 [in _Lg._]: 223: "Sound-types as members of a 
 phonemic class are called allophones". 
Bloch 1948 [in _Lg._]: 38: "A class of phonetically same segments, or 
 of spans consisting wholly of phonetically same segments in the 
 same order, is an _allophone_". 
Hocket 1958: 63: "In... case[s]... where the phonetically trained inves- 
 tigator can hear two or three [!] clearly distinct *types* of sound all 
 representing one and the same phoneme in a given language, the dif- 
 ferent *types* are often called _allophones_" [emphases added]. 
 But the most striking discussion from this period is (I would argue) an 
earlier one by Hockett, who makes it clear (see below) that, in his termi- 
nology, even _phone_ refers to a class. That is, he here seems to have 
followed Bloomfield in viewing transcriptional -- i.e., non-laboratory -- 
phonetics as already involving some classification, and not just segmen- 
tation of a physical signal; also Bloomfieldian are some no longer cur- 
rent uses of _feature(s)_ and _distinctive_. It is also clear from the fol- 
lowing, which uses the quite useful-seeming term _homeophone_, that 
not all terminological suggestions catch on: 
Hockett 1942 [in _Lg.]: �5.1-7.5: "...[B]iophysical analysis and classi- 
 fication... will be called... _alpha-phonetics_.... A _change-point_ 
 is any point at which any organ changes from one type of function 
 to another...[; t]he segment between two successive change-points is 
 an _alpha-sound_. Two utterances which have approximately the 
 same elements in the same arrangement are said to be _alpha-phoneti- 
 cally similar_.... An alpha-phonetic element which recurs in a cer- 
 tain position in all the members of some equivalence class is a _dis- 
 tinctive_ element...[; f]or example, in English, aspirate release of 
 stop consonants is a distinctive element...[, though] this variation is 
 subphonemic.... By going through all of the equivalence classes of 
 a range...[, one can] list all of the distinctive elements that oc- 
 cur.... [Often,] two or more elements clearly constitute a single unit, 
 one of the minimum units into which the utterances of the range may 
 be analyzed, ... termed _features_.... The determination of features 
 and of the postions in which they occur is the business of _beta-pho- 
 netics_.... A _beta-sound_ differs from an alpha-sound in...[that 
 o]nly distinctive features are concerned in the characterization of be- 
 ta-sounds.... The difference between a beta-sound and a _phone_ is 
 that a phone is a member of a phoneme...[; also,] a single phone may 
 include two or more adjacent beta-sounds, and conversely. _Phone- 
 mic analysis_ is the analysis of beta-phonetic material into phones, 
 and the classification of the phones into phonemes.... A _homeo- 
 phone_ is a class of phones such that all members of the class are 
 characterized by all the features that characterize any one member; 
 obviously all the members of a homeophone will belong to the same 
 phoneme. An allophone is a class of phones such that all are mem- 
 bers of the same phoneme and occur in simliar beta-phonetic envi- 
 ronments (the same position).... The homeophone and the allophone 
 may be identical...[, so that] all the members of one are members of 
 the other, and conversely.... The homeophone may include the allo- 
 phone and also have members which are not in the allophone.... The 
 allophone may include the homeophone, and also have members 
 which are not in the homeophone.... In some cases, finally, a home- 
 ophone and an allophone may intersect but each may include phones 
 which are not members of the other..." [see Hockett for examples]. 

 Enough of PABS (Post-Bloomfieldian American Structuralists) and 
their _Papst_ (Bloomfield); what about the alleged invariably assimila- 
tory nature of allophones? This issue links up with a controversy that 
arose in the late 70's and early 80's in connection with responses to so- 
called "natural" phonologies like Natural Generative Phonology and es- 
pecially Stampean Natural Phonology. In the course of argumentation 
presenting evidence that (and Why) "...Phonology Isn't [All] Natural", 
mention was made then of work like the following: 
Flege <1980 [references in Flege & Port 1980, cited in Anderson 1981]: 
 Saudi Arabic does *not* display vowel length differences dependent 
 on following obstruent voicing. [Here entirely paraphrased; empha- 
 sis added.] 
 From the same source comes an instance where at least one allophone 
seems to be not always assimilatory (though, as in some of the other ca- 
ses below, this may be the elsewhere-allophone): 
Anderson 1981 [in _LI_]: 509-512, after Einarsson 1945, Pe'tursson 
 1974: Modern Icelandic /k/ and /g/ are realized as "fronted velars 
 (palatals)" before front unrounded vowels (but not front rounded 
 vowels) *and* before a diphthong with a *non-front* nucleus, as in 
 'good luck' _gaefa_ ['cai:va] and 'call (' _kaeli_ 
 ['chai:li] (with aspiration here not superscripted). [Here mostly para- 
 phrased; emphases added.] 
 To this example can be added one that responds to an alleged assimila- 
tion discussed in a recent posting by Andreas Gather: 
Agreed, German /x/ is realized as a palatal fricative after front vowels 
 and /l, n/, but palatal-fricative allophones of /x/ are also found after 
 the r-phoneme -- whose various realizations, across the range of 
 varieties and styles, may include such non-front elements as a uvular 
 approximant, a uvular fricative, or even a uvular trill. How is this 
 synchronically assimilatory, if a back -- and non-high -- (uvular) 
 segment is followed by a front -- and high -- (palatal) fricative, even 
 in the same syllable? Laboratory-phonetic studies would be of par- 
 ticular value here. 
 In addition, the facts of /l/'s phonetic realizations in English are admit- 
tedly somewhat complex (see, e.g., Sproat and Fujimura 1993 in _Jour- 
nal of Phonetics_), but the following relies on a frequently encountered 
In many varieties of English, the /l/ phoneme has a velarized allophone 
 in onsets before back vowels, but also in codas after vowels of any 
 quality (e.g., in _fill_). This is probably the elsewhere-allophone 
 (given that, for many speakers, non-velarized [l] occurs only in on- 
 sets before front vowels), but its coda occurrences hardly seem as- 
 For that matter, I don't think that Marc Picard got a convincing answer 
to his question [Linguist List Vol-10-1863] about closed-syllable laxing 
of high vowels in Que'bec French (QF), either. For one thing, while 
laxness (and tenseness) is indeed often tied up with timing issues, there 
do seem to exist long lax vowels (and short tense vowels) in at least 
some languages. And there are so-called lengthening consonants in QF 
(and other varieties of French) that can in any case outweigh syllable-
structure considerations (though not everywhere in Que'bec). 
 And, to take off from Picard's query about English flapping, which 
provoked at least one strong response arguing that this process is assim- 
ilatory: does the same answer hold for British (and American -- e.g., 
 into a following vowel, since this would, by parity of reasoning, al- 
 so be assimilatory? (It could occur before diphthongs or long vow- 
 els, for example, so that a decent syllable nucleus would still be left.) 
 In any case, even if (and I remain skeptical about this) English aspira- 
tion were to be regarded as assimilatory, the case of the final unreleased 
stops in _yep_ and _nope_ is different from a simple lack of aspiration 
(which J. B. de C. views as assimilatory -- to the unaspirated silence 
that follows [?], though silence after a voiceless stop could actually be 
considered an incredibly long VOT lag, could it not?). The late Dwight 
Bolinger once pointed out that _yep_ and _nope_ stand out in normally 
having unreleased final pronunciations of /p/ (presumably to give em- 
phasis and finality, brevity and literal close-mouthedness, etc.). One 
can legitimately ask what these allophones are assimilating to, since a 
pause doesn't require a literally closed mouth. 
 But silence does require the end of an overt message. 
 Richard Janda 
 Department of Linguistics, The Ohio State University 
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue