WORD
ISSN: 0043-7956 (Print) 2373-5112 (Online) Journal homepage: https://www.tandfonline.com/loi/rwrd20
On the Description of Phonic Interference
Uriel Weinreich
To cite this article: Uriel Weinreich (1957) On the Description of Phonic Interference, WORD, 13:1,
1-11, DOI: 10.1080/00437956.1957.11659624
To link to this article: https://doi.org/10.1080/00437956.1957.11659624
Published online: 04 Dec 2015.
Submit your article to this journal
Article views: 1016
View related articles
Citing articles: 1 View citing articles
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=rwrd20
WORD==
VoLuME 13 April, 1957 NuMBER 1
ON THE DESCRIPTION OF PHONIC INTERFERENCE
ORIEL WEINREICH
In Chapter 2.2 of my book, Languages in Contact (New York,
1953; hereinafter LiC), I attempted to outline a plan for the
description of phonic interference in its many aspects. That
attempt was, I believe, not completely successful. The revised
plan proposed here is based on further consideration of the problem,
in which I have benefited particularly from discussions with
Messrs. William J. Nemser (Columbia University) and Einar
Haugen (University of Wisconsin) and from the latter's recently
published work, including his able and constructive review of my
book. 1
By phenomena of INTERFERENCE I mean those instances of
deviation from the norms of a language which occur in the speech
of a bilingual as the result of his familiarity with another language,
i.e. as a result of language contact (LiC, p. 1).
The problem of phonic interference concerns the manner in
which a speaker perceives and reproduces the sounds of one
language in terms of another. The language which causes the
interference is called primary (P); the language which suffers the
interference is designated secondary (S). P-Spanish means Spanish
as the primary language in a conta-ct situation; S-English, English
as the secondary language. We are examining foreign "accent"
in a broad sense--namely, language S spoken with a P-accent
(LiC, p. 14).
1 See Haugen's review or LiC in Language 30.380-8 (1954); "Problems or Bilingual
Description," Report of the Fifth Annual Round Table Meeting on Linguistics and
Language Teaching (Georgetown University Monograph Series on Languages and
Linguistics 7), 1954, 9-19; "'Problems of Bilingual Description," General Linguistics
1.1-9 (1955); Bilingualism in the Americas: a Bibliography and Guide to Research,
Publkations or the American Dialect Society n°. 26, 1957.
2 URIEL WEINRECH
In reconsidering the presentation of LiC, it seems useful to
make a forthright distinction between SYNTAGMATIC and PARA-
DIGMATIC factors in phonic interference. The former pertain
to relations between sounds in a sequence, i.e. in the spoken
chain. The latter concern relations between sounds in the pattern,
i. e. sounds which might occur at a given point in the spoken chain.
It is useful, furthermore, to distinguish between the bilingual's
ANALYSIS of the material in the secondary language and his
RENDITION of that material. One of the faults of the treatment
in LiC is the inadequate separation of these two planes. Of
course, it is possible to rend£r unfamiliar sounds well by pure
mimicry, yet such rendition without analysis hardly leads to pro-
ductive speech. The bilingual's analysis in the present discussion
is analogous to that process which enables us not merely to repeat
a series of numbers like 1, 4, 9, 16, 25, ... , but to continue it
consistently according to some hypothesis as to the formula.
Phonic interference may be due to factors on four levels:
I. Properly PHONIC FACTORS. This includes differences in the
stocks of phonemes of the languages in contact, in the compo-
nential analysis and in the distributional patterns of their
phonemes.
II. ExTRA-PHONIC FACTORS within the languages. Here we
may consider cases of interference (or, on the contrary, suppression
of interference) aimed at the avoidance of a particularly undesir-
able homophony, such as might cause serious ambiguity,
unwanted allusions to taboo words in either language, and the
like. These lexical considerations function as a control of inter-
ference in specific words and have nothing to do with the sound
systems as such. They are analogous to the semantic factors
which may occasionally upset an otherwise regular sound change.
Possibly the influence of spelling on the pronunciation of the
secondary language should also be classified in this category. 2
III. ExTRA-LINGUISTIC FACTORS. These comprise the moti-
vations for achieving intelligible, acceptable, or even native-like
• In LiC the influence of spelling is discussed only in connection with the form
in which loan words from language S are integrated in h:mguage P (p. 28), but spelling
as a model for the pronunciation or language S as such, is not dealt with. Yet in the
pronunciation or S-English in India, for example, the spelling has been ovcrwhelm::-
ingly influential. Cf. Ashok Kelkar's study of the English pronunciation of' speakers
of P-Marathi in a forthcoming issue of this journal.
ON THE DESCRIPTION 01' PHONIC INTERFERENCE
speech that are present in a concrete speech situation or in the
general sociocultural setting in which the language contact takes
place. (LiC, § 1.3, pp. 25, 67, and chapters 3 and 4.. )
IV. In any honest description we may also expect a residue
of ERRATIC CASES of interference (see Haugen, Language 30.384).
These are perhaps equivalent in status to the unaccountable
kind of slips of the tongue that a unilingual speaker makes.
Here I will be concerned only with properly phonic factors ( 1).
Syntagmatic Factors
SEGMENTATION. A given sequence of speech sounds may be
segmented differently by the speakers of different primary lan-
guages. Thus, a speaker of P-Hindi would segment the English
sequences [redhijr] adhere and [klabhaws] clubhouse as containing
the unit phonemes /@, ~f. The [ cc ·] of English bad might be
segmented as a cluster, freref, from the point of view of P-Finnish.
Similarly, the sequence [ts] at the end of English words might
-morphological considerations apart--be analyzed as a unit
phoneme by a speaker of P-German. True, if such phonemic
metanalysis is possible without leaving a remainder of segments
unaccounted for, the rendition of the S material by the foreigner
may be quite native-like in this respect. But it is submitted here
that the difference of analysis is a legitimate object of the linguist's
interest even when rendition is not affected adversely.
FAMILIAR SEQUENCEs. It is often observed that bilinguals,
in mispronouncing words of S, adapt unfamiliar sequences of
sounds to sequences more like those which occur in P. Thus,
the sequence fekst/ as in English extra is often simplified to fest/
by speakers of P-Italian, where the long cluster in not familiar
(LiC, p. 23). The implication of ~his for the description of phonic
interference is that the unit of interlingual identification should
not invariably be the phoneme (LiC, pp. 7, 14), but occasionally
a sequence of phonemes. On the other hand, in the interests
of econominal description no precise length of this sequence
should be specified; it should simply be a possibility held open,
like the possibility of occasionally equating not words, but stretches
of words ("idiomatic constructions") in a bilingual dictionary.
The whole matter of unfamiliar sequences is treated too casually
in LiC (p. 23) as a structural factor which merely "favors" inter-
ference.
1-1
UIUEL WEINRECH
Obviously a sequence of sounds may be familiar in language P,
but only in positions limited with respect to a morpheme or word
boundary. Thus, hn is a sequence common in Ukrainian which
does not occur anywhere in English; but the -vi- of an Indonesian
word is a sequence which, though it does not occur initially in
English, is quite common word-medially (singing). A sequence
in an unfamiliar position, as we all know, often strikes us as no
less exotic than a completely unfamiliar one. The relative value
of total and "positional" familiarity in governing interference is
capable of systematic investigation.
In general it is not yet clear, at the present stage of research,
what methods may profitahry be used in generalizing about classes
of sequences. We may anticipate a technique by which it will
be possible to say that or- in some language would be less liable
to interference from P-English than mr- because, though neither
cluster occurs initially in English, the former does approximate
occurring English clusters more closely. After all, there may
be holes in distributional patterns, too. Related to this point is
the possibility of expressing certain aspects of the combinability
of phonemes in quantitative terms, as established by the recent
work of Herbert H. Paper. 3 His procedures will enable us, by
comparing the distribution indices of pairs of equivalent phpnemes
of P and S, to make quantitative statements of difference of
distributional structure of the two languages. However, the
description of particular instances of interference will still have
to refer to discrepancies in specific familiar sequences in the
languages in contact.
It would seem that when we deal with unfamiliar sequences of
subphonemic segments (and perhaps in other cases, too), an
analysis is possible in both syntagmatic and paradigmatic terms;
cf. the chapter on the intersection of these two dimensions, p. 8
below.
A particular type of sequence discrepancy is that in which
order alone is involved. Metatheses of sounds in the integration
on unfamiliarly shaped loanwords are well documented in histor-
ical linguistics. It is also possible to conceive of cases in which
a segment of sound in language P is distinctive but the place of
its occurrence is not significant. This might lead to a mutilation
of sound sequences of S such that all the phonemes are rendered,
3 F. Harary and II. II. Paper, "Towards a General Calculus of Phonemic Distri-
bution", to appear in Language, vol. 33, n•. 2 (1957).
ON THE DESCRIPTION OF PHONIC INTERFERENCE 5
but not in the right order. A real if somewhat marginal instance
is that of a child which, on learning to pronounce r between a
consonant and a vowel at the age of 22% months, began, in
polysyllabic words requiring CrV, to insert the new sound in any
syllable at random; for a period, professor was likely to turn up
half the time as pofressor, and breadbox alternated with beadbrox.
Whether a corresponding phenomenon occurs systematically in
non-child speech is open to question; at any rate many of us from
our fieldwork know the feeling of identifying the constituent
sounds of a word in a strange language, but not being sure of the
order in which they were said.
Paradigmatic Factors
We consider first the paradigmatic aspects of the bilingual's
ANALYSIS of a sound. It is convenient, as Haugen has shown,
to deal with the allophone first, and the term 'sound' in the
following paragraphs is to be interpreted as referring to an allo-
phone, i.e. a contextual variant of a phoneme.
A sound of a given language S, e.g. the [b) of Spanish, has
many features: [b) is 1. voiced (rather than voiceless), 2. oral
(rather than nasal), 3. fricative (rather than occlusive), 4. apical
(rather than medial or dorsal), 5. dental (rather th;m interdental
or alveolar or postalveolar or cacuminal), 6. of a certain average
duration (rather than longer or shorter than that), 7. non-
pharyngalized (rather than pharyngalized), and so on. From
the point of view of Spanish, some of these features, viz, 1, 2,
and 4, are distinctive, while the rest are redundant, i.e. automati-
cally predictable. In an analysis of the sound l b] conforming to
the phonemic pattern of a diiTerent language, say of P-English.
feature 3 would also be distinctive. If P were Arabie, 6 and 7
might be distinctive in addition. ·Whenever a bilingual elassifies
a feature of sound as distinctive which in the original S-language
is redundant, we may speak of OVER-DIFFERENTIATION. In the
opposite case-as when a speaker of P-Spanish analyzes the
fricativity of an intervocalic English [b) as redundant-we have
UNDER-DIFFERENTIATION.
The advance represented by this statement compared to that
in LiC (p. 18) lies in the realizations that (a) over- and under-
diiTerentiation are matters of distinctive features rather than of
phonemes, since conceivably in the bilingual's analysis of a single
phoneme there may be both an under-diiTerentiated feature and
6 URIEL WEINRECH
an over-differentiated one: a speaker of P-French might consider
the lip rounding of [o]. in Russian [t'6t'a] 'aunt' as distinctive
(over-differentiation) and its stress as redundant (under-differen-
tiation); (b) over- and under-differentiation are matters of analysis
rather than rendition; even under-differentiation may lead to
impeccable renditions. In the previously cited example, the
speaker of P-Spanish may under-differentiate the feature of
fricativity which distinguishes English fd/ from fbf, but inter-
vocalically he would render a satisfactory [ b] nevertheless.
A situation of a special sort arises when language P classifies
a distinctive feature of a given sound of S as redundant, but it
does possess this distinctive f~ature in others of its own sounds.
Thus the voicing and the laxness of S-Romansh fzf are redundant
from the point of view of P-Schwyzertutsch phonemics, but the
laxness at least is distinctive in such other Schwyzertiitsch sounds
as /b, d, vf, etc. This may be conducive to a more correct analysis
of Romansh fzf by bilinguals than, say, of Romansh JA.f, whose
feature of palatality is not distinctive for any sound of P-Schwy-
zertiitsch. It is this which was mentioned as a factor of para-
digmatic structure that inhibits interference, and was related to
the concept of holes in the pattern, in LiC (p. 23). We may now
state more precisely that under-differentiation of a distinctive
feature may affect all, or only some, of the phonemes of the
S-language which are characterized by that feature.
If in a language S a certain distinctive feature is associated
with great regularity with a certain redundant feature (e.g.
backness and lip-rounding of [u]), and if another language, P,
manifests the same association except that the classification of
the two features as distinctive and rendundant is reversed, the
two sounds may be effectively identified astraddle of the languages,
producing an elementary case of reinterpretation of features
(cf. also p. 9 below). A workable reinterpretation should lead
to renditions which are faultless with respect to the features
concerned.
When we pass on from questions of analysis to those of RENDI-
TION, our point of view shifts. What we care about here is whether
a feature (be it distinctive or not) is always present in a sound,
or whether it occurs only occasionally as a matter of free variation,
or never. If the presence of a feature in a sound of S is a matter
of free variation, any treatment of it by the rules of another
language, P, ought to yield faultless renditions, whatever those
rules may be, since by definition the presence or absence of the
ON THE DESCRIPTION OF PHONIC INTERFERENCE 7
feature does not matter for S. Also, if a feature is always present
in the equivalent sound of P, the rendition will be satisfactory,
whatever the regularity or irregularity of this feature in the
corresponding sound in S. On the other hand, errors of varying
predictability may be expected in other cases. For a given
feature of the equivalent sounds of lan.,guages P and S, the results
are as follows if the feature is present:
IN TilE SOUND 1:'11 TilE SOUND OF LANGUAGE P
OF LAl'OGUAGE s Always Sometimes (free variation) Never
Always + Unpr. err. Pred. err.
Sometimes (free variation) + + +
Never Pre<l. cr1·. Unpr. err. +
+ = Satisfactory rendition
Prell. err. = Predictable errors
Unpr. err. = Unpredictahlc rrrors
For example, in S-English, final Jp/ is sometimes exploded, at
other times not; the feature of release is a matter of free variation.
In the English' speech of a bilingual in whose P-language the
release of final Jp/ is regular, a released [p] would perhaps be more
frequent than in the speech of an English unilingual, but there
would be no perceptible interference. In the reverse case, in
which English is the P-language, there would be unpredictable
errors on this point, since the bilingual would treat the release
of [p] as a matter of free variation contrary to the rules of the
S-language. Examples of the other possibilities are easily
invented.
At this point it is quite feasibl~ to proceed from allophones
(contextual variants) to a statement of how the PHONEME AS A
WHOLE is rendered by the bilingual. We then necessarily obtain
an additional class of errors, namely those which are predictable
or unpredictable of the phoneme in a particular context. For
example, if a speaker of a language P in which voicing of Jb/ is
a matter of free variation, were to render forms in S-German
(where voicing of Jbf is always present in intervocalic position),
we might expect errors in the rendition of Jbj in some contexts
(namely intervocalically) but not finally.
8 URIEL WEINRECH
Haugen's approach to the description of interference difTers
from the one proposed.here chiefly in that, while he occasionally
comments on the relative importance of a feature to a given lan-
guage (Norwegian Language in America, p. 432), he does not lake
the componential analysis of all phonemes as the basis of description.
But whatever role is assigned to the distinctive feature in analy-
zing the allophones of a given language, when the cross-language
equivalences between allophones have been established-by the
present method or by any other-there can be no objection to
the adoption of Haugen's procedure for classifying them into "dia-
phones" (interlingually identified phonemes) of various types:
convergent and divergent, sifilple, compound and complex. (See
Haugen's latest formulation in Bilingualism in the Americas,
§ 3.3.3.)
Intersection of Syntagmatic and Paradigmatic Factors
There are cases of interference in which syntagmatic and para-
digmatic factors intersect.
One type of intersection occurs when a phoneme of S has an
allophone which is subject to further segmentation in terms of
language P. Thus, Tamil fef has, in word-initial position, the
allophone [je ]. Faced with the English sequence [jel] yell, the
speaker of P-Tamil would resegment it so that the first two
segments would together constitute the word-initial variant of
his fef. Faced with felf ell, he would probably note a strange
initial allophone of fef and, in rendering it, he would supply the
additional "suballophonic" segment [j]-. From a syntagmatic
point of view, we have a difference in segmentation in the two
languages. Paradigmatically, we might speak of a discrepancy
in the regularity with which the "feature" of j-onset occurs-never
in English, always initially in Tamil. A bilingual's treatment
of discrepant functioning of subphonemic glides between vowels,
epenthetic vowels and consonants, and similar phenomena would
likewise involve a choice of syntagmatic and paradigmatic terms
in the description of the interference.
Another category of intersection is that in which a special
segment is inserted into S forms by the bilingual to manifest a
feature which in P does not occur distinctively in a particular
sound. Thus, a type of P-Ukrainian in which a distinctively
voiceless /f/ does not occur causes bilinguals to render [xv] for the
/-sound of secondary languages. The bilingual unwraps, as it
ON THE DESCRIPTION OF PHONIC INTERFERENCE 9
were, the bundle of simultaneous features which constitute [f)
and sets up a special segment for the display of voicelessness.
Conversely, the rendition of a labialized [kw] by a speaker of a
P-language which has distinctive labialization, in lieu of the
sequence [kw] of an S-language which does not, trades a segment
for a simultaneous distinction.
Perhaps the most interesting variety of intersection occurs
when a distinctive feature is transferred to another, existing
segment of S: it is taken out of one bundle of features and wrapped
in another. Thus the Romansh sequence [es ·] as in Messa 'Mass'
contains -a consonant which is distinctively long (or geminated)
and a vowel which is redundantly short, but whose shortness is
required by what follows. The same sequence may be interpreted
in terms of a neighboring type of Schwyzertiitsch as consisting
of a vowel which is distinctively short and a consonant redundantly,
but automatically, long. This is what was called "reinterpretation
of features" in LiC (pp. 18f.), but it is a more complex case than the
elementary one outlined above (p. 6), since a phoneme boundary
is crossed here. The features of each of a pair of phonemes are
reinterpreted· as to distinctiveness, and in addition, the two
allophones continue to imply each other despite the reinterpre-
tation. Another example is the opposition, in S-Arabic, of
faq-akf [aq-ak]; a speaker of P-French might consider the
frontness of the vowel distinctive rather than the pharyngalization
of the consonant: fak-akf. The parallelism with the concept
of the preservation of useful sound features will be apparent to
all who are familiar with diachronic phonemics.
Inference in Prosodic Features
Interference in prosodic features was trtated in LiC in a some-
what too fragmentary manner (pp. 17, 18, 38f.). In a more
systematic analysis, several types of stress system and the possible
interferences should be examined. We leave aside the bilingual's
problems that bear on the use of stresses and stress contours as
meaningful, morpheme-like units; only the distinctive function
of stress is touched on here. 4
We may consider languages which have culminative stress,
• On the conflicting functions of stress in a single language, see my paper, "Stress
and Word Structure in Yiddish," in The Field of Yiddish: Studies in Language, Folklore
and Literature, New York, 1954, 1-27.
10 URIEL WEINRECH
i.e. a stress which occurs in a determined but unpredictable place
in each word; demarcative stress, i.e. a stress which occurs
automatically in a determined place in each word in a position
related to the beginning or the end of the word (final, pre-final,
initial, etc.); and non-distinctive stress, i.e. a stress which plays
no role in identifying or segmenting words and which may occur
at any point of a word, without any necessary consistency.
Examples of languages with culminative stress are English,
Russian; demarcative stress occurs in Latin, Czech, etc.; non-
distinctive stress is the rule in French.
If the P-language has culminative stress, we may expect the
bilingual to pay attention to the stresses of the S-language and to
render them correctly. If he has not attained the automaticity
of stress placement which is possible in an S-language with demar-
cative stress, or if he places stresses consistently in an S-language
in which they are not distinctive and need not be consistently
placed, there is a lack of economy analogous to the over-differen-
tiation discussed under the heading of paradigmatic factors; but
the rendition may be perceived as faultless.
If the bilingual is not accustomed, by the rules of P, to place
the stress on a separately determined part of each word, errors in
rendition may be expected. When the required place of the
stress of a form in S happens to coincide with the automatic place
of demarcative stress in P, the rendition may be adequate by
accident; otherwise there will be faulty stressing. For example
if a speaker of P-Czech {demarcative initial stress) were to apply
his habits to S-English, he would happen to get discipline right
but disciple wrong (*disciple). A speaker of P-French, if he
treated the stress of S-English as if it were non-distinctive, would
sometimes happen to render words correctly, but according to no
predictable formula. The same word might be produced with
initial stress once and final stress at another time. Thus, it is
clear that there will be considerable errors in the rendition of S if the
stress system of Sis more rigorous than that of P(the progression
non-distinctive-demarcative-culminative being treated as increas-
ing in rigor). Also, the place of likely errors may ·be predictable
if either P or S or both operate with demarcative stress.
This is a mere outline of a plan for the description of interference
in stress. The bilingual's treatment of secondary and lesser
stresses should be analyzed in a similar way. The demarcative
• See Andre Martinet, "Accents et tons," Miscellanea phonelica 2.13-24 (1954).
ON THE DESCRIPTION OF PHONIC INTERFERENCE 11
type of stress may be divided further into word-demarcative and
stem-demarcative; and still other refinements may be introduced. 5
But no purpose would· be served by simply listing all the possible
categories of interference; the analogy with the description of
segmental-phoneme interference makes it clear how this classi-
fication could be extended.
Interference in other prosodic features, particularly tone
systems of various types, remains to be described on related
principles.
Finally, the analogy between interference in segmental pho-
·nemes and in prosodic features could be pursued further by
introducing such concepts as "allophones of stress." But no
terminological exercise is likely to make the already apparent
parallelisms any clearer, nor to eliminate the ultimate dissimilari-
ties between segmental phonemics and prosody.
Columbia University.
