PHONETIC APPROXIMATION

IN SECOND LANGUAGE ACQUISITION’

James Emil Flege

University of Florida

This instrumental study of the phonetic contrast between English /ptk/ and
/bdg/ produced by Saudi Arabians reveals that native-language phonetic norms
may carry over to production of target-language sounds. Despite the existence of
phonetic interference, however, the present cross-sectional study suggests that
Saudi learners gradually approximate the phonetic norms of English, at least
insofar as several temporal acoustic correlates of stop voicing are concerned.
The Saudis’ English speech, although not typically Arabic or English in phonetic
terms, seems to be the product of a fairly stable interlanguage phonetic system
which admits the possibility of phonetic strategies by individual speakers.

It seems very reasonable to think that abstract phonological differences

between the native language (Ll) and the target language (L2) are not the
only-or necessarily the most important-determinants of how an adult
language learner will pronounce a foreign language. To be sure, the phonologi-
cal organization of L1 and L2 will influence L2 speech production, but a
learner’s mispronunciations will not always match sounds2 found in L1 and
L2. In fact, language learners frequently produce a range of different
phonetic variants (including the correct realization) for a single L2 phoneme,
sometimes even producing sounds which are not typically found in either
L1 or L2 (Berger 1951, Kruatrachue 1960, Nemser 1971a, 1971b, Johansson
1973, Dickerson 1974). Like the child learning a first language, an adult L2
learner must slowly learn to articulate unfamiliar sounds and to extend
production of already familiar sounds to new phonetic contexts. But unlike
the child, the adult learner must also modify certain well-established articula-
tory habits in order to produce L2 phonemes according to phonetic norms
of the target language.

The author would like to acknowledge Robert Port for assistance in completing
this study. Support was provided in part by NICHD grant HD125 11 to Indiana Univer-
sity and by NIH grant NS 07102 to the University of Florida.
We discuss here the production of “sounds” (i.e., phonetic “segments” or
“phones”). However, this should not be taken to imply that the phonological organiza-
tion of several allophonic variants within a single phoneme does not have important
consequences for phonetic learning. This is an important empirical question which must
be further investigated.

117
118 Language Learning Vol. 30, No. I

In previous research a relatively abstract form of “interference” which

may derive from structural differences between the sound systems of L1 and
L2 has often been the focus of investigation. This emphasis may have tended
to obscure the fact that an accurate pronunciation of foreign language sounds
often requires a language learner to make subtle articulatory adjustments,
even when cognate L1 and L2 sounds share similar phonological functions
in the sound systems of L1 and L2. For example, to correctly produce
French /y/, /p/, and Is/ an American learner must show some measure of
articulatory motor learning in all three cases. To produce /y/ the learner must
acquire articulatory motor control for an entirely new phonetic segment;
production of English / q / must be extended to a new position in French3 ;
and /p/ must be implemented with specific modifications in order to be,
phonetically, a French / P / . ~
Relatively little attention, however, has been paid to the purely articu-
latory motor learning requirements of second language learning. Results of
studies of first language acquisition by young children suggest that limita-
tions on articulatory motor control severely constrain a chdd’s phonological
output (see, for example, Menn 1979:789ff.). In an experiment which was
meant to simulate adult foreign language learning Bribre (1966 :789) found
that L2 sounds which also occurred in L1 were generally easier for subjects
to learn to produce than L2 sounds which never appeared in the phonetic
surface of L1, even though the phonological status of the similar L1 and L2
sounds was different. This finding suggests that establishment of articulatory
motor control is itself a n important part of second language learning. The
language learner, it seems, must acquire complex new sets of highly automatic
articulatory gestures or modify existing patterns of phonetic implementation
in addition to acquiring control of an abstract, reorganized phonology.
The term “interlanguage” has been used to describe a learner’s foreign
language competence (Selinker 1972, cf. Corder 1967, 1971, Nemser 1971a,
1971b, Richards 1971). The existence of an “interlanguage” seems to depend
on the assumption that a learner’s L2 output reflects an evolving linguistic
competence which is reorganized at successive stages in the learning process,

The phoneme / q / does not occur in word-initial position in English, but in con-
nected speech a word-initial sound may occur in intervocalic position. For example, in
“Maisje . . .” French / q / occurs in the same phonetic environment (intervocalic, pre-
stress) as the English / q / in “azure.” In English, however, / q / never occurs in utterance-
initial position, as it does in French “Je viens . . .” (See Brikre 1966:782, Eckman
19 77 :373)
For example, French /p/ has shorter VOT values than English /p/ (i.e., is less heavily
aspirated) and is distinguished from /b/ by virtue of its closure duration, unlike English.
 Flege 119

only gradually approximating the linguistic competence of native speakers of

L2. Even if a learner’s control of the syntactic and semantic structures of L2
is best characterized by a series of “interlanguages” one might wonder if the
same notion is equally applicable to articulatory control of L2 sounds. The
development of pronunciation skills seems slow compared to that of lexical,
syntactic, and semantic skills and does not seem to admit the possibility
for insight, induction, and cognitive reorganization as do more abstract
aspects of linguistic competence. However, the seemingly slow improvement
of pronunciation may be due in part to the physiological component of
pronunciation as well as to the nature and method of gathering pronuncia-
tion data. In much previous research, especially that done within a phonemic
theory framework,’ the L2 sounds produced by a language learner have
often been viewed as discrete entities which are produced either correctly or
incorrectly instead of as a continuum of approximations to phonetically
accurate L2 sounds (Flege 1979). This approach is perhaps due in part to
the strong perceptual tendency of human listeners to perceive speech sounds
in terms of the phonemic categories of their native language (Studdert-
Kennedy 1976) which may itself be responsible for making improvement
in pronunciation appear to be slower than it actually is.
If examined in sufficient detail, however, there is some indication that a
learner’s L2 pronunciation can be characterized as the output of a relatively
stable phonetic system which is distinct from that of either L1 or L2, that is,
as an “interlanguage.” Dickerson’s study (1974) of Japanese-accented
English, for example, provides evidence of slow but systematic progress
towards English norms of pronunciation (cf. Wolfram 1978). By making a
close phonetic transcription of the full range of phonetic approximations
to several English phonemes produced by Japanese speakers at several stages
of learning, Dickerson found evidence of gradual improvement. Both the
frequency and the number of different types of mispronunciations decreased
over a year’s time, depending on both phonetic context and speech style.
Importantly, it seemed to be the case that the least correct mispronunciations
tended to disappear first from the learner’s speech, while the closer (but still
phonetically inaccurate) approximations to L2 phonemes remained longer.
Although a careful study of L2 pronunciation which is based on phonetic

Specification of speech sounds by means of binary distinctive features will yield

distinct phonetic segments but does not provide the means to characterize phonetic
differences which are not in themselves linguistically significant. Such “subphonemic”
phonetic differences may, however, be important to the process of speech perception
and contribute to the perception of foreign accent.
120 Language Learning Vol. 30, No. 1

transcription is important and useful, this method of data gathering has limi-
tations (cf. Berger 1951, Nemser 1971b, Johansson 1973). First, an experi-
menter may not be able to transcribe with great reliability a12 the phonetic
variants produced by a language learner. This is both because the number of
transcription categories is limited (while speech is continuous) and because
even highly trained listeners may continue to be influenced by the phonemic
categories of their native language. Second, data derived by phonetic tran-
scription may lead to an understanding of how frequently and where certain
pronunciation variants may occur but cannot, in themselves, provide a
quantitative estimate of how closely a variant approximates the phonetic
norms of the target language.
In the present study we used instrumental techniques t o examine phonetic
details of the foreign language speech sounds produced by adult learners. Our
goal was t o determine if it is reasonable to characterize a learner’s phonetic
control of the sounds of a foreign language as the output of an interlanguage
rather than simply as the product of interference between the phonetic
systems of L1 and L2. If L2 pronunciation is the output of an interlanguage
we might expect to find that:
1 . the L2 sounds produced by language learners are phonetically inter-
mediate to similar sounds produced in L1 and L2 by native speakers
of those languages,
2. L2 sounds (including mispronunciations) are produced fairly consis-
tently in the same phonetic context by language learners, and
3 . individual learners may adopt different phonetic strategies to produce
L2 sounds.
The focus of our study was the phonetic contrast between English /ptk/
and /bdg/ produced by Americans and two groups of Saudi Arabians. We
focused on the stop voicing contrast because it is functionally important to
communication in English and because several of its correlates can be easily
and reliably measured. More importantly for our purposes, however, is the
fact that voiced and voiceless stops are phonetically contrasted in different
ways in Arabic and English.
In word-initial position English /ptk/ and /bdg/ are distinguished in several
ways. For example, the duration of a vowel in tab is much longer than the
vowel in rap and, at the same time, the closure duration of /pi is longer than
the closure duration of /b/ (Lisker 1957, Raphael 1972). Sounds like English
/p/ and /b/ are ordinarily distinguished by the presence or absence of glottal
pulsing (voicing) during the stop closure interval (Lisker, Abramson, Cooper,
and Schvey 1969). In Saudi Arabian Arabic, however, the contrast between
 Flege I21

word-final voiced and voiceless stops seems to depend almost exclusively on

the presence or absence of glottal pulsing. Unlike English, in Arabic there is
no duration contrast between vowels preceding voiced vs. voiceless stops nor
is there a contrast in closure duration between post-stress voiced and voiceless
stops (Flege 1979).
The phonetic contrast between voiced and voiceless stops in word-initial
position also differs between Arabic and English. In Saudi Arabian Arabic the
closure interval of a voiceless stop is longer than that of a voiced stop; this is
not so in English. It is also the case that onset of voicing (VOT) occurs
somewhat sooner after stop release in Arabic than it does in English (Flege
1979).
These language-specific phonetic differences in stop voicing between
Arabic and English provide an opportunity to determine how well adult
language learners modify the phonetic specification of a phonological con-
trast when that contrast is phonetically implemented in different ways in
Ll and L2. They also provide an opportunity to determine whether individ-
uals learn phonetic characteristics of L2 at varying rates, whether experience
speaking L2 will influence phonetic learning, and whether L2 sounds pro-
duced by adult learners actually approximate the phonetic norms of L2
rather than simply matching the phonetic norms of either the native or the
target language. The present report is part of a larger study in which Saudi
Arabians produced very comparable phonetic material in both colloquial
Arabic and English (Flege 1979). Knowledge of how Saudi Arabians imple-
mented the stop voicing contrast in Arabic provides an important point of
comparison to the production of English stop voicing by native speakers of
Arabic. In the present report, however, we present only data for English
stops produced by individual Saudi and American speakers.

Methods

Two groups of male Saudi Arabians, differing principally according to

length of residence in the U.S., and one group of Americans (six in each
group) produced English CVC words (“bat,” “bad,” “back,” “bag,” “pat,”
“tap,” “dab,” cab,” “gab”) in the carrier sentence “I say-again to Bob.”
LC

Subjects were seated before a microphone in a soundproof booth in the

Phonetics Laboratory at Indiana University, where a recording was made
using a high-quality tape recorder (Revox Model A700). Spectrograms were
made of six tokens of each test word, and measurements made by hand to
the nearest 5 msec. of several acoustic correlates of phonological voicing:
122 Language Learning Vol. 30, No. 1

vowel duration, stop closure duration, and voice-onset time (VOT). Analyses
of variance and post hoc tests of paired comparisons (Scheffk, p < .Ol) were
performed to test for differences between speaker groups. (See Flege [1979]
for details.)

Results

We found clear differences in the voicing contrast produced by Saudis and

Americans (Table l), although measured values of certain phonetic correlates
of stop voicing produced by some of the Saudi learners generally fell within
the range of values produced by native English speakers. The Saudis’ specifi-
cation of stop voicing-as reflected by measurements of four acoustic corre-
lates-showed values which were intermediate to those found in the English
produced by Americans and the Arabic produced by Saudis (Table 2), sug-
gesting that the Saudi learners of English were approximating the phonetic
norms of English.
In Figure 1 we see that voice-onset time (VOT) values for /ptk/ produced
by Americans are generally longer than VOT values produced by Saudis.
To make the mean values produced by individual speakers in the three groups
more easily comparable, the mean values produced by the speakers in each
group are ranked according to ascending value of the acoustic dimension
being considered; group means are displayed at the right. Thus in Figure 1
we see that speakers in all three groups produced a similar range of VOT
values, but that the mean VOT values for /p/, It/, and /k/ produced by
Americans (Group AM) are greater than those produced by either Saudi
group (p < .01). We see an overlap in values produced by individual speakers
in the three speaker groups, such that some of the longer VOT values pro-
duced by individual Saudis in Groups Arl and Ar2 fall within the range of
VOT values produced by Americans despite the between-group differences in
means. Even though speakers in Group Ar2 had lived in the U.S. on average
39 months compared with only 8 months for speakers in Group Arl, VOT
values produced by the two groups are not different.
A similar pattern of between- and within-group differences was found
for the vowel duration correlate of stop voicing, as seen in Figure 2. Here the
durations differences between vowels preceding /ptk/ vs. /bdg/ are displayed.
We see that in minimal pairs ending in /p-b/, /t-d/, and /k-g/ the vowel
duration contrasts produced by Americans are generally larger than those
produced by Saudis. In fact, the vowel duration contrasts produced by the
Americans reach significance in all three minimal pairs, while those produced
 Flege 123

Table 1
Mean duration of four correlates of stop voicing
produced by speakers in three speaker groups, in msec.

Speaker Group

Am A12 Ar 1

Voice-onset time
/PI in pat
It/ in tab

/k/ in cab

Vowel duration
tap vs. tab

bat bad

back bag

Final stop duration

tap vs. tab

bat bad

back bag

Initial stop duration

pat vs. bat

tab dab

cab gab

“Am” means Americans; “Arl,” Saudis who had lived less than one year in the United
States; “Ar2,” Saudis who had lived more than two years in the United States.
An asterisk (*) between two mean durations indicates that the difference between them
is significant at the .01 level by Analysis of Variance.
124 Language Learning Vol. 30, No. 1

Table 2
Mean duration of four correlates of stop voicing
produced in Arabic b y Saudi Arabians, in msec.
(Data taken from Flege 1979.)

Voice-o nset time

/p/ (none in Arabic)

It/ in /kaas/

/k/ in /kaas/

Vowel duration
- /Saab/

Final stop duration

- /Saab/

Initial stop duration

/baas/ -

ltaasl /daas/

/kaas/ /gaas/

Standard deviations are in parentheses.

An asterisk (*) between two mean durations indicates that the difference is significant
at the .01 level by Analysis of Variance.
“n.s.” indicates “not significant.”
 Flege 125

speakers

100 -
M-
ai
0

60-

40-
.
0
126 Language Learning Vol. 30, No. 1

by the Saudis do so in only half the possible cases (butlbud, Arl ;buck/bug,
Arl, Ar2; p < .01). Again, we see overlap in the size of the contrast produced
by Saudis and Americans: greatest for but vs. bud, less for buck vs. bug, and
none at all for tap vs. tub. Just as for VOT, we see that some Saudis produced
a vowel duration contrast as large as that produced by some Americans.
In Figures 3 and 4 closure duration contrasts between /p-b/, /t-d/, and
/k-g/ are displayed. For both initial and final stops the duration of a voiced
stop (/b/, /d/, /g/) is subtracted from the duration of its voiceless cognate
(/pi, it/, /k/). In Figure 3 duration contrasts between final stops are dis-
played. No duration contrast between It/ and /d/ is displayed for Americans
because they-unlike the Saudis-neutralized this duration contrast by pro-
ducing /d/ and most /t/’s as flaps. As we might expect from the fact that
English but not Arabic possesses a duration contrast between word-final
voiced and voiceless stops (Tables 1 and 2 ) we find that closure duration
contrasts produced by Americans are generally larger than those produced by
Saudis. The Americans produced a significant contrast between final /p-b/
and /k-g/. The more experienced Saudi speakers of English (Ar2) produced a
significant contrast between all three stop voicing pairs (p < .Ol), but the
less experienced Saudis (Arl) failed to produce a significant contrast between
any of the stop voicing pairs. Despite these between-group differences, how-
ever, we find that speakers within all three speaker groups produced a similar
range of values, the larger contrasts produced by individual Saudis falling
within the range of values produced by Americans.
Finally, we see a similar pattern for the duration contrasts between word-
initial /ptk/ and /bdg/. For Saudis to produce an English-like voicing contrast
between initial /ptk/ and /bdg/ they must learn to suppress the duration con-
trast used in Arabic since a duration correlate is not characteristic of the
voicing contrast between initial stops in English. Americans make both
members of the /p-b/, /td/,and /kg/ pairs about equal in duration, whereas
the Saudis make /ptk/ longer than /bdg/ (p < .01, except for tub vs. dub
produced by Arl). Despite betweengroup differences, however, we once
again see a familiar pattern of phonetic approximation: The range of values
produced by the three speaker groups is similar, and some, but not all,
Saudis produced values as small as those produced by the Americans. This
suggests that some Saudis have learned to suppress an acoustic correlate
which is appropriate for initial stops in Arabic but not Enghsh when speaking
English, at least in the controlled conditions of this experiment.
 I + + +
N N 0
0 0 0

0 31 b 0 0 31 !a o 31

Figure 2. Mean duration difference between vowels in minimal pairs ending in voiced
and voiceless stops, in rnsec. For data points above the dotted line the vowel preceding
a voiced stop is longer than the vowel preceding a hornorganic voiceless stop.
128 Language Learning Vol. 30, No. 1

It/-/d/

+60- n(/-/g/
m

*40-
 Initial Stop Contrast in msec
I 4. + a + +
h) N + N
I I B
i 0 0 0 0 0 i
1 4 4 0
? I
I

\
x I
\
1 2
s\ \
\

!
p 0. 3 o i o 31
I !
I
,
Figure 4. Mean difference in closure duration between word-initial voiced and voice-
less stopS, in msec. For data points above the dotted line the voiceless stop is longer than
its voiced hornorganic cognate.
130 Language Learning Vol. 30, No. 1

Discussion

The present data show a clear carry-over of Arabic phonetic specification

of stop voicing to the English stops produced by Saudi Arabians. Values for
several correlates of stop voicing produced by Saudis in English resemble
values for the same phonetic dimensions found in Arabic. A phonetic study
of Arabic (Table 2) revealed a duration contrast between Arabic /t-d/ and
/k-g/ in word-initial but not word-final position. The same pattern of duration
contrasts was produced by Saudis in the present study for English /t-d/, /k-g/,
and /p-b/, even though English has a duration contrast for final but not initial
stops (i.e., just the opposite of Arabic). Saudis speaking English also produced
the same relatively small duration contrast between vowels preceding voiced
‘and voiceless stops which is found in Arabic, and the same relatively short
VOT values characteristic of Arabic.
Despite the Saudis’ maintenance of Arabic-like p h h e t i c correlates of
voicing in English we found some evidence that the Saudis were approxi-
mating phonetic norms of English. The more experienced Saudis (Ar2) but
not the less experienced Saudis (Arl) seem to have acquired a duration con-
trast between word-final English /ptk/ and /bdg/. In fact, speakers in Group
Ar2 extended their duration contrast to include /t-d/ even though Americans
largely neutralize this contrast by flapping. For most of the stop voicing pairs
and most of the phonetic correlates considered in this study some speakers
from both Saudi speaker groups produced values comparable to those pro-
duced by the Americans. One could argue that overlap of the values produced
by Saudis and Americans is not really the result of a genuine, goal-directed
approximation of L2 phonetic norms by the Saudis, but is simply the reflec-
tion of a normal distribution of values t o be expected of any measurement of
speech produced by agroup of speakers. However, the group means produced
by both Saudi groups do represent a clear departure from mean values found
in Arabic (Table 2) and lie in the direction of English values.6 The values
produced by the more experienced Saudis (Ar2), moreover, are generally
closer to values produced by native English speakers than are those of the
relatively less experienced Saudi speakers of English (AT 1).
The present study supports the notion that language learners may adopt

An apparent exception are VOT values, which are somewhat shorter in Arabic-
accented English than in Arabic instead of falling somewhere between the mean VOT
values which are characteristic of English and Arabic. The Saudis seem to have decreased
VOT in English (vis-8-vis Arabic) to compensate for the relatively long duration of their
word-initial English stops (as compared to Arabic stops). See Flege (1979) for details.
 Flege 131

individual strategies to produce phonological contrasts. Inspection of Figures

1 4 reveals that it was not always the same Saudis who produced the most
English-like values for VOT, vowel duration, and stop closure duration.
For example, speaker 5 in Group Ar2 produced VOT values for /p/, It/, and
/k/ which are as large as values produced by some Americans. Since this same
speaker also succeeded in suppressing a non-English duration contrast be-
tween word-initial /ptk/ and /bdg/, we might infer that his attention was
directed more closely to the phonetic contrast between word-initial rather
than word-final tops, an interpretation which is supported by the fact that
\
Saudis on the wh le were somewhat more successful at producing English
temporal correlates of voicing for final stops than for initial stops.
The existence of phonetic strategies might also be inferred from a compari-
son of how the voicing contrast was produced by the two groups of Saudis.
The relatively inexperienced Saudis (Ar 1) produced a significant vowel
duration contrast in but vs. bud but failed to produce a duration contrast
between /t-d/ in the same minimal pair. Conversely, the relatively experienced
Saudis (Ar2) produced a stop duration contrast but no vowel duration con-
trast in but-bud.
Thus, although we see a similar range of approximation to English tem-
poral correlates of voicing by individual Saudis, neither individuals nor groups
of Saudi speakers show the same degree of approximation to English values
for each of the voicing correlates examined. This suggests that the object of
phonetic learning during second language acquisition may be an ensemble of
discrete articulatory gestures which together represent the phonetic specifica-
tion of speech sounds in L2. This interpretation is supported by the finding
of Flege (1979) that although the Saudis in this study generally applied the
temporal correlates of stop voicing found in Arabic to their production of
English /p/ (a phoneme absent from Arabic) they were not successful in using
glottal pulsing (a correlate used to contrast Arabic /t-d/ and /k-g/) to contrast
English /p-b/.
Results of this experiment also indicate that a difference in phonological
inventories between Arabic and English has affected the Saudis’ specification
of English stop voicing. Inspection of Figures 1 4 reveals that fewer of the
Saudis produced phonetic correlates having values which fall within the
range of values produced by Americans for /p/ and /p-b/ than for other
stops and voicing contrasts. Thus it seems that the absence of /p/ in Arabic
affects how the Saudis produced English /p/. In this regard it is interesting to
consider the /p-b/ contrast produced by speaker 1 in Group Arl. T h s
speaker, like most of the other Saudis, generalized the duration contrast
132 Language Learning Vol. 30, No. 1

between It-d/ and /k-g/ found in Arabic to the English /p-b/ contrast. But in
so doing he exaggerated the size of this voicing correlate, producing a much
later duration contrast than any other Saudi or American in both word-
initial and word-final positions. He did not, however, produce a similar
exaggeration for the English /t-d/ and /k-g/ contrasts, suggesting that it is
the relative novelty of /p/ which causes this adult language learner to search
out relevant phonetic correlates to English stop voicing.
A tendency t o exaggerate just one of the several phonetic correlates of
a distinctive feature (such as voicing) may be due in part to its degree of
perceptual saliency or ease of articulation, or to phonetic differences be-
tween L1 and L2. Once language learners become “aware” (at some level
of processing) of phonetic differences between their present phonetic output
and that of the target language they may attempt to modify their pronuncia-
tion either by exaggerating a phonetic dimension they already control or by
exaggerating one which they have discovered to be a part of the target-lan-
guage phonetic system. It remains an important empirical question for
applied phonetic research to determine whether language learners can modify
phonetic specification of sounds without first exaggerating some phonetic
dimension, and whether such exaggeration may actually facilitate the process
of phonetic learning.
Whatever the cause of speaker-specific phonetic strategies of implementa-
tion it seems likely that their function is to insure intelligibility. The speaker
who exaggerates the /p-b/ contrast in duration, for example, probably does so
because his or her /p/ has been mistaken for /b/ by American listeners. In a
listening test based on the speech samples discussed here Americans perceived
about half. the word-final /p/’s produced by Saudis as /b/ (Flege 1979). Thus,
whether or not the Saudis produce an acceptable /p/ or implement the
English stop voicing contrast just like Americans, it is probably true that
they are aware of phonetic differences between their English and that of
native speakers and are attempting to modify their phonetic implementation
of stops.
Although many people seem t o accept the notion that adults and children
differ in their ability to modify existing phonetic patterns, the present study
points to several similarities between child first-language acquisition and
adult second-language learning. First, studies of child language reveal in-
stances of the exaggeration by child learners of a f i s t language of certain
phonetic dimensions found in the adult target language (see, for example,
Macken and Barton 1977, Simon 1978). Second, children (like adults in
this study) may adopt individual phonetic strategies to produce speech
 Fiege 133

sound contrasts. For example, both Ingram (1975) and Simon (1978) report
data which indicate that chldren may use phonetic dimensions not ordinarily
exploited by adult speakers of the target language. And third, both children
and adults seem to approximate new phonetic dimensions. In a longitudinal
study Macken and Barton (1977) showed that some chldren acquiring
English as a first language only gradually approximate the VOT values pro-
duced by adult English speakers. At some point in the phonetic learning
process children may produce VOT differences between voiced and voiceless
stops which, although significant, would probably not be sufficient in them-
selves to cue the voiced-voiceless distinction for adult English-speaking
listeners. Thus, adults-like children-may be aware of the importance of
certain phonetic dimensions but not be able to produce them reliably.

Conclusions

Phonetic specification of speech sounds in a foreign language seems to

be the output of an “interlanguage” rather than simply the product of
interference between L1 and L2. Although a Saudi’s contrast between English
/ptk/ and /bdg/ resembles the phonetic contrast between Arabic voiced and
voiceless stops, some Saudis approximate English phonetic norms for stop
voicing more closely than others, although not all the correlates produced
by a single speaker are equally English-like. The present study suggests that
individual adult learners of a foreign language may modify pre-existing
phonetic patterns at somewhat differing rates, make slow progress
in acquiring the phonetic norms of a target language, and adopt somewhat
different phonetic strategies for producing new or phonetically different
sounds in L2. Thus, adult foreign language learners appear to resemble
child learners of a first language, at least in regard to their general pattern
of phonetic learning.

REFERENCES
Berger, M.D. 195 1. The American English pronunciation of Russian immigrants. Unpub-
lished Ph.D. Thesis, Columbia University (University Microfilm No. 51-3872).
BriBre, E. 1966. An investigation of phonological interference. Language 42.768-798.
Corder, S.P. 1967. The significance of learners’ errors. IRAL 5.161-170.
Corder, S.P. 1971. Idiosyncratic dialects and error analysis. IRAL 9.
Dickerson, L.J. 1974. Internal and external patterning of phonological variability in
the speech of Japanese learners of English: toward a theory of second-language
acquisition. Unpublished F’h.D. Thesis, University of Illinois.
134 Language Learning Vol. 30, No. 1

Eckman, F. 1977. Markedness and the contrastive analysis hypothesis. Language Learn-
ing 27.315-330.
Flege, J.E. 1979. Phonetic interference in second language acqu
Ph.D. Thesis, Indiana University.
Fowler, C.A. 1977. Timing Control in Speech Production. Bloomington: Indiana Uni-
versity Linguistics Club.
Ingram, D. 1975. Surface contrast in children’s speech. Journal of Child Language
2.281-292.
Johansson, F.A. 1913. Immigrant Swedish Phonology, A Study in Multiple Contact
Analysis. Travaux de 1’Institut de linguistique de Lund, 9. Lund: CWK Gleerup.
Kruatrachue, F. 1960. Thai and English: a comparative study of phonology for peda-
gogical purposes. Unpublished Ph.D. Thesis, School of Education, Indiana University.
Lisker, L. 1957. Closure duration and the intervocalic voiced-voiceless distinction in
English. Language 33.42-49.
Lisker, L., A. Abramson, F. Cooper, and M. Schvey. 1969. Transillumination of the
larynx in running speech. J. Acoust. Soc. A m . 45.1544-46.
Macken, M., and D. Barton. 1977. A longitudinal study of the acquisition of the voicing
contrast in American English word initial stops, as measured by VOT. Papers and
Reports on Child Language Development 14.74-121.
Menn, L. 1979. Towards a psychology of phonology: child phonology as a first step.
Proceedings of the Conference on Applications of Linguistic Theory in the Human
Sciences. Department of Linguistics, Michigan State University.
Nemser, W. 1971a. Approximative systems of foreign language learners. IRAL 9.1 15-
123.
Nemser, W. 1971b. A n Experimental Study of Phonological Interference in the English
of Hungarians. Indiana University Series in Uralic and Altaic Studies, No. 105,
Bloomington, Indiana.
Port, R., and F. Mitleb. (In Press) Phonetic and phonological manifestations of the
voicing contrast in Arabic-accented English. J. Applied Psycholinguistics 1.
Raphael, L. 1972. Preceding vowel duration as a cue to the perception of the voicing
characteristic of word-final consonants in American English. J. Acoust. SOC. A m .
51.1296-1303.
Richards, J. 1971. Error analysis and second language strategies. Language Sciences
17.12-22.
Selinker, L. 1972. Interlanguage. IRAL 10.209-231.
Simon, C. 1978. Stop voicing in English and French monolinguals: some developmental
issues and experimental results. Paper delivered a t the NICHD conference on Child
Phonology, Bethesda, Maryland, May, 1978.
Studdert-Kennedy, M. 1976. Speech perception. In Lass, N. (ed.), Contemporary Issues
in Experimental Phonetics, 234-295. New York: Academic Press.
Wolfram, W. 1978. Contrastive linguistics and social lectology. Language Learning
28.1-28.