Flege, J. and MacKay, I (2011). What accounts for “age” effects on overall degree of foreign accent? In M.
Wrembel, M. Kul and Dziubalska-Kołaczyk, K. (Eds) Achievements and perspectives in the acquisition of second
language speech: New Sounds 2010, Vol. 2, Bern, Switzerland: Peter Lang, Pp. 65-82.
Flege, J. (1995) The Phonetic Study of Bilingualism. In G. Bloothooft, V. Hazan, D. Huber & J. Llisterri (Eds)
European Studies in Phonetics and Speech Communication. Utrecht: OTS Publications, Pp. 98-103.
Flege, J. (1992) Speech learning in a second language. In C. Ferguson, L. Menn, & C. Stoel-Gammon (Eds)
Phonological Development: Models, Research, and Application. Timonium, MD: York Press, Pp. 565-604.
Here you can download book chapters dealing with second-language (L2)
speech learning that were published by James Emil Flege & colleagues
Piske, T., Flege, J., MacKay, I., and Meador, D. (2011). Investigating native and non-native vowels produced in
conversational speech. In M. Wrembel, M. Kul and Dziubalska-Kołaczyk, K. (Eds) Achievements and
perspectives in the acquisition of second language speech: New Sounds 2010, Vol. 2, Bern, Switzerland: Peter
Lang, Pp. 195-205.
Mora, J., Keidel, J. and Flege, J. (2011). Why are the Catalan contrasts between /e/-
/eh/ and /o/-/oh/ so difficult for even early Spanish-Catalan bilinguals to perceive? In M.
Wrembel, M. Kul and Dziubalska-Kołaczyk, K. (Eds) Achievements and perspectives in
the acquisition of second language speech: New Sounds 2010, Vol. 2, Bern,
Switzerland: Peter Lang, Pp. 183-193.
Mora, J.C., Keidel, J.L. & Flege, J.E. (2015) Effects of Spanish use on the production of
Catalan vowels by early Spanish-Catalan bilinguals. In J. Romero & M. Riero (Eds.) The
Phonetics-Phonology Interface, Representations and methodologies. Amsterdam: John
Summary. Two studies by Mora et al. (here 2015, below, 2011) examined the
production and perception of Catalan vowels by individuals who had learned both
Spanish and Catalan by school age. Previous research focusing on immigrants who
learned an L2 after arriving in a predominantly L2-speaking environment have shown
that some early learners continue to pronounce the L2 with a foreign accent and may
produce certain L2 vowels inaccurately. Flege & MacKay (2004) focused on the
perception of English vowels by early Italian-English bilinguals who had lived in
Canada for decades. These authors noted inaccuracy in the perception of L2
(English) vowels by early learners who continued to use the L1 (Italian) frequently but
not by those who used the L1 infrequently.
But what about individuals who use both the L1 and the L2 frequently in a city like
Barcelona, where bilingualism is the norm? Sebastan-Gallès & Soto-Faraco (1999, p.
120) observed errors in vowel perception by early learners in Barcelona, which the
authors interpreted as possible evidence for a "lack of plasticity". Further, the
authors suggested that the malleability of the speech perception system may be
limited "severely" by school age because exposure to the L1 exerts a "very strong
constraint" on the "organization and acquisition of phonemic categories". Pallier et al.
(1997, p. B14) concluded that even early and frequent exposure to an L2 might be
insufficient to permit the learning of "two new phonetic categories which overlap" a
single L1 category. Bosch et al. (2000, pp. 215-216) inferred that early learners
continued to represent Catalan vowels as "foreign" speech sounds for which "stable
representations in long-term memory" were not established.
The results obtained by Mora et al. (2011, 2015), on the other hand, do not support
the view that severe limits exist on the ability of early learners to produce and
perceive vowel distinctions found in one of their two languages, but not the other.
The research by Mora et al. confirmed the existence of differences in accuracy
among bilinguals, but showed that the variation is largely determined by patterns of
Flege, J. (2009). Give input a chance! In T. Piske and Young-Scholten, M. (Eds) Input Matters in SLA. Bristol:
Multilingual Matters, Pp. 175-190.
Key words: input, LOR, length of residence, AOA, age of arrival, age, age-related, individual differences,
longitudinal study, long-term memory representation, early learners, late learners, foreign accent, sound
system, amount of input, quality of input, Polish, Chinese, language use estimates, percentage L1 use,
principle components analysis, language background questionnaire, perceived foreign accent,
pronunciation, immigrants, language dominance, neurological development, phonetic category,
native-speaker input, foreign-accented input, L1 proficiency, confounds, confounds, self-report, Experience
Summary. This chapter evaluated the role of input in L2 speech acquisition through a reexamination of data
obtained in large (n = 240) studies examining degree of foreign accent in English sentences spoken by
Italian and Korean immigrants to North America. The two studies yielded very similar results despite a
difference in speech materials and L1. The author notes that one common index of input, length of residence
(LOR), directly accounted for fairly little variance in part because it is not an actual measure of input. The
predictive strength of the immigrants' Age of Arrival (AOA), a macro variable, accounted for far more
variance not because age is intrinsically important but because of its association with other variables (e.g.,
self-reported percentage use) that are likely to be of importance. A series of principal components analyses
(PCA) suggested that the most important predictors of how well an L2 will be pronounced after decades of L2
ìimmersion are (a) how frequently on a daily basis the L2 is used, and (b) the extent to which the L1 is
maintained. The role of other variables such (e.g., specific aspects of neural maturation or cognitive
development await empirical testing.
Flege, J. (2007). Language contact in bilingualism: Phonetic system interactions. In J. Cole and Hualde, J.
(Eds.), Laboratory Phonology 9. Berlin: Mouton de Gruyter, Pp. 353-380.
Summary. This is the last comprehensive exposition of the Speech Learning Model (SLM), a model which
attempts to account for L2 speech learning. The fundamental tenets of the SLM are that (a) the processes
and mechanisms used in successful L1 learning, including the ability to create new long-term memory
representations called phonetic categories, remain intact across the life span, and (b) the phonetic
information needed to distinguish the sounds of an L2 from one another, and from previously established L1
sounds, remains accessible. The chapter devotes considerable time to elucuding what happens when new
phonetic categories are established, and when they are not established for L2 sounds that differ auditorily
from the closest L1 sound.
Key words: Speech Learning Model (SLM), early vs. late learners, production vs. perception, plasticity,
neural maturation, critical period, L2 input, ultimate attainment, Chinese, Korean, LOR, immigrants,
confounds with age, confounded factors, bilinguals, constraints, mutual influence, L2-on-L1 effect, language
dominance, phonetic interaction, perceptually assimilate, equivalence classification, phonetic category
formation, phonetic category assimilation, phonetic category dissimilation,
Key words: conversational speech, fluent early bilinguals, early vs. late bilinguals, vowel production accuracy,
native Italian, auditory evaluation, spelling pronunciation, L2 vowel production
Summary. to appear.
Key words: second language, L2, individual differences, age effects, age-related effects, age of L2 learning,
ultimate L2 proficiency, L2 input, amount of L2 input, quality of L2 input, early learners, late learners, AOA,
age of arrival, LOR, length of residence, immersion, immigrants, neural maturation, language background
questionnaire, plasticity, critical period, foreign accent, perceived foreign accent, VOT, native Italian, Korean,
frequency of L1 use, social networks, immigrants, delayed repetition, maturational constraint, phonological
acquisition, grammaticality judgment test, confounds, subgroup matching procedure, cognitive development,
phonetic category development, cross-language phonetic differences, SLM, Speech Learning Model,
perceived phonetic dissimilarity, input, motivation, L1 proficiency, macro variable, frequency of L2 use,
phonological short-term memory
Summary. To appear.
Key words: vowel perception, Catalan, Spanish, early learners, bilinguals, bilingualism, bilingual vowel
perception, frequency, frequency of L2 use, categorical, L2 performance, AOA, age of learning, late
learners, maturation, categories, new categories, single-category assimilation, robust contrasts, dialect
variation, neutralize, functional load, foreign-accented input, phonological variation, degree of robustness,
experience, AXB, production, read-aloud task, percent L2 use, identification, discrimination, category
Summary. See above.
Key words: second language phonology, contrastive, phonetic realization, perceptual foreign accent, L1 use,
L2 vowel perception, native Italian, new phonetic categories, AOA, native Italian, discrimination scores,
phonetic perception, SLM, word final consonants, unaspirated, functionally equivalent, interference,
frequency of L2 use, feature hypothesis, phonetic category formation, Swedish, vowel quantity, contrastive
use, spectral vs. temporal cues, English, Estonian, Spanish, phonological length distinction, feature
prominence, individual differences, age of arrival, word familiarity, L2 use
Summary. to appear.
Flege, J. (2003). Assessing constraints on second-language segmental production and perception. In A. Meyer
and N. Schiller (Eds) Phonetics and Phonology in Language Comprehension and Production, Differences and
Similarities. Berlin: Mouton de Gruyter. Pp. 319-355.
Key words: Speech Learning Model (SLM), age, age of learning, L2, second language, age of L2 learning,
input, use, L2 use, amount of L2 use
Summary. to appear.
Key words: second-language, production, adults, children, vowel production, English, Korean, perceived
cross-language similarity, perceived dissimilarity, equivalence classification, phonetic realization, phonetic
context, word frequency, word familiarity, lexical factors, segmental factors, picture-naming task, years of
English education, SLM,
Summary. to appear.
Flege, J. (2003). Methods for assessing the perception of vowels in a second language. In E. Fava & A. Mioni
(Eds) Issues in Clinical Linguistics. Padova: UniPress, Pp. 19-44.
Key words: second language, segmental perception, vowel perception, identification, discrimination, AXB,
ABX, triadic, guessing, AX, same-different, accuracy, relation between prodution and perception, Korean,
Japanese, Czech, Hungarian, Arabic, Portuguese, German, Dutch, individual differences, training,
longitudinal study, change trials, perceptual learning, vowel inventory, phonetic differences, response labels,
forced choice, oddity, phonetic symbols, IPA, auditory, phonetic level, long-term memory, categorial,
categorial discrimination, Perceptual Assimilation Model, Speech Learning Model, A', A-prime
Summary. to appear.
Flege, J. (2002). Interactions between the Native and Second-language Phonetic Systems. In P. Burmeister, T.
Piske and A. Rohde (Eds) An Integrated View of Language Development: Papers in Honor of Henning Wode.
Trier: Wissenschaftlicher Verlag, Pp. 217-244.
Key words: Speech Learning Model (SLM), critical period, CPH, critical period hypothesis
Flege, J. (1999). Age of learning and second-language speech, In D. Birdsong (Ed) Second Language
Acquisition and the Critical Period Hypothesis. Hillsdale, NJ: Lawrence Erlbaum. Pp. 101-132.
Key words: Speech Learning Model (SLM),confounds, confounded factors, age, age of L2 learning,
AOA, age of arrival, age of first exposure
Flege, J. (1997). English vowel production by Dutch talkers: More evidence for the "new" vs "similar"
distinction". In A. James and J. Leather (Eds) Second-language speech, Structure and process. Berlin: Mouton
de Gruyter, Pp. 11-52. [Studies in Second Language Acquisition, 13. ISBN 3-11-014126-4]
Bohn, O.-S. and Flege, J. (1997). Perception and production of a new vowel category by second-language
learners. In A. James & J. Leather (Eds) Second-language speech: Structure and process. Berlin: Walter de
Gruyter, Pp. 53-74. Studies in Second Language Acquisition, 13. ISBN 3-11-014126-4]
Key words: vowel production, vowel perception, new vowel, relation between production and perception,
German, individual differences, equivalence classification, experience, category formation
Flege, J., Munro, M., & MacKay, I. (1996). Factors affecting the production of word-initial consonants in a
second language. In R. Bayley & D. Preston (Eds) Second Language Acquisition and Linguistic Variation.
Amsterdam: John Benjamins, Pp. 47-73.
Flege, J. (1995). Second-language Speech Learning: Theory, Findings, and Problems. In W. Strange (Ed)
Speech Perception and Linguistic Experience: Issues in Cross-language research. Timonium, MD: York Press,
Summary. Many earlier studies led the the creation of the SLM, but this is the first full-blown presentation of
the model. When researchers refer to the model in a single citation, this is often the chapter that is cited
even though the model continued to be elaborated over the following decade. The stated aim of the SLM is
to account for age-related changes in the ability to produce and perceive L2 sounds, that is,
position-sensitive vowels and consonants. The model posits that the processes and mechanisms that permit
successful L1 speech acquisition -- including the ability to establish phonetic categories -- remain intact over
the life span. Gone is an earlier distinction between "identical", "similar" and "new" L2 sounds. Instead, the
likelihood of phonetic category formation is hypothesized to increase as a function of the perceived phonetic
dissimilarity of an L2 sound from the closest sound(s) in the L1 inventory. In the absence of phonetic
category formation, the properties of an L2 sound and a neighboring L1 sound may merge to form a
"composite" category. The model posits that the phonic elements present in the L1 and L2 inventories exist
in a common phonological space and may interact with one another, leading in some cases to a "deflection"
of neighboring L1, L2, and L1/L2 categories to maintain phonetic contrast in the bilingual's dual system.
Key words: Speech Learning Model (SLM)
Flege, J. (1988). Effects of equivalence classification on the production of foreign language speech sounds, In
A. James & B. Leather (Eds) Sound Patterns in Second Language Acquisition. Dordrecht: Foris, Pp. 9-39.
This aim of this 177-page chapter was to "identify and develop theoretical issues of importance with regard to
the production and perception of sounds found in a foreign language." It reviewed nearly all the literature
dealing with L2 speech learning that existed up to 1983. (Publication was, alas, delayed by five years.) The
findings summarized here formed the underpinnings of what later became the Speech Learning Model
Table of contents Introduction: Why it is important to study L2 speech, Defining the phenomena of "foreign
accent. Effects of a foreign accent: Accent detection, Diminished acceptability, Diminished intelligibility,
Negative evaluation. Factors causing accent: Age and amount of L2 experience, L2 input, Motivation and
affective factors, Style and social meaning, individual differences, listener effects. Phonological and phonetic
factors: A two-level model of production, Phonological rules, Phonological and phonetic features, New vs.
similar sounds. Perceiving L2 speech sounds: Factors affecting segmental perception, Interlingual
identification, Phonological filtering, Reinterpretation, Perceptual tuning: Maintenance, Learing, Loss,
Phonetic category prototypes Mechanisms causing accent: Maintenance, Learning, Loss, Hypercorrection
Linguistic factors: Orthographic effects, Word familiarity, Phonetic context, Coarticulation, Basis of
articulation Phonotactics and syllable structure: English /r/ and /l/, Production by Japanese learners,
Production and perception of /r/ and /l/, Word stress Vowel timing: Intrinsic timing, Extrinsic timing, English
differential vowel duration, English vowel tensity, Phonemic length distinctions Consonant timing: Voice
onset time, VOT in child speech, The modification of VOT, VOT in L2 speech acquisition, VOT in L2
perception Emerging issues and theories: Phonological translation, Peripheral sensory feedback, Attention
to speech, Strategies Conclusions
Flege, J. (1991) Perception and production: The relevance of phonetic input to L2 phonological learning. In T.
Heubner & C. Ferguson (Eds) Crosscurrents in Second Language Acquisition and Linguistic Theory.
Philadelphia: John Benjamins, Pp. 249-289.
Hammond, R. & Flege, J. (1990). The acquisition of second language phonological systems in a communicative
framework. In B. Lawton & A. Tamburri (Eds) Romance Languages Annual 1989, Volume I. Purdue: Purdue
Research Foundation, Pp. 671-676.
Flege, J. (1988). The production and perception of speech sounds in a foreign languages. In H. Winitz (Ed)
Human Communication and Its Disorders, A Review 1988. Norwood, N.J.: Ablex, Pp. 224-401.
Flege, J. (1992) The Intelligibility of English vowels spoken by British and Dutch talkers. In R. Kent (Ed.)
Intelligibility in Speech Disorders: Theory, Measurement, and Management. Amsterdam: John Benjamins, Pp.
Summary: This chapter had two aims: to introduce the Speech Learning Model (SLM), and then to examine
the intelligibility of English vowels spoken by native speakers of Dutch within the framework of that model.
The preliminary version of SLM presented here differs in important respect from the model formally
presented by Flege (1995). Click here for details.
The L2 production experiment examined six English vowels spoken by eight native speakers of British English
and 50 Dutch adults, all university students aged 20-25 years who had all begun to learn English in school at
the age of 12 years. Subgroups of the Dutch participants (n=16 each) were established based on overall
degree of foreign accent (FA) in the production of English sentences. The assumption was made that the
subgroups differed in experience in English, most for the Mild FA group and least for the Strong FA group. In
support of this assumption, all members of the former group were majoring in English whereas most members
of the latter group were engineering students who do not seem to have pursued the study of English beyond
the formal academic requirements.
L2 vowel production accuracy was assessed by determining the percentage of times the English vowels were
identified as intended by three native speakers of British English, who used keywords to signal their
identifications. One vowel examined, /ɪ/, was as operationally classified as an “identical” vowel and another
vowel, /æ/, was classified as a “new” vowel. The remaining four vowels examined, /i/, /u/, /˄/ and /ɒ/, the vowel
in hot) were classified as “similar”. The classifications just mentioned were based on a “phonetic symbol” test
(“new” if the L2 vowel is traditionally represented by an IPA symbol not used for an L1 vowel) and the
distances between L1 and L2 vowels seen when visually inspecting plots of the vowels in an F1-F2 acoustic
The study tested two specific hypotheses regarding the effect L2 experience on L2 vowel production. First,
participants in the three Dutch subgroups would not differ from one another or from the native English
speakers in producing the “identical” L2 vowel because an unmodified use of a Dutch vowel in English would
go undetected by native English speaking listeners. Second, production of the “new” vowel would improve
with experience. Both predictions were confirmed. The Mild FA and Moderate FA groups were found to
produce the new vowel /æ/ more accurately than the Strong FA group did, but did not differ from native
speakers of English.
Somewhat different results were obtained when the vowels were presented for identification to native
speakers of American English were asked to classify the same vowels using keywords. A between group
difference, suggesting an effect of L2 experience on L2 vowel production, was obtained was obtained for the
“new” vowel /æ/. However, an unexpected between group difference was also obtained for one of the “similar”
vowels, /ɒ/. This suggested the importance, in a study that relies of listener judgments to assess L2
production accuracy, of using listeners who are native speakers of the variety of the L2 that has been
Summary: This chapter, based on a presentation at Stanford University, further develops ideas later
incorporated into the Speech Learning Model. The aim of the chapter was to provide a "brief overview of
research that has described L2 learners' production and perception of L2 sounds". It starts with the
observation that learners of an L2 are generally "better able to control their speech apparatus" than [are]
children acquiring an L1" who already possess a "phonetic system for producing speech". That being the
case, L2 production errors may arise from an "inappropriate use of previously acquired [i.e., L1] structures",
that is, cross-language interference. The author notes the common view that learners often attempt to
"decompose" L2 words into sequences of L1 phonemes but stresses that the simple substitution L1 for L2
"sounds" is merely a starting point for, as demonstrated by evidence presented in the chapter even most
adults who learn an L2
... eventually modify previously established patterns of speech production and perception when attempting
to deal with the sound system of an L2. Thus, the differences one might observe between native and
non-native speakers often involve more than just the maintenance of old articulatory habits (p. 566).
Sections: Production vs. perception, IInterlingual identification, New vs. similar sounds, Equivalence
classification, Defining the "new" vs."similar" distinction, Vowel production, L2 consonant production, Early
versus late learners, Realization rules, Vowel production, Individual differences, A critical period? Changes
in speech learning, Linguistic factors, Summary
A recurrent them of the chapter is that many aspects of L2 production can be understood in terms of "how L2
sounds are categorized. L2 speech learning is not always just "old wine in new bottles" (p. 566)The author
hypothesizes that as L1 categories develop in children, they become ever more efficient for the production
and perception of L1 sounds. He suggests foreign accents are likely to be evident for individuals who begin
learning an L2 after "about the age of five to seven years" because, as the result of L1 category
development, L2 learners become less likely to establish new "phonetic categories for certain L2 sounds and
thus [to produce] them accurately" (p. 597). Although this hypothesis was incorporated into the first formal
presentation of the SLM in 1995 (in W. Strange, Editor) later research undermined the view that an important
shift occurs in the ability to learn L2 speech at about the ages of 5-7 years. Children's L1 phonetic
categories continue to develop slowly, in some cases not reaching adult levels until the age of about 12-14
years. Also, the presence of foreign accent has been observed in the speech of many children who began
learning an L2 before the age of 5 years, even after years of L2 use.
Flege, J. & Hillenbrand, J. (1987) Limits on Pronunciation Accuracy in Foreign Language Speech Production,
In G. Ioup & S. Weinberger (Eds) Interlanguage Phonology: The Acquisition of a Second Language Sound
System. Rowley, MA: Newbury House, Pp. 176-203.
This study addressed two research questions drawn from research examining phonological aspects of
second language acquisition and from work focusing on the remediation of articulatory errors in monolingual
children. The first question was whether "new" L2 sounds -- those without an obvious equivalent in the L1 --
would be produced more or less accurately than "similar" L2 sounds that differ phonetically from a
counterpart in the L1. The second question addressed by this study was whether accuracy in the production
of phonetic segments in an L2 depends on the amount of attention that can be allocated to the production of
Four groups participated in the three conditions of what was described to the participants as a study of "
linguistic creativity." Post-study debriefing indicated that none of the participants realized that the focus of the
study was actually pronunciation. In the Reading condition the participants were asked to read sets of seven
phrases each. The phrases in Set 1 (English) began with the word "two" (/tu/) whereas those in Set 3 and Set
4 (French) began with the words "tous" (/tu/) and "tu" (/ty/). In the Sentence Condition the participants were
asked to spontaneously produce complete sentences beginning with the phrases they had just read. Finally,
the task of the Story Condition was to tell a coherent story including sentences that began with each of the
phrases from the original reading task.
Four groups of young adults participated: (1) monolingual native speakers of English (2) native French
speakers who were residing in Chicago (3) students at Northwestern University who had spent the previous
academic year in Paris (4) native English speakers who taught French at an American university. The
American professors had resided in France somewhat longer than the American students and, unlike the
students, used French on a regular basis with native French-speaking colleagues.
The English monolinguals produced only English speech materials (Sets 1 and 2) whereas the remaining
three groups produced both English and French materials (Sets 3 and 4) in counterbalanced order. The
dependent variables measured acoustically were VOT (in msec) of word-initial tokens of French and English
/t/ and second formant frequencies (F2) in the production of three vowels. French /t/ was classified as a
similar L2 sound for native speakers of English inasmuch as it is produced with substantially shorter VOT
values than is the /t/ of English. French /u/ was also classified as similar inasmuch as its production differs, in
terms of F2 values, from the /u/ of English. Finally, the front rounded /y/ of French was classified as "new"
because it is labelled using an IPA symbol not employed in the phonetic description of vowels in English.
Moreover, in terms of F2, French /y/ differs far more from the English /u/ than does the French /u/.