Tsukada, K., Birdsong, D., Bialystok, E., Mack, M., Sung, H. and Flege, J. (2003). The perception and
production of English /ɛ/ and /æ/ by Korean Children and Adults living in North America. In M. Solé, D.
Recasens & J. Romero (Eds) Proceedings of 15th International Congress of Phonetics Sciences,
Barcelona: Casual Productions, Pp. 1589-1592.
Flege, J. (2002). No perfect bilinguals. In A. James and . Leather (Eds) New Sounds 2000: Proceedings of
the Fourth International Symposium on the Acquisition of Second-Language Speech. University of
Klagenfurt, Pp. 132-141.
Baker, W., Trofimovich, P., Mack, M. and Flege, J. (2002). The effect of perceived phonetic similarity on
non-native sound learning by children and adults. In B. Skarabela, S. Fish and A. Do (Eds) Proceedings of
Birdsong, D. & Flege, J. (2001). Regular-irregular dissociations in the acquisition of English as a second
language. In BUCLD 25: Proceedings of the 25th Annual Boston University Conference on Language
Development, Boston, MA: Cascadilla Press, Pp. 123-132.
Guion, S., Flege, J., & Loftin, J. (1999). The effect of L1 use on foreign accent ratings in Quicha-Spanish
bilinguals. In J. Ohala, Y. Hasegawa, M. Ohala, D. Granveille and A. Bailey (Eds) Proceedings of the XIVth
International Congress of Phonetics Sciences, Berkeley, CA: Department of Linguistics, UCLA, Pp. 1471-
Flege, J., Yeni-Komshian, G. & Liu, S. (1999). Age Constraints on learning L2 phonology and morphosyntax.
Proceedings of the the Joint Meeting of the 16th International Congress on Acoustics and the 137th Meeting
of the Acoustical Society of America, Berlin, 15-19 March, 1999.
Piske, T., Flege, J., MacKay, I., & Meador, D. (1999). Non-natives’ production of vowels in conversational
speech. Proceedings of the the Joint Meeting of the 16th International Congress on Acoustics and the 137th
Meeting of the Acoustical Society of America, Berlin, 15-19 March, 1999.
Flege, J. (1998). The role of subject and phonetic variables in L2 speech acquisition. In M. Gruber, D.
Higgins, K. Olsen and T. Wysocki (Eds) Papers from the 34th Annual Meeting of the Chicago Linguistic
Society, Volume II, The Panels. Chicago: Chicago Linguistic Society. Pp. 213-232.
Flege, J. (1998). Second-language Learning: The role of subject and phonetic variables. In Proceedingsof
the ESCA Workshop on Speech Technology in Language Learning (Marholmen, Sweden, May 24-27,
1998). Pp. 1-9.
Flege, J., Guion, S., Akahane-Yamada, R., & Downs-Pruitt, J. (1998). Categorial discrimination of English
and Japanese vowels and consonants by native Japanese and English subjects. In P. Kuhl and L. Crum
(Eds) Proceedings of the 16th International Congress on Acoustics and the 135th Meeting of the Acoustical
Society of America, Volume IV, New York: Acoustical Society of America, Pp. 2973-2974.
Nozawa, T., & Flege, J. (1998). Perception of English vowels by Japanese speakers residing in the United
States, In: Psychology and Learning of Language, Tokyo: Kinseido (Japan Society of Speech), Pp. 65-77.
Flege, J. (1997). The role of category formation in second-language speech learning. In J. Leather and A.
James (Eds) New Sounds 97, Proceedings of the Third International Symposium on the Acquisition of
Second-Language Speech, Klagenfurt, Austria: University of Klagenfurt, Pp. 79-89.
Hillenbrand, J.,and Flege, J. (1992). Application of acoustic techniques to the assessment of speech
disorders. In Assessment of Speech and Voice Production: Research and Clinical Applications. NIDCD
Monograph. National Institute of Deafness and Other Communication Disorders. Pp. 53-62.
Bohn, O., Flege, J., Dagenais, P., & Fletcher, S. (1991). Effects of bite-block and loud speech on tongue
heights in the production of German vowels. In Proceedings of the 12th International Congress of Phonetic
Sciences, Volume 3, Pp. 70-73.
Flege, J. (1984). The detection of foreign accentedness. In A. Cohen and M. van den Broecke (Eds)
Proceedings of the 10th International Congress of Phonetic Sciences, Dordrecht: Foris, Pp. 677-681.
Flege, J. (1982). English speakers learn to suppress final stop devoicing, In R. Chametzky, R. Hirzel and K.
Tuite (Eds) Papers from the 18th Regional Meeting, Chicago: Chicago Linguistic Society, Pp. 111-122.
Flege, J., Brown, W., Jr., & White, K. (1982). Voicing control in child speech, In T. Kastor-Bennett (Ed) Mid-
America Linguistics Conference Proceedings, Wichita State University, Dept. of English, Pp. 11-31.
Flege, J. & Hammond, R. (1981). Speakers' awareness of some non-segmental phonetic aspects of foreign
accent. In M. Henderson (Ed) 1980 Mid-America Linguistics Conference Papers, University of Kansas, Dept.
of English, Pp. 145-163.
Flege, J. (1980). Temporal correlates of [voice] in Arabic-accented English. In J. Wolfand D. Klatt (Eds)
Speech Communication Papers Presented at the 97th Meeting of the Acoustical Society of America, New
York, Acoustical Society of America, Pp. 171-175.
Flege, J. (2012). The role of input in second language speech learning. VIth International Conference on
Native and Non-native accent of English. Lodz, Poland 6-8 December 2012.
Note: This is a slightly edited version of the keynote I presented at New Sounds 2010. The aim of the talk
was to consider four hypotheses regarding age-related effects on L2 speech acquisition. A lot of territory
is covered here. However, I call your attention especially to the discussion of why the Johnson & Newport
(1987) study does not provide convincing evidence in support of a "maturational" account of age effect on
L2 learning. I've replotted some data from our JML article (Flege et al. 1999). The study tested the critical
period hypothesis by examining 240 Korean immigrants to the US. Several factors other than the age of
L2 learning, including education in the US and amount of English use, might well account for most of the
variance in the two outcome variables (foreign accent ratings, morphosyntax scores). Interestingly, the
cross-over from Korean-dominant to English dominant occured at an AOA of 12 years.
Flege, J. (2005). The origins and development of the Speech Learning Model. Keynote lecture at the
1st Acoustical Society of America Workshop on L2 Speech Learning, Simon Fraser University,
Vancouver, CA (April 14-15, 2005). [conference presentation]
Flege, J. (2010). "Age" effects on second language acquisition. New Sounds 2010, 1-3, 2010,
Note: This is a slightly edited version of the keynote I presented at New Sounds 2010. The aim of the
talk was to consider four hypotheses regarding age-related effects on L2 speech acquisiton. A lot of
territory is covered here. However, I call your attention especially to the discussion of why the Johnson
& Newport (1987) study does not provide convincing evidence in support of a "maturational" account
of age effect on L2 learning. I've replotted some data from our JML article (Flege et al. 1999). The
study tested the critical period hypothesis by examining 240 Korean immigrants to the US. Several
factors other than the age of L2 learning, including education in the US and amount of English use,
might well account for most of the variance in the two outcome variables (foreign accent ratings,
morphosyntax scores). Interestingly, the cross-over from Korean-domiant to English dominant occured
at an AOA of 12 years.
Aoyama, K., Flege, J., Guion, S., Akahane-Yamada, R. and Yamada, T. (2003). Foreign accent in
English words produced by Japanese children and adults. In M. Solé, D. Recasens and J. Romero (Eds)
Proceedings of 15th International Congress of Phonetics Sciences, Barcelona: Casual Productions, Pp.
Imai, S., Flege, J. and Walley, A. (2003). Spoken word recognition of accented
and unaccented speech: Lexical factors affecting native and nonnative listeners. In M. Solé, D.
Recasens and J. Romero (Eds) Proceedings of 15th International Congress of Phonetics Sciences,
Barcelona: Casual Productions, Pp. 845-848.
McAllister, R., Flege, J., & Piske, T. (1999) The acquisition of Swedish Long vs. Short contrasts by nativve
speakers of English, Spanish and Estonioan. In J. Ohala, Y. Hasegawa, M. Ohala, D. Granveille and A.
Bailey (Eds) Proceedings of the XIVth International Congress of Phonetics Sciences, Berkeley, CA:
Department of Linguistics, UCLA, Pp. 751-754.
Piske, T. & MacKay, I. (1999). Age and L1 Use effects on degree of foreign accent in English. In J. Ohala,
Y. Hasegawa, M. Ohala, D. Granveille and A. Bailey (Eds) Proceedings of the XIVth International
Congress of Phonetics Sciences, Berkeley, CA: Department of Linguistics, UCLA, Pp. 1433-1436.
Flege, J.. (1999). The relation between L2 production and perception. In J. Ohala, Y. Hasegawa, M. Ohala,
D. Granveille and A. Bailey (Eds) Proceedings of the XIVth International Congress of Phonetics Sciences
(Berkeley, CA: Department of Linguistics, Univ. of California at Berkeley), Pp. 1273-1276.
Abstract: It has been claimed that a correlation does not exist between how accurately experienced late
learners produce and perceive phonetic segments in a second language (L2). According to one theory,
learners of an L2 are no longer able to align segmental production and perception after the passing of a
critical period. This contribution reviews studies that have examined L2 production and perception. All of the
studies yielded significant, albeit modest,correlations. Possible explanations for why stronger correlations
have not been observed are presented.
Comment: The article lays out the SLM hypothesis regarding the relation between segmental production
and perception in an L2, namely: "The accuracy with which L2 segments are perceived limits how accurately
they will typically be produced (p. 1273). The article notes that "not all aspects of perceptual learning may be
incorporated into production. That is, production and perception may not be brought into perfect alignment,
as is the case in LI speech acquisition. Thus, the SLM predicts that modest correlations will exist between L2
segmental production and perception for highly experienced speakers of an L2."
For further discussion of the relation between production and perception, including analysis techniques, see
Discussion Topic 4.
Flege, J.E. (2016). The role of phonetic category formation in second language speech acquisition.
Eight International Conference on Second Language Speech, 10-12 June, 2016, Aarhus University,
Abstract: In 1995, after 15 years of preliminary work, I formally presented the Speech Learning Mode. A
key aspect of this L2 speech acquisition model was the hypothesis is that learners of any age, even late
learners, retain the capacities used in successful L1 speech learning, including the capacity to create
new phonetic categories for certain L2 sounds. In the first part of this talk I describe the both general and
specific properties of phonetic categories, referred to as "containers" in which learners store and
structure information in long term memory pertinent to classes of speech sounds. Then I present studies
providing evidence for the formation of phonetic categories for English /p/, by native speakers of
Spanish; for English /ɝ/, by native speakers of Italian; and for English /r/, by native speakers of
Japanese. I conclude the talk with a brief summary, indicating what it is that we still don't know about L2
category formation, and identifying obstacles to complete understanding of this crucial aspect of L2
Flege, J.E. (2017). Imparare la pronuncia di una lingua straniera: Possiamo farlo anche noi anziani?
Circolo Culturale «Enrico Pocci» Via G. Verdi, 7 Tuscania (VT)
Flege, J.E. (2017) The cross-language acquisition of stops differing in VOT: Historical overview and key
findings. Phonetic Teaching and Learning Conference 2017, University College London. 9-12 August
Part 1: Historical overview of VOT
A great deal of L2 research has focused on the Voice onset time (VOT) dimension in word-initial stop
consonants. Interest in the VOT dimension arises from the fact that languages differ in terms of how VOT
is used to distinguish phonemes (e.g., /b/ vs /p/) and to the fact that the VOT dimension provides a useful
bridge between production and perception. Other considerations are that it is easy to measure VOT and
to create perceptual continua consisting of stimuli that differ in VOT. Alas, it is also possible to make
errors when using VOT to evaluate segmental production and perception.
This talk reviews the VOT dimension from the perspective of cross language differences, inter-lingual
identification, phonetic theory, L1 acquisition and L2 learning. A proposal deriving from the L1 acquisition
research is that learners of an L2 may need as much as 10 years of native speaker input to produce and
perceive L2 stops accurately. The discussion of phonetic organization focuses on the nature of phonetic
categories and the language-specific realization rules used to motorically output the phonetic categories
in speech production. Also included is a consideration of individual differences and a review of studies
examining the capacity of adults to learn to use VOT differently when learning an L2.
Part 2: The cross-language acquisition of stops differing in VOT: Key findings
This review of the literature led to the following conclusions regarding the production and perception of
word-initial consonants by native speakers of Romance languages who learn English as an L2
1. Learning to perceive and produce the VOT dimension depends more importantly on the phonetic
input received than on age of L2 learning;
2.Learners of all ages retain the speech-learning capacities available to children who learn an L1,
including the capacity to make effective of use of input and establish new phonetic categories and
phonetic realization rules;
3. For native speakers of Romance languages learning of English /b d g/ and /p t k/ proceed differently
due to differing patterns of cross-language differences;
4. Native speakers of Romance languages do not establish new phonetic categories for English /b d g/.
Instead, they restructure their L1 phonetic categories for use in two languages on the basis of the input
they have received;
5. On the other hand, native speakers of Romance languages who receive adequate phonetic input do
establish new phonetic categories for English /p t k/;
6. Doing so does not cause them to “lose” their L1 phonetic categories for /p t k/. As a result, they have
three categories for stops consonants, not just two as is the case for monolingual speakers of English
and Romance languages;
7.Those who begin to produce L1 /b d g/ with short-lag VOT values as the result of long-term exposure
to such stops in the L2 slightly increase VOT in L1 /p t k/ in order to avoid producing L1 /b d g/ and /p t
k/ with the same short-lag VOT values
Flege, J.E. (2017) The cross-language acquisition of stops differing in VOT: Historical overview. SICSS
2017: 2017 Seoul International Conference on Speech Sciences, 10–11 November 2017, Seoul National
Flege, J.E. (2017) The role of input in the acquisition of L2 stops. SICSS 2017: 2017 Seoul International
Conference on Speech Sciences, 10–11 November 2017, Seoul National University, Korea
This talk examines the role of input in the perception and production of L2 stop consonants, focusing on
the VOT dimension. I review results obtained in research with native Italian learners of English in Canada.
The results convinced me that input is a far more important determinant of success in L2 speech learning
than the age at which L2 learning begins. Many believe that native vs. non-native differences are often due
to differences in age of L2 learning (AOL). Immigrants' AOA is more important in our research because it
conditions the kind of experience our participants had with English, defining both the quantity and quality of
input they received over the course of their lives
Flege, J.E. (2018). The Critical Period Hypothesis fails to predict L2 foreign accent and segmental
production accuracy. Research Frontiers of Second Language Speech, April 14-15 2018, Nanjing
University of Science and Technology, China.
Lenneberg (1967) was correct in saying that most people who begin learning an L2 after puberty usually
speak it with a foreign accent (FA), but predictions generated by his critical period hypotheses (CPH) do
not conform well to detailed analyses of FA. First, individuals who learned their L2 as young children and
have spoken it for decades as their primary language speak their L2 with FA. Second, the strength of a FA
grows tends to grow increasing strong after the closure of the hypothesized critical period. Further, the
focus on L2 learning that has been promoted by widespread acceptance of the CPH has diverted attention
away from a a very important aspect of FA, that is, the presence of an L2 inspired FA in the native
language of bilinguals. A consideration of segmental production and perception by L2 learners casts
further doubt on the validity of the CPH. Individuals who learned the L2 as young children differ significantly
from native speakers of the target L2. In some cases adults who began learning the L2 at the ages of 8
and 20 ages perform the same in the L2. Taken as a whole, the results of phonetically oriented research
has not supported the CPH, and so I propose that it be rejected.
Flege, J.E. (2018). How the revised Speech Learning Model (SLM-r) works for word-initial stop
consonants.Research Frontiers of Second Language Speech, April 14-15 2018, Nanjing University of
Science and Technology, China.
In this talk I review research examining the production and perception of word-initial stops by speakers
of two Romance languages, Spanish and Italian. In an aspiration language like English, the phonetic
basis of voiced-voiceless distinctions is the presence vs absence of aspiration following release while in
Romance languages it is the presence vs absence of glottal pulsing prior to release of word-initial stops.
Different tasks face L1 Romance learners of English for voiced and voiceless stops. The task for /p t k/
is to establish new “long lag” phonetic categories. As predicted by the SLM-r, the L2 learners examined
manage to do so if they obtained a sufficient amount of native speaker input. The learning task for /b d
g/. is different. In English these stops can realized with lead or short-lag VOT values. L1 Romance
learners cannot create new “short lag” phonetic categories for them because this portion of the phonetic
space (shared across languages) is already occupied by short-lag realizations of L1 /p t k/. It would be
easy for Romance L1 learners to simply continue producing English /b d g/ with full pre-voicing. After all,
some native speakers of English do so. However, as predicted by the SLM-r, Romance L1 learners of
English L2 begin showing the English pattern, producing: full pre-voicing, prevoicing that ends before
release, and short-lag VOT. Also as predicted by the SLM-r, the mix depends on the input they have
received, and effects of the (seemingly unnecessary) learning that has occurred in English is carried
back into the L1.
Flege, J.E. (2018) L2 speech learning: Time to change the paradigm. Center for Research on Bilingualism,
Stockholm University, June 11, 2018
Abstract. The critical period hypothesis (CPH) has exerted a strong influence on L2 speech research ever
since Eric Lenneberg (1967) observed that people who learn an L2 after puberty usually speak it with a
foreign accent. His observation was largely correct, but research going beyond a superficial level of
observation yields findings that are incompatible with the CPH. Importantly, the CPH also fails to predict an
important phenomenon regarding foreign accent: an influence of the L2 on how bilinguals pronounce their
native language. Evidence regarding segmental-level production and perception also poses serious
challenges for the CPH.
The primary problem is not that the CPH is wrong but that its widespread appeal has discouraged the search
for better, falsifiable explanations of age-related effects on L2 speech learning. The talk concludes with a
presentation of a method that might be used to obtain, for the first time, the crucial predictor variable that
has been missing from L2 research: estimates of the quantity and quality of input received by L2 learners.
Keywords: critical period hypothesis, Lenneberg, L2, speech, speech learning, foreign accent, confounds,
input, quantity of input, quality of input, Italians, Koreans, Experience Sampling Method, LOR, length of
residence, foreign-accented input
Flege, J. (2010). “Age” effects on Korean adults’ learning of English morphosyntax. Second annual
SLATE Conference, University of Illinois, Champaign, IL, May 6, 2010.
Summary. This talk presents the grammaticality judgment test (GJT) results obtained by Flege et al.
(1999) for 240 Korean immigrants in greater detail than was possible in the original JML article.
The GJT scores obtained for “Early” (AOA 1-8.5 years), “Mid” (9.5-14.5) and “Late” (15.5-22-5)
learners were all significantly lower scores than the NE comparison group and all three Korean groups
differed significantly from one another. The data pattern was clear but interpretation of the results was
difficult because the selection variable, AOA, was confounded with other variables (frequency of L1
and L2 use, LOR, and years of education in the US) that might reasonably be expected to have
influenced the GJT scores.
To “untangle” variables confounded with AOA we computed two GJT subscores. The Rule-based
subscores tested knowledge of regular, productive, and generalizable rules of the surface morphology
of English; the Lexically-based subscores tested the Koreans’ knowledge of irregular and
ungeneralizable aspects of English morphosyntax. Korean groups having mean AOAs of 13-21
obtained significantly lower Lexically-based than Rule-based subscores but not the groups having
mean AOAs ranging from 3-11 years.
Matched subgroup analyses compared groups in which AOA was held constant and either years of US
education or language use varied. When years of US education varied, matched subgroups differed
significantly for Rule-based but not Lexically-based subscores. When language use varied, matched
subgroups differed significantly for the Lexically-based but not Rule-based subscores.
Conclusion: These results effectively “explain away” AOA as the cause for Early vs Late differences,
in at least for the GJT used here. It’s time for SLA researchers to stop treating variation in L2 input as
a nuisance and start designing research to explore the role of quantity and quality of input on the
learning of L2 morphosyntax.
Keywords: morphosyntax, English, Korean, grammaticality judgment test, GJT, ciritcal period,
constaint, GJT, Johson & Newport (1989, age of arrival, AOA, earlier is better, confounds, input,
language use, LOR, length of residence, principal components analysis, lexically-based, rule-based.
Summary: The results reported here are inconsistent with the view that native
Japanese (NJ) adults' difficulty producing and perceiving English liquids is due to the
passing of a critical period or to the filtering out of phonetic information needed to
define English /r/ and to distinguish the English /r/ from English /l/. The results are,
however, readily understandable within the framework of the Speech Learning Model.
NJ speakers who are first exposed to English as children (Early learners) are
generally more successful in producing and perceiving English liquids than are NJ
speakers whose first everyday exposure to English occurs in adulthood (Late
learners). This is because Early learners generally get more and better input from
native speakers of English than Late learners do..
Both Early and Late learners are more successful at perceiving English /r/ than /l/
English because /r/ is perceptually more distant from the one liquid of Japanese (/R/)
than English /l/ is.
NJ speakers, even Late learners, manage to establish new phonetic categories for
English /r/ if they get abundant native speaker input. Their progress for English /l/ is
limited, however, because the phonetic properties of this sound are merged with the
properties of Japanese /R/. Regardless of exposure age, Japanese-English bilinguals
use a composite /R/-/l/ category when processing both the /R/ of their L1 and the /l/
of their L2.
Flege, J. E. (2018). The Speech Learning Model (SLM) account of how Japanese
speakers learn English /r/ and /l/. Sophia University, Tokyo, July 21, 2018.
Flege, J. E. (2018) The critical period hypothesis fails to predict foreign accent and
segmental accuracy. Sophia University, Tokyo, July 22, 2018.
Summary.. The critical period hypothesis (CPH) drawn from Lenneberg’s 1967
book continues to be accepted as the best explanation for age-related effects in
L2 speech research. However, the CPH generates incorrect predictions regarding
overall degree of foreign accent; segmental production and segmental perception.
In my view, input is the single most important predictor of success in L2 speech
learning. The problem we face, however, is that our measurements of quantity and
quality of L2 input are inadequate. The input hypothesis cannot be properly
evaluated (accepted or rejected) until we begin assessing input more precisely
Keywords: L2, speech learning, foreign accent, FA, strength, vowels,
consonants, production, perception, Italian, immigrants, perception, production,
Lenneberg, critical period, CPH, maturation, maturational state, error detection,