Professional Documents
Culture Documents
Examination of The Critical Period Hypot PDF
Examination of The Critical Period Hypot PDF
Alejandro E. Brice
University of South Florida-St. Petersburg
St. Petersburg, Florida
Roanne Brice
University of Central Florida
Orlando, Florida
4/0IIIMMINIONINIONV
second language. Each concept is briefly dis- Johnson and Newport (1989) assessed
cussed. A comprehensive review of this lit- AOA (i.e.. when optimal learning occurred
erature is not presented here due to space before or after the critical period) by means
limitations. However. the reader is referred of a grammaticality judgment task. They
to Birdsong (1999) who provides findings wished to test whether young children were
from numerous researchers across different able to acquire a second language earlier
language areas. than adults and. consequently. achieve high-
Age of arrival (AOA) is defined here as er levels of final proficiency in their second
the age at which participants came to the language. Participants included 46 Korean
new country and were exposed to the sec- and Chinese speakers learning English. All
ond language. that is. English. One of the ear- participants had lived in the United States
lier studies which investigated age of arrival for a period of at least five years. Participants
was conducted by Patkowski (1980). He in- were grouped according three AOA distinc-
vestigated syntactic language abilities of 67 tions (i.e., 5-7: 8-13: and 14-16 years).
immigrants via oral interviews and subse- Participants were then asked to provide
quent ratings. Participants were placed into grammaticality judgments on 276 English
two groups, namely, age of arrival before 15 sentences (half of which were grammatical
years and AOA after 15 years, thereby test- or ungrammatical). Results of their stud y in-
ing the age limitation constraint on learning dicated an advantage for earlier arrivals over
a second language. The ratings were con- the later arrivals (Johnson & Newport, 1989)
ducted by two English as a second language supporting the notion of earlier learning be-
(ESL) teachers trained over a period of two ing better or supporting the CPH. However,
weeks. The number of years that the partici- it should be noted that Johnson and New-
pants had resided in the United States ranged port (1989) did not assess the actual English
from six to 61 years. Patkowski found age of proficiency of their participants. It is gener-
arrival to he a major predictor of language ally agreed that learning a second language
proficiency concluding that age is a limiting earlier is better. however. there is disagree-
factor in ultimate L2 attainment, supporting ment as to the strength of the earl y language
the notion that language learning decreases learning. Because of the methodological dif-
as a function of age or absolute age limita- ficulties in the Johnson and Newport study
tion (i.e., in support of the critical period hy- (1989), the strength of their argument needs
pothesis). However. several flaws appeared to he tempered.
to he present in Patkowski's (1980) study. Length of residence (LOR) is the length
Some limitations of the study are noted in of time the individual has lived in the new
that the author does not provide sufficient country and is an indirect measure of second
information reporting the participants in the language proficiency. Moyer (1999) investi-
study . for example, their educational back- gated the significance of age of immersion.
ground, or when they learned English. In ad- age of instruction. years of teaching, years
dition, Patkowski's investigation was not a of immersion (similar in concept to LOR),
study of ultimate attainment as high profi- types of phonological feedback, profession-
ciency levels of his participants were never al motivation, and the importance of native
ascertained. However. the most serious lim- pronunciation on ultimate attainment in L2
itation of this study involves inappropriate phonology among adults. Age of immersion,
analysis of the data. Patkowski violated the age of instruction. professional motivation.
principles of normality of distribution and and type of phonological feedback all corre-
homogeneity of variance. He used ANOVA lated with L2 pronunciation ratings. Age of
parametric statistics but instead should have immersion and age of instruction yielded the
used nonparametric statistics. Consequent- highest correlations. Therefore, it appeared
ly. the results and conclusions from the Pat- that age did play a major role in L2 phonolog-
kowski (1980) study are questionable. ical attainment: however. motivation (19% of
146 ASIA PACIFIC JOURNAL OF SPEECH, LANGUAGE, AND HEARING, VOL. 11, NO. 3
the total variance was accounted for by this lish speakers produced five sentences which
variable) and amount of instruction (22% of were presented to a panel of native English
the total variance was accounted for by this speakers. The panel rated the sentences for
variable) did not yield highly significant out- degree of foreign accent. The native English
comes to language learning. Moyer's data speakers received the highest ratings than
suggest a trend toward a linear relationship most Italian speakers whose pronunciation
between age of immersion, age of instruc- ratings decreased with increased age of ar-
tion, and L2 phonology. However, no dis- rival. Flege et al.'s (1995) study demonstrat-
continuity between age of immersion, age ed a near-linear relationship between AOA
of instruction, and L2 phonological attain- and foreign accent ratings. In sum, there
ment was observed; thus, the CPH was not was no discontinuity or cutoffs between the
supported with regard to these aspects. ratings and AOA as would be predicted by
Flege, Munro, and MacKay (1995) inves- the critical period hypothesis. Therefore,
tigated the critical period hypothesis relat- the CPH was not supported by the findings
ed to second language pronunciation. Ac- from Flege, Munro, and MacKay (1995).
cording to Flege et al. (1995), difficulties
in L2 pronunciation have more to do with
the Speech Learning Model (SLM) than the Ultimate Attainment
critical period hypothesis. According to the
Speech Learning Model, difficulties in pro- Ultimate attainment is the ability to reach
nunciation result from the inability to prop- high levels of proficiency in the second lan-
erly establish new second language phonet- guage which may resemble nativelike com-
ic representations of sounds because the L2 petence (White & Genesee, 1996). White
phoneme is similar to one in the first lan- and Genesee stated that, "What is in ques-
guage, yet significantly different enough to tion is whether native-like competence is at-
cause pronunciation difficulties. Research tainable in individuals who acquire their L2
has shown that contrasting allophonic varia- after a certain age" (p. 234), that is, do sec-
tions (sounds that are similar yet which do ond language learners achieve the same lev-
not carry different meaning, for example, el of proficiency as first language speakers
English /b/ in "bat" versus the Italian /b/ in after the "critical period." Nativelike compe-
"babbo") are more difficult to fully acquire tency is the ability to communicate at high
than completely new phonemes in the sec- levels of ability and proficiency near that of
ond language (Flege, 1987; Meador, Flege, a native first language speaker.
& MacKay, 2000). For example, learning to Bongaerts, van Summeren, Planken,
pronounce the English 1d3/ (not found in and Schils (1997) investigated ultimate at-
Italian) is easier than learning to pronounce tainment with regard to English pronuncia-
the contrasting English /b/ versus the Italian tion. Their participants consisted of 10 na-
/b/ in the initial word position. The above tive speakers of British English, 11 native
mentioned difficulties in acquiring the Eng- speakers of Dutch who were highly success-
lish /b/ versus the Italian /b/ may have to do ful English speakers, and 20 native speak-
with the VOT aspects of these two Romance ers of Dutch with varying degrees of Eng-
language sounds (Brice, Castellon-Perez, & lish proficiency. English speech samples
Ryalls, 2004; Zampani, 1998). were elicited from all the participants and
Flege, Munro, and MacKay (1995) stud- judged by native speakers of British Eng-
ied English pronunciation of English sen- lish. The judges rated the samples on a Lik-
tences by 240 Italian speakers who had ert five-point scale from very strong foreign
moved to Ottawa, Canada between two and accent to no foreign accent at all. The native
23 years of age. The participants had lived in English speakers received very high scores
Canada for at least 15 years. The Italian par- with the very successful Dutch speakers al-
ticipants and a control group of native Eng- so receiving high scores, albeit, not as high
EXAMINATION OF THE CRITICAL PERIOD HYPOTHESIS AND tILTINIATE ATTAINMENT 147
as the native English speakers. The third & Soares, 1986; Li, 1996) in determining
group of non-English speakers with varying which language was activated during bilin-
degrees of proficiency did not score highly. gual perception tasks. In bilingual speech
It should be noted that some English learn- perception tasks it is important to recognize
ers in the successful Dutch speakers group the acoustic differences inherent between
did attain nativelike attainment. Bongaerts the two languages.
et al. (1997) concluded that some learners When comparing Spanish and English,
are not bound by critical period constraints, there are acoustic differences based on the
which he considered to be a "fairly excep- perception of certain phonemes and tim-
tional phenomena" (p. 462). ing features of those phonemes. For exam-
The majority of studies in ultimate at- ple, speech perception at the phonetic level
tainment have focused on morphosyntac- may vary as a result of when the phoneme
tic attainment (Johnson & Newport, 1989); is voiced. Is the phoneme voiced prior to
universal grammar in late language learning articulatory release (i.e., prevoicing), at ar-
(White & Genesee, 1996); phonological at- ticulatory release (i.e., zero onset), or after
tainment (Bongaerts et al., 1997); or speech articulatory release (i.e., lag)? These timing
production (Flege & Schmidt. 1995). The aspects have been quantified with voice on-
question of which factors contribute most set time (VOT) measures.
significantly to second language learning Voice onset time is defined as the time
have been actively debated over numerous between the air pressure release of an artic-
years. In research regarding localization of ulatory constriction and the onset of voic-
two language within the brain, Birdsong ing. In Spanish, the stop consonants /p,t,k/
(2006) stated that are classified as unaspirated stops with
shorter-lag voice onset time (VOT) values
The basic research issue addressed in this (+20 milliseconds [msec]) (Thornburgh &
area of cognitive neuroscience is whether Ryalls, 1998; Zampini, 1998). In English,
processing in the L2 is accomplished in the
these aspirated stops are produced with
same way as processing in the Li. The de-
gree of observed similarity hinges on three
longer-lag VOT values (+60 msec) (Lisker &
principal factors: the age at which L2 ac- Abramson, 1964). The Spanish stop conso-
quisition is begun. the level of L2 proficien- nants of /b,d,g/ are produced with a nega-
cy, and the type of task demanded of the tive or zero VOT (e.g., -20 msec or 0 msec)
subjects (p. 24). or with a shorter lag (+20 msec) than the
typically longer lags (+30 msec) in English
(Zampini, 1998). Because of the longer lags
Bilingual Speech Perception for voiceless and voiced consonants that are
produced in English (i.e., + 30 or +60 msec,
Ultimate attainment in language learn- respectively), English speakers may need
ing is influenced by numerous internal and longer VOT values to perceive the voiceless
external factors. Bilingual speech percep- or voiced stops than do Spanish speakers
tion has not been extensively investigat- (Flege & Eefting, 1986).
ed. Grosjean (2001) emphasized this point A number of studies have used a gating
when he stated that, "There has been far methodology in studying monolingual and
less (if any) systematic research on language bilingual word recognition tasks (Cotton &
mode bilingual language model in the Grosjean 1984; Grosjean 1988; Tyler & Wes-
domain of perception" (p. 10). In addition, sels, 1985). Grosjean (1996) stated that in
few studies have investigated speech percep- the gating task " . . . a spoken language stim-
tion, age of arrival, and ultimate attainment ulus is presented in segments of increasing
in the second language (Flege & Schmidt, duration and subjects are asked to propose
1997). Some studies have used code-mixed the word being presented and to give a con-
utterances (Grosjean, 1998, 2001; Grosjean fidence rating after each segment" (p. 597).
148 ASIA PACIFIC10IJRNAI, OF SPEECH, IAN(;UAGE, ANI) HEARING, VOL. 11, NO. 3
Therefore, gating appears to he useful in de- and Spanish with Spanish). Thus, the partici-
termining the amount of phonetic-acoustic pants were judged to be balanced bilingual
information that a bilingual listener needs speakers perceiving all language conditions,
for the correct identification of the word that is, Spanish, English, and code-mixing.
(Li, 1996). Therefore, the purpose of this pa- In addition, Brice and Ryalls found that nei-
per is to investigate two separate, yet relat- ther group (early, middle, or late bilinguals)
ed studies studying speech perception, age significantly showed an overall faster pro-
of arrival, and a comparison of bilingual and cessing ability. Age of arrival did not seem
monolingual speakers on English speech to impact overall on a balanced bilingual's
perception measuring the critical period of perception of one language over the other
learning and ultimate attainment in the sec- indicating that no absolute age restriction
ond language. on learning was noted (i.e., rejecting no-
tions of an overall effect related to the CPH).
Descriptive differences were noted among
The Research Paradigm: the groups with the middle AOA group per-
Summary of Relevant Research forming better on all four language presenta-
tion conditions further negating the critical
The research design and paradigm for period hypothesis (Lenneberg, 1967).
this study was based on the combination Brice, Castellon-Perez, and Rvalls (2004)
and extension of two previous studies with studied the recognition of code-mixed
Spanish-English bilingual speakers (Brice, words among Spanish-English bilingual in-
Castellon-Perez, & Ryalls, 2004; Brice & Ry- dividuals, specifically investigating the pho-
alls, 2004) and the results from a preliminary notactic aspects of speech and the effects
study with monolingual and bilingual Span- on the dual activation of the two languages
ish-English speakers (Gladwell, 2002). All in speech perception. Do bilingual speakers
studies involved a gating speech identifica- demonstrate a marked bias for one language
tion paradigm. A brief summary of this pre- when presented with bilingual input? By in-
vious research is presented here. vestigating dual language activation in bilin-
Brice and Ryalls (2004) investigated guals, the notion of equal perceptual abili-
speech perception in a group of Spanish- ties related to achieving ultimate attainment
English fluent and equally proficient bilin- was investigated. Thirty Spanish-English flu-
guals divided into three groups of proficien- ent and proficient participants with no re-
cy based on their age of arrival (AOA) or age ported speech, language, or hearing deficits
at which they came to the host country (i.e., heard sentences with the last word divided
the United States in this study). The partici- into gated portions of approximately 70 mil-
pants were exposed to English and grouped liseconds. Target words consisted of voiced
on these ages (i.e., early AOA, 3-8 years; mid- and voiceless initial stop consonants (CC-
dle AOA, 9-15 years, and late AOA, 16-22 voiced, CC-voiceless, CV-tense vowels, and
years) = 30); therefore, assessing the is- CV- lax vowels). The participants were in-
sue of early versus late learning and the in- structed that the target word could be ei-
fluence of the CPH on second language ther in language, that is, Spanish or English.
learning. They were required to verbally name the
The results of this study indicated that word they perceived with 100% confidence
the participants, as a group, were able to and also to identify the word as being either
discriminate between their two languages. Spanish or English.
Descriptively the participants were able to The results indicated that bilingual lis-
differentiate the different code-mixed lan- teners were not able to identify items with
guage conditions under which the stimuli regard to the specific language being heard
were presented (i.e., Spanish with English, when the initial consonant-consonant (CC)
English with English, English with Spanish, was voiceless. The listeners were able to dif-
EXAMINATION OF THE CRITICAL PERI( )I) HYPOTHESIS AND ULTIMATE ATTAINMENT 149
ferentiate items with regard to the specific attainment by the bilingual speakers equiva-
language being heard when the initial CC lent to that of native English speakers.
was voiced. In addition, the bilingual partici-
pants were able to differentiate items with
regard to the specific language being heard Purpose of This Study
when the target vowel was tense. Results of
these comparisons indicated that consonant The purpose of this study was to inves-
voicing and vowel tenseness were signifi- tigate the recognition of Spanish-English
cant factors in speech perception and word and English words in speech among Span-
identification. Lax vowels are not present in ish-English bilingual and monolingual Eng-
Spanish, therefore, the bilingual participants lish speaking individuals in two separate yet
may have been able to disengage their Span- related studies. The first study investigated
ish word list when presented with tense phonotactic aspects and age of arrival on
English words and focus solely on their Eng- speech perception among three groups of
lish word list in word identification and re- bilingual individuals (n = 45). In essence,
trieval. For some conditions. it appeared this study was a measure of L2 proficiency
that fluent bilinguals could apply their im- for speech perception for the three groups
plicit phonotactic knowledge to make pho- of fluent and proficient bilingual Spanish-
notactic decisions. English speakers based on both phonotactic
Balanced bilingual listeners did not seem features and AOA thus assessing the critical
to show an overall marked bias for one lan- period hypothesis (CPH).
guage over the other in making these lan- The second investigation, compared 30
guage decisions; thus, it appeared that they Spanish-English bilinguals with 30 monolin-
recognized each language equally adding gual English participants. This study asked
support to the notion of equal activation in whether fluent and proficient bilingual
both languages indicating the possibility of speakers showed equal perceptual abilities
ultimate attainment in English. when compared to monolingual English
Gladwell (2002) studied speech percep- speakers on English only gated words, thus
tion among bilingual (fluent and proficient measuring their ultimate attainment abilities
Spanish-English speakers) and monolingual in English.
(English speakers) on a gating speech per-
ception task (n = 60); thereby, directly as-
sessing ultimate attainment abilities in bi- Methods
lingual versus monolingual speakers. The
stimuli were identical to that used in the
Brice and Ryalls (2004) study. However, on- Participants: Study One
ly the final target words that ended in Eng-
lish were analyzed. Forty-five Spanish/English fluent bilin-
The results indicated that the Spanish- guals with no reported speech, language, or
English speakers were able to identify voiced hearing deficits participated in this experi-
initial consonants quicker than their coun- ment. Each participant (i.e., an university stu-
terpart English monolingual speakers. Eng- dent in a large metropolitan city in Florida,
lish speakers were able to identify lax vowels U.S.) was required to he between the ages of
quicker than the bilingual speakers. Spanish 18 to 40 years and to have lived in the United
does not contain lax vowels; thus, this pho- States for at least six years. The participants
netic feature is unable to transfer from Span- were selected and divided into three groups
ish to English. Overall, both the bilingual and of 15 participants according to their age of
monolingual speakers perceived the stimuli arrival (AOA) to the United States (i.e., ear-
with equal capacity. Therefore, for speech ly AOA, 3-8 years; middle AOA, 9-15 years;
perception, there seemed to exist ultimate and late AOA, 16-22 years). The closer the
150 ASIA PACIFIC JOURNAL OF SPEECH. LANGUAGE, AND HEARING. VOL. 11, NO. 3
board sampling sound at 44.1-kHz and 16-bit each gate. The setting consisted of a speech
quantization. Response forms were used to laboratory in a quiet room.
collect the participants' responses as to the
language and the word and whether or not
the participants were 100% confident. Results from Study One
VA have different sample sizes, the sum of sumption was maintained. With sphericity
squares for effect plus error do not equal assumed, results indicated significant differ-
the total sum of squares. This causes tests ences for Phonotactic Construction [F(3,56)
of main effects and interactions to be corre- = 114.418, p = .000]. Using partial eta
lated. SPSS offers an adjustment for unequal squared as the measure of effect size, 73%
sample sizes in MANOVA" (p. 2). This study of the total variability was accounted for by
used a repeated measures multivariate analy- Phonotactic Construction. Using Cohen's
ses. Hence, the statistical package that was (1988) criteria (i.e., small = .10-.29; medi-
used for this study (Statistical Package for um = .30-.49; large >.50), the effect size for
the Social Sciences—SPSS) offered a correc- the Phonotactic Construction variable was
tion for the unequal Phonotactic Construc- large. The interaction of Phonotactic Con-
tion data cells. struction by Group was not significant =
.738). Refer to Table 1.
Assessment of Phonotactic
Features: Study One Assessment of Age of Arrival
Features: Study One
In order to answer the primary questions
of the effects of word recognition pertaining Prior expectation that significant results
to Phonetic constructions (CC-voiced; CC- would occur between Groups (early, mid-
voiceless; CV-tense; CV-lax) and age of arriv- dle, late AOA) justified the use of a one-tailed
al (early, middle, late bilinguals), a general test (alpha set at p <.05). Results of the test
linear model multivariate repeated measures of between subjects effects indicated signif-
or a randomized blocks (grouping the ex- icant differences between Groups [F(2,42)
perimental participants into blocks) analy- = 2.520, p = .046]. These results are pre-
sis of variance (RBANOVA) using the Wilks' sented in Table 2. The middle AOA Group
Lambda F test was conducted. demonstrated the lowest mean scores (i.e.,
Mauchly's Test of Sphericity was not sig- lower scores indicated faster perception
nificant (p = .212), thus, the sphericity as- and quicker word recognition) on all Pho-
Partial Eta
Type III Squared
Sum of Mean (Effect
Source Squares DF Square F Significance Size)
Phonotactic
Construction
(Greenhouse- 3,
Geisse r) 5611.101 56 1870.367 114.418 .000* .731
Phonotactic
Construction
by Group
Interaction
(Greenhouse- 6,
Geisser) 57.884 56 9.647 .590 .738 .027
* Significant at p <.05.
154 ASIA PACIFIC JOURNAL OF SPEECH. LANGUAGE, AND HEARING. VOL. 11. NO. 3
Partial Eta
Type Ea Sum Squared
iource of Squares DF Mean Square F Significance (Effect Size;
* Significant at p <.05.
notactic Constructions. Using partial eta Spanish than the CV-tense items in English.
squared as the measure of effect size, 10% Refer to Table 4.
of the total variability was accounted for by
Age of Arrival Groups. Thus, the effect size
for Groups was small. There was no signifi- Discussion of Study One
cant difference according to interaction of
Phonotactic Construction by Age of Arrival
Group (p = .73).
Phonotactic Comparisons
* Significant at p <.05.
restricted set of vowels than English (Brice, kowski • (1980) of an absolute age limit for
Goldstein, Anderson, & So, 1996; Goldstein, second language learners, at least, in the
2001). Spanish vowels are more consistent realm of speech perception.
in length and are typically shorter in dura- Instead, it appears that the balanced bi-
tion. In addition, Spanish vowels are always lingual middle AOA group benefited from
tense. Therefore, these contrasts between having at least six years of exposure in both
Spanish and English vowels may have aided languages. This finding seems to suggest
quicker identification in Spanish. positive transference between the two lan-
The results indicated significant differenc- guages at the speech perceptual level. How-
es between the AOA groups, that is, between ever, it should be noted that the late AOA
subject contrasts indicated differences for group performed the slowest in word iden-
CC-voiced by the middle AOA group. In ad- tification. Therefore, a sensitive period or
dition, differences were found between the window of optimal learning may be present
middle and late AOA groups. The AOA group with the late bilinguals.
means indicated that for all Phonotactic Con-
trasts the middle AOA group (9-15) had the
quickest identification times followed by the Results from Study Two
early (3-8), and late (16-22) AOA groups.
These results did not support a critical
period occurring around puberty as better Independent and Dependent
identification scores occurred with the mid- Variables: Study Two
dle AOA, and presence of an absolute cut-
off age was not noted. Therefore, this study The independent variables for this study
does not support the assertion made by Pat- were Phonotactic Construction (CC-voiced;
15(6 ASIA PACIFIC JOURNAL OF SPEECH, LANGUAGE, AND HEARING, VOL. 1 1, NO. 3
CC-voiceless; CV-tense; CV-lax) and the bi- struction by Group interaction. Using Co-
lingual versus monolingual Groups . The hen's (1988) criteria (i.e., small = .10-.29;
dependent variable (DV) represented a cal- medium = .30-.49; large >.50), the effect
culated percentage of the total number of size for the Phonotactic Construction vari-
gates participants needed to be 100% confi- able was large, whereas, the effect size for
dent to correctly identify the word. interaction of Phonotactic Construction by
Group was small. Refer to Table 5.
Multivariate Analyses:
Spanish-English and Assessment of Phonotactic Features:
English-English Sentences
Study Two
A general linear model multivariate re-
peated measures or a randomized blocks A one-way ANOVAs was conducted to
analysis of variance using the Wilks' Lamb- determine whether there were differences
da F test was conducted. Mauchly's Test of in scores between Groups for the Phonotac-
Sphericity was significant = .01), thus, the tic constructions of CC-voiced; CC-voiceless,
sphericity assumption was not maintained. CV-tense, and CV-lax. Results of the ANOVA
The more conservative Greenhouse-Geiss- analyses indicated significant differences
er adjustment was used. Results indicated between Groups for the Phonotactic Con-
significant differences for Phonotactic Con- structions of CC-voiced [1(1,58) = 16.207,
struction [F(3,56) = 181.58,p = .000]. There p = .000]; and CV-lax [F(1,58) = 5.114, p =
were no significant differences according to .0281. Using r as the measure of effect size,
Group (p = .41). A significant interaction of 46% of the variability was accounted for by
Phonotactic Construction by Group [F(3,56) the Phonotactic Constructions of CC-voiced,
= 17.208, p = .000] was found. Using eta whereas, 56% was accounted for by CV-lax.
squared as the measure of effect size, 76% Thus, the effect sizes for the Phonetic con-
of the total variability was accounted for by structions of CC-voiced and CV-lax were me-
Phonotactic Construction and 23% was ac- dium to large. The ANOVA results are pre-
counted by interaction of Phonotactic Con- sented in Table 6.
Eta
Type III Squared
Sum of Mean (Effect
Source Squares DF Square F Significance Size)
Phonetic
construction
(Greenhouse-
Geisser) 13815.683 3, 56 83.77 181.58 .000* .76
Phonetic
construction
by Group
(Greenhouse-
Geisser) 1309.221 3, 56 25.71 17.208 .000* .23
Significant at p <.05.
EXAMINATION OF THE CRITICAL PERIOD HYPOTHESIS AND ULTIMATE ATTAINMENT 157
Table 6. Word Identification Between Groups by Phonotactic Conditions for Stud y Two
Type HI
Sum of Mean
Source Squares DF Square F Significance r effect size
Significant atp
Table 7. Post Hoc Analyses: One Way ANOVA for Spanish-English and English-English
Sentences Between Groups (One-Tailed Test)
Type III
Sum of Mean effect
Source Squares DF Square F Significance size
CC-voiced
Spanish-English 1510.017 1, 59 1510.017 13.391 .000* .427
CC-voiceless
Spanish-English 28.935 1, 59 28.935 .230 .316 .062
CV-tense
Spanish-English 119.709 1, 59 119.709 1.997 .081 .179
CV-lax Spanish-
English 446.901 1, 59 446.901 3.249 .038* .227
CC-voiced
English-English 1173.363 1, 59 1173.363 16.207 .000* .460
CC-voiceless
English-English 13.485 1, 59 13.485 .165 .343 .052
CV-tense
English-English 5.653 1, 59 5.653 .161 .345 .052
CV-lax English-
English 222.980 1, 59 222.980 5.114 .014* .280
• Significant at p <.05.
and perhaps phonemic awareness skills may all, these results seem indicate that the bilin-
also transfer between the two languages gual, Spanish-English speakers have reached
(e.g., identification and phonemic manipula- a level of ultimate attainment in English with
tion of voiced stop consonants). regard to speech word identification.
The Spanish-English speakers performed
better under the tense vowel conditions (i.e.,
for the Spanish-English sentences and the Overall Implications
English-English sentences). English speak-
ers have greater exposure and experience
producing and perceiving lax vowels; there- It appears that exposure to two lan-
fore, this may account for their better perfor- guages over a period of six or more years
mance under the Spanish-English sentences can yield enhanced abilities in the second
and the English-English sentences. This fea- language, that is, positive language transfer-
ture did not transfer as easily between Span- ence. However, higher levels of attainment
ish and English and has to be learned. As a of speech perception in a second language
consequence, the Spanish speakers were seem to occur from 9 to 15 years of age. This
less able to identify the CV-lax sentences. Bi- is in direct contrast to when abilities should
lingual students acquiring English may have decrease according to the CPH and reveal an
difficulty with lax vowels in speaking or absolute age restriction in terms of language
classroom tasks involving speech perception growth. Therefore, the results reported here
(such as phonemic awareness tasks). Over- have not supported the notion of the critical
EXAMINATION OF THE CR MCA!. PERIOD HYPOTHESIS AND ULTIMATE ATTAINMENT 159
Flege, J. E. (1987). The production of -new" ver- Lisker, L., & Abramson, A. (1964). A cross-lan-
sus "similar" phones in a foreign language: guage study of voicing in initial stops: Acous-
Evidence for the effect of equivalence classifi- tical measurements. W ord, 20, 384-422.
cation. Journal of Phonetics, 15, 47-65. Magiste, E. (1979). The competing language sys-
French, A., Poulsen, J., & Yu, A. Multivariate anal- tems of the multilingual: A developmental
ysis of variance. (2002). Retrieved October 19, study of decoding and encoding processes.
2004, from http://userwww.sfsu.edu/-efc/ Journal of V erbal Learning and V erbal Be-
classes/bio1710/manova/manovanew.htm havior, 18, 78-89.
Gladwell, A. M. (2002). Recognition of code- Meador, D., Flege, J., & MacKay, I. (2000). Fac-
mixed-words by bilinguals and monolin- tors affecting the recognition of words in a
guals: A n extended investigation of profi- second language. Bilingualism: Language
ciency levels. Unpublished Masters Thesis, and Cognition, 3(1), 55-67.
University of Central Florida, Orlando, FL. Moyer, A. (1999). Ultimate attainment in L2 pho-
Goldstein, B. (2001). Transcription of Spanish nology. Studies in Second Language A cquisi-
and Spanish-influenced English. Communica- tion, 21, 81-108.
tion Disorders Quarterly, 23(1), 54-60. Otero, M. (2002). Recognition of code-mixed
Grosjean, F. (1988). Exploring the recognition of words by Spanish/English bilinguals: A fo-
guest words in bilingual speech. cus on proficiency levels. Unpublished Mas-
Grosjean, F. (1996). Gating. Language and Cog- ter's Thesis. University of Central Florida,
nitive Processes, 11(6), 597-604. Orlando.
Grosjean, F. (2001). The bilingual's language Penfield, W. & Roberts, L. 1959. Speech and
modes. In J. Nicol (Ed.), One mind, two lan- brain mechanisms. Princeton, NJ: Princeton
guages: Bilingual language processing (pp. University Press.
1-22). Oxford, UK: Blackwell. Thornburgh, I). F., & Ryalls, J. H. (1998). Voice
Grosjcan, F., & Soares, C. (1986). Processing onset time in Spanish-English bilinguals: Ear-
mixed language: Some preliminary findings. ly versus late learners of English. Journal of
In J. Vaid (Ed.), Language processing in bi- Communication Disorders, 31, 215-229.
linguals: Psycholinguistic and neuropsycho- Tyler, L., & Wessels, J. (1985). Is gating an on-
logical perspectives (pp. 145-179). Hillsdale, line task? Evidence from naming latency data.
NJ: Lawrence Erlbaum Associates. Perception and Psychophysics, 38, 217-222.
Johnson, J., & Newport, E. (1989). Critical pe- White, L. & Genesee, F. (1996). How native is
riod effects in second language learning: The near-native? The issue of ultimate attainment
influence of maturational state on the acquisi- in adult second language acquisition. Second
tion of English as a second language. Cogni- Language Research 12, 233-265.
tive Psychology, 21, 60-99. Wylie, E., & Ingram, D. E. (1999). International
Lenneberg, E. H. (1967). Biological foundations second language proficiency ratings. Bris-
of language. New York: Wiley. bane, Australia: Griffith University.
Li, P. (1996). Spoken word recognition of code- Zampini, M. L. (1998). The relationship between
switched words by Chinese-English bilin- the production on perception of L2 Spanish
guals. Journal of Memory and Language, stops. Texas Papers in Foreign Language
35,757-774. Education, 33,85-99.