SSLA, 22, 499533. Printed in the United States of America.

THE ROBUSTNESS OF CRITICAL

PERIOD EFFECTS IN SECOND
LANGUAGE ACQUISITION

Robert M. DeKeyser
University of Pittsburgh

This study was designed to test the Fundamental Difference Hypothesis (Bley-Vroman, 1988), which states that, whereas children are
known to learn language almost completely through (implicit) domain-specific mechanisms, adults have largely lost the ability to
learn a language without reflecting on its structure and have to use
alternative mechanisms, drawing especially on their problem-solving
capacities, to learn a second language. The hypothesis implies that
only adults with a high level of verbal analytical ability will reach
near-native competence in their second language, but that this ability will not be a significant predictor of success for childhood second
language acquisition. A study with 57 adult Hungarian-speaking immigrants confirmed the hypothesis in the sense that very few adult
immigrants scored within the range of child arrivals on a grammaticality judgment test, and that the few who did had high levels of
verbal analytical ability; this ability was not a significant predictor for
childhood arrivals. This study replicates the findings of Johnson and
Newport (1989) and provides an explanation for the apparent exceptions in their study. These findings lead to a reconceptualization of
the Critical Period Hypothesis: If the scope of this hypothesis is lim-

First and foremost, I thank Anna Fenyvesi and Donald Peckham, who helped with the test design and
were in charge of the data collection, and Mary Connerty who helped with the data collection. I am
grateful to Jacqueline Johnson and Elissa Newport for sharing their original grammaticality judgment
test items with me, to Zoltan Dornyei for providing me with the Hungarian Language Aptitude Test
designed by his student Istvan Otto, and to Istvan Otto himself for providing me with a copy of his
thesis. David Birdsong, Michael Long, Barry McLaughlin, Brian MacWhinney, Christina Bratt Paulston, Liliana Sanchez, G. Richard Tucker, and three anonymous SSLA reviewers provided constructive criticism on drafts of this paper. I gratefully acknowledge the financial support of the University
of Pittsburghs Central Research Development Fund. A preliminary version of this paper was presented at the Second Language Research Forum, University of Arizona, Tucson, October 26, 1996.
Correspondence concerning this article should be sent to: Robert M. DeKeyser, Department of
Linguistics, University of Pittsburgh, Pittsburgh, PA 15260; e-mail: RDK1@pitt.edu.
2000 Cambridge University Press 0272-2631/00 $9.50

499

500

Robert M. DeKeyser
ited to implicit learning mechanisms, then it appears that there may
be no exceptions to the age effects that the hypothesis seeks to
explain.

The popular belief that adults are much worse at learning a second language
than children has been supported in part by the professional literature, especially since Lenneberg (1967). Evidence for the Critical Period Hypothesis,
which states that individuals past a certain age are worse at learning a language than younger individuals, has been accumulating, for both pronunciation (e.g., Asher & Garcia, 1969; Munro, Flege, & MacKay, 1996; Oyama, 1976;
Tahta, Wood, & Loewenthal, 1981) and grammar learning (e.g., Coppieters,
1987; Harley, 1986; Harley & Hart, 1997; Johnson, 1992; Johnson & Newport,
1989, 1991; Patkowski, 1980; Schachter, 1990; Sorace, 1993). Empirical research
has refined the popular concept in a number of ways, however: Children have
an advantage in ultimate attainment, not in rate of learning (e.g., Krashen,
Long, & Scarcella, 1979; Slavoff & Johnson, 1995); the decline of language
learning ability does not suddenly occur around puberty but seems to take
place gradually from ages 6 or 7 to 16 or 17 and beyond (e.g., Bialystok &
Hakuta, 1994; Johnson & Newport, 1989, 1991; Oyama, 1978).
In spite of this body of research, the concept of a critical period for second
language acquisition continues to be a controversial topic. Not only is there
no agreed-upon explanation (see, e.g., Goldowsky & Newport, 1993; Harley &
Wang, 1997; Long, 1990; for a recent overview of explanations, see Birdsong,
1999), but the very existence of the phenomenon is denied or played down by
some researchers (e.g., Bialystok, 1997; Bialystok & Hakuta, 1994, 1999; Birdsong, 1992, 1999; Flynn & Manuel, 1991; Ioup, Boustagui, El Tigi, & Moselle,
1994; White & Genesee, 1996, in the domain of morphosyntax; Bongaerts, 1999;
Bongaerts, Planken, & Schils, 1995; Bongaerts, van Summeren, Planken, &
Schils, 1997; Flege, 1999; Flege, Takagi, & Mann, 1995; Moyer, 1999; Neufeld,
1978, in the domain of pronunciation).
A new impetus for research in this area came from Bley-Vromans (1988)
formulation of the Fundamental Difference Hypothesis. According to this hypothesis, adults can no longer rely on the innate mechanisms for implicit language acquisition and must, therefore, rely on alternative, problem-solving
mechanisms.1 Even though the Fundamental Difference Hypothesis is well
known in the field of second language acquisition, no empirical study on age
effects has been conducted specifically to test the strong prediction it impliesthat is, that only adults with a high level of verbal ability are expected
to succeed fully at second language acquisition. Children, of course, all learn
their native dialect completely, regardless of their level of verbal ability (except in cases of a clear handicap), because they rely on language-specific
mechanisms of implicit learning instead of on general mechanisms for explicit
learning. If the implicit learning mechanisms used by the child are no longer
available, then the adult must bring alternative, verbal-analytic problem-solv-

Critical Period

501

ing skills to the process of language acquisition in order to succeed, and these
analytical verbal skills are characterized by strong individual differences.
Therefore, the Fundamental Difference Hypothesis predicts that those adults
who appear to be successful at learning a second language will necessarily
have a high level of verbal ability. In other words, if the Critical Period Hypothesis is interpreted as applying only to implicit language acquisition, no
exceptions should be foundthat is, no adults should be found who are successful in acquiring a second language without having a high level of verbal
ability, which allows for explicit learning.
Testing this hypothesis was the main goal in carrying out the experimental
research described in this article. A secondary aim was to replicate Johnson
and Newports (1989) study, which found that ultimate attainment in the second language (L2) was strongly correlated with age of acquisition for people
who started acquiring the language before age 17; no correlation with age was
found for those who arrived past age 17. Adult acquirers showed a wide
spread, mostly scoring far below the level of early childhood acquirers, but
with a sizable number of apparent exceptions who scored within the range of
these early acquirers. That study is probably the best known critical period
investigation in the area of L2 morphosyntax, probably largely because of its
very clear-cut results. A number of methodological criticisms of Johnson and
Newports work have been voiced in the last few years, however. The present
study, then, is an attempt at replicating their findings with a very different
population, while avoiding certain methodological problems and testing the
hypothesis that the exceptions can be explained by some adults use of alternative, explicit mechanisms of learningmechanisms not accessible to the
majority of learners.
REVIEW OF THE LITERATURE
Age Effects on the Acquisition of Morphosyntax
in a Second Language
Several researchers (e.g., Eubank & Gregg, 1999; Long, 1990; Schachter, 1996;
Scovel, 1988; Seliger, 1978; Walsh & Diller, 1981) have argued that multiple
critical periods may exist for various aspects of L2 skills; more specifically,
the decline in learners ability to acquire a native-sounding pronunciation may
have different causes and show different age effects compared to the ability
to acquire a nativelike mastery of grammar. The literature review presented
here focuses on ultimate attainment in morphosyntax. Recent reviews for age
effects on ultimate attainment in pronunciation can be found in Bongaerts et
al. (1997), Flege (1999), Harley and Wang (1997), Long (1990, 1993), and Patkowski (1994); a literature review for age effects on rate of acquisition can be
found in Slavoff and Johnson (1995).
Relatively few studies have investigated the effect of age of first exposure
on ultimate attainment in morphosyntax. Patkowski (1980, see also 1990) ob-

502

Robert M. DeKeyser

tained global syntactic proficiency ratings for 67 nonnative speakers of English. He had native-speaking judges rate the subjects on the basis of written
transcripts corresponding to 5-minute recorded segments, thus avoiding a
confounding influence of accent. Subjects who had been exposed to English
before age 15 received much higher ratings on average than those who arrived in the United States after age 15. The pre-puberty group showed a
strong ceiling effect, whereas the ratings for the post-puberty group were
normally distributed. The results for a grammaticality judgment test administered to the subjects were essentially the same as for the syntax ratings.
Hyltenstam (1992) analyzed written retelling and transcriptions of oral retelling of stories by 24 speakers of L2 Swedish whose native language was either Spanish or Finnish, as well as the corresponding data from 12 native
speakers. He found no overlap in the distribution of number of errors between
the native speakers and those who acquired the L2 after age 7; the distribution for those who acquired the L2 before age 6 overlapped with the distributions for both the native speakers and the late acquirers.
Johnson and Newport (1989) administered a grammaticality judgment test
with orally presented sentences, covering a wide variety of basic morphosyntactic structures of English, to a group of 46 native speakers of Chinese and
Korean who had immigrated to the United States at various ages (339). They
found a gradual decline of the level of proficiency from ages 67 to 1617;
proficiency leveled off past that age. Adults showed a wide variety in proficiency but no clear age effect within their group. The correlation between age
of arrival and test score was .77 for all the subjects together and .87 for
those who arrived before age 17. Individual elements of grammar varied
widely in their correlation with age, even though this correlation was significant for all 12 rule types examined.
This study has drawn widespread attention. It is one of the most frequently
cited references in the second language acquisition literature of the last 10
years, but it has also been criticized on a number of points. The length of residence (minimum 5 years) may not have been enough in all cases for the learners to have reached ultimate attainment levels (Bialystok & Hakuta, 1994;
Juffs & Harrington, 1995). The test may have been too long (276 items) for
the participants to concentrate on every item; the test required mental vigor
(Bialystok & Hakuta, 1994, p. 70). Age of arrival was to some extent confounded with age at test taking, which may explain to some extent the lower
scores of the older learners, who may have lost some of their attentional vigilance (Bialystok & Hakuta, 1994, p. 70; see also Bialystok, 1997). The biggest
change may be around age 20 rather than around age 16, and a statistically
significant decline continues into adulthood; therefore there is no strong qualitative change around age 16 (Bialystok & Hakuta, 1994; see also Bialystok,
1997; Bialystok & Hakuta, 1999). Research on learners of an L2 more closely
related to their L1 shows less dramatic age effects (Kellerman, 1995). The latter point is no threat to the concept of a critical period; clearly, the more
closely related the L1 and L2 are, the fewer structures have to be acquired

Critical Period

503

from scratch, and the fewer structures, therefore, are eligible to show an age
effect. The exact shape of the age function is a difficult point; it is hard to find
enough subjects in the critical age range of 1520,2 and the exact point in the
curve where the decline has bottomed out may depend on the structures investigated (see the hypothesis of multiple critical periods mentioned previously) and on individual differences among learners. The other three
criticisms, however, should certainly be taken into account by anybody trying
to replicate Johnson and Newports (1989) findings: Participants should preferably have been using the L2 for 10 years or more,3 any confound of age of
arrival with age of test taking should be avoided, and a shorter test may be
advisable in order to avoid excessive fatigue.
Johnson (1992) replicated the previous study with the same subjects a year
later, this time in a written format. The results were essentially the same as
those obtained with orally presented stimuli, even though the correlation between age and test score was somewhat lower than in the previous study (r =
.54 for all subjects together; .73 for those who arrived before adulthood),
and the number of morphosyntactic structures that showed a correlation between age and test score was smaller.
Johnson and Newport (1991) found very much the same results for oral
grammaticality judgments concerning subjacency in L2 English among 21 native speakers of Chinese who had their first exposure to English between ages
4 and 38. Performance was negatively correlated with age of arrival for those
who immigrated before adulthood (r = .63) and then leveled off to barely
above chance.
Schachter (1990; see also 1989) also investigated subjacency as a function
of age, this time with 79 Korean, Chinese, Indonesian, and Dutch speakers of
L2 English, who had their first exposure to the language after age 12. She
found that many learners in the first three groups had not acquired the constraints on wh-movement, referred to as subjacency, even where the various
forms of wh-movement themselves had been acquired. Korean speakers, in
particular, whose native language shows no evidence of subjacency, performed at chance level. The Dutch speakers, however, whose native language
shows the same range of subjacency phenomena as English, performed at the
same level as the native speaker control group.
Lee (1992, quoted in Schachter, 1996) found not only that Korean learners
ability to judge the grammaticality of reflexives in L2 English, after three years
of exposure, was negatively correlated with age of first exposure after age 14,
but also that it was positively correlated with age of first exposure before age
11. In other words, for this particular element of grammar, there seemed to be
not just a critical period with an end point, but one bounded on both sides,
which Schachter called a window of opportunity (p. 185). Given the relatively short period of exposure, however, this is probably a matter of speed
of acquisition rather than of ultimate attainment.
Even when the L1 and L2 are more closely related, strong age effects have
been documented. Coppieters (1987) administered a grammaticality judgment

504

Robert M. DeKeyser

test on subtle points of French grammar to 21 highly educated and very highly
proficient speakers of French as an L2, six of whom did not even have a detectable accent. The statistical analysis of the results, corroborated by qualitative data from follow-up interviews, showed that there was not even any
overlap in the distribution of the scores from 21 nonnatives with those from
native-speaking controls.
Sorace (1993) obtained similar results in a study with 24 L1 English and 20
L1 French near-native speakers of L2 Italian, who had not been exposed to the
L2 before age 18. Her study was more narrowly focused than Coppieters
(1987) experiment (it dealt exclusively with auxiliary choice) but similar in its
methodology (elicitation of grammaticality judgments). Just as in Coppieterss
study, the results indicated that, even for these near-native L2 speakers whose
length of residence in Italy ranged from 5 to 15 years, intuitions about grammaticality were substantially different from those of a control group of 36 native speakers.
Four studies published so far appear to strike a dissenting note. They all
assess the linguistic competence of highly proficient L2 speakers to determine
whether their knowledge of L2 grammar is indistinguishable from that of native speakers, and all of the studies have been quoted as evidence against the
Critical Period Hypothesis.
Ioup et al. (1994) documented the successful acquisition of Arabic as an L2
by two adult native speakers of English, one tutored, the other untutored. The
differences between the two were very small and both were relatively close to
native norms on a variety of tests. The tutored learner had a doctorate in Arabic. The other individual was a teacher of ESL who had studied Latin and who
paid close attention to inflectional morphology from the beginning of her
learning. At the time of testing, she had been in Egypt for 26 years. Although
both learners did very well on their tests, they were still far from a perfect
approximation of the native norm and, therefore, do not really provide evidence against the Critical Period Hypothesis. The focus of the study, in fact,
was on the role of instruction.
Bialystok (1997) reported briefly on two studies of L2 acquisition as a function of age, one with university students who were native speakers of English
or German and who had started studying L2 French at different ages, and one
with native speakers of Chinese who had immigrated to Canada and learned
L2 English at different ages. In both cases, those who learned L2 at a later
age (after age 15) did better than the younger learners. Bialystok takes this as
evidence against the Critical Period Hypothesis, but the fact that the older
learners did better suggests that what was measured was rate of learning
rather than ultimate attainment. As no minimal length of residence is mentioned in Bialystoks report, it is possible that many learners had not reached
their ultimate level of attainment yet. In that case, the level of performance
is to some extent a function of the rate of learning, and several studies have
documented rate advantages for older learners (Ekstrand, 1976; Krashen,
Long, & Scarcella, 1979; Snow & Hoefnagel-Hohle, 1978). Moreover, for the L2

Critical Period

505

French learners, no residence in a French-speaking environment is mentioned;

it appears that they learned the language largely through study only. For this
reason also, then, the findings do not contradict the Critical Period Hypothesis: This hypothesis only applies to age of acquisition, not age of instruction,
as shown in various studies (e.g., Johnson & Newport, 1989; Oyama, 1978).
The findings of two other studies are somewhat harder to refute. White and
Genesee (1996) tested 19 native, 44 near-native, and 45 nonnative speakers of
English on their knowledge of certain aspects of Universal Grammar. The nearnatives (defined as those who scored in the upper one-eighteenth part of the
assessment scale for a variety of linguistic domains) were indistinguishable
from native speakers on grammaticality judgment tasks for a variety of whmovement constraints. White and Genesee interpreted this as evidence that
at least these UG elements are not subject to critical-period effects, but they
acknowledged that the similarity between the L1 and L2 may have played a
role (74 out of 89 of the L2 English speakers were L1 speakers of a Germanic
or Romance language, and the proportion for the near-native speakers was
probably even higher). Furthermore, the role of age was largely tested by
means of two-way analyses of variance (age group proficiency group); given
the very narrow range of proficiency in scores on the grammaticality test that
can be expected in the near-native category (as per White and Genesees definition) compared to the nonnative category, a main effect for age in that proficiency group is virtually excluded, and only age proficiency interaction
effects can be expected; a couple of significant or marginally significant interaction effects were indeed found. Finally, for any study in which the linguistic
structure under scrutiny is as narrowly defined as in White and Genesee, generalizability to other structures is a serious concern. Even if all learners, regardless of age, learned a specific structure perfectly, this would not be
evidence against the Critical Period Hypothesis: The hypothesis states that
learners past a certain age have trouble learning many structures, not that
all structures are problematic for them. (See also Eubank & Gregg, 1999, for
discussion of White & Genesee.)
Birdsong (1992) administered grammaticality judgment tests on various elements of French morphosyntax to 20 native and 20 advanced nonnative
speakers. Birdsongs findings were strikingly different from those obtained by
Coppieters (1987), whose aforementioned study was similar in design: Although the native speakers did significantly better as a group than the nonnatives, there was considerable overlap in the distribution. The fact that even
native speakers judgments show a high degree of interindividual variability
for certain items in Birdsong might account for the overlap between natives
and nonnatives (see Pulvermuller & Schumann, 1994)only 515% of responses from natives deviate from the norm in the Coppieters study; this percentage varied from 16.7 to 31.6 in Birdsongs studybut the numerous small
differences in methodology, structures tested, and subject populations make
it difficult to point to the reason(s) for the difference between the Birdsong
and the Coppieters findings with any degree of confidence.

506

Robert M. DeKeyser

In conclusion, the preponderance of the evidence is consistent with the

view that the critical period effect is pervasive in L2 acquisition, and, therefore, typically found in the morphosyntactic competence of even advanced
nonnative speakers, but that for the most advanced speakers it may not be
visible, except with subtle testing formats such as those used by Coppieters
(1987).
The Role of Verbal Aptitude
Most research on the role of foreign language learning aptitude in L2 learning
has been carried out in classroom contexts, with teenagers and young adults,
and has consistently found a high correlation between foreign language learning aptitude and success in the L2, typically ranging from .4 to .7, which is
somewhat higher than the correlations typically found between verbal intelligence and success in the L2 (see, e.g., Carroll, 1981, 1990; Sasaki, 1993;
Skehan, 1989; Wesche, Edwards, & Wells, 1982). For grade school students,
however, much lower correlations have been found (e.g., Harley, 1986; Harley & Hart, 1997).
The role of language learning aptitude in untutored acquisition is less clear,
especially in the long run. Harley and Hart (1997), in particular, after showing
that analytic verbal ability was the only significant predictor of various measures of L2 proficiency for students whose immersion started in grade 7, but
generally not a significant predictor of the same proficiency measures for students who entered an immersion program in grade 1, stated that the significant role of verbal ability found in late immersion students may have been
due to the nature of the late immersion programs in question (i.e., their degree of initial focus on the structure of the language to be learned), and that
it remains to be seen whether analytic ability will prove to be an important
predictor of success for older learners outside of the classroom context.
Skehan (1986, 1989, 1998) strongly argued that it should play an important
role there too, because it reflects a generalized capacity to handle language
structure. Perhaps it should play an even bigger role in untutored acquisition
than in classroom learning because, without instruction, the burden of discovering grammatical structures is completely on the learner (see McLaughlin, 1990), and research in educational psychology on aptitude-treatment
interaction has consistently shown that, as more of a specific, informationprocessing burden is put on the student, the corresponding aptitude becomes
a more important predictor of successful learning (see, e.g., Corno & Snow,
1986; Cronbach & Snow, 1981; Snow, 1989; Snow & Lohman, 1984).
Krashen (1981), on the other hand, claimed that analytic ability (aptitude) predicts success in conscious learning, whereas affective variables
(attitude) are the best predictors of subconscious acquisition. It would follow from that point of view that verbal ability usually plays a more important
role in adult learners who receive traditional form-focused instruction and
less of a role in informal acquisition by most adults. The conscious-uncon-

Critical Period

507

scious, formal-informal, or learning-acquisition dichotomy, however, does not

completely coincide with the instructed-naturalistic distinction because some
classroom learners (e.g., in some early immersion programs and some classroom teaching that follows the natural approach inspired by Krashens theories) largely engage in acquisition without reflection on linguistic structure,
whereas some adults with high verbal ability or substantial experience with
linguistic analysis through their education, or both, will engage in such reflection and use it as a tool for learning the language, regardless of the learning
context.
Thus it remains an empirical question how important a role analytic ability
plays in L2 learning by adults whose skills develop over a long period of constant exposure to the L2 and what the consequences are for those adults who
lack high verbal ability, educational experience with language analysis, or
both. Although none of the studies reviewed in the previous section mention
verbal aptitude scores, it is likely, given the populations they worked with
(mostly university faculty for Coppieters, 1987; faculty and students for Johnson and Newport, 1989; all college-educated people in Birdsong, 1992, most
having had many years of formal instruction in French; people who held professional positions or were continuing their education in Patkowski, 1980, p.
451; language professionals in Ioup et al., 1994; university students of language
in Bialystoks 1997 study), that the role of verbal aptitude explains the apparent exceptions to a strong age effect. A high level of verbal aptitude may allow
L2 speakers to perform morphosyntactically like native speakers. This may explain the right tail of the distributions in Patkowski as well as Johnson and
Newport, the impressionistic data that led to the inclusion of subjects in Coppieterss study, the partial overlap of the native and nonnative distributions
in Birdsong, and the lack of age effects in at least one of Bialystoks studies
(along with the aforementioned factors).
A major goal of this study, then, besides replicating Johnson and Newports
(1989) study to assess the effect of age on ultimate attainment in L2 morphosyntax, is to assess the effect of verbal abilitymore specifically, foreign language learning aptitude, on ultimate attainment. Studying the effects of age
and verbal aptitude, and their interaction, should be more revealing than
studying them separately: If the aptitude effect varies with age and the age
effect varies with aptitude, this interaction can shed light on the cause of each
effect and allow for a test of the Fundamental Difference Hypothesis.
Hypotheses
The literature presented in the previous section, in particular Johnson and
Newport (1989) and Bley-Vroman (1988), led to the following hypotheses:
1. Participants in this study will show a strong negative correlation between age of
arrival and performance on a grammaticality judgment test, but with some overlap
in range between child and adult acquirers.

508

Robert M. DeKeyser

2. Those adult acquirers, however, who score within the range of child acquirers will
all have high verbal aptitude, which may have allowed them, at least in part, to
learn the L2 grammar through explicit reflection on rules.
3. Different elements of grammar will show different correlations with age of acquisition; not all structures are equally sensitive to the critical period effect.

METHOD
Participants
Fifty-seven native speakers of Hungarian participated in this study. Hungarian
was chosen because it was desirable to have speakers of one language in order to eliminate variability due to the L1 within one study, and because even
the basic structures of English morphosyntax present many elements that
have to be acquired by speakers of a non-Indo-European language such as
Hungarian. Other desiderata were a long period of residence, to eliminate the
risk of confounding ultimate attainment with rate of acquisition, and a wide
spread in the two main independent variables, age of arrival and socioeconomic status (the latter being a first approximation of verbal ability). The Hungarian community around Pittsburgh met all these requirements.
Of the 57 participants, 32 were male and 25 female. They all lived in a 200mile radius around Pittsburgh (western Pennsylvania, eastern Ohio, northern
West Virginia, western Maryland) and were recruited by advertising in Hungarian-American newsletters, by distributing fliers at community events, and
by word of mouth in the community. All participants were volunteers and
were paid $20 for their efforts. Sixty Hungarian Americans volunteered, but
the data from three (elderly) people had to be discarded owing to problems
with hearing the test stimuli or concentrating on the test. Of the 57 remaining
participants, 42 (25 males and 17 females) were older than 16 when they immigrated; 15 (8 females and 7 males) were younger than 16 at the time of their
arrival in the United States (or, in a few cases, a different English-speaking
country before moving on to the United States); the range of age of arrival was
140. All participants had resided in the United States for at least 10 years;
average length of residence was 34 years (35.6 for the younger and 33.7 for
the older group). Average age at testing was 55 (43.2 for the younger and 60
for the older group), ranging from 16 to 81. Level of education and occupational status also varied widely, from blue-collar workers with an eighth-grade
education to professionals with doctoral degrees; the average number of
years of schooling was 14.6 for the younger and 13.4 for the older group.
Although all the participants were native speakers of Hungarian, some had
also had varying degrees of exposure to other languages such as Russian,
Czech, Slovak, Romanian, German, or Hebrew before moving to an Englishspeaking country. None reported any substantial exposure to English before
emigrating. When asked about their proficiency in Hungarian compared to English now, 22 said they still felt more comfortable in Hungarian than in English

Critical Period

509

(1 in the younger and 21 in the older group), 20 said they were more comfortable with English (10 in each group), and 15 said it made no difference (4 in
the younger and 11 in the older group).
The measures for age of arrival, age at time of testing, aptitude, years of
schooling, and grammar test scores for all individuals are given in Appendix A.
Instruments
4

Grammaticality Judgment Test. The grammaticality judgment test used

by Johnson and Newport (1989) was made available to me by the authors and
was adapted in a variety of ways. The test was shortened to include 200 instead of 276 items; most subcategories now include six rather than eight items
(three- rather than four-item pairs). Four practice items were added at the beginning to ensure the testing procedure was understood perfectly by all participants. A few items were deleted or changed on J. Johnsons advice (personal
communication, May, 1996). Several subcategories were deleted or reorganized; a few were added to include structures that tend to be problematic for
Hungarian speakers. The reliability coefficient (KR-20) obtained was .91 for
grammatical items and .97 for ungrammatical items.
The complete set of items is given in Appendix B; the items in the Appendix
are grouped according to the structure being tested. Incorrect items are
marked with an asterisk. Items from Johnson and Newport (1989) are marked
JN; new items are marked DK. A more detailed list of changes compared to
the test in Johnson and Newport is given in Appendix C.
Language Learning Aptitude Test. Carroll and Sapons (1959) Modern Language Aptitude Test is usually considered the best verbal aptitude test in
terms of its predictive validity for L2 learning. Some of the minor technical
problems discovered over the years do not affect the Words in Sentences
part (Carroll, 1990), which is specifically aimed at measuring grammatical sensitivity and therefore should be the best predictor of grammar learning. (This
subtest has been shown to correlate highly with verbal intelligence or even
general intelligence, however. Wesche, Edwards, and Wells, 1982, for instance,
found a correlation of .56 with the overall score on Thurstone and Thurstones
[1962] Primary Mental Abilities.) Because the participants in this study were
native speakers of Hungarian, a corresponding test in Hungarian was needed.
The Hungarian Language Aptitude Test, Words in Sentences (Otto, 1996b), is
an adaptation of the Words in Sentences part of the Modern Language Aptitude Test (Carroll & Sapon). The test consists of 20 five-way, multiple-choice
items; these were selected as the best items from a pool of 50 items piloted
with 177 high school students in Budapest. The difficulty level for the 20 items
ranged from .23 to .68 with a mean of .52. Item-total correlations were all significant at the p < .001 level, with a mean of .52 (Z. Dornyei, personal communication, October 21, 1996; see also Otto, 1996a).

510

Robert M. DeKeyser

Background Questionnaire. All participants filled out a two-page questionnaire about their language background, educational background, age of arrival
in North America, and age at the time of the test.

Procedure
The tests and the questionnaire were administered individually to the participants at their home (or, in some cases, a room on campus). They filled out
the background questionnaire first, with the help of the experimenter.
The grammaticality judgment test came immediately after the questionnaire. The 200 items were tape-recorded by a male native speaker of English,
with one of the two items of each correct-incorrect pair in the first 100 items,
and the other in the second 100. Within each set of 100, the items came in a
fixed random order, with the exception that no two items of the same category ever occurred consecutively. Each item was read twice in a row, with a
3-second interval between repetitions and a 6-second interval between items.
Participants indicated whether they considered an item to be correct or incorrect by blackening the corresponding circle on an standardized answer sheet
for computer processing. The tape recording guided them through the first
four items, giving feedback on which were the correct answers. Participants
could ask the experimenter to stop the tape at any time if they were tired;
there was always a break of a couple of minutes after the first 100 items, while
the tape was turned around and rewound. The test took about 55 minutes.
Finally, the aptitude test was administered; 20 minutes were allotted for
this test. The participants read the stimulus sentences on a test sheet, and
marked their answers on a specially designed multiple-choice answer sheet
(Otto, 1996b).
RESULTS
Age of Acquisition and Ultimate Attainment
The correlation between the grammaticality judgment test score and age of
acquisition was .63 (p < .001) for the group as a whole (N = 57). For adult arrivals (n = 42), the correlation was .04 (ns); for participants who arrived before the age of 16 (n = 15), the correlation was .26 (ns). Figure 1 presents a
scatterplot for age of acquisition and grammaticality judgment test score.
Separate analyses were performed for high- and low-aptitude learners. The
average score on the aptitude test was 4.7 out of 20 (4.3 for the younger and
4.9 for the older group), with a standard deviation of 2.79. A somewhat arbitrary cut-off point was established between high- and low-aptitude learners, at
a place where the high-aptitude group would be sufficiently different from the
average, but without making it too small for statistical analysis. High aptitude
was operationalized as having a score of 6 or more (at least .46 standard deviations above average); the resulting high-aptitude group was composed of 15

Critical Period

511

Figure 1. Scores on the grammaticality judgment

test (out of 200) as a function of age of acquisition.
individuals, roughly the upper third of the distribution.5 For high-aptitude
learners, the correlation between age of acquisition and grammaticality judgment test score was .33 (ns); for the group with average or low aptitude (n =
42), the correlation was .74 (p < .001).
The Role of Aptitude
The correlation between the grammaticality judgment test score and the aptitude score was .13 (ns) for the group as a whole (N = 57). For adult arrivals
(n = 42), the correlation was .33 (p < .05); for participants who arrived before
the age of 16 (n = 15), the correlation was .07 (ns). Aptitude scores did not
correlate with age of arrival (r = .09, ns).
Individual Structures
For grammatical items, one cannot know which part of the sentence a test
taker does not like in a case where the item is rejected; for accepted items, of
course, every part was accepted. Only ungrammatical items allow for the anal-

512

Robert M. DeKeyser

ysis of specific types of errors. Although it is never 100% certain that a given
test taker has rejected an item because of the error built in by the test designer (see, e.g., Kellerman, 1995), this built-in error is certainly by far the
most likely reason for the rejection and can safely be interpreted as such
when summing results over individuals, with only a small margin of error.
Therefore, following Johnson and Newport (1989), I only included the 100 ungrammatical items in the analysis of errors for individual structures.
Contrary to Johnson and Newport (1989), however, in order to avoid the
language-as-fixed-effect fallacy (Clark, 1973), that is, the error of assuming
that all instances of a given structure are equally good tests of that structure,
scores were not summed over items testing the same structure. Instead, the
scores for individual items were correlated with age of arrival, and the items
were ranked according to this correlation. In this way, one can see whether a
given structure correlates moderately with age of arrival because all items
correlate moderately, or because some correlate and others do not. Table 1
shows the ranking of items from highest to lowest correlation with age of arrival.
Dividing the items into three groups, depending on whether the correlation
of the corresponding scores with age of arrival was high (r significant at p <
.01), marginal (.01 < p < .05), or low (r not significant at p .05), consistently
or almost consistently high correlations with age of arrival were found for the
following structures: present progressives with auxiliary omitted (PPAO: 3/3
items in top group), determiners omitted (DEOM: 3/3 items in top group), determiners used with abstract nouns (DEAB: 4/4 items in top group), plurals
marked on mass nouns (PLMN: 3/3 items in top group), wh-questions without
do-support (WHNA: 3/3 items in top group), wh-questions without subject-verb
inversion (WHNI: 3/3 items in top or in-between group), irregular plurals regularized (3/3 items in top or in-between group), wrong subcategorization of
verb for gerund, infinitive, to + infinitive (SUBC510: 5/6 items in top group),
and adverbs between the verb and the object (WOAD: 2/3 items in top group).
Consistently low correlations were found for word-order problems in declarative sentences not involving adverbs (WODO, WOIO, WOPP, WOVS: 12/12
items in bottom group), yes-no questions that lack do-support (YNVS: 3/3
items in bottom group), and gender errors in pronouns (PRGE: 4/4 items in
bottom or in-between group).
DISCUSSION
Hypothesis 1
The hypothesis of a strong negative correlation between age of acquisition
and score on the grammaticality judgment test was confirmed. A correlation
of .63 was found, which is similar to the .77 reported by Johnson and Newport (1989). More importantly, the extent of overlap between the range of
scores for adult acquirers and for acquirers below age 16 is very small.

Critical Period

513

Table 1. Grammaticality judgment test items ordered by size of correlation

with age of arrival
p < .01
.5710 PPAO02
.5417 TPSO01
.5249 SUBC07
.5171 DEAB02
.5168 WOAD02
.5146 YNAA01
.5142 PLMN01
.5105 PLMN02
.5101 PLMN03
.4958 SUBC10
.4837 DEAB04
.4669 YNDT03
.4639 PLIR03
.4386 SUBC06
.4358 DEOM02
.4354 DEAB03
.4204 TPSO03
.4135 WHNA01
.4106 WHNI02

.4071 PPAO03
.4026 PTII01
.4026 YNDT02
.3964 WHNA02
.3961 PTIR03
.3941 SUBC08
.3940 PMTF02
.3896 PTMO02
.3871 DEOM01
.3833 PROM03
.3821 WHNA03
.3803 PPAO01
.3767 TPSM02
.3743 WOAD01
.3729 DEOM03
.3632 TPSM01
.3592 DEAB01
.3521 SUBC05
.3410 PLIR01

.01 < p < .05

.3396 PMSE01
.3339 SUBC03
.3387 SUBC02
.3344 YNAV03
.3316 PTIR02
.3295 WHNI03
.3246 YNAV01
.3244 PTMO03
.3141 WHNI01
.3072 PPMO01
.2767 PLIR02
.2743 PRGE04
.2696 PTII02
.2693 PMSE03
.2599 PLMO02
.2573 YNAA02

p .05
.2565 PPMO02
.2530 TPSM04
.2516 YNDT01
.2441 PTII03
.2423 PROM01
.2408 TPSO02
.2407 PROM04
.2360 PMTF01
.2326 PLMO03
.2289 PMTF04
.2281 YNVS01
.2279 WOIO03
.2259 SUBC09
.2180 YNVS02
.2156 WOVS03
.2078 WODO03
.2071 PRAC03
.1918 TPSM03
.1903 PTMF03
.1832 TPSM05
.1777 WOIO02
.1766 PMSE04
.1757 PRGE01

.1735 PTMO01
.1719 PPMO03
.1685 PROM02
.1609 YNAA03
.1504 WOPP03
.1491 WOAD03
.1458 WOPP01
.1452 WOIO01
.1294 SUBC01
.1265 PLMO01
.1220 PTIR01
.1093 WOPP02
.1052 PRAC01
.0984 WODO01
.0915 YNVS03
.0814 YNAV02
.0649 PRGE03
.0572 SUBC04
.0267 PMSE02
.0218 WOVS02
.0189 WODO02
.0116 PRGE02
WOVS01

Note. PTMO = past tense marking omitted; PTIR = irregular verbs regularized; PTII = regular ending on irregular stem;
PLMO = plural marking omitted; PLIR = irregular plurals regularized; PLMN = mass nouns used with plural marker;
TPSO = third-person -s omitted; TPSM = third-person -s marked on main verb after modals; PPMO = progressive -ing
omitted; PPAO = progressive auxiliary omitted; DEOM = determiner omitted; DEAB = determiner used with abstract
nouns; PROM = Pronoun omitted (transfer from Hungarian object pro-drop); PRGE = gender errors; PMSE = phrasal
verb separation not allowed; PMTF = phrasal verb separation allowed, but particle moved too far; SUBC = subcategorization; YNAA = Aux Aux order; YNAV = Aux Verb order; YNVS = Verb Subject order; YNDT = double tense marking;
WHNI = no aux inversion; WHNA = no aux; WODO = S V DO order violated; WOIO = S V IO DO order violated; WOVS =
S V order violated; WOPP = S V PP order violated; WOAD = incorrect adverb placement.

Whereas roughly half of the adult arrivals in the Johnson and Newport study
scored within the range of the pre-16 arrivals, only a few adult arrivals performed as well in the present study. As Figure 1 shows, the dividing line between those who started acquiring English before age 16 and those who
started later falls around a score of 180 (out of 200). Among the younger arrivals, only 1 out of 15 scored below that point (a person with an eighth-grade
education and a blue-collar job); among the older arrivals only 3 out of 42
scored above that point.
Johnson and Newport (1989) reported an even stronger correlation between age of arrival and test performance for the early arrivals (r = .87) than
for the group as a whole, but that result was not replicated here; among those
who arrived before age 16 the correlation was .26 (ns). This discrepancy is
hard to interpret because, in both the present study and Johnson and New-

514

Robert M. DeKeyser

ports, the number of data points in the age range of 1216 was very small
(2 and 5, respectively).2 It should be pointed out, however, that Johnson and
Newports data show a dip in proficiency scores for the age range 1216 compared to both earlier acquirers and adults (see their Figure 2, p. 80). As there
is no theoretical explanation for such a decline in adolescent learners only,
the dip in their Figure 2 is probably an artifact, and the correlation coefficient
of .87 probably strongly overestimates the true r.
To assess Bialystoks (1997; see also Bialystok & Hakuta, 1994) interpretation of Johnson and Newports (1989) correlation coefficients (i.e., test scores
correlated strongly with age of arrival, because the latter was confounded
with age at the time of the test), partial correlation coefficients were calculated. The correlation between test score and age of arrival remained a robust
.54 (p < .001) after age at time of test was partialed out; the correlation between test score and age at time of test with age of arrival partialed out was
.13 (ns). These results confirm Johnson and Newports interpretation of their
correlation coefficients and refute Bialystoks criticism.
Length of residence, which was another potential confound with age of arrival, turned out not to be correlated with test scores at all in my data; r = 0.00
(sic). This clearly shows that length of residence no longer plays a role past
the first 10 years (see Oyama, 1978).
Hypothesis 2
It was predicted that no adult acquirers would score within the range of child
acquirers unless they had high verbal aptitude. Using the same operational
definition of high aptitude as above (a score of 6 or higher on the aptitude
test), it was found that, among the participants who started acquiring English
after age 16 but obtained a high score on the grammaticality test (over 180,
i.e., in the upper right quadrant of Figure 1; or even including those who came
close, with a score over 175), all but one had an aptitude score of 6 or above.
Their aptitude scores and grammaticality judgment test scores were 11/177,
10/190, 8/194, 7/184, 6/176, and 3/186. This last case, the only exception, is
that of a participant in the study who was doing postdoctoral studies in the
natural sciences; this suggests that he must be of above-average analytical
ability and that his aptitude test score is not indicative of his analytical abilities. It should also be pointed out that from age 15 to age 26 he lived in Israel
and that he felt most comfortable in Hebrew, followed by English, followed by
Hungarian. (Of the other five highly successful high-aptitude learners, two said
they were more comfortable in English, one in Hungarian, and two equally proficient in both.) Even though the data are not 100% clear-cut, one cannot escape the conclusion that there were few adult acquirers who scored within
the child acquirers range or even came close (6/42) and that none of those
who did were clearly of average or below-average aptitude. This finding is in
line with what the Fundamental Difference Hypothesis predicts: The only way
that an adult learner can achieve grammatical competence similar to that of a

Critical Period

515

native is by using analytical, problem-solving abilities, because the implicit

learning mechanisms of the child are no longer available or accessible. Therefore, only learners with above-average analytical abilities can eventually reach
a near-native level, at least on the rather easy kind of test used here, which
does not include rare constructions or convoluted sentences and does not require participants to perform under time pressure. On a more subtle test of
the kind used by Coppieters (1987), even they, of course, may clearly fall
short of native standards.
Further evidence for the different role that analytical abilities play in second language acquisition by children and adults can be found in the correlations between aptitude and grammaticality judgment scores for the child
arrivals (r = .07; ns) and the adult arrivals (r = .33; p < .05). These figures are
remarkably similar to those that Harley and Hart (1997) found for the relationship between aptitude and proficiency: .15 to .09 for early immersion and .41
to .45 for late immersion. Whereas the younger acquirers in the present study
all reached a native or near-native level regardless of aptitude, only the adults
with above average aptitude eventually became near native.
For the interpretation of these results, it is useful to know that years of
schooling did not predict grammaticality judgment test scores (r = .006 for the
younger and .08 for the older group). Clearly, aptitude plays a role independent of schooling, and it only plays that role for adult learners.
Hypothesis 3
As hypothesized, different structures showed different degrees of correlation
with age of arrival. Structures for which the test scores correlated highly with
age of arrival were the present progressive auxiliary, articles (determiners),
wh-questions, plurals, subcategorization, and adverb placement. Structures
that did not show differential proficiency as a function of age were word order
in declarative sentences (with the exception of adverb placement), do-support
in yes-no questions, and pronoun gender. A comparison of these results with
the ones obtained by Johnson and Newport (1989, p. 87, Figure 3) shows many
similarities:6 Scores for articles, plurals, and subcategorization strongly reflected age of arrival in both studies; scores for word order and for do-support
in yes-no questions hardly correlated with age of arrival in either study (see
Johnson & Newport, 1989, p. 87, fn. 8).7 Adverb placement, present progressive auxiliary, and pronoun gender were not analyzed separately by Johnson
and Newport but lumped together respectively with other word-order problems, missing progressive -ing, and missing pronouns. Finally, Johnson and
Newports wh-question included items in which the error is one of a missing
argument rather than of wh-movement, which may explain why those items
appeared relatively easy in their study compared to the present one, in which
the wh-question category consisted of two types of items only: lack of auxiliary and lack of inversion. In conclusion, then, to the extent that the item classification in the two studies can be compared, all structures that clearly fell

516

Robert M. DeKeyser

into either the easy (low correlation with age) or difficult (high correlation
with age) category in this study fell into the same category in Johnson and
Newports.
The fact that certain structures seem impervious to otherwise strong age
effects begs for an explanation. What characteristics do pronoun gender, subject-verb inversion and do-support in yes-no questions, and basic word order
have in common that could explain why they are learned easily, that is, by
virtually everybody, even by adults whose verbal ability is below average?
Transfer from Hungarian cannot be an explanation because Hungarian makes
no gender distinction in personal pronouns and has no do-support, no subjectverb inversion for questions, and a rather variable word order (see, e.g., Kenesei, Vago, & Fenyvesi, 1998). What errors against the three structures in question do have in common is that they are all perceptually salient. Pronoun
gender errors are so irritating to native speakers that they will almost always
correct them when their nonnative interlocutors make such mistakes, even
though overt correction of grammar errors is otherwise rare in adult nativenonnative interaction; therefore the gender distinction becomes very salient
to the learner. In the case of basic word order, errors lead to differences from
the norm that are very salient themselves: sentences that begin with a (lexical) verb (WOVS, e.g., *Bites the dog), end with a verb (WODO, e.g., *The girl
the movie likes; WOPP, e.g., *The students to the movies went), or begin with
two consecutive noun phrases (WOIO, e.g., *The woman the policeman asked
a question). In all three cases a salient position (sentence initial or sentence
final) is occupied by a syntactic constituent that can never occupy that position in English (WOVS, WODO, WOPP) or it is occupied twice (WOIO). Finally,
in the case of yes-no questions without do-support or subject-verb inversion,
the error has the same highly salient effect: a lexical verb in initial position.
This last case is particularly interesting because it contrasts clearly with
errors in wh-questions, which are similar and yet go undetected by most
adults: lack of inversion (WHNI, e.g., *What Martha is bringing to the party?)
and lack of do-support (WHNA, e.g., *What they sell at the corner store?). In
these cases the errors are less salient because they occur after the question
word that occupies the initial, salient position; as a result they go undetected.
It is known that salience of specific structures plays a role in the ease or
difficulty with which they are acquired in an L2. Bardovi-Harlig (1987), for instance, showed in a study of 95 college-age learners acquiring L2 English that
wh-questions with preposition stranding (e.g., Whom did John give the book
to?) were acquired before pied piping (e.g., To whom did John give the book?),
even though the latter is unmarked and would therefore be expected to be
acquired first. As Robinson (1996) pointed out, preposition stranding is characterized by the fact that the question word and the preposition occupy the
two most salient positions in the sentencefirst and last. Salience clearly prevailed over markedness in Bardovi-Harligs study. In the same way, in the
present study, the salient inversion pattern in yes-no questions was learned
before inversion in wh-questions, even though the latter is considered the

Critical Period

517

least marked (Eckman, Moravcsik, & Wirth, 1989; Greenberg, 1963). Doughty
(1991) also provided evidence for the importance of the salience factor in L2
acquisition by showing that learners who had received treatment consisting
of bringing to prominence (p. 462) the structural components of relative
clauses did as well on a subsequent production task as the learners who were
taught the rules.
It is also known that perceptual salience interacts with the implicit-explicit
learning distinction. Reber, Kassin, Lewis, and Cantor (1980, Experiment 1, pp.
494497), in particular, showed that the structure of letter strings representing an artificial grammar was better learned explicitly than implicitly when the
stimuli were organized in a way that made their similarities salient, and it was
better learned implicitly than explicitly when the similarities were not made
salient. The explicit-salient combination was by far the best of the four. Therefore, it is not surprising that more salient patterns are learned before less salient, yet otherwise similar, structures by adult learners, who, as argued
above, have to rely on explicit learning.
The nature of the few structures that were learned well by the adults in the
present study is further evidence, then, for the fact that L2 grammar learning
in adults is largely limited to explicit processes: Not only are the learners with
high ability for explicit learning the only adults who are successful overall, but
those structures whose salience makes them particularly good candidates for
explicit learning are also the only ones that are learned by all immigrants, regardless of age.
In conclusion, perceptually very salient aspects of morphosyntax such as
basic word order and pronoun gender appear to have been acquired by most
learners at any age of arrival, whereas many other basic structures such as
the use of articles, the use and position of auxiliaries, the position of adverbs,
certain elements of verb subcategorization, and even some uses of the plural
morpheme had not been learned well by many of the learners who arrived
after the age of 16, not even after decades of residence in an English-speaking
environment.
CONCLUSIONS AND IMPLICATIONS
This study has addressed a number of questions and challenges in the SLA
literature. It has accepted Longs (1990) challenge to look at what grammaticality judgments for specific structures reveal about the linguistic competence
of very advanced nonnative speakers who learned the L2 as adults.
It replicated Johnson and Newports (1989) landmark study and found remarkably similar results, having addressed the various methodological criticisms leveled at Johnson and Newport (1989) by Bialystok and Hakuta (1994),
Bialystok (1997), Kellerman (1995), and others, and having investigated a population that differed significantly from Johnson and Newports in terms of native language, education, and socioeconomic status.
This study provided an explanation for why certain learners and certain

518

Robert M. DeKeyser

structures appear to be exceptions to the critical period effect. Learners with

high verbal ability can use explicit learning mechanisms to bypass the increasingly inefficient implicit mechanisms, and certain structures, by virtue of their
saliency, can be learned explicitly by virtually all learners, regardless of verbal
ability.
It provided evidence for Bley-Vromans (1988) Fundamental Difference Hypothesis by showing that no adults reached a native level of competence in L2
morphosyntax unless they had been able to rely on explicit, analytic, problemsolving capacities.
Also, this study answered Harley and Harts (1997) question about the role
of language-learning aptitude in naturalistic acquisition by showing that aptitude, as measured by a test of grammatical sensitivity, is a predictor of ultimate attainment in L2, even after decades of exposure to the language in
nontutored contexts. More importantly, this study showed that age and aptitude interact in the sense that (a) age of arrival makes a clear difference for
those who have average or below-average verbal ability and (b) verbal ability
makes a difference for those who start to learn an L2 as adults. Aptitude plays
no role in ultimate attainment by child learners because they can rely on implicit learning mechanisms, which do not draw on their problem-solving abilities. It is, however, a necessary (but not a sufficient) condition for the ultimate
attainment of near-native competence by adults because they have, to a large
extent, lost these implicit mechanisms and, therefore, have to be able to draw
on alternative, explicit mechanisms to achieve a high level of success.
Most importantly, however, this study suggests that there really is a critical, and not just a sensitive or optimal, period for language acquisition, provided that the Critical Period Hypothesis is understood narrowly enough, that
is, applying only to implicit learning of abstract structures. As long as L2 competence is assessed without regard for the learning mechanisms that produced it, it may appear that there is merely an optimal age for language
learning, in the sense that there is a sizable negative correlation between age
of acquisition and ultimate attainment. If the Critical Period Hypothesis is constrained, however, to implicit learning mechanisms, then it appears that there
is more than just a sizable correlation: Early age confers an absolute, not a
statistical, advantagethat is, there may very well be no exceptions to the
age effect. Somewhere between the ages of 67 and 1617, everybody loses
the mental equipment required for the implicit induction of the abstract patterns underlying a human language, and the critical period really deserves its
name.
Such a conclusion comes as no surprise to those familiar with the study of
artificial grammar learning in experimental psychology, where more and more
evidence suggests that adults capacity to learn abstract patterns implicitly is
extremely limited (see, e.g., Perruchet & Pacteau, 1991; Redington & Chater,
1996; Shanks & St. John, 1994), even though, of course, adults can induce abstract patterns explicitly and can learn to associate concrete elements implicitly.

Critical Period

519

This restriction of the concept of critical period does not automatically answer the question of its ultimate cause, but it is an important step in that direction. Most literature reviews mention four different hypotheses: (a) Adults
are more self-conscious and less able to identify with speakers of the target
language than children, (b) adults receive less adequate input than children
do, (c) adults have different cognitive mechanisms, and (d) adults show the
effects of reduced neurological plasticity (see, e.g., Birdsong, 1999; Harley,
1986; Harley & Wang, 1997; Long, 1990; Singleton, 1989). If the critical period
is defined as an absolute phenomenon, in the sense of unavoidable loss at a
certain age of the capacity to induce abstract patterns implicitly, then affective variables are clearly an inadequate explanation. They may explain part of
the large variance typically seen for adults (see Figure 1; see, e.g., also Johnson & Newport, 1989; Krashen, 1981; Patkowski, 1990); in other words, they
may correlate with ultimate achievement, but they show too much interindividual variation to account for as absolute a phenomenon as the critical period. (For other arguments against affect as an explanation for age effects, see
Long, 1990.)
Input to adults must also vary from one individual situation to another.
More importantly, input differences are not a good explanation for age effects,
because it is precisely in the linguistic domain where input varies the least
phonologythat the age effects are most readily apparent, and it is at the
stage where the comprehensibility of input should be the least problematicin the later stages of acquisitionthat adults clearly perform worse
than children. (For other arguments against input as an explanatory factor,
see Long, 1990.)
As neither input nor affective variables can explain the range of phenomena observed in the literature, differences in cognitive functioning must certainly be involved; to what extent these follow from independent neurological
changes is still an open question. It may be that the severe decline of the ability to induce abstract patterns implicitly is an inevitable consequence of fairly
general aspects of neurological maturation and that it simply shows up most
clearly in language acquisition, because any human language is an exceedingly
complex web of highly abstract patterns. It may also be that (not neurologically determined) developmental differences in memory are involved, either
quantitative (Deacon, 1997, ch. 4; Elman et al., 1996; Goldowsky & Newport,
1993; Newport, 1990; Kareev, Lieberman, & Lev, 1997) or qualitative in the
sense of cognitive restructuring (Birdsong, 1994; Flege, 1992; Karmiloff-Smith,
1992; Wode, 1994). To what extent the change that researchers call the critical period phenomenon follows from deeper cognitive principles, and to what
extent these principles may follow from aspects of neurological maturation is
a question that will probably be with us for some time to come.
Further research should be carried out with other structures in other languages to ascertain that the most salient structures are learned by virtually
all adult learners, and the less salient abstract patterns only by those with
high verbal ability. If that research confirms the patterns found in this study,

520

Robert M. DeKeyser

the search for a neurological basis for the maturational decline in learning will
come into clearer focus: Is there anything in developmental cognitive neuropsychology that can explain why the ability to induce abstract patterns implicitly declines with age?
In the meantime, it is important that the practical implications of age effects in L2 learning not be overstated. The maturational effects described in
this study do not mean, for instance, that children should simply learn a foreign language in elementary school rather than in high school (Patkowski,
1994). Children are better than adults at acquiring a language implicitly, not
at figuring out its structure explicitly (on the contrary, adolescents and adults
are far better at that because of their higher level of general cognitive maturity). Implicit acquisition processes, however, require massive amounts of input, which only a total immersion program can provide, not a program
consisting of a few hours of foreign language teaching per week. The findings
of this study do imply, however, that explicit learning processes are a necessary condition for achieving a high level of competence in a nonnative language after childhood. Therefore foreign language teaching policies that deny
explicit focus on form to academically oriented adults, who can handle such
analytical approach of linguistic structure, should be considered as fundamentally flawed. They deny learners with high analytic ability the use of the only
mechanism at their disposal to master certain basic structures in the L2.
(Received 6 August 1999)
NOTES
1. Felix (1985) also saw age effects as the inevitable result of a shift from language-specific implicit learning mechanisms to explicit learning through general problem-solving skills, but in his view
explicit learning is a competitor that, as soon as it is fully developed, starts interfering with the
implicit mechanisms. This view seems to imply, though, that the stronger a persons problem-solving
skills are, the worse language learning will be; this implication is clearly not borne out by the facts
as discussed later in this paper.
2. Learners in the 1218 age range are too young to immigrate by themselves, whereas their parents are likely to be around 40 years of age already and therefore too old to be likely to immigrate
(note that neither the Johnson and Newport [1989] study nor the present one included any participants whose age of arrival was above 40). Bialystok and Hakuta (1999) reported a continuous decline
throughout the lifespan in a study with thousands of subjects, presumably including a fair number
in the 1218 range, but the outcome measure was a global self-rating, which is likely to have been
influenced by the popular conception that one learns language skills less well as one gets older.
3. Oyama (1978) found no difference in aural comprehension at various levels of noise interference between Italian learners of ESL who had been in the United States for 511 years and those
whose length of residence was 1218 years.
4. It is well known that grammaticality judgments do not constitute a perfectly valid measure of
linguistic competence (for thorough discussions see, e.g., Cowart, 1997, and Schutze, 1996). One of
the most frequently made objections is the role that metalinguistic competence plays in grammaticality judgments. The importance of this (generally valid) concern is mitigated in this study because
of the simple nature of both the stimuli (in terms of vocabulary and syntax) and the format (yes-no).
Furthermore, the lack of significant correlation between performance on the grammaticality judgment test and the language-learning aptitude test (see the following section) for the prepuberty arrivals seems to confirm that metalinguistic competence at the time of taking the test was not a major
causal factor in the grammaticality judgment scores of the participants in this study.
5. The low scores on the aptitude test may have been due in part to the fact that the participants

Critical Period

521

did not use Hungarian as extensively as monolinguals. Otto (1996b) did not give average scores for
his monolingual sample, but given that the mean level of item difficulty was .52, the average scores
must have been around 10. Given the low average aptitude scores in this study, they are to be
treated as relative rather than absolute values.
6. Figure 3 in Johnson and Newport (1989, p. 87) is a better basis for comparison with the correlations in the present study than the correlation coefficients they list on p. 88, because these coefficients obscure whether a lower correlation is due to older learners doing well or to younger arrivals
not doing as well as one might expect. As the proportion of young learners in my study is much
smaller than in Johnson and Newports, the difference between the two oldest (>16) and four youngest groups (<16) in their Figure 3 is a better point of comparison for the correlation coefficients in
the present study, which largely reflect the difference between those who arrived after age 16 and
those who arrived before.
7. Johnson and Newport (1989, p. 87, fn. 8) interpreted the items with missing do-support as a
subcategory of yes-no question problems that reflects word-order errors. In fact, in these sentences,
the word order always shows the correct inversion pattern in the sense that the tensed verb comes
before the subject; the error is one of lack of do-support (which entails the erroneous tensing of the
lexical verb). It is true, however, that the order of the subject and the lexical verb is incorrect, and
learners may, therefore, perceive the items as evidencing an incorrect word order. The fact that the
items in question fall into the easy group along with other word-order items, and not into the difficult
group with other items in which the auxiliary is lacking (WHNA) or not inverted (WHNI), lends support to Johnson and Newports point of view that the learners experience these yes-no questions
without do-support as word order errors rather than as auxiliary errors.
REFERENCES
Asher, J. J., & Garca, R. (1969). The optimal age to learn a foreign language. The Modern Language
Journal, 53, 334341.
Bardovi-Harlig, K. (1987). Markedness and salience in second-language acquisition. Language Learning, 37, 385407.
Bialystok, E. (1997). The structure of age: In search of barriers to second language acquisition. Second Language Research, 13, 116137.
Bialystok, E., & Hakuta, K. (1994). In other words: The science and psychology of second-language acquisition. New York: BasicBooks.
Bialystok, E., & Hakuta, K. (1999). Confounded age: Linguistic and cognitive factors in age differences
for second language acquisition. In D. Birdsong (Ed.), Second language acquisition and the Critical
Period Hypothesis (pp. 161181). Mahwah, NJ: Erlbaum.
Birdsong, D. (1992). Ultimate attainment in second language acquisition. Language, 68, 706755.
Birdsong, D. (1994). Decision making in second language acquisition. Studies in Second Language Acquisition, 16, 169182.
Birdsong, D. (1999). Introduction: Whys and why nots of the Critical Period Hypothesis for second
language acquisition. In D. Birdsong (Ed.), Second language acquisition and the Critical Period
Hypothesis (pp. 122). Mahwah, NJ: Erlbaum.
Bley-Vroman, R. (1988). The fundamental character of foreign language learning. In W. Rutherford &
M. Sharwood Smith (Eds.), Grammar and second language teaching: A book of readings (pp. 19
30). Rowley, MA: Newbury House.
Bongaerts, T. (1999). Ultimate attainment in L2 pronunciation: The case of very advanced L2 learners. In D. Birdsong (Ed.), Second language acquisition and the Critical Period Hypothesis (pp. 133
159). Mahwah, NJ: Erlbaum.
Bongaerts, T., Planken, B., & Schils, E. (1995). Can late starters attain a native accent in a foreign
language? In D. Singleton & Z. Lengyel (Eds.), The age factor in second language acquisition (pp.
3050). Clevedon, UK: Multilingual Matters.
Bongaerts, T., van Summeren, C., Planken, B., & Schils, E. (1997). Age and ultimate attainment in the
pronunciation of a foreign language. Studies in Second Language Acquisition, 19, 447465.
Carroll, J. B. (1981). Twenty-five years of research on foreign language aptitude. In K. C. Diller (Ed.),
Individual differences and universals in language learning aptitude (pp. 83118). Rowley, MA: Newbury House.
Carroll, J. B. (1990). Cognitive abilities in foreign language aptitude: Then and now. In T. S. Parry &
C. W. Stansfield (Eds.), Language aptitude reconsidered (pp. 1129). Washington, DC: Center for
Applied Linguistics.

522

Robert M. DeKeyser

Carroll, J. B., & Sapon, S. (1959). Modern Language Aptitude Test: Form A. New York: The Psychological
Corporation.
Clark, H. (1973). The language as fixed effect fallacy: A critique of language statistics in psychological
research. Journal of Verbal Learning and Verbal Behavior, 12, 335359.
Coppieters, R. (1987). Competence differences between native and near-native speakers. Language,
63, 544573.
Corno, L., & Snow, R. E. (1986). Adapting teaching to individual differences among learners. In M. C.
Wittrock (Ed.), Handbook of research on teaching (pp. 605629). New York: Macmillan.
Cowart, W. (1997). Experimental syntax: Applying objective methods to sentence judgments. Thousand
Oaks, CA: Sage.
Cronbach, L. J., & Snow, R. E. (1981). Aptitudes and instructional methods: A handbook for research on
interactions. New York: Irvington.
Deacon, T. W. (1997). The symbolic species: The co-evolution of language and the brain. New York:
Norton.
Doughty, C. (1991). Second language instruction does make a difference. Evidence from an empirical
study of SL relativization. Studies in Second Language Acquisition, 13, 431469.
Eckman, F. R., Moravcsik, E. A., & Wirth, J. R. (1989). Implicational universals and interrogative structures in the interlanguage of ESL learners. Language Learning, 39, 173205.
Ekstrand, L. H. (1976). Age and length of residence as variables related to the adjustments of migrant
children, with special reference to second language learning. In G. Nickel (Ed.), Proceedings of
the Fourth International Congress of Applied Linguistics (pp. 179197). Stuttgart, Germany: Hochschul.
Elman, J., Bates, E., Johnson, M., Karmiloff-Smith, A., Parisi, D., & Plunkett, K. (1996). Rethinking innateness: A connectionist perspective on development. Cambridge, MA: MIT Press.
Eubank, L., & Gregg, K. (1999). Critical periods and (second) language acquisition: Divide et impera.
In D. Birdsong (Ed.), Second language acquisition and the Critical Period Hypothesis (pp. 6599).
Mahwah, NJ: Erlbaum.
Felix, S. (1985). More evidence on competing cognitive systems. Second Language Research, 1, 4772.
Flege, J. E. (1992). Speech learning in a second language. In C. A. Ferguson & C. Stoel-Gammon (Eds.),
Phonological development: Models, research, implications (pp. 565604). Timonium, MD: York
Press.
Flege, J. E. (1999). Age of learning and second language speech. In D. Birdsong (Ed.), Second language
acquisition and the Critical Period Hypothesis (pp. 101131). Mahwah, NJ: Erlbaum.
Flege, J., Takagi, N., & Mann, V. (1995). Japanese adults can learn to produce English /r/ and /l/ accurately. Language and Speech, 38, 2555.
Flynn, S., & Manuel, S. (1991). Age-dependent effects in language acquisition: An evaluation of critical period hypotheses. In L. Eubank (Ed.), Point counterpoint: Universal Grammar in the second
language (pp. 117145). Amsterdam: Benjamins.
Goldowsky, B. N., & Newport, E. L. (1993). Modeling the effects of processing limitations on the acquisition of morphology: The less is more hypothesis. In E. Clark (Ed.), The proceedings of the 24th
annual Child Language Research Forum (pp. 124138). Stanford, CA: CSLI Publications.
Greenberg, J. H. (1963). Some universals of grammar with particular reference to the order of meaningful elements. In J. H. Greenberg (Ed.), Universals of language (pp. 5890). Cambridge, MA: MIT
Press.
Harley, B. (1986). Age in second language acquisition. Clevedon, UK: Multilingual Matters.
Harley, B., & Hart, D. (1997). Language aptitude and second language proficiency in classroom learners of different starting ages. Studies in Second Language Acquisition, 19, 379400.
Harley, B., & Wang, W. (1997). The Critical Period Hypothesis: Where are we now? In A. M. B. de
Groot & J. F. Kroll (Eds.), Tutorials in bilingualism: Psycholinguistic perspectives (pp. 1951). Mahwah, NJ: Erlbaum.
Hyltenstam, K. (1992). Non-native features of near-native speakers: On the ultimate attainment of
childhood L2 learners. In R. J. Harris (Ed.), Cognitive processing in bilinguals (pp. 351368). Amsterdam: North-Holland.
Ioup, G., Boustagui, E., El Tigi, M., & Moselle, M. (1994). Reexamining the Critical Period Hypothesis:
A case study of successful adult SLA in a naturalistic environment. Studies in Second Language
Acquisition, 16, 7398.
Johnson, J. S. (1992). Critical period effects in second language acquisition: The effect of written
versus auditory materials on the assessment of grammatical competence. Language Learning,
42, 217248.
Johnson, J. S., & Newport, E. L. (1989). Critical period effects in second language learning: The influ-

Critical Period

523

ence of maturational state on the acquisition of English as a second language. Cognitive Psychology, 21, 6099.
Johnson, J. S., & Newport, E. L. (1991). Critical period effects on universal properties of language:
The status of subjacency in the acquisition of a second language. Cognition, 39, 215258.
Juffs, A., & Harrington, M. (1995). Parsing effects in second language sentence processing: Subject
and object asymmetries in wh-extraction. Studies in Second Language Acquisition, 17, 483516.
Kareev, Y., Lieberman, I., & Lev, M. (1997). Through a narrow window: Sample size and the perception of correlation. Journal of Experimental Psychology: General, 126, 278287.
Karmiloff-Smith, A. (1992). Beyond modularity: Developmental perspective on cognitive science. Cambridge, MA: MIT Press.
Kellerman, E. (1995). Age before beauty. In L. Eubank, L. Selinker, & M. Sharwood Smith (Eds.), The
current state of interlanguage: Studies in honor of William E. Rutherford (pp. 219231). Amsterdam:
Benjamins.
Kenesei, I., Vago, R. M., & Fenyvesi, A. (1998). Hungarian. New York: Routledge.
Krashen, S. D. (1981). Aptitude and attitude in relation to second language acquisition and learning.
In K. C. Diller (Ed.), Individual differences and universals in language learning aptitude (pp. 155
175). Rowley, MA: Newbury House.
Krashen, S. D., Long, M., & Scarcella, R. (1979). Age, rate, and eventual attainment in second language
acquisition. TESOL Quarterly, 13, 573582.
Lee, D. (1992). Universal Grammar, learnability, and the acquisition of English reflexive binding by L1
Korean speakers. Unpublished doctoral dissertation, University of Southern California, Los
Angeles.
Lenneberg, E. H. (1967). The biological foundations of language. New York: Wiley.
Long, M. (1990). Maturational constraints on language development. Studies in Second Language Acquisition, 12, 251285.
Long, M. (1993). Second language acquisition as a function of age: Research findings and methodological issues. In K. Hyltenstam & A. Viberg (Eds.), Progression and regression in language (pp. 196
221). New York: Cambridge University Press.
McLaughlin, B. (1990). The relationship between first and second languages: Language proficiency
and language aptitude. In B. Harley, P. Allen, J. Cummins, & M. Swain (Eds.), The development of
second language proficiency (pp. 158174). New York: Cambridge University Press.
Moyer, A. (1999). Ultimate attainment in L2 phonology: The critical factors of age, motivation, and
instruction. Studies in Second Language Acquisition, 21, 81108.
Munro, M. J., Flege, J. E., & MacKay, I. R. A. (1996). The effects of age of second language learning on
the production of English vowels. Applied Psycholinguistics, 17, 313334.
Neufeld, G. (1978). On the acquisition of prosodic and articulatory features in adult language learning. Canadian Modern Language Review, 34, 163174.
Newport, E. (1990). Maturational constraints on language learning. Cognitive Science, 14, 1128.
Otto, I. (1996a). The development of a language aptitude test for Hungarian learners of foreign languages. Unpublished masters thesis, Eotvos Lorand University, Budapest.
Otto, I. (1996b). Hungarian language aptitude test: Words in sentences. Unpublished manuscript, Department of English Applied Linguistics, Eotvos Lorand University, Budapest.
Oyama, S. (1976). A sensitive period for the acquisition of a nonnative phonological system. Journal
of Psycholinguistic Research, 5, 261283.
Oyama, S. (1978). The sensitive period and comprehension of speech. Working Papers on Bilingualism, 16, 117.
Patkowski, M. (1980). The sensitive period for the acquisition of syntax in a second language. Language Learning, 30, 449472.
Patkowski, M. S. (1990). Age and accent in a second language: A reply to James Emil Flege. Applied
Linguistics, 11, 7389.
Patkowski, M. S. (1994). The Critical Age Hypothesis and interlanguage phonology. In M. Yavas (Ed.),
First and second language phonology (pp. 205221). San Diego, CA: Singular.
Perruchet, P., & Pacteau, C. (1991). Implicit acquisition of abstract knowledge about artificial grammar: Some methodological and conceptual issues. Journal of Experimental Psychology: General,
120, 112116.
Pulvermuller, F., & Schumann, J. H. (1994). Neurobiological mechanisms of language acquisition. Language Learning, 44, 681734.
Reber, A., Kassin, S., Lewis, S., & Cantor, G. (1980). On the relationship between implicit and explicit
modes in the learning of a complex rule structure. Journal of Experimental Psychology: Human
Learning and Memory, 6, 492502.

524

Robert M. DeKeyser

Redington, M., & Chater, N. (1996). Transfer in artificial grammar learning. Journal of Experimental
Psychology: General, 125, 123138.
Robinson, P. (1996). Learning simple and complex second language rules under implicit, incidental,
rule-search, and instructed conditions. Studies in Second Language Acquisition, 18, 2767.
Sasaki, M. (1993). Relationships among second language proficiency, foreign language aptitude, and
intelligence: A structural equation modeling approach. Language Learning, 43, 313344.
Schachter, J. (1989). Testing a proposed universal. In S. M. Gass & J. Schachter (Eds.), Linguistic
perspectives on second language acquisition (pp. 7388). New York: Cambridge University Press.
Schachter, J. (1990). On the issue of completeness in second language acquisition. Second Language
Research, 6, 93124.
Schachter, J. (1996). Maturation and universal grammar. In W. C. Ritchie & T. K. Bhatia (Eds.), Handbook of second language acquisition (pp. 159193). San Diego, CA: Academic Press.
Schutze, C. (1996). The empirical base of linguistics. Chicago: University of Chicago Press.
Scovel, T. (1988). A time to speak: A psycholinguistic inquiry into the critical period for human speech.
Rowley, MA: Newbury House.
Seliger, H. W. (1978). Implications of a multiple critical periods hypothesis for second language learning. In W. C. Ritchie (Ed.), Second language acquisition research: Issues and implications (pp. 11
19). New York: Academic Press.
Shanks, D. R., & St. John, M. F. (1994). Characteristics of dissociable human learning systems. Behavioral and Brain Sciences, 17, 367395.
Singleton, D. (1989). Language acquisition: The age factor. Clevedon, UK: Multilingual Matters.
Skehan, P. (1986). The role of foreign language aptitude in a model of school learning. Language Testing, 3, 188221.
Skehan, P. (1989). Individual differences in second language learning. London: Edward Arnold.
Skehan, P. (1998). A cognitive approach to language learning. Oxford: Oxford University Press.
Slavoff, G. R., & Johnson, J. S. (1995). The effects of age on the rate of learning a second language.
Studies in Second Language Acquisition, 17, 116.
Snow, C., & Hoefnagel-Hohle, M. (1978). The critical period for language acquisition: Evidence from
second language learning. Child Development, 49, 11141128.
Snow, R. E. (1989). Aptitude-treatment interaction as a framework for research on individual differences in learning. In P. L. Ackerman, R. J. Sternberg, & R. Glaser (Eds.), Learning and individual
differences: Advances in theory and research (pp. 1359). New York: Freeman.
Snow, R. E., & Lohman, D. (1984). Toward a theory of cognitive aptitude for learning from instruction.
Journal of Educational Psychology, 76, 347376.
Sorace, A. (1993). Incomplete vs. divergent representations of unaccusativity in non-native grammars
of Italian. Second Language Research, 9, 2247.
Tahta, S., Wood, M., & Loewenthal, K. (1981). Foreign accents: Factors relating to transfer from the
first language to a second language. Language and Speech, 24, 265272.
Thurstone, L. L., & Thurstone, C. T. (1962). Primary Mental Abilities Test. Chicago: Science Research
Associates.
Walsh, T., & Diller, K. C. (1981). Neurolinguistic considerations on the optimum age for second language learning. In K. C. Diller (Ed.), Individual differences and universals in language learning aptitude (pp. 321). Rowley, MA: Newbury House.
Wesche, M., Edwards, H., & Wells, W. (1982). Foreign language aptitude and intelligence. Applied Psycholinguistics, 3, 127140.
White, L., & Genesee, F. (1996). How native is near-native? The issue of ultimate attainment in adult
second language acquisition. Second Language Research, 12, 233365.
Wode, H. (1994). Nature, nurture, and age in language acquisition: The case of speech perception.
Studies in Second Language Acquisition, 16, 325345.

Critical Period

525

APPENDIX A
Table A1. Main variables for all participants
Age of
arrival

Age
at test

Grammaticality
judgment test score

Aptitude
test score

Years of
schooling

28
8
11
20
23
22
30
27
27
9
11
13
27
27
23
26
21
23
17
4
1
25
20
34
25
22
33
12
35
29
28
26
3
40
33
37
21
26
26
10
28
25
26
22
20

68
46
51
59
60
56
75
50
64
54
57
58
67
66
63
66
60
62
56
53
40
65
60
70
39
58
73
28
75
69
64
65
49
54
59
60
61
40
51
55
72
35
76
68
30

136
170
181
132
139
153
141
147
175
198
194
196
170
126
110
119
134
118
167
193
199
122
126
143
129
190
143
195
175
153
184
129
197
125
146
111
127
123
151
194
143
176
176
164
152

3
4
5
5
10
8
5
9
9
4
8
4
4
4
3
5
5
6
4
5
4
2
2
4
8
10
7
2
3
0
7
4
4
2
3
3
4
4
3
2
2
8
6
5
3

6
6
12
14
12
14
17
18
17
16
16
20
18
16
8
8
6
8
12
18
18
12
8
16
8
16
14
16
18
17
17
12
16
16
12
10
14
16
16
20
16
12
12
7
13

526

Robert M. DeKeyser
6
21
38
24
26
32
11
5
23
9
27
3

16
70
52
35
37
81
51
21
55
20
73
49

183
162
177
174
186
132
183
196
155
197
155
199

3
4
11
3
3
5
4
4
4
7
6
4

8
12
20
18
20
12
12
12
16
13
9
16

Critical Period

527

APPENDIX B
GRAMMATICALITY JUDGMENT TEST ITEMS GROUPED BY STRUCTURE
Practice Items
1. A snake bit she on the leg.*
2. Susan is making some cookies for us.
3. The baby bird has fall from the oak tree.*
4. The little boy was counting all his pennies last night.
Past Tense
A. Past tense marking omitted in obligatory context (PTMO).

1. JN
JN
2. DK
DK
3. JN
JN

Last night the old lady die in her sleep.*

Last night the old lady died in her sleep.
Sandy fill a jar with cookies last night.*
Sandy filled a jar with cookies last night.
John sing for the church choir yesterday.*
John sang for the church choir yesterday.

1
2
2
1
1
2

B. Irregular verbs regularized (PTIR).

4. JN
JN
5. JN
JN
6. DK
DK

Janie sleeped with her teddy bear last night.*

Janie slept with her teddy bear last night.
Last night the books falled off the shelves.*
Last night the books fell off the shelves.
A policeman gived Alan a ticket for speeding yesterday.*
A policeman gave Alan a ticket for speeding yesterday.

2
1
1
2
2
1

C. Regular ending on irregular stem (PTII).

7. JN
JN
8. DK
DK
9. JN
JN

A bat flewed into our attic last night.*

A bat flew into our attic last night.
Yesterday the teacher sented Allison to the principal.*
Yesterday the teacher sent Allison to the principal.
Mr. Murphy hidded his money under his mattress.*
Mr. Murphy hid his money under his mattress.

1
2
2
1
1
2

Plural
A. Plural marking omitted in obligatory context (PLMO).

1. JN
JN
2. JN
JN
3. JN
JN

Three boy played on the swings in the park.*

Three boys played on the swings in the park.
Many house were destroyed by the flood last week.*
Many houses were destroyed by the flood last week.
The farmer bought two pig at the market.*
The farmer bought two pigs at the market.

2
1
1
2
2
1

B. Irregular plurals regularized (PLIR).

4. JN
JN
5. JN
JN
6. JN
JN

A shoe salesman sees many foots throughout the day.*

A shoe salesman sees many feet throughout the day.
Two mouses ran into the house this morning.*
Two mice ran into the house this morning.
The boy lost two teeths in the fight.*
The boy lost two teeth in the fight.

1
2
2
1
1
2

528

Robert M. DeKeyser

C. Mass nouns used with plural marker (PLMN).

7. DK
DK
8. DK
DK
9. DK
DK

I need to get some informations about the train schedule.*

I need to get some information about the train schedule.
Our neighbor bought new furnitures last week.*
Our neighbor bought new furniture last week.
Teachers often give useful advices to their students.*
Teachers often give useful advice to their students.

2
1
1
2
2
1

Third-Person Singular
A. Third-person -s omitted in obligatory context (TPSO).

1. JN
JN
2. JN
JN
3. JN
JN

Johns dog always wait for him at the corner.*

Johns dog always waits for him at the corner.
Mrs. Sampson clean her house every Wednesday.*
Mrs. Sampson cleans her house every Wednesday.
Every Friday our neighbor wash her car.*
Every Friday our neighbor washes her car.

1
2
2
1
1
2

B. Third-person -s marked on main verb after modals (TPSM).

4. DK
DK
5. DK
DK
6. DK
DK
7. JN
JN
8. JN
JN

John can plays the piano very well.*

John can play the piano very well.
Our new neighbor should turns his radio down a bit.*
Our new neighbor should turn his radio down a bit.
Mary will goes to Europe next year.*
Mary will go to Europe next year.
The Johnsons may are moving to Chicago this fall.*
The Johnsons may be moving to Chicago this fall.
Mrs. Newport will is leaving the party early.*
Mrs. Newport will be leaving the party early.

2
1
1
2
2
1
1
2
2
1

Present Progressive
A. Progressive -ing omitted in obligatory context (PPMO).

1. JN
JN
2. JN
JN
3. JN
JN

The little boy is speak to a policeman.*

The little boy is speaking to a policeman.
Janet is wear the dress I gave her.*
Janet is wearing the dress I gave her.
The boy has been lie to his father.*
The boy has been lying to his father.

1
2
2
1
1
2

B. Progressive auxiliary omitted (PPAO).

4. DK
DK
5. DK
DK
6. DK
DK

Tom working in his office right now.*

Tom is working in his office right now.
The children playing in the garden till dark these days.*
The children are playing in the garden till dark these days.
Bob trying to fix Jims car with his new tools.*
Bob is trying to fix Jims car with his new tools.

2
1
1
2
2
1

Determiners
A. Determiner omitted in obligatory context (DEOM).

1. JN
JN
2. JN
JN

Tom is reading book in the bathtub.*

Tom is reading a book in the bathtub.
Mrs. Johnson went to library yesterday.*
Mrs. Johnson went to the library yesterday.

1
2
2
1

Critical Period
3. JN The boy is helping the man build house.*
JN The boy is helping the man build a house.

529
1
2

B. Determiner used with abstract nouns (DEAB).

4. DK
DK
5. DK
DK
6. DK
DK
7. JN
JN

The beauty is something that lasts forever.*

Beauty is something that lasts forever.
After a life like that he will go straight to the hell.*
After a life like that he will go straight to hell.
The red is a beautiful color.*
Red is a beautiful color.
The men played the basketball in the backyard.*
The men played basketball in the backyard.

2
1
1
2
2
1
1
2

Pronominalization
A. Pronoun omitted in obligatory context
(transfer from Hungarian object pro-drop) (PROM).

1. DK
DK
2. DK
DK
3. DK
DK
4. DK
DK

Peter made out the check but didnt sign.*

Peter made out the check but didnt sign it.
Mary looked at the flowers but didnt buy.*
Mary looked at the flowers but didnt buy them.
John took a sweater along but didnt put on.*
John took a sweater along but didnt put it on.
Mike wrote the letter but didnt send.*
Mike wrote the letter but didnt send it.

2
1
1
2
2
1
1
2

B. Gender errors (PRGE).

5. JN
JN
6. DK
DK
7. DK
DK
8. DK
DK

The girl cut himself on a piece of glass.*

The girl cut herself on a piece of glass.
Peter did not have any money on her.*
Peter did not have any money on him.
Mary fell but he did not break any bones.*
Mary fell but she did not break any bones.
John knew but she did not tell.*
John knew but he did not tell.

2
1
1
2
2
1
1
2

Particle Movement
A. Phrasal verb separation not allowed (PMSE).

1. JN
JN
2. DK
DK
3. DK
DK
4. DK
DK

The man climbed the ladder up carefully.*

The man climbed up the ladder carefully.
The drunk slept his hangover off in the guest room.*
The drunk slept off his hangover in the guest room.
The new neighbors carried a long conversation on.*
The new neighbors carried on a long conversation.
This plastic gives a weird smell off.*
This plastic gives off a weird smell.

2
1
1
2
2
1
1
2

B. Phrasal verb separation allowed, but particle moved too far (PMTF).

5. JN
JN
6. JN
JN
7. JN
JN

Kevin called Nancy for a date up.*

Kevin called Nancy up for a date.
The man looked the new cars yesterday over.*
The man looked the new cars over yesterday.
She broke her shoes very carefully in.*
She broke her shoes in very carefully.

2
1
1
2
2
1

530

Robert M. DeKeyser

8. DK Mary took her coat quickly off.*

DK Mary took her coat off quickly.

1
2

Subcategorization (SUBC)
1. JN
JN
2. JN
JN
3. DK
DK
4. DK
DK
5. JN
JN
6. JN
JN
7. JN
JN
8. JN
JN
9. JN
JN
10. JN
JN

George says much too softly.*

George says his prayers much too softly.
The little boys laughed the clown.*
The little boys laughed at the clown.
John said me that his wife was ill.*
John told me that his wife was ill.
The student was learning in his room until late last night.*
The student was studying in his room until late last night.
I want you will go to the store now.*
I want you to go to the store now.
I hope you to go to the store now.*
I hope you will go to the store now.
The man allows his son watch TV.*
The man allows his son to watch TV.
The man lets his son to watch TV.*
The man lets his son watch TV.
The girls want watching TV.*
The girls want to watch TV.
The girls enjoy to watch TV.*
The girls enjoy watching TV.

2
1
1
2
2
1
1
2
2
1
1
2
2
1
1
2
2
1
1
2

Yes-No Questions
A. *aux Aux s[ . . . (YNAA).

1. JN
JN
2. JN
JN
3. JN
JN

Will be Harry blamed for the accident?*

Will Harry be blamed for the accident?
Has been the King served his dinner?*
Has the King been served his dinner?
Is being the baby held by his mother?*
Is the baby being held by his mother?

2
1
1
2
2
1

B. *aux Verb s[ . . . (YNAV).

4. JN
JN
5. JN
JN
6. JN
JN

Can ride the little girl a bicycle?*

Can the little girl ride a bicycle?
Will wear Harry his new shirt to the party?*
Will Harry wear his new shirt to the party?
Is waiting Sally in the car?*
Is Sally waiting in the car?

1
2
2
1
1
2

C. *V s[ . . . (YNVS).

7. JN
JN
8. JN
JN
9. JN
JN

Knows John the answer to that question?*

Does John know the answer to that question?
Swam Janet in the race yesterday?*
Did Janet swim in the race yesterday?
Danced Bill at the party last night?*
Did Bill dance at the party last night?

2
1
1
2
2
1

D. Double tense marking (YNDT).

10. JN Where did Arnie hunted last year?*

JN Where did Arnie hunt last year?

1
2

Critical Period
11. JN
JN
12. JN
JN

Did Bobbie stayed at home last night?*

Did Bobbie stay at home last night?
Does Martha uses her microwave oven?*
Does Martha use her microwave oven?

531
2
1
1
2

Wh-Questions
A. No aux inversion (WHNI).

1. JN
JN
2. JN
JN
3. JN
JN

What Martha is bringing to the party?*

What is Martha bringing to the party?
Where Ted is working this summer?*
Where is Ted working this summer?
When Sam will fix his car?*
When will Sam fix his car?

2
1
1
2
2
1

B. No aux (WHNA).

4. JN
JN
5. JN
JN
6. JN
JN

Who you meet at the park every day?*

Who do you meet at the park every day?
What they sell at the corner store?*
What do they sell at the corner store?
When they leave for Mexico?*
When do they leave for Mexico?

1
2
2
1
1
2

Word Order
A. S V DO order violated (WODO).

1. JN
JN
2. JN
JN
3. JN
JN

The dinner the man burned.*

The man burned the dinner.
The ball the boy caught.*
The boy caught the ball.
The girl the movie likes.*
The girl likes the movie.

2
1
1
2
2
1

B. S V IO DO order violated (WOIO).

4. JN
JN
5. JN
JN
6. JN
JN

The woman the policeman asked a question.*

The woman asked the policeman a question.
The boy carrots feeds the rabbits.*
The boy feeds the rabbits carrots.
Linda a cake baked John.*
Linda baked John a cake.

1
2
2
1
1
2

C. S V order violated (WOVS).

7. JN
JN
8. JN
JN
9. JN
JN

Bites the dog.*

The dog bites.
Drinks the man.*
The man drinks.
Paints the woman.*
The woman paints.

2
1
1
2
2
1

D. S V PP order violated (WOPP).

10. DK
DK
11. DK
DK

The students to the movies went.*

The students went to the movies.
The children with the dog play.*
The children play with the dog.

1
2
2
1

532

Robert M. DeKeyser

12. DK All our friends in the suburbs live.*

DK All our friends live in the suburbs.

1
2

E. Adverb placement (WOAD).

13. DK
DK
14. DK
DK
15. DK
DK

The student eats quickly his meals.*

The student eats his meals quickly.
Kevin rides usually his bicycle to work.*
Kevin usually rides his bicycle to work.
My neighbor enjoyed slowly his dessert.*
My neighbor slowly enjoyed his dessert.

2
1
1
2
2
1

Critical Period

533

APPENDIX C
ADAPTATIONS MADE TO THE TEST ITEMS IN JOHNSON
AND NEWPORT (1989)
1. Four practice items were added at the beginning of the test.
2. In cases (pertaining to the categories Word Order and Subcategorization) where there were
three items or options rather than two, the number of items was reduced to two (one correct,
one incorrect), in line with all other cases.
3. Three sentences were modified on the advice of Johnson and Newport (J & N).
4. Two items were deleted on J & Ns advice (these items were not counted in their analysis).
5. Two J & N items had a violation of agreement on top of another grammar problem; this agreement error was eliminated.
6. Some items that were correct, but slightly awkward, were modified.
7. Three items labeled as Inappropriate Context in J & N were deleted because of the semantic
rather than morphosyntactic nature of the error (past tense, third-person singular, present
progressive).
8. A few items labeled as Questions with Tense, a subdivision of both the Past Tense and
Third-Person Singular categories, were moved to the category Yes-No Questions.
9. Some nouns used to test the mass-count distinction were changed to items known to lead to
transfer problems in Hungarian speakers.
10. A few items with erroneous -s marking on the main verb following a model were added to the
category Third-Person Singular, including two items that J & N listed in the subdivision Tense
Placement of the Aux Rule category.
11. The Present Progressive category now includes two items from the Be Takes -ing subdivision
of J & Ns Aux Rule category; two simpler items were dropped; items in which the progressive
auxiliary is omitted were added.
12. Part B of the category Determiners was made more homogeneous; all items are now abstract
nouns (transfer of the determiner from Hungarian is expected).
13. In the category Pronominalization, the nominative- versus objective-case items were dropped,
as well as the possessive adjective items; examples of pronoun omission were included
(transfer from Hungarian object pro-drop is expected); reflexives were dropped as a category;
gender errors were added (including one item from J&Ns Reflexives category).
14. In the category Particle Movement, items that concerned prepositional phrases rather than
particles were deleted and replaced by others (some for which separating is okay, some for
which it is not).
15. The category Aux Rule was eliminated (but two items from its subcategory Tense Placement
are now classified under Third-Person Singular).
16. In the category Subcategorization, items 5 and 6 were replaced by other items specifically
chosen to catch transfer from Hungarian.
17. In the category Yes-No Questions, the subcategory Invert + Stay was replaced with items
showing double marking (present and past).
18. In the category Word Order, items were added in which the SVPP order is violated, and (erroneous) SVAO items were added.
19. In many categories, the number of items was reduced in order to limit the test to 200 items
(from mostly four- to mostly three-item pairs per subcategory).