THE CRITICAL PERIOD HYPOTHESIS

The notion that a biologically determined period exists during which language

acquisition must occur, if it is to occur at all, is known as the critical period

hypothesis. Nature provides many examples of biologically determined

deadlines, the best known of which is probably the critical period for imprinting

in birds. Some species of birds walk as soon as they hatch. Chicks

and ducklings, for example, will follow the first moving thing they see, and

then they will follow it forever. Normally, the first thing a baby sees when it

hatches is its mother, with the result that chicks and ducklings then follow

their mothers everywhere. When this occurs, the chicks and ducklings are said

to be imprinted on their mother. Imprinting cannot happen any time; it must

occur within a few hours after hatching. (This requirement is not always absolute,

and the term sensitive period is sometimes substituted for critical period

[Lieberman, 1993].)

There are well-documented human examples of critical periods as well.

For example, some cells in the brain respond to input from both eyes in the

normal adult, but if these cells fail to receive input from two eyes during the

first year or two of life, they lose this capacity. Thus, the features common to

all examples of critical or sensitive periods are these: some environmental input

is necessary for normal development, but biology determines when the

organism is responsive to that input. That period of responsivity is the critical

period. If one were to design a study to test the critical period hypothesis, one

would deprive children of exposure to language during the normal period of

language development, provide the exposure later, and examine the languag~

development that occurs. Of course, such an unethical experiment could

never be done deliberately, but history has provided a few such cases.

..Wild" children

Victor of Aveyron was such a case in which first exposure to language, so far

as we know, came late. As mentioned in Chapter 1, Victor never acquired norrnallanguage.

There are a few other cases of such wild children who also did

not learn language under the circumstance of early deprivation and late first

exposure, but it is difficult to learn much from such cases because they are so

poorly,documented. Furthermore, when such children fail to acquire language,

we cannot be sure whether the failure was due to the late start or to

some impairment the child might have had previously. Reviewing the evidence

in 1967, Lenneberg came to the conclusion that "the only safe conclusions

to be drawn from the multitude of reports is life in dark closets, wolves'

dens, forests, or sadistic parents' backyards is not conducive to good health

and normal development" (p. 142).

There is one success story among such children. In the 1930s, a 6-year-old

child named Isabelle was discovered living hidden away in a dark room with

only her deaf-mute mother for contact. After her discovery, Isabelle was trained

intensively to speak, and she did learn to talk. Isabelle's success makes it clear

that she was cognitively normal, but her deprivation was also less extreme than

that of the other cases of wild children. Furthermore, we do not have the sort

of psycholinguistic details about Isabelle we would like. Although at age 8 she

was described as having "a normal IQ" and "not easily distinguished from ordinary

children of her age" (Brown, 1958a,p. 192), no one administered the tests

of linguistic competence that would allow detailed comparisons of her language

competence with that of children with normal experience.

There is one modern case of a wild child who was discovered after linguistics

and neurology were sufficiently advanced to allow us to ask questions

that were not asked of the earlier cases.

The case of Genie

In 1970 a woman who is known to most of the world only as "Genie's mother"

was lookiIig for the office of services for the blind in downtown Los Angeles.

She was nearly blind, was seekiIig help for herself, and had only recently

managed to escape virtual captivity by her mentally ill husband. By mistake,

she entered the general social services office. She brought with her a 13-yearold

daughter, Genie. The eligibility worker at the social services office noticed

the small, frail-looking child with a strange gait and posture and called her supervisor,

who, after questioning Genie's mother, called the police. The police

took Genie into custody and admitted her into the hospital for severe malnutrition

(Curtiss, 1977; Rymer, 1993).

The story of Genie's background that was eventually revealed was horrific.

From the time Genie was 20 months old until her mother's escape when

she was 13, Genie spent her time alone, strapped to a potty chair in a small

bedroom. She was fed hurriedly, with minimal interaction and no talk. If

Genie made any noise, her father would beat her with a large piece of wood

he kept in the room for that purpose. Like the Wild Boy of Aveyron before

her, Genie had no language when she was discovered. Also like Victor of

Aveyron, Genie was immensely interesting to the scientific community. The

story of Genie's life and treatment both before and after her discovery have

been described by Curtiss (1977) and by Rymer (1993). We shall confine ourselves

here to the investigation of Genie's language development, described

by Susan Curtiss in her dissertation and subsequent papers (Curtiss, 1977,

1985, 1988, 1989).

Genie did not talk at all when she was first discovered. Four years later,

she scored in the range of a normal 5-year-old on standardized vocabulary

tests. She combined words into complex utterances, and she could express

meanings. However, her language was far from normal. As the examples of

Genie's speech in Box 2.4 show, her vocabulary and semantic skills far ex

wash hair in sink.

At school scratch face.

I want Curtiss play piano.

Like go ride yellow school bus.

Father take piece wood. Hit. Cry.

Source' Curtiss, 1977.

ceeded her syntactic skills. Her grammar was deficient in both production and

comprehension. In production, her utterances were telegraphic, lacking most

grammatical morphemes. In comprehension tests, she failed to understand

passive constructions and distinctions marked by tense, and she had other difficulties

as well.

Another fact about Genie's language might be related to her grammatical

limitations. Dichotic listening tests showed that language was a right-hemisphere

activity for Genie. In fact, the nature of her grammatical limitations has been

compared to the grammatical deficiencies of patients who have recovered language

after surgical removal of the left hemisphere. One possible explanation

of this phenomenon is that Genie was exposed to language too late for the

normal process of acquisition of language as a left-hemisphere function. She

acquired language with the right hemisphere, and-as we have seen in the

aphasia data-the right hemisphere is not as good at language as the left.

Genie's conversational competence was also extremely limited, and she often

ignored the speech addressed to her. As Curtiss (1977) described it:

Verbal interaction with Genie consists mainly of someone's asking

Genie a question repeatedly until Genie answers, or of Genie's

making a comment and someone else's responding to it in some way.

...Except for those instances where Genie exerts control over the

topic through repetition, verbal interaction with Genie is almost

always controlled and/or "normalized" by the person talking to

Genie, not by Genie. (p. 233)

Curtiss attributes this conversational incompetence to Genie's lack of

early socializing experience.

The study of Genie is certainly more informative than earlier reports on

isolated children. Evidence that language was a right-hemisphere function for

Genie suggests that by age 13, a left hemisphere that has never been used for

language has lost that capacity. However, interpretation of Genie's outcome is

still hampered by the fact that we do not know with certainty that Genie was

a normal child except for her experiences. Once, when Genie was seen by a

doctor as an infant, she was diagnosed as mentally retarded. But there was

never any follow-up to see whether that pediatrician's impression was correct,

and even before Genie was totally isolated, she had something less than an

ideal environment. Susan Curtiss, who worked most closely with Genie, vehemendy

disagrees with the possibility that Genie could be retarded (Rymer,

1993), but we simply do not know for sure.

Late acQuisition of American Sign Language

A better test of the critical period hypothesis is provided by individuals who

have normal early experience except for being deprived of exposure to language.

This is the circumstance of many children born deaf to hearing parents.

These children have no language input at home because they cannot

hear the language their family speaks, and their parents do not know sign language

(and historically have been discouraged from learning it to communicate

with their children, although this has changed in recent years). Many of

these children are eventually exposed to sign language when they meet other

deaf children, some of whom have deaf parents and have been exposed to

sign language from infancy. Comparing the sign language acquisition of children

who learned sign from infancy to that of children who were first exposed

to it later in childhood or in adulthood provides a very nice test of the critical

period hypothesis. If the young brain is better at language acquisition, deaf individuals

who began to acquire sign as older children should be less proficient

than those who acquired it in infancy.

Newport (1990) studied the sign language proficiency of deaf adults who

ranged in age from 35 to 70, who used American Sign Language (ASL) in their

everyday communication, and who had done so for more than 30 years. Some

of these adults had acquired ASL as infants from their deaf parents. Some had

first been exposed to ASL when they entered a school for the deaf between

the age of 4 and 6; some had first been exposed only after the age of 12, when

they entered the school as teenagers, or later, when they made friends with

or married someone from that school. Newport administered a battery of comprehension

and production tests to assess how well these deaf adult.5 had

mastered the grammar of ASL. She found that adults who were first exposed

to ASL after early childhood-even after 30 years of using the language every

day-<iid not perform as well as those who had been exposed as infants. Similarly,

Mayberry and Eichen (1991) found that early learners of sign had an advantage

over late learners in recalling and reproducing ASL sentences that

were presented to them. This evidence suggests there is some benefit to being

a young language learner.

,;1

The evidence from second language acQuisition

A far mor~ frequently occurring test of the effect of age on language acquisition

occurs in the realm of second language acquisition, and to the casual observer,

the results are obvious. Children learn a new language readily, and

soon after moving to a new lan~age community they are indistinguishable

from their native-born peers. Adults, in contrast, master a new language only

with difficulty and never quite sound like native speakers. The gist of these

everyday observations has been frequently supported by experimental tests:

the younger one is when exposed to a second language, the better one's ultimate

proficiency in that language. That fact does not necessarily support the

critical period hypothesis, however. The hypothesis that biology sets a cutoff

age after which language cannot be learned to native-like competence predicts

an abrupt age-related change in the success of language learning that

cannot be explained by other factors. As we shall see, the details of the findings

regarding age and second language learning do not consistently support

that strong hypothesis.

Age effects on second language acquisition. Among adults who have

emigrated to the United States from a non-English speaking country, their age

of arrival in the United States predicts the degree to which they will have a

foreign accent and their performance on grammatical judgment tasks. In both

cases, the younger the immigrants were when they were first exposed to English,

the more like native speakers they sound and perform.

In a study by Oyarna (1976), the English speech of60 Italian immigrants

was tape-recorded, and two judges scored those records on a five-point scale

ranging from no foreign accent to heavy foreign accent. Oyama then analyzed

the influence of two variables: (1) the age of the immigrants at arrival in the

United States and (2) the number of years living in the United States. Oyama

found a strong effect of age at arrival, with young arrivals showing less

accent than older ones. The number of years had no effect. Oyama (1978)

similarly found strong age-of-arrival effects on second language users' ability

to repeat English sentences presented to them under noisy conditions (i.e.,

static on the tape).

Other studies have similarly found that age of arrival affects the ability to

speak 'a second language without an accent and that native-like performance

depends on exposure beginning in early childhood (Flege, 1987; Flege &

Fletcher, 1992). Sometimes the benefit of youth to acquiring unaccented speech

in a second language is explained as the effect of age on acquiring a motor

skill. Speech production involves moving the lips, tongue, and mouth in ways

particular to each language, and that may be what is difficult for a late learner.

However, there is more to knowing the sound system of a language than just

the motor skill, and Oyama's (1978) finding of effects of age of arrival on com

suggests that not just the mouth, but also the brairi, is involved iri

explaining age effects on the mastery of second language phonology .

A similar effect has been observed for measures of grammatical competence.

Johnson and Newport (1989) presented grammatical and ungrammatical

English sentences to Chinese and Korean natives who were living in the

United States and who had learned English as a second language. As a group,

they did less well on identifying the ungrammatical utterances than a comparison

group of native English speakers. The interesting results with respect

to the critical period hypothesis come from an analysis of subgroups of the

second language speakers. Those who were between 3 and 7 years old when

they arrived in the United States were not different from the native speakers

of English. Those who were between 8 and 15 at airival performed less well

on the test than native speakers, but the younger they were at airival, the

more nearly they approx.imated native competence. Those who were 17 or

older performed least well and did no better than those who were 30. As was

the case for accent, differences were observed between those exposed as

young children, under the age of 7, and those exposed as older, but still prepubescent,

children. Additional evidence of age-of .airival effects comes from

Coppieters (1987), who found that native speakers outperformed adult second

language learners on tests of grammatical competence-even when the second

language learners worked as authors and professors, writing in their second

language.

Continuity or discontinuity in age-of-arrival effects. Although it is clear

that the age at which one starts to learn a language is related to one's ultimate

level of proficiency in that language, it is not clear that the cause of this relation

is a change in the brain's specific capacity for language acquisition. The

hypothesis that there is a biologically defined window during which language

acquisition must occur predicts a discontinuity in the relation of age to ultimate

proficiency. Proficiency should be good for those exposed during the

critical period and less good for those exposed outside of the critical period.

Some evidence suggests that, to the contrary, the function that relates age of

first exposure to ultimate proficiency is a smooth one.

One study that found a continuous function mapping age of arrival to language

competence used census data collected from 2 million immigrants, all of

whom had been living in the United States for at least 10 years (Hakuta, Bialstok,

& Wiley, 2003). The res~ndents reported how proficient they considered

themselves to be in English, their age when they immigrated to the United

States, and the level of education they had attained. The findings are presented

in Figure 2.6 for native speakers of Chinese and Spanish. There is clear evidence

that age of immigration makes a difference, but there is no clear

evidence of a discontinuity .The observed effect of education level also argues

against a biological account of proficiency differences. Although one potential

criticism of these data is that the self-report measure was not very sensitive, it

Native Chinese speakers

>-4

u

" ' .:= 3

" ' 0

." ',::: College t..

~'. ---,

"'~ --:::-:::::~HSgra d .J":': 2

."..'::: --

: <8 yrs ed bQI)

<5 yrs ~d ~

>-.4

u

~0;

.0

~ 3

8

p.

.!:

.~ 2

~

1

0 20 40 60 0

Age of immigration

Figure 2.6 The relation between age of immigration and English proficiency

for native Chinese and Spanish speakers who emigrated to the United States

Source. From "Critical Evidence. A Test of the Critical-Period Hypothesis for Second-Language

Acquisition." Psychological Science, 2003, 14, 31-38. Copyright 2003 American Psychological

Association.

20 40 60

Age of immigration

was sufficiently sensitive to reveal the age of immigration effect. More seriously,

it has been suggested that Hakuta et al.'s (2003) data analytic techniques

might have obscured evidence of a discontinuity in the age effect (Stevens,

2004). On the other hand, other studies have also found that even after puberty

, younger is better with respect to second language acquisition and that

some adults (approximately 5 to 25 percent, depending on the study) can

achieve native-like proficiency in a second language. These other findings similarly

suggest that a simple "puberty closes the window of opportunity" account

of age effects on second language acquisition does not fully explain the data.

Nonbiological influences on second language acQUisition

Another argument against the critical period hypothesis is based on evidence

that factors other than biological age contribute to the age-related decline in second

language learning achievement. Typically, the input conditions for children

and adults are different. Children attend school in the new language, whereas

adults must do work that their limited language skills allow, thus limiting their

exposure to the new language. Jia and Aaronson studied Chinese immigrants to

the United States who were between 5 and 16 years old for their first year in

their new country (Jia & Aaronson, 2003). They found that the children under

age 2 were exposed to a richer English-language environment than the older

children. These very young children watched more English-language television,

were exposed to more English-language books, and had more English-speaking

friends. The older children had richer Chinese language environments. Although

all the children reported they were more comfortable using Chinese when they

first arrived, by the end of the first year the younger children, but not the older,

reported being more comfortable using English. The younger children also

scored higher than the older children on tests of English profidency .If young

children, but not older children or adults, switch their dominant language when

inimigrating, then the studies that look for effects of age on language-learning

ability by comparing younger and older arrivals have a problem.

To the extent that comparison of the language proficiency of young to adult

learners of a second language is a comparison of language proficiency in speakers

for whom that language is dominant to speakers for whom that language is

not, the validity of that research is compromised. It may be that switching the

dominant language is what causes the differences, not age per se. To investigate

this dominant language switch hypothesis, Jia and Aaronson (2003) gave

grammaticality judgment tasks in English and Chinese to native Chinese who

had come to the United States between the ages of 1 and 38. Like Johnson and

Newport, they found that younger arrivals performed better on the test of

English ~. They also found that the younger arrivals performed worse on

the test of Chinese grammar. In fact, scores on the two tests were negatively

related-.,-the better a subject did on the English test, the worse he or she did on

the Chinese test. This finding supports the hypothesis that better second language

acquisition by younger children occurs at least in part because they switch

to the second language as their dominant language.

Further support for the contention that the language-learning advantage

of young children is not purely a reflection of a greater language-learning ability

comes from a study of the acquisition of Dutch as a second language by

English speakers who moved to the Netherlands (Snow & Hoefnagle-Hohle,

1978). The subjects ranged in age from 3 years to adult. Snow and Hoefnagel-

Hohle tested their participants' mastery of Dutch using a variety of measures

of pronunciation, vocabulary, grammar, and text comprehension, first 6

months after their arrival and then two more times at 4- to 5-month intervals.

Contrary to the prediction of the critical period hypothesis, they found that the

youngest children scored the lowest on every test and that the 12- to 15-yearolds

showed the most rapid acquisition.

Also contrary to a biologically based account of the advantage of youth

in second language acquisition, there is evidence that social psychological

variables playa role in second language acquisition. Johnson and Newport

found two such variables that affected the performance in English of their Korean

and Chinese participants: self -consciousness and American identification.

T'ne participaflts who were less self -conscious about making errors and those

who identified themselves as Ainerican showed greater mastery of English.

Both the characteristics of not being self -conscious and of identifying with the

new country are more likely to be true of children than adults, and thus these

noribiological factors may also contribute to the observed age difference in

second language acquisition.

The data on second language acquisition suggest that if you give younger

children and older children the same experience with a second language, the

b I

older children will learn more rapidly. However, in most immigrant situations,

it is the younger children who actually learn more and ultimately achieve

higher level~ of proficiency. Some of the disadvantage that older learners have

may be the result of general cognitive declines that come with age, including

declines in memory , attention, and speed of processing. Some of the disadvantage

of older learners is likely due to other, noncognitive factors including

the conditions of input, their initial high level of proficiency in their native language,

and their reluctance to switch cultural identities and language.

In sum, although it is widely believed that language acquisition is "a

mysterious skill that seems to shut off automatically around the age of 12"

(Osborne, 2003, p. 40), this view is simply not supported by the evidence. The

evidence from Genie, from studies of sign language acquisition, and from recovery

from aphasia show that children have an advantage over adults. However,

that general conclusion leaves open many questions about the basis and

the time of any critical period. Lenneberg (1967), who originally hypothesized

the existence of a biologically based critical period for language acquisition,

proposed puberty as the deadline. According to Bates (1993), however, the capacity

for recovery from aphasia begins to decline after age 5. Newport's 1990

data on the acquisition of American Sign Language show a decline in ultimate

achievement between those first exposed at 6 and those first exposed at 12,

and another decline from 12 to adulthood. None of these sources of data supports

the notion of puberty as a deadline or of an abrupt deadline at any age.

The latgest source of data on the critical period hypothesis, which is also

the source of most people's belief in it, is the observation of age-related differences

in second language acquisition. What these data appear to show,

however, is a continuous decline in ultimate language proficiency related to

the age of first exposure-not an abrupt shutting-off of ability. Furthermore,

some of the decline in language achievement may have nothing to do with

language-learning ability. Rather, other sources of influence including conditions

of exposure and willingness to identify with the new language group

give children an advantage over adults. To the extent that the decline does reflect

a decline in ability, it is not necessarily a language-specific ability. All

sorts of general cognitive processes decline with age (Hakuta et al. , 2003). Additionally,

it has been proposed that one way in which children's cognitive

abilities are limited relative to adults' may provide a language-learning advantage'.

Newport (1991) has argued that it is easier to figure out the structure

of language if you analyze small chunks than if you analyze longer stretches

of speech. Small chunks are all that children can extract from input and store

in memory. Adults, in contrast, extract and store larger chunks, thereby giving

themselves a more-difficult analytical task. Newport refers to this argument as

the "less is more" hypothesis. This hypothesis has also received support

from a computer simulation of language acquisition done within the connectionist

approach, which found that the computer was more successful if fed

~(J

shorter sequences as input (Elman, 1993, 2001), although that finding has

been disputed (Rohde & Plaut, 1999).

One possible reconciliation of the data with respect to the critical or, more

likely, sensitive period hypothesis is that there is a biological preparedness for

language acquisition that is maximal in early childhood. Language acquisition

also depends on other factors, including general cognitive processes and motivation.

The benefit of these other factors typically decline more gradually than

the biological advantage and are more variable after puberty. Thus, early childhood

is the easiest time to acquire language, but it is doable after early childhood

if other factors are maximally supportive.

THE GENETIC BASIS OF LANGUAGE DEVELOPMENT

The genetic basis of language universals

All normal children in anything remotely like a normal environment learn to

talk. Furthermore, the course of Janguage development is, in broad outline,

constant across varying environments. The argument has been made that this

invariance and robustness of language development reflects a genetic plan at

work. Language development is to a large degree a maturational process according

to this view-its course and timing determined by the unfolding of a

genetic blueprint CGilger, 1996; Gleitman, 1981).

On the other hand, it has been argued that the universal acquisition of

language is the result of universal features of human environments. Although

the circumstances under which children acquire language vary

widely, all environments appear to provide two forms of support for the

process of language acquisition. All environments show children that language

is used to communicate with other people, and all environments provide

children with samples of speech in a manner that allows children to figure

out the relations between the sound and meaning CCrago, Allen, &

Hough-Eyamie, 1997; Lieven, 1994). This is not to say that language acquisition

does not also depend on biological properties of humans. Rather, the

biologically provided language-acquisition device depends on finding a certain

sort of environment-just as the biologically provided program for

physical growth depends on nutrition.

Thehetitabillty of Individual djfferentes

Just as the fact that all humans acquire language has been attributed to our

shared genetic blueprint, the individual differences among children in their rate

of language development have been attributed, to some degree, to genetic differences.

The field that studies the genetic basis of individual differences in behavioral

characteristics is behavior genetics, and the tools of behavior genetics