Critical Period Hypothesis


The only language men ever speak perfectly is the one they learn in babyhood,
 when no one can teach them anything!

—Maria Montessori, Italian educationist, The Absorbent Mind (1949:ch. 1)

Children are generally taken to acquire languages rapidly and seemingly with little effort, where adults are believed to be doomed to failure, their task frequently considered challenging and frustrating: “[t]he lack of general guaranteed success is the most striking characteristic of adult foreign language learning” (Bley-Vroman 1989:43). The Critical Period Hypothesis, or its idea, was first introduced by Penfield and Roberts, who argued that language acquisition is most efficient before “the human brain becomes … stiff and rigid” (1959:236). It posits the existence of an age during which genetically endowed language acquisition can take place spontaneously, relatively effortlessly, and characteristically meeting a high degree of success, but at—or before—the onset of puberty (12-14 years of age)[1] acquiring a language naturally, automatically and with complete success becomes impossible (cf. e.g. Coppieters 1987; Johnson & Newport 1989; Patkowsky 1990). “The earlier the better” rule of thumb captures the negative correlation between the age of LA onset and subsequent asymptotic attainment. Most evidence to support the claim was supplied by Lenneberg (1967), who suggested the existence of an ‘age of resonance’, a period during which the brain retains heightened plasticity and LA can take place as a genetic heritage, but that this innate propensity (LAD or, as he put it, latent language structure) atrophies once maturation sets in[2], when this plasticity begins to disappear as a consequence of the firm lateralisation of the language function in the left hemisphere (LH) of the brain (associated with analytic and linguistic skills and creative language use for most people). Analysing the results of aphasia studies Lenneberg found that injuries to the right hemisphere of the brain (RH, associated with non-linguistic and holistic skills) caused more language problems in children than in adults. He also provided evidence that in the cases of children who had undergone surgery of the LH no speech disorders resulted, the brain compensated for the damage and they rapidly recovered total language control after such operations, whereas with adults almost total language loss occurred and they continued to display permanent linguistic impairment. This suggested different neurological bases of language in the two groups.

Apart from research on localisation, neurobiology has also focused on the process of myelination – the covering of neural axons with myelin, a process which occurs after birth and allows for more efficient transport of neural impulses (Jacobs 1988). The slowing of this process “results in reduced neural plasticity and, consequently, in difficulty in learning” (Pulvermüller & Schumann 1994:719).

Thus the maturational process in which the brain seems to almost totally lose its plasticity apparently coincides with the withering away of our access to UG—or its maturation to the point of rigidity—both for L1A and SLA (cf. e.g. Johnson & Newport 1989; Long 1990; Birdsong & Molis 2001; this has been termed the No Access Hypothesis[3]): “just as the milk teeth drop out, so UG becomes defunct” (Cook 1988). While a child can acquire additional languages in a way parallel to his/her mother tongue—accumulating new structures via verbal communication through “cooperation” alone (Vygotsky 1934a; A. N. Leontiev 2005:18)—and become a truly multilingual speaker[4], a post-Critical-Period learner no longer has the same capacity to acquire an L2 naturally and automatically. This has been borne out by a body of research (e.g. Swain & Lapkin 1989; Turnbull et al. 1998) indicating that adults who have found themselves in an immersion situation (immigrants, sojourners) never acquire full linguistic competence without the benefit of additional earlier or ongoing persistent methodical learning (cf. e.g. Strozer 1994:187); moreover, some fail dismally without developing even rudimentary grammatical competence (e.g. failing to apply inflectional affixation). Mechelli et al. (2004) found that overall operationality on L2 reading, writing, speech comprehension and production correlates negatively with age of acquisition (p<.01, r=-0.855), with L1 proficiency correlated with grey-matter density in the left inferior parietal cortex of the brain (Z=4.1; p<.05),[5] which in turn stayed in inverse proportuion to the age of SLA (Z=3.2; p<.05).

However, Lenneberg’s supposition that language acquisition is easier before puberty was only partially correct. Further neurofunctional research found that the adult’s brain never loses the plasticity of the new-born baby’s brain, with the result that in cases of damage to specific areas of neural tissue (as in aphasia), the associated functions need not be irreversibly lost, but may be transferred to other areas. Other research found permanent irreversible linguistic deficits after left hemispherectomy in children and adults alike. In some studies, postmaturational effects were found (Birdsong 1992; Bialystok & Hakuta 1994, 1999; Flege 1999), while in others significant numbers of late learners performing on various linguistic tasks in the range of native controls (Cranshaw 1997; Bongaerts 1999; Van Boxtel 2005). Only when pronunciation is concerned is an early commencement an advantage (e.g. Lopata 1963), and even then only in terms of success, not rate of acquisition (children do not acquire phonetic skills as rapidly as older learners, although they may be more successful in detecting sound differences, especially in the case of tonal languages. Cummins (1981a), Fullana and Muñoz (2003), Leather (2003), and García Lecumberri and Gallardo (2003) found that—in a formal context—early exposure does not correlate to a ‘near-native’ accent). One way to account for such a state of affairs is the differential fossilisation hypothesis developed by Seliger (1978), borne out by neurolinguistic evidence. This asserts that various levels of language processing (syntax and pronunciation, for instance) are linked to distinct neural mechanisms, with the result that various aspects of language fossilise independently (learners with native-like syntax but foreign pronunciation are not uncommon). The process of the lateralisation of macroneural circuits and localisation of language functions is gradual and lengthy, with discrete aspects of language affected at different stages, and only several becoming unacquirable after puberty (phonology doubtless being most noticeable and therefore frequently quoted – cf. the ‘Henry Kissinger effect’ alias ‘Joseph Conrad phenomenon’).[6] Lamendella (1977) chose the term ‘sensitive period’ to emphasise the fact that acquisition may be more efficient during childhood, but not restricted to that period.

While the CPH may hold for acquisition, adults still seem to do at least as well as YLs in learning a foreign language (and, more significantly, they manifest a similar developmental route). For instance, Matias (2006) discovers that the more expert language interpreters are those who have—proficiently—learnt the foreign language after their L1 was well-supported by the surrounding environment, thus endorsing the potential for FLL accomplishment on the one hand, and emphasising the importance of well-developed NL competence on the other.

Important results with respect to the CPH cause célebre could be delivered by a good combination of functional- and structural magnetic resonance imaging (MRI[7]) and psycholinguistic research (van den Noort et al. 2006; but see next subsection). In any case, there seems to be at least one difference between the L1 and further mastered languages, whether ‘acquired’ or ‘learnt’. House (2004) observes that multilinguals frequently declare that under stress, the influence of drugs or alcohol, cognitive and emotional overburdening, or fatigue they experience attrition in their Ln competence much more frequently than in the L1.[8] This apparently happens both during production and reception, e.g. when listening to foreign radio or television broadcasts (when a volume increase is frequently necessary; ibid.).

The CPH re-examined

While the existence of a critical period is generally accepted for L1A, attempts to extend it to SLA still arouse a good deal of contention, and the controversy around the maturational model of L2 attainment remains as intense as ever (Birdsong 1999; Marinova-Todd et al. 2000; Scovel 2000; Hyltenstam & Abrahamsson 2001; Singleton 2001; Wartenburger et al. 2003; Van Boxtel 2005; van den Noort et al. 2006). More recent studies relating age to language acquisition, such as those contained in the volume edited by del Pilar García Mayo and García Lecumberri (2003), as well as close re-examinations of earlier research, e.g. Marinova-Todd et al. (2000), seem to indicate that although older learners are indeed less likely than children to master an L2 at a high level of operationality, this is due to differences in the learning situation rather than their (in)capacity. Researchers such as Gisela Jia argue that the role of age in L2 learning is not so much about brain maturation, but a by-product of other age-related factors: kids are simply quicker to integrate into—and interact with—the new language community, absorbing much more language in the process. At the same time, no constraint was found on the possibility of adults becoming nativelike L2 users, and the widespread belief that adults cannot master FLs is wide of the mark; Marinova-Todd et al. (ibid.) contend that the followers of the CPH were labouring under three major misconceptions:
-          Misinterpretation of observations of child and adult learners relating to speed and efficiency of acquisition. Contrary to popular belief, continually confirmed hard data (e.g. Snow & Hoefnagel-Höhle 1977, 1978; Genesee 1987; Johnson & Newport 1989; Rivera 1998; and summarised in McLaughlin 1984, 1985) focusing on immersion learners demonstrate that children learn new languages slowly and effortfully, and older learners are by and large more rapid and efficient in the initial stages, including phonological acquisition; a fact admitted even by researchers who argued that younger learners tend to eventually achieve greater proficiency (e.g. Krashen et al. 1979). Krashen (1973) also challenged Lenneberg’s claim demonstrating that brain lateralisation may be completed by the age of 5. Other studies (e.g. Oyama 1976; Birdsong 1992; Bialystok & Hakuta 1994) have similarly shown deterioration in subjects’ proficiency well after physiological changes in the brain or biological changes associated with puberty.
-          Misattribution of age differences in linguistic abilities to neurobiological factors. While some researchers have turned to neuroscience in the hope of discovering fresh and more conclusive evidence, whose extrapolations were often readily accepted by the public, they made an error assuming that differences in the location of two (or more) languages in the brain or in the speed of processing account for differences in proficiency levels and the poorer performance of older learners. Wuillemin et al. (1994) and Furtado and Webster (1991) proved that the different localisation of the L1 and L2 in a bilingual brain cannot account for differences in proficiency. Moreover, Pulvermüller and Schumann (1994) assent that even if plasticity were related to learning, it would not explain away the great variation in ultimate achievement in older L2-ers.
-          Misemphasis on poor adult achievers and ignoring nativelike L2 speakers. This has most probably been due to an inner imperative that:

We should not be paying too much attention to the select handful of specially gifted individuals who can arguably pass for natives, but should take heed of the vast majority of people who are distinctive L2 users; we would not make the mistake of basing the study of human speech on a specially talented group such as opera-singers or mimics. (Cook 2002b:6)

which was doubtless followed by most of the parties involved. Researchers and laypersons alike have mistakenly fallen victim to the fallacy of somehow extrapolating from the frequent considerable problems with language learning of many adults over all adults. Yet, grown-ups are not a homogeneous group of linguistically lame ducks; despite overwhelming evidence that adult L2-ers on average achieve lower levels of proficiency, this evidence is insufficient to conclude the existence of a CP. In a significant in-depth study, Birdsong (1992) showed that a group of near-native speakers of French subsumed L2-ers who performed well above some of the natives. He also questioned the long-standing belief in the eventual fossilisation of adults’ L2 skills, which are believed to plateau at some point short of achieving native-like proficiency.

These three factors certainly merit consideration when the issue of the CP is raised. Additionally, Marinova-Todd et al. (2000:24) make the point that NSs’ perception of L2 speakers may be influenced by the amount of information given about the speaker and by the generally held belief about the unfeasibility of attaining native-like pronunciation. They also rightly warn against extrapolating to the conclusion that adults have learning problems because they are adults, as there are myriad factors at play that can impinge on L2 proficiency: social, psychological, educational, not least the environment: “most adult learners fail to engage in the task with sufficient motivation, commitment of time or energy, and support from the environment in which they find themselves to expect high levels of success” (op. cit.:27). For instance, a series of studies by Neufeld (1979) demonstrated that adult L2-ers could attain native-like pronunciation after being exposed to a silent period when they were only asked to listen to TL speech (thus, essentially, replicating the learning situation of young learners). Singleton (1995) found that children who take up English after the early elementary years—even as late as during high school—can become nativelike if their instructional environment is well structured, nurturing and motivating. On the other hand, Yeni-Komshian, Flege, and Liu (2000) concluded that learners who live in a foreign-language ecosystem do not automatically achieve—or consistently retain—nativelike pronunciation in their mother tongue if they left their L1 environment before age 8, which suggests that prepubescent learners may attain high levels of L2 operationality only at the expense of their L1 (and vice versa in the case of older learners, who are more likely to maintain their L1 at a high level, while the former are more prone to switch to L2 dominance if not TL monolingualism; Jia & Aaronson 1999). In the same vein, Flege (1999:125) explained the general decline in L2 pronunciation with age not as resulting from a loss of an ability, but as “a function of how well one pronounces the L1, and how often one speaks the L1.”

Still, whilst offering consolation to those who keep aiming at near-native proficiency, these accounts need not invalidate the claim that natural acquisition (without the crutch of conscious learning) gradually becomes crippled. Rather, they are consistent with the assumption that adult FLL operates tapping—or having access to—general cognitive problem-solving abilities rather than a language-specific module: “the end of the critical period is the point at which the nature of learning changes from being an automatically engaged process to one in which it becomes yet another cognitive ability” (Skehan 1998:283f.):

The presence of a definite article in an L2 English utterance by a Polish speaker may reflect the ultimate attainment of English morphosyntax with respect to the determiner system or it might have been slipped in there through the speaker’s skilful use of metagrammatical knowledge [especially as] metalinguistic ability may be fast and efficient, and … the results of metalinguistic learning may often be identical to those guided by UG. (Sharwood Smith 2004:262, 276)

Thus, the results of studies compatible with the presence of UG cannot disregard the possibility that the learners arrived at their current competence in a different manner (Prof. Romuald Gozdawa-Gołębiowski, p. c.). This will have bearing on our discussion of the learner in Appendix K.1.

source: Paradowski, Michał B. (2007) Exploring the L1/L2 Interface. A Study of Polish Advanced EFL Learners. Institute of English Studies, University of Warsaw, pp. 5–8.

[1] There is no consensus between the advocates of this claim concerning the exact age when natural acquisition becomes impossible. Penfield and Roberts (1959) argued that the optimum time for SLA falls within the first 9 years of life; Krashen situated the critical period at 5-6 years of age; Felix and Weigl (1991) at around 7; while Weber-Fox and Neville (1996) discovered brain responses to some grammatical anomalies suggesting age 4 while to others age 11 – hence, highly inconclusive. More recently, Hyltenstam and Abrahamsson (2003) argue that in general it becomes more and more difficult to acquire ‘native-like-ness’, but that there is no particular cut-off point, additionally presenting a number of cases where native-like FL proficiency was acquired during adulthood.
[2] Just as a carrier rocket which falls off when a satellite has been successfully put into orbit – it has fulfilled its role and would now only constitute unnecessary burden (Prof. Jerzy Zybert, p.c. Dec. 8 2006).
[3] A more fitting label in my opinion would be ‘the No Control Hypothesis’, since the issue concerns the extent to which the principles of UG constrain the formal properties of IL (or any human grammar, for that matter; the empirical question being, “What does UG consist of, precisely?” Samuel D. Epstein, p.c., Mar 9 2007).
[4] Susan Curtiss (1995) says in this respect:

The power to learn a language is so great in the young child that it doesn’t seem to matter how many languages you seem to throw their way… They can learn as many spoken languages as you can allow them to hear systematically and regularly at the same time. Children just have this capacity. Their brain is ripe to do this… there doesn’t seem to be any detriment to… develop[ing] several languages at the same time.

(but see also fn. 56 with respect to the last sentence). Actually, children’s receptivity of the mind seems to go beyond the domain of language alone, as it appears that they can be successfully trained to develop other cognitive abilities, such as that of playing chess – the example of the Polgar sisters probably comes to mind first, but practically all world chess champions had been spending hours over the board already as very young children. (This, of course, should be taken to imply that the language faculty is not a separate, dedicated module.)
[5] Though they conclude that this relationship could be an instance of a more general principle extending beyond the domain of language, whereby the human brain changes structurally in response to environmental demands as a function of learning (Maguire et al. 2000; Draganski et al. 2004).
[6] An explanation of why no advantage in such a salient hallmark of an individual’s linguistic make-up as pronunciation is observed in the adult learner may be that this aspect of language is probably least amenable to conscious manipulation (Ellis 1985a:109f.) – it is generally not considered to be under the adult’s conscious or immediate cognitive control (Moyer 2004); for instance, the average speaker is unaware of applying final devoicing (Prof. Jerzy Zybert, p.c. Dec. 8 2006), just as stress perception and production have been found not to correlate with proficiency in other aspects of the FL (Boyle 1987; Archibald 1998). In other words, we can generally do little about the way we sound – even if we are able to mimic the sounds of the FL. Some support for this account comes from the observation by Czyżak et al. (2007) that phonetically ‘hopeless cases’ are not always as hopeless as meets the eye, and that their low phonetic proficiency stems mainly from the failure of group teaching to address their individual needs and abilities; the fact that it is possible to rid oneself of the traces of one’s dialect, as evidenced by Sting’s forsaking his Geordie pronunciation (Gralińska-Brawata 2007), adds to the optimists’ camp. Additionally, while a speaker can avoid certain syntactic structures and this pass undetected, such avoidance is not possible with pronunciation. Another account may be the claims that infants limit the phoneme distinctions they hear to those present in their environmental language until about 1 year of age (e.g. Werker & Tees 1984; Kuhl 1994): “it is possible that representations of languages in Broca’s area that are developed by exposure early in life are not subsequently modified” (Kim et al. 1997:173); but see Marinova-Todd et al. (2000:15f.) for counteranalysis.
There are, of course, also exceptions to this rule, where individuals who embarked on learning foreign pronunciation (even if not the core of the language system, which could have been internalised earlier) after puberty nonetheless managed to master it at a native-speaker level, occasionally even surpassing indigenous models. These, however, are savants equipped with unique predispositions, as rare in everyday life as Olympic sportsmen; there are reasons to believe that FL pronunciation correlates with good sound perception, which is often identified with a bent for music (Źrałka 2006).
[7] A procedure for scanning brain activity during specific tasks.
[8] In the L1, this is most frequently visible where lexical items from the same semantic or phonetic field are unwittingly exchanged; cf. e.g. a drunkard’s question: “Przepraszam, gdzie ten pociąg… znaczy tramwaj… jedzie?” Of course, this happens in the FL as well; two control-group students from the experiment discussed in the empirical part of this dissertation transformed the direct question “Where’s the nearest post office?” into the reported “Whould you happen to know… where the nearest police station is?” (emph. added).