Which of the following is evidence for a critical period for language acquisition?

Cognition. Author manuscript; available in PMC 2022 Jan 1.

Published in final edited form as:

PMCID: PMC7736136

NIHMSID: NIHMS1638678

Abstract

Hartshorne, Tenenbaum and Pinker [2018] used a very large sample in order to disentangle the effects of age, years of experience, and age of exposure from each other in context of second-language acquisition. Participants were administered an online test of English grammar. Results revealed a critical period ending around 17 years of age for the most effective acquisition of a second language [L2]. The findings of a late cutoff indicate the age range of late childhood to late adolescence as crucial for learning an L2. In this piece, we argue that these results can be conceptualized by emergentist models of language acquisition in which both behavior and brain interactively reorganize across development.

Keywords: critical periods, second language acquisition, development, emergentism

Introduction

Hartshorne, Tenenbaum, and Pinker [2018], henceforth HTP, present evidence for a critical period of second-language [L2] syntax acquisition that extends into late adolescence, only declining around the age of 17. We, as well as HTP, have come to the conclusion that the proposed critical period may be representative of a change in a behavioral skill with more breadth than simply L2-syntax learning. In other words, we propose that the window closing on syntax acquisition in L2 [or, for that matter, any other sub-skill of language acquisition] falls under the closing of a larger biological window of opportunity of some kind.

To better operationalize a ‘critical period,’ Johnson [2011] proposes a framework distinguishing between neurobiological and behavioral views of the critical period. Johnson suggests that, while it may be unreliable to reduce behavior to brain or biological processes, modelling the underlying hardware to a behavior [in this case, L2 syntax learning] could provide insight on the emergence of higher levels of behavior. Johnson further hypothesizes that increased performance in different behavioral domains could be explained by the development of supporting neural “hardware.” HTP rightly point out that “it is hard to say whether any of the identified neural maturational processes might correspond to the changes in syntax acquisition that we observed” [p. 275], so a framework for examining neural changes alongside behavioral performance may be of use in analyzing the developmental trajectory of a skill. In other words, combining neural and behavioral approaches would allow us to connect the development of one to corresponding changes in the other. Using this approach we attempt to elucidate what broader skills may share similar neural and behavioral developments and so underlie the changes in language-specific skills. To this end, we propose the use of two models to connect neural development with behavioral development: Interactive Specialization [IS] [M. H. Johnson, 2011] and Neuroemergentism [NM].

Our argument regarding HTP’s critical period results is that they may not be confined to the narrow domain of L2 syntax learning but represent larger behavioral and neural changes. Therefore, given the power of their study from such a large sample size, their findings may serve as an effective tool for further research into the ways in which behavior and neurobiology organize and reorganize across development. These IS and NM frameworks in concert with the critical period proposed by HTP are likely to shed light on the nature of development. The predictions listed here, based on each framework given HTP’s data, would be instrumental in further testing this critical period. In order to more clearly flesh this argument out, we will review HTP’s findings.

A short summary of HTP’s findings

Hartshorne, Tenenbaum, and Pinker [2018] use a much larger than normal sample [680,333 participants] to investigate the nature of a potential critical period for second language acquisition. HTP argue that a large sample was required in order to make sense of the potential effects that the age of an individual, age of first exposure to L2, and duration of exposure to L2 could have on their results. The use of this approach allowed the authors to arrive at a model that is better able to determine the shape of any decline in L2-learning ability. Analysis puts the onset of this decline at 10-12 years of age, marking the end of the putative “critical period” for L2 learning. Support for this conclusion comes from the small amount of variation in task performance between individuals who began L2 learning in the first decade of life. A sharp decline represents the end of ability in the ‘ultimate’ acquisition of L2 syntax [where abilities could ever reach native-like] beginning around 17.4 years of age. The authors attribute this decline to changes specific to late adolescence, rather than developmental changes that occur in early childhood or during the onset of puberty [for further discussion see Johnson & Newport, 1989].

Further support for HTP’s critical period comes from another recent study using online tests to amass an enormous sample size like HTP. In 2015, Hartshorne & Germine found that language task performance [word-pairs in long and short-term memory; word lists] dropped starkly in the late teens. Few other tasks [i.e., Reversed Lists, Long-Term Memory for Faces, Arithmetic, Vocabulary knowledge, etc.] had such early drop-offs in performance—most declines seem to occur far later in life. This [and HTP’s complimentary data] contradicts the tendency to focus on early childhood and infancy as the critical periods for language acquisition: developmentally, the experience-dependent ‘word boom’ in early childhood has always represented the peak of language acquisition, and performance afterwards is a trough [Dale & Fenson, 1996; Fenson et al., 1994; Weisleder& Fernald, 2013]. Though, even early-life critical periods of language learning have led to the proposal of a shared mechanism [as do the present proposal and HTP] with other forms of early learning [Watson et al., 2014].

Finally, as an aside, it is worth noting that critical periods in L2 have a more nuanced nature than critical periods in a first language [L1], especially when sequentially acquired [languages not learned simultaneously]. Namely, the learning of an L2 builds upon other existing skills that are already present at the time that L2 learning begins [Newport et al., 2001]. Any developmental model of language learning must contend with this while also considering the neural processes across the lifespan that may affect [or, more specifically, limit] L2 acquisition.

So far, we have reviewed HTP’s results indicating a steep decline of L2 syntax-learning ability around 17.4 years of age. HTP argue that this finding is compatible with the view that language learning ability is the only ability to show a drop so early in life. We also noted that HTP’s findings may actually be consistent with the view that general cognitive functions may be declining in late adolescence. Recall that this does, however, contrast with the far earlier critical periods for language learning that have been previously discussed in the literature. This raises the need for an explanation of the later-closing critical period; one that takes into account the complications of Unlearning through the lenses of neural development and behavioral development.

Implications of HTP’s findings

In the remaining section of this piece we will consider the implications of HTP’s findings focusing on three specific questions: 1] Why does HTP’s critical period close so much later than those generally used in the language acquisition literature? 2] What role might the decline of other broader cognitive skills play in the decline of L2 learning abilities? 3] What models may best explain the late close of HTP’s critical periods in terms of said other skills? Here, we present two models, Interactive Specialization [IS] and Neuroemergentism [NM] which focus on the relationship between the brain, behavior and development. Taken together these two models may resolve Question 3 and aid in explanations of Questions 1 and 2 using HTP’s data.

The IS framework is centered on the interaction between neighboring brain regions that are specialized to certain tasks leading to behaviors. It proposes a number of clear hypotheses on how this may play out in terms of regional specialization and inter-regional interaction: a] as brain regions specialize, the behaviors associated with activation in said area become more specific and their response to unimportant stimuli diminishes; b] amount of localization for any particular skill can be linked directly to the degree of specialization for that skill; c] a region of the brain with high levels of specialization will be less plastic in the wake of injury than it would if it were less specialized [as it may have been at earlier points in life]; d] as cognitive skills develop behaviorally, so too physically do their associated regions; e] as regions specialize and develop, so too do their networks, and vice-versa, leading to smaller, more specialized networks; f] as regions specialize, they are influenced by and so influence their neighbors’ specializations.

IS suggests that behavioral and neural development are a reciprocal process, and over time regions and networks will become more specialized on their own, as networks interconnected within themselves, and in neighborhoods [all reciprocally]. This framework has been used to account for language development. For instance, the comparison in brain activity observed before and after the period of extreme vocabulary growth during early childhood revealed language-related activity that was more generalized both in regions of activation and in stimuli to which they would react [see Johnson [2011, p. 14] for more]. Applied to HTP’s critical period data, the IS framework would make a number of predictions.

First, it would predict that activity seen during L2 syntax learning [such as event-related potentials seen in electroencephalograms or cortical blood flow activity changes in functional magnetic resonance imaging scans] would become more specialized over the course of HTP’s critical period [approximately 10-17 years]— i.e., that areas specializing in syntax-learning would show less activity to irrelevant or incorrect stimuli. Second, as key regions further specialize within the behavior, their neighboring regions will retain and refine pertinent abilities and the behavior’s neural networks will shrink. Third, injured neighboring regions’ abilities would not be as readily adopted by other regions in later years of the critical period as in earlier.

Applied to our critical questions, these might offer some solutions. In response to Question 1, the later-closing critical period may be attributed to a longer period of neural development than previously expected: perhaps the hyper-specialization of related regions continues long after heightened plasticity of infancy typically associated with L2 learning ability. To test this, a longitudinal study of language networks and/or activation [fMRI, ERP, etc.], or a cohort-based study of the same regions of interest would help to disambiguate this possibility. In particular, differences before and after early childhood and before and after puberty could be compared and tracked longitudinally. In doing so, we may also be able to answer Question 2 and observe the development of specialized regions [related skills] that were clearly involved in the network in early childhood but pruned out over adolescence and now only tangentially involved. By this method, we may be able to see what skills’ rise and fall could lead to the consequent abilities in L2 learning.

Such a methodology has recently been leveraged by Váša and colleagues in order to understand how the brain changes during adolescence [Váša et al., 2020]. To do this they tracked changes in functional connectivity in a group of 14- and 26-year-olds with a focus on cortical and subcortical brain regions over the course of more than 1.5 years. The results of their analyses revealed two different forms of change across development. Some areas showed conservative growth in that they led to stronger connectivity across time whereas others were disruptive in that the interconnectivity changed [either by strengthening or weakening] across development. The areas of the brain represented in these two different types of networks has the potential to shed light on HTP’s data. Areas that were conservative where those involved in sensorimotor processing including but not limited to domains such as movements of body parts and sensory modalities. Areas that showed disruption were associated with higher-level cognition such as memory, theory of mind, and language or sentence processing. Areas that had a disruptive role included those involved in cortical-subcortical communication as well as association areas involved in cortical-cortical processing such as the parietal and prefrontal cortex. Future studies could observe whether the closing of a critical period for syntax does or does not align with the neuroanatomical changes that are occurring during this phase. In a similar vein, studies could also ask participants to complete a number of cognitive, perceptual and motor tasks as well as some second language tasks at different ages. In this way, it would help to see the extent to which second language acquisition might rely on different underlying factors. Using both a behavioral and a neural approach, future studies could investigate the ways in which changes in the ability to learn a second language may shed light on the type of window that is closing during adolescence.

A second framework that may be of use is Neuroemergentism [NM] [Hernandez et al., 2019]. NM combines the concepts of a number of different emergentist approaches to neurobehavioral development over the lifespan, including Neuronal Recycling, Neural Reuse and Neuroconstructivism [Anderson, 2010; Dehaene & Cohen, 2007; Goldberg, 2006; Tomasello, 2009; Anderson, 2016; Karmiloff-Smith, 2006]. Essentially, NM provides a developmental framework to account for the specialization of regions and networks in the brain supporting new skills [like language]. It considers ontogenetic [individual] development in the discussion of phylogenetic [evolutionary] predispositions of certain regions to certain large-scale skills. These phylogenetically predisposed regions [for instance, the primary auditory cortex] can then be recruited during developmental periods [i.e., HTP’s critical period] to perform new functions in a broader ability [like leveraging skills from the auditory systems in language comprehension].

For example, the Visual Word Form Area [VWFA] is specialized for reading and located in the fusiform gyrus across multiple language groups. Although specialized reading had no reason to exist until very recently, reading results in very uniform patterns of neural activity [Dehaene & Cohen, 2007]. As was eloquently stated by the authors of the NM framework when discussing Neural Reuse and the VWFA: “The presence of a clearly defined and consistently located VWFA in humans provides evidence this area adapted functions from one set of areas for a very different purpose. The original area was likely involved in more elementary functions such as detecting lines, curves, and intersections. The fact that it was dedicated to such basic functions makes it, according to Anderson [2010], an ideal circuit for neural reuse and the development of human language and reading. Anderson points out that areas like the VWFA show how the brain can take evolutionarily older areas and reuse them for newer purposes.” [Hernandez et al., 2019].

NM also highlights the idea that no developmental change is isolated to any one region or skill [behavioral or cognitive]. This builds on previous work by Elizabeth Bates [1979] who used an emergentist approach to describe the nature of language representation and its neural bases. This can be seen in her classic example of a giraffe which makes use of the same neck bones present in humans. This new neck is not its own new organ, though, nor is its change isolated; the giraffe’s cardiovascular system must adapt to the new height of the head, its back legs must shorten to accommodate the new, higher center of gravity. In the same way, no brain region changes on its own, and to consider it to be singularly responsible for that skill [like a long neck for eating from high trees, or the Fusiform gyrus area for face recognition] would be short-sighted. Should there be an injury [genetic, neural, etc.] to any given skill at some point in its development, that injury would snowball across any number systems, no matter how tangentially involved.

NM complements the approach taken by IS. Both consider the development of specialization: IS via the effect of pruning [localization] of skills to ever-smaller regions and networks, and how those regions affect their neighbors’ specializations; NM via the evolutionary and individual developmental backgrounds of any given ‘old’ system that may be co-opted for ‘new’ skills and the diffuse effects of injury. NM differentiates itself by including the evolutionary background of regions as well as individual changes that occur across the lifespan.

Answers to the questions surrounding HTP’s critical period using the NM framework would be more general than IS. First, it would predict that early-life injury [genetic or neural] to any system involved in syntax learning, no matter how distal, would have an observable effect on the critical period development of that skill. Taken with IS, this could be observed by ERP or fMRI activity over that period when compared to neurodevelopmentally typical L2 learners [Question 1]. This could also help elucidate what other cognitive processes are involved in L2 syntax learning through the effect of their injury [again, no matter how tangentially connected] [Question 2]. Also, as is illustrated in the original paper, studies on the evolutionary bases of those involved systems may elucidate how they were adapted to L2 syntax learning [also Question 2]. Again, these predictions are far more abstract, but this matches the more general framework of NM.

Much discussion of this mysteriously late-closing CP is beyond the scope of this paper, which only seeks to incorporate new frameworks into elucidating its causes. Future research should also consider the age at which this CP would begin, and how that may be modulated by bilingual or monolingual early-life experience. Following this line of thought, research into the difference between early/simultaneous bilinguals and late/sequential bilinguals would also be interesting. IS predicts that during early-life, neighboring regions and networks would be more interactive and interconnected. Thus, an early bilingual could develop and incorporate various skills for use in learning and controlling their languages, while a late bilingual would be limited in their adaptive abilities by the lower plasticity of later life. Research comparing the functional connectivity of highly specialized regions involved in syntax between early and late bilinguals would highlight such incorporation [or exclusion] of other skills or networks. Research of this kind would also address how age of acquisition of L2 influences later-life skills and the differences between monolinguals and bilinguals in HTP’s original work. Understanding how early bilinguals use other skills to improve their language abilities from such a plastic period of life would aid in explaining the differences they show in those skills when later compared to monolinguals and late bilinguals.

These frameworks are neurobiological, and this discussion is meant to encourage linking cognitive and biological approaches, but the effect of environment cannot be ignored when considering critical periods and cognitive development. In fact, inherent in the discussion of genetic predisposition is the need to understand how it interacts with environmental influence. Perhaps such factors as socioeconomic status may have a strong effect on how late a CP like HTP’s closes. Considering how the effects of such factors may differ between early and late bilinguals will also play a role in understanding their influence on the development of skills such as syntax learning.

HTP’s publication involving a large number of participants led to the identification of a critical period for syntax with the window for the acquisition closing at 17.4 years of age. In this piece we describe two frameworks, Interactive Specialization and Neuroemergentism, that could be leveraged in further investigations of HTP’s evidenced critical period and the mechanisms underlying the critical period’s closure. We hope to stimulate future studies that seek to more closely link the closing of this window with neurobiological indices and accompanying psychological processes.

Acknowledgments

The authors confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome. Research reported in this publication was supported by the National Institutes of Health to the National Institute On Aging under Award Number R21AG063537 as well as to the National Institutes of Child Health and Human Development under the Award Number P50HD052117. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

• Anderson ML [2010]. Neural reuse: A fundamental organizational principle of the brain. Behavioral and Brain Sciences, 33[4], 245–266. 10.1017/S0140525X10000853 [PubMed] [CrossRef] [Google Scholar]
• Anderson ML [2016]. Neural reuse in the organization and development of the brain. Developmental Medicine & Child Neurology, 58[S4], 3–6. 10.1111/dmcn.13039 [PubMed] [CrossRef] [Google Scholar]
• Bates E [1979]. The Emergence of Symbols: Cognition and Communication in Infancy. Academic Press. [Google Scholar]
• Dale PS, & Fenson L [1996]. Lexical development norms for young children. Behavior Research Methods, Instruments, & Computers, 28[1], 125–127. 10.3758/BF03203646 [CrossRef] [Google Scholar]
• Dehaene S, & Cohen L [2007]. Cultural Recycling of Cortical Maps. Neuron, 56[2], 384–398. 10.1016/j.neuron.2007.10.004 [PubMed] [CrossRef] [Google Scholar]
• Fenson L, Dale PS, Reznick JS, Bates E, Thai DJ, & Pethick SJ [1994]. Variability in early communicative development. Monographs of the Society for Research in Child Development, 59[5], 1–173; discussion 174–185. [PubMed] [Google Scholar]
• Goldberg AE [2006]. Constructions at Work: The Nature of Generalization in Language. Oxford University Press. [Google Scholar]
• Hartshorne JK, & Germine LT [2015]. When Does Cognitive Functioning Peak? The Asynchronous Rise and Fall of Different Cognitive Abilities Across the Life Span. Psychological Science, 26[4], 433–443. 10.1177/0956797614567339 [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• Hartshorne JK, Tenenbaum JB, & Pinker S [2018]. A critical period for second language acquisition: Evidence from 2/3 million English speakers. Cognition, 177, 263–277. 10.1016/j.cognition.2018.04.007 [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• Hernandez AE, Claussenius-Kalman HL, Ronderos J, Castilla-Earls AP, Sun L, Weiss SD, & Young DR [2019]. Neuroemergentism: A framework for studying cognition and the brain. Journal of Neurolinguistics, 49, 214–223. 10.1016/j.jneuroling.2017.12.010 [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• Johnson JS, & Newport EL [1989]. Critical period effects in second language learning: The influence of maturational state on the acquisition of English as a second language. Cognitive Psychology, 21[1], 60–99. 10.1016/0010-0285[89]90003-0 [PubMed] [CrossRef] [Google Scholar]
• Johnson MH [2011]. Interactive Specialization: A domain-general framework for human functional brain development? Developmental Cognitive Neuroscience, 1[1], 7–21. 10.1016/j.dcn.2010.07.003 [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• Karmiloff-Smith A [2006]. The tortuous route from genes to behavior: A neuroconstructivist approach. Cognitive, Affective and Behavioral Neuroscience, 6[1], 9–17. Scopus 10.3758/CABN.6.1.9 [PubMed] [CrossRef] [Google Scholar]
• Newport EL, Bavelier D, & Neville HJ [2001]. Critical thinking about critical periods: Perspectives on a critical period for language acquisition. In Language, brain and cognitive development: Essays in honor of Jacques Mehler [pp. 481–502]. The MIT Press. [Google Scholar]
• Tomasello M [2009]. Constructing a Language. Harvard University Press. [Google Scholar]
• Váša F, Romero-Garcia R, Kitzbichler MG, Seidlitz J, Whitaker KJ, Vaghi MM, Kundu P, Patel AX, Fonagy P, Dolan RJ, Jones PB, Goodyer IM, the NSPN Consortium, Vértes PE, & Bullmore ET. [2020]. Conservative and disruptive modes of adolescent change in human brain functional connectivity. Proceedings of the National Academy of Sciences, 117[6], 3248–3253. 10.1073/pnas.1906144117 [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• Watson TL, Robbins RA, & Best CT [2014]. Infant perceptual development for faces and spoken words: An integrated approach: Development of Face and Spoken Language Recognition. Developmental Psychobiology, 56[1], 1454–1481. 10.1002/dev.21243 [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• Weisleder A, & Fernald A [2013]. Talking to Children Matters: Early Language Experience Strengthens Processing and Builds Vocabulary. Psychological Science. 10.1177/0956797613488145 [PMC free article] [PubMed] [CrossRef] [Google Scholar]