January 2019: Critical Periods in Language Learning

Hartshorne, J. K., Tenenbaum, J. B., & Pinker, S. (2018). A critical period for second language acquisition: Evidence from 2/3 million English speakers. Cognition, 177, 263–277.

Summary by Dr. Arturo E. Hernandez


A critical period for language learning refers to the notion that certain ages are ideal for language learning and that others are less than ideal. This notion also has been applied to the learning of a second language. One problem with the literature is that studies have been unable to differentiate between the age at which one has to start learning a second language and the time it takes to become proficient at a second language. In other words, if a person learns a second language at age 10, we won’t know whether they can attain a native-like level until many years later. To figure this out, we would need a large group of people who had learned at many different points in the past but who were many different ages at the time of testing. This is precisely the question that Hartshorne, Tenenbaum, and Pinker (2018) set out to test in their study.

Authors’ Approach

The authors begin with a discussion of learning rate (or learning ability, which they use interchangeably) and critical periods as they pertain to second-language learning. Hartshorne et al. define a critical period as a “theory-neutral descriptor of diminished achievement by adult learners, whatever its cause” and learning rate as the difficult-to-test issue of a consistent decline in the ability to learn a second language over time. They point out that previous research confounded learning rate reduction with actual performance reduction. For example, the data may show a steep drop in the ability for those who learned English at 12 or older when making judgments about grammar as adults relative to younger learners of English as adults. On the surface, this would appear to indicate that age 12 is the beginning of the critical period and that any learning of a second language becomes more difficult after that point. The problem is that researchers would need a lot of data from people who began to learn at age 12 but are now different ages and thus have had more years to learn the language. Only by gathering data from thousands of individuals would it be possible to see how learning at age 12 differs from individuals who are age 14, 16, 18, 22, and so on when years of experience in the second language are the same. 

The authors identified three key variables to untangle this question: age of acquisition (or age of first exposure to the language), years of experience, and age of testing. By looking at all three of these variables they would have a better chance to capture the point at which learning rate began to diminish because part of this analysis would involve those who were some years past the potential critical period. The research team decided to test hundreds of thousands of participants to be able to isolate the contribution of each of the three key variables and evaluate their change over time.


The test consisted of two parts. First, participants were asked to list their languages learned from birth, current place of residence, first exposure to English, how long they had lived in an English-speaking environment, and family members fluent in English. From there, a “shotgun approach” grammar test examined overall English capability—questions included “critical items” designed to weed out cultural and other-language influence, including common mistakes for first speakers of other widely used languages and slang from different English-speaking nations. This test was distributed over Facebook and went “viral,” collecting more than 680,000 test subjects.


Of the total sample of 680,333, a sample of 558,265 subjects was used for the final analysis. Subjects were dropped for a number of methodological reasons. Then the researchers used models that would help to capture the change occurring. The authors report that a critical period occurs at 17.4 years of age with an exponential decay after that. In short, those who learned after the age of 17.4 showed a steep drop on their scores on the English test.


The primary question behind the Hartshorne et al. (2018) study is why later learners do not reach the same language attainment as mature native speakers. In answering this question, it is important to bear in mind that the latest age of native-like attainment is not equivalent to the age when ability declines. Native-like attainment can be reached several years after the age of 17.4, although the learning rate will decrease after that age. The present study suggested that native speakers in immersion settings score the best on language tests at about 30 years old. Therefore, someone must start learning between 10 and 12 years old to attain native-level proficiency; otherwise, they run out of time before the sharp decrease in learning rate at 17.4 years. In summary, this study found that grammar-learning ability is intact throughout childhood but decreases swiftly from late adolescence and into early adulthood.

Secondary Analysis

Hartshorne et al.’s study also helped to shed light on the ideal age to learn English, depending on a speaker’s first language. The authors limited themselves to three language groups that had a large number of first-language, Romance language, Germanic language, and Mandarin Chinese speakers. The results revealed that the ideal age to learn English as a second language depended on the first language. Romance language speakers benefited the most from simultaneous acquisition of English. Germanic speakers benefited from learning between ages 1 and 5. Mandarin Chinese speakers showed the greatest benefit when learning English between the ages of 6 and 10. Other analyses revealed demographic effects, including higher accuracy given postsecondary education and being female. Thus, secondary analyses suggest that the particular language history, education, and gender also play a role in second-language attainment.


This study had to use a written comprehension test, given the large sample size. Additionally, because people make few language errors, there may be a ceiling effect in the data. That is, people did well in general, which might impair the ability to distinguish between different levels of performance. In addition, it’s not clear whether the observed test focuses entirely on grammar. Many questions on the test involved general problem-solving strategies, which potentially could be attributed to cognitive strategies. This might partially explain the fact that the decline in second-language learning ability occurred so late relative to other studies. Adolescence is a time of great change and flexibility. Thus, the continued ability to learn a second language might be due to general changes and not to those specific to the learning of grammar. 


This study has implications for the learning of language and literacy in the United States. First, a great number of nonnative English speakers are entering the school system. The Hartshorne et al. (2018) study shows us that the nature of the first language may matter for the timing and the way in which English is introduced to English language learners. For example, it may be the case that Spanish speakers benefit from early exposure to English but also would benefit from continued exposure to their native language. On the other hand, Mandarin speakers may not show any penalty for learning English later in childhood. Second, the study shows that language learning is a protracted process that requires a considerable amount of time. The bolstering of basic language abilities across development may be paramount, especially for those who come from households with relatively low levels of parental education. 


The authors found that second-language learning ability diminishes at a much later age, 17.4 years, than previously thought. Secondary analyses showed that first-language type, education, and gender also played roles in second-language ability level. Future studies should focus on using different types of grammar tests and investigating the role of problem-solving ability as a possible factor in the critical period effect observed. Nevertheless, this study shows us how learning a second language successfully is possible even in very late adolescence.