Hernandez, A.E., & Li, P. (2007). Age of acquisition: It’s neural and computational mechanisms. Psychological Bulletin, 133, 638-650.
This article explores possible neural and computational underpinnings of AoA effects, and presents an argument for specific aspects of language acquisition that are more susceptible to AoA effects. Hernandez and Li first present a few of the existing theoretical accounts of AoA. Brown and Watson (1987) propose AoA effects in word learning are the result of “phonological completeness,” whereby early-learned words are stored holistically (thus more easily retrieved), while late-learned words are fragmented and need to be reconstructed. However, this theory has been disputed based on the fact that in a segmentation task reaction times were found to be faster for early-learned words than late-learned words, which the authors suggest is counter to what should be the case if the late-learned words were already fragmented. (This argument was a little difficult to follow simply because the article does not explain what a “segmentation task” actually is.) Another theory of AoA effects is the “cumulative frequency” hypothesis. This theory proposes early-learned words are more easily accessed because of the additive effects of their frequent usage over time. Late-learned words, according to this theory, would have been encountered less frequently and therefore would be less easily accessed. However, research in older adults did not find AoA effects of specific words to increase with age.
Another theory is the “semantic locus hypothesis,” which posits that early-learned words have semantic advantage over late-learned words because they are represented in the semantic network first, and affect the way later learned words are semantically processed. This would suggest a semantic “map” that is formed to early words and affects later words. Hernandez and Li point out that, if this was the case, bilingual speakers would match semantic concepts to two separate forms, that AoA would transfer from one language to the next, and that L2 lexical items should inherit L1’s lexical AoA. This does not turn out to be the case, however, as L2 lexical speed is associated with the age at which the word was learned in L2 and not the corresponding age it was learned in L1. For the semantic locus hypothesis to work, therefore, there would have to be separate semantic stores for each language. Hernandez and Li therefore suggest AoA exerts effects at the lexical level, not the semantic level. Further, evidence from computational modeling reviewed in the article suggests increased rigidity in word learning with age, which would be more suggestive of AoA effects at the lexical level, since one could assume semantic processing would be more enriched with age given the increased capacity to manipulate concepts and infer semantic meaning as one gets older.
The article goes on to present some intriguing evidence from neuroimaging studies of increased activation of Heschl’s gyrus, implicated in auditory processing, when making lexical decisions to early learned words, whereas increased inferior prefrontal cortex activation, implicated in effortful activation of semantic meaning, is evidenced for late learned words. The authors suggest this may represent a reliance on auditory processing for early learned words and a reliance on semantic processing for late learned words. If the lexical/semantic distinction in AoA is correct, this would make sense, in that early encoding of lexical information would involve increased reliance upon auditory processing to detect subtleties between phonemes and morphemes in phrase construction in order to learn lexical rules. In line with this, another study cited in the article found that participants coactivated auditory representations when making lexical decisions pertaining to early learned words, even though these words were presented visually. Hernandez and Li conclude that the neural substrates of early learned lexical information appears to represent more automatic processing, whereas late learned words require additional processing, mapping lexical information onto semantic information.
This has interesting implications for the learning of a second language. Hernandez and Li review studies that have investigated semantic and lexical processing between L1 and L2. In neuroimaging studies, tasks involving syntactic processing evidenced greater AoA effects than tasks involving semantic processing. In a study of Chinese-English bilinguals, in which individuals read sentences containing syntactic violations, differences between bilinguals and native speakers arose for both syntactic and semantic violations, but were apparent at different ages. Differences between L2 learners and native speakers in the detection of syntactical violations appeared in participants who had learned English as early as 2 years of age, whereas detection of semantic violations appeared only after age 11. In addition, late learners rated themselves as less proficient in L2. Another study found increased activation relative to early learners in areas implicated in motor planning and articulatory effort when late learners of L2 processed grammatical violations, suggesting early learners were able to process words more automatically whereas late learners required additional, more effortful processing. These differences were not found in semantic processing, and instead differences in semantic processing were only found when comparing late learners with high proficiency to late learners with low proficiency. The authors suggest, therefore, that AoA effects exist for grammatical violations, whereas proficiency is associated with semantic violations. Further, the article goes on to discuss how the magnitude of AoA effects in L2 acquisition are greatest when the differences between L1 and L2 are large. For example, AoA effects in judging grammatical errors in Spanish-English learners are smaller than AoA effects in judging grammatical errors in Chinese-English learners.
The authors suggest this difference in the magnitude AoA effects is the result of the ability to detect subtle differences in syntax between the two languages, or what would constitute a syntactic violation in L2, which require phonological and articulatory demands, which in turn require a certain level of auditory and motor processing. They argue that it is here where AoA exerts its effects – that perhaps the “window” for learning lexical aspects of a second language has more to do with sensorimotor processing. They relate this to the critical period evidenced in the development of sensorimotor processing in visual domains, and in the behavioral and neural domains of learning music (perfect pitch; synchronization of motor movements to visual presented stimuli). They posit that language, in essence, reflects a sensorimotor ability. Because areas involved in sensorimotor processing undergo rapid organization and reorganization early in life, and evidence a loss of plasticity, perhaps the difficulty a late learner of L2 has in mastering the subtleties of L2 (such as noun-verb agreement in Chinese-English speakers, or verb irregularities in Spanish-English speakers), is the result of reduced plasticity in processing and articulating these subtle differences. The study by Jia, Aronson, & Wu reviewed in the next posting lends further support to this hypothesis.
I immediately think of my own step-mother, who came to America in 1975 from her native Vietnam. After 33 years speaking English and 25 years living with an English-speaking husband in an English-speaking household, she has an excellent grasp of the semantics of English, and has even written books and poetry about her experiences using English. However, she still struggles with the grammatical subtleties of English, such as dropping the last sound of words, using correct noun-verb agreements, or using the wrong verb form in sentences. By contrast, her oldest daughter, who was 6 years old when they arrived in America, has no trace of a Vietnamese accent, and sounds no different than a native speaker of English. This account may also lend support the emotional context of learning hypothesis of Harris, Gleason, and Aycicegi (2006) reviewed in the last posting: One could argue that regardless of the ability to master the grammatical or lexical aspects of language, the ability to master the semantic aspects of a second language continues throughout life, and grasping the semantic meaning would also imply a relating of that meaning to the self through emotional processes. Therefore, one could develop a richer emotional connection to certain words from L2 learned late, and a critical window for pairing emotion and words may not necessarily exist. Instead, as Harris, Gleason & Aycicegi (2006) suggest, it may be the case that the greater emotional salience of early learned words accounts for differences found between the emotional intensity of reprimands in L1 versus L2, rather than AoA effects being at play.