Singularities in inflection: Linguistic goslings or resource limitations?

Human beings today communicate in 7,000 different languages. Although many languages are expected to go extinct in the future because there are not enough people left who keep them alive, the ability to speak more than one language will continue to be in high demand. In fact, by some estimates, more of us are bilingual (i.e., speak two languages, viz. 43% of the world population) than are monolingual (40%).

It is unsurprising, therefore, that bilingualism has been a topic of considerable research in cognitive science, as attested by earlier posts on this website herehere, and here.

Of particular interest in bilingualism is the fact that many learners of a second language have persistent difficulties with particular grammatical aspects of a language. For example, consider the agreement between verbs and subjects of sentences in terms of number or gender. In English, singular and plural verb forms are largely identical in the present tense: We say “I walk”, “you walk”, “we walk”, and “they walk” without having to worry about the form of the verb. There is only one exception: We say “he (or she or it) walks”. This is far simpler than, say, in German, where you’d have to remember to say “Ich gehe”, “Du gehst”, “er geht”, “wir gehen”, or “sie gehen” depending on who does the walking.

But here is an intriguing fact: Despite the overt simplicity of English inflection, Chinese people who learn English as a second language (abbreviated to L2 in the literature) have great difficulty in managing the inflections of English verbs, notwithstanding their apparent simplicity. This difficulty can persist even after many years of immersion in an English-speaking environment.

recent article in the Psychonomic Society’s journal Memory & Cognition examined the possible reasons for this difficulty. Researcher Henrietta Lempert compared two candidate explanations that have been advanced in the literature.

On the first view, there is a “critical period” in our childhood during which we can acquire a language, and unless our native language (abbreviated L1) contains the same grammatical peculiarities as another language that we learn later in life, then our L2 proficiency will never match the competence with which we have mastered L1. It’s a bit like Konrad Lorenz’s famous geese, who “imprinted” on the mother goose or on Prof. Lorenz, depending on who took them for a walk (or even a swim) shortly after hatching:

When the goslings missed out on this early experience with Prof. Lorenz, their opportunity to acquire him as a father figure faded rapidly.

On the second view, grammatical infelicities exhibited by L2 speakers are a consequence of resource limitations. That is, although speaking seems really easy, it actually isn’t: Even L1 speakers find it difficult to speak while sharing their cognitive resources with other task demands at the same time. It is for this reason that mobile phone conversations can make drivers less safe, as we have noted in a recent post. So if driving can be impaired even in L1 speakers, imagine how readily competing task demands may interfere with L2 speakers.

To differentiate between those two candidate explanations, Lempert relied on native Chinese speakers who differed in age of immersion in English. Chinese is ideally suited for research of this type because it does not use inflections on either verbs or nouns to signal person or number. Instead, the equivalent grammatical function is conveyed through context or word order or special “function” words. It is not surprising that some Chinese L1 speakers find the inflections in L2 English highly challenging, to the point where correct production of inflections for verbs can be exceedingly rare—in one reported case, correct production was as low as 5% after 10 years of immersion in English, with no improvement after a further 8 years.

Lempert used a preamble completion task in her study. In this task, participants are presented with a sentence fragment such as:

The letter from the lawyer . . . lost,


The officers from the station . . . suspicious,

and the participant has to read out the preamble and supply the missing word. Each of the preambles was presented in a singular and a plural form, so the “letter” might be “letters” and the “officers” might be “officer” and so on.

To differentiate the resource account from the critical-period hypothesis, Lempert manipulates processing load in a subtle but effective manner by varying the animacy of the subject of the preambles. The distinction between animates—such as “officers” or “lawyer”—and inanimates—such as “station” or “letter”—is fundamental to human cognition. In general, animates are remembered better than inanimates, and many languages, including English and Chinese, have a preference for sentences that follow an animate-inanimate order, rather than the other way round. Thus, “the letters from the lawyer…” is less common and perhaps slightly more challenging to process than “the officers from the station …”, all other variables being equal.

Partial results from the preamble task are shown in the figure below.

The overall pattern of results was quite complex and nuanced, but the main findings are nicely conveyed by this figure: When the subject of the preamble was plural, the number of verb errors during production—that is, saying “was” instead of “were”—increased with the age of arrival in an English-speaking society. Bilinguals who arrived early in life (up to age 10) committed fewer errors than those who arrived in their teens (middle; 11-15) or even later (16-25). However, this overall effect was non-significant in the case of sentences commencing with an animate subject: this is represented by the panel labeled “AI” on the left. By contrast, when the subject was inanimate (panels “IA” and “II”), the age of arrival effect was particularly pronounced.

Lempert concluded that the data are “…precisely what would be expected if processing II and IA preambles consumes more resources than is the case for AI, and consequently leaves fewer resources for computing verb number agreement…” In other words, when sufficient processing resources are available, L2 proficiency can fully reveal itself, to the point that age of arrival no longer matters. And if age of arrival does not matter, then the critical-period hypothesis is unlikely to provide a fully satisfying account of L2 inflection errors.

Lempert’s finding is somewhat analogous to the idea that Konrad Lorenz’s goslings might have learned to follow him at a later age, had they not been distracted by other events in their lives. This seems highly unlikely, thus demonstrating that not everything that works for a human is good for a goose.

You may also like