A baby hears the chair in the presence of a chair and attempts to map language onto the world. But, as most famously illustrated by philosopher Willard Van Orman Quine, what in the world is the chair? Is it an armrest? A cushion? The elongated, chair-shaped shadow on the wall? Given the complexity of our environments, how are such mappings ever successful?
Language-acquisition researchers have given this question plenty of attention. Some argue that infants are born ready to map labels to entire objects, rather than parts or features of objects. Others suggest that encountering words many times, in many contexts (a phenomenon I describe here), helps infants home in on correct mappings. But the reverse question is less often raised: Which word should be mapped to said object? Speech to children rarely consists of isolated words, with nouns often pairing with articles. So is the chair called chair or the?
Ruling out the would be trivial if infants had some way of knowing which words are so-called function words. (Instead of describing concepts, function words—which also include prepositions, pronouns, and conjunctions—describe grammatical relationships between concepts; hence their meanings map only uneasily onto objects in the world.)
Function words do sound somewhat different from other words—they tend to be very short and unstressed—so infants could theoretically use this characteristic to discount the words as potential object labels.
An even more promising avenue is frequency. Function words are among the most common in languages. A growing body of research—including a study published this year by Harvard University’s Jean-Rémy Hochmann—indeed suggests that infants may resist labeling objects with very common syllables.
In Hochmann’s study, 17-month-olds listened to a stream of artificial speech consisting of syllables, some of which occurred frequently and the rest infrequently. After two minutes of this, the babies then heard a nonsense word composed of a high-frequency syllable (vo, for instance) and a low-frequency one (mu). Vomu vomu vomu: the infants heard this word over and over again. All the while, a strange-looking 3D object rotated slowly on a monitor. (One such object looked not unlike a plus sign made of balloons; another resembled the torso of the Michelin man.) Before long, the infants had learned to associate vomu with the object.
But say vomu wasn’t a single word at all. Say it was an article like the, followed or preceded by a noun like chair. Would babies, if given evidence of this, prefer to treat the low-frequency mu as the object label over the high-frequency vo?
To test this, Hochmann created two new nonsense words, one containing the high-frequency syllable and the other containing the low-frequency one (e.g., gimu and vona). The words were played for the babies in the presence of two 3D objects, one of which they’d seen earlier. The infants better preferred the familiar object when it was paired with gimu than vona, suggesting that they considered mu—the lower frequency syllable—the superior object label. (A follow-up study found a similar effect even when the lower frequency syllable occurred at the beginning of the training word, the equivalent of chair the.)
From where might this tendency arise? Perhaps infants become overly familiarized to the high-frequency syllables and simply tune them out, much as we do the humming of lights or the whooshing of cars down a busy street. Or perhaps these familiar syllables serve as mental bookends of sorts, allowing the babies to distinctly perceive and remember where other words—potential object labels—begin and end. Or perhaps by 17 months, children have already formed the grammatical category “function words” in their language, learned that function words tend to be common, and now expect all highly frequent syllables to be function words. To know for sure, we’ll have to go younger. Open question: Do babies younger by half—old enough to understand language to be referential, but still in the earliest stages of recognizing function words—also refrain from searching the world for the’s?
Permission required for reprinting, reproducing, or other uses.