It is often stated that children are precocious language learners, and word learners. Although children typically learn many words between 2 and 5 years of age, misleading claims are often made regarding this: In fact we know embarrassingly little about the rate at which children acquire words. For example, we do not know whether, in a ‘head-to-head’ test, children learn words faster than adults, or how many times children must hear a word before they have some lasting memory for the word.
We have shown that young children often are rather slow to learn words, and actually are faster to learn other kinds of information than words, in carefully controlled tests.
An early study (Deák & Wagner, 2003) showed that 4-5-year-old children were slow to learn several new words for novel “creatures” in an attractive play-set. During two play sessions the experimenter named the categories in controlled but conversationally natural ways, mentioning the features that defined each category and encouraging children to repeat the names. Yet after hearing the words more than a dozen times, with clear examples, definitions, and encouragement, 4-5-year-olds had learned only about one out of four words, and did not even know which features were important for that category. The task was not especially difficult: 6-7-year-olds could learn most of the words and definitions, given the same experience. This finding made us question whether young children are actually precocious at figuring out word meanings.
We then did a series of studies (Deák & Toney, 2013) to investigate fast mapping – a presumed ability of children to learn words after one exposure – plus other purported word-learning tendencies like the mutual exclusivity bias (an assumption that each thing has just one word) and the taxonomic bias (an assumption that words refer to categories). To test whether these traits are specialized for word-learning, we taught 3- to 5-year-old children either words, facts, or pictograms (i.e., abstract graphical visual symbols; see Figure) for unfamiliar objects (see Figure).
In three experiments we found no evidence of specialized word learning:
- After 1 to 4 exposures, children showed less ability to pick out which object was associated with a word than with a fact or a pictogram. This was true after only 1 or 2 exposures, indicating that fast-mapping was not specialized for words.
- Similarly, after 4 or 5 exposures, children were produced the names they’d learned for the objects less accurately than the facts, and no more accurately than pictograms (although they were asked to draw the pictograms, and it is admittedly difficult to compare speaking and drawing accuracy).
- Children generalized the words less systematically (Experiment 1) or equally systematically (Experiments 2 & 3) as facts or pictograms to new variants of the objects.
- When taught a scheme that violated one-to-one associations between objects and words/facts/pictograms, to test the ‘mutual exclusivity bias,’ children showed a modest slowing of learning, but only in one of two conditions, and with no greater slowing for words than for facts or pictograms.
Conclusion: In a ‘head-to-head’ test, children learned new facts, or new pictograms, as fast or faster than words. There was no evidence that children used any different biases to learn words than to learn facts or pictograms. They just did not learn the words as quickly.
Another question: are children really like ‘little sponges’ that learn new languages much better than adults do? We tested this by giving a group of adults the same word-learning test. Adults vastly outperformed children, learning words much faster and more accurately. Of course, word learning is only one aspect of language learning, but in that aspect at least, children are perhaps a bit more like sieves than sponges.