When I consider my years spent learning to spell, I remember trying to master the curiously silent h, the gentle g, and the i before e rule with its litany of exceptions. I remember the anxieties of weekly quizzes and the satisfaction of correcting my first-grade teacher, who once had the unparalleled misfortune of spelling hamster with a p.
What I do not recall is a time when my spelling was prephonological, which is to say entirely free from the sound-letter correspondences that anchor it today. I don’t recall not knowing what sound an h or a g usually made, or when slipping in an extra p would have been the least of my problems because I’d written out “jabcsp” for hamster. But though I don’t recall it, chances are good that I went through a prephonological phase. Many children attempt to write out words and sentences (using real, recognizable letters) before they are burdened with knowledge about what the letters mean or even, in some instances, that the letters mean anything at all.
Meaningless letter strings, however, turn out to be anything but random. In a study published three years ago, the researchers Tatiana Cury Pollo, Brett Kessler, and Rebecca Treiman of Washington University in St. Louis analyzed American and Brazilian children’s spellings of a number of common English and Portuguese words. The study focused on those children whose spelling accuracy was not significantly better than chance (about half of the sample of three- to five-year-olds). The researchers wished to address a fundamental question: What are children’s earliest conceptions of written language?
Though the children were egregiously inaccurate, they nonetheless demonstrated some sensitivity to the statistical properties of their language (not unlike the sorts of properties that monkeys can exploit to identify words). For instance, children chose to spell most words using three or four letters, a perfectly respectable word length that is typical of the words they are likely to encounter. Children were also sensitive to variability in word length; the proportion of words spelled with one or two letters approximately matched their proportion in the wild.
Some revealing differences occurred, however, between the American and Brazilian preschoolers. Though English and Portuguese share an alphabet, the languages differ somewhat in how letters are used. English words are likelier to contain double letters, for example, while Portuguese words have a higher proportion of vowels and vowel-consonant alternations, as in luz and gota. Sure enough, the spellings generated by the American children had more double letters, while the Brazilian children produced more vowels and (even when controlling for this) more vowel-consonant alternations.
The researchers also observed some predictable—and adorable—errors. The children tended to overuse letter sequences from the alphabet (e.g., fgh), as well as the letters in their own name. Clearly, they’d gathered that in their language a small number of symbols are combined and recombined to make words and sentences, so they went with the symbols (and sequences) they knew best.
Both English and Portuguese have alphabets. What about a language like Mandarin, in which a word is written using a (non-phonological) character? At this year’s Society for the Scientific Study of Reading meeting, Treiman and her collaborator Li Yin presented evidence (not yet published) that Mandarin’s littlest scribes also produce garbled word look-alikes before they learn to write properly. They do not, however, hijack parts of their own name—either entire characters, or components of them—in the process. The researchers suspect that this is because the children don’t observe much repetition in the text they see: instead of a small number of symbols repeating often, Mandarin children are exposed to a much larger number of symbols that rarely repeat.
Our earliest scribbles, then, are faithful to our native tongues—whether English or Portuguese or Mandarin.
