Skip to main content
Skip to main content
Florida Learning Disabilities Research Center



Book, Room, Flood: How statistical learning helps children figure out how to pronounce things

By: Mia Daucourt

Why is it important to read with our children? Although the reasoning may seem circular, learning to read requires reading. At its core, learning to read is a developmental process that is built up through repeated exposure to reading. The main reason for this is that reading ability is built through familiarity with words. In order to recognize words automatically, you must be able to draw from a mental store of previously encountered words. By reading with our children, we expose them to many words, along with the sounds, written features, and meanings of the words they read. It is only with this repeated exposure that children are able to build familiarity with words and their component features so that their reading can eventually become automated.

In the same way that books are grouped together in a library by their genres, authors, and topic areas, our mental library of words organizes words into meaningful categories based on the visual and sound-related features they have in common. For example, the two words ‘head’ and ‘tread’ can be evaluated and grouped together based on three different features. They have the same vowel combination (-ea), end in the same letter (d), and are pronounced with the same letter sound (with an /ē/ sound). Therefore, they would likely be “stored on the same shelf” in a mental library of words, along with other words that share the same written and sound-related features. Importantly, repeated exposure to reading allows children’s word libraries to grow both in terms of absolute size and organizational complexity, resulting in more connections between words based on a variety of features, and enabling children to recognize more words, more quickly.

In order to allow for their mental library of words to fill up and become properly organized, children must first undergo the developmental process of learning to read. To start off, before they can learn to read a full story or even a sentence, children have to learn how to read words by decoding them. Decoding is the process by which a word is broken down into its component letters, vowel combinations, and sounds so the full word can be incrementally sounded out and read. Fundamentally, this process allows children to match up written forms with their spoken sounds, a process which is formally called grapheme (the written part) to phoneme (the spoken sound) correspondence. As children become familiar with more words, it supports the development of their grapheme to phoneme correspondence (GPC) ability, allowing them to match up written and spoken word forms automatically.

Reading with our children is also important because it provides a context for teaching them about complex GPC relationships. This is especially important for reading instruction in a language, like English, that has what is called a “semi-regular orthography.” Basically, this means that English has a lot of different ways to pronounce the same vowel combinations, depending on the word in which the vowels are found. Because of this, children need to be exposed to a lot of different words in order to learn all of the potential pronunciations for a given vowel combination. For example, although “book” and “blood” have the same -oo, the vowel combination is pronounced differently for each word—/o͝o/ for book and /ə/ for blood. Based on this example, you can see that a vowel combination’s pronunciation varies depending on the vowels’ context—the letters that come before and after it. As a result, when encountering an unknown word, children can use their knowledge of similar previously-encountered words that they have stored in their mental libraries to deduce how a word should be pronounced and read based on their previous reading experience. For example, if most -oo words that also end in -k, like nook, look, and crook, usually have the same /o͝o/ sound as book, a child is likely to guess that unknown words that share the same features (-oo in the middle and end in -k) are probably pronounced in the same way based on their previous knowledge of GPC relationships. In doing so, children use the other letters in a word, or “contextual cues,” to help them read new words.

Reading with our children also supports an understanding of GPC relationships because it provides repeated word exposure that enables children to pick up on word-level regularities. They are able to get used to what words should look and sound like based on the vowels and other syllables in a word. This happens through a passive process called statistical learning, in which multiple encounters with a word allow children to pick up on common, repeated patterns in GPC relationships. As a result, children start to store information in their word libraries about the frequency of different GPC relationships for the same vowel combinations, like which GPC relation is the most likely (the one they encounter most often) and which GPC’s are less common. They can then use this contextual information on the frequency of GPC relationships to help them read words they have never seen before.

Importantly, the level of understanding of GPC relationships in children usually increases with reading experience. This makes sense because younger readers that are still developing have had less exposure to words both in terms of number and frequency and are more likely to read simpler books. In statistical learning terms, this would mean that less experienced readers have not had the experience or time to pick up on less common word patterns that may only be available in more complex books. Thus, they do not have as much information in their word libraries from which to draw from. As a result, inexperienced readers tend to rely more heavily on the most common GPC relationships that are less likely to vary based on the other word features because of their limited exposure.

A recent study by Steacy and colleagues (2018) tested whether children in grades 2-5 use their knowledge of GPC regularities, which they acquire through statistical learning and store in their word libraries, to read words they have never seen before. In order to ensure that the words were completely new to the readers, the researchers used made-up words. This also allowed the researchers to test children’s knowledge of GPC relationships because children would have to apply their knowledge of real English words to guess how a fake word should be pronounced. The way children pronounced these fake words would show whether they used the other letters and letter combinations in a word as context cues to figure out how a vowel or vowel combination should be pronounced. It would also reveal how much children knew about GPC relationships, based on whether children were able to apply both common and uncommon GPC rules to reading unknown words. The researchers also tested children’s reading-related skills, including their ability to read words, their knowledge of letter sounds (phonological awareness), and their ability to correct mispronounced spoken words, in order to see if any of these child characteristics affected how children read unknown words.

The researchers found evidence that statistical learning plays a part in children’s reading development. Specifically, results showed that less experienced readers, which have smaller, less complex mental word libraries and less exposure overall to uncommon GPC relationships, were more likely to use common, rather than uncommon, GPC relationships when pronouncing unknown words than more experienced readers. On the other hand, based on their expanded reading experience and increased exposure to nuanced GPC relationships, more experienced readers were more likely to use a variety of uncommon pronunciations for unknown words. In general, these results show that as children’s reading experience grows, children become more sophisticated in their ability to read words. This is most likely a consequence of their ability to use more information from the word (the other letters in a word) to figure out how to read an unknown word by comparing it to their larger, more complex word libraries. This positive relationship between exposure and reading skill sophistication supports the notion that statistical learning plays a part in learning to read. It is important to read with our kids because it helps us expose them to all of the nuances in the English language. As a result, they are able to build up rich, complex word libraries that include information about nuanced GPC relationships, which they can draw in order to learn to read automatically. Statistical learning through reading helps our kids transition from learning to read to reading to learn.

Steacy, L. M., Compton, D. L., Petscher, Y., Elliott, J. D., Smith, K., Rueckl, J. G., Sawi, O., Frost, S. J., & Pugh, K. R. (2018). Development and prediction of context-dependent vowel pronunciation in elementary readers. Scientific Studies of Reading, 1-15.