I recently published a paper called “Which Words Are Special? Identification of “Sight” Words in Educational Resources” in Reading Research Quarterly with Lauren Schilling, a current graduate student at Harvard and former research assistant of mine at the University of Wisconsin, and Mark Seidenberg, my graduate adviser.
Early reading instruction typically prioritizes a set of “special” words (often called “sight,” “trick,” “snap,” or “high frequency” words) that are deemed instructionally important based on some combination of factors related to their irregularity/ inconsistency (in terms of orthographic-phonological structure) and/ or their frequency (a property that is related to consistency but most often conceptualized as something that is theoretically distinct). Importantly, there has historically been a lack of consensus on which words actually belong in this category and the properties that should constitute a word’s putative status as one that deserves special attention.
The study analyzes six influential early reading resources: the Dolch list, the Fry list, Fountas & Pinnell’s “Words to Know,” Fundations’ “Trick Words,” Wonders’ “High-Frequency Words,” and Kilpatrick’s “Irregular Words.”
There is substantial disagreement across these six lists, a total of 973 unique words were identified. Remarkably, only 28 words (3%) appeared in all six lists, while more than half of the words appeared in only a single resource. The word lists differ significantly in terms of a number of lexical properties, including frequency, spelling-sound consistency, age of acquisition, imageability, and word length. For example, Kilpatrick’s list is an outlier featuring lower-frequency, later-acquired, and longer words, while others like Fry focus tightly on very high-frequency terms. However, the differences go well beyond this.
We provide a number of ideas in closing the article that attempt to move away from solutions that rely primarily on intuition rather than explicit, scientific theories of learning, as well as proposing an alternative, reproducible method that jointly evaluates word frequency and spelling-sound consistency using Euclidean distance to select the most instructionally valuable words. Instead of requiring children to memorize rule-violating “sight words” by rote, we advocate for instruction that groups words in a way that highlights overlapping statistical patterns (e.g., “have,” “had,” and “has”) even when the overlap doesn’t fall under classic “phonic” type patterns associated with them. This approach will engage children’s natural statistical learning mechanisms, making early reading acquisition more efficient and mutually supportive of word learning in other modalities (in spoken language comprehension and production, writing, or otherwise).
The article is open access and can be found here.