A statistical learning perspective on children's learning about graphotactic and morphological regularities in spelling.
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
We put forward our view on how children learn to spell through a review of recent research on English and French children's spelling development. We examine children's learning of graphotactic conventions (legal combinations of letters), followed by a more in-depth treatment of how children learn the place of morphemes (smallest units of meaning in language) in spelling. We contrast findings from recent research with those of traditional models that suggest that children use both graphotactic and morphological information relatively late in their spelling careers and that the end-point of development, particularly for morphological conventions, lies in rule-based performance. Instead, it seems that quite young children's spellings are influenced by both graphotactic and morphological patterns and that writers do not rely (at least exclusively) on rules. We consider the possibility that children might use statistical learning to gain knowledge of both graphotactic and morphological features of the orthography. Keywords: morphology, spelling acquisition, orthographic learning, spelling development The English and French orthographies adhere to several types of regularities: specifically, phonological, graphotactic, and morphological. In this article, we put forward our view on how children's spellings are influenced by these consistencies, with a special focus on learning of graphotactic and morphological regularities. Our approach draws on the statistical learning perspective, in which it is important to consider the regularities in the (written) input on which children could rely. Accordingly, we begin by considering the kind of information represented in the English and French orthographies. French and English, like most alphabetic orthographies, rely in part on the regular connections between letters and sounds to generate word spellings. This is the case for the words cat and non (no). And yet this phonological basis does not explain the variability with which sounds are represented across different words. For example, the /eI/ sound can be spelled with -ay, -ey, -eigh, -a, and -ea at the ends of words in English and with -ez, -ai, -e, -ef, -ee, and -aie at the end of words in French. A second source of information useful to spelling comes from graphotactic regularities about the legal combinations of letters.' In French, words cannot end in a consonant doublet on its own (e.g., pomme, but not *pomm_ for the word apple in French), but they can in English. In English it is extremely unusual for words to start with consonant doublets (e.g., full, but not *fful), and such patterns never occur in French. These graphotactic patterns determine, in part, the specific letters that make up the spellings of words in both English and French. A third level of consistency in spelling comes from the smallest units of meaning in language, or morphemes. The use of morphological information determines the choice between several plausible representations of a given sound. The decision between the many different spellings of/e/ (as in bed, head, and said) is simple for some words when morphology is considered (e.g., ea in health driven by the root heal). Similarly, the sound /et/ has many different spellings in French (e.g., -aite, -ete or -ete), but a word ending in a diminutive suffix is always spelled -ette (as in fillette, little girl). Morphology even sometimes overrides the phonological basis of spelling. English regular past tense verbs end in the letters -ed, despite the pronunciation of these letters as either /t/, /d/ or /Id/ (as in walked, warned, and spotted). Similarly, in French the letters -ai are used for the sound 3 in faiseur (/fazr/, producer) to preserve the spelling of the root (faire /fer/, to do), despite the fact that e is the typical spelling of this sound. In the above examples, morphology could be useful, but success in spelling could also be achieved through word-specific memorization. …
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it