Pseudoword spelling: insights into sublexical representations and lexical interactions

ABSTRACT

In this work we introduce a new tool for measuring English spelling-sound consistency, the PG Toolkit, which we use to conduct detailed analyses of pseudoword spellings that provide new insights into the nature of sublexical and lexical representations. There are several key findings: first, sound-spelling consistency measured at two different “grain sizes”, phonographeme and onset/rime, each explained unique variance in pseudoword spelling. Second, lexical skill was more related to pseudoword accuracy at the onset/rime level than at the phonographeme level, and individuals who chose more consistent mappings to spell pseudowords tended to have better lexical skill. Finally, no unique contribution of consistency in the reading direction (“feedback”) was found after controlling for consistency in the spelling direction. Taken together, the results validate the various measures provided by the PG Toolkit and establish new evidence that supports an interpretation of sublexical processes as operations over hierarchically-structured representations.

KEYWORDS:

• Sublexical
• spelling
• pseudowords
• PG Toolkit
• consistency
• dual-route

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 In some literatures, the spelling “rhyme” is used to refer to the phonological representation and “rime” to the orthographic. Here we use “rime” throughout to refer to the mapping between the two, and phonographeme to refer to the mapping of individual phonemes to graphemes (note that graphemes include multi-letter units such as the digraph CH).

2 For example, the six-letter word THRIFT decomposes to three units when measured as onset, nucleus, and coda (THR, I, and FT respectively), compared to five units when measured at the phonographeme level (TH, R, I, F, and T).

3 Specifically, word-initial /æ/ is spelled A_E in AXE, ANNE, and the two-syllable pronunciation of AVERAGE (i.e., AVE-RAGE). Word-initial /æ/ is spelled AU in AUNT/AUNTIE (the common American pronunciation /ænt/ rather than /ɑnt/).

4 Irregular segments were defined as those with ≤ 18% P→G consistency per the PG Toolkit (i.e., given the phoneme in that position, it is spelled with those letters no more than 18% of the time).

5 Given that a participant may correctly spell regular segments using sublexical processes (e.g., the G in “gauche”), accuracy on those segments was not used in analyses because it would be a less valid measure of lexical knowledge.

6 For example, in Experiment 1 the dispersion ratio ≈ 9.2, indicating significant overdispersion p < 0.001, when including random intercepts only by-phonological target. When nesting spellings within phonological targets, the dispersion ratio was ≈ 0.2, no significant overdispersion.

7 VIF were even higher if the mean probabilities, instead of the minimum, were used. Even without adding in the G→P variables, the P→G variables had VIFs > 17 when using the means, compared to <6 when using the minimums.

8 We also ran the stepwise regression with entering the lexical variables last, in Step Three. Those results are presented in Supplementary Materials, Tables S3 and S4. The relative importance of the lexical predictors was diminished, but importantly they still explained significant amounts of variance even after first controlling for sublexical measures.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.