Deep Dives

The Issue

Educators sometimes witness difficulties in students’ reading development that can’t be accounted for by the “big five” skills they are usually testing and monitoring (i.e., phonemic awareness, phonics, fluency, vocabulary, comprehension). One such example can be illustrated by two students reading a text with an unfamiliar, irregular word. One student may read the last word in the line ‘He put on his socks and shoes’ as “/sh…ō…s/…Shows?... Shoes!” Another student, however, may never make it to the correct pronunciation of the word, instead either asking for help or proceeding despite the sentence not making sense. The ability the first student demonstrated to alter the phonemes implied by the printed graphemes to a familiar pronunciation of an oral word is especially important for languages such as English that have a deep orthography. Students need to be able to demonstrate flexibility to reconcile the oral pronunciation of printed words with ambiguous spellings (e.g., choosing the correct vowel sound for <ea> in each, head, or steak), irregular words (e.g., choir), or even the stress and vowel reduction patterns of regular multisyllabic words (e.g., allow). This skill is usually referred to as set for variability (SfV). Despite being discussed in the research literature for decades, SfV remains largely absent from most curricula, assessments, and professional conversations.

Should educators be paying more attention to it?

The Research

Set for variability was originally identified (initially as set for diversity) as a necessary skill that students must exhibit in order for phonics instruction to be successful, due to the frequent mismatch in a word’s printed form and oral pronunciation (Gibson, 1965). Seidenberg and McClelland’s triangle model (Plaut et al., 1996; Seidenberg & McClelland, 1989) is often referenced as a framework for understanding the relationship between SfV and word reading. The triangle model describes word reading as the integration of spelling, sound, and meaning rather than as a one-way process from letters to sounds. Set for variability can be understood as a kind of ‘phonological cleanup’ process: the reader uses lexical and semantic knowledge to adjust an imperfect spelling-based pronunciation until it matches a familiar spoken word.

Despite the time elapsed since first identifying this skill and its assumed necessity, set for variability remains a relatively understudied construct in reading research and has received limited attention in educational practice. One contribution to this lack of awareness and acknowledgement might be the inconsistencies in how it is defined, labeled, and measured. The skill of set for variability is often measured through mispronunciation correction tasks, in which students hear or generate a spelling-based pronunciation (e.g., iz-land) and must adjust it to a real word (e.g., island). While set for variability and mispronunciation correction remain the most commonly used terms in research, SfV has also been referred to as semantic and phonological ability to adjust recoding, or SPAAR (Kearns et al., 2016). Additionally, some of the research that contributes to our understanding of SfV investigates related but slightly different concepts, such as:

• linguistic/lexical/phonological flexibility- broader, less standardized terms that may or may not represent the same skill described in this article
• spelling pronunciation- the over-regularized pronunciation produced by applying grapheme–phoneme correspondences, which is often the problem SfV operates on, but not the ability to correct this pronunciation itself
• variable vowel pronunciation- one mechanism involved in SfV, but does not fully encompass the variability needed for the broader skill needed in deep orthographies, including the resolution of silent, ambiguous, or irregular consonants, syllabic stress patterns, etc.

The skill of SfV has primarily been studied as a mechanism for the ability to read irregular words, although research methodologies sometimes include ambiguously spelled words (those with graphemes that represent more than one phoneme, such as <oo>) and multisyllabic words, which are relevant for not only their variable vowel pronunciations (including vowel reduction) but also the inclusion of syllabic stress (Steacy et al., 2022). Even regularly spelled words can be tricky for some students, due to variables such as vowel length and the effects of coarticulation (Elbro et al., 2012).

Recent research suggests that set for variability may be best understood as an integrative skill rather than a purely phonological one. Successful SfV performance appears to draw on phonological knowledge, orthographic knowledge, and lexical knowledge, with the relative contribution of these sources varying by reader and by word (Edwards et al., 2022; Edwards et al., 2025; Elbro et al., 2012; Steacy et al., 2019a). These contributions make sense when we consider how a child might succeed at reading an unfamiliar irregular word. Imagine a child attempting the word debt in a text for the first time. After initially attempting to sound out the word, the child would need to first recognize that it is not a word they are familiar with (vocabulary) and that they may need to flex their pronunciation. They may also recognize that /bt/ is not a phoneme combination typically found in English words (phonology), and that the letter <b> can sometimes be silent (orthography) before confirming the corrected pronunciation is both a real word and one that makes sense in the context of the sentence (vocabulary). Interestingly, although research does acknowledge the contributions of vocabulary, phonology, and orthography to SfV, it appears to be an aspect of reading not fully captured by these other skills (Elbro et al., 2012; Kearns et al., 2016; Tunmer & Chapman, 2012). In statistical terms, this means SfV appears to account for unique variance in reading outcomes: across studies, it explains some differences in children’s word reading that remain even after researchers have considered related skills such as vocabulary, phonological awareness, decoding, or orthographic knowledge (Elbro et al., 2012; Kearns et al., 2016; Steacy et al., 2019a).

Put more simply, SfV is connected to those skills, but it does not seem to be reducible to them. 

One unresolved question is whether set for variability is primarily a cause of stronger word reading, a consequence of reading experience, or both. As children read more words, they likely gain more stored pronunciations, spelling patterns, and examples of variable grapheme-phoneme relationships; at the same time, stronger SfV may help children make better use of those encounters with print. Savage et al. (2025) add an important nuance to this question. In their analysis, SfV was most clearly related to reading growth after intervention had ended, predicting word-reading outcomes from an immediate posttest to delayed posttest (three months after the intervention had concluded) rather than simply explaining who improved immediately during instruction. This suggests that SfV may matter most as a tool for continued word learning: once students have received instruction, those with stronger flexibility may be better able to use later encounters with print to refine and expand their word-reading knowledge.

Set for variability appears to be most relevant to word-level reading outcomes, particularly real word reading, irregular and complex word reading, and word identification (Elbro et al., 2012; Kearns et al., 2016; Steacy et al., 2019b; Steacy et al., 2022; Tunmer & Chapman, 2012). Some studies also suggest that SfV may contribute to decoding or broader word-reading performance, not just exception-word reading, because even regular words can require flexibility in pronunciation (Elbro et al., 2012). However, although these studies are promising, we have not accumulated the level of evidence necessary to begin including SfV as a construct in reading screeners or diagnostic tests. Other constructs (such as phonemic awareness, nonsense word decoding, RAN, etc.) have extensively replicated research establishing them as predictors of reading outcomes as compared to set for variability. It boils down to the fact that we do not yet know if or how including SfV might improve the validity of tests attempting to identify the risk or presence of reading difficulties.

There are also practical and conceptual assessment challenges. SfV is usually assessed by mispronunciation correction tasks, which can vary in whether students hear spelling-based pronunciations, random mispronunciations of isolated words, or mispronunciations embedded in sentence contexts, and these variations may not measure exactly the same thing. In addition, many SfV tasks require individual oral administration, which may make them harder to use efficiently in schoolwide screening systems than measures such as nonsense word reading, RAN, or oral reading fluency. A related open question is whether weak SfV is especially characteristic of certain reader profiles, such as students with dyslexia or students with developmental language disorder. This question may also connect to how children understand unfamiliar accents or other nonstandard pronunciations: in both cases, the listener or reader must map an unexpected pronunciation onto a known word. At present, the research is not yet developed enough to know whether SfV should be treated as a broad instructional concern or as a more specific explanatory factor for some struggling students.

Importantly, we also have not yet fully established whether SfV is a construct that is amenable to training (such as phonemic awareness) or more primarily situated for screening/assessment purposes (such as nonsense word reading and RAN). In other words, we don't know if the time spent teaching the skill would be well spent. Some research studies show that students can increase their set for variability skills with training; however, this progress usually leads to minimal (Dyson et al, 2017) or no (Zipke, 2016) transfer to reading words that weren’t specifically targeted during the interventions. Other studies suggest that students may benefit just as much from other approaches that build flexibility into word reading instruction, such as teaching variable vowel pronunciations (Steacy et al., 2016) or morphology (Dunn et al., 2025; Savage et al., 2025).

In Sum

The current evidence places set for variability in an important but still unsettled position. It appears to capture something essential in the move from decoding to word recognition, especially when a printed word does not map neatly into its spoken pronunciation. But the research has not yet posited set for variability as a skill demanding inclusion in screening, assessment, or intervention protocols. For now, its value may be simply in helping educators understand and interpret a specific kind of reading difficulty, and reminding them not to assume that knowing grapheme-phoneme correspondences and being able to blend them is always sufficient. While it comes naturally to some, other students need to learn how to evaluate, adjust, and confirm their pronunciation attempts.

Questions we just don’t have the answers to:

1. How should set for variability best be operationalized if used to identify the presence or risk of reading difficulties?
   1. Should mispronounced words be presented in isolation or in context?
   2. Should mispronunciations of words be limited to “regularized” pronunciations of spellings? Or should it encompass a broader scope of possible mispronunciations?
2. Would including tasks measuring set for variability improve the validity of screeners and tests attempting to identify the risk or presence of reading difficulties?
3. Are children with dyslexia more likely to experience difficulty with set for variability?
   1. Developmental language disorder?
4. Do children with difficulty with SfV also experience challenges with simple CVC decoding?
   1. When are their challenges first identified?
5. How much does SfV training help students generalize this skill to reading other, nontrained words?
6. Does increased reading growth naturally increase SfV? Or is it the other way around?
   1. Does training vocabulary improve SfV? Morphology?
   2. Are the relationships between these skills based on causation or only correlation?

Nitty-gritty of some of the research:

Tunmer & Chapman, 2012

Participants

• 140 children (with a mean age of 5.9 at the time of the first testing period) from 16 different schools

Study Design

• Students were tested at four time periods over the course of three years. Set for variability was assessed by children’s ability to correct a mispronunciation made by a puppet, both in isolation and when produced within a sentence.

Findings

• Students’ ability to correct a mispronunciation produced in isolation (without sentence support) was primarily responsible for the relationship between SfV and irregular word reading. The study found a relatively strong correlation between concurrent SfV and phonemic awareness skills and a moderate correlation between concurrent SfV and vocabulary. Also, SfV had a small but significant amount of unique variance in irregular word reading (when tested a year later), and influenced student decoding and word recognition abilities after three years.

Notes/Limitations

• This research is seen as one of the core foundational studies of set for variability, suggesting that SfV is a construct beyond the manifestation of vocabulary knowledge. Importantly, this study was not experimental, and the relationships suggested are correlational only, without suggestion of what variables may have caused a skill or deficit.

Kearns, et al., 2016

Participants

• 94 English-speaking third- and fourth-grade students from five schools, approximately half with typical achievement and half with reading difficulty. A larger sample of 206 students in grades 2-4 was also used for the mispronunciation correction task.

Study Design

• Researchers administered a mispronunciation correction task to measure set for variability (which the authors termed semantic and phonological ability to adjust recoding, or SPAAR), along with assessments of vocabulary, morphological awareness, attention, phonological awareness, phonological working memory, decoding, and word reading.

Findings

• SfV predicted students’ word reading abilities even after accounting for vocabulary, morphological awareness, attention, phonological awareness, phonological working memory, and decoding. Notably, once SPAAR was included in the models analyzing student performance, phonological awareness no longer independently predicted word reading. This suggests that SfV may capture a reading-related skill that overlaps with phonological awareness, but is not fully explained by it.

Notes/Limitations

• The study design suggests a correlational relationship, so it cannot establish whether SfV causes stronger word reading, whether stronger reading experience improves SfV, or whether the relationship is reciprocal. It is also noted that the study did not separately assess regular from irregular word reading.

Savage, et al., 2025

Participants

• 273 struggling readers in Canada (mean 7.7 years)

Study Design

• Students received 4× 30-minute intervention sessions per week for 15 weeks, totalling an average of 26 hours. Classrooms were randomly assigned to one of two intervention conditions. One condition focused on set for variability: teaching irregular words through mispronunciation correction. The other focused on morphology: teaching irregular words by identifying affixes and word bases and building word sums with word matrices. All students’ intervention also included a focus on phonological awareness and systematic synthetic phonics. Students were tested prior to intervention, immediately following the intervention, and again approximately 3 months after the intervention had concluded. Whereas Dunn et al. (2025) report on the main intervention comparison, this study analyzes which student skills predicted responsiveness to intervention over time.

Findings

• This study suggests that the benefits associated with SfV may not be immediately evident during or right after instruction. In their study, SfV was most clearly related to reading growth from posttest to delayed posttest, suggesting that stronger SfV may support students’ continued word learning after intervention has ended.

Notes/Limitations

• As students who scored within one standard deviation from the norm in word reading skills were excluded from the study, the results should not be generalized to typically developing readers. The study also does not show that SfV instruction caused the delayed reading growth; rather, it shows that individual differences in SfV after intervention predicted later reading growth.