Frost has gathered compelling data from Semitic languages showing that skilled readers are sensitive to letter order. He therefore argues that paradigms such as transposed-letter priming (TLP) cannot tell us anything of universal significance regarding the brain and lexical processing. Here I argue that developmental data from an orthography showing TLP (English) lead to the same conclusion. Children do not learn to read, nor spell, in a cognitive vacuum. Instead, their errors during learning demonstrate that “orthographic” learning cannot be separated from other aspects of language development. Learning to recognize printed words involves complex interactions with phonology, morphology, and meaning from the very beginning of learning.
When children are learning to spell, letter-order sensitivity is mandatory. If young children produce spellings that are insensitive to letter order, perhaps writing “GAOL” for “goal” or “LOIN” for “lion,” these spellings are wrong. In orthographically inconsistent languages like English, children make many such errors, and therefore rote-learning of spelling conventions and weekly spelling tests are a ubiquitous part of primary (elementary) education. Indeed, children who are sensitive to letter order but who produce spellings that are not phonetically acceptable (e.g., “COUGE” for “cough” or “SUARCH” for “search”) are those children who are educationally at risk for disorders of literacy (Frith Reference Frith and Frith1980). Frith reported that 12-year-old children whose spelling errors were phonetically acceptable (e.g., “COFF” for “cough,” “SURCH” for “search”) usually resolved their spelling problems after further rote-learning. In contrast, children who had produced spellings like “COUGE” went on to have literacy difficulties.
Early reading errors show the strong influence of meaning in English children's early orthographic behaviour. For example, Seymour and Elder (Reference Seymour and Elder1986) noted visual word recognition errors like “TIGERS” for “lions,” “WHEELS” for “car,” and “GIRL” for “children.” These substitutions are semantically appropriate, but they ignore the letter–sound correspondences in the target words: Meaning can trump phonology in early reading. However, at the same time, when learning to spell, young children will spell words accurately that they cannot read (e.g., “bun,” “mat,” “leg”; Bradley & Bryant Reference Bradley and Bryant1979; Bryant & Bradley Reference Bryant, Bradley and Frith1980). This suggests that in early spelling, phonology is the trump card. Children can build up a spelling pattern like “bun” by analyzing the sounds in the target word, but they then fail to recognize the visual word that they have just written down when asked to read it aloud.
Many of children's spelling errors show the importance of phonology. Very young spellers use phonological insights that conventional English orthography ignores. Thus, words like “chair” and “truck” are spelled with the same onset (“CH”). This is phonologically appropriate because the first sound in the word “truck” is affricated, hence children misspell “truck” as “CHRAC.” Other examples include “ASCHRAY” for “ashtray” and “CHRIBLS” for “troubles” (Read Reference Read1986). Spelling errors will even differ by English dialect. American 7-year-olds (rhotic dialect) will misspell “girl” as “GRL,” whereas British 7-year-olds (non-rhotic dialect) will misspell “girl” as “GEL” (Treiman et al. Reference Treiman, Goswami, Tincoff and Leevers1997). The rhotic dialect of American phonology pronounces the letter “R” after the vowel, and this linguistic knowledge affects orthographic learning. Very young spellers add more letters than conventional to reflect phonology; for example, “SOWEMEG” for “swimming” (complex onset SW). They also use letter names in spelling (as in “TOM NICTA CR” for “Tom nicked a car”; Read Reference Read1986; Treiman Reference Treiman1993). The letter name of “R” sounds like the phonological rime of “car.”
Children's spelling errors also show that morphological knowledge affects orthographic learning. For example, young children produce errors like “KIST” for “kissed” and “CLAPT” for “clapped.” However, after learning the “ed” rule, some children over-apply the rule to spell words that are not verbs, for example, producing “SOFED” for “soft” and “NECSED” for “next” (Nunes & Bryant Reference Nunes and Bryant2006). This is not a rare anomaly: More than half of the children in this longitudinal study produced errors of this kind. Gradually, children realize that words like “kissed” have two morphemes, whereas “mist” has only one morpheme, and stop producing these spelling errors (by the age of around 9 years). Nunes and Bryant (Reference Nunes and Bryant2006) argue that this is not because of direct teaching, as they measured the teachers' morphological knowledge and found they were unaware of which verbs added “ed” (“kiss–kissed”) and which did not (“sleep–slept”). Rather, such errors decline via implicit learning of the complex interactions between phonology, morphology, and meaning.
Clearly, therefore, young learners apply multiple linguistic strategies to the processing of orthographic form. Implicit orthographic learning also affects the recovery of phonology from print, showing that implicit learning is bi-directional. English 7-year-olds are better at reading non-words like “BICKET,” which use orthographic segments of real words like “TICKET” to represent phonology, than matched non-words like “BIKKET” (51% versus 39%; see Goswami et al. Reference Goswami, Porpodas and Wheelwright1997). Implicit knowledge also affects silent reading. In a proofreading task, 12-year-olds detected significantly more silent-letter omissions (“SISSORS” for “scissors”) than pronounced-letter omissions (“SCARELY” for “scarcely”; Frith Reference Frith1978). This shows the influence of implicit morphological knowledge. Silent letters often reflect morphological roots (Latin “scissor” for “cutter”). Similarly, 8-year-olds recall more words on the basis of a silent-letter prompt (“C” for “SCISSORS”) than a pronounced-letter prompt (“C” for “SCRIBBLE”; Ehri & Wilce Reference Ehri and Wilce1982). Just as for skilled adults, the task in hand matters for the strategy that is employed. Letters are treated differently by the cognitive system, depending on the specific linguistic environment created by the (experimental) task.
Data like these provide a wealth of evidence supporting Frost's argument that models of visual word recognition must reflect the basic fact that words have phonological, semantic, and morphological characteristics as well as orthographic characteristics. The errors made by young learners of English show that children learning to recognize visual word forms seek linguistic coherence throughout the skill acquisition process. A universal model of reading hence needs to take account of developmental data as well as cross-linguistic data. Even though TLP is found in skilled readers of English, data from English learners support Frost's theoretical view. Letter-order insensitivity is an idiosyncratic strategy used by skilled readers to optimize encoding resources in certain linguistic environments.
Frost has gathered compelling data from Semitic languages showing that skilled readers are sensitive to letter order. He therefore argues that paradigms such as transposed-letter priming (TLP) cannot tell us anything of universal significance regarding the brain and lexical processing. Here I argue that developmental data from an orthography showing TLP (English) lead to the same conclusion. Children do not learn to read, nor spell, in a cognitive vacuum. Instead, their errors during learning demonstrate that “orthographic” learning cannot be separated from other aspects of language development. Learning to recognize printed words involves complex interactions with phonology, morphology, and meaning from the very beginning of learning.
When children are learning to spell, letter-order sensitivity is mandatory. If young children produce spellings that are insensitive to letter order, perhaps writing “GAOL” for “goal” or “LOIN” for “lion,” these spellings are wrong. In orthographically inconsistent languages like English, children make many such errors, and therefore rote-learning of spelling conventions and weekly spelling tests are a ubiquitous part of primary (elementary) education. Indeed, children who are sensitive to letter order but who produce spellings that are not phonetically acceptable (e.g., “COUGE” for “cough” or “SUARCH” for “search”) are those children who are educationally at risk for disorders of literacy (Frith Reference Frith and Frith1980). Frith reported that 12-year-old children whose spelling errors were phonetically acceptable (e.g., “COFF” for “cough,” “SURCH” for “search”) usually resolved their spelling problems after further rote-learning. In contrast, children who had produced spellings like “COUGE” went on to have literacy difficulties.
Early reading errors show the strong influence of meaning in English children's early orthographic behaviour. For example, Seymour and Elder (Reference Seymour and Elder1986) noted visual word recognition errors like “TIGERS” for “lions,” “WHEELS” for “car,” and “GIRL” for “children.” These substitutions are semantically appropriate, but they ignore the letter–sound correspondences in the target words: Meaning can trump phonology in early reading. However, at the same time, when learning to spell, young children will spell words accurately that they cannot read (e.g., “bun,” “mat,” “leg”; Bradley & Bryant Reference Bradley and Bryant1979; Bryant & Bradley Reference Bryant, Bradley and Frith1980). This suggests that in early spelling, phonology is the trump card. Children can build up a spelling pattern like “bun” by analyzing the sounds in the target word, but they then fail to recognize the visual word that they have just written down when asked to read it aloud.
Many of children's spelling errors show the importance of phonology. Very young spellers use phonological insights that conventional English orthography ignores. Thus, words like “chair” and “truck” are spelled with the same onset (“CH”). This is phonologically appropriate because the first sound in the word “truck” is affricated, hence children misspell “truck” as “CHRAC.” Other examples include “ASCHRAY” for “ashtray” and “CHRIBLS” for “troubles” (Read Reference Read1986). Spelling errors will even differ by English dialect. American 7-year-olds (rhotic dialect) will misspell “girl” as “GRL,” whereas British 7-year-olds (non-rhotic dialect) will misspell “girl” as “GEL” (Treiman et al. Reference Treiman, Goswami, Tincoff and Leevers1997). The rhotic dialect of American phonology pronounces the letter “R” after the vowel, and this linguistic knowledge affects orthographic learning. Very young spellers add more letters than conventional to reflect phonology; for example, “SOWEMEG” for “swimming” (complex onset SW). They also use letter names in spelling (as in “TOM NICTA CR” for “Tom nicked a car”; Read Reference Read1986; Treiman Reference Treiman1993). The letter name of “R” sounds like the phonological rime of “car.”
Children's spelling errors also show that morphological knowledge affects orthographic learning. For example, young children produce errors like “KIST” for “kissed” and “CLAPT” for “clapped.” However, after learning the “ed” rule, some children over-apply the rule to spell words that are not verbs, for example, producing “SOFED” for “soft” and “NECSED” for “next” (Nunes & Bryant Reference Nunes and Bryant2006). This is not a rare anomaly: More than half of the children in this longitudinal study produced errors of this kind. Gradually, children realize that words like “kissed” have two morphemes, whereas “mist” has only one morpheme, and stop producing these spelling errors (by the age of around 9 years). Nunes and Bryant (Reference Nunes and Bryant2006) argue that this is not because of direct teaching, as they measured the teachers' morphological knowledge and found they were unaware of which verbs added “ed” (“kiss–kissed”) and which did not (“sleep–slept”). Rather, such errors decline via implicit learning of the complex interactions between phonology, morphology, and meaning.
Clearly, therefore, young learners apply multiple linguistic strategies to the processing of orthographic form. Implicit orthographic learning also affects the recovery of phonology from print, showing that implicit learning is bi-directional. English 7-year-olds are better at reading non-words like “BICKET,” which use orthographic segments of real words like “TICKET” to represent phonology, than matched non-words like “BIKKET” (51% versus 39%; see Goswami et al. Reference Goswami, Porpodas and Wheelwright1997). Implicit knowledge also affects silent reading. In a proofreading task, 12-year-olds detected significantly more silent-letter omissions (“SISSORS” for “scissors”) than pronounced-letter omissions (“SCARELY” for “scarcely”; Frith Reference Frith1978). This shows the influence of implicit morphological knowledge. Silent letters often reflect morphological roots (Latin “scissor” for “cutter”). Similarly, 8-year-olds recall more words on the basis of a silent-letter prompt (“C” for “SCISSORS”) than a pronounced-letter prompt (“C” for “SCRIBBLE”; Ehri & Wilce Reference Ehri and Wilce1982). Just as for skilled adults, the task in hand matters for the strategy that is employed. Letters are treated differently by the cognitive system, depending on the specific linguistic environment created by the (experimental) task.
Data like these provide a wealth of evidence supporting Frost's argument that models of visual word recognition must reflect the basic fact that words have phonological, semantic, and morphological characteristics as well as orthographic characteristics. The errors made by young learners of English show that children learning to recognize visual word forms seek linguistic coherence throughout the skill acquisition process. A universal model of reading hence needs to take account of developmental data as well as cross-linguistic data. Even though TLP is found in skilled readers of English, data from English learners support Frost's theoretical view. Letter-order insensitivity is an idiosyncratic strategy used by skilled readers to optimize encoding resources in certain linguistic environments.