![Sample our Behavioral Sciences journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/110801/TOC-Subject-Sample-2021-Behavioral-Sciences.jpg"})
ABSTRACT
This research describes vowel letter dyslexia, a new type of dyslexia characterized by impaired reading of vowel letters. We report a multiple case study of 23 Hebrew readers with vowel letter dyslexia (1 acquired, 22 developmental). They made vowel-letter migrations, omissions, and additions in reading, with significantly fewer errors on consonants, and without vowel errors in speech production. Based on 24 tests that examined various components and the analysis of errors in reading 33,483 words, we ruled out deficits in the orthographic-visual analysis and phonological-output stages, as well as visual, morphological, and auditory deficits. We concluded that vowel letter dyslexia results from a selective deficit in a vowel-letter tier in the sublexical route. Indeed, vowel errors occurred predominantly when the participants read via the sublexical route. Thus, vowel letter dyslexia provides cognitive neuropsychological evidence for the separate processing of vowels and consonants in the sublexical reading route.
Acknowledgements
We thank Max Coltheart for the inspiration for this research, for asking some difficult questions that we had to solve, and for suggesting some of the tests in this study. We are deeply grateful to Aviah Gvion with whom we walked the first steps into acknowledging that vowel dyslexia exists and into the beginning of the journey to understand it. We thank Dror Dotan, Maya Yachini, Hadas Nahmias, and Yair Lakretz for their helpful comments, and Gali Ellenblum for referring R.O. to us.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 We use the term “vowel dyslexia” as a short for “dyslexia for vowel letters” — namely, a dyslexia that affects vowel letters rather than vowel phonemes.
2 In reading familiar words via the lexical route, this orthographic under-representation of vowels does not pose any special problem—the word entry in the orthographic input lexicon activates the entry in the phonological output lexicon, where vowels are fully specified, and the word is pronounced correctly. The difficulty starts when reading via the sublexical route—in this case, when consonants appear consecutively without a vowel between them, the reader can either rely on the morphological structure of the word, which provides information about the vowels in the word, or insert a vowel (probably in the phonological output buffer), and receive support from the phonological output lexicon, if such word exists there.
3 Hebrew also has a writing system with diacritical marks, which provides almost full information about the vowelling of the word (and about the conversion of the four ambiguous consonant letters). Within this diacritical system, vowels are signified by points and short strokes placed below, within, and above the printed letter. However, the use of diacritical marks in Modern Hebrew is very limited, and they appear mainly in children's books, poetry, and sacred texts (Shimron, Citation1993). Children learn to read Hebrew in first grade with diacritic marks, and these marks are abandoned in the second or third grade. After third grade Hebrew readers become less and less acquainted with the diacritic marks, and many adult readers are not fully familiar with this diacritic system and cannot use it for accurate reading (Seroussi, Citation2011). Thus, Hebrew readers after third grade read almost exclusively the orthography without the diacritics, and this is the system we used in this study.
4 These 23 participants were the first 23 participants that we identified with vowel dyslexia. Since then we continued testing, and by now we have seen 202 Hebrew-speaking individuals who show this pattern: a selective impairment in the sublexical route that affects reading vowel letters when reading sublexically.
5 The age of our sixth-grader participant (Y.X.) was like the age of the seventh-graders in the control group, so we present his comparisons to this control group. We also did all the comparisons of his reading to the fifth-grader control participants, and all the results remained the same—all comparisons that were significant compared to the seventh-graders were also significant in the comparison to the fifth-graders. We also compared the performance of the fourth-grader (E.M.) to a subgroup of the fourth–fifth-grade control group that only included fourth-graders, and the results did not change—he made significantly more vowel errors than them. He made also significantly more vowel errors than a control group of 32 second-graders in the screening test, the migratable word test, and the additional word pair reading test.
6 We excluded participants from this study if their vowel errors could be fully explained by a deficit in the orthographic–visual analyser, in letter position encoding, letter-to-word binding, or letter identity (or a combination of these). When a participant made more vowel errors than the matched controls, but all of these errors were vowel position errors, and she or he also made consonant–consonant position errors, a letter position encoding deficit could explain the vowel errors, and this participant was excluded from the study. When a participant made more vowel errors than the matched controls, but all of these errors were vowel migrations between words, and she or he also made consonant migrations between words, a letter-to-word binding deficit could explain the vowel errors, and this participant was excluded from the study; similarly, participants who made more vowel identity errors than the controls but also made more consonant identity errors, and did not make other vowel errors, were excluded; see Appendix 1 for details.
7 There are very few words in Hebrew in which a migration of two adjacent vowel letters or of two vowel letters across a consonant letter creates another word, so we could not analyse vowel–vowel migrations.
8 To be on the safe side, this analysis excluded migrations (both consonant-consonant [CC] migrations and migrations that involved a vowel and a consonant [VC migrations]) for participants with letter position dyslexia, it excluded between-word migrations (of both V and C) for participants with attentional dyslexia, and, like all other analyses, it excluded surface errors altogether.
9 VC migrations were analysed out of items that included middle V and C, and CC migrations out of items including at least two middle Cs; as we explain above in 3.2 point 6, the analysis of existing words included only words in which these migration types could create another existing word.
10 We coded the various errors on vowels according to the following guidelines: ה and א: correct conversion to vowels—a, e; vowel substitution—i, o, u; ו: correct conversion to vowels—o, u; vowel substitution—a, e, i; ו : correct conversion to vowels—i; vowel substitution—a, e, o, u. When no vowel was present (two consecutive consonants): correct conversion to vowels—a, e, or no vowel; pronouncing o, i, u in these cases was coded as letter addition. Reading no vowel when one of the vowel letters was present was coded as vowel letter omission. These guidelines held for both nonwords and words (“correct conversion to vowels” constituted the phonologically plausible conversions, which in this case indicated no vowel error, but could include surface errors in irregular words).
As reported in Section 3.2 point 6, we analysed the rate of each error type out of the number of words in which an error of that type creates an existing word. For example, the list of 100 words in Section 4.2.1 included 72 words with a lexical potential for vowel migration, 88 for vowel omission, 74 for vowel substitution, and 35 words for vowel addition. Therefore, vowel migrations were calculated out of 72 words, omissions out of 88 words, and so on.
11 Our analysis showed that vowel migrations were real migrations and did not result from an accidental omission of a vowel letter and an addition of the same vowel letter in a different position within the word. If this were the case, we would expect to see a general pattern of many error responses that involve both a vowel addition and a vowel omission, and approximately 25% of these should be omission and addition of the same letter, given that there are 4 vowel letters in Hebrew. The results, however, show that vowel migrations (the suggested omission+addition of the same letter errors) were the decisive majority of the errors that involved both an omission and an addition of a vowel letter. Out of 576 cases in the reading of all participants together that could be interpreted as vowel omission+addition, 562 responses (98%) were vowel migrations (i.e., omission of a letter and addition of the same letter in a different position), and only 14 errors involved omission and addition of different vowel letters within the same word. The same result was observed also when analysing only the 12 participants who had vowel dyslexia without LPD: A total of 95% (171 responses) of the responses that could be interpreted as vowel omission+addition were migrations (i.e., omission and addition of the same letter), and only 9 were errors of vowel omission+addition of different vowel letters. Hence, vowel migrations are clearly a true phenomenon and cannot be explained as coincidental mixed error of omission and addition of the same letter in a different position.
Furthermore, vowel letter migrations included both forward migrations, in which the vowel letter moved forward with respect to its position in the target word (form→from), and backward migrations (from→form). The number of forward migrations (276) and backward migrations (286) did not differ significantly, t(22) = 0.31, p = .76.
As for addition errors, there was a total of 342 vowel letter additions in word reading. Of these, only 27% (93 additions) were of letters that existed in the target word; given that there are 4 vowel letters, it seems that vowel additions were really additions rather than perseverations.
12 Further support for the conclusion that the vowel dyslexia of our participants did not result from an orthographic–visual analyser impairment comes from their vowel errors being insensitive to letter form. In Hebrew, 5 letters have different forms in word-final and non-final positions. Earlier studies indicated that individuals with orthographic–visual analyser dyslexias do not make errors that cause a letter-form change (letter position dyslexia, Friedmann & Gvion, Citation2005; neglect dyslexia, Friedmann & Gvion, Citation2005). We did not create a word list a priori to assess whether letter form affects vowel errors in our participants. However, a few of the words in the lists were such that a vowel error would cause a change in letter form (for example:ריף→ריפא, RIFA→RIF, where vowel omission caused F to change from non-final to final form; שטפו→שטוף, STOF→STFO, where vowel migration caused F to change from final to non-final form). Eight of our participants made errors that changed letter forms in these words, unlike in orthographic–visual analyser dyslexias. This finding (although requiring replications with tests built for that purpose) further places their source of vowel deficit in a stage that is no longer sensitive to letter forms, later than the orthographic–visual analyser. It is interesting to note that of the 17 LPD participants analysed in Friedmann and Gvion (Citation2005), a single LPD participant did make 5 letter-form-changing migrations. In retrospect, when we analyse this participant's error pattern, we see that she actually had vowel letter dyslexia.
13 This pattern also holds when we look separately at letter identity errors (the participants with pure vowel dyslexia and intact lexical route make vowel identity errors only on nonwords; the other participants make more vowel identity errors on nonwords than on words, Appendix 5). The same holds for vowel migrations (Appendix 3).
14 Support from the orthographic input lexicon cannot account for the better reading of words than nonwords, because the words had lexical neighbours differing in a vowel letter, so support from the lexicon could have caused a vowel error in words just like it could result in vowel errors in nonwords.
15 We can safely conclude that the participants who made surface errors in reading indeed had surface dyslexia, and did not make vowel errors that we misinterpreted as surface errors: Firstly, they made errors of phonologically plausible yet lexically incorrect conversion also of the consonant letters that can be ambiguously converted. Secondly, they made errors in reading potentiophones (and understanding homophones) that differ only in homophonic consonant letters and have the exact same vowel letters.
16 Steiger's Z (Steiger, Citation1980) analysis indicates that the correlation between pv × ps and vowel error rate in word reading is significantly higher than the correlation between only pv and this rate, ZH = 3.53, p < .001.
17 As to vowel substitutions, whereas some vowel letters can be converted to two vowels, they can only be converted to these two, and not to others. Hence, reading the vowel letter as one of the other vowels, which is not a phonologically plausible conversion, would count as a vowel substitution (such as reading the letter ו, instead of /o/ or /u/, as /i/). Vowel omissions and additions can be treated the same way—whereas /a/ and /e/ in the middle of the word are usually not written, the other vowels are encoded with a letter, and, hence, omitting or adding them is an impossible conversion, and, hence, a vowel error.
18 An interesting question is what would happen in languages in which the conversion of some consonants depends on the identity of the following vowel. For example, in Italian and Spanish (and also in English), the pronunciation of the letters C and G depends on the following vowel. A question arises as to whether individuals with vowel dyslexia would have difficulties correctly converting consonants that depend on adjacent vowels when reading nonwords. This would probably depend on whether these vowel letters are part of a digraph to be converted as a unit, or whether they are separate from the preceding consonant and are part of the (impaired) vowel tier, in which case consonant conversion would be affected. This further raises the issue of whether multiletter graphemes are part of the C and V tiers or whether there is a separate tier for multiletter graphemes. [It might also be that multiletter units that correspond to a single consonant (sh) or vowel (ea) phoneme are part of the corresponding tiers, whereas mixed multiletter units (like ough) are stored in a separate tier.]
19 We excluded from this analysis N.T. and R.X., who only made errors in reading nonwords.
20 This analysis was made for the data collected from 15 participants with LPD and without vowel dyslexia reported in Keidar and Friedmann (Citation2011), 14 participants reported in Faran, Gvion, and Friedmann (Citation2013), 11 participants reported in Friedmann and Gvion (Citation2012), and 8 participants collected by Salner, Friedmann, and Chajut (Citation2013). The locus of impairment of each of these LPD participants was in the letter encoding function in the orthographic–visual analyser: Each of them made transpositions in both oral reading and silent reading tasks, and did not make vowel errors other than transpositions; they did not make more vowel letter omissions, additions, or substitutions than age-matched controls.
21 The analysis (summarized in ) includes data for all the participants who had vowel dyslexia and LPD and who participated in tasks that examined the orthographic–visual analyser without involving the sublexical route. For N.T., who had letter position encoding impairment in the orthographic–visual analyser but no impairment in the lexical route, we could analyse CC versus VC migrations in reading aloud of words, as this task, for him, involved the orthographic–visual analyser but not the sublexical route. His reading aloud of words, which was affected by his LPD but not by his vowel letter dyslexia, showed more CC migrations than VC migrations. For the other participants with LPD+vowel letter dyslexia, we examined CC versus VC migrations in the same–different task, which examines the visual analyser directly, without involving the sublexical route. For the three participants who did not do the same–different task (but who had surface dyslexia in the connection between the orthographic and phonological lexicons and hence showed good comprehension of homophones), we used migratable word comprehension tasks (from the TILTAN test battery, Friedmann & Gvion, Citation2003), which would be affected by a letter position encoding deficit in the orthographic–visual analyser but did not require reading aloud through the sublexical route. The migratable word association task is described in Lukov et al. (Citation2015); the migratable word–picture task is described in Keidar and Friedmann (Citation2011).