Phonological Precedence in Dyslexia: A Case Study

Abstract

Developmental dyslexia is believed to involve a phonological deficit of which the exact properties have not been clearly established. This article presents the findings of a longitudinal case study that suggest that, at least for some people with dyslexia, the fundamental problem involves a disturbance of temporal-spatial ordering abilities. A very frequent reading error of the case is to misread CVC syllables in word internal position as CV syllables instead. Specifically, there is a phonological conspiracy in her reading such that she avoids (word internal) codas and maps them as something else via epenthesis, deletions, metathesis, etc. It is noted that this overproduction of CV units evinces a problem with the temporal-spatial ordering of salient items (vowels) versus nonsalient items (consonants). This can be interpreted from a Minimalist perspective as the reflex of difficulties in linearizing phonological material at the sensorimotor (SM) interface. The analysis is couched in terms of the coupling hypothesis of syllable structure and Optimality Theory.

ACKNOWLEDGMENTS

I am indebted to Fusa Katada for her comments and discussions concerning this work. I am deeply appreciative of Jean-Roger Vergnaud, in memory, for his thought-provoking comments and discussions about clocks and linear precedence in grammar. I am also grateful to Nancy Hedberg, Greg Hickok, Nicole Nelson, Associate Editor Heather Goad, and four anonymous reviewers for their valuable comments, insights, criticisms, and discussions. I thank Allyson Walker for her careful transcriptions, attention to detail, and patience in checking and rechecking transcriptions. Finally, I acknowledge the generous assistance of Bridget A. E. Richter for her vigilant and careful formatting of references in this text. Errors and inadequacies in this work are entirely my own.

Notes

¹In this literature, phonological can refer to various levels of phonological representation, from that of the syllable, to the division of onset-rhyme, to the unit phoneme, and awareness refers to the ability to manipulate, discriminate, and/or identify phonological representation of various types.

²In a RAN task, a subject is presented with a list of digits, letters of the alphabet, or pictures of common objects and is required to name them as rapidly as possible.

³Vision therapy was prescribed and Tara participated in the prescribed number of therapies.

⁴She was diagnosed with dyslexia at this time. This diagnosis was reconfirmed at age 19;05 by an independent psychologist.

⁵These tests were repeated again when she was 12 years and 7 days old. The second round of scores was very similar to the first round of testing.

⁶At retest (12;00,07); in the first round of testing she was in the 63rd percentile for both tests.

⁷Although Tara falls in the average range for expressive knowledge of word opposites, she self-reported great difficulty with expressive use and receptive comprehension for temporal and directional words such as before/after and above/below.

⁸It is possible to use a visual route to decoding as well. People with dyslexia who rely on a purely visual route read via knowledge of spelling patterns. They have difficulties reading words that have irregular pronunciation. For example, they might read yacht as [jækt].

⁹She scored slightly higher in rapid naming on the earlier test (11;10). At age 12 years, there is a different norm reference for the raw scores so the percentiles changes even with similar raw scores.

¹⁰Tara's knowledge of her second language would not have been adequate to pass a university-level proficiency test.

¹¹However, an anonymous reviewer points out that the Incomplete Words subset of the WJC might not really be a test of phoneme awareness although it is typically thought of this way. The reviewer notes that the Incomplete Words subtest lacks a metaphonological component (unlike phoneme counting or phoneme deletion tests) and instead might really be measuring ability to retrieve “the phonological representation of a word in the presence of a degraded stimulus.” In light of this consideration, particularly given Tara's good score on Sound Blending, it remains unclear how weak Tara's phonemic awareness is as an adult.

¹²Although I will ultimately argue that the data indicate a problem with mapping to the SM interface, rather than a problem of phonological awareness, it must be noted that most of the data that is considered comes from Tara's reading, and reading is a task that requires phonological awareness. Therefore, as noted by an anonymous reviewer, caution is required in drawing conclusions about phonological representation based on data from such a task.

¹³For some nonwords, the vowels had more than one possible pronunciation, particularly when the vowel is reduced. A vowel pronunciation counted as the wrong quality if it was not plausible.

¹⁴For a number of nonwords, stress could have plausibly fallen on either the penultimate or antepenultimate syllable of the word.

¹⁵Suppose substitutions were not structure preserving in the sense discussed in the text. In that case, it would look like a vowel were deleted and a consonant inserted in the same position in the word. There were no reading errors that looked like that.

¹⁶It is still possible that this error is not an inversion but rather a partial word substitution as a familiar spelling pattern from a word like hospital or capital. If that is the case, reversals/inversions are even rarer for Tara than asserted in the main text.

¹⁷There are a few examples where it is difficult to code the reading error because it is consistent with more than one analysis—for example, the error could be due to either metathesis or insertion. In that case, the error is coded as the most common process on the assumption that a process is more common because it is more parsimonious. Therefore, since metathesis is a relatively uncommon process, it is not coded when an analysis of a different (commoner) process is possible.

¹⁸Indeed, accommodation facts show graphemes are the input to the operations involved in syllable repair such as deletion and metathesis, etc., even when syllable structure repair is not relevant.

¹⁹Here we see that word internal codas are possible. However, in these examples, the spelling patterns are quite familiar and Tara immediately associates them with the words they resemble, appearing to read them as compound words. In fact, the patterns are so familiar that even a reader without dyslexia has to pause to notice in these cases that the words are in fact nonwords. Recall that the patterns of reading errors that we discuss here are most relevant to words that have rare spelling patterns, as otherwise Tara can simply exploit her knowledge of the visual patterns of words instead of employing a phonological route (cf. section 3.1.1).

²⁰Not every mistake Tara made led to elimination of a word internal coda. The Appendix summarizes the other types of errors that were observed. Substitution of a different consonant for the target consonant is by far the most common type of error when syllable contact avoidance is not involved. Substitution took place for several reasons. Tara often read the letter c and g wrong. She read c as [s], the less frequent pronunciation of c but the value that is indicated in the name of the letter [si]. She also sometimes read g as [ʤ], also the less frequent pronunciation of g but the one indicated by the name of the letter [ʤi]. Finally, she read ch with its rarer pronunciation—the value it would have in words from the Greek—namely [k] instead of the more expected [ʧ]. In addition to phonics motivated substitutions, some consonant substitutions had obvious phonological motivations such as assimilation in voicing or place to surrounding sounds or due to perseveration of or anticipation of a sound elsewhere in the word. There were nine examples like this, which accounts for 52.9% of the substitution errors we are considering here. There were three other consonant substitution errors. They did not have an obvious motivation in the data. My focus is on Tara's coda related problem, and therefore I do not further discuss this particular data here.

²¹Tara does not always eliminate anything that would be a coda in the output. This is especially true of highly sonorant codas. I will discuss sonorant+obstruent syllable contact in some detail shortly.

²²There were three additional word substitutions. Two of them overcame a problem I have not discussed yet: vowel hiatus *V.V. The substitutions were nave for naïve and nativity for naiveté. The only exception to the word substitution being motivated for some reason of syllable contact was adequate for aquatic, where there were no syllable contact problems in either word.

²³Tara's pronunciation of glides that were in onset position in the input was also unexpected. To limit the scope of this article, I set aside an investigation of behavior that was unique to glides.

²⁴On a cautionary note, closer examination of the dyslexia data indicates that simply embracing markedness as very highly ranked might not account for all the data or make the correct predictions about Tara's reading. Here is one problem I tentatively identify under such an approach. Kaye & Lowenstamm (1981) propose an implicational universal between the two possibilities CCV and CVC, such that if a language allows for a complex onset, it must also allow for a coda. That is, if a language allows for CCV syllables, then it must also allow for CVC syllables, but not the reverse. In other words, CCV syllables are marked with respect to CVC syllables. As we saw in sections 3.2.4, and specifically, 3.2.4.7, Tara's reading clearly does not adhere to this implicational universal in that it allows the presumably more marked CCV form to occur, but not the less marked CVC form in word internal position. This seems to defy markedness generalizations concerning syllable structure when child language is involved. Therefore, it might not be that markedness per se defines Tara's reading. However, I have not been able to find sufficient data to clarify if this universal really holds for word internal syllables (although it clearly holds for word margins) as I looked through the child language data. This is because most of the words in databases tend to be short and therefore have too few syllables, and in particular, too few word internal syllables to throw any light on this issue.

²⁵Pater's *N constraint says that a no nasal plus voiceless obstruent sequence can exist.

²⁶/l/ also participated in metathesis as the following example illustrates where Tara pronounces the word segilcap (expected pronunciation: [ˈsɛɡɪɫkæp]) as [ˈsɛlɪkæp]. In this example, the /l/ has moved into prevocalic position.

²⁷Tara (age 21;11) was asked to judge the accuracy of someone else reading nonwords with the types of syllable contact that she always avoids in her own reading. This was to determine if her own production was somehow implicated in the problem, as this observation would not depend on her production, but instead on her auditory perception of another reader and her own decoding. Tara was able to look at a list of the words that were being read while listening. The list was an abbreviated list of the same nonwords she had read at an earlier time. The misread words were all misread by applying the various strategies implicated in the CV conspiracy. Tara was unable to accurately judge whether the reader pronounced a word inaccurately or not.

²⁸He notes that an example of an automatic physical effect is the Bernoulli effect. He suggests one possibility is a pressure to minimize opening and closing gestures of the jaw, where CV (an opening) has fewer gestures than VC (an opening and closing). And therefore, CV would be the automatic ordering of {CV} due to Vergnaud's proposed minimizing of gestures effect.

²⁹Anti-phase coupling is inherently less stable in that it can easily be disrupted. Upon disruption it goes into in-phase coupling. The reader can easily convince him- or herself of this by tapping against a surface the index finger on his/her right hand in anti-phase to the tapping of his/her index finger on the left hand against the same surface. If this tapping is speeded up sufficiently, the finger tapping will spontaneously fall into in-phase coupling.

³⁰Louis Goldstein suggests (personal communication, May 1, 2011) that the less stable anti-phase coupling of the coda to the vowel might not be the real problem. Rather knowing how to use asynchronous clocks to trigger CV in the right order might be the problem or perhaps the clock settling time is the relevant point. He believes that experimental evidence can be interpreted as showing that it takes longer to stabilize the clocks in the anti-phase mode and suggests that perhaps the dyslexic has some cognitive limitation that prevents him/her from having enough time to do so. This seems to me to be a very promising direction for future research and might prove relevant to linking naming deficits to phonological deficits.

³¹An anonymous reviewer suggests that the presence of word final Cs and word initial Vs in Tara's reading is consistent with Lowenstamm's (1996, 1999) CV/government phonology. CV phonology would also claim that vowel-initial words begin with an empty C and that C-final words end with an empty V. However, it is not completely clear that Tara's reading supports CV phonology given that she appears to have no problems with complex onsets, but Lowenstamm's framework would require an empty nucleus to interrupt the consonants in a cluster. A clearer understanding of this issue awaits a more systematic exploration of complex onsets in Tara's reading.

³²I assume, following Ramus et al. (2010), that the lexicon contains three levels of representation: a lexical phonological level, a lexical orthographic representation, and a lexical semantic representation.

³³Tara was again exposed to the rules of Pig Latin and appropriate examples a second time where she was able to produce some words that did follow the rules of Pig Latin, but she also still produced some Pig Latin words that had the same pattern illustrated above.

³⁴However, as pointed out by an anonymous reviewer, since Tara is able to delete the onset in the Pig Latin task, there is something unusual in her inability to use the onset in the syllable added at the end of the rhyme unit. It is possible that these tasks require different abilities to some extent: “decommissioning” the onset, versus “redeploying” it.

³⁵That is, they do not treat a vowel immediately followed by its coda as being closely connected.

³⁶As discussed earlier, Goad & Brannen (2003) conclude there is actually an intermittent CV+CV stage, with the final V being silent.

³⁷Given proposals about the status of final Cs in early language acquisition as possible onsets of silent vowels, I ignore word final C.

³⁸As pointed out earlier, the literature appears to have nothing to say specifically about early acquisition involving words of more than two syllables. Moreover, there appears to be few data bearing on the issue in the corpus I examined (Smith 1973; MacWhinney 2000). Establishment of accurate generalizations concerning word internal syllable development in early language acquisition would clearly be a worthwhile goal for future research. I note the fact that the words I examined for Tara's reading tended to be longer than two syllables.