*

Temporal-Spatial Sequencing in Prosodic Development: The Case of Dyslexia Fusa Katada and Patricia Schneider-Zioga Waseda University/MIT and California State University, Fullerton

1. Introduction The reading disorder called developmental dyslexia is believed to involve deficits in phonological awareness, leading to weak phonological representations of words in the mental lexicon (Goswami 2003). We note a previously unidentified characteristic of dyslexia, namely inaccurate reading marked by an overproduction of CV units in the absence of VC units. We argue that the core problem lies in a general disturbance of temporal-spatial ordering abilities, evinced in atypical prosodic development. We suggest that the atypical prosodic development leads the person with dyslexia to be not able to exploit the unit rhyme (VC in CVC) in reading. This creates a rather severe problem for English because the pronunciation of vowel graphemes in English is opaque, but most predictable, when the context of the coda grapheme(s) is taken into account (Treiman et al 1995). To account for heavy reliance on CV units in dyslexia, we adopt the coupled oscillator planning model of syllable structure (Goldstein et al. 2006, Goldstein, Nam, Saltzman & Chitoran 2008, Nam & Salzman 2003, Saltzman & Byrd 2000). This model accounts for syllable structure in terms of modes of coupling vocalic and consonantal gestures which are each associated with a planning oscillator (clock). We extend this model to the mechanisms of foot construction and propose that, for people with dyslexia, there is a disturbance in the inherently unstable anti-phase coupling, while there is no disturbance in the inherently stable in-phase coupling. Finally, we ask why dyslexia is usually not apparent in the spontaneous speech of people with dyslexia. We note that there is a silent phonology, implicated whenever we speak to ourselves “in our head.” This speech clearly requires linear ordering as there are silent copies and ellipsis. We note that effective reading crucially requires silent phonology. We propose that silent phonology is disordered in dyslexia.

2. The role of the unit rhyme in reading in English The statistical analyses carried out by Kessler & Treiman (1997) reveal a strong tendency of English orthography to be rhyme-based. That is, the coda, and not the onset, provides useful information about the possible pronunciation of the vowel grapheme. Treiman et al. (1995) documents VC as a more predictable environment for orthographic pronunciation than CV. For example, the digraph ea is almost always pronounced [ɛ] before d (the orthographic rhyme ead as in br.ead, tr.ead, h.ead, and d.ead), and as [i] before p (the orthographic rhyme eap as in h.eap, l.eap, and ch.eap). On the other hand, onset graphemes do not offer a clue concerning the potential pronunciation of a vowel as we can see by the varying pronunciation of the vowel digraph when the onset is held the same, but the coda varies: cf. hea.d vs. hea.p. ____________________________ * This work has gone through various stages of its development. We thank audiences at the following meetings for their valuable comments and suggestions: the 78th Annual Meeting of the Linguistic Society of America (2004), the Fourth International Conference on the Mental Lexicon (2004), the 80th Annual Meeting of the Linguistic Society of America (2006), CUNY Phonology Forum (2007), and the 33rd Annual Meeting of the Berkley Linguistics Society 33 (2007). The usual disclaimers apply.

An important finding in this respect is that of Goswami (1999, 2001), who argues that beginning readers strongly benefit from exploiting the unit rhyme to read analogically. Other studies such as Wise, Olson, and Treiman (1990) confirm this position. Moreover, Duncan et al. (1997) finds that poor rhymers were poor readers and good rhymers were good readers. These facts indicate that in order to successfully read English, not only must one know the various possible mappings of the vowel grapheme to its appropriate phoneme, but, more importantly, the reader must exploit the unit rhyme in the process in order to be able to take into account the maximally informative context for accurate pronunciation of the vowel grapheme. In orthographically transparent languages such as German, it is not necessary to exploit the unit rhyme in reading that language since the mapping between grapheme and phoneme is consistent.

3. Brief profile of our case and the method of experiments We undertook a case study of an English-speaking girl with dyslexia, henceforth called Tara, a pseudonym, born in 1988. Her language development seemed completely normal in her spontaneous spoken language. However, concerned about her slow acquisition of reading and spelling, Tara’s mother had her tested on a number of psychometric tests at relevant stages in her development. She demonstrated well above average language abilities in the Peabody Picture Vocabulary Test and in forward digit repetition. Nonetheless, she looked predictably atypical in the language measure of rapid naming, a problem which consistently correlates with dyslexia. The experiments were conducted between March-May 2002 when Tara was 14 years old. Our observations are concerned with words that Tara does not already know how to read. The words she reads for these observations are nonsense words, or Latin or vocabulary items that are at a very advanced reading level. Since these words are not in Tara’s reading vocabulary, she must use a phonological route to arrive at the correct pronunciation. All words were presented to her as individual words on a reading list, and therefore she could not use contextual clues to arrive at a plausible pronunciation.

4. CV as a fundamental unit in dyslexia We demonstrate a striking contrast in Tara’s reading: the lack of codas versus presence of onsets. We show that her reading can be characterized as involving a decided lack of VC units and an overproduction of CV units, leading to highly inefficient decoding.

4.1. Observation 1: The avoidance of codas Our first observation is that Tara avoids pronouncing codas in all positions except word finally. In all of the examples in (1), Tara pronounces the codas as onsets. She achieves this either by epenthesizing a vowel immediately after a word internal coda (1a, b, c, and d), or by metathesizing a vowel from a nearby vowel digraph so that a vowel follows the coda (1e, f, g), or by simply deleting a word-internal coda (1h). (1) Avoidance of coda via epenthesis, metathesis, or deletion presented words Tara’s pronunciation process a regatleb [ri.ˈgɑ.t.lɛb] epenthesis of [] b bailesbud [be.ˈli.sɪ.bəd] epenthesis of [ɪ] c besaubgant [bi.ˈsɑ.bɪ.gænt epenthesis of [ɪ] d aipcid [ˈe.pɪ.sɪd] epenthesis of [ɪ] e maudlin [ˈmɑ.du.lɪn] metathesis of u and d f leegtab [le.ˈgɑ.tə.bə] metathesis of e and g g weipgan [ˈwi.pɪ.gæn] metathesis of i and p h wepgad [ˈwe.gæd] deletion of p

4.2. Observation 2: The presence of onset including complex onsets By contrast, onsets, including word-internal complex onsets, are intact for Tara. In the examples in (2), we see that if the orthographic sequence CC occurs word-internally, several things can happen. Most commonly, if the sequence is a possible onset according to the phonotactics of English, it was pronounceable as an onset. In these cases the word internal sequences: st, sp, and tr are mapped as complex onsets and the words are read accurately. (2) The presence of onsets including complex onsets presented words Tara’s pronunciation a feestary [ˈfi.stə.nɑ.ri] b reastal [ˈri.stəl] c toispar [ˈtoɪ.spɑr] d degatrab [di.ˈgæ.træb] e roipsar [ˈroɪ.spɑr] f tobrawnest [tobrɑwənɛs] g pelaustible [plɑ.sɪ.tə.bl̩ ]

complex onsets st st sp tr (ps → sp) br st → sɪtə

Example (2e) is noteworthy. The word internal sequence is ps. According to the phonotactics of English, ps cannot be interpreted as a complex onset. The only way to interpret this is as a coda of the first syllable followed by an onset of the second syllable. However, Tara does not allow codas and the observed prohibition on word-internal codas must be satisfied. This prohibition could be observed via epenthesis of a vowel; that is, a […C1C2…] sequence would be pronounced as […C1V.C2…]. However, there is another possibility which Tara could and does exploit for this example: the sequence C1C2 [ps] can be metathesized to C2C1 [sp] and form a licit complex onset [ps → sp ]. We see then that Tara’s knowledge of English onset phonotactics is intact. Example (2f) has two word-internal sequences in the orthographic input: br and wn. Tara pronounced the first sequence br as a complex onset in accordance with the phonotactics of English. The second sequence: wn, which could not be interpreted as a complex onset, was pronounced with an epenthetic vowel: wə.n in order to avoid pronouncing a coda. The final example (2g) illustrates that a licit complex onset, the sequence st, can also map as CV.CV. This example demonstrates that the reorganization of syllables is not only driven by a need to repair syllables that violate Tara’s phonotactic constraints.

4.3. Observation 3: The retention of VV sequences: diphthongs and tense vowels Tara has no trouble with diphthongs, VV sequences which occur syllable-internally. As examples (2c) and (2e) in the previous section show, she properly pronounces diphthongs [oɪ]. Moreover noted are numerous examples of tense vowels, which Tara pronounces accurately. The examples are [i] as in ‘be’ or ‘bee’ and [u] as in ‘food’. Tense vowels are analyzed as VV that is syllable-internal, and classified as forming a natural class with diphthongs for our present purpose.

4.4. Summary of Tara’s syllables We see that Tara can have syllables that are more complex than the unmarked CV syllable; however, essentially only complex onsets are possible. Codas appear only in word final position. The retention of syllable-internal vowels, as well as tense vowels, is also noted. In short, we introduced and underscored the fact that Tara’s reading errors reveal her reliance on CV as a unit of sequential organization of the melody of the words she reads. In the following sections we consider the source of her reliance on this unit.

(3) Tara’s syllables syllable type CV.CV CVC CVV

possibility possible impossible possible

remarks presence of onsets/complex onsets absence of codas (possible word finally) presence of diphthongs/tense vowels

5. Difficulties with rhythmic awareness in dyslexia Thomson et al (2006) and Goswami et al (2002) report that children and adults with developmental dyslexia have difficulties with auditory cues to speech rhythm and stress. Both expressive and receptive abilities seem affected. Also non-linguistic areas are affected such as manual tapping to a metronome, and perception of amplitude modulating tones as having a beat or not. It is known that rhythmic periodicity in speech is related to the onset of vowels in stressed syllables (i.e., the onset of vowels constitutes a P-center cue, where P stands for perceptual as in, for example, perceptually aligning beats of a metronome to syllables). Researchers suggest that difficulty with rhythmic awareness affects the ability to gain access to segmental organization within syllables. We develop this idea further. We point out that our analysis can account for the fact that this is a co-morbid finding in dyslexia. We hypothesize that the temporal-spatial ordering problem that is suggested by the difficulties with rhythmic awareness has a metrical correlate. This is because the task of the metrical component of the grammar is to temporally organize grammatical units. In the rest of the paper we propose that Tara’s overproduction of CV units is the reflex of a problem with the temporal ordering of vocalic items in the construction of feet.

6. The temporal-spatial ordering decision in syllable and foot structure 6.1. Syllable structure: VC versus CV as a constituent In a CVC syllable, the most widespread assumption since Hockett (1955) is that the nucleus and following coda (VC) form a constituent, the rhyme, as in (4a). However, the dyslexia data show that in fact for Tara, CV, and not VC, seems to form a tight unit, and hence functions as a constituent. With regard to CV as a constituent, we note the relevance of the findings of articulatory phonology in Browman & Goldstein (2000) as discussed and developed in Vergnaud (2007), that the onset of a syllable is a co-articulation on the vowel. We also note the fundamental idea of Lowenstamm’s (1996) CV only phonology, that CV is the only available syllable type, despite any surface appearances to the contrary. The observation that the onset consonant forms a tight unit with a following vowel in the Japanese mora-based system is also relevant here (cf. Kubozono 1989; Katada 1990, 2008; Poser 1990, Ito & Mester 1992, among others). This observation is based on a variety of evidence, among them the fact revealed by a Japanese language game called Shiritori (hip-slinging or tail-taking), that the (initial) moraic unit in a CV sequence is not just the vowel, but the onset C behaves as part of the mora as well (Katada 1990). A competing view of syllable structure to (4a) is then as in (4b). (4)a. VC as a constituent S / \ / R / / \ C V C

b. CV as a constituent S / \ M \ / \ \ C V C

Tara’s reading seems to conform more with (4b) than with (4a). We will develop the view that CV forms a constituent.

6.2. The coupled oscillator model of syllable structure (COMOSS) The coupled oscillator model of syllable structure (Goldstein et al. 2006, Goldstein, Nam, Saltzman & Chitoran 2008, Nam & Salzman 2003, Saltzman & Byrd 2000) accounts for syllable

structure in terms of modes of coupling vocalic and consonantal gestures which are each associated with a planning oscillator (clock). In this model onset gestures and coda gestures are coupled in different ways. Onset consonant gestures are hypothesized to be coupled in-phase (a relative phase of 0o) to the vowel and coda consonant gestures are coupled in an anti-phase (a target relative phase of 180o) pattern. The in-phase coupling is inherently stable and the anti-phase one is not. Given the idea that CV forms a unit, and in particular following Vergnaud’s (2007) implementation of Browman & Goldstein (2000), we can think of the onset consonant (C) as an autosegment of the vowel (V), where the onset is coupled in-phase to the vowel, represented by a vertical line ( | ) in (5a); along the temporal axis, C and V are represented as having no time elapse between the two. By contrast, the coda consonant is not a co-articulation of the vowel, and hence cannot as easily be integrated into the syllable. Under the coupling model illustrated above, the consonantal gesture of a coda is coupled to the preceding vowel through an inherently unstable anti-phase relation, represented by a horizontal line ( — ) in (5b); along the temporal axis, V and C are represented as having a time lag between the two. Under the above view, a C1VC2 syllable has the representation in (5c), following Vergnaud (2007). In (5c), there is essentially no ordering decision to make for the onset; the onset consonant C is necessarily part of the salient unit V. Therefore, the ordering in CV of C with respect to V is for free. This representation, on the other hand, requires a temporal ordering decision about the order of CV and C. (5) a. In-phase coupling of C to V C | V

b. Anti-phase coupling of C to V V—C

time

time

c. C1VC2 syllable C1 | V—C2 time

6.3. Moraic trochees and the role of the syllable Kager (2007) observes that the moraic trochee captures the idea that a single heavy syllable [CVC] is quantitatively and metrically equivalent to two light syllables [CV.CV] (Allen 1973; Halle & Vergnaud 1978; McCarthy 1979; Prince 1983). Intuitively, this is directly relevant to Tara’s reading since, through epenthesis, deletion, metathesis, and other phonological processes, she produces two light syllables exactly where a typical reader would produce one heavy syllable. A more insightful description of her reliance on CV units then may be that she produces moraic trochees that consist only of light syllables. In what follows we undertake this theoretical tact. In the foot typology of Hayes (1995), (6a) is the representation of a foot containing two light syllables CVCV and (6b) for a foot containing a single heavy syllable CVC. Both representations involve a bi-moraic trochee; in (6a) two moras (μs) are constructed into an F node with a syllable node σ interpolated between each μ and F; in (6b) two moras are constructed directly into a single syllable node σ and that the single syllable node is constructed into a foot. (6) Types of moraic trochees (F = foot, σ = syllable, μ = mora) a. F b. F / \ | σ σ σ | | / \ μ μ μ μ | | | | CV CV CV C The two structures in (6) make evident a difference between two light syllables (6a) and a single heavy syllable (6b). In (6a), syllable (σ) and mora (μ) are virtually identical to each other so that the node σ adds very little (if anything) to the representation of the moraic trochee containing CVCV. In

(6b), on the other hand, the syllable node (σ) plays a significant role in that it is the node which syllabifies the second mora together with the first. In other words, (6) maintains the moraic trochee properly without a need for syllables as in (7a) below, whereas (6b) requires the construct syllable to keep the second mora legitimate in the structure. This is because, for a heavy syllable (CVC), without a syllable node as in (7b), the second mora μ, a coda, is illegitimate in the structure since F is not a construct that participates in syllabification and hence is not a representation that is typical or grammatical. (7) a.

F / μ | CV

\ μ | CV

b. * F / μ | CV

\ μ | C

We thus assume that the construct syllable exists only when it is not identical in weight to the construct mora, hence (7a) for CVCV and (6b) for CVC, which are repeated below as (8a) and (8b), respectively. (8) Types of moraic trochees (F = foot, σ = syllable, μ = mora) a. F b. F / \ | μ μ σ | | / \ CV CV μ μ | | CV C

6.4. The linearization of moras within a foot The prosodic properties we view as foundational to our analysis are the following: (i) CV is the basic organizing unit of phonological melody and this basic CV unit is a mora; (ii) the construct bimoraic trochee captures the relation between a foot consisting of two light syllables, CVCV as in (8a) and a foot consisting of a single heavy syllable, CVC or CVV as in (8b). Both have a bi-moraic trochee and hence are equal in terms of weight. The difference between the two lies in that a moraic trochee consisting of two light syllables does not require the postulation of a syllable, whereas the one consisting of a single heavy syllable requires it. Conforming to these prosodic properties, we postulate that moras are organized in the following way so that a higher prosodic construct foot emerges, one with a syllable as in (8b), and the other without as in (8a): (9) Foot Construction Mechanism (FCM) (i) Every mora is a CV, where C and V are autosegments of each other. (ii) Two moras make up a potentially quantity sensitive unit called a moraic trochee: [CVCV]. (iii) This unit can be headed by a functional category f, in which case a syllable (σ) emerges: σ[f [CVCV]]; otherwise, a syllable does not emerge: [CVCV]. 1 (iv) If f is present, the final V or final C must be silent, yielding σ[f[CVCV]] or σ[f[CVCV]], where 2 silent elements are indicated by double strike through of V and C. (v) Each mora (CV), whether C or V is silent or not, is ordered with respect to the other through the coupling of vocalic gestures (cf. Schneider-Zioga & Vernaud 2009). (vi) [CVCV] is then constructed into a foot (F), yielding F[CVCV] or F[σ[f [CVCV]]], depending on the presence or absence of a functional category (f), and hence a syllable (σ).

1

This accords with Goad & Brannen (2003), who present phonetic evidence that in English, word final consonants in early stages of child language acquisition are actually syllabified as onsets in CV syllables where V is silent. 2 This is essentially along the lines of van der Hulst’s radical CV phonology.

The mechanisms of the coupling of gestures outlined in section 6.2 are directly relevant here. For (9i), CV is an autosegment formed by the stable in-phase coupling of C to V, as represented in (5a). In the system we are proposing, every unit is underlyingly a CV, hence, the anti-phase coupling of a coda C to V does not exist. Moreover, we extend this model to account for the higher metrical construct foot (F). Specifically, in (9v-vi), F emerges via the coupling of vocalic gestures, as in (10), where CV as an in-phase coupled autosegment is represented as CV and an anti-phase coupled moraic trochee CV-CV is represented as CV—CV. (10)a.

F /

\ μ μ | | C V—CV

b. F | σ / \ μ μ | | C V—CV

7. Linearization and Phonology of Dyslexia Given the Foot Construction Mechanism (FCM) in (9), Tara’s possible and impossible syllables summarized in (3) have the abstract structures in (11), which make transparent the source of Tara’s reliance on CV units. (11) a. [CV.CV](possible) F / \ μ μ | | C V—CV

b. [CVC](impossible) F | σ / \ μ μ | | C V—CV

c. [CVV](possible) F | σ / \ μ μ | | C V—CV

Assuming that these abstract structures are present in her knowledge, we see that Tara has no problem with the linearization of non-tautosyllabic moras within a foot without a syllable (11a). However, within a syllable, she apparently is not able to linearize a pronounced consonant and its autosegmentally associated silent vowel (11b), whereas she is able to successfully linearize a silent consonant and its autosegmentally associated pronounced vowel (11c). It seems then that Tara is in principle able to engage in anti-phase coupling of articulatory gestures, but is unable to successfully order a pronounced and a silent phonological element that are coupled in an anti-phase relation. This applies to the vocalic gesture as in (11b). The silence of the consonantal gesture that is part of the mora as in (11c) is irrelevant to Tara’s phonology since it does not directly participate in the coupling between moras (accomplished via anti-phase coupling of vowels) and subsequent linearization. Its sole relation is to the coupled vowel with which it forms an autosegment. Tara’s [VCV] is a subset of [CVCV] with the initial C being silent: [CVCV], in which coupling of the two pronounced Vs are done on a par as in (11a). Consonant clusters arise in the COMOSS outlined in section 6.2 via the anti-phase coupling of one consonant to another along with the simultaneous inphase coupling of all consonantal gestures in the complex onset to the vowel. Since all anti-phase coupled gestures are overt under this condition, it poses no problems for Tara. We see that Tara’s overproduction of CVs is the only way for her to ensure successful coupling and hence ordering between moras within syllables and within feet. Anti-phase couplings are inherently unstable, and we have noted that this is precisely where Tara has problems. We observe then that Tara requires all elements of this key to syllable organization to be pronounced. A reasonable conjecture then would be that not just Tara, but the dyslexic English population in general has weaknesses in executing the particular type of linear ordering involved in anti-phase coupling within a syllable.

Our analysis is consistent with the low prevalence of phonological dyslexia in the Japanese speaking population: a figure as low as two percent (Ueno 2006), as opposed to ten to fifteen percent in the English speaking population. Reading in mora-based, non-rhyme oriented languages such as Japanese does not require the person, whether dyslexic or not, to exploit the unit rhyme, and hence phonological dyslexia simply does not surface.

8. Dyslexia as selectively impaired silent phonology The atypicality of Tara’s phonology is not evident in her spontaneous spoken language Why then do the atypicalities manifest themselves only in her reading? First we note that the primary mature mode of reading is “reading to oneself.” This is a subtype of what we can call self-directed or internalized speech. This type of speech is silent to the outside world but audible to the speaker. It appears to have all of the characteristics of externalized speech, including extremely silent elliptical phrases (J.-R. Vergnaud, p.c.). It seems very difficult to “silently” read without hearing at least most of the sentence(s) that we are reading actually pronounced in our mind. If a subjective attempt is made to completely suppress this activity, reading seems impossible. The experience is like scanning a collection of letters with no sense related to them whatsoever. That is, “silent” reading seems to require internalized speech, which appears to have a phonology identical to that of externalized speech even to the point of requiring linearization. In addition, reading necessarily requires the understanding that silent symbols (graphemes) stand for pronounced sound and that they stand for externally inaudible yet internally pronounced sound. We conjecture that in dyslexia it is the silently pronounced language whose phonology is affected, while the externalized phonology remains intact. We recognize that “reading aloud” is also affected by dyslexia, and indeed it is the atypical characteristics of reading aloud that allow someone other than the reader to notice the dyslexia. We conjecture that reading aloud is also mediated by the silent, internalized phonology via the required use of the silent symbols, we call graphemes. At this point in time, we do not have anything more concrete to offer on this aspect of dyslexia beyond indicating a direction for further research regarding this selective impairment of language.

9. Conclusion Developmental dyslexia as a phonologically-based disorder has been primarily explored in the literature as a deficit in phonemic awareness. Here we have attempted to establish that the core problem does not lie in phonemic awareness. Instead, we have argued that its source is temporal-spatial ordering problems which find a particularly strong reflex in prosodic development. The starting point for our theoretical account of dyslexia was the coupled oscillator planning model of syllable structure (COMOSS) as developed by Vergnaud (2007) and Schneider-Zioga & Vergnaud (2009) who propose that the COMOSS can be generalized to account for foot in addition to syllable, and who argue for a planning as well as execution level of the phonology relevant to COMOSS. We have presented a possible implementation of this model which illumines the temporal-spatial ordering problems that are evident in dyslexia. In particular we have established in our case study that a great many phonological reading errors can be accounted for if establishing anti-phase coupling between phonological gestures is difficult for the person with dyslexia.

References Allen, William Sydney. 1973. Accent and rhythm. Cambridge: Cambridge University Press. Browman, Catherine P. and Louis Goldstein. 2000. Competing constraints on intergestural coordination and selforganization of phonological structure. Les Cahiers de l’ICP, Bulletin de la Communication Parlée 5, 25-34. Duncan, Lynne G., Phillip H.K. Seymour, and Shirley Hill. 1997. How important are rhyme and analogy in beginning reading? Cognition 63, 171-208. Goad, Heather and Kathleen Brannen. 2003. Phonetic evidence for phonological structure in syllabification. In Jeroen van de Weider, Vincent J. van Heuven, and Harry van der Hulst (eds.), The Phonological Spectrum. Amsterdam: John Benjamins, 3-30.

Goldstein, Louis, Dani Byrd and Elliot Saltzman. 2006. The role of vocal tract gestural action units in understanding the evolution of phonology. In Michael Arbib (ed.), From action to language: The mirror neuron system. Cambridge: Cambridge University Press, 215-249 Goldstein, Louis, Hosung Nam, Elliot Saltzman and Ioana Chitoran. 2008. Coupled oscillator planning model of speech timing and syllable structure. In Proceedings of the 8th Phonetics Conference of China and the International Symposium on Phonetic Frontiers. (CD-Rom) Goldstein, Louis, Hosung Nam, Elliot Saltzman and Ioana Chitoran. 2008. Coupled oscillator planning model of speech timing and syllable structure. In Proceedings of the 8th Phonetics Conference of China and the International Symposium on Phonetic Frontiers. (CD-Rom) Goswami, Usha. 1999. Causal connections in beginning reading: the importance of rhyme. Journal of Research in Reading 22(3), 217–240. Goswami, Usha. 2001. Rhymes are important: A comment on Savage. Journal of Research in Reading 24(1), 1929. Goswami, Usha, Jennifer Thomson, Ulla Richardson, Rhona Stainthorp, Diana Hughes, Stuart Rosen, and Sophie K. Scott. 2002. Amplitude envelop onsets and developmental dyslexia: a new hypothesis. Proceedings of the National Academy of Sciences 99(16), 10911-10916. Goswami, Usha. 2003. How to beat dyslexia: The broadbent lecture 2003. The Psychologist 16(9), 462-465. Hays, Bruce. 1995. Weight of CVC can be determined by context. In Jennifer Cole & Charles Kisseberth (eds.), Perspectives in Phonology, Center for the Study of Language and Information, Stanford, CA, 61-79. Hockett, Charles F. 1955. A Manual of Phonology. Baltimore: Waverly Press, Inc. Ito, Junko and Armin Mester. 1992. Weak layering and word binarity. Linguistic Research Center, LRC-92-09, University of California, Santa Cruz Kager, René. 2007. Feet and metrical stress. In Paul de Lacy (ed.) The Cambridge Handbook of Phonology. Cambride: Cambridge University Press, 195-227. Katada, Fusa. 1990. On the representation of moras: evidence from a language game. Linguistic Inquiry 21, 641646. Katada, Fusa. 2008. Word reversing by a person with Williams syndrome: more evidence for mora as independent unit. Paper presented at CUNY Phonology Forum 2008. Kessler, Brett and Rebecca Treiman. 1997. Syllable structure and the distribution of phonemes in English syllables. The Journal of Memory and Language 37, 295-311. Kubozono, Haruo. 1989. The mora and syllable structure in Japanese: evidence from speech errors. Language and Speech 32, 249-278. Lowenstamm, Jean. 1996. CV as the only syllable type. In Jacques Durand & Bernard Laks (eds.), Current Trends in Phonology: Models and Method. European Studies Research Institute, University of Salford Publications, 419-442. McCarthy, John. 1979. On stress and syllabification. Linguistic Inquiry 10, 443-466. Nam, H., and Saltzman, E. 2003. A competitive, coupled oscillator model of syllable structure. Proceedings of the XIIth International Congress of Phonetic Sciences, 2253-2256. Poser, William. 1990. Evidence for foot structure in Japanese. Language 66. 78-105. Schneider-Zioga, Patricia and Jean-Roger Vergnaud. 2009. Feet and their combination. Paper presented at CUNY Phonology Forum 2009. Thomson, Jennifer, Ben Fryer, James Maltby, and Usha Goswami. 2006. Auditory and motor rhythm awareness in adults with dyslexia. Journal of Research in Reading 29(3), 334-348 Treiman, Rebecca, John Mullennix, Ranka Bijeljac-Babic, and E. Daylene Richmond-Welty. 1995. The Special Role of Rimes in the Description, Use, and Acquisition of English Orthography. Journal of Experimental Psychology: General 124 (2), 107-136. Ueno, Kazuhiko. 2006. LD to Disurekushia (LD and Dyslexia). Tokyo: Koudansha. Wise, Barbara, Richard Olson, and Rebecca Treiman. 1990. Sub-syllabic units as aids in beginning readers word learning onset-rime versus post-vowel segmentation. Journal of Experimental Child Psychology 49, 1-19. Vergnaud, Jean-Roger. 2007. Some preliminary remarks on the syntactic organization of phonology. Paper presented at CUNY Phonology Forum 2007.