Overcoming the effect of letter confusability in letter-by-letter reading: A rehabilitation study

This article was downloaded by: [University of Waikato]On: 10 July 2014, At: 08:31Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,UK

NeuropsychologicalRehabilitation: An InternationalJournalPublication details, including instructions for authorsand subscription information:http://www.tandfonline.com/loi/pnrh20

Overcoming the effect of letterconfusability in letter-by-letterreading: A rehabilitation studyLara Harris a , Andrew Olson a & Glyn Humphreys ba University of Birmingham , Birmingham , UKb University of Oxford , Oxford , UKPublished online: 01 Mar 2013.

To cite this article: Lara Harris , Andrew Olson & Glyn Humphreys (2013) Overcomingthe effect of letter confusability in letter-by-letter reading: A rehabilitation study,Neuropsychological Rehabilitation: An International Journal, 23:3, 429-462, DOI:10.1080/09602011.2013.776500

To link to this article: http://dx.doi.org/10.1080/09602011.2013.776500

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all theinformation (the “Content”) contained in the publications on our platform.However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, orsuitability for any purpose of the Content. Any opinions and views expressedin this publication are the opinions and views of the authors, and are not theviews of or endorsed by Taylor & Francis. The accuracy of the Content shouldnot be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions,claims, proceedings, demands, costs, expenses, damages, and other liabilitieswhatsoever or howsoever caused arising directly or indirectly in connectionwith, in relation to or arising out of the use of the Content.

http://www.tandfonline.com/loi/pnrh20

http://www.tandfonline.com/action/showCitFormats?doi=10.1080/09602011.2013.776500

http://dx.doi.org/10.1080/09602011.2013.776500

This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expresslyforbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

http://www.tandfonline.com/page/terms-and-conditions

http://www.tandfonline.com/page/terms-and-conditions

Overcoming the effect of letter confusability in

letter-by-letter reading: A rehabilitation study

Lara Harris1, Andrew Olson1, and Glyn Humphreys2

1University of Birmingham, Birmingham, UK2University of Oxford, Oxford, UK

Patients who read in a letter-by-letter manner can demonstrate effects of lexicalvariables when reading words comprised of low confusability letters,suggesting the capacity to process low-confusability words in parallel acrossthe letters (Fiset, Arguin, & McCabe, 2006). Here a series of experiments ispresented investigating letter confusability effects in MAH, a patient withexpressive and receptive aphasia who shows reduced reading accuracy withlonger words, and DM, a relatively “pure” alexic patient. Two rehabilitationstudies were employed: (i) a word-level therapy and (ii) a letter-level therapydesigned to improve discrimination of individual letters. The word-level treat-ment produced generalised improvement to low-confusability words only, butthe serial processing treatment produced improvement on both high and lowconfusability words. The results add support to the hypothesis that letter con-fusability plays a key role in letter-by-letter reading, and suggest that a rehabi-litation method aimed at reducing ambiguities in letter identification may beparticularly effective for treating letter-by-letter reading.

Keywords: Letter-by-letter reading; Letter confusability; Letter-level readingtreatments.

Correspondence should be addressed to Dr Lara Harris, School of Psychology, University of

Birmingham, Edgbaston, Birmingham B15 2TT. E-mail: [email protected]

This work was conducted in partial fulfilment of a PhD at the University of Birmingham by

the first author. The work was supported by the Stroke Association and the NIHR.

We are grateful to patients MAH and DM for their enthusiasm and energy in participating in this

work.

Lara Harris is now working as a Post-doctoral Researcher in Neuropsychology at Academic

Neuropsychiatry, Institute of Psychiatry, King’s College London.

Neuropsychological Rehabilitation, 2013

Vol. 23, No. 3, 429–462, http://dx.doi.org/10.1080/09602011.2013.776500

# 2013 Taylor & Francis Group

Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

INTRODUCTION

Letter-by-letter (LBL) reading is a strategy developed by some dyslexicpatients that results in a pronounced word length effect in reading (Gesch-wind, 1965a, 1965b, Price & Humphreys, 1992). Several studies haveshown that patients adopting a letter-by-letter reading strategy can show sen-sitivity to lexical properties of words even before serial processing of theletters takes place. For example, studies of implicit reading have found thatLBL patients can make above-chance lexical and semantic decisionsdespite exposure durations that are insufficient for explicit word naming(Coslett & Saffran, 1989; Coslett, Saffran, Greenbaum, & Schwartz, 1993;Shallice & Saffran, 1986). Recent work into the effect of letter confusability(a measure of the visual similarity between the capital letters of the alphabet)in these patients has further suggested that there is activation of lexical rep-resentations not based on serial processing. (Although most matrices arebased on capital letters, some use lower-case letters, e.g., Mueller & Weide-mann, 2012.) Arguin and Bub (2005) reported interactive effects of confusa-bility and orthographic neighbourhood (N) size on reading in three alexicpatients: When letter confusability was low there was a facilitatory effectof N size. With high confusability words there was no significant effect ofN size. There is evidence that effects of N reflect parallel processingof words. For example, normal participants do not show facilitatory effectsof N when words are presented in sequential fragments (Snodgrass &Mintzer, 1993). N effects in normal participants are also more pronouncedwith low frequency relative to high frequency words (Andrews, 1989,1992; Arguin, Bub, & Bowers, 1998; Sears, Hino, & Lupker, 1995),suggesting that the effects emerge when word processing is made more diffi-cult. The positive effects of N found with low confusability words in LBLpatients indicate that there can be parallel processing of letters in words,but such parallel processing also remains relatively problematic.

Inhibitory effects of high N size have also been demonstrated. Pugh,Rexer, Peter, and Katz (1994) showed that neighbours had a detrimentaleffect on reading in normal subjects when the letter distinguishing the itemfrom its neighbours occurred after a 100ms delay. Also, in the neuropsycho-logical literature, a patient presenting with (left) neglect dyslexia showed adetrimental effect of N size when a neighbour differed from the targetword by the leftmost letters (the first two letters in 4-letter words), particu-larly when the first letter was visually similar to that of a neighbouringitem (Arguin & Bub, 1997). Therefore, it may be that under conditions ofserial presentation (e.g., delayed letter presentation in normal participants),or processing (e.g., neglect dyslexia or letter-by-letter reading), high Nwords may exert a deleterious effect. Presumably this effect would beespecially detrimental when neighbours of a given item differed by a letter

430 HARRIS, OLSON, AND HUMPHREYS

Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

at the terminal positions. Given that LBL readers show strong serial com-ponents in reading, it is possible that negative effects of N could be apparentunder some conditions.

Rehabilitation studies

Much of the rehabilitation work with patients showing letter-by-letter behav-iour has sought to improve residual capacities in parallel processing (Coslettet al., 1993). However, since pre-lexical deficits may also be present in thesepatients (Chialant & Caramazza, 1998), improvements may additionally bebrought about by improving letter discrimination. Several rehabilitationstudies have used letter identification training with LBL patents. MultipleOral Reading (MOR) methods require the repeated reading of text in orderto improve reading speed (e.g., Beeson, 1998; Moody, 1988; Moyer, 1979;Tuomainen and Laine, 1991). Both Tuomainen and Laine (1991) andMoyer (1979) applied an MOR technique to LBL patients, whilst also incor-porating training in letter discrimination and identification. The patients’reading rates improved following therapy, but it was unclear whether thistreatment effect was due to the letter identification or parallel processingaspects of the training. Arguin and Bub (1994) more specifically targetedthe letter processing capacities of their patient with a method that usedspeeded same–different letter matching and overt reading of pronounceableletter strings. They found that while identification did not improve at the letterlevel, there was a generalised improvement to whole-word reading. Theauthors concluded from this result that a visual processing deficit was a keycontributing factor to letter-by-letter reading. A recent study has indicatedthat positive treatment effects following MOR treatments may be driven byrepeated learning of specific words rather than general improvements intop-down processing, limiting the generalisability of these effects toreading text containing practised words (Lacey, Lott, Snider, Sperling, &Friedman, 2010).

Kinaesthetic approaches have been used in letter-by-letter reading patients(Lott, Carney, Glezer, & Friedman, 2010; Sage, Hesketh, KinastheticLambon Ralph, 2005). Sage et al. (2005) contrasted the effectiveness of par-allel and serial processing treatments in their rehabilitation study of theirpatient FD, who frequently used letter-by-letter reading. Using an ABACAdesign the authors applied a word-level treatment (e.g., pairing orthographicrepresentation of a word with its phonological form, as read by the therapist)which aimed to promote parallel processing, and a letter-level treatmentwhich focused on distinguishing letter forms using kinaesthetic information(tracing letters) and improving letter naming. While both treatmentsimproved accuracy on directly treated sets, only the letter-level therapy pro-duced generalised improvement to control items. At initial baseline, FD

LETTER CONFUSABILITY IN LBL READING 431

Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

showed a preponderance of omission errors, although after the word-leveltreatment, visual and semantic errors dominated, leading the authors at thisstage to characterise him as a deep dyslexic patient. The letter-level treatmentreduced the number of visual errors when compared to the word-level treat-ment. These data suggest that a letter-level approach may be more effective inovercoming letter confusability than a word-level treatment.

The current study

To our knowledge, this work is the first to directly investigate whetherimproving letter discrimination processes in rehabilitation can reduce the det-rimental effect of letter confusability on reading, and whether there are widerconsequences of improved letter discrimination on the effects of word lengthand neighbourhood size.

Our experimental tasks assessed the role of letter confusability in letter-by-letter reading behaviour in DM and MAH. DM presented with pure alexiawhile MAH showed letter-by-letter reading in the context of wider languageproblems. Both patients showed strong effects of word-length and frequentlynamed letters while reading. We tested for interactive effects between confu-sability, word length, and orthographic neighbourhood size at initial baseline.These experiments test the idea that some letter-by-letter patients are capableof parallel processing under conditions of low perceptual demands, but mayemploy a laborious serial strategy when discrimination is hard.

We further tested this idea through a rehabilitation study, treating readingbehaviour on a set of high and low confusability words through two therapies:a parallel processing (word-level) approach and a letter-level technique toimprove letter discrimination (the speed and accuracy of serial reading)using an ABACA design. We tested generalised improvement in readingafter each therapy on (i) untreated high and low confusability words, (ii)untreated words with high and low neighbourhood sizes, and (iii) untreatedwords of varying length, to test the effect of each therapy on the wordlength effect in reading. A study timetable is provided in Table 1.

We predicted that word-level therapy would improve the reading of lowbut not high confusability words, because the treatment is geared at proces-sing the whole word form, and this should be effective under conditions oflow but not high perceptual demands. The letter-level therapy shouldimprove reading of all words, as the treatment focus is on improving letterdiscrimination, which should help overcome effects of letter confusability.Our specific aims were to assess: (i) the relative effects of treating parallelprocessing (using a word-level treatment) compared with serial reading(through a letter-level treatment); (ii) the efficacy of these treatments in apure alexic patient (DM) as well as in a patient with letter-by-letter readingin the context of central alexia (MAH); (iii) generalised improvement from


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

each therapy on low and high confusability words; (iv) the effect of parallelprocessing and serial reading treatments on the length effect; (v) generalisedtreatment gains on high and low neighbourhood reading, and (vi) possibleinteractive effects of letter confusability and N size.

CASE DESCRIPTIONS

DM

Patient DM, a 55-year-old female, experienced left medial and inferior occipito-temporal brain damage as a result of an abscess in her brain caused by multiplearteriovenous malformations (AVMs) in her lungs. MRI images for DM,showing damage to the left inferior medial occipito-temporal region, are pro-vided in Humphreys, Riddoch, and Price (1997). She formally worked as anEnglish teacher in a secondary school. Functionally she presented with homon-ymous hemianopia, alexia without agraphia (Osswald, Humphreys, & Olson,2002) and anomia with a particular impairment in naming living things(Humphreys et al., 1997). Aside from some occasional naming difficulties,DM’s free speech was fluent and grammatically well-formed. There was no evi-dence of a semantic impairment. Data on DM’s reading performance have beenpublished previously (Osswald et al., 2002), and her performance on readingtests from PALPA is summarised in Table 2. Broadly, DM presents with accu-rate, but slow, reading, which is strongly modulated by length. Her length-sensitive reading performance is set against (mostly) preserved visual percep-tion and recognition, auditory processing, semantic processing, and spelling,leading us to characterise DM as a pure alexic patient.

TABLE 1Study timetable

Time Description

Weeks 1–6 Baseline assessments, and reading performance on (i) treated word set; (ii) untreated

words – high and low confusability; (iii) untreated words – high and low

confusability and N size; (iv) untreated words of varying word length

Weeks 7–17 Rehabilitation phase 1: Word-level therapy: Encouraging fast parallel processing

reading under short presentation

Weeks 18–22 Baseline assessments: Reading (i) treated word set; (ii) untreated words – high and

low confusability; (iii) untreated words – high and low confusability and N size;

(iv) untreated words of varying word length

Weeks 23–33 Rehabilitation Phase 2: Letter-level therapy

Weeks 34–38 Baseline assessments: Reading (i) treated word set; (ii) untreated words – high and

low confusability; (iii) untreated words – high and low confusability and N size;

(iv) untreated words of varying word length


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

MAH

MAH was referred to the Behavioural Brain Sciences unit at the University ofBirmingham in 2008, and was diagnosed with chronic receptive and expres-sive aphasia after a cerebrovascular accident (CVA). MAH managed thefamily business in selling non-metallic fixings before he retired and he hadalready retired at the time of his CVA. At the time of testing he was aged76–77 years. Lesion analysis using SPM and MRIcron revealed damage tothe left superior and middle temporal gyrus, left superior temporal pole,left insula and the head of left caudate (Figure 1).

Compared to DM, MAH presented with much wider problems in language.MAH’s free speech contained frequent semantic and phonological parapha-sias, with frequent attempts to self-correct. Such self-correction tendenciessuggest a problem not due to a semantic deficit per se but rather a problemin access to the phonological output system. This was also suggested by hisrelatively good performance on tests of semantic access, especially withwritten presentation.

His reading performance was marred by frequent, largely phonological,errors, although he was fairly accurate with shorter words, and 5-letterwords that were highly frequent and imageable (Table 2). As with DM, hisreading was affected by length.

Like DM, MAH had preserved visual perception and recognition, achiev-ing good scores on all assessments in the Visual Object and Space PerceptionBattery (VOSP; Warrington & James, 1991) and the Birmingham Object Rec-ognition Battery (BORB; Riddoch & Humphreys, 1993). There was no evi-dence of a hemianopia and he could identify single letters briefly exposedalone in his right visual field.

We assessed letter recognition capabilities in both patients using two letteridentification tasks. Both tasks used letters with very different upper and lower-case forms (e.g., q-Q but not S-s). In one task, we presented patients with 10upper and 10 lower-case letters and asked them to write each letter in the oppo-site case. DM made no errors, and MAH made only one error (q-P). In thesecond task patients performed same–different letter judgements, where 60pairs of letters were displayed on a computer. The two letters in a pair appearedin different cases, and the patients indicated whether the two letters corre-sponded to the same grapheme (e.g. a-A) or not (e.g., a-B), using one of tworesponse keys. On this task DM scored 59/60, and MAH scored 57/60.Noting the long response times for this task in both patients, we ran anotherversion of the task where the letters appeared for only 200ms and were followedby a masking pattern (######). Here, performance accuracy fell dramatically,with DM and MAH scoring 38/60 and 35/60, respectively.

Both patients also performed a computer-based lexical decision task.Twenty frequent (log10 frequency per million .2) words and 20 consonant


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

strings, matched in length, were displayed for 200ms, followed by a maskingpattern (######). Patients decided whether the stimuli were real words or not.DM made four and MAH made six errors. Both patients showed good, albeitslow, identification of single letters. Under short display conditions, bothcross-case matching and lexical decision was impaired. In sum, there was evi-dence for pre-lexical deficits in both patients.

MAH’s auditory processing skills were largely preserved on tests using realwords, e.g., on same–different judgements with written (67/72, PALPA 2,Kay, Lesser, & Coltheart, 1992) and picture (38/40, PALPA 4, Kay et al.,1992) selection, but not with same–different judgements with nonwords(52/72, PALPA 1, Kay et al., 1992). This pattern of performance suggestsfacilitated auditory discrimination with meaningful stimuli. MAH showedsome impairment at making fine auditory discriminations, generally sparedauditory lexical access, but strong effects of syllable length on repetition(PALPA 7: 1 syllable: 5/8, 2 syllables: 4/8, 3 syllables: 1/8, Kay et al., 1992).

For MAH, there appeared to be no effect of syllable length on reading whenword length was controlled (PALPA 30, Kay et al., 1992, Table 2, Chi Squaretest, p ¼ .135). Although impaired in both conditions, MAH read regularlyspelled words better than irregular items, 15/30 and 5/30, respectively, ChiSquare test, x2(1) ¼ 20.000, p , .0001. Across the different stimuli themost frequent errors were phonologically related to targets (85.38%, nonwords55.08%, and real words 29.29%) and MAH made no semantic errors.

MAH was impaired across a range of spelling conditions. On a spelling testmanipulating word length, MAH scored 10/24. There was no effect of wordlength on accuracy (3-letter 2/6, 4-letter 3/6, 5-letter 2/6, 6-letter 3/6,PALPA 39, Kay et al., 1992) but there were signs of frequency and imageabil-ity effects (7/20 for High Imageability, High Frequency (HIHF), 4/20 forHigh Imageability, Low Frequency (HILF) and 0/20 on both Low Imageabil-ity, High Frequency (LIHF) and Low Imageability, Low Frequency (LILF)words, PALPA 40, Kay et al., 1992). However, there was no effect of regu-larity (regular 6/20; exception 6/20), a finding reproduced in a homophonespelling test (regular 2/20; exception 3/20). A test assessing performanceon words of different grammatical classes produced very low performance(1/20, “bell”). MAH scored 7/30 on a lexical morphology spelling test,

Figure 1. Images from MAH’s structural MRI scan. N.B. Grey matter lesion appears in red and white

matter lesion in green. The lesion was created in SPM and added as an overlay onto a standard multi-

slice template in MRIcron. The SPM analysis was a one-sample t-test with the covariates healthy (140

brains aged 40+) vs. age and gender.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

TABLE 2DM’s and MAH’s scores on tests of reading

DM MAH

Published normal

cut-off scores

PALPA 29 Reading: word

length

3-letter: 6/6, M RT ¼

0.62


1.13


1.56


2.65

3-letter: 5/6∗

4-letter: 3/6∗

5-letter: 2/6∗

6-letter: 0/6∗

3-letter: M ¼ 6

4-letter: M ¼ 6

5-letter: M ¼ 6

6-letter: M ¼ 6

PALPA 30 Reading: syllable

length

1 syllable: 8/8, M RT

¼ 1.44

2 syllables: 7/8∗, M

RT ¼ 2.20

3 syllables: 8/8, M

RT ¼ 2.91

1 syllable: 1/8∗

2 syllables: 2/8∗

3 syllables: 1/8∗

1 syllable: M ¼ 7.83

(0.38)

2 syllables: M ¼ 8

3 syllables: M ¼ 7.90

(0.31)

PALPA 31 Reading:

imageability × frequency

HIHF 20/20, M RT

¼ 1.76

HILF 20/20, M RT

¼ 2.05

LIHF 20/20, M RT

¼ 2.58

LILF 19/20, M RT ¼

3.20

HIHF 17/20∗

HILF 13/20∗

LIHF 1/20∗

LILF 2/20∗

HIHF: M ¼ 19.94

(0.25)

HILF: M ¼ 19.94

(0.07)

LIHF: M ¼ 20

LILF: M ¼ 19.52

(0.68)

PALPA 32 Reading:

grammatical class

Nouns: 20/20

Adjectives: 20/20

Verbs: 20/20

Functors: 17/20∗

Nouns: 2/20∗

Adjectives:

2/20∗

Verbs: 3/20∗

Functors: 2/20∗

Nouns: M ¼ 19.87

(0.43)

Adjectives: M ¼ 19.97

(0.18)

Verbs: M ¼ 19.97

(0.18)

Functors: M ¼ 19.93

(0.37)

PALPA 35 Reading:

regularity

Regular words:

30/30

Irregular words:

30/30

Regular words:

15/30∗

Irregular words:

5/30∗

Regular words: M ¼

29.96 (0.20)

Irregular words: M ¼

29.85 (0.37)

PALPA 36 Reading:

nonwords

3-letter 6/6, M RT ¼

3.14


3.11


4.75


7.64

3-letter 2/6∗

4-letter 3/6∗

5-letter 1/6∗

6-letter 0/6∗

3-letter: M ¼ 5.77

(0.71)

4-letter: M ¼ 5.89

(0.43)

5-letter: M ¼ 5.57

(0.90)

6-letter: M ¼ 5.65

(0.85)

∗impaired performance. HIHF ¼ High Imageability, High Frequency; HILF ¼ High Imageabil-

ity, Low Frequency; LIHF ¼ Low Imageability, High Frequency; LILF ¼ Low Imageability, Low

Frequency.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

with no effects of item morphology. Again, successfully spelled words werehighly imageable (hairy, bees, rug, mice, daisy, sailor, and pram). MAH wasimpaired on a number of tests of semantic access. On picture naming (18/40,PALPA 53 Kay, Lesser, & Coltheart, 1992), most (22) errors were phonolo-gically related to targets, resulting in 13 nonword and 6 word productions.The remaining errors were nonword responses that were phonologically unre-lated to the target. On the PALPA test of spoken word–written word match-ing (PALPA 52), MAH scored 8/15, with 5 errors due to selecting thesynonym and 2 to selecting the semantic foil. MAH’s performance on the Pyr-amids and Palm Trees Test (Howard & Patterson, 1992) was within thenormal range when stimuli were three written words, but below the normalcut-off when three pictures were used (47/52).

Case summaries

DM represented a case of “pure” alexia. She was within the normal range ontests of auditory processing, spelling, semantic access and comprehension.Under short display conditions, DM made errors in cross-case matchingand lexical decision. Her reading accuracy was usually at ceiling, althoughher reading times showed effects of word length and she frequently namedletters aloud before attempting to read the whole word. She demonstratedeffects of imageability and frequency on reading at pre-therapy baseline.

MAH showed poor comprehension in reading and with auditory input (audi-tory sentence comprehension, spoken word–picture, and spoken word–written word matching). Phonological processing was better preserved,although performance was impaired with meaningless stimuli. Performancein reading, spelling and comprehension was also influenced by lexical vari-ables (frequency and imageability). A regularity effect was observed inreading only, and there was an effect of word length (linear decline in wholeword accuracy as word length increases) in reading (Table 2). Also, MAHwas impaired in cross-case matching and lexical decision under shortdisplay conditions. Broadly, the results suggest that MAH suffers central def-icits in language, with a profile of deep dysgraphia for spelling, and readingperformance showing some aspects of phonological dyslexia (difficultiesreading nonwords, the absence of semantic errors and a preponderance ofvisual paralexias). In addition, MAH’s length-sensitive reading, and his fre-quent naming aloud of letters, is consistent with a profile of reliance onserial letter processing. It is possible that MAH’s ability to fall back on thisreading strategy obscured the semantic errors that would characterise hisreading as deep dyslexic. The data suggested that MAH was a deep dysphasic,deep dysgraphic and possible deep dyslexic patient.

The current paper is comprised of two parts: Part 1 was an experiment con-ducted before any treatment took place to test the extent to which reading is


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

modulated by letter confusability, and to explore any interactive effect letterconfusability might have with effects of neighbourhood size. Part 2 aimed totest the efficacy of two therapies: one a word level, parallel processing andanother a letter-level, serial reading approach, and to assess the effect eachtherapy had on confusability, N size, and length effects in reading.

PART 1. EXPERIMENTAL INVESTIGATION: CONFUSABILITYAND NEIGHBOURHOOD SIZE

The current rehabilitation study aimed to test whether promoting letter identi-fication skills led to increased parallel processing in DM and MAH. Parallelprocessing was assessed by testing for effects of orthographic neighbourhood(see Arguin et al., 1998). The effects of this variable, and of letter confusabil-ity, were initially tested as a baseline measure. Effects of orthographicneighbourhood size were tested in DM and MAH pre-therapy, and whetherthese effects were modulated by letter confusability was assessed.

Experimental investigation method

Materials

The experiment used 180 3–7 letter words that were categorised as lowconfusability, low neighbourhood size (LCLN), low confusability, high neigh-bourhood size (LCHN), high confusability, low neighbourhood size (HCLN), orhigh confusability, high neighbourhood size (HCHN). These sets were matchedon a range of lexical variables (see Appendix A for stimulus details).

Method

Measures were based on the mean confusability (visual similarity withother letters) for each word. The confusability scores were taken fromletter-confusion matrices obtained in previous studies (Gilmore, Hersh, Car-amazza, & Griffin, 1979; Loomis, 1982; Townsend, 1971; Van Der Heijden,Malhas, & Van Den Roovaart, 1984), developed from rating data from normalparticipants. Letter confusability ranged from .24 (letter L) to .73 (letter B),with an average of .48. For each word, we calculated the summed confusabil-ity of letters, and then the mean confusability per letter in the word.

Words with , .45 mean confusability were categorised as low confusability,and . .52 as high confusability (as in Fiset et al., 2006). Words with smallorthographic neighbourhoods were selected as those with zero to one neigh-bour, and the high N word set comprised words with three or more neighbours.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

Procedure

The experimental stimuli were created using E-Prime stimulation software(E-Prime 1.2, Psychology Software Tools, 2002) and were displayed on a1024 × 768 Samsung monitor. Words were presented to the left of fixationin 18 point Arial font to counter right hemianopia in both patients. The soft-ware recorded accuracy and response times. Vocal onsets were used as ameasure of response latency, and these were recorded using a microphoneattached to a Stimulus Response (SR) box. After each trial the examinertyped in whether the patient’s response was correct (1) or incorrect (0).The examiner manually recorded erroneous responses verbatim. Prior to ana-lysing correct response latencies, the following values were excluded: (i)trials where there was technical difficulty with the response means (e.g.,the voice key failed to register the participants’ vocal onset), (ii) valuesless than 200ms, and (iii) values more than three standard deviations awayfrom the mean in that condition for that patient. The words in the experimentserved as the untreated word set, and the experiment was run with bothpatients at each baseline phase.

Results

Accuracy analyses

DM made no reading errors in any neighbourhood or confusability con-dition. MAH’s accuracy data are provided in Figure 2.

MAH’s accuracy data were submitted to a loglinear analysis. The testrevealed a significant three-way interaction between confusability, neigh-bourhood size and score, x2(1) ¼ 28.029, p , .001, reflecting a significantfacilitative effect of neighbourhood for the low confusability items, x2(1)¼ 16.843, p , .001, but a deleterious effect of neighbourhood for high con-fusability items, x2(1) ¼ 11.562, p ¼ .001, Figure 2.

RT analyses

Due to the poor accuracy rates in MAH’s performance in some conditions,his RT data were not analysed. DM’s mean RT data are presented in Figure 3.

Normality tests conducted on DM’s RT data were not significant (Shapiro-Wilk test p . .1) and so the data were entered into a univariate ANOVA.There was a significant interaction between confusability and N size, F(1,338) ¼ 53.433, p , .001. The interaction reflected significantly better per-formance with high N relative to low N words in the low confusability set(p ¼ .001), but a significantly poorer performance with high comparedwith low N words in the high confusability set (p , .001). There was adeficit for high neighbourhood words with high confusable letters, t(159) ¼


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

Figure 2. MAH’s parallel processing accuracy in reading words (n size and letter confusability). N.B.

LCLN ¼ low confusability, low neighbourhood size, LCHN ¼ low confusability, high

neighbourhood size, HCLN ¼ high confusability low neighbourhood size, and HCHN ¼ high

confusability, high neighbourhood size.

Figure 3. DM’s response latencies in reading words (n size and letter confusability). Error bars are

based on 95% confidence intervals.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

4.906, p , .0001, and a facilitation with low confusability letters, t(179) ¼5.450, p , .0001. The pattern of facilitative effect of higher N in low butnot confusability words is consistent with other work on the subject (e.g.,Arguin et al., 1998), and more generally with the idea that high N size is facil-itative when letter discrimination is good, but it disrupts performance whenperceptual demands increase (with highly confusable letters), where thehigher neighbourhoods of items compete for selection.

Discussion

The effect of orthographic neighbourhood was mediated by letter confusabil-ity in MAH and DM. A facilitative effect of high orthographic neighbourhoodwas found only when words were composed from low confusability letters.However, for words with high confusability letters, there was a detrimentaleffect of high neighbourhood. This pattern of results was found in the accu-racy data for MAH and the RT data for DM. The negative effect of highneighbourhoods can be attributed to disproportionately increased competitionfrom other neighbours when the letters are themselves high in confusability.In contrast, the better discrimination of low confusability letters may arisebecause representations for these words can be activated in parallel acrossthe letters present, enabling positive effects of supporting (high neighbour-hood) words to emerge. In sum, it appears that large neighbourhoods are sup-portive when discrimination is good but they introduce increased competitionwhen discrimination is poor.

PART 2. REHABILITATION STUDY

Despite their vastly different cognitive profiles, both patients showed letter-by-letter reading behaviour which was affected by letter confusability. Inboth MAH and DM there was a better performance with low confusabilitywords, which elicited “normal” facilitative effects of high neighbourhoodsize. There was also impaired reading of high confusability words, whichactually showed a deleterious effect of high N, possibly reflecting greatercompetition from these many neighbours where letter discrimination is poor.

The rehabilitation study aimed to remediate the letter confusability effectin DM and MAH and also to assess the effects on reading of improving dis-crimination between confusable letters. We used two rehabilitation tech-niques based on the word-level and letter-level therapies described in Sageet al. (2005). In particular, we were interested in whether the treatmentsimproved parallel processing, and whether this improvement was specificto particular word sets. For instance, it was hypothesised that the word-level therapy would promote parallel processing, but that this may only bemanifest in selective sets with low perceptual demands (low confusability


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

words). In contrast, letter recognition therapy, focused on distinguishingbetween visually similar competitor letters, aimed to improve letter-by-letterprocessing, which may in turn improve the reading of words with high con-fusability letters. Parallel processing was assessed by analysing accuracyand response times in high and low confusability words, the presence ofneighbourhood effects in low and high confusability words (under theassumption that improved reading times in high neighbourhood conditionsindicate parallel processing), and length effects in reading (with a levelling-out of reading time and accuracy on short and long words suggesting parallelprocessing). The baseline assessments used before and then after each treat-ment were (i) high and low confusability words; (ii) high (HN) and low neigh-bourhood (LN) words comprised of high (HC) and low confusability (LC)letters (yielding four reading conditions: LCLN; LCHN; HCLN; HCHN)and (iii) 4–7-letter words.

Rehabilitation study method

Participants

Patients DM and MAH participated in the rehabilitation study.

Materials

There were four word sets, the set of treated words, and three untreated setsto assess generalised improvement – (i) untreated words which were eitherhigh or low in confusability; (ii) untreated words which varied in confusabil-ity and N; and (iii) untreated words of varying word length. The treated wordscomprised 40 4- and 5-letter words. Half of the set contained high confusabil-ity and half contained low confusability letters (see Appendix B for details ofthe stimuli).

1. Untreated words – high and low confusability: Another untreated set ofhigh and low confusability words were also used (see Appendix C fordetails) to assess the effect of confusability over the baseline phaseoverall.

2. Untreated words – high and low confusability and N size: The stimulidescribed in the Experimental investigations section were reserved asuntreated items and were tested at each baseline (untreated confusabil-ity by neighbourhood set, see Appendix A for details of the stimuli).

3. Untreated words – word length: Words comprised of 4–7 letters, matchedon a range of lexical variables (see Appendix D for information).

For the tests manipulating confusability, words with , .45 mean confusa-bility were categorised as low confusability, and . .53 as high confusability,


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

using Gilmore et al. (1979). Words in each condition were matched across arange of lexical variables (see Appendices A, B, C and D for information onthe stimulus sets).

Procedure

An ABACA design was used. As in Sage et al. (2005), the word-level treat-ment was employed first and the letter-level therapy second. An error-reducing paradigm was adopted to prevent reinforcement of errors (Wilson,Baddeley, Evans, & Shiel, 1994).

Word-level treatment

The first therapy aimed to promote parallel processing of words. In bothtreatments, words were presented on individual cards in upper case in 16point Arial font. The treatment was conducted at home and also duringone-hour weekly rehabilitation and assessment sessions at the University.Patients’ partners were given detailed instructions about the methodologyand were happy to assist with the treatment. Each treatment phase lastedfor 10 weeks.

The procedure for the word-level treatment was as follows:

1. A treatment trial began with presenting the participant with a word card.2. The therapist repeated the word five times while the patient studied the

word on a card. Patients were asked to listen rather than attempt pro-duction at this stage.

3. The patient was then requested to repeat the word five times while stillstudying the word card. DM never made errors at this stage, but MAHmade frequent phonological paraphasias, which may have interferedwith the intervention (see Fillingham, Hodgson, Sage, & LambonRalph, 2003, for a discussion of this in relation to errorless learning).In these instances, the therapist read the word aloud, and asked MAHto repeat the word a further five times.

Serial reading treatment

The second treatment aimed to promote letter identification, and overcomeletter confusability by emphasising differences between visually similarletters.

1. The participant was presented with each word card, but this time onlyone letter was visible at a time, using a “moving window” card posi-tioned over the word card. While showing the participant each letter,the therapist read each letter name and asked the participant to repeatit. In addition, the participant was asked to trace the letter shape with


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

his/her finger. If any letter was repeated incorrectly, the therapist readthe letter again and asked the participant to repeat it. When a letter namewas repeated accurately, the next letter in the word was treated.

2. When this step was completed successfully, the participant was asked toread each letter before reading the whole word. If an error was made,the participant was taken back to step 1.

Results

Performance on the treated set across baselines

DM’s reading of words was at ceiling at both pre- and post-therapy base-lines (20/20 in each confusability and baseline condition), and so her accu-racy data were not analysed. MAH’s reading accuracy on the treated wordsis shown in Figure 4.

MAH’s reading accuracy data on low and high confusability treated wordsat the three baseline points (initial baseline, after word-level treatment andafter letter-level treatment) were entered into a loglinear analysis. Therewas a significant main effect of baseline, x2(1) ¼ 114.371, p , 0001, d ¼9.015, but not confusability (p ¼ .360). We summed across low and high con-fusability words and submitted the data to McNemar tests. The analyses indi-cated improved reading accuracy after word-level treatment (McNemar test,18/40 vs. 38/40, p , .0001), which was retained after letter-level treatment(McNemar test 18/40 vs. 40/40, p , .0001).

Reading times for the words in the treated set for both patients are providedin Figure 5. Normality tests conducted on DM’s correct reading time data (leftpanel) were not significant (Shapiro-Wilk test p . .8), and so the data wereanalysed using a univariate ANOVA. DM’s data revealed a highly significant

Figure 4. MAH’s reading accuracy on the treated set across baselines and confusability conditions.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

effect of baseline, F(2, 114) ¼ 16.027, p , .0001, d ¼ 0.730. Reading timesimproved after both therapies compared with initial baseline (word-leveltreatment, t(78) ¼ 4.270, p , .0001, d ¼ 0.9171; letter-level treatment,t(78) ¼ 4.448, p , .0001, d ¼ 0.9311, although there were no significantdifferences in RTs between the two post-therapy baselines, p ¼ .845. Therewas also significant effect of confusability, F(2, 114) ¼ 5.613, p ¼ .02,d ¼ 0.043, due to the better performance with low relative to high confusa-bility words overall, t(118) ¼ 2.118, p ¼ .036, d ¼ 0.477. This reflected anon-significant trend at initial baseline, t(38) ¼ 1.575, p ¼ .062, d ¼ 0.423,but not after word-level treatment (p ¼ .581), and a significant advantagefor low confusability words in the set after letter-level therapy, t(38) ¼2.425, p ¼ .020, d ¼ 0.531.

MAH’s data were normally distributed (Shapiro-Wilk test ps . .05), andso were also examined using univariate ANOVA. MAH showed a significanteffect of baseline, F(2) ¼ 4.761, p ¼ .010, d ¼ 0.476, with significantlyfaster RTs after word-level therapy, t(78) ¼ 2.633, p ¼ .010, d ¼ 0.59, andletter-level treatment, t(78) ¼ 2.474, p ¼ .016, d ¼ 0.477, compared toinitial baseline. There was no significant effect of confusability (p . .8).

In summary, both therapies improved DM and MAH’s reading times. DMshowed an advantage for low over high confusability items while MAHshowed similar performance on low and high confusability words.However, these data are based on a restricted set of 40 items, and so theabsence of confusability effects may be due to repeated training improvingperformance on all words. The next section describes performance onuntreated sets, assessing confusability, neighbourhood and length effects ateach baseline phase.

Figure 5. Mean reading times for the treated set across baselines (for each data point, N ¼ 20). DM

left panel, MAH, right panel). Error bars are based on 95% confidence intervals.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

Generalised improvement to untreated items following each treatment

Reading performance of the patients on untreated words was tested at eachbaseline phase, focusing on a comparison of generalised improvements withlow and high confusability words.

Confusability effect across treatments

DM performed at ceiling under all conditions, making just three errors intotal, all of which were phonologically similar word substitutions. MAH’saccuracy rates across low and high confusability words are provided inFigure 6.

MAH’s accuracy data over the baseline phase were analysed using a log-linear analysis. There was a significant interaction between test time, confu-sability and score, x2(2) ¼ 31.141, p , .0001 d ¼ 0.345. We analysed theresults for the low and high confusability sets separately. For low confusabil-ity words, performance improved reliably after the word-level treatment,x2(1) ¼ 15.421, p , .0001, d ¼ 0.423, but not after the letter-level treatment(p ¼ .833), when compared to the initial baseline. Also there was a significantdecline in accuracy on low confusability words between word-level andletter-level treatment post-therapy baselines, x2(1) ¼ 13.837, p , .0001,d ¼ 0.399.

Figure 6. MAH’s accuracy scores for high and low-confusability words from the untreated set, across

baseline phases.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

When the results for high confusability words only were analysed perform-ance did not improve after the word-level treatment (Chi Square test, p ¼.358). However reading was better after the letter-level treatment when com-pared with both initial baseline, x2(1) ¼ 24.381, p , .0001, d ¼ 0.539, andpost-word-level treatments, x2(1) ¼ 16.386, p , .0001, d ¼ 0.437. MAH’sand DM’s mean RT data are provided in Figure 7.

Normality tests conducted on both patients’ correct reading time data werenot significant (Shapiro-Wilk test ps . .8), and so both data sets were sub-jected to within-subjects ANOVAs. DM’s RT data were entered into anANOVA with the factors confusability (low and high) and baseline phase(initial baseline, word-level treatment, letter-level treatment). There was asignificant interaction between confusability and baseline phase, F(2, 194)¼ 12.565, p , .001, d ¼ 0.2157. Bonferroni post-hoc tests for the low con-fusability set showed significant differences between initial baseline andafter both treatments (all ps , .0001). There was also a significant differencein performance on the low confusability words between the baseline after theword-level and letter-level treatments (p , .0001). For the high-confusabilitywords, there were no differences between the initial baseline and after theword-level treatment (p ¼ .178), but there was a significant improvement(reduced RTs) for the baseline after the letter-level treatment compared toboth the initial baseline (p , .0001) and performance after the word-level(p ¼ .005) treatment.

A similar analysis of MAH’s RT data revealed a significant interactionbetween confusability and baseline phase, F(2, 358) ¼ 16.327, p , .0001,d ¼ 0.2459. Bonferroni post-hoc tests showed a significant improvement

Figure 7. Patients’ mean correct reading times for high and low-confusability words from the

untreated set, across baseline phases (DM, left panel, MAH, right panel). Error bars are based on

95% confidence intervals.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

on the low confusable word set between initial baseline and performance afterthe word-level (p , .0001) and letter-level (p , .0001) treatments. No sig-nificant differences existed between performance at the pre-therapy baselineand performance following the letter-level treatment (p ¼ 1.00). For the highconfusability sets, no differences existed between the initial baseline and thebaseline that followed the word-level treatment (p ¼ .058), but there weresignificant improvements between the baselines that followed the word-level and letter-level therapies (p , .0001), and between initial baselineand post-letter-level therapy baseline (p , .0001). At initial baseline, no sig-nificant differences existed in reading times for high- and low-confusabilitywords (p ¼ .070) – a pattern which emerged after word-level treatment (p, .0001). After the letter-level treatment, no differences existed betweenthe high and low confusability sets (p ¼ .233).

Neighbourhood and confusability effects across treatments

Accuracy analyses

DM scored at ceiling across baseline phases, and so her accuracy data werenot considered further. MAH’s percent accuracy scores at each baseline phaseacross neighbourhood and confusability conditions are provided in Figure 8.

MAH’s data collected at each baseline point and across the neighbourhoodand confusability conditions were entered into a loglinear analysis. The analy-sis returned a significant four-way interaction between baseline phase, confu-sability (high or low), neighbourhood size (high or low) and accuracy, x2(2)¼ 41.576, p , .001. To unpack this interaction, separate loglinear analyseswere performed on the data from each baseline phase. As detailed in theExperimental investigations section, the initial baseline data showed a signifi-cant three-way interaction between confusability, neighbourhood size and

Figure 8. MAH’s percent accuracy scores across confusability and neighbourhood conditions for

each baseline test left to right (initial baseline, after word-level treatment, after letter-level treatment).


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

accuracy, x2(1) ¼ 28.029, p , .001. This reflected the significant facilitativeeffect of neighbourhood for the low confusability items, x2(1) ¼ 16.843, p ,

.001, but the deleterious effect of neighbourhood for the high confusabilityitems, x2(1) ¼ 11.562, p ¼ .001, Figure 8, left panel. We attribute this dele-terious effect of neighbourhood size for the high confusability words to heigh-tened interference from competing neighbours when letter discrimination waspoor.

When the data from the baseline phase that followed the word-level treatmentwere subjected to a loglinear analysis, a significant three-way interaction wasfound between confusability, neighbourhood size and accuracy, x2(1) ¼137.138, p , .001. As with the initial baseline data, this interaction was dueto facilitative effects of high N with low confusability words, x2(1) ¼ 83.198,p , .001, but detrimental effects of high N with high confusability words,x2(1) ¼ 55.648, p , .001, Figure 8, central panel. Therefore, while his accuracyimproved overall, MAH’s reading of high confusability words was still nega-tively affected by high neighbourhood size after word-level therapy.

When a loglinear analysis was performed on the data from the baselinephase after the letter-level treatment, there was also a significant 3-way inter-action, x2(1) ¼ 10.676, p ¼ .001. This was due to the facilitative effect of Nsize with low confusability words, x2(1) ¼ 41.026, p , .001, and a facilita-tive effect of N size with high confusability size that did not reach signifi-cance, x2(1) ¼ 3.192, p ¼ .074, Figure 8, right panel. In sum, theletter-level treatment was the most effective strategy for overcoming a dele-terious effect of N size in high confusability words. Improved performance onhigh relative to low neighbourhood words is thought to indicate parallel pro-cessing (Arguin & Bub, 2005).

Figure 9. DM’s mean correct RT across confusability and neighbourhood conditions for each

baseline test left to right (initial baseline, after word-level treatment, after letter-level treatment).

Error bars are based on 95% confidence intervals.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

RT analyses

MAH achieved fairly low scores in some of the neighbourhood by confu-sability conditions (Figure 9) and so correct RT analyses were not possible.DM’s mean reading latencies over confusability and neighbourhood con-ditions across baseline phases are provided in Figure 9.

Normality tests conducted on DM’s correct reading time data were not sig-nificant (Shapiro-Wilk test ps . .2) The same words were repeated at eachbaseline, and so DM’s RT data from each condition and at each baselinephase were entered into a between-subjects ANOVA, using the factors con-fusability, neighbourhood size and baseline phase. The analysis revealed asignificant three-way interaction between time, confusability and N size,F(2, 126) ¼ 44.762, p , .001.

Bonferonni tests revealed that, at the initial baseline, high N words wereread faster than low N words for the low confusability sets (p , .001), butthe opposite pattern was seen with high confusability words (p ¼ .027).After both word-level and letter-level treatments, RTs were significantlyfaster for high N words in both low and high confusability sets (all ps ,

.001, Figure 9). These results suggest that word-level and serial-readingapproaches were similarly effective in producing facilitative effects of Nsize, indicative of parallel processing, for high confusability words.

Length effects across treatments

In order to test for changes in the length effect over the treatments, perform-ance on 4, 5, 6, and 7- letter words was analysed in both cases over baselinephase. The sets were matched for frequency, confusability, and neighbour-hood size (Appendix D). MAH’s accuracy data across length and baselinephase are provided in Figure 10.

Figure 10. MAH’s percent accuracy scores across word lengths for each baseline test left to right

(initial baseline, after word-level treatment, after letter-level treatment).


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

A loglinear analysis was performed on MAH’s accuracy data across lengthand baseline phase. Although there appeared to be a levelling out of thelength effect in accuracy at the post-word-level therapy baseline (Figure 10),there was no significant three-way interaction between accuracy, length andbaseline phase (p ¼ .564). Overall, in the three-way analysis, there were signifi-cant two-way interactions between accuracy and length, x2(6) ¼ 49.463, p ,

.0001, and accuracy and baseline phase, x2(2) ¼ 13.572, p ¼ .001. Theformer interaction reflects the decline in accuracy over length, observed tosome extent, at all baseline phases (Figure 10). To unpack the accuracy × base-line phase interaction, Chi Square tests were used to compare performancebetween baseline phases. There was no significant improvement in accuracyoverall between the initial baseline and the baseline that followed the word-level therapy (p ¼ .127). However, there were significant improvementsbetween scores at the initial baseline and after the letter-level treatment, x2(1)¼ 12.168, p , .0001, and between the two post-therapy baselines, x2(1) ¼4.271, p ¼ .039. This indicates that, for MAH, letter-level therapy was moresuccessful, relative to a word-level therapy, in improving reading accuracy.

The data from each baseline phase were entered into three separate log-linear analyses, with length and accuracy as the factors. A loglinear analysison the data from the post-letter therapy baseline showed no significant inter-action between accuracy and length (p ¼ .416), despite significant two-wayinteractions at initial baseline, x2(4) ¼ 24.114, p , .0001, and at the post-word therapy baseline, x2(4) ¼ 28.169, p , .0001. These results suggestthat, for longer words, letter-level therapy was more successful in improvingMAH’s reading accuracy than word-level treatment. This may be due to thetraining speeding the serial identification of each letter, or it may be due toenhanced letter identification facilitating the parallel processing of theletters in the words. Whichever the case, the result occurred irrespective ofletter confusability. This finding contrasts with the results from the word-level

Figure 11. DM’s correct RT scores across word lengths for each baseline test left to right (initial

baseline, after word-level treatment, after letter-level treatment). Error bars are based on 95%

confidence intervals.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

treatment, which promoted efficient processing only when the perceptualdemands allowed (e.g., low confusability). In sum, the letter-level treatmentwas more successful in diminishing the length effect than the word-leveltreatment for MAH. DM’s mean correct RT data across word length and base-line phase are provided in Figure 11.

Normality tests conducted on DM’s correct reading time data were not signifi-cant (Shapiro-Wilk test p . .1) and so DM’s RT data were analysed acrosslengths and baseline phase using a within-subjects ANOVA, with the factorslength (4, 5, 6, and 7) and baseline phase (initial baseline, after word therapy,after letter therapy). The analysis returned a significant interaction between base-line and length, F(6, 1021) ¼ 2.672, p ¼ .014. There were also highly signifi-cant main effects of both factors, length: F(3, 1021) ¼ 45.541, p , .0001, andbaseline phase: F(2, 1021) ¼ 20.681, p , .0001. The effect of baseline phaseon RT reflects the facilitated mean RT across baseline sessions (Figure 11). Bon-ferroni post-hoc analyses performed on these data revealed a significant effect oflength at initial baseline (p , .0001) and at the post-word therapy baseline (p ,

.0001), but not at the post-letter therapy baseline (p ¼ .154). The mean differ-ences in DM’s reading times of 4 and 7 letter words were 827ms at initial base-line, 627ms after word-level treatment, and 214ms after the letter-leveltreatment, although reading times were slowest on 6-letter words (Figure 11).Although the effects of word length remained substantial they did reduce afterthe letter-level treatment compared to the other baselines (Figure 11). Similar

Figure 12. The distribution (% of total errors) of MAH’s errors across baseline sessions. Data are

from the untreated reading set.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

to MAH, DM’s performance improved over baseline phase, and again, the letter-level therapy led to a remediation of the length effect post-therapy.

MAH’s error types across baseline tests

Figure 12 shows the distribution of error types for MAH at each baseline phase.When the data were entered into a loglinear analysis, a three-way inter-

action was found between the factors error type frequency, accuracy, andbaseline phase, x2(4) ¼ 28.906, p , .0001.

Phonologically similar responses were the most frequent error type forMAH at each baseline. When the phonological error frequency data wereentered into a loglinear analysis, there was a significant three-way interactionbetween error type, accuracy, and baseline phase, x2(4) ¼ 28.906, p ,

.0001. This interaction reflected a significant reduction in phonologicalerrors between the initial baseline and after the word-level treatment (log-linear analysis on the phonological error data with the factors accuracy andbaseline phase, x2(1) ¼ 7.760, p ¼ .021. In contrast to this there was no sig-nificant difference in the frequency of phonological errors following letter-level treatment, when compared to the baseline that followed the wordtherapy; loglinear analysis on the phonological error data with the factorsaccuracy and baseline phase: p ¼ .098; nor was there a difference betweenperformance after the letter-level treatment and the initial baseline; loglinearanalysis on the phonological error data with the factors accuracy and baselinephase: x2(1) ¼ 1.347, p ¼ .510.

Similar analyses were performed using the total semantic errors. There wasagain a significant interaction between baseline phase, accuracy, and fre-quency of semantic errors, x2(2) ¼ 18.236, p , .0001. The increase insemantic errors following word-level therapy was significant; loglinear analy-sis on the semantic error data with the factors accuracy and baseline phase(initial baseline vs. word-level), x2(1) ¼ 12.362, p ¼ .002, and there was asignificant decline in semantic errors when comparing word-level andletter-level baseline phases; loglinear analysis on the semantic error datawith the factors accuracy and baseline phase (letter-level vs. word-level)x2(2) ¼ 12.735, p ¼ .002. Comparisons of semantic errors at the initial base-line and after letter therapy were not significant; loglinear analysis on thesemantic error data with the factors accuracy and baseline phase (initial base-line vs. letter-level), p ¼ .999.

GENERAL DISCUSSION

These results demonstrate differential effects of word-level and letter-leveltherapy on reading in one patient with pure alexia and another with length-sensitive reading in the context of wider deficits in language. The word-level


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

treatment improved reading performance in terms of both reading accuracyand speed for MAH, and for reading speed in DM. This applied to alltreated items and to untreated items containing low, but not high, confusabil-ity letters. The letter-level treatment led to improvements in reading perform-ance on both untreated low and high confusability words. From these findings,we conclude that the word-level therapy improved parallel processing ofletters, generating a general benefit for words with low confusability letters.The effect of the letter-level treatment however was to produce a generalisedimprovement to words with both low and high confusability letters. Thesefindings suggest either that letter-level treatment improved parallel wordidentification or promoted the speed and accuracy of letter-by-letterreading. A key feature of the letter-level treatment was to highlight dis-tinguishing perceptual features of letters, and this may have been crucial inovercoming the letter confusability effect in the patients.

DM

DM, the pure alexic patient, showed highly accurate, but slow and length-sensitive reading at initial baseline. Word-level therapy did not facilitateher performance in every reading condition, producing facilitated RTs onlyon low-confusability words. Letter-level treatment, on the other hand, wasextremely effective and improved DM’s mean reading time substantiallyacross different word types.

When initially tested, DM showed a pattern of facilitative neighbourhoodsize (N) effects with low confusability words, and a deleterious effect of highN with high confusability words. These initial effects were attributed to anability to process words in parallel for words with low confusability letters(i.e., low perceptual demands), and reliance on a serial strategy with high con-fusability letters (high perceptual demands), (see also Arguin & Bub, 1994,2005). However, after both treatments there were facilitative effects of Nsize in both confusability conditions (Figure 9). This finding is at odds withthe lack of generalised improvement to high confusability words after theword-level treatment observed in the overall analysis. As N effects are indica-tive of parallel processing (Arguin & Bub, 2005), these data suggest someimprovement in parallel processing even on high confusability words afterthe word-level therapy. The divergent patterns of N effects evidencedbetween these patients over baseline sessions could be explicable by thedifferences inherent in the patients’ clinical profiles. For instance, DM’sreading performance was better relative to MAH’s, and so the effect of thefirst treatment may have been effective enough to produce parallel processingeven on the high confusability words, but only in high N conditions. There-fore, high confusability items are still problematic, but words with high Nare recognised faster.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

MAH

MAH had wider-spread language problems relative to DM, including someapparent semantic deficits and deficits in name retrieval, in addition to hisreading impairment. MAH showed some treatment gains in reading accuracyfor both treated and untreated items, after both word-level and letter-leveltherapies. As with DM, generalised improvement was modulated by item con-fusability and treatment type.

The direction of the neighbourhood effect in MAH’s accuracy data dif-fered as a function of condition and baseline phase. While low confusabilitywords showed a facilitative effect of N size throughout, MAH’s reading ofhigh confusability words was detrimentally affected by high N size. Thispattern was present at initial baseline and remained after word-levelreading treatment. However, the letter-level treatment produced an advantagefor high N words comprised of both low and high-confusability letters(Figure 9). So, for DM, high neighbourhood items were facilitated afterword treatment and this was maintained after letter-level therapy. ForMAH, high neighbourhood words remained problematic after the word-level treatment, but were facilitated after the letter-level therapy. Thisfinding was not, however because there was no improvement after word treat-ment for MAH, as word treatment did benefit low neighbourhood words. Wepropose that for MAH, either by promoting parallel processing or by improv-ing the speed and accuracy of serial letter processing, the letter-level treat-ment improved reading of confusable letters. Therefore, where wordscomprised of high confusability words showed a deleterious effect of highneighbourhood at baseline (likely due to increased interference from compet-ing neighbours where letter identification is poor), that persisted after theword-level treatment, the letter-level treatment improved letter identificationsuch that reading was facilitated by high neighbourhoods.

Theoretical implications

The findings have some general implications for theories of LBL reading. Pre-therapy there was evidence of parallel processing in these patients, as Neffects interacted with confusability. Although this pattern of results followedthe pattern demonstrated by three alexic patients in Arguin and Bub (2005),no other studies to our knowledge have directly investigated the N effect inalexic patients (although note that, if the N effect interacts with confusability,mixed sets of confusability words may obscure any effect of N in descriptionsof these patients).

A number of studies have reported improved parallel processing in LBLreaders as a function of word-level therapy (e.g., Coslett et al., 1993).However, investigations of parallel processing following letter-level


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

treatment are relatively sparse. Arguin and Bub’s (1994) serial reading reha-bilitation study found improvement on whole word, but not serial reading,suggesting that improving a serial strategy may enable parallel processing.Sage et al. (2005) showed that the reading of untreated items only improvedafter the letter-level treatment. The current data are consistent with theseprevious findings. Both of the current patients showed improved performanceon only low confusable sets after a word-level therapy, but both high and lowconfusable words improved after letter-level therapy. The results are consist-ent with the idea that letter-level treatments particularly improve the identifi-cation of high confusability words. Further, the reduction of the length effectin the reading behaviour of both patients indicates that letter-level treatmenteither facilitated parallel letter processing and word identification or moreefficient serial identification of letters. In contrast, word-level therapyimproved performance on low but not high confusability words which weattribute to the treatment being geared at processing the whole word form,which is effective under conditions of low but not high perceptualdemands. In contrast the letter-level therapy improved letter discrimination,which helped remediate the effect of letter confusability on reading. Thesedata are helpful for understanding the nature of the functional deficit inthese patients. The finding that the largest benefit observed was due toletter-level treatment, along with the strong effects of confusability atinitial baseline, suggests that employment of a letter-by-letter reading strategymay be driven by a pre-lexical perceptual deficit.

Distribution of MAH’s error types over baseline phases

While DM made very few errors throughout the sessions, MAH’s profile wasless clear-cut. At initial baseline he produced frequent semantic errors withauditory and picture inputs but never when reading written words, where pho-nological errors dominated. At least one letter-by-letter reading patientreported in the neuropsychological literature has shown aspects of a deep dys-lexic profile, suggesting that a serial processing strategy may have developedin response to partially recovered deep dyslexia (Buxbaum & Coslett, 1996).It has also been noted that alexic patients may initially produce semanticerrors in reading which may decrease over time, as patients adopt a letter-by-letter reading strategy (Landis, Graves, & Goodglass, 1982). Strikingly,word-level treatment led to an increase in semantic errors in MAH. Thesudden appearance of semantic errors after word-level treatment was alsofound in Sage et al.’s patient FD, whose error types transformed from 60%to 10% omissions, 20% to 40% visual, and 3% to 17% semantic errorswhen compared with the initial baseline and post-word-level treatment per-formance. Letter-level treatment brought about the preponderance of


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

omission errors observed at initial baseline. MAH’s data closely replicatesthis pattern.

We suggest that the therapy reduced the likelihood of MAH using a serialreading process, responding instead to information coded across the wholeword. Coslett and Saffran (1992) have argued that the semantic errors canreflect right hemisphere reading, and letter-by-letter reading a compensatoryleft-hemisphere strategy. According to this argument, the word-level trainingencouraged the use of right hemisphere reading at the expense of the lefthemisphere compensatory process. It would clearly be interesting to assessthis possibility using functional brain imaging.

In contrast to the effects after word-level therapy, semantic errors were nolonger produced after letter-level therapy. This is consistent with letter-leveltreatment leading to more accurate reading responses based on facilitatedletter identification and naming, perhaps operating through the left hemi-sphere. It is interesting to note that irregular words were included in thetreated set. This may have been important for MAH, forcing him to rely onletter identification rather than sounds to support his reading (as letter–sound processing would lead to regularisation errors – no regularisationerrors were observed). Nevertheless, given MAH’s generally poor spokenrelative to written word production, the data indicate a general problemwith spoken output in addition to the apparent problems in visual aspectsof reading, treated here. These additional output problems likely served tolimit improvement – especially for the word-level reading strategy if thisled to MAH using a lexical–semantic reading route reliant on impaired pho-nological retrieval.

Study limitations

MAH and DM presented with very different cognitive profiles. In particular,MAH presented with several additional language deficits that may haveaffected the outcome of therapy. For instance, it is possible that some ofMAH’s difficulties in reading longer words could be due to the greater rela-tive difficulty of producing these words pre-therapy. Indeed, although thetherapies aimed to improve the visual processing of words, the sessionsalso trained word and letter production. This could explain some of the differ-ences in therapy outcomes between the patients, as MAH showed a betterremediation of the length effect than DM, due possibly to MAH’s improvedletter/word production accuracy post-therapy. We maintain that he alsoshowed pre-lexical deficits, demonstrated by his impaired performance onletter identity tasks and his pronounced effect of letter confusability pre-therapy, and that the letter-level strategy was successful in improvingletter-discrimination (inferred by the absent letter confusability effect afterthis treatment). The current data suggest generalisability of treatment


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

effects, showing that reading therapy, and the letter-level strategy in particu-lar, can improve reading in patients with wider cognitive deficits as well as inpure alexia.

Following Arguin and Bub (2005), an effect of orthographic neighbour-hood size was taken to indicate parallel letter processing. However, Fisetet al. (2006) showed that it is the number of phonographic neighbourhoods(words differing in one letter and one phoneme, e.g. race–rate) rather thanorthographic neighbourhoods (race–rack) that exerts a facilitatory effect onword naming. We did not control for this variable, and further work isneeded to investigate the role of phonological information in the facilitatoryeffect of N in LBL readers. Nonetheless, our initial baseline data are consist-ent with previous findings of high-level effects in letter-by-letter reading(Arguin, Bowers, & Bub, 1996; Arguin, Fiset, & Bubb, 2002; Montant &Behrmann, 2001), particularly effects of lexical frequency and imageability,in words comprised of low confusability letters (Arguin et al., 2002; Fisetet al., 2006), and that these effects may be brought about on high confusabilityitems following targeted therapy.

Conclusion

The current paper described two patients with signs of letter-by-letter readingwho showed effects of letter confusability and orthographic neighbourhoodon reading at initial baseline. A word-level treatment improved performanceon low-confusability words, but only the letter-level treatment was successfulin producing facilitated performance for both high confusability and low confu-sability words in the patients. These results are consistent with research showingthat letter confusability contributes significantly to letter-by-letter behaviour.

The deleterious effect of high N words in high confusability sets can beattributed to increased competition following the serial letter processing strat-egy required to decipher these words, where interference from orthographicneighbours with shared letters to those already processed exert a detrimentaleffect. In high confusability word sets, letter-level therapy, but not word-leveltherapy, led to a facilitative effect of N on accuracy in MAH, whereas theword-therapy was sufficient to produce a supportive effect of high N sizeon RT in DM, and this effect persisted after letter-level treatment.

Only the letter-level treatment was successful in remediating the wordlength effect in both patients. Taken together, the confusability effects atinitial baseline, and the highly beneficial effect of the letter-level therapy,lead us to attribute the patients’ letter-by-letter reading to damage at a pre-lexical level, i.e., in letter perception and discrimination processes that areneeded before a word can be read.

The current study demonstrated that a therapy geared at overcomingconfusability through training letter discrimination was more successful than


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

a word-level approach, giving better generalisation over different word sets.The letter-based therapy also reduced RTs to the same degree as the word-level approach, and reduced the length effect, suggesting that it either facilitatedparallel word processing or a more efficient serial word identification process.Improving perceptual abilities through improving the speed and accuracy ofletter-level appeared to be the most effective strategy for letter-by-letter readers.

REFERENCES

Andrews, S. (1989). Frequency and neighbourhood effects on lexical access: Activation or search?

Journal of Experimental Psychology: Learning, Memory and Cognition, 15, 802–814.

Andrews, S. (1992). Frequency and neighbourhood effects on lexical access: Lexical similarity

or orthographic redundancy? Journal of Experimental Psychology: Learning, Memory and

Cognition, 18, 234–254.

Arguin, M., & Bub, D. (1994). Pure alexia: Attempted rehabilitation and its implications for

interpretation of the deficit. Brain and Language, 47, 233–268.

Arguin, M., & Bub, D. (1997). Lexical constraints on reading accuracy in neglect dyslexia. Cog-

nitive Neuropsychology, 14, 765–800.

Arguin, M., & Bub, D. (2005). Parallel processing blocked by letter similarity in letter-by-letter

dyslexia: A replication. Cognitive Neuropsychology, 22, 589–602.

Arguin, M., Bub, D., & Bowers, J. (1998). Extent and limits of covert lexical activation in letter-

by-letter reading. Cognitive Neuropsychology, 15, 53–92.

Arguin, M., Bowers, J. S., & Bub, D. (1996). Implicit lexical access in letter-by-letter reading is

not mediated by the right hemisphere. Brain and Cognition, 30, 275–277.

Arguin, M., Fiset, S., & Bub, D. (2002). Sequential and parallel letter processing in letter-by-

letter dyslexia. Cognitive Neuropsychology, 19, 535–555.

Beeson, P. M. (1998). Treatment for letter-by-letter reading: A case study. In N. Helm-Estab-

rooks & A. L. Holland (Eds.), Clinical decision making in aphasia treatment (pp. 153–177).

Madrid: Singular Press.

Buxbaum, L., & Coslett, H. B. (1996). Deep dyslexic phenomenon in a letter-by-letter reader.

Brain and Language, 54, 136–167.

Chialant, D., & Caramazza, A. (1998). Perceptual and lexical factors in a case of letter-by-letter

reading. Cognitive Neuropsychology, 15(1/2), 167–201.

Coslett, H. B., & Saffran, E. M. (1989). Preserved reading in pure alexia. Brain, 112, 327–359.

Coslett, H. B., & Saffran, E. M. (1992). Optic aphasia and the right hemisphere: A replication

and extension. Brain and Language, 43, 148–161.

Coslett, H. B., Saffran, E. M., Greenbaum, S., & Schwartz, H. (1993). Reading in pure alexia:

The effect of strategy. Brain, 116, 21–37.

Fillingham, J. K., Hodgson, C., Sage, K. E., & Lambon Ralph, M. A. (2003). The application of

errorless learning to aphasic disorders: A review of theory and practice. Neuropsychological

Rehabilitation, 13, 3.

Fiset, D., Arguin, M., & McCabe, E. (2006). The breakdown of parallel processing in letter-by-

letter dyslexia. Cognitive Neuropsychology, 23(2), 240–260.

Geschwind, N. (1965a). Disconnexion syndromes in animals and man. I. Brain, 88, 237–294.

Geschwind, N. (1965b). Disconnexion syndromes in animals and man. II. Brain, 88, 585–644.

Gilmore, G. C., Hersh, H., Caramazza, A., & Griffin, J. (1979). Multidimensional letter simi-

larity derived from recognition errors. Perception and Psychophysics, 25, 425–431.

Howard, D., & Patterson, K. (1992). Pyramids and Palm Trees: A test of semantic access from

pictures and words. Bury St Edmunds, UK: Thames Valley Test Company.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

Humphreys, G. W., Riddoch, M. J., & Price, C. J. (1997). Top-down processes in object identi-

fication: Evidence from experimental psychology, neuropsychology and functional

anatomy. Philosophical Transactions of the Royal Society of London Series B: Biological

Science, 352, 1275–1282.

Kay, J., Lesser, R., & Coltheart, M. (1992). PALPA: Psycholinguistic Assessments of Language

Processing in Aphasia. Hove, UK: Lawrence Erlbaum Associates.

Lacey, E. H., Lott, S. N., Snider, S. F., Sperling, A., & Friedman, R. B. (2010). Multiple oral re-

reading treatment for alexia: The parts may be greater than the whole. Neuropsychological

Rehabilitation, 20, 601–623.

Landis, T., Graves, R., & Goodglass, H. (1982). Aphasic reading and writing: Possible evidence

for right hemisphere participation. Cortex, 18, 105–112.

Loomis, J. M. (1982). Analysis of tactile and visual confusion matrices. Perception and Psycho-

physics, 31, 41–52.

Lott, S. N., Carney, A. S., Glezer, L. S., & Friedman, B. (2010). Overt use of a tactile/kinaes-

thetic strategy shifts to covert processing in rehabilitation of letter-by-letter reading. Apha-

siology, 24, 11.

Montant, M., & Behrmann, M. (2001). Phonological activation in pure alexia. Cognitive Neu-

ropsychology, 18, 697–727.

Moody, S. (1988). The Moyer reading technique re-evaluated. Cortex, 24, 474–476.

Moyer, S. B. (1979). Rehabilitation of acquired alexia: A case study. Cortex, 15, 139–144.

Mueller, S. T., & Weidemann, C. T. (2012). Alphabetic letter identification: Effects of percei-

vability, similarity, and bias. Acta Psychologica, 139, 19–37.

Osswald, K., Humphreys, G. W., & Olson, A. (2002). Words are more than the sum of their

parts: Evidence for detrimental effects of Whole-word information in alexia. Cognitive Neu-

ropsychology, 19, 675–695.

Price, C. J., & Humphreys, G. W. (1992). Letter-by-letter reading: Functional deficits and com-

pensatory strategies. Cognitive Neuropsychology, 9, 427–457.

Pugh, K., Rexer, K., Peter, M., & Katz, L. (1994). Neighborhood effects in visual word recog-

nition: Effects of letter delay and nonword context difficulty. Journal of Experimental Psy-

chology: Learning, Memory and Cognition, 20, 639–648.

Riddoch, M. J., & Humphreys, G. W. (1993). BORB: Birmingham Object Recognition Battery.

London: Psychology Press.

Sage, K., Hesketh, A., & Lambon Ralph, M. A. (2005). Using errorless learning to treat letter-

by-letter reading: Contrasting word versus letter-based therapy. Neuropsychological Reha-

bilitation, 15, 619–642.

Sears, C. R., Hino, Y., & Lupker, S. J. (1995). Neighborhood size and neighborhood frequency

effects in word recognition. Journal of Experimental Psychology: Human Perception and

Performance, 21, 876–900.

Shallice, T., & Saffran, E. (1986). Lexical processing in the absence of explicit word identifi-

cation: Evidence from a letter-by-letter reader. Cognitive Neuropsychology, 3, 429–458.

Snodgrass, J. G., & Mintzer, M. (1993). Neighborhood effects in visual word recognition: Facil-

itatory or inhibitory? Memory and Cognition, 21, 247–266.

Townsend, J. T. (1971). Theoretical analysis of an alphabetic confusion matrix. Perception and

Psychophysics, 9, 40–50.

Tuomainen, J., & Laine, M. (1991). Multiple oral rereading technique in rehabilitation of pure

alexia. Aphasiology, 5, 401–409.

Van Der Heijden, A. H. C., Malhas, M. S. M., & Van Den Roovaart, B. P. (1984). An empirical inter-

letter confusion matrix for continuous-line capitals. Perception and Psychophysics, 35, 85–88.

Warrington, E. K., & James, M. (1991). Visual object and space perception battery. Bury

St. Edmunds, UK: Thames Valley Test Co.


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

Wilson, B. A., Baddeley, A. D., Evans, J. J., & Shiel, A. (1994). Errorless learning in the reha-

bilitation of memory impaired people. Neuropsychological Rehabilitation, 4, 307–326.

Manuscript received June 2012

Revised manuscript received February 2013

First published online February 2013

APPENDIX

APPENDIX AM scores on lexical variables for the confusability ∗ N size stimuli used in the Experimental

Investigation sections, and as the untreated set in the Rehabilitation Method

Confusability

Celex

Frequency Regularity Length Imageability

Orthographic

N Size N

Low

Confusability,

Low

Neighbourhood

0.26 13.68 20 regular 5.13 381.85 .035 108

Low

Confusability,

High

Neighbourhood

0.27 15.53 18 regular 5.55 388.33 6.528 72

High

Confusability,

Low

Neighbourhood

0.53 15.31 22 regular 5.23 336.89 0.341 82

High

Confusability,

High

Neighbourhood

0.53 15.89 19 regular 5.80 351.04 5.568 81

APPENDIX BM scores across lexical variables for the treated set

Confusability

Celex


Orthographic

N Size N

Low

Confusability

0.28 30.129 12 regular M ¼ 4.5

Range ¼

4-5

331.15 7.4 20

High

Confusability

0.53 39.73 10 regular M ¼ 4.5

Range ¼

4-5

332.1 8.4 20


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

APPENDIX DM scores across lexical variables for the untreated letter length experiment

Letter

Length Confusability

Celex

Frequency Regularity Imageability

Orthographic

N Size N

4 0.38 15.16 40 regular 292.9 4.18 80

5 0.43 12.77 50 regular 257.5 2.06 80

6 0.37 16.58 45 regular 195.3 1 80

7 0.47 12.702 42 regular 164.9 0.41 80

APPENDIX CM scores across lexical variables for the untreated confusability experiment

Confusability

Celex


Orthographic

N Size N

Low

Confusability

0.26 18.87 42 regular M ¼ 5.5

Range ¼

4-7

201.54 2.8 180

High

Confusability

0.53 17.74 53 regular M ¼ 5.5

Range ¼

4-7

212.83 3.07 180


Dow

nloa

ded

by [

Uni

vers

ity o

f W

aika

to]

at 0

8:31

10

July

201

4

Date post:	27-Jan-2017
Category:	Documents
Upload:	glyn
View:	215 times
Download:	0 times

Overcoming the effect of letter confusability in letter-by-letter reading: A rehabilitation study

Documents