Does poor spelling equate to slow reading? The relationship between reading, spelling, and orthographic quality

Page 1

Does poor spelling equate to slow reading?The relationship between reading, spelling,and orthographic quality

Sandra Martin-Chang • Gene Ouellette •

Melanie Madden

Published online: 16 February 2014! Springer Science+Business Media Dordrecht 2014

Abstract High quality lexical representations in memory, characterized byaccuracy and stability, are said to underpin fluent reading. Here, the relationshipbetween orthographic quality and reading speed was examined by asking under-graduates (N = 74) to repeatedly read and spell words. Spelling performance overfive trials indicated orthographic quality. Single word reading speed was measuredusing E-Prime technology. A within-participant repeated measures analysis revealedthat words which participants spelled consistently accurately, were read faster thanwords which were misspelled. This pattern also held in a within-word analysis; thesame words were read faster by individuals who always spelled them correctly,compared to those who did not. Further, it was found that when words were spelledusing the same incorrect letter patterns across trials (i.e., in the same erroneousway), they were read faster than when they had an incorrect but less stable repre-sentation (i.e., inconsistent spelling across trials). Hence, the difference in readingspeed appears to be a function of both the accuracy and stability of the orthographicrepresentations stored in memory, rather than due to characteristics of individualparticipants or words. These results lend support for a central role of lexical qualityin both spelling and reading, and are discussed with reference to the lexical qualityhypothesis.

Keywords Lexical quality ! Orthography ! Spelling ! Word readingspeed

S. Martin-Chang (&)Department of Education, Concordia University, 1455 de Maisonneuve Boulevard West, Montreal,QC H3G 1M8, Canadae-mail: [email protected]

G. Ouellette ! M. MaddenDepartment of Psychology, Mount Allison University, Sackville, NB, Canada

123

Read Writ (2014) 27:1485–1505DOI 10.1007/s11145-014-9502-7

Page 2

Introduction

When a word is stored in memory, it can be accessed in terms of how it sounds, howit is written, or the concepts it represents (Perfetti, 2007; Rosenthal & Ehri, 2011).However, the sounds in the spoken word are not always indicative of which lettersare contained in its spelling, and often a word’s lexical tag bears no directresemblance to its semantic meaning.1 Word learning, then, is the process ofcreating and refining these individual components of word identity, as well as thelinks between them. The concept that three aspects of word identity (sound, print,and meaning) are integral for expert reading is not new, in fact, it has formed acentral premise in both connectionist (the triangle model: see Plaut, McClelland,Seidenberg, & Patterson, 1996; Seidenberg & McClelland, 1989) and cognitivemodels (lexical quality hypothesis (LQH): see Perfetti & Hart, 2001). As aconceptual framework, it has gained popularity because it creates several testablehypotheses. To illustrate, it is hypothesized that phonology, orthography, andsemantics are all necessary contributors to lexical quality (Perfetti & Hart, 2001).An extension of this logic would suggest that a deficiency in any one of thesecomponents would hamper lexical access and result in reduced word reading. Forexample, words with lower quality orthographic representations should be accessedless efficiently and consequently read more slowly, than words with higher qualityrepresentations. One way to measure orthographic quality is via spelling;consistently accurate spelling is taken to reflect high quality orthographicrepresentations. Yet whether orthographic quality directly influences reading speedremains unclear. It was this question that motivated the current investigation. Ourgoal was to examine the connection between the quality of orthographicrepresentations as indexed by spelling accuracy and consistency, and single wordreading speed.

The lexical quality hypothesis

The LQH describes how new word-specific representations are added to the mentallexicon, and once there, how they are refined (Perfetti, 1997; Perfetti & Hart, 2001).The LQH states that there are two main considerations to bear in mind whenconceptualizing lexical quality: (1) the quality of the components of word identitythemselves (phonological, orthographic, and semantic), and (2) the coherenceamong the components (Perfetti, 2007). For the present purposes we will befocusing on the first of these considerations: specifically, the quality of orthographicrepresentations.

Reichle and Perfetti (2003) describe a high quality word representation as onethat enables the reader to arrive at the word’s specific pronunciation (e.g., /s//p//I//l//t/), or meaning (e.g., ‘‘poured out’’), based on a given letter string (e.g., ‘spilt’). Incontrast, a low quality representation may only bring the reader to a near

1 E.g., ‘large’ is pronounced with four phonemes, and written with five letters; it signifies something big.‘Little’ is also pronounced with four phonemes however, it is written with six letters including a doubledconsonant separating a closed syllable and a final stable syllable. It signifies something small.

1486 S. Martin-Chang et al.

123

Page 3

approximation of the word (e.g., ‘sp_ _ t’ = ‘‘spilt’’, ‘‘split’’, ‘‘splat’’), which is thendependent on context to be directed to the correct pronunciation (e.g., ‘The milk wassp_ _ t’). A similar statement can be made for writing: a high quality representationis one that allows a writer to consistently spell a word accurately.2 It is notuncommon for children to fluctuate between the accurate and inaccurate spellings ofwords, sometimes within the same piece of writing (Frith, 1978). According toPerfetti (2007), this variability is a reflection of a lexical representation that is stilldeveloping. Spelling a word consistently is only possible when the letters andphoneme constituents become stable in memory; an orthographic representationwould be considered stable when it produces consistent spelling patterns overseveral instances.

According to the LQH, consistent, accurate spelling and fluent reading arereflections of the quality of, and the coherence among, the three facets of lexicalrepresentation (phonology, semantics, and orthography; Perfetti, 2007). There is aconsiderable body of research supporting the central role of phonology in spellingand reading (e.g., Melby-Lervag, Lyster, & Hulme, 2012). Indeed, even in proficientreaders, stressing the phonological aspects of words heightens orthographic andsemantic learning (Rosenthal & Ehri, 2011). In contrast, in cases such as selectivelanguage impairments when phonological representations are impoverished, readingdeficits are more likely (Catts, Bridges, Little, & Tomblin, 2008; Catts, Fey,Tomblin, & Zhang, 2002). Congruent with the LQH, there is also a growingliterature supporting the role of semantic representation in literacy skills (e.g.,McKay, Davis, Savage, & Castles, 2008; Ouellette & Fraser, 2009; Ouellette, 2010;Taylor, Plunkett, & Nation, 2011) and likewise, the role of orthographicrepresentations in verbal word learning (Burki, Spinelli, & Gaskell, 2012).However, while ‘orthography’ occupies a prominent place in many developmentalreading theories (e.g., Ehri, 1995; Perfetti, 2007; Share, 2008), the empiricalevidence detailing how orthographic representations contribute to skilled wordreading is far less conclusive.

Orthographic representations, spelling, and reading

The fact that incomplete orthographic representations may suffice for accuratereading has been referred to as ‘partial cue reading’ (Frith, 1980, 1985). Forexample, the word ‘‘occasion’’ may still be read, even if some information ismissing from the orthographic representation (e.g., oc?as?ion). In contrast, this lackof orthographic specification or ‘quality’ renders spelling difficult (Conrad, 2008;Holmes & Castles, 2001). Likewise, a representation may be complete yetinaccurate, preventing correct spelling but allowing enough of a partial match tovisual words to allow pronunciation in reading. The LQH predicts that having betterorthographic knowledge, while not mandatory for reading accuracy, shouldnevertheless be advantageous to the reader. Higher orthographic quality is predicted

2 Spelling that is accurate is written with the correct letters in the precise order according to localconvention. Hereafter, ‘accurate’ and ‘correct’ are used as synonyms.

Reading, spelling, and orthographic quality 1487

123

Page 4

to result in word reading that is more fluent and less influenced by context,compared to reading that stems from lower quality representations (Perfetti, 1997).

There are a limited number of studies that can be used to support this claim. Forexample, the relationship between spelling skill and top-down, context driven,cognitive processes was explored in a series of experiments by Andrews andcolleagues (e.g., Andrews & Bond, 2009; Andrews & Reynolds, 2013; Hersch &Andrews, 2012). The authors concluded that it was ‘‘good spelling rather than goodreading that was associated with reduced top-down reliance’’ (Andrews & Bond,2009, p. 707). With regards to speed of lexical access, Holmes and Castles (2001)matched good spellers and ‘unexpectedly poor spellers’ for word reading ability andvocabulary knowledge. They found that both groups were equally accurate whenasked to perform a lexical decision task, however, ‘unexpectedly poor spellers’ tooksignificantly longer to complete it.

While there is evidence to support the notion that ‘skilled spellers’ can recognizewords more quickly and are less reliant on top-down processing than less skilledspellers, it is important to note that the experimental designs listed above did notevaluate spelling for the same target words used in the reading tasks. In all cases,scores from general reading and spelling tests were used to form the experimentalgroups. Therefore, while acknowledging that lexical representations are developedincrementally on a word-by-word basis (e.g., Conrad, Harris, & Williams, 2013;Hersch & Andrews, 2012) these studies did not address whether the specificity of agiven lexical representation related directly to reading performance for that item.This type of detailed data is imperative to building theoretical models of skilledreading.

Two studies have directly examined the relationship between spelling accuracyand single word reading speed. Holmes and Carruthers (1998) gave adultparticipants a spelling test in which they had one opportunity to spell each word.These same words were then used in a silent reading task that required participantsto press a response key when they had finished. This design allowed Holmes andCarruthers to evaluate reading speed for each word, as a function of its spellingaccuracy. The results indicated that participants could read words that they couldnot spell just as quickly as words that they were able to spell accurately, which leadto the conclusion that reading proceeds unaffected by incomplete or inaccurateorthographic representations. This finding is in contrast to what would be predictedby the LQH and other developmental theories that give a prominent role toorthographic representations in reading fluency (e.g., Ehri’s, 1995 phase theory;Share’s, 2008 self-teaching theory).

Burt and Tate (2002) also directly evaluated the connection between spellingaccuracy and reading speed within a sample of skilled, adult readers. As was doneby Holmes and Carruthers (1998), the same words were used in both the spellingand reading assessments, allowing for a direct evaluation of the impact of thequality of orthographic representations as reflected by spelling performance on wordreading. Instead of the silent reading task employed in the earlier study, Burt andTate relied on a lexical decision task as a measure of single word reading fluency. Incontrast to Holmes and Carruthers, the findings from this paradigm revealed thatparticipants recognized words that they could personally spell faster than words they

1488 S. Martin-Chang et al.

123

Page 5

could not spell accurately. Thus, Burt and Tate’s data support the assertion thatwhile incomplete orthographic representations may suffice for reading accuracy,speed of word recognition is detrimentally affected by the lack of accurateorthographic representations, as inferred from their single trial spelling task.

Although the conflicting results of Holmes and Carruthers (1998) and Burt andTate (2002) may seem difficult to reconcile, it is worth noting that neither studyused a direct measure of word reading speed; rather this was inferred from a silentreading task in the former study and from a lexical decision task in the latterparadigm. While silent reading may be ecologically valid in reading research itpresents challenges when it comes to measurement. The measures used in thesestudies required other cognitive and motor demands (including manually pressing aresponse key), and thus are arguably not solely reflective of single word readingspeed. Moreover, both Burt and Tate (2002) and Holmes and Carruthers (1998)required participants to spell words correctly only once, which is not a sufficientindicator of lexical representation. According to Perfetti (1992, 2007) spelling aword accurately (or inaccurately) only once does not necessarily indicate the lexicalquality of that item. Rather, lexical quality is reflected in both the accuracy anddegree of stability that is observed as words are spelled over several trials. In thissense, a single spelling attempt cannot differentiate between words that wouldalways be spelled accurately from words that may be spelled correctly onlyoccasionally due to unstable representations. Moreover, a single exemplar does notallow one to distinguish between whether an incorrect spelling is due to a completebut erroneous representation (i.e., the word would always be misspelled in the sameway) or a very unstable representation (i.e., the word may be misspelled in differentways by the same individual). Therefore, further investigation is merited to(a) clarify whether the accuracy of orthographic representations impacts wordreading speed, and (b) to evaluate the influence of the stability of theserepresentations on word reading. In particular, it has yet to be directly evaluatedwhether having unstable representations (as reflected in inconsistent misspellingsacross trials) would result in faster or slower reading than what may occur as a resultof having stable, yet erroneous, representations (as reflected in misspellings that areconsistent across attempts) that would present the reader with a mismatch betweentheir stored representations and the visual word form encountered in print.

Current investigation

The goal of the present investigation was to examine the relationship between thequality (accuracy and stability) of orthographic representations in memory, asindexed by spelling accuracy over repeated trials, and directly measured wordreading speed. The LQH claims that spelling productions reflect the quality of theorthographic representation, and those words with high quality representationsshould be accessed more efficiently than words with lower quality representations.This hypothesis is directly tested by using spelling accuracy and stability asindependent variables and reading speed as the dependent variable in a within-participant design. Given that spelling a word correctly over a number ofopportunities has been suggested as a strong indicator of lexical quality (Perfetti,

Reading, spelling, and orthographic quality 1489

123

Page 6

1992), we defined ‘idealized spelling performance’ as spelling that was accurate andstable over five trials. Words that were always spelled correctly (i.e., on all fivetrials) were assumed to have the highest quality representations within our word set.Words that were never spelled correctly were taken to have the lower qualityrepresentations. Single word reading fluency was operationalized as word recog-nition that was precise and fast (Martin-Chang & Levy, 2006).

Method

Participants

Seventy-four undergraduate students (54 female, 20 male) with a mean chronolog-ical age of 19.44 years (SD = 1.72 years) took part in this study in exchange forcourse credit in an introductory psychology course at a small undergraduateuniversity in Eastern Canada. All participants were native English speakers.

Materials

Control measures

Each participant was administered the spelling subtest of the Woodcock-JohnsonTest of Achievement—Third Edition (WJ–III; Woodcock, McGrew, & Mathers,2001) to measure general spelling ability; the WJ-III has been shown to have highsplit-half reliability (median r = .90; Schrank, McGrew, & Woodcock, 2001). Inthis subtest, participants were asked to spell orally presented words of increasingdifficulty until six errors were made or until all items had been spelled.

Participants also completed the Test of Word Reading Efficiency (TOWRE:Wagner, Torgesen, & Rashotte, 1999). The reported reliability for the TOWRE isexcellent, with all forms reported above .90. For the first part of this test, eachparticipant read aloud as many real words as possible in 45 s, from a list of 104words. All participants subsequently read aloud as many non-words as possible in45 s, from a list of 63 non-words. These two subtests provide a word reading andphonemic decoding proficiency score, respectively.

Spelling accuracy and reading response time stimuli

In order to increase the likelihood that participants would experience difficultyspelling some of the items, 10 very difficult words were chosen from a list pilotedon 175 undergraduate students. These items were misspelled by more than half(60 %) of the students initially tested. In order to create a list that would offer arange of difficulties, 10 easier words were added to the stimuli. Ideally, these wordscould have been chosen from the piloted list. However, words that students foundless challenging to spell during the pilot study were not only more transparent (i.e.,they came closer to a one-to-one correspondence between sounds and letters), butthey were also used substantially more frequently in American English. Therefore,

1490 S. Martin-Chang et al.

123

Page 7

to avoid the confound of frequency effects on spelling, an additional 10 words werematched list-wise to the difficult items in terms of their frequency of use inAmerican English (Brysbaert & New, 2009). The additional 10 words were deemedless difficult to spell because they contained an equal number of letters and syllablesas the difficult items, but they had more consistent grapheme-to-phoneme pairings.Finally, an additional 10 words, that were very easy to read and spell, were includedin the list. These filler words were short and very common. They were selected tokeep participants from getting discouraged and were not scored or analyzed. All ofthe items can be found in Table 5 in the Appendix section.

Procedure

The participants completed the standardized tests (control measures); they thencompleted the reading response time task. Using E-Prime 2.0 (Schneider,Eschmann, & Zuccolotto, 2002), the words (20 target; 10 filler) were presented inrandom order, one at a time on a computer screen, preceded by a brief fixation point(?). Participants completed three randomized blocks, so that every target word wasread three times in total. Reading response time was recorded for each individualword by E-Prime via a connected voice key. The time was recorded from thepresentation of the word to the onset of the reader’s voice, which also triggered thedisappearance of the word from the computer screen. The researcher keyed in one ofthree codes after each word was read aloud: (1) correct, (2) incorrect, or (3) acomputer error (e.g., microphone was tripped by a sound other than the participantreading). This input on the keyboard then triggered the presentation of the next word(again preceded by the fixation point). While reaction time data tend to be positivelyskewed, the transformations used to reduce this characteristic pattern impact powerand Type II error rather, than Type I error (Ratcliff, 1993; Whelan, 2008).Following the recommendations of Ratcliff, the data were not transformed, resultingin a more conservative and stringent test of the current hypothesis. The data weretrimmed by removing responses that were faster than 100 ms under the contentionthat such reaction times cannot reflect the processes under study and thus are notconsidered legitimate (Luce, 1986). No responses exceeded 2000 ms. Invalid trialsdue to technical failures were also discarded. Further, only trials in which the wordwas read correctly were used to compute each participant’s mean reading timeacross trials for each word. In all, these procedures resulted in 5.5 % of thecomputed data points being discarded from the analyses. The distributions for eachword were then examined and 12 reading response times were identified as potentialoutliers exceeding a z-score criterion of 3.5 (see Cousineau & Chartier, 2010) andwere replaced with the next highest response time.

Participants returned 7 days later to complete a spelling dictation test of thewords. The spelling test was administered individually, one week after the readingtask to minimize any interference between the tasks. The spelling test was the onlytask administered in this second session so that the visual exposure to the spellingsgenerated by the participants (which may or may not be spelled correctly) could notinfluence response rates for that item in the reading task. During the spellingdictation the participants heard the words, presented one at a time, and immediately

Reading, spelling, and orthographic quality 1491

123

Page 8

wrote them down on numbered and lined paper. The test was self-paced. The 20target words were presented five times each, in random order, along with 10 fillerwords. During the testing, each word was concealed by an opaque cover after it wasprinted so that the participant could not reference previous spellings of a given wordon subsequent presentations. No feedback was given during this assessment.

Results

Standardized tests of reading and spelling were administered to ensure thatparticipants fell within the expected range. The standardized test scores also allowedfor a correlational analysis of skills in these areas and performance on the experimentalmeasures of reading response time and spelling. The means and standard deviationsfor performance on the TOWRE Word, Non-Word and the Woodcock-JohnsonSpelling Subtest are reported in Table 1, along with the mean performance on theexperimental reading response time task and the accuracy of spellings for theexperimental words. As evident from these data, the participants were functioningwithin an expected range and the literacy measures were moderately to stronglycorrelated with each other. The negative correlations with reading response rate reflectthe faster response time associated with higher reading and spelling scores. Of specialnote are the moderate (negative) correlations between reading response rate and boththe standardized and experimental measures of spelling. The experimental measure ofspelling was also highly correlated with the standardized spelling test.

Accuracy of orthographic representations and reading speed

Within-participant analysis

The first question of interest in this study was: Does reading speed differ as afunction of the accuracy of orthographic representations? To address this question,words were first classified according to their spelling accuracy (how many times outof 5 that they were spelled correctly). A spelling accuracy code of 5 indicated thatthe word was ‘always’ spelled correctly. An accuracy score of 0 indicated that theword was ‘never’ spelled correctly. Scores from 1 to 4 indicated that the word wasspelled accurately on some, but not all, of the trials. Mean reading times were thencalculated for each of these classifications. This was done separately for eachparticipant to reflect individual spelling performance (i.e., each participant had adifferent number of items that were spelled 0, 1, 2, 3, 4, or 5 times correctly andincorrectly, as well as different words in each classification). As there were very fewobservations in the 1, 2, 3, or 4 times correct classifications (and most participantshad no observations in at least one of these), only the always (5/5 trials spelledcorrectly) and never (0/5 trials correct) classifications were included in thefollowing analysis. The overall means and standard deviations for the always correctand never correct spelling accuracy classifications are shown in Table 2.

As indicated in Table 2, words that were always spelled correctly were readfaster than the words that were never spelled correctly. This pattern was confirmed

1492 S. Martin-Chang et al.

123

Page 9

in a within-participant, repeated measures analysis of variance that comparedreading speed as a function of spelling accuracy, F (1,73) = 9.20, p = .003,d = .363. The correlations in Table 1 indicate a moderate relation between overallspelling and reading performance, where better spelling is, on average, associatedwith faster reading speeds. In addition, this within-participant analysis demonstratesthat the association between orthographic quality, as indexed by spelling accuracyand reading speed, holds within the individual. Participants read words they couldpersonally spell, faster than words they could not.

However, as previously noted, in this analysis each participant had differentwords in the never and always categories, leaving open the question of whether thewords in the never category were made predominately of the words that weremisspelled by the majority of the undergraduates in the pilot study. This raises thequestion of whether the same words would be read faster by those who spelled themaccurately versus incorrectly. This query was explored in a within-word analysis.

Within-word analysis

An alternative way of examining these data is to ask whether the same words (e.g.,plagiarism, diaphragm) are read faster by people who can spell them accuratelycompared to those who cannot (i.e., a within-word analysis). To do this, meanreading times were again calculated for each spelling accuracy classification, but

Table 1 Descriptive statistics and correlation coefficients

Variable 1 2 3 4 5

1. TOWRE words –

2. TOWRE non-words .521** –

3. WJ-III spelling .432** .729** –

4. Reading ratea -551** -.685** -.623** –

5. Spelling accuracyb .294* .624** .756** -.540** –

Mean 96.22 55.05 51.43 614.46 3.26

SD 6.70 6.85 4.25 98.72 0.88

Range 79–104 30–63 40–59 456–890 1.2–5.0

* p \ .01; ** p \ .001a Reading response rate in ms averaged across all wordsb Number of correct spelling attempts out of 5, averaged across all words

Table 2 Overall mean reading response times by spelling accuracy

Spelling accuracy N Meana SDa

Always correct (5/5) 74 605.63 100.92

Never correct (0/5) 74 626.63 113.42

a Response time in milliseconds. The mean response time was calculated separately for each participant,for each spelling classification. The means reported here are averaged across participants

Reading, spelling, and orthographic quality 1493

123

Page 10

this time separately for each word. Note that reorganizing the data in this manneryields a different matrix of cell means as the focus is now within-words rather thanparticipants and the number of participants in each spelling classification may varyfrom one word to the next. Table 3 lists the mean reading times of each spellingclassification for each word, along with the number of participants in each, derivedfrom this data reorganization.

As evident in Table 3, some words were more likely to be misspelled than others,confirming the range of difficulty of the employed word set; words specificallychosen because of their expected difficulty were misspelled far more often thanwere the words anticipated to be only moderately difficult (F (1, 73) = 192.98,p \ .001, d = 2.03). This raises two competing explanations for the results reportedin the first set of analyses: (1) words with more accurate orthographic represen-tations are read faster than words with less accurate representations, and (2) wordswith many grapheme-to-phoneme mappings that are inconsistent or irregular mighttake both longer to read (e.g., Jared, 2002; Taraban & McClelland, 1987), and beharder to spell, than words with fewer inconsistent grapheme-to-phoneme pairings.That is, the extended reading times might be a consequence of word-specificcharacteristics rather than an outcome based on the accuracy of the orthographicrepresentation. A second set of analyses was conducted to tease apart thesecompeting hypotheses; here we asked whether the same words were read fasterwhen individuals could spell them correctly compared to when they could not. If so,specific word characteristics cannot be used to explain the difference in reading timebecause they are controlled for (it is a within-word comparison); therefore, theorthographic representation hypothesis would gain support.

This question was addressed in a within-items ANOVA comparing the meanresponse times for the always correct and never correct spelling classifications asshown in Table 3 (again due to insufficient observations in the 1-4 spellingclassifications they were not included in the analysis and ‘disruption’ was excludedas no participant scored 0/5 for this word). This analysis confirmed that there was astatistically significant difference between the spelling accuracy classifications:When a word was always spelled correctly it was read significantly faster than whenit was never spelled correctly (F (1, 18) = 28.25, p \ .001), with a very large effectsize (d = 1.307).

Stability of orthographic representations and reading speed

The second research question of interest in this study was: Does reading speed differas a function of the stability of orthographic representations? To further investigatethe nature of the spellings for the words that were always or never spelled correctly,each word was recoded according to how stable the spellings were over the 5 trials.This lexical stability score reflects how consistent the orthographic representationsare in memory. A stability score of 5 indicates that the word was both highlyconsistent and accurate because it was spelled in the same, correct, manner acrossall 5 trials: Words that were always spelled correctly were automatically given aperfect stability score. In contrast, there are numerous potential ways for a word tobe misspelled. A word that was never spelled correctly could have a stable, yet

1494 S. Martin-Chang et al.

123

Page 11

Tab

le3

Mea

nre

adin

gre

spo

nse

tim

es(a

ndn

um

ber

of

par

tici

pan

ts)

per

wo

rd(i

nm

illi

seco

nd

s)as

afu

nct

ion

of

spel

lin

gac

cura

cysc

ore

s

Word

sS

pel

ling

accu

racy

score

sa

01

23

45

NM

(SD

)N

M(S

D)

NM

(SD

)N

M(S

D)

NM

(SD

)N

M(S

D)

Bu

reau

crac

y4

56

54

.38

(12

5.9

8)

26

36

.08

(55

.74

)

0–

25

92

.00

(83

.44)

75

77

.71

(10

0.2

7)

13

56

0.4

0

(40

.22)

Co

mm

iser

ate

19

66

7.7

8

(16

3.0

3)

45

08

.00

(92

.19

)

14

71

.33

(0.0

0)

66

32

.61

(50

.92)

69

41

.67

(21

7.0

2)

31

66

6.8

1

(16

0.2

6)

Co

nv

erg

ent

16

70

.50

(0.0

0)

18

59

.50

(0.0

0)

0–

29

54

.17

(14

0.2

4)

56

24

.13

(11

7.5

6)

64

62

2.9

8

(13

1.8

3)

Co

urt

eous

21

65

3.1

8

(15

2.1

8)

26

86

.67

(3.7

7)

25

76

.50

(98

.76

)

34

55

.00

(19

.09)

10

55

2.5

9

(81

.59)

31

55

4.8

0

(93

.58)

Dia

phra

gm

23

595.0

6

(10

2.2

8)

15

13

.00

(0.0

0)

0–

26

33

.33

(54

.21)

10

55

9.6

0

(80

.24)

36

52

1.0

6

(83

.39)

Dis

rup

tio

nb

0–

0–

0–

0–

55

50

.67

(35

.64)

67

58

0.1

5

(11

0.6

9)

Div

iden

d8

67

4.3

1

(10

8.7

3)

0–

14

99

.33

(0.0

0)

46

80

.33

(20

6.4

2)

12

76

8.5

9

(18

5.3

3)

47

57

1.8

0

(13

0.1

5)

Fil

trat

ion

18

71

.33

(0.0

0)

0–

0–

0–

57

37

.97

(17

5.5

5)

68

61

2.9

9

(13

5.9

0)

Flu

ore

scen

t3

26

30

.46

(15

2.0

3)

35

99

.78

(92

.82

)

36

33

.89

(68

.29

)

55

90

.30

(61

.37)

36

38

.56

(96

.34)

28

58

4.2

4

(84

.73)

Gra

die

nt

27

42

.08

(12

.14

)

19

19

.33

(0.0

0)

26

03

.00

(21

.21

)

0–

65

69

.23

(98

.86)

63

54

8.2

4

(85

.46)

Ho

oli

gan

14

72

9.2

3

(11

1.6

6)

0–

19

55

.67

(0.0

0)

36

42

.00

(68

.64)

46

67

.67

(10

3.3

2)

51

63

5.3

1

(12

6.1

6)

Reading, spelling, and orthographic quality 1495

123

Page 12

Tab

le3

con

tin

ued

Word

sS

pel

ling

accu

racy

score

sa

01

23

45

NM

(SD

)N

M(S

D)

NM

(SD

)N

M(S

D)

NM

(SD

)N

M(S

D)

Lo

llip

op

55

30

.47

(67

.83

)

0–

16

59

.00

(0.0

0)

0–

25

88

.00

(25

.46)

66

53

6.1

6

(88

.13)

Par

affi

n4

76

58

.48

(14

8.0

5)

35

75

.89

(96

.68

)

0–

24

54

.83

(19

.56)

55

66

.47

(11

7.0

7)

16

57

9.7

3

(12

1.6

6)

Pin

nac

le2

06

77

.51

(12

1.3

4)

55

60

.43

(66

.12

)

25

41

.00

(63

.64

)

46

47

.89

(15

4.7

7)

65

24

.11

(55

.98)

36

54

6.9

1

(83

.38)

Pla

gia

rism

47

65

0.1

2

(12

7.5

0)

0–

15

31

.00

(0.0

0)

0–

25

64

.83

(3.5

4)

23

58

1.8

8

(83

.93)

Pro

pel

ler

19

60

5.6

0

(89

.50

)

55

42

.43

(99

.46

)

35

74

.72

(11

9.5

8)

46

19

.13

(17

0.6

9)

97

14

.28

(20

2.7

6)

34

61

3.8

7

(84

.00)

Sal

uta

tion

2633.5

(12

4.6

9)

15

92

.50

(0.0

0)

0–

15

45

.00

(0.0

0)

66

01

.94

(11

6.1

3)

64

59

9.8

6

(14

3.8

1)

Sil

ho

uet

te3

46

64

.52

(17

4.8

4)

35

58

.44

(13

5.8

0)

25

85

.00

(12

7.2

8)

45

91

.54

(67

.88)

35

46

.89

(82

.23)

28

59

4.4

0

(10

0.9

6)

To

bo

gg

an2

56

67

.93

(14

5.6

6)

46

28

.83

(14

1.1

5)

16

09

.33

(0.0

0)

56

43

.17

(84

.32)

75

41

.00

(83

.72)

31

56

2.7

9

(77

.86)

Zu

cch

ini

36

66

3.0

1

(13

2.1

4)

16

33

.33

(0.0

0)

16

44

.00

(0.0

0)

0–

15

75

.50

(0.0

0)

30

61

9.6

2

(15

2.5

0)

aN

um

ber

of

corr

ect

spel

ling

atte

mpts

out

of

5b

Giv

enth

atD

ISR

UP

TIO

Nw

asn

ever

spel

led

inco

rrec

tly

on

all

5tr

ials

,it

was

excl

ud

edfr

omal

lan

alys

es

1496 S. Martin-Chang et al.

123

Page 13

erroneous, representation in memory (i.e., it was misspelled the same way on eachtrial) or a more unstable representation (i.e., it was misspelled differently ondifferent trials). Therefore, a scoring system was devised to capture this variabilityin which one point was deducted for each unique spelling. A score of 4 indicatesthat the word was spelled consistently, but incorrectly on each trial (5 trials-1 uniquespelling = 4). A score of 0 indicates that the representation for that word was highlyvariable because it was misspelled 5 unique ways (i.e., 5 trials-5 uniquespellings = 0). These stability scores are presented in Table 4.

As depicted in Table 4, in the vast majority of cases when a word was misspelledon all 5 trials, it was written the same incorrect way each time (stability score 4), orit was spelled incorrectly in two different ways across the trials (stability score 3).Note again that the stability score of 5 represents instances where the word wasalways spelled correctly. To evaluate whether the stability of the orthographicrepresentation impacted reading speed for each individual word, a within-wordanalysis of variance was run to compare reading speed as a function of stabilityscores, using the data from the stability scores of 3, 4, and 5 (as these were thescores with the vast majority of observations), followed by planned orthogonal(Helmert) comparisons. This analysis confirmed that there was a statisticallysignificant difference in reading speed between the stability classifications (F (2,13) = 5.89, p = .015). Planned orthogonal contrasts indicated that when a wordhad a stable and accurate representation (stability score of 5) it was readsignificantly faster than when it had a fairly stable yet inaccurate representation(stability scores of 3 and 4), F (1, 14) = 12.08, p = .004, d = .899; and that when aword had a highly stable yet inaccurate representation (stability score of 4) it wasread faster than when it had a less stable inaccurate representation (stability score of3), F(1, 14) = 4.14, p = .061, d = .539.

Discussion

The present study evaluated the relation between underlying orthographic represen-tations, as reflected in spelling accuracy and stability, and reading speed. Accordingto the premise of the LQH, inaccurate or unstable lexical representations should leadto slower activation and hence slower reading times. The data presented here largelysupport this hypothesis; words that could not be spelled accurately were read moreslowly than words that could. Further, words with more stable representations werealso read more rapidly than were words with less stable representations.

The participants’ ability to read words that they could not accurately spell, whichwas the case for all of the words in the never category, supports the notion thatreading can still occur with partial or erroneous representations (Burt & Tate, 2002;Conrad, 2008; Holmes & Carruthers, 1998). As Burt and Fury (2000) proposed, it ispossible that reading representations need only be accurate enough to distinguish theitem from other words. For accurate spelling, however, a fully specifiedorthographic representation is needed (Hilte & Reitsma, 2011). The present studyindicates that while inaccurate representations may suffice for reading, they do notpromote reading speed to the same degree as fully stable, accurate representations.

Reading, spelling, and orthographic quality 1497

123

Page 14

Tab

le4

Mea

nre

adin

gre

spo

nse

tim

es(a

nd

nu

mb

ero

fp

arti

cip

ants

)p

erw

ord

(in

mil

lise

con

ds)

asa

fun

ctio

no

fst

abil

ity

sco

res

Wo

rds

Sta

bil

ity

sco

resa

01

23

45

NM

(SD

)N

M(S

D)

NM

(SD

)N

M(S

D)

NM

(SD

)N

M(S

D)

Bu

reau

crac

y2

71

7.7

5

(11

3.4

9)

67

49

(17

6.2

)

12

61

1.3

6

(22

5.8

3)

13

64

0.5

7

(11

4.7

4)

12

59

9.9

5

(96

.09

)

13

56

0.4

0

(40

.22)

Co

mm

iser

ate

17

37

(0.0

0)

0–

0–

10

54

5.2

5

(32

5.5

6)

86

45

.33

(17

0.8

1)

31

66

6.8

1

(16

0.2

6)

Co

nv

erg

ent

0–

0–

0–

0–

16

70

.5

(0.0

0)

64

62

2.9

8

(13

1.8

3)

Co

urt

eous

0–

0–

0–

76

40

.71

(20

9.3

)

14

59

9.0

8

(20

6.2

9)

31

55

4.8

0

(93

.58)

Dia

phr

agm

0–

16

38

(0.0

0)

17

34

.67

(0.0

0)

65

83

.61

(76

.96)

15

58

7.4

8

(11

2.6

1)

36

52

1.0

6

(83

.39)

Dis

rupt

ionb

––

––

––

––

––

––

Div

iden

d0

–0

–1

60

6

(0.0

0)

36

17

.55

(71

.93)

45

65

.37

(38

8.7

9)

47

57

1.8

0

(13

0.1

5)

Fil

trat

ion

0–

0–

0–

0–

18

71

.33

(0.0

0)

68

61

2.9

9

(13

5.9

0)

Flu

ore

scen

t1

89

4.7

(0.0

0)

27

62

.83

(25

4.7

9)

15

61

.33

(0.0

0)

96

75

.05

(16

0.9

1)

19

58

5.1

4

(12

4.2

6)

28

58

4.2

4

(84

.73)

Gra

die

nt

0–

0–

0–

17

33

.5

(0.0

0)

17

50

.67

(0.0

0)

63

54

8.2

4

(85

.46)

Ho

oli

gan

17

92

.5

(0.0

0)

17

14

(0.0

0)

16

87

(0.0

0)

57

75

.53

(14

7.8

7)

65

68

.14

(29

0.3

3)

51

63

5.3

1

(12

6.1

6)

1498 S. Martin-Chang et al.

123

Page 15

Tab

le4

con

tin

ued

Wo

rds

Sta

bil

ity

sco

resa

01

23

45

NM

(SD

)N

M(S

D)

NM

(SD

)N

M(S

D)

NM

(SD

)N

M(S

D)

Lo

llip

op

0–

0–

0–

0–

55

30

.46

(67

.82)

66

53

6.1

6

(88

.13)

Par

affi

n0

–1

55

2

(0.0

0)

0–

12

67

1.3

8

(16

4.2

4)

34

64

4.3

5

(20

0.1

4)

16

57

9.7

3

(12

1.6

6)

Pin

nac

le0

–0

–1

62

3

(0.0

0)

47

46

.08

(17

5.3

6)

15

63

2.5

3

(22

6.4

)

36

54

6.9

1

(83

.38)

Pla

gia

rism

0–

0–

0–

11

71

9.6

3

(19

2.0

2)

36

62

2.8

(17

4.6

)

23

58

1.8

8

(83

.93)

Pro

pell

er0

–1

69

8

(89

.50

)

0–

35

67

.13

(13

.87)

15

60

7.0

9

(96

.67)

34

61

3.8

7

(84

.00)

Sal

uta

tio

n0

–0

–1

72

1.6

6

(0.0

0)

0–

15

45

.33

(0.0

0)

64

59

9.8

6

(14

3.8

1)

Sil

ho

uet

te2

59

1.8

3

(8.7

2)

57

27

.76

(21

1.4

)

36

16

.22

(14

1.2

3)

12

71

3.2

2

(19

3.8

2)

12

61

3.6

3

(15

9.4

4)

28

59

4.4

0

(10

0.9

6)

To

bo

gg

an0

–2

57

4.6

6

(37

.71

)

16

94

(0.0

0)

13

67

6.6

2

(12

9.2

9)

95

98

.98

(28

9.4

5)

31

56

2.7

9

(77

.86)

Zu

cch

ini

0–

0–

17

68

.5

(0.0

0)

11

59

6.8

6

(23

7)

24

66

1.3

(13

3.8

)

30

61

9.6

2

(15

2.5

0)

aO

ne

po

int

was

ded

uct

edfo

rea

chu

niq

ue

spel

lin

go

ver

fiv

etr

ials

,so

that

asc

ore

of

0in

dic

ates

rep

rese

nta

tion

sth

atw

ere

mos

tv

aria

ble

and

asc

ore

of

5in

dic

ates

repr

esen

tati

on

sth

atw

ere

mo

stst

able

;th

ese

sco

res

wer

eco

mp

ute

dw

ith

the

dat

afr

omth

e‘n

ever

’an

d‘a

lway

s’sp

elli

ng

accu

racy

clas

sifi

cati

on

sb

Giv

enth

atD

ISR

UP

TIO

Nw

asn

ever

spel

led

inco

rrec

tly

on

all

5tr

ials

,it

was

excl

ud

edfr

om

all

anal

yses

Reading, spelling, and orthographic quality 1499

123

Page 16

Studies have demonstrated that individuals who are skilled spellers also tend tobe faster readers (Holmes & Castles, 2001) with stronger bottom-up processingabilities (Andrews & Bond, 2009; Hersch & Andrews, 2012). The results from thecurrent investigation also show negative correlations between spelling ability andreading rate suggesting that better spellers have faster word activation in reading.These results support and extend previous work showing moderate to high positivecorrelations between the ability to read and spell (e.g., Burt & Fury, 2000; Mehta,Foorman, Branum-Martin, & Taylor, 2005).

While the studies listed above provide a cohesive fit with the LQH, they do notrule out the possibility that advanced spellers may be also be distinctive from poorspellers in other respects (e.g., faster rapid automatic recall, higher IQ). However, itis generally accepted that lexical quality varies not only ‘‘across individuals for agiven word’’, as suggested above but also ‘‘across words for a given individual’’(Perfetti, 2007, p. 380). This second aspect of lexical quality suggests that evengenerally good spellers can have words with lower quality representations, andgenerally poor spellers may have some words with higher quality representations.The strength of the current design is that the participants were not defined bygeneral ability, but rather within each participant the words were defined by howaccurately they were spelled on an individual basis. Using the same words in boththe spelling and reading tasks provides insight into how the lexical quality of eachword relates to single word reading speed. Indeed, the within-participant analysisrevealed that the association between lexical quality as indexed by spelling accuracyand reading speed holds within the individual: participants read words theypersonally could spell correctly faster than words they could not spell.

The within-word analysis was conducted to evaluate if word characteristicsthemselves were responsible for these results: if some words were more difficult toread—and consequently, to spell—then one would not expect to see a relationbetween spelling accuracy and reading speed within words. Yet this analysis revealedthat within-words, reading speed remained very much a function of spellingaccuracy. Said differently, the same words were read faster by individuals who couldspell them correctly, compared to individuals who could not. Taken together, ouraccuracy data provide convincing support that the difference in reading speedappears to be a function of the orthographic representations stored in memory, ratherthan due solely to the general characteristics of the participants or the words.

With regard to reading performance for misspelled words, longer readingresponse times could reflect either slower activations due to poor qualityrepresentations or a conflict between stable, but incorrect, representations andprinted words. Our stability scores reveal that both of these effects might be exertinginfluence. Words with representations that appeared fairly stable, but inaccurate(i.e., words with a stability score of 3 or 4) were read slower overall relative towords that had representations that were both stable and accurate (i.e., stabilityscore of 5). The fact that the words with fairly stable yet inaccurate representationswere read more slowly than words that were stable and accurate supports the ideathat discrepancies between word-specific representations in memory and visualinput might lead to less efficient word retrieval. Likewise, the LQH highlights theimportance of lexical representations that are ‘fully-specified’ and highly coherent

1500 S. Martin-Chang et al.

123

Page 17

(Perfetti, 2007); therefore, it follows that words with orthographic representationsthat are stable in memory should be advantaged over those that have highly variablerepresentations. A novel finding in the present study was the strong trend suggestingthat this may be the case; words with more stable yet inaccurate representations (i.e.,stability scores of 4) appear to be read faster than words with less stable inaccuraterepresentations (i.e., stability scores less than 4).

Holmes and Castles (2001) predicted that ‘‘compared with better spellers, [poorspellers] might also take an exceptionally long time to read connected text’’ (p. 343).Certainly, causal links have been suggested between spelling and reading, such thatimproving spelling results in parallel improvements in reading accuracy (Conrad,2008; Ehri & Wilce, 1987). Our data suggest that refining orthographic represen-tations, even for words that can be accurately read, might bolster reading speed forthose same words. This possibility remains open to experimental investigation.

Limitations and future research

The LQH posits that three facets of word identity are crucial in the quality of wordrepresentations in memory. Here we focused on the orthography (spelling) andphonology (naming words aloud from print) of a set of target words—however itshould be acknowledged that semantic understanding was not tested. As can be seenin the appendix, efforts were made to select words whose meanings were commonlyknown to undergraduate students. Nevertheless, future studies could include asimple multiple-choice or definition test to ensure that the words are understood byall of the participants.

In addition, due to the limited number of data points that fell in ‘sometimescorrect’ classifications (i.e., words with accuracy scores between 1 and 4), we werenot able to fully evaluate the emergence of orthographic representations withinindividuals and words. Lexical quality has been conceptualized in the literature asbeing established incrementally; therefore, future work should aim at gathering alarge enough data pool to examine lexical quality as it becomes more refined alongthis continuum, on a word-by-word basis.

Here, words were classified as being spelled correctly/incorrectly (accuracy scores)and consistently/inconsistently (stability scores) over the 5 trials, however, spellingerrors can also be classified based on how conspicuous they are (e.g., Holmes & Castles,2001). Scoring the degree of deviation between conventional spellings and errors is aninteresting and worthwhile endeavor; however it was beyond the scope of the currentpaper. Likewise, while we are able to measure response times experimentally at theprecision level of milliseconds, the strong effect sizes speak to the magnitude of thedifferences between spelling classifications. Whether or how these timing differenceswould impact everyday reading remains open for further study.

Implications

Perfetti noted that the prediction that word reading would be hampered by poororthographic quality had ‘‘not been tested. Partly because it implies a difficultseparation of spelling, pronunciation, and meaning knowledge for specific words

Reading, spelling, and orthographic quality 1501

123

Page 18

and readers […]’’ (2007, p. 379). Our results speak to this hypothesis. We report thatwords with stable and accurate spellings, and which were therefore presumed toembody higher quality mental representations, were read faster than words thatcould not be spelled correctly. This pattern held whether the data were organizedand analyzed within participants or within words. This finding should not be takenas self-evident; the reading times were taken from words that were read aloudcorrectly and that were most likely familiar to the participants. Nevertheless, it wasthe ability to spell each individual word that best predicted word reading speed.

The present findings build a strong case for the hypothesis that higher qualityorthographic representations allow for more rapid recognition/retrieval of words(Perfetti, 2007; Share, 2008). They also provide empirical support that clarifies thetheoretical underpinning of accurate spelling and fast word reading. More stable andaccurate orthographic representations mean that the reader does not have to resort tophonemically decoding the entire word (Share, 2008), which would be a slower andless efficient process (Nation, Angells, & Castles, 2007). The data described heresuggest that the integrity of stored orthographic information directly impactsreading speed. Future studies are needed to determine whether learning how to spellwords will offer practitioners an additional method for improving reading speed.

Acknowledgments We gratefully acknowledge financial support from the Social Sciences andHumanities Research Council to the first author. We would also like to thank Christopher Martin-Changfor programming E-Prime, and Allyson Haley for help with data collection.

Appendix

See Table 5.

Table 5 Words used in experimental reading and spelling tasks

Numberof letters

Numberof phonemes

Number ofsyllables

Frequencyper milliona

Easy words

Commiserate 11 8 4 0.18

Dividend 8 8 3 0.37

Filtration 10 9 3 0.45

Gradient 8 8 3 0.18

Hooligan 8 7 3 0.73

Propeller 9 7 3 1.53

Salutation 10 9 3 0.12

Lollipop 8 7 3 1.78

Disruption 10 9 3 1.0

Convergent 10 9 3 0.04

1502 S. Martin-Chang et al.

123

Page 19

References

Andrews, S., & Bond, R. (2009). Lexical expertise and reading skill: Bottom-up and top-down processingof lexical ambiguity. Reading and Writing, 22, 687–711. doi:10.1007/s11145-008-9137-7.

Andrews, S., & Reynolds, G. (2013). Why it is easier to wreak havoc than unleash havoc: The role oflexical co-occurrence, predictability, and reading proficiency in sentence reading. In M. A. Britt, S.Goldman, & J. F. Rouet (Eds.), Reading—From words to multiple texts (pp. 72–91). New York:Routledge.

Brysbaert, M., & New, B. (2009). Moving beyond Kucera and Francis: A critical evaluation of currentword frequency norms and the introduction of a new and improved word frequency measure forAmerican English. Behavior Research Methods, 41, 977–990. doi:10.3758/BRM.41.4.977.

Burki, A., Spinelli, E., & Gaskell, G. (2012). A written word is worth a thousand spoken words: Theinfluence of spelling on spoken-word production. Journal of Memory and Language, 67, 449–467.doi:10.1016/j.jml.2012.08.001.

Burt, J. S., & Fury, M. B. (2000). Spelling in adults: The role of reading skills and experience. Readingand Writing, 13, 1–30. doi:10.1023/A:1008071802996.

Burt, J. S., & Tate, H. (2002). Does a reading lexicon provide orthographic representations for spelling?Journal of Memory and Language, 46, 518–543. doi:10.1006/jmla.2001.2818.

Catts, H. W., Bridges, M., Little, T. D., & Tomblin, J. (2008). Reading achievement growth in childrenwith language impairments. Journal of Speech, Language, and Hearing Research, 51, 1569–1579.doi:10.1044/1092-4388(2008/07-0259.

Catts, H. W., Fey, M. E., Tomblin, J., & Zhang, X. (2002). A longitudinal investigation of readingoutcomes in children with language impairments. Journal of Speech, Language, and HearingResearch, 45, 1142–1157. doi:10.1044/1092-4388(2002/093.

Conrad, N. J. (2008). From reading to spelling and spelling to reading: Transfer goes both ways. Journalof Educational Psychology, 100, 869–878. doi:10.1037/a0012544.

Conrad, N. J., Harris, N., & Williams, J. (2013). Individual differences in children’s literacy development:The contribution of orthographic knowledge. Reading and Writing, 26, 1223–1239. doi:10.1007/s11145-012-9415-2.

Table 5 continued

Numberof letters

Numberof phonemes

Number ofsyllables

Frequencyper milliona

Difficult words

Bureaucracy 11 9 4 0.51

Courteous 9 6 3 1.22

Diaphragm 9 7 3 2.06

Paraffin 8 7 3 0.43

Pinnacle 8 6 3 0.43

Silhouette 10 6 3 0.57

Toboggan 8 7 3 0.08

Zucchini 8 7 3 0.96

Plagiarism 10 8 3 0.35

Fluorescent 10 9 3 0.59

Filler words

Happy Range Small Understand Paper

House Print Shoot Green Break

Book Sight Plant Rainbow Shirt

Walk Food Sister Boat Sound

a Brysbarert and New (2009)

Reading, spelling, and orthographic quality 1503

123

Page 20

Cousineau, D., & Chartier, S. (2010). Outlier detection and treatment: A review. International Journal ofPsychological Research, 3, 58–67.

Ehri, L. C. (1995). Phases of development in learning to read by sight. Journal of Research in Reading,18, 116–125. doi:10.1111/j.1467-9817.1995.tb00077.x.

Ehri, L. C., & Wilce, L. S. (1987). Does learning to spell help beginners learn to read words? ReadingResearch Quarterly, 22, 47–65. doi:10.2307/747720.

Frith, U. (1978). Spelling difficulties. Journal of Child Psychology and Psychiatry, 19, 279–285. doi:10.1111/j.1469-7610.1978.tb00470.x.

Frith, U. (1980). Unexpected spelling problems. In U. Frith (Ed.), Cognitive processes in spelling (pp.495–517). London, UK: Academic Press.

Frith, U. (1985). Beneath the surface of developmental dyslexia. In K. Patterson, J. Marshall, & M.Coltheart (Eds.), Surface dyslexia: Neuropsychological and cognitive studies of phonologicalreading (pp. 83–103). The Hague, Netherlands: Martinus Nijhoff.

Hersch, J., & Andrews, S. (2012). Lexical quality and reading skill: Bottom-up and top-downcontributions to sentence processing. Scientific Studies of Reading, 16, 240–262. doi:10.1080/10888438.2011.564244.

Hilte, M., & Reitsma, P. (2011). Activating the meaning of a word facilitates the integration oforthography: Evidence from spelling exercises in beginning spellers. Journal of Research inReading, 34, 333–345. doi:10.1111/j.1467-9817.2010.01442.x.

Holmes, V. M., & Carruthers, J. J. (1998). The relation between reading and spelling in skilled adultreaders. Journal of Memory and Language, 39, 264–289. doi:10.1006/jmla.1998.2583.

Holmes, V. M., & Castles, A. E. (2001). Unexpectedly poor spelling in university students. ScientificStudies of Reading, 5, 319–350. doi:10.1207/S1532799XSSR0504_02.

Jared, D. (2002). Spelling-sound consistency and regularity effects in word naming. Journal of Memoryand Language, 46, 723–750. doi:10.1006/jmla.2001.2827.

Luce, P. A. (1986). A computational analysis of uniqueness points in auditory word recognition.Perception and Psychophysics, 39, 155–158. doi:10.3758/BF03212485.

Martin-Chang, S., & Levy, B. (2006). Word reading fluency: A transfer appropriate processing account offluency transfer. Reading and Writing, 19, 517–542. doi:10.1007/s11145-006-9007-0.

McKay, A., Davis, C., Savage, G., & Castles, A. (2008). Semantic involvement in reading aloud:Evidence from a nonword training study. Journal of Experimental Psychology. Learning, Memory,and Cognition, 34, 1495–1517. doi:10.1037/a0013357.

Mehta, P. D., Foorman, B. R., Branum-Martin, L., & Taylor, W. (2005). Literacy as a UnidimensionalMultilevel Construct: Validation, sources of influence, and implications in a longitudinal study ingrades 1 to 4. Scientific Studies of Reading, 9, 85–116. doi:10.1207/s1532799xssr0902_1.

Melby-Lervag, M., Lyster, S., & Hulme, C. (2012). Phonological skills and their role in learning to read:A meta-analytic review. Psychological Bulletin, 138, 322–352. doi:10.1037/a0026744.

Nation, K., Angells, P., & Castles, A. (2007). Orthographic learning via self-teaching in children learningto read English: Effects of exposure, durability, and context. Journal of Experimental ChildPsychology, 96, 71–84. doi:10.1016/j.jecp.2006.06.004.

Ouellette, G. (2010). Orthographic learning in learning to spell: The roles of semantics and type ofpractice. Journal of Experimental Child Psychology, 107, 50–58. doi:10.1016/j.jecp.2010.04.009.

Ouellette, G., & Fraser, J. R. (2009). What exactly is a yait anyway: The role of semantics in orthographiclearning. Journal of Experimental Child Psychology, 104, 239–251. doi:10.1016/j.jecp.2009.05.001.

Perfetti, C. A. (1992). The representation problem in reading acquisition. In P. B. Gough, L. C. Ehri, & R.Treiman (Eds.), Reading acquisition (pp. 145–174). Hillsdale, NJ: Lawrence Erlbaum Associates.

Perfetti, C. A. (1997). The psycholinguistics of spelling and reading. In C. A. Perfetti, L. Rieben, & M.Fayol (Eds.), Learning to spell: Research, theory, and practice across languages (pp. 21–38).Mahwah, NJ: Lawrence Erlbaum Associates.

Perfetti, C. A. (2007). Reading ability: Lexical quality to comprehension. Scientific Studies of Reading,11, 357–383. doi:10.1080/10888430701530730.

Perfetti, C. A., & Hart, L. (2001). The lexical bases of comprehension skill. In D. Gorfien (Ed.), On theconsequences of meaning selection (pp. 67–86). Washington, DC: American PsychologicalAssociation.

Plaut, D. C., McClelland, J. L., Seidenberg, M. S., & Patterson, K. (1996). Understanding normal andimpaired word reading: Computational principles in quasi-regular domains. Psychological Review,103, 56–115. doi:10.1037/0033-295X.103.1.56.

1504 S. Martin-Chang et al.

123

Page 21

Ratcliff, R. (1993). Methods for dealing with reaction time outliers. Psychological Bulletin, 114,510–532. doi:10.1037/0033-2909.114.3.510.

Reichle, E., & Perfetti, C.A. (2003). Morphology in word identification: A word-experience model thataccounts for morpheme frequency effects. Scientific Studies of Reading, 7, 219-237.

Rosenthal, J., & Ehri, L. C. (2011). Pronouncing new words aloud during the silent reading of textenhances fifth graders’ memory for vocabulary words and their spellings. Reading and Writing, 24,921–950. doi:10.1007/s11145-010-9239-x.

Schneider, W., Eschmann, A., & Zuccolotto, A. (2002). E-Prime reference guide. Pittsburgh, PA:Psychology Software Tools.

Schrank, F. A., McGrew, K. S., & Woodcock, R. W. (2001). Assessment service bulletin number 2: WJ IIItechnical abstract. Itasca, IL: Riverside Publishing.

Seidenberg, M. S., & McClelland, J. L. (1989). A distributed, developmental model of word recognitionand naming. Psychological Review, 96, 523–568. doi:10.1037/0033-295X.96.4.523.

Share, D. L. (2008). Orthographic learning, phonological recoding, and self-teaching. In R. V. Kail (Ed.),Advances in child development and behavior (pp. 31–82). San Diego, CA: Elsevier Academic Press.

Taraban, R., & McClelland, J. L. (1987). Conspiracy effects in word pronunciation. Journal of Memoryand Language, 26, 608–631. doi:10.1016/0749-596X(87)90105-7.

Taylor, J. H., Plunkett, K., & Nation, K. (2011). The influence of consistency, frequency, and semanticson learning to read: An artificial orthography paradigm. Journal of Experimental Psychology.Learning, Memory, and Cognition, 37, 60–76. doi:10.1037/a0020126.

Wagner, R. K., Torgesen, J. K., & Rashotte, C. A. (1999). Comprehensive test of phonological processing(CTOPP). Austin, TX: PRO-ED.

Whelan, R. (2008). Effective analysis of reaction time data. The Psychological Record, 58, 475–482.Retrieved from http://opensiuc.lib.siu.edu/tpr/vol58/iss3/9.

Woodcock, R. W., McGrew, K. S., & Mather, N. (2001). Woodcock–Johnson" III test. Itasca, IL:Riverside Publishing.

Reading, spelling, and orthographic quality 1505

123