Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Rhyming ability, phoneme identity, letter-sound knowledge, and the use of orthographic analogy by prereaders Walton, Patrick D. 1993

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1993_fall_phd_walton_patrick.pdf [ 7.19MB ]
JSON: 831-1.0098869.json
JSON-LD: 831-1.0098869-ld.json
RDF/XML (Pretty): 831-1.0098869-rdf.xml
RDF/JSON: 831-1.0098869-rdf.json
Turtle: 831-1.0098869-turtle.txt
N-Triples: 831-1.0098869-rdf-ntriples.txt
Original Record: 831-1.0098869-source.json
Full Text

Full Text

RHYMING ABILITY, PHONEME IDENTITY, LETTER-SOUND KNOWLEDGE,AND THE USE OF ORTHOGRAPHIC ANALOGY BY PREREADERSbyPATRICK DAVID WALTONA THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIESDepartment of Educational Psychology and Special EducationWe accept this thesis as conformingto the required standardDr. R. WatsonDr. R. JarmanDr. N. KishorTHE UNIVERSITY OF BRITISH COLUMBIASeptember 10, 1993© Patrick David Walton, 1993In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature) Department of EcLcelf■ovicti Ps &kb/  op zst ( -a,cocht) YkThe University of British ColumbiaVancouver, CanadaDate^ OWDE-6 (2/88)11ABSTRACTRecent research in phonological awareness found a stronglink between rhyming ability in preschool children and laterreading achievement. The use of orthographic analogy, theability to make inferences from similarities in spelling tosimilarities in sound, was proposed as the mechanism to explainthis relationship (Goswami & Bryant, 1990). Literature waspresented that suggested the need for further research.Four research questions were examined. First, canprereaders learn to read unfamiliar words on the basis oforthographic analogy after brief training with rhyming words?The evidence supported the view that they could.Second, will the ability to read words by orthographicanalogy be enhanced by phonological training in onset and rime,and by the use of segmented text? The brief phonologicaltraining did not increase analogy word reading over the sametraining without it. However, using text segmented at the onset-rime boundary for training items did increase analogy wordreading.Third, will reading by orthographic analogy vary accordingto the level of prereading skills (rhyming ability, phonemeidentity, letter-sound knowledge)? The majority of children withhigh prereading skills learned to read analogy test words whereasmost children with low prereading skills found the task tooarduous.Fourth, will rhyming ability make an independentcontribution to reading achievement? The results were equivocal.Rhyming ability did make an independent contribution to thenumber of trials taken to learn the training items. It did notwhen analogy word reading was the dependent variable. Phonemeidentity accounted for most of the variance in analogy wordreading.Further analyses found that the ability to identify thefinal phoneme was the best discriminator between children wholearned to read analogy test words and those who did not. Apossible explanation was that children used the final phoneme todetermine the sound of the rime ending rather than the last twophonemes together.111ivTABLE OF CONTENTSCHAPTER^ PAGEiAbstract ^  ►iTable of Contents ^  ivList of Tables  viiList of Figures ^  ixAcknowledgments  xI. INTRODUCTION  ^1II. REVIEW OF LITERATURE  ^7Background  ^7Forms of Phonological Awareness ^  11Phonological Awareness Before Reading ^  18Phonological Skills and Beginning Reading  23Related Models of Beginning Reading ^  30Rhyme, Orthographic Analogy, and Reading ^ 34Summary ^  41III. PROBLEM  45Statement of the Problem ^  45Rationale ^  46Hypotheses  51IV. METHOD ^  54Pilot Study ^  54Subjects  54Design ^  55Procedure  56Condition I ^  57Condition II  59VCondition III ^  60Tests for Prereading Skills ^  60Rhyme Test ^  61Letter-sound Knowledge Test ^  62Phoneme Identity Test ^  63Tests for Dependent Variables  64Trials to Criterion ^  64Analogy Word Reading  66Letter-sound Correspondence Word Reading ^ 67Results ^  67Orthographic Analogy and Beginning Reading ^ 70Prereading Group and Treatment Effects ^ 72Discussion ^  74Additional Analyses ^  78Recommendations ^  82Main Study ^  86Hypotheses  86Subjects ^  88Design  90Procedure ^  91Spontaneous Analogy (SA) Condition ^ 95Phonological (PH) Condition ^  96Orthographic Segmentation (OS) Condition ^ 97V.^RESULTS ^  99Pretest Measures ^  99Data Analysis  103Data Screening ^  103viHypothesis 1 ^  107Testing Rival Explanations ^  109Planned Orthogonal Contrasts  113Hypothesis 2a ^  115Hypothesis 2b  116Hypothesis 2c ^  117Hypothesis 3  119Analysis of Covariance ^  122Hypothesis 4 ^  123Phoneme Identity and Analogy Word Reading ^ 126Summary ^  131VI. DISCUSSION  134Rhyme-Based Analogies ^  136Treatment Effects  137Prereading Skills ^  138Rhyme ^  139Educational Implications ^  142Limitations and Suggestions for Future Research ^ 143References ^  145Appendix A: Definition of Terms ^  157Appendix B: Training Words and Illustrations for thePilot Study ^  159Appendix C: Illustrations for the Rhyme Test ^ 163Appendix D: Examination of Mathematical Assumptions ^ 167Appendix E: Testing For Homogeneity of Slopes ^ 172viiLIST OF TABLESTABLE^ PAGEPilot Study1. Pretest Scores for Prereading Skills Groups ^ 682. Correlations Among Pretests and Trials toCriterion ^  693. Analogy and Letter-sound Wokd Reading ^ 714. ANOVA Table for Analogy Test Word Reading ^ 725. Analogy Words Read by the High Prereading SkillsGroup ^  736. Number of Subjects Reading Analogy Test Words ^ 747. Subject's Prereading Skills Rankings andReading Scores ^  798. Relation of Rhyme to Trials to Criterion AfterControlling for Related Variables ^  80Main Study9. Children's Background Measured by Father'sOccupation and Mother's Education ^  8910. Pretest Scores for Prereading Skills Groups ^ 9911. Pretest Scores for the Three Treatment Conditions ^ 10012. Pretest Scores for the Six Treatment Groups ^ 10113. Correlations Among Pretests, Trials toCriterion, and Analogy Word Reading ^ 10214. Screening for Extreme Residuals in DependentVariables ^  10415. Effects of Adjusting Outliers on Analogy WordReading ^  10516. Occasions Analogy and Letter-sound Test WordsRead Correctly ^  10817. Planned Orthogonal Contrast Matrix for AnalogyWord Reading ^  11418. Subject's Prereading Skills Rankings and ReadingScores ^  121viiiLIST OF TABLES19. Analogy Word Reading with Prereading Skillsas Covariates ^  12320. Relation of Rhyming Ability to Trials to CriterionAfter Controlling for Related Variables^ 12421. Relation of Rhyming Ability to Analogy WordReading After Controlling for Related Variables.... 12522. Predicting Analogy Word Reading with Three Formsof Phoneme Identity ^  12723. Error Analysis of Letter-sound CorrespondenceWord Reading ^  130LIST OF FIGURESFIGURE1. Three Ways to Divide Words Into Component Sounds ^2. Two Models of the Links Between PhonologicalAwareness, Letter-sound Knowledge, and Reading ^ixPAGE12323. Training and Reading Test Sets Used by Baron (1977)^.. 394. Baron's and the Proposed Training Items For Set 1 ^ 495. Training and Reading Test Sets For the Pilot Study ^ 576. Training and Reading Test Sets For the OS Condition ^ 937. Training and Reading Test Sets For the SA and PHConditions ^ 948. Analogy Word Reading Over Training Sets ^ 119xACKNOWLEDGMENTSI can see the face of every person who helped and I wouldlike to acknowledge their contributions. First, my wife Lonaedited the documents, critiqued the ideas, and raised our sonswhen I was absent.Dr. Rita Watson contributed ideas that helped to guide theconceptual framework of the thesis and kept the research theorydriven. Dr. Ron Jarman provided wisdom and assurance. Dr. NandKishor suggested the use of planned orthogonal contrasts.The teachers and administration at the Child Study Centre onthe University of British Columbia campus opened the doors totheir classrooms and allowed the thesis to be piloted with theirstudents. The teachers were Nancy Duggen, Lorna Kryszkiewicz,and Chava Rubenson, and the senior administrator was Dr. GlenDixon. Kate Andersen helped as a research assistant.A similar contribution was made by the teachers andadministration of School District #24 in Kamloops, BritishColumbia. The teachers were Linda Froese, Darlene Phelps, JuneNeufeld, and Barbie Williams, and the senior administrator wasTarry Grieve.The research was supported by a doctoral fellowship grantedby the Social Sciences and Humanities Research Council of Canada.Finally, I would like to thank the children for teaching meso much.1I. INTRODUCTIONThe functional role of words in language is to conveymeaning and the ultimate goal of instruction in reading is toenable the individual to acquire the ability to develop meaningfrom print. Although both reading and speech require somemastery of language, reading requires, in addition, a moreexplicit awareness of the phonological structure of words asrepresented by the alphabet. Unfortunately for the prereader,such phonological awareness is not an automatic consequence ofspeaking a language.One of the most consistent findings of recent research inreading is the strong link between a child's skill inphonological awareness and their progress in reading.Phonological awareness is conscious access to the constituentsounds of speech and the ability to manipulate these sounds.Phonological awareness is the best single predictor of success inlearning to read (e.g., Wagner & Torgesen, 1987) and this hasbeen demonstrated not only for English, but also for French,(Alegria, Pignot, & Morais, 1982), Norwegian (Skjelfjord, 1987),Russian (Elkonin, 1973), and Swedish (Lundberg, Olofson, & Wall,1980). The relationship holds even when extraneous variablessuch as age, language ability, IQ, social class, and memory arecontrolled (Bradley & Bryant, 1985). In addition, a candidateamong potential causes of reading disabilty is a deficit ofphonological processing ability (Stanovich, 1988; Vellutino &Scanlon, 1987).2Rhyming is a form of phonological awareness that involvesunits of sound called onsets and rimes (in the word "cat", "c" isthe onset and "at" is the rime), which are between syllables andphonemes in size. Words having the same rime will rhyme (e.g.,"cat", "fat") and to have this awareness, one must be able todetect the common two-phoneme segment "at". Four-year-oldchildren (Lenel & Cantor, 1981), and even some 3-year-olds, areable to make competent judgments about rhyme (MacLean, Bradley, &Bryant, 1987). Yet preschoolers are usually unable to isolatesingle phonemes (Liberman, Shankweiler, Fischer, & Carter, 1974),other than when the phoneme overlaps with the onset (e.g., "c" in"cat").There are strong grounds for believing that rhyming skill isan important predictor of later reading success. Rhyming skillemerged as the best predictor of reading in the longitudinalstudies of Bradley and Bryant (1983), of Lundberg et al. (1980),and of Ellis and Large (1987). However, exactly how rhymingskill may be related to the development of reading has receivedlittle research attention.Goswami and Bryant (1990, 1992) argue that a possibleexplanation of the link between rhyming and reading is that theability to recognize rhyming words may form the basis fornoticing that these words often share common spellings. A childwho can hear that "bat" and "mat" rhyme could find it easy torecognize that the spelling pattern at the end of these words isthe same. This ability to make inferences from similarities inspelling to similarities in sound has been referred to as the3ability to make orthographic analogies (Goswami & Bryant, 1990).Their argument is that the strong link found between rhyming andreading arises because children's experiences with rhyme helpthem to make orthographic analogies when they begin to read.However, it is difficult to determine whether children juststarting to read are able to make rhyme-based analogies about newwords. The difficulty is this. Non-readers would not beexpected to be able to make orthographic analogies because theability to read at least some rhyming words would be required inorder to identify the common spelling patterns. Sampling readersincludes the confound that rhyme and orthographic analogy maydevelop as a result of reading experience.Two possible solutions to this dilemma have been attempted.One solution was to show children words they are not unable toread (e.g., 'beak'), tell them what the word says, and then testto see if the child can use this clue word to read analogouswords that contain the same spelling sequence (e.g., 'weak').Studies of this type were conducted by Goswami (1986, 1988,1990a, 1990b, 1991) and Goswami and Mead (1992). A second methodwas to train prereading children to read a limited set of wordsthat can then be used, with the help of analogies, to read wordsthat rhyme with the training set. Studies by Baron (1977) and byPick, Unze, Brownwell, Drozdal and Hopmann (1978) followed thisdesign.Goswami consistently found that children were able to readwords that share rimes with clue words more easily than wordsthat did not. Her research provided evidence that young children4are able to make orthographic analogies to read new words but thelink between rhyme and orthographic analogy is less clearlyestablished. Rhyming ability was not measured in her studies(1986, 1988) on reading and analogy that included non-readers(defined as the inability to score on the Schonell Word ReadingTest). Her analogy research (Goswami, 1990b) that did measurerhyming ability was conducted with a sample of readers and thisprecluded isolating the relationship between rhyming andorthographic analogy from the effects of reading ability.From the second method of studying orthographic analogy andbeginning reading, Baron (1977) and Pick et al. (1978) bothconcluded that with only limited training children are able tomake orthographic analogies when reading unfamiliar words.However, Pick et al. did not take the precaution of using testwords that could be read in one way if the children were makinganalogies and in another if they were recoding individualletters. Without this precaution, Pick et al.'s claim that thechildren made analogies is not convincing.Baron (1977) taught kindergartners to rote memorize wordsand sounds and then tested transfer to reading new words thatcould be read by analogy to the trained words. Baron includedtwo types of reading test words in his experiment, one type thatcould be read by analogy and another type that could not.Children's performance on the analogy words was about 90%correct, compared to 15% for the other words. Baron argued thatkindergarten children are able to make orthographic analogieswhen beginning to read.5However, a caution is in order when considering Baron'sfindings. The source of the caution is that Baron's trainingincluded words presented as segmented text (e.g., 'b', 'at','ed'). According to Goswami and Bryant (1990) "We cannot be surewhether the children did so well [in reading analogy words]because they made a genuine analogy or ... because they applied arule that they had just learned about a spelling sequence whichrepresented a particular rime." (p. 68). In other words, thechildren may have read the analogy test words by assembling thesegmented portions of text presented in the training set ratherthan by making an orthographic analogy.The view that beginning readers can use orthographic analogyis contrary to the steps of reading proposed by Ehri (1991,1992a). Ehri argues that children can make orthographicanalogies only after first learning to read, which involvesconsiderable experience with recoding individual lettersequences.In summary, there is considerable evidence that rhymingability is related to later reading ability and some evidencethat this relationship exists because rhyming facilitates the useof orthographic analogy when reading new words. Existingresearch on orthographic analogy and beginning reading has a)sampled readers and so confounded reading ability with rhyme andthe use of orthographic analogy, or b) sampled prereaders but notmeasured rhyme and the use of orthographic analogy concurrently.As well, prereading skills implicated in the use of orthographic6analogy other than rhyming, notably phoneme identity and letter-sound knowledge, have received little research attention.The proposed study seeks to examine the question of whetherchildren are able to use orthographic analogy when firstbeginning to read. Further, the relationships between rhymingability, phoneme identity, letter-sound knowledge, and the use oforthographic analogy will also be studied. The findings couldlead to a more complete understanding of the skills involved inthe initial steps of reading acquisition.II. REVIEW OF LITERATUREBackgroundAnd so to completely analyze what we do when we read wouldalmost be the acme of a psychologist's achievement for itwould be to describe very many of the most intricateworkings of the human mind, as well as to unravel thetangled story of the most remarkable specific performancethat civilization has learned in all its history. (Huey,1908/1962, p. 6).Despite the enormous amount of research on reading in theeight decades since Huey made this observation, the story ofreading remains tangled.It is largely accepted by the educational community thatpreschool reading experience is beneficial for children, yetsurprisingly little research has been conducted on the subject.This was probably a result of the widely held view that readingshould not be taught until the child reached a mental age of6.5 years, the time at which they were said to bedevelopmentally ready (Morphett & Washburne, 1931). Gesell(1940) made an even stronger argument and stated "The attemptto force reading [before age six] frequently leads to temporaryor permanent maladjustment and more or less serious disturbancein the course of normal school achievement." (p. 208).Challenges to this maturational view were originally basedlargely on studies of children who had learned to read throughinformal parental instruction before schooling. The first7studies reflecting this change began to appear in the 1960's(Clay, 1966; Durkin, 1966; Goodman, 1967; Plessas and Oaks,1964; Reid, 1966) and relied mainly on data from theobservations and interviews of early readers and their parents.In 1966 Durkin wrote:The literature still shows some remnants of thematurational concept of readiness, but, as a whole,articles and books are now dominated by the oppositeconception, highlighting the contribution of environmentalfactors. (p. 48).The research on phonological awareness and reading whichreflected this changing view began with the important findingby two Soviet psychologists (Elkonin, 1963; Zhurova, 1963,cited in Ball & Blachman, 1991) that a relationship existedbetween phoneme segmentation abilities and subsequent successin reading. The first western psychologist to researchphonological awareness and reading was Bruce (1964), who foundthat prereaders were unable to detect sounds within words atthe phoneme level and that the ability to do so distinguishedreaders from non-readers. These findings were largely ignoreduntil the 1970's when a second set of research on phonologicalawareness was undertaken which focused mainly on therelationships between phonological skills and readingacquisition (Baron, 1977, 1979; Calfee, 1977; Fox & Routh,1975; Liberman et al., 1974). Consistent findings of a highlysignificant relationship between phonological skills andreading in turn stimulated an enormous amount of research which8is ongoing (see Goswami & Bryant, 1990, Sawyer & Fox, 1991, andShankweiler & Liberman, 1989, for reviews).Stanovich (1988) described the linking of phonologicalawareness and reading as a success story in cognitivepsychology and asked the following:How often in cognitive developmental psychology haveresearchers been able to discover converging ways ofisolating a theoretically intriguing process, link theprocess to the performance of a real-world task ofcritical importance, and show that the efficiency of theprocess in question can be brought under experimentalcontrol? (p. 7).The literature linking phonological skills to readingachievement is striking. A host of studies have shown that thebest single predictor of reading ability among preschoolers isphonological awareness (Share, Jorm, MacLean, & Mathews, 1984;Tunmer and Nesdale 1985; Wagner & Torgesen, 1987). Share etal. (1984) reported that phoneme segmentation at school entrywas the best of 39 measures in predicting reading success twoyears later. The better children are at rhyming (Bradley andBryant, 1983; Ellis & Large, 1987; Lundberg et al., 1980) or atmanipulating phonemes (Lundberg et al. 1980; Stanovich,Cunningham, & Cramer, 1984; Tunmer & Nesdale, 1985), thequicker and more successful is their progress in reading.Many of these studies have implicitly assumed that theskills prerequisite to reading require or at least benefit from9direct instruction. Lundberg and Hoein (1991) were moreexplicit when they emphasized:... the importance of explicit guidance for helpingchildren access the elusive, implicit segments oflanguage. The mere exposure to print and the developmentof adequate concepts of print functions do not seem to besufficient. (p. 89).An alternate view holds that surrounding the child in aliterate environment is sufficient for learning to read.Learning to read in this way "does not need to be 'patched'with skills instruction" and "one cannot reconcile directinstruction with natural learning." (Goodman, 1989, p. 69).The skills necessary for reading, including phonologicalawareness, are viewed as developing spontaneously andindependently in children in literacy enriched environments(e.g., Smith, 1971). This view will not be examined in theproposed study.The elusive segments of language referred to by Lundbergand Hoein (1991) are the sound units within whole words. Theseunits (syllables, onsets and rimes, phonemes) vary in size andin degree of difficulty to identify and manipulate.A great difficulty in the research on phonologicalawareness has been attempting to sort out the relationshipsbetween the various forms of phonological awareness anddetermining their precise roles in reading acquisition. Theforms of phonological awareness continue to be both a majorsource of interest and of confusion in the literature.10Forms of Phonological Awareness There are several sources of complication in the researchon phonological awareness which make it extremely difficult tocompare studies and thus come to a clear understanding of theliterature. These centre around the different sizes of thesound units under study (i.e., syllables, onsets and rimes,phonemes), the varying difficulties of manipulation of thesound unit (i.e., identification, blending, segmenting,deleting), and the lack of definitional consistency. The useof different tasks across studies that combined a variety ofsound units and manipulations has made the interpretation,consolidation, and comparison of research findings difficult.Most of the early research on phonological awareness wasbased on the implicit assumption that syllables are linearstrings of phonemes. Several studies (e.g., Leong & Haines,1978; Treiman & Baron, 1981) found that children achievedawareness of syllables before awareness of phonemes. The ideathat syllables have an internal structure suggests that theseearly studies may have overlooked an important form ofphonological awareness. Researchers examined children'sawareness of syllables and phonemes and did not consider theintermediate units of onsets and rimes.Current research suggests that there are at least fourunits of sound within words; namely syllables, onsets andrimes, and phonemes (see Figure 1). Note that the stimulusitems in phonological tasks are presented verbally, not aswritten text. Four-year-old children (Lenel & Cantor, 1981),11and even some 3-year-olds (MacLean et al., 1987), are able tomake competent judgments about rhyme (note that words which endin the same rime will rhyme, e.g., "fat", "cat"). Yetpreschoolers are usually unable to isolate single phonemes(Bruce, 1964; Lenel & Cantor, 1981; Liberman et al., 1974).Manipulation of words at the phoneme level involves smallerunits of sound than does the manipulation of onsets and rimes.Older children, or at any rate children who have learned toread an alphabetic script, find tasks that require themanipulation of phonemes to be reasonably easy (Mann, 1986).^WORD^SYLLABLE^ONSET and RIME^PHONEME"cat"^"cat" "c-at"^"c-a-t""string"^"string"^"str-ing"^"s-t-r-i-n-g""wigwam"^"wig-wam"^"w-ig-w-am"^"w-i-g-w-a-m"Figure 1. Three ways to divide words into component sounds.(from Goswami & Bryant, 1990, p. 2)The relationship between the onset-rime sound unit andreading was examined by Treiman (1983), who showed that wordgames that divide spoken syllables at the onset and rimeboundary are easier to learn than games which break these unitsup. These results were later confirmed (Goswami, 1986;Kirtley, Bryant, MacLean, & Bradley, 1989). Treiman (1992)summarized this research when she concluded that "... there maybe a point at which children are fairly good at analyzing12spoken syllables into onsets and rimes but have troubleanalyzing onsets and rimes into their component phonemes." (p.70). The only exception appears to be when the onset isrepresented by a phoneme.Treiman (1985) found that 4-year-olds could isolate thefirst sound in a word more successfully if it began with asingle consonant than if it began with a consonant cluster.She concluded that children could isolate and detect singlephonemes in cases where the phoneme coincides with the onset(e.g., the "b" sound in the word "bat"). This finding wasoffered as additional evidence for the existence of theintrasyllabic structures of onsets and rimes, as theidentification of phonemes should be easier when they overlapwith the onset. Thus the question answered by Treiman's (1985)analysis is not whether young children could detect phonemes,but whether some phonemes are easier to detect than others.Note that onset-rime segmentation would involve phonemeawareness if the onset is a single consonant (e.g., "c-at"),which is often the case, so that it may represent anintermediate step between the awareness of syllables andphonemes (Goswami & Bryant, 1990).Forms of phonological awareness become more complicated asa variety of manipulations with each of the sound units arepossible, and the degree of difficulty depends on both the sizeof the sound unit and the type of manipulation. Taskdifficulty increases as the size of the sound unit decreases.For example, several researchers (Bowey & Francis, 1991;13Goswami, 1986; Kirtley et al., 1989) found that onset-rimesegmentation (e.g., "cat" to "c-at") is much easier forchildren than phoneme segmentation (e.g., "cat" to "c-a-t").Yopp (1988) gave 10 phonological awareness tests tokindergarten children (mean age = 5 years, 10 months) andcomputed the relative difficulty of phonological tasks byaveraging the percentage correct for all subjects per test.The ranking from least to most difficult was rhyming, blending,identification, segmentation, and deletion. It is important,then, to accurately identify the unit of sound being examined(i.e., syllable, onset-rime, phoneme) and the type ofmanipulation involved in order to make comparisons acrossstudies.To further complicate the situation, Byrne and Fielding-Barnsley (1990) and Content (1985, cited in Morais, 1991) foundthat fricative consonants (i.e., "f", "h", "s") are easier forprereaders to identify than stop consonants (i.e., "d", "g","t"). Byrne and Fielding-Barnsley suggested that fricativesoffer an easier point of entry for sound identity training thando stops, and that stops may require special attention.Researchers in the past have produced confusing results bycombining items which use different sound units or requiredifferent manipulations without making these distinctionsclear. For example, Bruce (1964) had 5-year-old preschoolersattempt to delete phonemes and found that none of the childrenproduced correct answers in any of the 30 trials. Calfee14(1977), on the other hand, found that over 90% of the 5-year-old preschoolers he tested could delete phonemes.These apparently contradictory findings can be explainedby the differences in difficulty between the phoneme deletiontasks used in the two studies. Calfee had children deleteonsets from rimes (e.g., delete "m" in "mice" to get "ice")whereas children in the Bruce study attempted to deletephonemes from onsets (e.g., delete " s " in "spin" to get "pin").The Bruce phoneme deletion task was much more difficult andwell beyond the ability of prereaders while the Calfee taskobviously was not.Phonological tasks which appear to be similar can producemarkedly different results. Stanovich et al. (1984) observedpoor performance when kindergarten children attempted to deletea singleton onset (e.g., delete "c" in "cat" to get "at"), butgood performance on an apparently similar task in which theywere required to identify the singleton onset that had beendeleted for them ("at", "cat", what sound was deleted?). Inthis vein Blackman (1983) argued that:Tasks which on the surface appear to be measuring the samephenomenon may in fact require different degrees oflinguistic awareness, or may differ in their cognitiverequirements ... we must not talk about phonemesegmentation per se in relation to reading, butsegmentation within the context of a particular task. (pp.476-477).15A comprehensive analysis that examined the different formsof phonological awareness was completed by Yopp (1988). Yoppadministered 10 tests of phonological awareness to 104kindergarten children and reported high intercorrelations amongthe tasks, as did Stanovich et al. (1984).Further, Yopp used a multiple regression analysis thatused reading simple nonsense words as the dependent variable.Yopp's measure of reading acquisition consisted of a brieftraining session where children were told the sound of eachsegmented letter in a nonsense word like 'HOF', and were thengiven a demonstration on how to blend the letters together toform a word. This involved segmentation and blending at thephoneme level and letter-sound knowledge. The results of thestepwise regression analysis showed that Yopp's sound isolationtask, which measured phoneme identification, combined withphoneme deletion, proved to be the best predictors of readingnonsense words (R 2 = .62).Bryant and Goswami (1990) criticized Yopp's readingmeasure by stressing that children may use skills related torhyming (i.e., onset-rime segmentation) when beginning to read.Learning to read the nonsense words used by Yopp requiredphoneme awareness and letter-sound knowledge and precluded theuse of rhyming related skills. Further, Bryant and Goswamiargued that real words may be read differently from nonsensewords. Still, Yopp's idea of counting the trials taken tolearn to read a list of words is a potentially useful measureof reading acquisition in prereaders. Following the16recommendation of Bryant and Goswami (1990), Yopp's readingtask could be improved by using real words which are unfamiliarto the child and by including words that can be read with theassistance of rhyming skill (i.e., including words that rhyme).Such a task will be used in this study as a measure ofbeginning reading.To summarize what is known about forms of phonologicalawareness, words can be divided into at least three differentsized units of sound; syllables, onsets and rimes, andphonemes. Tasks using these sound units vary in difficultywith manipulations of the smallest units, phonemes, the mostarduous. Most prereaders can manipulate syllables, onsets andrimes, and phonemes that coincide with the onset. Themanipulation of phonemes, other than in the initial position ofthe word, is associated with reading.The ranking of sound unit manipulation from least to mostdifficult is rhyming, blending, identification, segmentation,and deletion. Rhyming and the ability to identify initialphonemes are typically present before reading begins. Manyprereaders are able to identify phonemes in medial and finalpositions of CVC words (Yopp, 1988), while phoneme blending andsegmentation appear to develop with reading. The ability todelete phonemes, other than at the onset-rime division, usuallydevelops well after reading has begun.Related to these findings is the possibility that somephonological skills using phonemes are consequences of readingor, the more common position, that these skills develop in a1718reciprocal relationship with reading (e.g., Goswami & Bryant,1990; Perfetti, Beck, Bell, & Hughes, 1987). This explanationdoes not in itself, however, describe how phonological skills andreading interact. It is apparent though that close attentionmust be paid to the forms of phonological awareness whenexamining the relationships between phonological skills andreading acquisition. The puzzle that has yet to be completed inthe relationship between phonological awareness and reading isdetermining which manipulations of which units of sound, incombination with text, are necessary for reading to begin.Phonological Awareness Before ReadingAn important consideration when studying reading withkindergartners is the necessity to determine their ability inrelevant phonological skills. This is based on the assumptionthat the phonological abilities already possessed by prereadersare likely involved in the beginning reading process. Further,more cognitively demanding phonological skills may be builtupon existing phonological abilities (e.g., Bryant et al.,1990).What phonological skills, then, do kindergartners possess?For most kindergartners, syllable segmentation poses littledifficulty (Liberman et al., 1974) and onset-rime segmentation(e.g., "cat" to "c-at") is only slightly more difficult(Kirtley et al., 1989). Few, however, are able to segmentphonemes (e.g., "cat" to "c-a-t") and those who can are usuallyreaders, whether preschool or adult (Bruce, 1964; Mann, 1986;Morais, 1991; Wagner, Torgesen, Laughon, Simmons, & Rashotte,1993).The few treatment-outcome studies involving phonologicalawareness with kindergartners indicate that they know, or canlearn, how to manipulate sounds at the onset-rime level (i.e.,recognize rhymes) and to form letter-sound relationships inbrief periods of time (Baron, 1977; Bryant, Maclean, Bradley, &Crossland, 1990; Byrne & Fielding-Barnsley, 1989, 1990; Calfee,1977; Content et al., 1982; Fox & Routh, 1976; Lundberg, Frost,& Petersen, 1988; Olofsson & Lundberg, 1983, 1985). Further,the rapidity with which they learn to perform these featssuggests that they already have the prerequisite skills.A related and also important matter to consider is whetherphonological awareness can be developed by children prior toinstruction in reading. Lundberg et al. (1988) and Olofssonand Lundberg (1983) provided evidence that it can. Olofssonand Lundberg (1983) trained 95 preschool children (six and 7-year-olds) in an eight week program designed to stimulatephonological awareness. The program featured daily exercisesand games which involved comparing words, syllablesegmentation, rhyming, and identifying the initial phonemes inwords. Their conclusions were that phonological awarenessamong prereaders can be stimulated by systematic training andthat preoccupation with letters was not of critical importanceto get conscious access to the phonological levels of language.The fact that phonological awareness can be developedwithout using the letters of the alphabet does not necessarily19mean that letters are unimportant. In fact, the preschoolteachers in the Lundberg et al. (1988) study commented on thedifficulty of teaching phonological awareness without usingtext. Hohn and Ehri (1983) found that phoneme segmentation wasbetter learned by using letters rather than blank markers andBradley and Bryant (1983) found that training preschoolers inletters and rhyming was significantly more effective inpromoting later reading achievement than training in rhymingalone.The Lundberg et al. (1988) study involved a larger numberof preschool children than their earlier work (some 400) andthe training period was more extensive as it included eightmonths of daily sessions. The training sequence of the gamesand exercises was easy listening games, rhyming exercises,segmenting sentences into words, segmenting words intosyllables, and segmenting initial phonemes.After a year of formal schooling following the training,the experimental group demonstrated significantly betterreading ability than the control group. In addition, theexperimental group made even greater gains in reading abilityover the control group during the next two years, thusdemonstrating Stanovich's Matthew Effect (Stanovich, 1986). Alarge number of skills were measured but the only significantpreschool factors in predicting word recognition and sentencereading in grade three were rhyming, letter knowledge, andphoneme segmentation. These three variables combined accountedfor 30% of the variance in word recognition and 36% of the20variance in sentence reading. Lundberg et al. (1988) concludedthat "phonological awareness can be developed before readingability and independently of it." (p. 282, italics in theoriginal).Byrne and Fielding-Barnsley (1990) claimed that phonemeidentification can be taught to preschoolers regardless ofwhether the phoneme is in the initial or final position in theword, or in a cluster. This finding contradicts other research(e.g., Bruce, 1964; Kirtley et al., 1989) and could be anartifact of the particular task used to measure phonemeidentity. In the Byrne and Fielding-Barnsley (1990) task,children were presented with a pair of pictures and asked whichone started or ended with the target phoneme. This is quitedifferent from being required to articulate which phoneme is ina specified position in a given word, a more stringentdemonstration of phoneme identity, and the procedure morecommonly used (e.g., Yopp's 1988 sound isolation task).Bradley and Bryant (1983) combined correlational andexperimental designs with a large group of four and 5-year-oldpreschoolers to examine the effects of phonological awarenesstraining on reading acquisition. First, the children weretested on their ability to categorize words according tosimilar rhyme patterns (e.g., "sit" different from "pin" and"win"). The children who had done most poorly on the rhymingtask were then divided into four groups which were trained fora total of 40 sessions over the next two years. The threetreatments were a) rhyming (categorizing words by sound), b)21rhyming and letter-sound knowledge, and c) categorizing wordssemantically (e.g., as farm animals). The fourth group was ano-treatment control.After controlling for age, IQ, and memory, Bradley andBryant found a highly significant relation between thechildren's test scores on rhyming ability after the trainingand reading achievement three years later (r = .52). The grouptrained in both rhyming and letter-sound knowledge scoredsignificantly higher than the other three groups. The groupwhich received the rhyming training alone scored higher thanthe categorizing and control groups on reading achievement butthe combination of rhyming training and letter-sound trainingwas far superior. Bradley and Bryant (1983) concluded that"the experience which a child has with rhyme before he goes toschool might have a considerable effect on his success later onlearning to read and to write." (p. 419). This claim issupported by research which found a link between experiencewith nursery rhymes and later reading achievement (Bryant,Bradley, Maclean, & Crossland, 1989).Two comments are in order when evaluating the Bryant etal. findings. First, the phoneme detection and readingmeasures were taken well after the rhyming tests, one and twoyears later, respectively. The relationships between rhymingand the other variables are based, then, on prediction ratherthan on concurrent development. Second, what Bryant et al.(1990) termed phoneme detection was actually a combined scoreof phoneme tapping (tapping out the number of phonemes in a22presented word as in the Liberman et al. 1974 task) and phonemedeletion. Phoneme deletion is a difficult form of phonemicawareness (Yopp, 1988) and given the age of the subjects, aneasier form of phonemic awareness may have been moreappropriate (e.g., phoneme identity). As well, it isunfortunate that Bryant et al. (1990) did not measure letter-sound knowledge so the that relationships between rhyme,phoneme awareness, and letter-sound knowledge could not beexamined.In summary, the research on phonological awareness beforereading indicates that phonological awareness can be developedin preschoolers without the use of text and that the units ofsound which benefit from training include syllables, onsets andrimes, and phonemes in the initial position of words. There islimited evidence that prereaders benefit from instruction inphoneme awareness other than when the phoneme overlaps with theonset. There is strong evidence that a connection existsbetween children's sensitivity to rhyme before they read andtheir reading ability some time later, although there is littleresearch on the mechanisms which may explain this relationship.Phonological Skills and Beginning ReadingResearchers who have examined the development ofphonological skill and reading with children who are justbeginning to read have produced findings that are unclear. Thecontroversies centre around which phonological skills areprerequisite to reading, which develop reciprocally with23reading, and which are a consequence of reading. As Wagner andTorgesen (1987) state:It is no longer enough to ask whether phonological skillsplay a causal role in reading. The question now is whichaspects of phonological processing are causally related towhich aspects of reading ... at which point in their co-development, and what are the directions of these causalrelations? (p. 192).Unfortunately, there is no clear consensus on either theforms of phonological skill necessary for beginning reading(e.g., rhyme, phoneme blending, phoneme segmentation) or on therelationship between phonological skills, letter-soundknowledge, and reading. The examination of the prereadingskills necessary for reading began with the role of letters.Several studies in the 1970's (e.g., Jenkins, Bausell, &Jenkins, 1972; Samuels, 1972) examined whether instruction inletter-sound knowledge was sufficient to begin reading and theevidence supported the view that it was not (see Ehri, 1983,for a critique). Recent research on this topic has centred onexamining which type of phonological skill in combination withletter-sound knowledge may be sufficient for reading, againwith no consensus.Tunmer and Rohl (1991) argue that segmentation skills arenecessary for learning letter-sound correspondences and formaking blending possible. Morais (1991) suggested the reverseorder of development when he indicated that "... it is bylearning the associations of sounds to letters that the child24usually initiates the acquisition of segmental awareness." (p.51).Perfetti et al. (1987) argued that reading enables thediscovery of parallel phonological principles and thatdifferent components of phonological awareness have differentrelations to reading progress. The conclusions from theirlongitudinal study of first grade readers were that phonemeblending acts as a prerequisite for reading and that phonemesegmentation had a reciprocal relation to reading. Gains inreading enable gains in phoneme segmentation, which enablefurther gains in reading. Others (Wagner & Torgesen, 1987)agree that phoneme blending is a prerequisite to reading butargue that phoneme segmentation skills are a consequence ofreading acquisition.Many of the studies which have examined phonologicalawareness and beginning reading have sampled children afterthey have begun formal schooling and have some reading ability.This research has provided useful information but samplingreaders precludes determining the direction of cause and effectbetween particular forms of phonological awareness andbeginning reading. As well, teachers typically use a varietyof methods to teach reading within the same classroom, makingit difficult to attribute gains in phonological awareness andreading to specific training.Recent studies that sampled prereaders and examined therelationship between phonological awareness and beginningreading was carried out by Byrne and Fielding-Barnsley (1989,251990, 1991, 1993 ). These researchers examined the roles ofphoneme segmentation, phoneme identity, and letter-soundknowledge in the first steps of learning to read.In their first study, preliterate 3 to 5-year-oldpreschoolers were taught to read short words (e.g., 'mat' and'sat') and then asked to choose between "mow" and "sow" aspronunciations for the written word 'mow'. According to Byrneand Fielding-Barnsley (1989), the ability to select "mow" forthe correct pronunciation demonstrated acquisition of thealphabetic principle. The alphabetic principle was defined asusable knowledge that phonemes can be represented by letters,such that whenever a particular phoneme occurs in a word, andin whatever position, it can be represented by the same letter.Their most important finding was that many children who couldidentify phonemes and had letter-sound knowledge could acquirethe alphabetic principle, while those who had only one of theseskills could not.Byrne and Fielding-Barnsley (1990) then used a trainingprocedure similar to their previous study and confirmed theearlier results. In addition, they examined how two forms ofphonemic awareness, phoneme identity and phoneme segmentation,influenced acquisition of the alphabetic principle. Theirevidence indicated that training in phoneme identity was moreeffective than training in phoneme segmentation. Phonemeidentity was successfully taught to preschoolers in their(1991) study and Byrne and Fielding-Barnsley also found thatthe increased levels of phonemic awareness occurred with26untrained as well as trained sounds. Byrne and Fielding-Barnsley used the same forced-choice word recognition test asin their previous research and claimed that most of thechildren who possessed phonemic awareness and who knew relevantletter sounds could use their knowledge to decode unfamiliarwords. The claim about decoding ability must be viewed withcaution. The source of the caution is that in the wordrecognition task, children were asked to select a pronunciationprovided by the researcher for a written word. It can beargued that this is unlike the independent reading ofunfamiliar text.Their most recent study evaluated a program to teachphonemic awareness to preschoolers and then measured readingand spelling at the end of kindergarten. Byrne and Fielding-Barnsley (1993) found that "... the clearest differentiation inall measured aspects of reading and spelling resulted fromdividing the children into those who understood phonemeidentity at the end of preschool and those who did not." (p.109).The Byrne and Fielding-Barnsley research providesconvincing evidence that awareness of phonemes, in combinationwith letter-sound knowledge, is related to beginning reading.Rhyming ability was measured in their 1991 study butinexplicably was not included in any analyses. Unfortunately,as a result the relationships between rhyming skill, phonemeidentity, letter-sound knowledge, and beginning reading werenot examined.272 8The difficulty that prereaders have in isolating singlephonemes has led to the claim that reading leads to phonemeawareness (Morais, 1991). The results of several studies (Bryantet al., 1990; Kirtley et al., 1989; Treiman, 1985) on rhyme havedemonstrated that prereaders do have difficulty detectingphonemes, but also that they are capable of doing so, and thatthe onset-rime distinction provides an explanation for theirsuccesses and failures. They succeed when the phoneme that theyhave to detect represents the onset (e.g., deleting the "c" in"cat") but not when the phoneme is only part of such a unit(e.g., deleting the "t" in "string").The rime unit is also a factor in recognizing sounds.Prereaders recognize that "mat" and "cat" end in the same way,but not that "mat" and "pit" share the same ending (Bradley &Bryant, 1985; Kirtley et al., 1989). Thus, the relative ease ofidentifying phonemes is at least partially determined by therelationship of the target phoneme to onset and rime units theunits implicated in rhyming ability.Another interesting finding of the Kirtley et al. (1989)study was the particularly strong connection between reading andthe ability to classify words by only their final consonant.They concluded that a major step in learning to read may takeplace when the child learns to break the rime into itsconstituent sounds by detaching the preceding vowel from thefinal consonant.29The hypothesized relationships between rhyming skillspossessed by prereaders and learning to read were summarized byGoswami and Bryant (1990) as follows:Children are sensitive to the sounds in words long beforethey learn to read ... but these sounds are not phonemes, orat any rate not always phonemes. The important phonologicalunits for young children are onset and rime. Thephonological skill that they bring to reading and writing isthe ability to divide a word into its onset and its rime,and also to categorize words which have the same onset orthe same rime. (p. 147).The ability to divide a word into onset and rime (e.g.,"bat", "b-at") is based on rhyme and it is this skill thatallows the analogy to be made between common rime endings(Bryant et al., 1990; Goswami, 1988). Goswami and Bryant(1990, 1992) propose that orthographic analogy, the ability tomake inferences from similarities in spelling to similaritiesin sound, is the mechanism that explains the relationshipbetween rhyme and reading. Further, rhyming ability is viewedas making an independent contribution to reading.The literature on the relationship between rhymingability, phoneme awareness, and reading has been summarized asfollows by Bowey and Francis (1991):The conceptualization of onset-rime sensitivity as a naturaldevelopmental phenomenon facilitating comprehension ofreading instruction, which in turn fosters phonemicsensitivity (where phoneme and onset/rime units do not30coincide) appears to provide a more economical synthesis ofthe existing literature. (p. 100, italics are in theoriginal).In summary, despite the proliferation of research onphonological awareness in the past decade, the exact role ofphonological skills in the acquisition of reading remainsunclear. The positions include the view that rhyming skillsbased on the units of onset and rime are the importantphonological skills children bring to reading (Goswami &Bryant, 1992) and the view that phoneme identity and letter-sound knowledge are sufficient for reading to begin (Byrne &Fielding-Barnsley, 1991).Related Models of Beginning ReadingThe robust and consistent findings demonstrating a strongrelationship between phonological skills and reading aresufficiently recent that comprehensive models to incorporatethem are only in their genesis, although they have beenanticipated. For example, Barron (1986) summarized empiricalevidence for rejecting the standard dual-route model ofbeginning reading and its associated hypotheses of direct andindirect access. He concluded that the most promisingalternative model is a single process lexical model in whichacquisition of word recognition would be accounted for byinteractions among orthographic and phonological units ofvarious sizes in the lexicon. The phonological units Barrondescribed included phonemes and rimes.The robust findings on the relationship betweenphonological skills and reading place empirical constraints onfuture models and can be used to evaluate existing ones.Recent models (Just & Carpenter, 1987; Rayner & Pollatsek,1989) which are comprehensive in their attempt to account forreading are notable for incorporating recent research oninformation processing during reading, but both models areintended to describe the cognitive processes involved duringfluent reading by adults, not those of beginning reading byyoung children.Current models of reading which do incorporate theresearch in phonological awareness are few, tend to be datadriven, and centre on the phonological and letter-sound skillsnecessary for reading to begin. Two of the central models thatinclude the prereading skills measured in the proposed study(Byrne & Fielding-Barnsley, 1991; Bryant et al., 1990; Ehri,1991) will be considered here.These models are presented in Figure 2 as Models 1 and 2and show only the proposed relationship between rhyming skill,phoneme identity, letter-sound knowledge, and beginningreading. One model is used to represent those proposed byByrne & Fielding-Barnsley and by Ehri because of theirsimilarity.Model 1, based on Byrne and Fielding-Barnsley (1989, 1990,1991) and Ehri (1991) holds that phoneme identity incombination with letter-sound knowledge are sufficient forreading to begin. According to this position:31Phoneme^WordIdentity ReadingLerter-scur,ctKnowledge32Neither phonemic awareness nor knowledge of thecorrespondences between letters and phonemes is sufficientfor the emergence of initial insights into the alphabeticprinciple. But both in combination seem, on these results,to firmly promote its acquisition in otherwise preliteratechildren. (Byrne and Fielding-Barnsley, 1989, p. 317).Model 1 Byrne and Fielding-Barnsley, 1991; Ehri, 1991Model 2 Bryant et al., 1990Figure 2. Two models of the links between phonologicalawareness, letter-sound knowledge, and reading.Rhyming ability was not included in even the most recentdescriptions of beginning reading provided by these researchers(e.g., Byrne, 1992) and so it is not included in Model 1. Infact, Ehri (1992a) argues that considerable experience withalphabetic recoding at the phoneme level is necessary beforeorthographic analogies involving larger units can be made.Rhyming ability was measured in a recent study by Byrne andFielding-Barnsley (1991) but surprisingly was not included inany analyses.Model 1 predicts that phoneme identity and letter-soundknowledge make independent contributions to reading and thatrhyming ability is of no additional benefit.Model 2 is based on research described in Bryant et al.(1990) who argued that children's sensitivity to rhyme makesindependent contributions to phoneme identity and to reading.It predicts that rhyming ability will make an independentcontribution to reading after controlling for phoneme identityand letter-sound knowledge.Bryant et al. (1990) did not include letter-soundknowledge in the three models they tested and the role for thisvariable is inferred from their other research. Bradley andBryant (1983) found that training in phonological awareness andletters was more successful in producing later reading thantraining in phonological awareness alone. They presented thisfinding as evidence that training in letters and phonologicalawareness is more effective for later reading than training inphonological awareness alone.It is important to note that proposing and even finding arelationship between variables does not provide, in itself, an33explanation of the relationship. This principle isparticularly relevant to the proposed study when examining therelationships presented in Model 2. The mechanism proposed toaccount for the relationship between rhyming and reading is theuse of orthographic analogy (Bryant et al., 1990; Goswami &Bryant, 1990, 1992). This is discussed in detail in thefollowing section.Rhyme, Orthographic Analogy, and ReadingRhyming was found to be the best predictor of reading inseveral longitudinal studies that measured phonological awarenessand later reading achievement (Bradley & Bryant, 1983; Ellis &Large, 1987; Lundberg et al., 1980). However, exactly howrhyming skill is related to the development of reading hasreceived little research attention and has been more difficult toestablish.Goswami and Bryant (1990, 1992) argue that the link foundbetween rhyming and reading arises because children's experienceswith rhyme help them to make orthographic analogies when theybegin to read. Rhyme has a distinctive effect by making childrenaware that words share segments of sounds (e.g., "-at" segment in"bat" and "cat"). This, in turn, prepares them for learning thatrhyming words often have similar spelling sequences, and the useof orthographic analogy is proposed to be based on thesesimilarities.The implication is that the use of orthographic analogy isrelated to existing prereading skills, especially rhyming343 5ability, and potentially phoneme identity and letter-soundknowledge as well. The increased use of analogy by children whoare more familiar with the relevant knowledge is supported byresearch on analogical reasoning with children. Two consistentfindings from that literature are that children are able to solveanalogies when, a) the operations are familiar (i.e., they arestrong in rhyme, phoneme identity, and letter-sound knowledge)and, b) similarities in the analogous tasks are pointed out byproviding hints or instructional examples (Brown, 1989; Brown,Kane, & Long, 1988).Convincing evidence supporting the claim that rhymingability is linked to reading through the use of orthographicanalogy, however, is limited and has proven difficult to obtain.The source of the difficulty is that preliterates would not beexpected to make orthographic analogies because they would nothave the prerequisite experience with text. Sampling readerspotentially confounds rhyming skill and the use of orthographicanalogy with reading ability. It could be argued that rhymingand making orthographic analogies were linked in young readersbecause both are implicated in reading.Research attempting to examine the use of orthographicanalogy by beginning readers has taken two forms. In a series ofstudies by Goswami (1986, 1988, 1990a, 1990b, 1991; Goswami &Mead, 1992), children were shown words which they were not ableto read (e.g., 'beak'), told what the word said, and then testedto see if they could use this clue word to read analogous wordscontaining the same spelling sequence (e.g., 'weak'). Success in36the word reading which resulted was used as evidence by Goswamithat the children were able to use orthographic analogy. Two ofher studies will be considered in more detail.Goswami (1986), in a study of 5 to 7-year-old children,provided a clue word (e.g., 'beak') which remained visiblethroughout the trial, and then asked the children to read threedifferent types of words. The first type of word had a commonrime (e.g., 'weak'), the second type had only part of the rime(e.g., 'bean'), and the third type were control words (e.g.,'bask'). The 5-year-olds did better only on the words which hada common rime (mean = 0.89 out of 6 words read) and did equallypoorly on the words that contained only part of the rime and thecontrol words. The older children, on the other hand, did manageto read some words which contained only part of the rime, butwere able to read far more words which contained the completerime. Goswami concluded that prereaders (defined as children whodid not score on the Schonell Graded Word Reading Test) are onlyable to use spelling sequences which represent complete rimes.Can prereaders make analogies about onsets as well? Toexamine this question, Goswami (1986) gave the 6-year-olds a tasksimilar to the previous experiment except that the clue wordsshared an onset (always a consonant cluster) with the clue word(e.g., 'trim', 'trap') or only part of a rime (e.g., 'wink','tank'). Goswami found that children read many more of the firstkind of analogy words than the second and she concluded thatonsets, as well as rimes, play a significant role in children'sanalogies. However, this second study is inconclusive regarding3 7the use of analogy in the first steps of reading because manychildren in the sample were readers and this precludesdetermining the direction of effect between analogy use andreading.Goswami's research seems to demonstrate that even very youngchildren can successfully use analogy to decode new words, but acaution is in order. The experience of hearing a rhyming wordmight encourage children to think of other rhyming words so thata phonological priming effect may have occurred. The possibilitythat phonological priming explained the results of the 1986 studywas examined in a later study by Goswami (1990a) with 6 to 8-year-olds. Goswami used the same technique as the 1986 study butincluded words that shared orthography (e.g., 'most', 'cost') andwords that shared phonology (e.g., 'most', 'toast'). Goswamifound that phonological priming was insufficient to account forthe analogy reading she observed. Unfortunately, Goswami did notmeasure rhyming ability in her studies that sampled prereaders,which she indicated formed the basis of the orthographicanalogies, or other prereading skills which were implicated suchas phoneme identity and letter-sound knowledge. Measuring theseprereading skills would have allowed for an examination of thebasis for the use of orthographic analogy.Goswami's conclusions concerning rhyming ability and readingwere supported by Wise, Olson, and Treiman (1990), who used twolearning conditions to train children to read words. In onecondition the words were separated at the onset-rime border(e.g., "f-ork") and in the second condition words were separated38within the rime (e.g., "co-rn"). As they predicted, the childrenwere better at reading the words which had been presented at theonset-rime division and they concluded that the onset-rimeboundary is an important one for children learning to read.A second method to examine the relationship betweenorthographic analogy and reading has been to train young childrento read a small set of words and then test them on reading wordsthat rhyme with the training set. Studies by Baron (1977) and byPick et al. (1978) followed this design.Pick et al. (1978) trained 6-year-olds to read 12 simpleconsonant-vowel-consonant (CVC) words and then asked the childrento read CVC nonsense words which shared combinations of the sameletters, including rimes. Thirteen of the 17 children in thestudy read some nonsense words correctly on the first transfertrial. Pick et al. (1978) concluded that children are able tomake orthographic analogies with only limited training. However,Pick et al. did not use test words that could be read by letter-sound correspondence in addition to the analogy test words andwithout this precaution, Pick et al.'s claim that the childrenmade analogies is not convincing.The study by Baron (1977) is important to the proposed studyand will be examined in detail. Baron had kindergartners rotememorize words and sounds (e.g., 'b', 'at', 'bat', 'ed', 'red')and then tested transfer to reading new words such as 'bed' and'rat' (which can be read by analogy to the rimes of the trainedwords), and 'bad' and 'bet' (which do not share rimes, but onlyletters with the trained words, and therefore can only be read,39according to Baron, by combining letter-sound correspondenceslearned in the training words). Figure 3 shows the training andtest items used by Baron (1977).Children's performance on the analogy words was about 90%correct, compared to 15% for the other words, and Baron arguedthat kindergarten children are able to make orthographicanalogies when beginning to read. Goswami and Bryant (1990) citeBaron (1977) as support for their view that beginning readersmake orthographic analogies as they state "By and large, Baron'sstudy does seem to demonstrate that children take to analogiesvery well and very soon" (p. 68). They do, however, highlight anelement of Baron's training which may have allowed children toread the analogy test words by using spelling rules.Training Sets^1^2^3^4b^d rat ug^in^ugbat^bug pin muged am^at^anred^dam sat ranReading Test SetsType 1. Analogy^bed (9)^dug (11) sin (13) rug (13)Analogy rat (11) bam (11) pat (13) man (12)Type 2. Letter-sound^bad (1)^bum (1)^pit (1)^rag (1)Letter-sound bet (1)^bag (0)^sit (2)^run (5)Note. The number of occasions the reading test words were readcorrectly is in parenthesis (max. = 13).Figure 3. Training and test words used by Baron (1977).40The source of the problem is that the letters whichrepresented the rimes were presented as segmented text (e.g.,'at' and 'ed' in set 1, Figure 3), and then children were givenexplicit instruction about the sounds. Baron also includedexplicit training in letter-sound knowledge (e.g., 'b' in set 1).Goswami and Bryant (1990) argue that children may have simplyassembled the segmented portions of text to read the analogy testwords, which would not require the use of orthographic analogy.This critique is decisive in determining the treatment conditionsused in this study. Still, an equally likely explanation ofBaron's results is that instruction in the spelling sequence ofthe rimes helped draw attention to the similarities between therhyming words and thus facilitated the making of orthographicanalogies.It could also be argued that training with the stimulusitems used by Baron (e.g., 'b', 'at', 'bat', 'ed', 'red') alsotrained phonological awareness skills (e.g., onset-rimesegmentation). The single training condition in Baron's studyprecluded being able to examine the effects of phonologicalawareness training from the training in segmented text, as theyare confounded. As well, Baron did not measure rhyming abilityor other prereading skills such as phoneme identity or letter-sound knowledge which may have formed the basis for the use oforthographic analogy. Finally, the sample was not screened forchildren who could read. These considerations are essential tothe pilot and to the main study that was conducted.SummaryResearch into reading with preschool children is recentand was inhibited by the belief that teaching reading skills toyoung children was potentially harmful (Gesell, 1940). As thisview changed, researchers investigating the phonological skillsof preschool children found that prereaders were unable tosegment phonemes, while young readers could (Bruce, 1964). Theimportance of this finding remained unappreciated until thefollowing decade when interest in phonological awareness becamewidespread. The consistent finding was that phonologicalawareness proved to be the best single predictor of success inbeginning reading.A credible hypothesis of the relationship betweenphonological awareness and reading appears to be that childrenare sensitive to the sounds of spoken language before theybegin reading, and that they can use these phonologicalabilities, especially rhyming, when learning to read (e.g.,Bryant et al., 1990). It has also been argued that awarenessof units of sound smaller than those involved in rhyming, inparticular phonemes, are both necessary and sufficient forreading to begin (e.g., Byrne & Fielding-Barnsley, 1991).There are at least two views of the relation betweenphonological awareness and beginning reading. One view is thatchildren begin reading when the ability to analyze words at thephoneme level is accompanied by letter-sound knowledge. Themechanism for reading is the application of letter-soundcorrespondences and the necessary skills are phoneme awareness4142and letter-sound knowledge (Byrne, 1992; Byrne & Fielding-Barnsley, 1989, 1990, 1991, 1993; Ehri, 1991).A second view is that rhyme makes an independentcontribution to reading and that rhyme related skills (i.e.,onset-rime segmentation, initial phoneme identity), incombination with letter-sound knowledge, are related tobeginning reading. Orthographic analogy, the ability to makeinferences from similarities in spelling to similarities insound, is proposed to be the mechanism which explains thisrelationship. Rhyming skill facilitates the use oforthographic analogy as words that rhyme often share similartext, the rime ending.There have been few studies on the use of orthographicanalogy in beginning reading, which is surprising given thepotential theoretical and practical implications. However, thisarea of research is recent and has inherent difficulties relatedto sampling. If readers are sampled, then the use oforthographic analogy is confounded with reading ability and thedirection of the relationship between reading and analogy cannotbe determined. Prereaders, on the other hand, would not beexpected to use orthographic analogy as some experience withreading is necessary before children could benefit from notingsimilarities in text.One attempted solution to this dilemma has been to teachchildren to read a clue word and then test to if this enablesthem to read other words, some of which share a spelling patternwith the clue word (e.g., Goswami, 1986, 1991). These studies43found that children were able to read more words that shared acommon rime with the clue word than test words that did not. Theuse of orthographic analogy was inferred on the basis of thesimilarities in text between the clue word and the successfullyread test word. However, in Goswami's studies which includedprereaders, the prereading skills upon which the use oforthographic analogy were proposed to be based (rhyming skill,initial phoneme identity, letter-sound knowledge) were notmeasured.Another approach to studying orthographic analogy inbeginning readers was demonstrated in a study by Baron (1977).Baron trained kindergarten children to read a set of letters andwords and then immediately asked them to read two types of words.One type of word could be read by analogy to the training setwhereas the other test words required the recoding of individualletters. Baron found that over 90% of the children were able toread the analogy words while only 15% read the words thatrequired letter-sound correspondence.Baron's claim that the children read the words by analogywould be convincing except for two cautions. First, the trainingitems contained segmented text and Goswami and Bryant (1990)argued that the children may have simply assembled the spellingsequences to read rather than have read using orthographicanalogy. Even if Baron's claim that the children read by analogyis accepted, the use of segmented text in training itemsimplicitly gave children phonological training with onsets andrimes. As a result, the effects of segmented text andphonological training are confounded.Secondly, Baron did not screen his sample for readers.Finally, Baron did not measure potentially relevantprereading skills (i.e., rhyme, phoneme identity, letter-soundknowledge) so that the relationship between ability in theseprereading skills and the use of analogy to read was notexamined.44III. PROBLEMStatement of the ProblemThe review of literature illustrated the need for furtherresearch on the use of orthographic analogy and beginningreading. More specifically, the following three questions havereceived little research attention. First, can prereaders useorthographic analogy in the beginning steps of reading? Second,what is the relationship between training with rhyming text andprereaders' ability to use orthographic analogy. Third, how isthe use of orthographic analogy related to the prereading skillsof rhyming, phoneme identity, and letter-sound knowledge?Existing research has provided evidence that young readerscan use orthographic analogy (Ehri, 1992a; Goswami, 1988).However, sampling readers has made it unclear whetherorthographic analogy can be used in the first steps of reading,or is the result of reading experience. Studies samplingprereaders (Goswami, 1986, 1988) have concluded that somechildren can use orthographic analogy when first starting toread. Unfortunately, these studies are few and the potentiallyrelevant prereading skills of rhyming, phoneme identity, andletter-sound knowledge were not measured, so the basis for theuse of orthographic analogy was not examined.Baron (1977) was extremely successful in trainingkindergarten children to read words by orthographic analogy (90%correct) but the segmented text used in his training allowed foran alternative explanation of the results. Goswami and Bryant4546(1990) cite Baron's (1977) study in support of their views butalso suggest that Baron's readers may have read the analogy wordsby assembling the segmented text, rather than by usingorthographic analogy. In addition, Baron did not screen thesample for readers or measure potentially relevant prereadingskills.The question of the use of orthographic analogy in the firststeps of reading has important theoretical and practicalimplications, yet it remains largely unstudied.Rationale There is evidence that kindergarten children are able tomake orthographic analogies when beginning to read (Baron, 1977;Goswami, 1986, 1990a). Also, there appears to be a link betweenrhyming ability and orthographic analogy in young children whoare readers (Goswami, 1990b). These researchers and others(e.g., Bryant et al., 1990) argue that beginning readers abilityto use orthographic analogy is based on rhyming skill.To examine the question of whether children are able to makeorthographic analogies when they first start to read, it is clearthat a study is required that, a) samples prereaders, b) includestest words that could identify if the children were makinganalogies or if they were recoding individual letters, and c)examines conditions that may instill and enhance the use oforthographic analogy. In addition, measuring skills such asrhyming ability, letter-sound knowledge, and phoneme identity47would allow for an examination of the relationship between theseprereading skills and the use of orthographic analogy to read.Sampling prereaders is preferred as it allows therelationships between prereading skill and the use oforthographic analogy to be studied without the confounding effectof reading ability. The proposed study will sample prereaders.Testing for reading with words that can identify if childrenwere making analogies or if they were recoding individual lettersis necessary or else any claim that children used orthographicanalogy to read would not be convincing. Baron (1977)accomplished this by using one set of test words which could beread by analogy (analogy test words rhymed with the trainingwords) and another set which could not (letter-sound test wordsdid not share ending sequences with training words). Theproposed study will use the same two types of reading test wordsas Baron (1977) to provide evidence to identify the strategychildren used to read the test words. This will also allowcomparisons to be made with Baron (1977).Baron's training was extremely effective in promotingbeginning reading regardless of the skills used by the childrento read the analogy words. It is likely that the children madeorthographic analogies and Baron's (1977) training provides auseful model which could be modified to study orthographicanalogy and rhyming in beginning reading. However, the stimulusitems used in Baron's training sets make it impossible todetermine whether the analogy word reading resulted from thebenefits of presenting the words in visual segments, from the48implicit phonological training children would receive in learningthese segmented words, or the combination. As well, the use ofsegmented text allows for the possibility that the analogy testwords were read by piecing together the segmented stimulus items,not by the use of orthographic analogy (Goswami & Bryants'scriticism). The study seeks to examine the use of orthographicanalogy by prereaders without the confound of segmented text.For this reason, Baron's original training sets were modified tocontain only whole word text.Modifying Baron's study to comply with the previously statedcriteria would mean, a) screening the sample for readers, b)providing conditions that can examine the effects of orthographictraining with whole words, and c) measuring rhyming ability,phoneme identity, and letter-sound knowledge. To satisfy b), itis necessary to provide a training condition or conditions whichuse whole words while still following Baron's training as closelyas possible. Any new conditions would require, then, deletingthe segmented text from the training sets used by Baron (e.g.,'b', 'at', 'ed' from set 1, Figure 4).The reduced training sets, however, would not allow fortraining in orthographic analogy as the remaining two words donot share common text (e.g., reduced set 1 = 'bat', 'red'). Thisproblem can be solved by adding words that rhyme with the tworemaining words in the training set. This would allow for thepossibility of experience in orthographic analogy before thechildren are tested on this ability (see Figure 4). Toaccommodate these changes, five new words were added to Baron's49four training sets. The endings upon which the analogies couldbe made were unchanged as were the words in Baron's reading test.Figure 4 shows set 1 of Baron's (1977) four training sets and theproposed set 1 for the proposed training conditions.Set 1 Training Items^ Baron's Proposedb^batat matbat^reded tedredSet 1 Reading Test WordsType 1. Analogy^ bed^bedAnalogy rat ratType 2. Letter-sound bad^badLetter-sound bet betFigure 4. Baron's and the proposed training items for set 1.Successfully training prereaders to read analogy test wordsusing only whole words during training would preclude Goswami andBryant's (1990) caution that subjects may have read the analogywords by assembling segments of training text. Trainingconsisting solely of teaching prereaders to read whole words,some of which rhyme, describes one of the proposed trainingconditions.An inspection of Baron's training reveals that the segmentedtraining items (e.g., 'b', 'at', 'ed') divide the whole trainingwords at the onset-rime boundary (see Figure 4). It is possiblethat learning to read these training items provided phonological50training with onset and rime. If this was true, then thephonological training could have been a key element in theanalogy reading that followed. This analysis is consistent withthe view that orthographic analogy in beginning readers isrelated to their rhyming ability (e.g., Goswami, 1990b). Thepossibility exists that the phonological training with onset andrime was the key feature of Baron's training and accounted forthe subsequent analogy reading.To examine this question, two proposed training conditionswill incorporate phonological training in onset and rime. Again,the Goswami and Bryant (1990) caution of Baron's training isremoved as the training words are not presented as segmentedtext. This suggestion would likely be supported by Treiman(1991) who recommended that "Research is needed to determinewhether phonological awareness training programs that include anonset/rime step are more successful than those that do not." (p.164).The third recommended modification to Baron's study is tomeasure the prereading skills of rhyming, phoneme identity, andletter-sound knowledge. The reasons for this addition are thatGoswami and Bryant (1990, 1992) argue for a specific relationshipbetween rhyming and the ability to use orthographic analogy.Measuring rhyming ability would allow for an examination of thishypothesis. Also, making an orthographic analogy on the basis ofCVC rhyming words (e.g., 'bat', 'rat') would potentially involvephoneme identity and letter-sound knowledge, as the onsets ofthese words are phonemes represented by letters. Measuring51phoneme identity and letter-sound knowledge as well as rhymingskill would permit a study of the relative contribution of theseprereading skills to the use of orthographic analogy.In summary, the proposed study responds to the need forfurther research on beginning reading by training prereadersunder conditions that can examine the use of orthographic analogyin the first steps of reading. In addition, the prereadingskills of rhyme, phoneme identity, and letter-sound knowledgewill be measured to examine their roles. The study will allowfor an examination of whether prereaders are able to makeorthographic analogies, the conditions under which this abilitymay be instilled or enhanced, and the relationships between therelevant prereading skills and the use of orthographic analogy.A pilot study was carried out to provide a preliminary testof the proposed hypotheses and to practice the testing andtraining procedures. The hypotheses tested in the pilot studyare presented in the following section.Hypotheses The following hypotheses were tested in the pilot study.The hypotheses for the main study are very similar but alsoinclude the recommendations of the pilot study.1. Prereaders can learn to read words on the basis oforthographic analogy.Two conditions must prevail before evidence can be providedin favour of this hypothesis. First, prereaders must be able toread significantly more analogy reading test words following52training compared to their pretest scores (subjects wereinitially screened for reading ability). Second, they will learnto read significantly more analogy reading test words thanletter-sound correspondence test words. This second condition isalso necessary because it provides evidence that children wereusing orthographic analogy to read the analogy reading test wordsrather than by recoding individual letters.A positive finding would provide evidence to support theview that children can use orthographic analogy when firstbeginning to read (Goswami & Bryant, 1990, 1992) and would runcounter to the position that children are able to useorthographic analogy only after considerable experience withalphabetic recoding at the phoneme level (Ehri, 1991, 1992a).2. The ability to read words by orthographic analogy will beenhanced by training in orthographic analogy, and by phonologicaltraining in onset and rime.a) Whole word training (Condition I) will instill theability to read words by orthographic analogy.b) Phonological training in onset and rime (Condition II)beyond whole word training will enhance the ability to read wordsby orthographic analogy more than whole word training alone.c) Phonological training in onset and rime with referenceto text (Condition III) will enhance the ability to read words byorthographic analogy more than Condition II training alone.Findings in support of these hypotheses would provideevidence that limited training with whole words is sufficient forprereaders to learn to read words using orthographic analogy, and53that this ability is further enhanced by phonological trainingwith onset and rime.3. The ability to read words by orthographic analogy will varyaccording to the level of prereading skills (Low or High).A finding that children in the High prereading group areable to read more analogy test words than the Low group would beevidence that one or more of the measured prereading skills(rhyming ability, phoneme identity, letter-sound knowledge) areimportant to the use of orthographic analogy. It would alsosupport the view that children perform better on analogical taskswhen they are familiar with the operations and objects introduced(Brown, 1989; Brown, Kane, & Long, 1988).54IV. METHODA pilot study was conducted at the Child Study Centre on theUniversity of British Columbia campus. The method and results ofthe pilot study will be presented before describing the mainstudy.Pilot StudySubjects The initial sample included 21 preschool children. Thechildren were pre-tested for reading ability, and any able toread more than one of eight words selected from Baron's (1977)word reading test (bed, man, bet, dug, sin, pit, bag, rag) werescreened from the study. The eight words were printed in 1"lower case letters on an 8" by 11" card. The card was presentedto the children individually and they were asked to look at thewords and were encouraged to try to read them. Eighteen childrenwere unable to read any words, one child read one word, and twowere excluded from the study because they read two or more words.There was insufficient time to complete the training withtwo children. Two others declined to leave their classroom ontwo consecutive occasions and were dropped from the experiment atthat point. These declines appeared to reflect the child'sreluctance to leave an interesting classroom activity rather thana negative response toward the training, which they seemed toenjoy. Results are reported for the remaining 15 children.The average age of the 15 children (9 girls and 6 boys) was4 years 11 months and the age range was 4 years 3 months to 555years 4 months. Based on preschool teacher information, allchildren had English as a first language and none had any knownlanguage impairments.Expressive language and vocabulary skills were assessed withthe Peabody Picture Vocabulary Test-Revised (PPVT-R) (Dunn &Dunn, 1981). The mean standard score (average for the populationis 100) on the PPVT-R was 114.67 (SD = 13.42). The high PPVT-Rscores of the children likely reflects the high socioeconomicstatus of the population from which the children were drawn.Stratified random assignment, based on the pretest scores ofrhyming, phoneme identity, and letter-sound knowledge was usedwith the preschool sample. The pretest scores were converted toz scores and then added for each subject. Subjects were thenassigned to High or Low prereading skill groups based on the signof the added z scores with positive z scorers going to the Highgroup and negative scorers to the Low group. Then, children wererandomly assigned to one of the three treatment conditions fromwithin each stratified prereading skill group.DesignThe study used an experimental design with three treatmentconditions. The treatment conditions were crossed withprereading ability levels (Low and High) based on the pretestscores of rhyming, phoneme identity, and letter-sound knowledge.This produced a 2 by 3 (2 prereading skill groups by 3 treatmentconditions) factorial design.56The dependent variable for the ANOVA was the ability to readthe analogy reading test words. Reading the letter-soundcorrespondence test words and the number of trials taken to learnto read the training words to criterion were used as dependentvariables in related analyses.ProcedureThe researcher and an assistant visited the children at theparticipating preschool for three days before beginning anytesting or training.The children were initially screened with words selectedfrom Baron's (1977) reading test and were then given the PPVT-R.In the second session, children were pretested for the threeprereading skills of rhyming, phoneme identity, and letter-soundknowledge. The children were grouped (Low or High) according totheir prereading skills and then assigned to one of the threetreatment conditions.The treatments consisted of teaching the children four setsof words and then, following each training set, testing thechildren for word reading. All training was provided by theresearcher or an assistant. There were four sets of words ineach training condition (see Figure 5) and the order of trainingset presentation was random within each treatment condition.The training sessions lasted less than 20 minutes each andone or two sets of training words were taught per session.Testing for reading immediately followed the training. Theprimary emphasis of the training was that the sessions with thechildren were enjoyable and only took place when the childrenwere amenable.Training Sets 1 2 3 4bat bug pin mugmat mug tin bugred dam sat ranted ram mat panReading Test SetsType 1^Analogy bed dug sin rugAnalogy rat bam pat manType 2. Letter-sound bad bum pit ragLetter-sound bet bag sit runFigure 5. Training and reading test sets for the pilot study.Condition I In the first training condition, children were presentedwith a set of training words and accompanying drawings (seeAppendix B) which illustrated the referent of each printed word.The training words were printed in lower case on 2" by 4" cardsin a 1" san serif font, one word per card. The illustrations,one for the referent of each printed word, were on separate 2" by4" cards. The stimulus items were identical for all trainingconditions.The four words in each training set were presented in acolumn with the rhyming words adjacent to each other and theillustrations directly to the left of the printed words. Theresearcher said "We are going to play a game with pictures and5758words. The words will help you to read other words." Theresearcher then introduced two puppets and offered one to thechild, which was invariably accepted. The researcher, throughthe puppet, then named each object pictured and drew attention tothe accompanying text. Children were asked to try and see howthe words were the same.After two namings of the illustrations and paired text, theillustrations were removed and the children were invited to readthe text, with corrective feedback, to a criterion of twosuccessive readings of the four training words without error.The number of trials taken to reach the criterion was recordedfor each training set to a maximum of 12 trials.Immediately after reaching criterion with the four trainingwords, two training words were removed (e.g., 'mat' and 'ted' inset 1) to match the whole words use by Baron (1977). Theremaining two training words were named by the researcher as areminder. Then the children were asked to read the test words(i.e., 'bed', 'rat', 'bad', and 'bet' in set 1), which werepresented individually and in the sequence shown in Figure 5.Following Baron, the training words (e.g., 'bat' and 'red' in set1) remained visible as a reference for the children, who weretold that the training words could help to read the test word.If children said they did not know the test word, they wereencouraged to compare the test card to the training cards.Incorrect and correct responses were recorded for each test wordand, following Baron, corrective feedback was given followingeach incorrect response. In Baron's study the correctivefeedback consisted of being given the correct answer with anexplanation, while in the pilot study, the correct answer wasgiven without an explanation.Compared to Baron's (1977) training, Condition I did notpresent the training words as segmented text and providedconsiderably less information regarding the similarity amongtraining and test words. These changes preclude using thestrategy of reading the test words by assembling segmentedspelling sequences (Goswami and Bryant's criticism of Baron'straining). As such, Condition I would be a stringent test ofbeginning reader's ability to use orthographic analogy.Condition II The procedures for Condition I were followed in Condition IIexcept that the trials to criterion was followed by phonologicaltraining in onset and rime. The children were invited to play agame with sounds and were told that the puppet could say thetraining words so that they have two sounds. One puppet(operated by the researcher) pointed to the whole word and saidthe onset of the word (e.g., "b"). The other puppet (operated bythe child) was asked to imitate the researcher. This process wasfollowed by training in segmenting the rime (e.g., "-at") and inonset-rime blending (e.g., "b-at", "bat"). This training wasrepeated twice for each word in sequence.Immediately following the onset-rime training, two trainingwords were removed (as in Condition I) and the children wereasked to read the test words.5960Condition III The procedures for Condition II were followed in ConditionIII except that during the phonological training in onset andrime, the researcher pointed to the segment of the word thatrepresented the onset and rime as they were being segmented andblended. This allowed for a more explicit connection to be madebetween the representation of the spellings of onsets and rimesand their corresponding sounds. Note that unlike Baron (1977),whole words were kept intact and were not presented as segmentedtext.In summary, the three training conditions differed in thefollowing ways:1. In Condition I, children heard only the entire words andreceived no phonological training.2. In Condition II, training in whole words was followed byphonological training in onset and rime.3. In Condition III, training in whole words was followed byonset-rime training with attention drawn to the relevant text.Tests For Prereading Skills Bowey and Francis (1991) found that task order was a factorwhen both rhyming and phoneme awareness tasks were used. Theyfound that performance on the rhyme task was lower when itfollowed the phoneme task than when it preceded it. Theirexplanation was that when the phoneme task was presented first,prereaders attempted to solve the rhyme task by focusing onphonemic units. To preclude this interference, the testing order61for the pretests and was rhyming ability, letter-sound knowledge,and then phoneme identity.Rhyme Test A version of the rhyme-oddity task used by Bradley andBryant (1983) was used in the pilot study to measure rhymingability. Following Bryant et al. (1990), the added feature wasthe use of pictures to remove the memory load. The initialletter was changed in seven of the 48 stimulus items to allow forillustrations. None of the non-rhyming words were altered.The test consists of 2 practice trials with correctivefeedback and then 10 experimental trials without correctivefeedback. In each trial the child was given four words withpictures, where three rhyme and the fourth does not (see AppendixC). The child's task is to detect the word that does not rhymeand say it back to the researcher.The directions followed Bradley and Bryant (1983). First,the experimenter asked the child if they knew any nursery rhymes.Then the experimenter suggested a rhyme and encouraged the childto produce rhyming words with the following conversation:Do you know Hickory dickory dock?Hickory dickory dock, The mouse ran up the ... ?Do you know Jack and Jill?Jack and Jill, Went up the ... ?Then, the experimenter and the child alternately producedrhyming words, until the experimenter introduced a word that wasblatantly incorrect (e.g., "hat", "rat", ... "table"). If there62was no quick negative response from the child, the error waspointed out. Then the experimenter said:Now I am going to show you four words with pictures, and Iwant you to tell me which word does not sound like theothers. Wait until I have said all the words before youtell me which one it is. Fan, cat, hat, mat.The practice items were:fan cat hat mat^leg peg hen kegThe test items were:pin tin sit fin^doll hop top pop^bun but gun sunmap cap tap pal^pack tack sad back wig pig pin digweed peel seed lead men red bed fed^sand hand land banksink mint pink winkThe mean for rhyming in the pilot study was 5.73 (maximumscore = 10) and the standard deviation was 2.52. In Bradley andBryant's (1983) sample, which included only 5-year-olds, rhymingwas normally distributed with a mean of 6.67 and a standarddeviation of 2.33. Note that their test for 5-year-olds did notuse illustrations although they recently revised their test for4-year-olds to include illustrations in order to remove thememory load (Bryant et al., 1990).Letter-Sound Knowledge Test Children were presented with two 8" by 11" cards whichlisted the letters of the alphabet in alphabetical order in a 1"san serif font. One card listed the letters from 'a' to 'o' andthe second card listed the remaining letters. The letters were63presented in lower case to match the stimulus items used in thetraining and reading testing sets. The children were asked toprovide the sound of individual letters. Children who respondedwith the name of the letter were asked for the sound of theletter. Vowels were scored correctly if they were sounded longor short and the letter 'c' was scored correctly if sounded as"k" or "s".Children in the pilot study either had little or no letter-sound knowledge (9 children scored 0 or 1), or knew many letter-sounds (the remaining 6 children knew 10 to 21 letter-sounds).The mean for letter-sound knowledge was 6.93 and the standarddeviation was 9.06.Phoneme Identity Test The Yopp (1988) sound isolation test was used to measurephoneme identity. In Yopp's study it had the highest predictivecorrelation with a subsequent test of learning to read novelwords (r = .72) and the reliability with 5-year-olds was .84(Cronbach's alpha).The test measures the ability to identify phonemes in theinitial, final, and then medial positions. The scores are summedto produce a total score for phoneme identity (maximum score =15).The test consisted of one practice trial with correctivefeedback and then 15 experimental trials, also with correctivefeedback. In each trial the child was given a word and then64asked to identify the initial, final, or medial sound, with fivewords given for each condition.Directions for the practice item following Yopp (1988) were:I am going to say a word, and you tell me what sound theword starts with. Let's try one for practice: Jack. Whatsound does Jack start with?The directions for the identifying the final and medialphoneme paralleled the directions for identifying the initialphoneme, with the words "food" and "sat" as practice items,respectively.The test items were:Initial sound - car rose name you sleepFinal sound - dog pencil late bean goMedial sound - hot than keep cup pigIdentifying phonemes was difficult for many of the childrenin the pilot study and three were unable to identify anyphonemes. The mean for phoneme identity was 3.87 and thestandard deviation was 3.91. More phonemes were identified inthe initial position (36) than in the final (13) or medial (9)positions. In Yopp's study, which included only 5-year-olds, themean for phoneme identity was 8.77 and the standard deviation was3.74.Tests For Dependent Variables Trials to CriterionThe researchers noticed during the first day of trainingthat the number of trials taken to learn to read the four65training words appeared to vary with the level of prereadingskills. Some children in the High prereading skills group wereable to read the training words immediately after the twopractice trials whereas others in the Low group were unable toread the training words even after 12 practice trials. Allchildren received corrective feedback during this training.The decision was made to record the number of trials thechildren took to reach the criterion of reading the trainingwords (data was not collected on the first training set for the 8children trained on the first day). This measure of readingacquisition provided an additional dependent variable to studythe relationships between the prereading skills and the abilityto learn to read words. A similar measure was used by Yopp(1988) as a test of initial reading acquisition.The words in the training sets were the same for alltreatment conditions. After two namings of the illustrations andpaired text, the illustrations were removed and children wereinvited to read the text with corrective feedback to a criterionof two successive readings of the four training words withouterror. The number of trials taken to reach the criterion wasrecorded for each training set to a maximum of 12 trials.Scores on the test could range from 0 to 10. Children whosuccessfully read each word on the first two trials obtained ascore of 10. Children who successfully read each word on thethird trial received a score of 9, and so on. A score of 0signified that the child was unable to read the words correctlyeven after 12 trials. The number of trials taken to learn the66training words for each set were averaged to obtain an overalltrials to criterion score.Four of the children in the Low prereading skill groupoccasionally adopted a strategy of memorizing the words withoutlooking at the text during one or two of the training sets. Thisstrategy was more effective for these children than attempting toread the text, as it reduced their trials to criterion on theoccasions it was used. This was problematic because it reducedthe amount of experience with the text which would later benecessary to read words in the reading test.The mean score for the trials to criterion for the pilotstudy was 7.34 (SD = 2.72) out of 10, with higher scores denotingfewer trials.Analogy Word ReadingThe reading test words were presented to the childrenimmediately after they had reached the trials to criterion on thetraining words, or were unable to learn the training words after12 trials with corrective feedback. The test words werepresented one at a time on individual cards in 1" san serif font.If children said they did not know the test word, they wereencouraged to compare the test card to the training cards.Corrective feedback was given following each incorrect response.Two training words were visible to the child (e.g., 'bat'and 'red' in set 1) during the attempt to read the analogy testwords. The analogy test words ('bed' and 'rat' in set 1) rhymedwith one of the training words and so could be read by analogy.67If the child could make the inference from similarities inspelling to similarities in sound (i.e., the common '-ed' and '-at' text represents common sounds), this would help to read theanalogy test words. Note that the analogy is based on rhyme andtherefore on the spelling sequence that represents the words'rime.Letter-sound Correspondence Word ReadingThe letter-sound reading test words were presented to thechildren immediately after the analogy reading test words. Theletter-sound reading test words (e.g., 'bad' and 'bet' for set 1)do not share any ending spelling sequences with the trainingwords, and therefore cannot be deciphered by analogy based onrhyme. There are two possible explanations that could accountfor the reading of the letter-sound test words. First, thechildren can read these words by recoding individual letters.Second, they may make analogies about segments of speech andspelling patterns that cut across the onset-rime division (e.g.,the training word 'bat' and the test word 'bad' both begin with'ba-'). Segmentation which cuts across onsets and rimes is muchmore difficult for children than segmentation at the onset-rimeboundary (Kirtley et al., 1989). If a significant number ofletter-sound words are read, both of these explanations will beconsidered.Results Table 1 presents the pretest data for the 15 children in thepilot study. The means and standard deviations are shown68separately by prereading skill groups. The prereading groupswere determined by the scores on rhyme, phoneme identity, andletter-sound knowledge (z scores for each variable were summed).Table 1. Pretest Scores for Prereading Skills GroupsPrereading Skills GroupLow High TotalMeasure Mean(SD) Mean(SD) Mean(SD)Agea 57.12(4.67) 61.57(2.94) 59.20(4.46)PPVT-Rb 108.63(10.21) 121.57(13.94) 114.67(13.42)Rhyme 4.75(2.12) 6.86(2.61) 5.73(2.52)Phonemeidentity 1.25(1.17) 6.86(3.81) 3.87(3.91)Letter-sound .25(.46) 14.57(7.98) 6.93(9.06)Note. n = 15.aAge is in months.bPPVT-R scores are standardized.Two-tailed independent t-tests found that the groupsdiffered significantly on age, t(13) = 2.17, p < .05, phonemeidentity, t(13) = 3.98, p < .01, and letter-sound knowledge,t(13) = 5.10, p < .001 but not on the PPVT-R, t(13) = 2.02, p =.06, or on rhyme, t(13) = 1.76, p = .11.69Rhyme appeared to be normally distributed although cautionis in order due to the small sample size. Phoneme identity had aslight floor effect and letter-sound knowledge appeared to bedistributed bimodaly rather than normally. Five children wereable to identify phonemes but had no letter-sound knowledge,while the reverse was true for only one child.Table 2 shows the correlations among the pretest scores andthe trials to criterion reading measure. These correlations mustbe viewed with caution due to the small sample size (n = 15).Table 2. Correlations Among Pretests and Trials to CriterionMeasure^1^2^3^4^5^61. Rhyme -2. Letter-sound^.39^-3. Phoneme id.^.51^.89***^-4. PPVT-Ra^.66**^.54*^.68**^-5. Age .72**^.40^.54*^.81***6. Trials to^.65**^.49^.53*^.49^.47CriterionaPPVT-R are raw scores.*p < .0 5 .^ **p < .01. ***p < .001.The analogy and letter-sound measures of reading were notincluded in the correlation analysis because they seriouslyviolated the normality assumption (10 children did not score onanalogy or letter-sound reading).70Predictably, age was highly correlated with the PPVT-Rscores and to rhyming ability as well. Rhyming also produced thehighest correlation with the trials to criterion measure. Notealso that phoneme identity was highly correlated with letter-sound knowledge.Orthographic Analogy and Beginning ReadingA primary question examined in the pilot study was whetherbeginning readers are able to use orthographic analogy in theinitial steps of reading. Two conditions would have to presentbefore evidence for using orthographic analogy would beconvincing. First, children would have to benefit from thetraining and be able to read a significant number of the analogyreading test words compared to their pretest scores (recallsubjects were initially screened for reading). Second, theywould also have to be less successful at reading the letter-soundcorrespondence test words. The second condition is alsonecessary because it is evidence that the children usedorthographic analogy to read the analogy test words rather thanby recoding individual letters.The first question of whether the prereaders were able tolearn to read the analogy test words was examined by seeing ifthe scores for analogy word reading differed significantly from 0(recall children were initially screened for reading ability).The results of the dependent t-test analysis (2-tailed) indicatedthe analogy reading scores did differ significantly from 0, t(14)= 2.13, p < .05.71The second question of whether children had more successreading the analogy words than the letter-sound correspondencewords was examined by subjecting the analogy and letter-soundreading scores to a dependent t-test analysis (2-tailed). Thedifferences between the two types of reading failed to reachsignificance with the total sample, t(14) = 1.73, p = .11 or withthe 5-year-olds analyzed separately, t(7) = 2.20, p = .06.However, the observed differences were in the hypothesizeddirection and would have been significant with a slightly largersample (i.e., one more 5-year-old).Table 3 shows the analogy and letter-sound word readingresults. In all, children were able to read the analogy testwords on 17 occasions compared to seven for the letter-sound testwords. Children were clearly less successful at reading theletter-sound test words.Table 3. Analogy and Letter-sound Word ReadingTest Word Type Test Words TotalAnalogy bed (4) dug (0) sin (0) rug (3) 7Analogy rat (2) bam (2) pat (3) man (3) 10Letter-sound bad (2) bum (1) pit (0) rag (0) 3Letter-sound bet (1) bag (1) sit (1) run (1) 472All four children who read the analogy test words were 5-year-olds. Half of the 5-year-olds (n = 8) then, learned to readthe analogy test words (mean = 4.25 words, maximum = 8 words)whereas none of the 4-year-olds read any analogy or letter-soundtest words.Prereading Group and Treatment Effects To examine the effects of prereading skills group membershipand treatment conditions on analogy reading, the analogy readingtest scores were analyzed by crossing the two prereading skillgroups (Low and High) with the three treatment conditions. Theresults of this 2 by 3 factorial design are presented in Table 4.This analysis should be viewed with caution due to the smallsample size (n = 15).Table 4. ANOVA Table for Analogy Test Word ReadingSource SS DF MS F omega2 pPrereadingskills group 20.82 1 20.82 4.97 .36 .05Treatmentcondition.02 2 .01 .003 .00 .99Prereading*treatment.02 2 .01 .003 .00 .99Error 37.67 9 4.19As shown in Table 4, there was a statistically significantdifference on analogy word reading between the Low and High73prereading skill groups (none of the children in the low groupread any analogy words), but not between the treatmentconditions. The means and standard deviations for the ConditionI, II, and III respectively were 1.4 (2.19), 1.25 (2.50), and .83(2.04). Nor was the interaction between prereading skill groupand treatment condition significant.To examine the lack of a main effect due to treatmentcondition, two contingency tables are presented. In Table 5, theanalogy word reading results are shown by treatment condition andby trials. As Table 5 shows, there were no clear differences inthe number of analogy words read between treatment conditions.Table 5. Analogy Words Read by the High Prereading Skills GroupCondition^Set 1 Set 2^Set 3^Set 4^Totals I^2^1^2^2^7II 2^1^1^1^5III^0^1^2^2^5Totalsa^4^3^5^5^17Note. n = 3 in Condition I, n = 2 in Conditions II and III.aMaximum word reading per trial = 14.However, analogy word reading increased over trials with onechild in Condition III. It is impossible to make a strong caseon the basis of one child but this data should not be discountedas it does provide evidence, however limited, that analogy wordreading increased over trials in Condition III.Another perspective on analogy word reading is provided byTable 6, which presents the number of subjects who read analogytest words. Note that all successful analogy word readers were5-year-olds.Table 6. Number of Subjects Reading Analogy Test WordsCondition^Set 1^Set 2^Set 3^Set 4I^ 2 1 2 1II 1^1^1^1III^0^1^1^1Totals 3^3^4^3As with Table 5, there were no apparent differences inanalogy word reading among treatment conditions. As well, thenumbers of children reading analogy words did not appear to varysignificantly over trials.DiscussionThe finding that prereaders were able to read the analogytest words after a few brief training sessions with rhyming wordssuggests that beginning readers are able to use orthographicanalogy. For this argument to be convincing however, anothercondition had to be met; significantly more analogy test wordswould have to be read than letter-sound correspondence testwords. The observed difference between reading analogy andletter-sound test words was in the hypothesized direction (more7475analogy words read than letter-sound words) but failed to reachstatistical significance.Given a larger sample and a statistically significantdifference between analogy and letter-sound word reading, such afinding would be evidence for the Goswami and Bryant (1990) viewthat children can use orthographic analogy when first beginningto read. Such evidence would also run counter to the positionthat children are able to use orthographic analogy only afterconsiderable experience with alphabetic recoding at the phonemelevel (Ehri, 1991, 1992b).The finding that children in the high prereading skill groupwere the only ones to benefit from the training suggests there isa relationship between learning to read words using orthographicanalogy and skills in rhyming, phoneme identity, and letter-soundknowledge. It also supports other research in analogicalreasoning that children perform better on analogical tasks whenthey are familiar with the operations and objects introduced(Brown, 1989; Brown, Kane, & Long, 1988).The treatment conditions used in the pilot study were basedon the assumption that the segmented text used by Baron (e.g.,'b', 'at', 'bat', 'ed', 'red') implicitly provided phonologicaltraining in onset and rime and that this phonological trainingwas a key element in the analogy test word reading that followed.However, no statistically significant differences on analogy wordreading were found among treatment conditions. This suggeststhat the ability to read words by orthographic analogy was notenhanced by the phonological training in onset and rime76(Conditions II, III) beyond the whole word training without it(Condition I).An important consideration is that there were only 15children in the pilot study and the finding of no statisticallysignificant differences among treatments was based only on thefour children who read analogy words (2 subjects in Condition I,1 subject in each of Conditions II and III). The followingdiscussion must be considered in light of the small sample size.Given the results of the pilot study, the use of theconditions which had phonological training in onset and rime(Conditions II, III) should be re-evaluated, and anothercondition be considered. There are three feasibleinterpretations.First, the pilot findings could be accepted as being true(at least for brief periods of training) and the recommendationbe made that the main study not include a condition which usesphonological training in onset and rime. Second, the pilotfindings could be considered an insufficient test of the possiblebenefits of phonological training on the use of orthographicanalogy due to the small sample size. A third alternative, giventhe pilot evidence for the use of orthographic analogy, is thatusing segmented text for training items would enhance the readingof analogy test words more than training with whole text, with orwithout phonological training with onset and rime.The pilot study provided evidence that prereaders can useorthographic analogy to read after only brief training with wordspresented as whole text. The use of whole text precluded the77possibility that the children used segmented text to read theanalogy words, the rival account Goswami and Bryant (1990)provided for the successful reading in Baron's study. Theimportant question that remains is whether the use of segmentedtext in training items (i.e., Baron's training sets) will resultin more analogy word reading than training with whole text.In order to compare the effects of training with segmentedtext with that of whole words, and to re-examine the possibleeffects of phonological training with onset and rime, it isrecommended that three treatment conditions be represented in themain study. One treatment condition would replicate pilot studyCondition I and use whole text for training words but not havephonological training in onset and rime. A second conditionwould also use whole text but would include phonologicaltraining. The third condition would use segmented text fortraining items (i.e., Baron's training) and include phonologicaltraining. These three conditions implemented together will allowfor an examination of whether the use of orthographic analogy isenhanced by phonological training and by training with segmentedtext over training with whole text alone.The use of phonological training in the third treatmentcondition would be a departure from Baron's original training butit is recommended so that the central distinguishing featurebetween conditions two and three will be the use of segmentedtext. If phonological training is not included in conditionthree, then there will be two central differences between theconditions (phonological training and segmented text), making the78effects of segmented text unclear. The cost of addingphonological training to the third condition is that a comparisonto Baron's (1977) study would not be exact.Additional Analyses The analyses up to this point have involved mainly groupperformances, and although they are sensitive tests of groupdifferences, they do not permit an examination of whether or notparticular phonological, letter-sound, and reading skills havebeen mastered by individual children. Analyses that examinedthese individual differences were conducted by other researchers(e.g., Bowey & Francis, 1991; Byrne & Fielding-Barnsley, 1990,1991) by giving each child a pass or fail rating on each relevantprereading measure and then examining the relationship betweensuccess on prereading skill and subsequent word reading ability.Following these researchers, children's performances wereclassified as High or Low on the three prereading skills ofrhyme, phoneme identity, and letter-sound knowledge. Subjectswere rated on each prereading skill by considering positive zscores to be High performance and negative z scores to be Lowperformance. The results of rating subjects on each prereadingskill and their subsequent word reading are presented in Table 7.As can be noted from Table 7, most of the analogy andletter-sound word reading was produced by children who were ratedHigh in all three prereading skills. It is also notable thatletter-sound knowledge, in combination with either rhyming orphoneme identity, was sufficient to produce some, albeit limited,word reading as well. Interestingly, five subjects were ratedHigh on rhyming skill alone but no subjects were rated High oneither phoneme identity or letter-sound knowledge alone.Table 7. Subject's Prereading Skills Rankings and Reading ScoresPrereadingSkill Ranking nReading Test ScoresAnalogy^Letter-soundHigh R, L, P 3 15 6High R, L only 1 2 0High L, P only 2 0 1High R, P only 0 0 0High R only 5 0 0High L only 0 0 0High P only 0 0 0Low R, L, P 4 0 0Note. R = rhyme, L = letter-sound knowledge, P = phonemeidentity.As previously mentioned, researchers noted during the pilotstudy that children in the Low prereading skills groups appearedto take many more trials to learn to read the training words tocriterion (2 successive trials without error) than children inthe High prereading skills group. The trials to criterion score,described by Yopp (1988) as a measure of the rate of readingacquisition, was used as the dependent variable to examine aprediction made by Goswami and Bryant (1990).7980Goswami and Bryant (1990) predicted that rhyming makes anindependent contribution to reading after accounting for theeffects of other variables such as age, language ability (PPVT-R), phoneme identity, and letter-sound knowledge (see Table 2 forthe correlations among the pretests and the trials to criterionscores).For a stringent test of the hypothesis that rhyming abilitycontributes independently to reading, Bryant et al. (1990)recommend entering the relevant variables into a fixed-ordermultiple regression analysis with rhyming skill as the lastvariable entered. This analysis would show whether thechildren's rhyming scores predicted reading after the influencesof the other variables were removed. The results of thisanalysis, shown in Table 8, must be viewed with caution due tothe small sample size.Table 8. Relation of Rhyme to Trials to Criterion AfterControlling for Related Variables Variable^ Cumulative R 2^R2 ChangeStep 1. Age .22 (.00)^.22 (.00)Step 2. PPVT-R^ .26 (.22)^.03 (.22)Step 3. Phoneme identity^.33 (.48)^.08 (.26)Step 4. Letter-sound knowledge^.34 (.48)^.01 (.00)Step 5. Rhyming^ .49 (.92)^.16 (.44)Note. Results are reported for the total sample (n = 15) and forthe 5-year-olds (n = 8) separately in parenthesis.81As shown in Table 8, rhyme accounts for 16% (44% with 5-year-olds) of the variance in the trials to criterion measureafter differences age, language skills, phoneme identity, andletter-sound knowledge were accounted for. These results lendsupport to Goswami and Bryant's claim that rhyming makes anindependent contribution to reading.The data was also analyzed with all possible subsetsregression (BMDP) where the best subset of independent variablesor variable is selected on the basis of Mallow's CP' the sampleR 2 , or the adjusted R 2 . The best subset selected according toMallow's CP was rhyming ability alone. The best subset accordingto the R 2 values were rhyme combined with letter-sound knowledge(R 2 = .49) followed by rhyming combined with phoneme identity (R 2= .48). These results also support the view that rhyming abilityis significantly related to beginning reading.A final additional analysis was performed on the phonemeidentity scores and is based on the prediction that children willbe able to identify more phonemes in the initial position ofwords than in medial or final positions. A positive findingwould support the view that rhyming skill facilitates theidentification of phonemes in the initial position (Goswami &Bryant, 1990). This is based on evidence that in CVC rhymingwords, the onset is represented by a phoneme so that identifyinginitial phonemes should be easier than when the phoneme is inanother position.In the pilot study, identifying the initial phoneme wasfound to be much easier than identifying the medial or finalphonemes, F(2, 12) = 6.97, p < .01). This finding is in linewith previous research (Kirtley et al., 1989).Recommendations The pilot study was instructive in many regards and thereare several improvements which are recommended for the mainstudy. First, 5-year-olds are the preferred age group for themain study. Seven of the subjects in the pilot study were 4-year-olds and most of these children had difficulty learning thetraining words and none were able to read any analogy or letter-sound test words. As well, four of the 4-year-olds knew noletter-sounds, which created a bimodal distribution for thatvariable.Most importantly, half of the 5-year-olds (4 out of the 8)benefited from the training and were able to read the analogyand/or letter-sound test words. The central question examined inthe study is whether prereaders can use orthographic analogy whenbeginning to read and based on the pilot study, many 5-year-oldsappear to have the prerequisite skills.The use of orthographic analogy is inferred on the basis ofsuccessful reading of the analogy test words and fewer readingsof the letter-sound test words. Several children in the pilotstudy revealed the strategy they used to read the test words byorally segmenting words and by pointing with their fingers. Itis recommended that children who read words correctly be askedfor the strategy they used to pronounce the test words. This8283information would be useful in determining the strategy they usedfor reading.There are five changes recommended for the treatmentconditions used in the pilot study. First, many children wereconfused by the initial letters in the training and test wordsfor set 2. Two training words began with the letters 'b' or 'd'and all four test words begin with 'b' or 'd'. The recommendedchange is to replace the initial 'd's with 'h's to eliminate thisconfusion. The training word 'dam' becomes 'ham' and the testword 'dug' changes to 'hug'. In addition, several children inthe pilot study appeared surprised and when the researchers namedthe test word 'bum'. 'Bum' will be changed to 'hum'.Second, it is recommended that the illustrations not be usedto accompany the training words. The illustrations focusedattention away from the text and seemed to encourage the use ofthe strategy of memorizing the order of the words without payingattention to the text.Third, to preclude the use of memorizing the word order inthe trials to criterion task, and to balance the order of readingtest items, the main study should use the Latin squares method ofbalancing, a) the order of training set presentation acrosssubjects, and b) the order of test word presentation within sets.The pilot study randomized a) but not b).Fourth, Condition II should be dropped so that only onecondition which provided onset-rime training with whole wordswould be retained for the main study. Condition III is preferred84over Condition II because the phonological training is moreexplicit.Finally, it is recommended that a treatment condition beadded that follows Baron (1977) and contains segmented text andonset-rime training.The three recommended treatment conditions would be asfollows. One condition would be limited to training with wholeword text and without phonological training in onset and rime(spontaneous analogy condition, SA). A second condition wouldtrain children to read whole word text and have onset-rimetraining (phonological condition, PH). The third condition wouldtrain children to read segmented text and include onset-rimetraining (orthographic segmentation condition, OS). Thesetreatment conditions build upon the pilot evidence that childrenmay learn to use orthographic analogy when beginning reading andadd examinations of the effects of phonological training andtraining with segmented text.Rating each prereading skill high or low on the basis of zscores and then examining the relationship between prereadingskill and word reading permits an examination of whetherparticular skills had been mastered by individual children (seeTable 7). It is recommended that this analysis also be used inthe main study.The analysis of the relationship among the pretests and thetrials to criterion measure was an effective method to determineif rhyming ability made an independent contribution to readingacquisition (see Table 8). If the trials to criterion measure is85normally distributed, it is recommended that this analysis beretained for the main study.To summarize, the recommendations for the main study are asfollows:1. The sample should consist of prereaders who are at leastfive years old.2. Children who read test words correctly should bequestioned about the strategy used.3. Items for set 2 should be altered to avoid confusionbetween the initial letters 'b' and 'd'.4. The training should not include illustrations.5. The order of training sets and reading test wordpresentation should be balanced by the Latin square method.6. Conditions I and III should be retained but Condition IIshould be dropped.7. A condition should be added that uses segmented textfollowing Baron (1977).8. Analyzing the mastery of particular skills by individualchildren should be added.9. The relationships between the prereading skills andlearning to read should be studied using the number of trials tocriterion as the dependent measure.Main StudyHypotheses The hypotheses for the main study are similar to those inthe pilot study but also incorporate the pilot recommendations.1. Prereaders will learn to read more words on the basis oforthographic analogy than by letter-sound correspondences. Ifprereaders can read more analogy test words than letter-soundcorrespondence test words, this hypothesis is supported. Apositive finding would provide evidence for the view thatbeginning readers can use orthographic analogy to begin reading(Goswami & Bryant, 1990, 1992) but not the position thatconsiderable experience with alphabetic recoding at the phonemelevel is a prerequisite (Ehri, 1991, 1992a).2. The ability of prereaders to learn to read words byorthographic analogy will be differentially enhanced by differenttraining conditions.a) Phonological training in onset and rime in addition totraining using whole word text (PH condition) will enhance theability to read words by orthographic analogy more than trainingusing whole word text alone (spontaneous analogy condition, SA).b) Phonological training in onset and rime using segmentedtext (OS condition) will enhance the ability to read words byorthographic analogy more than phonological training in onset andrime using whole word text (PH condition).c) Phonological training in onset and rime using segmentedtext (OS condition) will enhance the ability to read words by8687orthographic analogy more than training with whole word text (SAcondition).Comparisons of the training conditions will provide evidencefor the relative contribution of phonological and segmentedorthography training to reading by orthographic analogy. Thiswas not possible in Baron's (1977) study where effects ofphonological training in onset and rime, and training usingsegmented text were conflated.3. The ability of prereaders to learn to read words byorthographic analogy will vary according to their level ofprereading skills. Specifically, children in the High prereadingskills group will read more analogy reading test words thanchildren in the Low prereading group. A positive finding wouldbe evidence that one or more of the prereading skills of rhyming,phoneme identity, and letter-sound knowledge were related toreading words by orthographic analogy. It would also add to theexisting evidence in the analogical reasoning literature thatyoung children perform significantly better on analogical taskswhen they are familiar with the operations and objects introduced(Brown, 1989; Brown, Kane, & Long, 1988).4. Rhyming ability of prereaders will make an independentcontribution in learning to read after accounting for the effectsof age, vocabulary (PPVT-R), phoneme identity, treatment groupmembership, and letter-sound knowledge.A positive finding would support the position taken byGoswami and Bryant (1990) that rhyming makes an independentcontribution to beginning reading.88Subjects The children were solicited from six kindergarten classroomsin the city of Kamloops, British Columbia, Canada. Oneparticipating school was selected because of its proximity to theuniversity and the three other schools were selected by theSuperintendent of the school district because of the largenumbers of kindergarten children in attendance. The sixclassrooms were taught by four different teachers (two teacherstaught separate morning and afternoon classrooms), all of whomcheerfully agreed to participate in the study.The teachers sent notices home with 88 children to requestthe consent of the parent or guardian to have their childparticipate in the study. All 74 of the children who were givenconsent were selected to participate in the study (return rate of84%). From the original sample, two children moved during thecourse of the study, three were screened for reading ability, andthree were dropped because they declined twice consecutively tocontinue the training. The children who declined did not seem tofind the training unpleasant but rather appeared to find otherkindergarten activities, especially the sand box, moreattractive. The final sample consisted of 66 children.Based on pre-study information from kindergarten teachers,the children had no known language impairments and had English asa first language. The average age of the 66 children (35 girlsand 31 boys) was 5 years 8 months (SD = 4.15, range = 5 years 1month - 6 years 5 months).89Following Bryant et al. (1990), the social background of thechildren was estimated using information collected from parents(see Table 9). The national percentages are from 1986 censusdata.Table 9. Children's Background Measured by Father's Occupationand Mother's EducationPercentageMeasure^ National Group's^nFather's OccupationProfessionalIntermediate managerTechnical101322151212655Manual skilled 31 40 16Manual partlyskilled or unskilled23 18 7Unemployed 2 1Mother's EducationUniversity 8 40 16Vocational/technical 32 37 15High school 42 13 5No qualifications 18 10 4Note. The years of study at each level was requested formother's education.The father's occupation and the mother's education, wasrequested from the parents. Of the 66 children in the study, 40families returned the self-addressed and stamped envelopescontaining this information (return rate of 61%). The original90intention was to include the mother's education in some analyses(hypothesis 4) but the low return rate eliminated thispossibility. The social background data on the 38 respondents ispresented in Table 9.Caution must be observed when estimating the socialbackground information collected is used to determine thegeneralizability of the findings as information was not receivedon 28 of the children. However, Table 9 suggests thatprofessional occupations are overrepresented in the sample andintermediate managerial and technical occupations areunderrepresented. Mothers who have attended university arehugely over represented in the sample as are mothers withvocational or technical training. High school graduates andmothers with no qualifications are very much underrepresented inthe sample. Again, these estimates of the social background mustbe viewed with caution as there were 28 families who did notreturn the questionnaires on social background.Design The study used an experimental design with three treatmentconditions; spontaneous analogy (SA), phonological training inonset-rime (PH), and orthographic segmentation (OS) training.Two prereading skills groups (Low and High) based on the pretestscores of rhyming, phoneme identity, and letter-sound knowledge,were crossed with the three treatment conditions. This produceda 2 by 3 (2 prereading skills groups by 3 treatment conditions)factorial design.Rhyming ability, phoneme-identity, letter-sound knowledge,age and PPVT-R served as independent variables. The threedependent variables in the study were the average number oftrials to criterion taken to learn to read the words in thetraining sets, reading orthographic analogy test words, andreading letter-sound correspondence test words.Procedure The researcher visited the children at the participatingschools before beginning any testing or training. A trainedassistant who followed the same procedure was hired to collectthe data in one of the six classrooms. The decision to beginworking with the children was based on the recommendation of theparticipating kindergarten teachers. All testing and trainingwas carried out in rooms adjacent to or inside the classroom.The children were initially screened for reading abilitywith eight words selected from Baron's (1977) reading test (bed,pit, bet, sin, hug, man, bag, rag). Children able to read morethan one word were screened from the study. Three children whoread only one word were included. Sixty-three of the childrenwere unable to read any of the screening test words, in spite ofencouragement. Most children commented that they could not readand several named individual letters in test words but wereunable to read the word. The finding that only three childrenwere readers was consistent with related research indicating thatfew children this age can read (Bryant et al., 1990; Byrne &Fielding-Barnsley, 1991), as well as with the pilot study data.9192Children were then given the Peabody Picture Vocabulary Test(PPVT-R) to estimate their receptive language and vocabularyskills. No child scored lower than one standard deviation belowthe standardized mean of 100 (sample standardized mean = 103.6,SD = 11.2).In the second session, children were tested for the threeprereading skills of rhyming, phoneme identity, and letter-soundknowledge. Stratified random assignment based on these pretestscores was used to assign children to Low or High prereadingskills groups. The pretest scores were converted to z scores andthen summed for each subject. Subjects were placed into High orLow prereading skill groups based on the sign of the summed zscores. Children with negative z scores were assigned to the Lowprereading skills group and children with positive z scores wereassigned to the High group. Children were then randomly assignedfrom within prereading skills groups to one of the threetreatment conditions, spontaneous analogy (SA), phonologicaltraining (PH), or orthographic segmentation (OS). The pilot dataprovided empirical support for splitting the group according tolevels of prereading skill, as only children in the High groupwere able to read any test words. Results from the main studyprovided further support for splitting the group in this manner,as prereading skills group membership proved to be an excellentpredictor of test word reading.As in the pilot study, the training consisted of teachingthe training items and then, following each training set, testingthe children for analogy and letter-sound correspondence word93reading. The training sessions lasted less than 20 minutes each,with one or two training sets taught per session.A primary emphasis of the training was to make the sessionsenjoyable for the children and they only took place when thechildren were amenable. Puppets were introduced during thetraining to assist in making the training more enjoyable.The order of training set and reading test word presentationwere balanced within groups using the Latin square methodfollowing Ott (1984). Children who began training with set 1were tested for reading with the first reading test word ('bed').Children who began training with set 2 were tested for readingwith the second reading test word ('bam'), and so on.Following the recommendations from the pilot study, theorthographic segmentation (OS) condition was like Baron's (1977)training but with two minor changes (see Figure 6).Training Sets 1 2 3 4b h s rat ug in ugbat bug pin muged am at anred ham sat ranReading Test SetsType 1. Analogy bed hug sin rugAnalogy rat bam pat manType 2. Letter-sound bad hum pit ragLetter-sound bet bag sit runFigure 6. Training and reading test sets for the OS condition.94First, words starting with the letter 'd' in set 2 werealtered to avoid the confusion children in the pilot study foundbetween the seven items beginning with 'd' or 'b' in that set.Second, children were given phonological training in onset andrime, the training which could have been implicitly taught bylearning the segmented text in Baron's (1977) study. The readingtest words were identical for all treatment conditions.The phonological condition (PH) followed the same procedureas the OS condition but used whole words in the training, notsegmented text (see Figure 7). The use of whole words precludedGoswami and Bryant's (1990) caution of Baron's (1977) results.The phonological training (PH) also explicitly taughtsegmentation and blending of onsets and rimes (e.g., "bat", "b-at",^"bat").Training Sets 1 2 3 4bat bug pin mugmat mug tin bugred ham sat ranted ram mat panReading Test SetsType 1^Analogy bed hug sin rugAnalogy rat bam pat manType 2. Letter-sound bad hum pit ragLetter-sound bet bag sit runFigure 7. Training and reading test sets for the SA and PHconditions.The spontaneous analogy (SA) condition was like thephonological (PH) condition but did not include phonological95training or explanations for incorrect responses to reading testwords. The spontaneous analogy condition (SA) consisted solelyof teaching prereaders to read two pairs of rhyming words. Thusit is a stringent condition for the use of orthographic analogy.Children who were able to read test words in all treatmentconditions were asked to explain how they arrived at the correctpronunciation. This information was gathered to help determinethe strategy the children used to read the test words correctly.Spontaneous Analogy (SA) Condition The SA training was similar to Condition I in the pilotstudy but for two minor changes. First, illustrations were notused so the training words were presented as text alone. Second,the order of training set and reading test word presentation werebalanced using the Latin square method.Briefly, children in the SA condition were introduced to thepuppets and then presented with the four training words in theset. They received two practice trials where the researchernamed each word. The number of trials the child took to read thefour training words to a criterion of two consecutive trialswithout error (to a maximum of 12 trials) was recorded. Thisformed the score for the trials to criterion reading measure. Asin the pilot study, the relationship between the number of trialstaken and the criterion score were inverted so that fewer trialswere represented by higher scores. Children were asked to readthe analogy and letter-sound correspondence reading test wordsimmediately following the trials to criterion.96Following Baron (1977), corrective feedback was given aftereach incorrect response during the reading test. Unlike Baron,the corrective feedback for the SA condition consisted only ofgiving the correct reading of the word responded to incorrectly.No explanation was provided.Phonological Training (PH) ConditionThe PH training condition was the same as the SA conditionexcept for two additions. First, the trials to criterion wasfollowed by phonological training with onset and rime. Second,following Baron, corrective feedback consisted of namingincorrectly read words and providing an explanation.Following the trials to criterion, children were invited toplay a game with sounds and were told that the puppet can say thetraining words so that they have two sounds. One puppet(operated by the researcher) pointed to the letter representingthe onset of the training word and said the sound (e.g., "b").The other puppet (operated by the child) was asked to imitate theresearcher. The researcher then pointed to the rime of the word,said the sound (e.g., "-at"), and then asked the child to repeatthe sound. This process was followed to teach onset-rimeblending (e.g., "b-at", "bat) and was repeated twice for eachpair of rhyming words. The analogy and letter-soundcorrespondence reading test words, balanced by Latin squares,were presented following the phonological training.The second feature added to the PH training from the SAtraining was that the corrective feedback provided after an97incorrect reading of a test word included an explanation. Theexplanation consisted of pointing out the similarities betweenthe test word and the training words. Following the explanationprovided by Baron (1977), the sounds of individual letters orspelling sequences within words were not given. The explanationconsisted only of pointing out common letters and rime sequencesbetween the test and training words.In summary, the PH treatment condition is like the SAtreatment condition but includes phonological training andprovides for an explanation after incorrect readings of a testword. The main difference between the PH and orthographicsegmentation (OS) condition is that the OS condition usessegmented text in training items.Orthographic Segmentation (OS) Condition The OS training condition followed the same procedures asthe PH training. The only difference between the two conditionswas in the stimulus items used during training, as the trainingitems for the OS condition contained segmented text.Briefly, children were taught to read the training set itemsby repetitions with corrective feedback to a criterion of twosuccessive trials without error (to a maximum of 12 trials). Asin the PH condition, the children received phonological trainingin onset and rime on the whole words in the training set.Finally, the children were asked to read the analogy and letter-sound reading test words. The feedback for incorrectly readingtest words consisted of naming the word correctly, and pointingout similarities between test words and training items (withoutproviding sounds), as in the PH condition.98V. RESULTSPretest MeasuresThe children were given the PPVT-R in the first session andthen tested for rhyming ability, phoneme identity, and letter-sound knowledge in the second session. Assignment to prereadinggroups was based on the summed z scores for rhyming ability,phoneme identity, and letter-sound knowledge. Children were thenrandomly assigned from within each stratified prereading skillsgroup to one of the three treatment conditions. The results ofthe pretests for the 66 children in the study are presented inTables 10, 11, and 12.Table 10. Pretest Scores for Prereading Skills GroupsMeasurePREREADING SKILLSLow^HighMean(SD) Mean(SD)TotalMean(SD)Agea 67.15(4.21) 69.03(3.93) 68.09(4.15)PPVT-Rb 101.21(13.42) 106.06(10.26) 103.60(11.20)Rhyme 5.70(2.14) 8.06(1.17) 6.88(2.09)Phonemeidentity2.39(2.12) 9.00(3.54) 5.70(4.41)Letter-soundknowledge4.70(4.08) 16.49(6.02) 10.59(7.83)99aAge is given in months.bPPVT-R scores are standardized.100All pretest measures appeared to be normally distributedwith the exception that phoneme identity was positively skewed.Ten of the 66 children were unable to identify any phonemes.Two-tailed independent t-tests found that the prereading skillsgroups differed significantly on rhyme, t(64) = 5.56, p < .001,on phoneme identity t(64) = 9.21, p < .001, letter-soundknowledge, t(64) = 9.32, p < .001, and on the PPVT-R, t(64) =2.50, p < .05. Surprisingly, the prereading skills groups didnot differ significantly on age, t(64) = 1.88, p > .05, the soledirect measure of maturation.Table 11 presents the pretest scores for the three treatmentconditions.Table 11. Pretest Scores for the Three Treatment ConditionsTREATMENT CONDITIONSA^PH^OSMeasure^Mean(SD)^Mean(SD)^Mean(SD) Agea 69.05(4.08)^67.82(4.07)^67.41(4.31)PPVT-Rb^102.59(13.54) 101.73(10.03)^106.59(12.80)Rhyme^7.14(2.15)^6.73(2.00)^6.77(2.18)Phoneme^5.14(4.34)^5.50(4.44)^6.45(4.55)identityLetter-^16.68(8.54)^10.41(7.42)^10.68(7.58)soundknowledgeaAge is given in months.bPPVT-R scores are standardized.101Children were randomly assigned to treatment conditionsusing a random number table and as expected, the pretest meansand standard deviations for the three treatment conditions weresimilar.Table 12 presents the pretest scores for the six treatmentgroups.Table 12. Pretest Scores for the Six Treatment GroupsTreatmentConditionMeasureLowPREREADING SKILLSHighSAMeanSDPHMeanSDOSMeanSDSAMeanSDPHMeanSDOSMeanSDAgea 68.00 67.36 66.09 70.09 68.27 68.734.20 4.65 3.91 3.86 3.55 4.45PPVT-Rb 100.73 98.59 104.36 104.46 104.91 108.8210.77 10.08 14.36 10.46 9.35 11.28Rhyme 6.27 5.36 5.46 8.00 8.09 8.092.37 1.92 2.21 1.55 .83 1.14Phoneme 1.91 2.36 2.91 8.36 8.64 10.00identity 1.92 2.34 2.17 3.59 3.78 3.47Letter- 3.00 5.09 6.00 18.36 15.73 15.36soundknowledge2.41 3.89 5.22 4.18 6.18 7.37aAge is given in months.bPPVT-R scores are standardized.Table 13 shows the correlations among the pretest scores,identity the trials to criterion measure, and analogy word102reading. The letter-sound reading test scores were not includedbecause of apparent violations of normality.Table 13. Correlations Among Pretests, Trials to Criterion, andAnalogy Word ReadingMeasure^1^2^3^4^5^6^71. Rhyme -2. Letter-sound^.42***^_3. Phonemeidentity^.57*** .70 ***^-4. PPVT-Ra^.38**^.33**^.35**^-5. Age^.23^.25^.34**^.49***6. Criterion^.63*** .50*** .58 *** .31 *^.20^-Average7. Analogy^.42*** .61 *** .76 *** .24^.22^.51*** —ReadingNote. These correlations are descriptive not inferential. Basedon a Bonferroni correction, r critical = .36, p < .05.aPPVT-R are raw scores.*p < .05. **P < .01. ***p < .001.The trials to criterion was the number of times taken by thechild to learn to read the training sets, with correctivefeedback. As in the pilot, the criterion scores were inverted sofewer trials are represented by higher scores.The prereading skills were highly and significantlyinterrelated. Age was the notable exception, but for themoderate relationship between age and the PPVT-R scores. Also103notable is the strong relationship between the three variablesused to classify the subjects according to High or Low prereadingskills groups (rhyming ability, letter-sound knowledge, andphoneme identity) and both the trials to criterion and analogyword reading measures.The reliabilities of the prereading skills tests werecalculated using Cronbach's alpha. The reliability estimateswere .73 for rhyming ability, .91 for phoneme identity (.84, .87,and .88 respectively for identifying phonemes in the initial,medial, and final positions), and .84 for letter-sound knowledge.These estimates were consistent with previous research (e.g.,Bryant et al., 1990; Yopp, 1988) and were sufficiently high tosuggest the prereading skills tests were measured reliably.The reliabilities of the reading measures were .93 fortrials to criterion, .86 for analogy word reading, and .71 forletter-sound correspondence word reading.Data Analysis Data ScreeningPedhazur (1982) and others (e.g., Shavelson, 1988) recommendscreening the data for extreme residuals, as extreme residualsmay seriously distort the results. Extreme residual are definedby Pedhazur (1982) as standardized residuals greater than 2.00(i.e., z > 2.00).The three dependent variables for the hypotheses to betested, the trials to criterion, analogy word reading, andletter-sound word reading, were all screened for extreme104residuals. Three children had scores that were detected asextreme residuals on six occasions, two occasions per child. Theresults of the screening for extreme residuals is presented inTable 14.Table 14. Screening for Extreme Residuals in Dependent VariablesCriterion^Analogy Letter-soundOutlier Group Membership Average Reading Reading1^SA - Higha^ 2.24^4.752^PH - High 2.35^2.053^OS - High^2.16^-2.10aHigh prereading skills group.Two of these children (outliers 1 and 2 in Table 14) mayhave been readers who were failed to be screened from the studyfor reading ability, as both read near the maximum number ofanalogy and letter-sound correspondence reading test words. Onewas a boy and the other a girl.The third child was in the High orthographic segmentationcondition (OS) and was the only member of that group who failedto read any analogy reading test words. This child was shy andappeared to reluctant to make a response unless she was confidentof the answer. She took the maximum number of trials to learnthe training sets, perhaps for the same reason, and was detectedas an outlier for the trials to criterion measure as well.105The hypotheses that used analogy word reading or trials tocriterion as dependent variables were first analyzed with theoutliers included and then again with the outliers deleted. Allfindings that were statistically significant at p < .05 remainedso in both sets of analyses and in fact the mean differencesincreased with the outliers deleted. As well, all findings thatwere not significant remained so after the outliers were dropped.Pedhazur (1984) recommends that extreme scores be correctedor deleted if they cannot be corrected because they may seriouslydistort the results. The scores in question were not data entryerrors and could not be corrected. Rather than deleting thescores, a more conservative procedure is to retain outliers butto assign them the value of the subject one less than the nextsubject closer to the mean. The analogy word reading scores foroutliers 1 and 2 were reduced to 5 from original scores of 7 and8. The score of the third outlier was increased to 2 from 0.The effects on the means and standard deviations of the affectedgroups for analogy word reading are presented in Table 15.Table 15. Effects of Adjusting Outliers on Analogy Word ReadingOutliers^OutliersIncluded Adjusted^DifferenceGroup^Mean SD^Mean SD^Mean SDSA High 2.55 (2.58) 2.36 (2.29) -.19 (-.29)PH High 2.82 (2.75) 2.55 (2.30) -.27 (-.45)OS High 4.73 (2.49) 4.91 (2.16) +.18 (-.33)Note. There was one outlier in each condition.106Adjusting the analogy word reading scores of the threeoutliers in the manner described increases the total varianceaccounted for by the model (2 prereading skills groups by 3treatment conditions) from 43% to 50% and decreases the meansquare error from 4.00 to 3.13. These differences, according toPedhazur (1984), would be conservative estimates of the amount ofdistortion generated by the extreme outliers. These improvementsin the fit of the model combined with the lack of change in thestatistical significance of the results led to the decision toadjust the scores of the outliers in the manner described. Thedata analyses reported that use analogy word reading as thedependent measure are with the scores of the three outliersadjusted.Adjusting the trials to criterion score for the one outlier(increased to 5.75 from 0) increased the mean for that group(orthographic segmentation condition, High prereading skills)from 6.84 to 7.36 and decreased the standard deviation from 2.01to 1.83. The analyses reported that used trials to criterion asthe dependent measure were calculated with the score of the oneoutlier adjusted. Adjusting the letter-sound correspondencereading scores for the two outliers (both scores of 5 wererescored as 3) reduced the mean from .57 to .53 and the standarddeviation from 1.36 to 1.19.Data screening revealed that both the analogy and letter-sound word reading test scores were positively skewed (skewness =1.06 and 2.44 respectively). Thirty-three children were unableto read any analogy reading test words and 49 children were107unable to read any letter-sound correspondence test words. Thissuggested that the population normality assumption of the t-testfor these analyses was violated.In spite of evidence that t-tests are robust for violationsof normality and homogeneity of variances given equal samplesizes (Glass & Hopkins, 1984), all t-tests that had eitheranalogy or letter-sound correspondence word reading were re-analyzed using transformed scores to adjust for positiveskewness. Following Tabachnick and Fidell (1989), three separatetransformations were used to adjust both word reading testscores. The three transformations used the following formulae:1. X = Square root (X).2. X = Log (X + 1)/Log (10).3. X = 1/(X + 1).The t-test for hypothesis 1 was re-analyzed following eachtransformation and in all cases the differences remainedstatistically significant at p < .05. As a result, the t-testreported for hypothesis 1 was calculated with the original data.Hypothesis 1 Prereaders will learn to read more words on the basis oforthographic analogy than by letter-sound correspondences.Thirty-three children were able to read the analogy readingtest words on 124 occasions (mean = 1.83, SD = 2.40) whereas only17 children were able to read the letter-sound correspondencetest words on 35 occasions (mean = .53, SD = 1.19). A 1-taileddependent t-test found that the difference between these means108was statistically significant, t(65) = 5.74, p < .001. Thisfinding was evidence that the prereaders used orthographicanalogy to read the analogy test words and that reading byletter-sound correspondence was much more difficult. Table 16shows the number of analogy and letter-sound reading test wordsread correctly.Table 16. Occasions Analogy and Letter-sound Test Words ReadCorrectlyWord Type Test Words TotalsAnalogy bed (18) hug (13) sin (13) rug (12) 56Analogy rat (15) bam (15) pat (19) man (19) 68Letter-soundbad (7) hum (4) pit (2) rag (2) 15Letter-soundbet (3) bag^(4) sit (7) run (6) 20Note. The number of occasions the word was read correctly are inparenthesis.The analogy word reading scores ranged from 12 to 19 and todetermine if these differences were significant, the analogy testwords were coded and the means were compared. There were apossible 528 occasions for these words (8 words by 66 subjects)An ANOVA of mean correct readings per analogy test word revealedthat the differences among words for correct readings was notstatistically significant, F(7, 520) = .67, p > .05, MS e = .18.Children were asked to explain how they read the test wordfollowing each successful reading. Many children did not provide109explanations that described their strategy (e.g., "I just readit", "I don't know", "I can read now") whereas several childrengave clear explanations. The most common explanation, given by15 children, was that the test word rhymed with a training word.This was additional evidence that the analogy children made wasbased on rhyming ability.Testing Rival Explanations Four rival explanations to the use of orthographic analogythat could possibly account for the significant differencebetween analogy and letter-sound test word reading were examined.They will be discussed in order.First, the argument could be made that the majority of theanalogy test word reading was accomplished by children in theorthographic segmentation condition, and if Goswami and Bryant'scritique is true then the reading could have occurred without theuse of analogy. If this is true, then the prediction would bethat children in the spontaneous analogy (SA) treatment conditionwould not have read more analogy than letter-sound reading testwords. A 1-tailed dependent t-test that compared the analogy andletter-sound reading scores for the SA condition found that thesignificant difference remained, t(21) = 2.81, p < .01. Thefirst rival explanation was not supported.Second, the analogy reading test words were the first to bepresented to children who began with training sets 1 or 2 whilechildren beginning sets 3 and 4 were tested first with letter-sound words. Balancing the set and word reading presentation110order had this effect. The analogy test words rhymed with one ofthe training words and children beginning with training sets 1and 2 may have simply guessed a rhyming word correctly for thetest word without the use of text, or even looking at the testword. Children who began the training with sets 3 or 4 would notbe rewarded for this strategy, at least not immediately. If thisguessing strategy was used by children who were presented withtraining sets 1 and 2 first, a significant effect for the orderof training set presentation is predicted, with childrenbeginning with training sets 1 and 2 scoring higher on analogyword reading. Also, it is more likely that the analogy readingtest words in training sets 1 and 2 would be read more often thanin sets 3 and 4 where the letter-sound test words are presentedfirst. The order of presentation of the training sets was notsignificant, F(3, 62) = .15, p > .05, MS e = 6.00. There were nosignificant differences among the means for the number of timeseach analogy test word was read, F(7, 520) = .67, p > .05, MS e =.18. The explanation that significant analogy word reading wasaccomplished by systematic guessing by children beginning withtraining sets 1 and 2 was not supported.Another possible guessing strategy which could be viewed asa rival hypothesis to explain significantly more analogy thanletter-sound word reading is that the children guessed a wordthat rhymed with a training word on every occasion a test wordwas presented. This explanation predicts that there would be nosignificant differences in analogy word reading between the Lowand High prereading skills groups or among the three treatment111conditions. Later analyses will provide evidence that thereindeed were significant differences on analogy reading testscores between the Low and High prereading skills groups as wellas between two treatment conditions. The rival explanation thatchildren constantly guessed a rhyming word for analogy readingtest words was not supported.The final rival explanation examined was tested and rejectedby Baron (1977) but will be reconsidered here. It was possibleto read an analogy reading test word correctly after sounding outtwo parts accurately (the initial letter and the last two as one,representing the rime). In order to correctly read a letter-sound test word, however, it was necessary to make three soundsaccurately, one for each letter. If the same number of wordparts were read correctly for both the analogy and letter-soundreading test words, the result would be significantly moreanalogy word reading.This possibility can be tested by counting the number ofparts of analogy and letter-sound test words read correctly. Ifthis explanation is true, there would be no difference in thenumber of parts of analogy and letter-sound test words readcorrectly, or more letter-sound test words parts would have beenread correctly. This is a stringent test because the last twosounds of the analogy test word both had to be correct to countas a single correct part, and because the maximum number ofcorrect parts would be two for each analogy test word but threefor each letter-sound test word. In spite of this lower ceilingfor the analogy test words, a 2-tailed dependent t-test found112that there were more parts of these words read correctly than ofthe letter-sound test words, t(65) = 2.77, p < .01. Theexplanation that fewer number of letter-sound test words wereread correctly because they required the correct reading of moreword parts than analogy words was not supported.There were two ways to examine the effects of treatmentconditions on word reading while controlling for prereadingskills. One alternative was to split the sample according tolevels of prereading skills, as in the pilot study. Thisprocedure would follow previous research that categorizedprereaders as pass/fail on prereading skills and then examinedreading within these groups (e.g., Bowey & Francis, 1991; Byrne &Fielding-Barnsley, 1990, 1991). Also, the pilot study providedevidence that prereaders' ability to learn to read analogy testwords was bimodaly rather than normally distributed. The maindisadvantage of this procedure was that variance within treatmentconditions would be reduced.The second alternative was to use the prereading skills ascovariates and retain the variance on analogy word reading withintreatment conditions. Prereading skills would be analyzed ascontinuous variables but analogy word reading could not beexamined under different levels of prereading skills.To provide for the most complete examination, both analyseswere conducted and as expected they were consistent. Theanalyses that split the sample according to the level ofprereading skills are presented first.113Planned Orthogonal Contrasts The method of planned orthogonal contrasts (POC) was used toexamine hypotheses 2a, 2b, 2c, and 3. The POC method wasselected because it is considered to be the most powerful test ofmean differences (Glass & Hopkins, 1984, Shavelson, 1988). Thisincreased power comes at the price of two constraints. First,hypotheses about differences among specific means must be setforth prior to the data collection and this condition was met.Second, the contrasts must be orthogonal. In the POC method, allthe sums of squares between groups is decomposed into J - 1orthogonal parts, each part associated with a contrast. Thecondition of orthogonal contrasts was also met. As there weresix groups in the study (2 prereading skills groups by 3treatment conditions), there were five possible orthogonalcontrasts. The POC matrix used to test hypotheses 2a, 2b, 2c,and 3 along with the results are presented in Table 17.POC uses a contrast-wise Type I error rate so that the errorrate increases with each contrast. To control for this it isrecommended that the specified Type I error rate be divided bythe number of contrasts made (Shavelson, 1988). As fivecontrasts were made in the study (see Table 17), the Type I errorrate was set at p < a/5, p < .01.Briefly, the first four contrasts made comparisons among thethree treatment conditions and were designed to estimate theeffects of phonological training and the use of segmented text onanalogy word reading. The hypothesis 2a contrast compared theanalogy word reading means of the spontaneous analogy (SA) and114phonological training (PH) treatment groups. Contrast 2bcompared the analogy word reading means of the phonologicaltraining (PH) and the orthographic segmentation (OS) conditions.The 2c contrasts compared the spontaneous analogy (SA) andorthographic segmentation (OS) means.Table 17. Planned Orthogonal Contrast Matrix for Analogy WordReadingPREREADING SKILLSLow^ HighTreatment SA^PH^OS^SA^PH^OSConditionMean^.18^.09^.91^2.36^2.55^4.91SD .41^.30^1.81 2.29^2.30^2.17Hypotheses t2a^0^0^0^1^-1^0^.08 ns2b^0^0^0^0^1^-1^2.98**2c^0^0^0^1^0^-1^3.06**2c^1^0^-1^0^0^0^.97 ns3 1^1^1^-1^-1^-1^6.61***Note. t-tests were independent 1-tailed, MS e = 3.13.ww--p < .01. ***p < .001. ns = not significant.Contrast 3 compared the mean of the Low prereading skillsgroup with the mean of the High prereading skills group. Itexamined the possible effects of High or Low prereading skills onanalogy word reading, regardless of treatment condition.115Hypothesis 2a Phonological training in onset and rime in addition totraining using whole word text (PH condition) will enhancethe ability to read words by orthographic analogy more thantraining using whole word text alone (spontaneous analogycondition, SA).This hypothesis examines whether phonological rhyme-basedtraining contributes to analogy word reading over the limitedtraining of simply learning to read two rhyming pairs of words.The contrast was made only with the High prereading skillsgroups. The mean of analogy word reading by the spontaneousanalogy (SA) group was 2.36 (SD = 2.29) while the mean for thephonological training (PH) group was 2.55 (SD = 2.30). The 1-tailed independent t-test found that the difference between thesemeans was not statistically significant, t(20) = .08, p > .05.This was evidence that phonological training did not benefitanalogy word reading over training in whole words alone, in spiteof the extra time children in the PH condition experienced withthe training words.Also, the explanation children in the PH group receivedfollowing an incorrect reading of a test word (similarities intext between the test and training words were pointed out) didnot increase analogy word reading.This was not a stringent test of the hypothesis because thephonological training was very limited (approximately 10 minutesin total).116Hypothesis 2bPhonological training in onset and rime using segmented text(OS condition) will enhance the ability to read words byorthographic analogy more than phonological training inonset and rime using whole word text (PH condition).This contrast was made with the High prereading skillsgroups only. The mean number of analogy words read by theorthographic segmentation (OS) group was 4.91 (SD = 2.17) whilethe mean for the phonological training condition (PH) was 2.55(SD = 2.30). Ten of the 11 children in the High prereadingskills OS group read two or more analogy reading test words butonly 7 of the 11 children in the High prereading skills PH groupaccomplished this task. An independent 1-tailed t-test foundthat the difference between the means was statisticallysignificant, t(20) = 2.98, p < .01, and it was evidence that thesegmented orthography used in the OS group enhanced analogy wordreading over phonological training with whole words.A rival explanation that may account for the observedsignificant difference between the means is that perhaps thesegmented orthography training (OS condition) was more difficultthan learning to read whole words (PH condition). If this wastrue, then the extra training time experienced by the children inthe OS condition could account for their increased analogyreading scores. This possibility was examined by comparing theaverage number of trials taken to learn to read the trainingwords. It predicts that OS group would have taken significantly117more trials to reach criterion (2 successful trials withouterror) than the PH group.The High prereading skills OS group took an average of 2.64trials to learn the training words to criterion (SD = 1.83) andthe High prereading PH group took an average of 2.11 trials (SD =1.57). An independent 2-tailed t-test found that the differencebetween these means was not statistically significant , t(20) =.72 , p > .05. The explanation that extra experience for the OSgroup with the training words accounts for the significantlyhigher analogy reading scores was not supported.Hypothesis 2cPhonological training in onset and rime using segmented text(OS condition) will enhance the ability to read words byorthographic analogy more than training with whole word text(SA condition).This question was examined separately for the Low and Highprereading skills groups. The mean of analogy word reading forthe OS High prereading skills group was 4.91 (SD = 2.17) and 2.36for the SA High prereading group (SD = 2.30). An independent 1-tailed t-test found that the difference between these means wasstatistically significant, t(20) = 3.06, p < .01. This findingwas evidence that the combination of phonological training andtraining with segmented text led to more analogy word readingthan training with whole words alone, at least for children withabove average prereading skills.118The mean of analogy word reading for the OS Low prereadingskills group was .91 (SD = 1.81) and .18 for the SA Lowprereading group (SD = .41). An independent 1-tailed t-testfound that the difference between these means was notstatistically significant, t(20) = .98, p > .05. This findingwas evidence that the combination of phonological training withsegmented text was not sufficient to enhance analogy word readingwith children that were below average on the prereading skills.A rival hypothesis that may account for the observeddifference among treatment group means was that one gender mayhave significantly more natural ability to learn to read than theother. This possibility was examined following Goswami (1986) byincluding gender as a variable in a 2 by 3 by 2 ANOVA (2prereading skills groups by 3 treatment conditions by 2 genders)with analogy word reading as the dependent variable. The maineffect for gender was not statistically significant, F(1, 54) =.36, p > .05, MS e = 2.30, and neither were the interactions.The hypothesis that differences in gender accounted for theobserved differences in analogy word reading was not supportedand gender was dropped from further analyses.Figure 8 graphs the total analogy word reading scores foreach training set over the four training sessions. As thetraining sets were balanced for order of presentation using theLatin square method, the trial number represents the order inwhich the sets were presented to the child, not the set number asin Figure 7. This difference along with unequal time intervalsbetween trials precluded trend analysis.14tu 12-E2 10O6co2 3^4160 ^Trial Number4 -2 -•■-441111111111400,..OS High0^ PH HighSA High^ OS Low• SA Low^ PH Low119Figure 8 graphically illustrates the superior analogy wordreading demonstrated by the OS High prereading skills groupcompared to the other groups. Note that the three groups withLow prereading skills are collapsed along the x axis. The 33children in the Low prereading skills groups read only 13 analogyreading test words in total.Figure 8. Analogy Word Readina Over Training Sets Hypothesis 3 The ability of prereaders to learn to read words byorthographic analogy will vary according to their level ofprereading skills.120This hypothesis makes two predictions. First, hypothesis 3predicts that children in the High prereading skills group willread significantly more analogy reading test words than childrenin the Low prereading group. Second, it predicts that childrenin the High prereading skills group will take fewer trials tolearn the words in the training sets than children in the Lowgroup.Hypothesis 3 was examined with analogy word reading as thedependent variable using the method of planned orthogonalcontrasts. The independent 1-tailed t-test found that the Lowand High prereading skills groups differed significantly on theability to read the analogy test words, t(64) = 6.61, p <.001.Children in the High prereading group read analogy test words on111 occasions (mean = 3.27, SD = 2.48) whereas children in theLow prereading group read only 13 analogy test words (mean = .39,SD = 1.12).To examine hypothesis 3 using the trials to criterionaverage scores as the dependent variable (average number oftrials taken to learn the training sets), an independent 1-tailedt-test was performed on the criterion average means of the Highand Low prereading skills groups. The average criterion scorefor the High prereading skills group was 2.11 (SD = 1.77) and6.54 (SD = 3.29) for the Low group. Children in the Lowprereading skills group took, on the average, over three times asmany trials to learn to read the training items as children inthe High group. Not surprisingly, the difference between thesemeans proved to be significant, t(64) = 6.81, p < .001).121The findings from the t-tests using analogy word reading andthe average trials to criterion scores as dependent variablesconverged. This convergence coupled with the magnitude of thedifferences in performance between groups was convincing evidencethat one or more of the prereading skills measured was importantfor the successful reading of the analogy test words.The methods of data analyses to examine hypothesis 3 up tothis point used group means for comparisons (planned orthogonalcontrasts, t-tests). Following the pilot, children were rated oneach prereading skill by considering positive z scores to be Highperformance and negative z scores to represent Low performance.Table 18 summarizes the results.Table 18. Subject's Prereading Skills Rankings and Reading ScoresREADING TEST SCOREPrereading^ Letter-Skill Ranking Analogy^sound^nHigh R, L, P^89 29^22 33High R only 9 1^11 17High R, P only 12 0 6 9High R, L only 5 0^4 6High L only 6 4 3 5High L, P only 1 0^1 2High P only 0 0 1 2Low R, L, P 2 1^18 27Note. R = rhyming ability, L = letter-sound knowledge, P =phoneme identity.122Note that most of the words that could be read by analogywere read by children who were rated as High on all three skills.Further, these 22 children accounted for almost all of theletter-sound word reading. By contrast, the 18 children who wereranked Low on all three prereading skills read a combined totalof only three analogy and letter-sound reading test words. Thisis further empirical support for splitting the sample into Highand Low prereading skills groups based on the added z scores.It is also noteworthy that the High letter only group readabout equal numbers of analogy and letter-sound reading testwords, suggesting that they read by letter-sound correspondence.Analogously, the High rhyme only group had some success readingthe analogy reading test words but almost none reading theletter-sound test words.Analysis of CovarianceThe effects of the treatment conditions were also beexamined by using the prereading skills as covariates and notsplitting the group into Low and High prereading skills groups.This analysis treated the prereading skills as continuousvariables and controlled for differences in these skills whileexamining treatment effects. The homogeneity of slopesassumption was examined prior to this analysis (see Appendix D)and the evidence suggested that it was plausible.The results of the ANCOVA with analogy word reading as thedependent variable and rhyming ability, letter-sound knowledge,and phoneme identity as covariates are presented in Table 19.123Table 19. Analogy Word Reading With Prereading Skills asCovariatesSS DF MS F ID R2SourceTreatment 21.75 2 10.88 4.88 .01 .05Rhyme .09 1 .09 .04 .84 .00Letter-sound 7.23 1 7.23 3.24 .08 .04Phoneme id. 48.43 1 48.43 21.72 .00 .24All covariates 201.15 3 67.05 30.07 .00 .35Error 133.80 60 2.23Note. R2 = .68.The adjusted means for the SA group was 1.45 (SD = 1.96),for the PH group was 1.38 (SD = 2.03), and for the OS group was2.66 (SD = 2.83). The adjusted means for the PH and the OStreatment conditions were contrasted using POC. The 1-tailedindependent t-test found that the difference between these meanswas statistically significant, t(42) = 2.84, p < .01. Consistentwith the POC analyses, the ANCOVA results demonstrated that thetreatment effects remained significant even after controlling forrelevant prereading skills. This is a stringent test for theeffects of the treatment conditions in the study.Hypothesis 4 Rhyming ability of prereaders will make an independentcontribution in learning to read after accounting for theeffects of age, vocabulary (PPVT-R), phoneme identity,treatment group membership, and letter-sound knowledge.124For a stringent test of this hypothesis, Bryant et al.(1990) recommended entering the relevant variables into a fixed-order multiple regression equation with rhyming ability enteredas the last variable. If rhyming ability did make an independentcontribution, it would account for a significant portion of thevariance in reading after the effects of the related variableswere removed. Treatment group membership was included to controlfor differences in training among treatment conditions.There were two possible dependent variables for examininghypothesis 4; the trials to criterion averages and the analogyword reading test scores. The analyses will be considered inthat order.The results of the fixed-order multiple regression analysisusing the criterion average scores are presented in Table 20.Table 20. Relation of Rhyming Ability to Trials to CriterionAfter Controlling for Related VariablesVariable R2 R2 Added FTreatment group .07 .07 5.09*Age .10 .03 1.79PPVT-R .16 .06 4.80**Phoneme identity .47 .31 35.78***Letter-sound .48 .01 .63Rhyming ability .56 .08 11.01******p <.001.^**p <.01.^*p <.05.125The fact that rhyming ability accounted for 8% of thevariance in the number of trials taken to learn to read thetraining sets is evidence that it does make an independentcontribution to beginning reading. This finding is consistentwith Bryant et al. (1990) where rhyming ability accounted for 7%in a similar analysis but with longitudinal data (rhyming abilitymeasured at 4 years 7 months, reading measured at 6 years 7months).Note that phoneme identity accounts for a sizable amountof variation even after the effects for treatment group, age, andPPVT-R have been removed. Treatment group membership was enteredas a variable to control for differences in the training amongtreatment groups.Analogy word reading was another dependent variable toexamine if rhyming made an independent contribution to reading.The results of are presented in Table 21.Table 21. Relation of Rhyming Ability to Analogy Word ReadingAfter Controlling for Related VariablesVariable R2 R2 Added FTreatment group .08 .08 5.48*Age .15 .07 5.31**PPVT-R .16 .01 .60Phoneme identity .60 .45 68.86***Letter-sound .63 .02 3.46**Rhyming ability .63 .00 .24***p <.001. **p <.01. *p <.05.126Following the fixed-order multiple regression analysis forthe criterion average scores, the pretest variables were enteredfirst followed by rhyming ability. Contrary to the regressionanalysis with the criterion average scores as the dependentvariable, rhyming ability did not make an independentcontribution to analogy word reading. This finding wassurprising given the evidence from previous hypotheses thatsupported the idea that the children used orthographic analogiesbased on rhyming. Note that phoneme identity accounted for anoverwhelming majority of the variance in analogy word reading.The finding that phoneme identity accounted for the majorityof the variation in analogy word reading adds to the puzzle inthis way. If phonemes were the units of sound that explainedmost of the word reading, then why were there not equal numbersof letter-sound test words read? In fact, the letter-sound testwords were read on only 35 occasions.The following section presents the results of two lines ofinquiry designed to further examine the findings that phonemeidentity accounted for a significant amount of variation inanalogy word reading and that rhyming ability did not appear tomake an independent contribution.Phoneme Identity and Analogy Word ReadingTwo questions drove the further examination of therelationship between phoneme identity and analogy word reading.First, children had been pretested for the ability to identifyphonemes in the initial, medial, and final positions of short127words presented orally. Is the ability to identify phonemes inone of these positions more predictive of analogy word readingthan the others? Second, what role did phoneme identity play inthe reading of the analogy test words? The results of theanalyses to examine these questions will be presented in order.To examine the question of which form of phoneme identitybest predicted analogy word reading, a discriminant analysis wasperformed. The ability to identify phonemes in the initial,medial, and final positions served as independent variables in astepwise discriminant analysis to estimate their ability todiscriminate between children who could read the analogy testwords and those who could not. Given the size and equality ofthe dependent variable (n = 33 in each group) and an evaluationof normality, there was no threat to multivariate analysis (seeAppendix D).The results of the discriminant function analysis arepresented in Table 22.Table 22. Predicting Analogy Word Reading with Three Forms ofPhoneme IdentityStandard. StructureVariable Coeff. Coeff. R2Final phoneme identity 35.22*** .70 .88 .35Initial phoneme identity 6.81 *** .51 .76 .06Medial phoneme identity .21Note. Wilk's Lambda = .58, F(2,63) = 22.61, p < .001.***p <.001.128The ability to identify phonemes in the final positionproved to be the best single discriminator among the forms ofphoneme identity. The ability to identify initial phonemes inthe initial position was statistically significant as well. Withthe use of a jackknifed classification procedure for the totalsample of 66 children, 77.3% were classified correctly as analogyreaders or non-analogy word readers.The second question asked concerned the role of phonemeidentity in the reading of the analogy test words. The evidenceso far suggests that children able to identify the last phonemein the analogy test words somehow benefited from thisinformation. But how so? A possible explanation is that thefinal phoneme, represented by the final letter, served as a clueto the sound of the rime ending. For example, the knowledge thata test word ended in the "t " sound would be a useful clue toestablishing the sound of the rime, even if the medial phonemewas ignored. If this was the strategy used to read the analogytest words, identifying the final phoneme may have been the firststep the children took to use make an orthographic analogy basedon the rime ending.For example, if the response for the test word 'bad' was"bed", and if "bed" rhymed with one of the training items, thiswould be evidence that children were making rhyme-based analogiesto read. However, the orthography upon which the analogy wasbased would be the final letter, which is representedphonologically by a single phoneme, not the rime. Note that thesame strategy produces a correct response to an analogy test word129but an incorrect response to a letter-sound correspondence testword.If this explanation is true, it predicts children would haveresponded incorrectly to the letter-sound test words with wordsthat ended with the final phoneme sounded correctly, and at thesame time rhymed with a training word. It also predicts thatthese words would probably be the most frequently made incorrectresponse.This is a stringent test of the explanation for two reasons.First, it places two constraints on an incorrect response beforeit is considered an example of the strategy (i.e., the finalphoneme must be sounded correctly and the incorrect response mustrhyme with a training word). Second, it allows for all otherpossible incorrect responses to fall outside of this category andbe counted as evidence against the explanation. All attemptschildren made to read test words were recorded. The errorschildren made attempting to read the letter-sound correspondencetest words are presented in Table 23.The means for incorrect responses to letter-sound test wordsthat both sounded the final phoneme correctly and rhymed with atraining word was 12.87 (SD = 3.50). The mean for all otherincorrect responses was 6.88 (SD = 2.67). A 1-tailed dependentt-test found that the difference between these means wasstatistically significant, t(7) = 3.29, p < .01. This isevidence in support of the explanation that children made arhyme-based analogy to the training words that was also based onthe sound of the final phoneme. This strategy would be130successful for reading the analogy test words but not for readingthe letter-sound test words.Table 23. Error Analysis of Letter-sound Correspondence WordReadingIncorrect Word ResponsesTestWord Rhyminaa Non-rhymingbad 17 7bet 23 11hum 9 5bag 6 9pit 14 6sit 11 7rag 9 7run 14 3Totals 103 55aThe incorrect response rhymed with a training word.Table 23 shows that on 103 occasions children responded to aletter-sound knowledge test word with a word that ended with thefinal phoneme sounded correctly, and at the same time rhymed witha training word. For example, the most common incorrect responsefor the letter-sound test word 'bad' was "bed" (8 occasions).The most common error for the test word 'run' was "ran" (10occasions). If the child identified the sound of the finalphoneme first, and then attempted to make an analogy based on the131sound of the rime in a training word, the 103 occasions where aword that rhymed with a training word was incorrectly given for aletter-sound test word would be explained. The analogy betweenthe test word and the training word would still be rhyme-based,but it would be made on the basis of sound, not text. Anorthographic analogy could only be made between the analogy testwords and the training words.SummaryHypothesis 1 questioned whether prereaders will learn toread more words on the basis of orthographic analogy than byletter-sound correspondences. Children read significantly moreanalogy reading test words than letter-sound correspondence testwords, t(65) = 5.74, p < .001.Hypotheses 2a, 2b, and 2c examined the treatment effects.Phonological training with onset and rime using whole word textdid not increase analogy word reading over the same trainingwithout it, t(20) = .08, p > .05. However, prereaders trainedwith segmented text read significantly more analogy test wordsthan did children trained with whole word text, t(20) = 2.98, p <.01. Finally, the combination of phonological training andtraining with segmented text led to more analogy word readingthan training with whole word text alone for children with aboveaverage prereading skills, t(20) = 3.06, p < .01, but not forchildren with low prereading skills, t(20) = .98, p > .05.Hypothesis 3 questioned whether the ability of prereaders tolearn to read words by orthographic analogy will vary according132to their level of prereading skills. Prereaders in the Highprereading skills group read significantly more analogy testwords than did children in the Low prereading skills group, t(64)= 6.61, p <.001. As well, children in the High group took fewertrials to learn to read the training items than did children inthe Low prereading skills group, t(64) = 6.81, p < .001).The ANCOVA that examined treatment condition effects onanalogy word reading after controlling for prereading skillsconfirmed the results from the planned orthogonal contrasts. TheANCOVA analysis was complemented by the multiple regressionanalysis in hypothesis 4 where the effects of prereading skillson analogy word reading could be examined after controlling fordifferences in treatment conditions and other pretests.Hypothesis 4 examined the claim that rhyming ability made anindependent contribution to reading. The two possible dependentvariables were the number of trials taken to learn the trainingitems and analogy test word reading. The results were equivocal.Using the trials to criterion as the dependent variable,rhyming ability contributed a statistically significant 8% ofvariance after accounting for the effects of all other pretests(i.e., age, treatment group membership, PPVT-R, phoneme identity,and letter-sound knowledge). However, rhyming ability did notmake an independent contribution to reading when analogy wordreading was the dependent variable. In fact, rhyming did notaccount for any variance after controlling for the effects of thepretests. Phoneme identity, on the other hand, accounted for 45%of the variance in analogy word reading.133Further analysis using the three forms of phoneme identity(i.e., in initial, medial, and final positions) found that theability to identify phonemes in the final position was the bestsingle discriminator between readers and non-readers.134VI. DISCUSSIONTwo explanations of the relationship between phonologicalawareness were examined. One account argues that the importantunits of sound are phonemes and that a combination of phonemeidentity and letter-sound knowledge is sufficient for reading tobegin (e.g., Byrne & Fielding-Barnsley, 1993; Ehri, 1991). Theproposed strategy for reading is recoding letter-sounds at thephoneme level.Another view argues that the important units of sound areonset and rime; units that are key to the rhyming abilitypossessed by prereaders (Bryant et al., 1990; Goswami & Bryant,1990). The proposed reading strategy is the use of orthographicanalogy based on rhyming ability. To accomplish this, childrenmust recognize that the rime endings of rhyming words typicallyare common text. This study examines the Goswami and Bryant(1990) model but has implications for the Byrne and Ehri model aswell.Convincing evidence about the relationship betweenphonological awareness and beginning reading requires informationon both the kind of phonological and orthographic units involvedand how these units figure in the strategies children use. Also,the sample should consist of prereaders to rule out the possiblereciprocal effects between phonological skills and readingability.Previous research on the use of orthographic analogy foundthat young readers can use orthographic analogy (Ehri, 1992a;Goswami, 1988; Peterson & Haines, 1992) but when the study135samples readers, it is unclear whether ability in orthographicanalogy precedes reading or follows it. Studies samplingprereaders have found that many prereaders can use orthographicanalogy when first starting to read (Goswami, 1986, 1988) butthese studies did not measure potentially relevant prereadingskills (i.e., rhyming ability, phoneme identity, letter-soundknowledge). A unique feature of this study on beginning readingwas that it, concurrently, sampled prereaders, measured relevantprereading skills, and gathered evidence on reading strategy.The design for the current study was derived from Baron(1977). Baron successfully taught kindergartners to readunfamiliar CVC words after very brief periods of training thatconsisted of rote memorizing rhyming words presented as segmentedtext. He tested strategy by using two types of test words; onetype could be read by analogy to training words and the otherrequired using letter-sound correspondences. He concluded thatchildren could use analogy in the first steps of reading.However, it was not clear what caused the effect that he found.Although Goswami and Bryant were largely in agreement withBaron's conclusion that the children used an analogy strategy toread new words, they argued that it was possible that thechildren did not read new words by making an orthographic analogybetween training and test words, but merely assembled thesegmented pieces of training text. In that case, no analogywould be required. The current study unpacks Baron (1977) byarticulating the effects of phonological training from effectsdue to segmented text.136Rhyme-Based Analogies The prereaders in the main study demonstrated repeatedlythat learning to read a new word was relatively easy if a rhyme-based analogy could be used to read the new word, but was anarduous task if it could not. Of the 66 children in the study,33 read analogy test words on 124 occasions while only 17children read letter-sound correspondence test words on 35occasions. Also, two children who each read five of the letter-sound test words were identified as outliers and may have beenreaders that were not screened from the study. If this was true,the already large difference between the observed orthographicanalogy and letter-sound word reading was a conservativeestimate.The already strong finding that the children used a rhyme-based analogy to read was upheld in the accounts they providedwhen asked to explain how they read the test words. Thisinformation was requested after each successful reading. Whilemany children did not provide explanations that described theirstrategy (e.g., "I just read it", "I don't know", "I can readnow"), several children gave clear explanations. The most commonexplanation, given by 15 children, was that the test word rhymedwith a training word.In addition, four conceivable rival explanations were testedthat could explain the results and none were supported.The study provided convincing evidence that prereaders canuse orthographic analogy in the first steps of reading (Baron,1977; Goswami and Bryant, 1990), but not for the view thatconsiderable experience recoding letter-sounds at the phonemelevel is a prerequisite (Ehri, 1992a).What, then, are the effects of phonological training inonset and rime, and of segmented text?Treatment EffectsResults showed that phonological training with onset andrime using whole word text did not increase analogy word readingover the same training without it. The current study providedevidence for the claim that brief phonological training withonset and rime did not increase analogy word reading over thesame training without it.This also suggested that the analogy word reading observedby Baron (1977) may not have been a result of the phonologicaltraining in onset and rime that may have been implicitly taughtby the use of segmented text.This finding was not consistent with that of Peterson andHaines (1992) who found significant gains in the use oforthographic analogy following phonological training with onsetand rime. A possible explanation for this disagreement is thatthe children in the Peterson and Haines (1992) study receivedconsiderably more training (seven 15-minute sessions over 1month) than the children in the current study.The study found that prereaders trained with segmented textread significantly more analogy test words than did childrentrained with whole word text. In fact, all but one child who hadhigh prereading skills and received training with segmented text137138read analogy test words. The lone exception was detected as anoutlier.What then was the role of segmented text in the use oforthographic analogy? A possible explanation is that whenchildren are trained with text that is segmented at the boundaryof onset and rime, the similarity between rime ending and commontext, the basis of orthographic analogy, is highlighted. Thisvisual segmentation of rhyming words aide the mapping of the rimeending directly onto text.The examination of training effects in the current studyprovided strong evidence in support of the view that the use oforthographic analogy by beginning readers is related to rhymingability (Goswami & Bryant, 1990). In addition, training withtext segmented at the onset and rime boundary facilitated theprocess.Prereading Skills The level of prereading skills proved to be highlypredictive of learning to read. The 33 children in the Lowprereading skills group read analogy test words on only 13occasions compared to 111 times for children in the Highprereading skills group. The magnitude of this difference wasconvincing evidence that one or more of the prereading skillsplayed an essential role in learning to read. Further, itsuggested that prereaders' ability to learn to read words usingrhyme-based analogies was bimodaly distributed.139However, the results from the analysis that compared theprereading skills groups on reading did not reveal the relativecontribution of rhyming ability, phoneme identity, and letter-sound knowledge. This is important because the two models ofphonological awareness and beginning reading that guided thestudy make different predictions about the primary sound unit;whether phoneme, or onset and rime. Further, as phonemes areassociated with letter-sound recoding, and onset and rime withthe use of orthographic analogy, identification of the primarysound unit is also indicative of reading strategy.RhymeBryant et al. (1990) claimed that rhyming ability made anindependent contribution to reading. Testing this claim meantexamining the relative contribution of all three prereadingskills (i.e., rhyming ability phoneme identity, letter-soundknowledge) while controlling for other potentially relatedvariables.The two possible dependent variables that could be used toexamine the hypothesis that rhyming ability makes an independentcontribution to reading were the number of trials taken to learnthe training items (an estimate of the rate of learning to readaccording to Yopp, 1988), and analogy test word reading. Theresults were equivocal.Using the trials to criterion as the dependent variable,rhyming ability contributed an additional 8% of variance afteraccounting for the effects of all other variables (i.e., age,140treatment group membership, PPVT-R, phoneme identity, and letter-sound knowledge). This result supported the Bryant et al. (1990)claim that rhyming ability contributed independently to reading.However, rhyming ability did not make an independentcontribution to reading when analogy word reading was thedependent variable. In fact, rhyming did not account for anyvariance after controlling for the effects of other variables.Phoneme identity, on the other hand, accounted for 45% of thevariance in analogy word reading. Bryant et al.'s (1990) viewthat rhyming ability makes an independent contribution to readingwas not supported.There are at least two explanations for the discrepancybetween the findings of the current study and the Bryant et al.,(1990) findings. One explanation may lie in the tasks used tomeasure phoneme awareness. Bryant et al. estimated phonemeawareness with tasks that required phoneme deletion or phonemetapping. The current study estimated phoneme awareness with atask that required phoneme identity. According to Yopp (1988),phonological manipulations requiring identity are easier thanthose requiring deletion. It is possible that prereaders areable to use their ability to identify phonemes when firstlearning to read but not the ability to delete or tap phonemes.If this explanation is true, it illustrates the importanceBlackman (1983), Wagner and Torgesen (1987) and others haveaccorded to identifying particular forms of phonemic awarenesswhen studying its relationship to reading.141A second reasonable explanation is based on the fact thatthe Bryant et al. (1990) evidence was longitudinal whereas thecurrent study measured rhyming ability and word readingconcurrently. It is possible that children require moreexperience with reading than was provided by the study before anindependent relationship between rhyming ability and readingcould be detected.Still, this result was surprising because the analogy thatcould be made between the analogy reading test words and thetraining words was based on rime endings. The common soundsamong rhyming words were represented by common text, the last twoletters. If phoneme identity was the primary phonological skillin beginning reading, then an equal number of letter-sound wordsshould have been read. This was not so.Further analysis using the three forms of phoneme identity(i.e., in initial, medial, and final positions) found that theability to identify phonemes in the final position was the bestsingle discriminator between readers and non-readers. Given thatfinding, what role did the ability to identify final phonemesplay in analogy word reading?A possible explanation was that the final phoneme,represented by the final letter, was used to determine the soundof the rime ending, and the medial vowel played a minor role. Ifthis was true, children did not differentiate the phonologicalrepresentation of the final letter (i.e., the final phoneme) fromthe rhyme. The children may not have used the information thatthe medial vowel is what distinguishes "ed" from "ad".142If this explanation was true, it predicted children wouldhave responded incorrectly to letter-sound test words with wordsthat ended with the final phoneme sounded correctly and thatrhymed with a training word. The error analysis revealed that on103 occasions children responded to a letter-sound test word thatmet these criteria compared to 55 all other incorrect responses.The explanation that children primarily used the final phoneme toread test words was supported. This explanation can account forboth the successful readings of analogy test words and the mostcommon incorrect responses to the letter-sound test words.The finding relating phoneme identity to the use of rhyme-based orthographic analogy by beginning readers represents anoriginal and significant contribution to the literature linkingphonological awareness and beginning reading.Educational Implications The training that so easily taught prereaders to readunfamiliar words was notable for both its simplicity and brevity.Learning to read the two pairs of rhyming training words tookseconds for many children, not minutes. That they were then ableto read unfamiliar words is remarkable.An important pedagogical implication of the current study isthat prereaders may learn to read new words easily and quicklygiven two conditions. First, the children must have sufficientlevels of relevant prereading skills. Second, the teachingshould be based on words that rhyme. Further, the process may beenhanced by using text segmented at the onset and rime boundary.143Limitations and Suggestion For Future ResearchThere are several limitations to the study. First of all,social background information was received on only 38 of the 66families in the study. As a consequence, caution must beexercised in generalizing any findings.Second, the sample was adequate in size (i.e., 11 subjectsper treatment group) to obtain satisfactory estimates of thevariables in the study, but a larger sample would provide morestable estimates and so would produce more confident results.The findings must be viewed in relation to the size of thesample.It should be noted that the study addresses the question ofreading mono-syllabic words presented singly to kindergartenchildren. The findings are not easily generalized to the fluentreading of connected text.The finding that phoneme identity, especially of the finalphoneme in CVC words, may be related to reading rhyming wordsinvites further research. A study that followed the design ofthe current study but that varied several medial consonants inreading test words while keeping final consonants intact couldexamine this question more completely.Another future research question generated by the currentstudy is the possibility of a relationship between rhymingability and identifying the final phoneme. If rhyming abilitywere to be enhanced by training (e.g., Peterson & Haines, 1992),would this instill or increase the ability to identify the finalphoneme?144Finally, about half of the prereaders took to reading byanalogy quickly and the remaining children found the task tooarduous. The implication for future research is to collectinformation that could possibly account for this difference.Experience with reading provided by parents or preschools arelikely candidates.In summary, the current study makes several importantcontributions to the literature on phonological awareness andbeginning reading. In particular, the study provided convincingevidence that prereaders can use orthographic analogy based onrhyming ability in the first steps of reading. The importance ofusing text segmented at the onset-rime boundary for trainingitems was also convincingly demonstrated. Finally, evidence forthe theoretically intriguing possibility of a link between finalphoneme identity and the use of rhyming-based orthographicanalogy was presented.145REFERENCESAlegria, J., Pignot, E., & Morais, J. (1982). Phonetic analysisof speech and memory codes in beginning readers. Memory and Cognition, 10, 451-456.Ball, E. W., & Blachman, B. A. (1991). Does phoneme awarenesstraining in kindergarten make a difference in early wordrecognition and developmental spelling? Reading ResearchQuarterly, 26, 49-66.Baron, J. (1977). Mechanisms for pronouncing printed words: Useand acquisition. In D. LaBerge & S. J. Samuels (Eds.), Basicprocesses in reading: Perception and comprehension (pp. 175-216). Hillsdale, NJ: Lawrence Erlbaum Associates.Baron, J. (1979). Orthographic and word-specific mechanisms inchildren's reading of words. Child Development, 50, 60-72.Barron, R. W. (1986). Word recognition in early reading: A reviewof the direct and indirect access hypotheses. Cognition, 24,93-119.Blackman, J. (1983). The role of psycholinguistic skills inreading acquisition: A look at early readers. ReadingResearch Quarterly, 18, 466-479.Bowey, J. A., & Francis, J. (1991). Phonological analysis as afunction of age and exposure to reading instruction. AppliedPsycholinguistics, 12, 91-121.Bradley, L. & Bryant, P.E. (1983). Categorizing sounds andlearning to read: A causal connection. Nature, 301, 419-421.146Bradley, L. & Bryant, P.E. (1985). Rhyme and reason in readingand spelling (IARLD Monographs, No. 1). Ann Arbor:University of Michigan Press.Brown, A. L. (1989). Analogical learning and transfer: Whatdevelops? In S. Vosniadou & A. Ortony (Eds.), Similarity andanalogical reasoning (pp.369-409). Cambridge: CambridgeUniversity Press.Brown, A. L., Kane, M. J., & Long, C. (1989). Analogical transferin young children: Analogies as tools for communication andexposition. Applied Cognitive Psychology, 3, 275-293.Bruce, D. J. (1964). The analysis of word sounds by youngchildren. British Journal of Educational Psychology, 34,158-170.Bryant, P. E., Bradley, L., Maclean, M., & Crossland, J. (1989).Nursery rhymes, phonological skills and reading. Journal of Child Language, 16, 407-428.Bryant, P.E., Maclean, M., Bradley, L., & Crossland, J. (1990).Rhyme and alliteration, phoneme detection, and learning toread. Developmental Psychology, 26, 429-438.Byrne, B. (1992). Studies in the acquisition procedure forreading: Rationale, hypotheses, and data. In P. B. Gough, L.C. Ehri, & R. Treiman (Eds.), Reading Acquisition (pp. 1-34). Hillsdale, NJ: Lawrence Erlbaum.Byrne, B., & Fielding-Barnsley, R. (1989). Phonemic awareness andletter knowledge in the child's acquisition of thealphabetic principle. Journal of Educational Psychology, 81,313-321.147Byrne, B., & Fielding-Barnsley, R. (1990). Acquiring thealphabetic principle: A case for teaching recognition ofphoneme identity. Journal of Educational Psychology, 82,805-812.Byrne, B., & Fielding-Barnsley, R. (1991). Evaluation of aprogram to teach phonemic awareness to young children.Journal of Educational Psychology, 83, 451-455.Byrne, B., & Fielding-Barnsley, R. (1993). Evaluation of aprogram to teach phonemic awareness to young children: A 1-year follow-up. Journal of Educational Psychology, 85, 104-111.Calfee, R. C. (1977). Assessment of independent reading skills:Basic research and practical applications. In A. S. Reber &D. L. Scarborough (Eds.), Toward a psychology of reading(pp. 289-323). Hillsdale, NJ: Lawrence Erlbaum.Clay, M. M. (1966). Emergent reading behavior. Unpublisheddoctoral dissertation. University of Auckland, Auckland, NewZealand.Content, A. (1985). L'analyse segmentale de la parole chez l'enfant. Unpublished doctoral dissertation. Universite'Libre de Bruxelles.Content, A., Morais, J., Alegria, J., & Bertelson, P. (1982).Accelerating the development of phonetic segmentation skillsin kindergartners. Cahiers de Psychologie Cognitive, 2, 259-269.Dunn, L., & Dunn, L. (1981). Peabody Picture Vocabulary test-revised. Circle Pines, MN: American Guidance Service.148Durkin, D. (1966). Children who read early. New York: TeachersCollege Press.Ehri, L. C. (1983). A critique of five studies related to letter-name knowledge and learning to read. In L. M. Gentile, M. L.Kamil, & J. S. Blanchard (Eds.), Reading research revisited. Columbus, OH: Charles Merrill.Ehri, L. C. (1991). Development of the ability to read words. InR.^Barr, M. L. Kamil, P. B. Mosenthal & P. D. Pearson(Eds.), Handbook of reading research: Volume II (pp. 383-417). New York: Longman.Ehri, L. C. (1992a). Beginners need some decoding skill to readwords by analogy. Reading Research Quarterly, 27, 12-26.Ehri, L. C. (1992b). Reconceptualizing the development of sightword reading and its relationship to recoding. In P. B.Gough, L. C. Ehri, & R. Treiman (Eds.), Reading Acquisition (pp. 107-143). Hillsdale, NJ: Lawrence Erlbaum.Elkonin, D. B. (1963). The psychology of mastering the elementsof reading. In B.. Simon & J. Simon (Eds.), Educational psychology in the U.S.S.R. London: Routledge & Kegan Paul.Elkonin, D. B. (1973). U.S.S.R. In J. Downing (Ed.), Comparativereading. New York: Macmillan.Ellis, N. C., & Large, B. (1987). The development of reading: Asyou seek so shall you find. British Journal of Developmental Psychology, 78, 1-28.Fox, B., & Routh, D. K. (1975). Analyzing spoken language intowords, syllables, and phonemes: A developmental study.Journal of Psycholinguistic Research, 4, 331-342.149Fox, B., & Routh, D. K. (1976). Phonemic analysis and synthesisas word-attack skills. Journal of Educational Psychology,68, 70-74.Gesell, A. (1940). The first five years of life. New York: Harper& Sons.Glass, G. V. & Hopkins, K. D. (1984). Statistical methods ineducation and psychology. New Jersey: Prentice-Hall.Goodman, K. S. (1989). Whole language is whole: A response toHeymsfeld. Educational Leadership, 46, 69-70.Goodman, Y. M. (1967). A psycholinguistic description of observedoral reading phenomena in selected young beginning readers.Unpublished doctoral dissertation, Wayne State University,Detroit.Goswami, U. (1986). Children's use of analogy in learning toread: A developmental study. Journal of Experimental ChildPsychology, 42, 73-83.Goswami, U. (1988). Orthographic analogies and readingdevelopment. Quarterly Journal of Experimental Psychology,40A, 239-268.Goswami, U. (1990a). Phonological priming and orthographicanalogies in reading. Journal of Experimental ChildPsychology, 49, 323-340.Goswami, U. (1990b). A special link between rhyming skills andthe use of orthographic analogies by beginning readers.Journal of Child Psychology and Psychiatry,  31, 301-311.150Goswami, U. (1991). Learning about spelling sequences: The roleof onsets and rimes in analogies in reading. ChildDevelopment, 62, 1110-1123.Goswami, U., & Bryant, P. (1990). Phonological skills andlearning to read. Hove: Lawrence Erlbaum Associates.Goswami, U., & Bryant, P. (1992). Rhyme, analogy, and children'sreading. In P. B. Gough, L. C. Ehri, & R. Treiman (Eds.),Reading Acquisition (pp. 49-64). Hillsdale, NJ: LawrenceErlbaum.Goswami, U., & Mead, F. (1992). Onset and rime awareness andanalogies in reading. Reading Research Quarterly, 27, 152-162.Hohn, W. E., & Ehri, L. C. (1983). Do alphabet letters help pre-readers acquire phonemic segmentation skill? Journal of Educational Psychology, 75, 752-762.Huey, E. B. (1962). The psychology and pedagogy of reading.Cambridge, MA: MIT Press. (Original work published 1908).Jenkins, J. R., Bausell, R. B., & Jenkins, L. M. (1972).Comparison of letter name and letter sound training astransfer variables. American Educational Research Journal,9, 75-86.Just, M. A., & Carpenter, P. A. (1987). The psychology of readingand language comprehension. Boston: Allyn & Bacon.Kirtley, C., Bryant, P., MacLean, M., & Bradley, L. (1989).Rhyme, rime and the onset of reading. Journal of Experimental Child Psychology, 48, 224-245.151Lenel, J. C., & Cantor, J. H. (1981). Rhyme recognition andphonemic perception in young children. Journal of Psycholinguistic Research, 10, 57-68.Leong, C. K., & Haines, C.F. (1978). Beginning readers' analysisof words and sentences. Journal of Reading Behavior, 10,393-407.Liberman, I. Y., Shankweiler, D., Fischer, F. W., & Carter, B.(1974). Explicit syllable and phoneme segmentation in theyoung child. Journal of Experimental Child Psychology, 18,201-212.Lomax, R. G., & McGee, L. M. (1987). Young children's conceptsabout print and reading: Toward a model of word readingacquisition. Reading Research Quarterly, 22, 237-256.Lundberg, I., & Hoien, T. (1991). Initial enabling knowledge andskills in reading acquisition: Print awareness andphonological segmentation. In D. J. Sawyer and B. J. Fox(Eds.), Phonological awareness in reading: The evolution ofcurrent perspectives (pp. 73-95). New York: Springer-Verlag.Lundberg, I., Frost, J. & Petersen, 0. P. (1988). Effects of anextensive program for stimulating phonological awareness inpreschool children. Reading Research Quarterly, 23, 263-284.Lundberg, I., Olofsson, A., & Wall, S. (1980). Reading andspelling skills in the first school years predicted fromphonemic awareness skills in kindergarten. ScandinavianJournal of Psychology, 21, 159-173.152MacLean, M., Bryant, P. E., & Bradley, L. (1987). Rhymes, nurseryrhymes and reading in early childhood. Merrill-PalmerQuarterly, 33, 255-282.Mann, V. (1986). Phonological awareness: The role of readingexperience. Cognition, 24, 65-92.Morais, J. (1991). Phonological awareness: A bridge betweenlanguage and literacy. In D. J. Sawyer and B. J. Fox (Eds.),Phonological awareness in reading: The evolution of currentperspectives (pp.31-71). New York: Springer-Verlag.Morphett, M. V., & Washburne, C. (1931). When should childrenbegin to read? Elementary School Journal, 31, 496-508.Olofsson, A., & Lundberg, I. (1983). Can phonemic awareness betrained in kindergarten? Scandinavian Journal of Psychology,24, 35-44.Olofsson, A., & Lundberg, I. (1985). Evaluation of long termseffects of phonemic awareness training in kindergarten.Illustrations of some methodological problems in evaluationresearch. Scandinavian Journal of Psychology, 26, 21-34.Ott, L. (1988). An introduction to statistical methods and data analysis. Boston: PWS Kent.Pedhazur, E. J. (1982). Multiple regression in behavioral research: Explanation and prediction. New York: Holt,Rinehart, & Winston.Perfetti, C. A., Beck, I., Bell, L. C., & Hughes, C. (1987).Phonemic knowledge and learning to read are reciprocal: Alongitudinal study of first-grade children. Merrill-Palmer Quarterly, 33, 283-319.153Peterson, M. E., & Haines, L. P. (1992). Orthgraphic analogytraining with kindergaten children: Effects on analogy use,phonemic segmentation, and letter-sound knowledge. Journal of Reading Behavior, 24, 109-127.Pick, A., Unze, M., Brownell, C. A., Drozdal, J. G., & Hopmann,M. K. (1978). Young childrens's knowledge of word structure.Child Development, 49, 669-680.Plessas, G. P., & Oaks, C. R. (1964). Pre-reading experiences ofselected early readers. Reading Teacher, 17, 241-245.Rayner, K., & Pollatsek, A. (1989). The psychology of reading.Englewood Cliffs, NJ: Prentice-Hall.Reid, J. (1966). Learning to think about reading. Educational Research, 9, 56-62.Samuels, S.J. (1972). The effect of letter-name knowledge onlearning to read. American Educational Research Journal, 9,65-74.Sawyer, D. J., & Fox, B. J. (1991). Phonological awareness inreading: The evolution of current perspectives. New York:Springer-Verlag.Shankweiler, D., & Liberman, I. Y. (1989). Phonology and readingdisability: Solving the reading puzzle. Ann Arbor:University of Michigan Press.Share, D. L., Jorm, A. R., MacLean, R., & Matthews, R. (1984).Sources of individual differences in reading acquisition.Journal of Educational Psychology, 76, 1309-1324.Shavelson, R. J. (1988). Statistical reasoning for the behavioral sciences. Boston: Allyn & Bacon.154Skjelfjord, V. J. (1987). Phonemic segmentation: An importantsubskill in learning to read. Scandinavian Journal of Educational Research, 31, 41-57.Smith, F. (1973). Psycholinguistics and reading. New York: Holst,Rinehart, and Winston.Stanovich, K. E. (1986). Matthew effects in reading: Someconsequences of individual differences in the acquisition ofliteracy. Reading Research Quarterly, 21, 360-406.Stanovich, K. E. (1988). Children's reading and the developmentof phonological awareness. Detroit: Wayne State UniversityPress.Stanovich, K. E., Cunningham, A., & Cramer, B. (1984). Assessingphonological awareness in kindergarten children: Issues oftask comparability. Journal of Experimental Child Psychology, 38, 175-190.Tabachnick, B. G., & Fidell, L. S. (1989). Using multivariatestatistics. New York: Harper & Row.Treiman, R. (1983). The structure of spoken syllables: Evidencefrom novel word games. Cognition, 15, 49-74.Treiman, R. (1985). Onsets and rimes as units of spokensyllables: Evidence from children. Journal of Experimental Child Psychology, 39, 161-181.Treiman, R. (1991). Phonological awareness and its role inlearning to read and spell. In D. J. Sawyer and B. J. Fox(Eds.), Phonological awareness in reading: The evolution of currentperspectives (pp. 159-189). New York: Springer-Verlag.155Treiman, R. (1992). The role of intrasyllabic units in learningto read and spell. In P. B. Gough, L. C. Ehri, & R. Treiman(Eds.), Reading Acquisition (pp. 65-106). Hillsdale, N. J.:Lawrence Erlbaum.Treiman, R., & Baron, J. (1981). Segmental analysis ability:Development and relation to reading ability. In G. E.MacKinnon & T. G. Waller (Eds.), Reading research: Advances in theory and practice (Vol. 3, pp. 159-198). SanDiego, CA: Academic Press.Tunmer, W. E., & Nesdale, A. R. (1985). Phonemic segmentationskill and beginning reading. Journal of Educational Psychology, 77, 417-427.Tunmer, W. E., & Rohl, M. (1991). Phonological awareness andreading acquisition. In D. J. Sawyer and B. J. Fox (Eds.),Phonological awareness in reading: The evolution of currentperspectives (pp. 1-30). New York: Springer-Verlag.Wagner, R. K., & Torgesen, J. K. (1987). The nature ofphonological processing and its causal role in theacquisition of reading skills. Psychological Bulletin.101, 192-212.Wagner, R. K., Torgesen, J. K., Laughon, P., Simmons, K., &Rashotte, C. (1993). Development of young reader'sphonological processing abilities. Journal of Educational Psychology, 85, 83-103.Wise, B. W., Olson, R. K., & Treiman, R. (1990). Subsyllabicunits as aids in beginning readers' word learning: Onset-156rime versus post-vowel segmentation. Journal ofExperimental Child Psychology, 49, 1-19.Yopp, H. K.. (1988). The validity and reliability of phonemicawareness tests. Reading Research Quarterly, 23, 159-176.Zhurova, L. E. (1963). The development of analysis of wordsinto their sounds by preschool children. Soviet Psychologyand Psychiatry, 72, 17-27.157Appendix A: Definition of Terms This appendix provides definitions of some terms used in thecurrent study. Written letters and words presented to the childare in single quotation marks (e.g., 'cat') and spoken letters orwords are in double quotation marks (e.g., "c", "cat").Alphabetic Principle - useable knowledge that phonemes can berepresented by letters, such that whenever a particularphoneme occurs in a word, and in whatever position, it canbe represented by the same letter.Continuant Consonants - consonants produced by a continued flowof air at their places of articulation (e.g., "f", "h", "s",v").Graphemes - letters or letter clusters (e.g., 'c', 'th')corresponding to phonemes.Onset - the sound of the initial consonant or consonant clusterof a syllable (e.g., "c" in "cat", "str" in "string").Onset-Rime Segmentation - articulating the onset of a syllableseparately from the rime (e.g., "cat" as "c-at").Orthographic Analogy - the ability to make inferences fromsimilarities in spelling to similarities in sound.Phones - speech sounds (e.g., the sound of the letter 't' in'ten').Phonemes - a group of phones that speakers of a language considerto be variations on the same sound, (e.g., the sounds of theletter 't' in 'ten' and 'stop') and having a distinctfunction in determining meaning.158Phoneme Blending - blending isolated sounds into words, e.g., "c-a-t" into "cat" (Yopp, 1988).Phoneme Deletion - deleting phonemes from words, e.g., deletingthe "c" sound from "cat" (Yopp, 1988).Phoneme Segmentation - articulating the phonemes of a wordseparately and in order, e.g., "cat" into "c-a-t" (Yopp,1988).Phonemic Awareness - "conscious access to the phonemic level ofthe speech stream and some ability to cognitively manipulaterepresentations at this level" (Stanovich, 1986, p. 362).Phonological Awareness - conscious access to the constituentsounds of speech and the ability to manipulate these sounds.Note that the difference between phonological and phonemicawareness is the that latter requires access to sounds atthe phoneme level.Rime - the sound of the vowel and any following consonants, e.g.,"at" in "cat" (Treiman, 1991).Stop Consonants - consonants produced by momentarily obstructingthe flow of air at their places of articulation (e.g., "b","d", "g", "k", "p", and "t").Appendix B: Pilot Study Training Illt Istrations and Words bug159mug•^a•^00.damramAppendix B: Pilot Study Training Illustrations and Wordsbatmatredted160pin161tinsatmat162mugbugranpanAppendix C: Illustrations for the Rhyme Test (Reduced 10%) 163N01■PAppendix D: Examination of Mathematical AssumptionsH^TEST^ASSUMPTIONS^EXAMINATION^TENABLE?^COMMENTS1^Depend.^1.Normalityt-testHistogram No^- t-test robust- conservative forskewness- scores transformedwith no change inresults2.Random^ YesSampling3.Pop. var.^ Yesunknown2abc POC^l.Planned YesContrasts2.Orthogonal^ YesContrasts2a^Indep.^1.Normality^Histogram^Yes^- small nt-test2.Homogen.^Levene F^YesVariances3.Independent^ YesObservations2b^Indep.^1.Normality^Histogram^Yes^- small nt-test2.Homogen.^Levene F^YesVariances3.Independent^ YesObservationsH TEST ASSUMPTIONS EXAMINATION TENABLE? COMMENTS2cHi.Indep.t-test1.Normality Histogram Yes - small n2.Homogen. Levene F YesVariances3.Independent YesObservations2cLowIndep.t-test1.Normality Histogram No - t-test robust- conservative forskewness- scores transformedwith no change inresults2.Homogen.^Levene F^No^- robust if nl = n2Variances3.Independent YesObservations3^Indep.^1.Normalityt-test2.Homogen.VariancesHistogramLevene FNo^- Low group positivelyskewed- t-test robust- conservative forskewnessNo^- robust if nl = n23.Independent^ YesObservationsH^TEST^ASSUMPTIONS^EXAMINATION^TENABLE?^COMMENTS2b ANCOVA^1.Independence Yes2.Normality^- Resid.^Yes^- robust when samplePlots sizes are equal anderror df > 203.Homogen.Variances4.Linearitya)Cov. & DV- F Max^Yes- Resid.^YesPlots- Cov. by^Yes^- except for Rhyme inY Plots SAb)Among IV's^- Cov.Scatterplts Yes5.Homogen.^-^YesSlopes Treat Int.- F BMDPV1^YesH^TEST^ASSUMPTIONS^EXAMINATION^TENABLE?^COMMENTS4 Trials to CriterionMR^1.Independence^ Yes2.Normality - Resid.Plot- Norm. P.PlotYesYes3.Homoscedas. - Resid.^YesPlot4.Linearity^- Resid.^YesPlot4 Analogy Word ReadingMR^1.Independence^ Yes2.Normality - Resid.Plot- Norm. P.PlotYesYes3.Homoscedas. - Resid.^YesPlot4.Linearity - Resid.PlotYesH^TEST^ASSUMPTIONS^EXAMINATION^TENABLE?^COMMENTSFinal^DA 1.Multiv.^- No Test^ - robust when df > 20,Phoneme Normality^Available n1 = n2 , 2-tailed2.Outliers^- Multiv.^NoneMahalan.- Univar. Two^- following TabachnickRes. z retained unless alsomultivariate3.Homogen.^- Canonical^Yes^- equality of plot forVar/Cov Plot each group, tenable- robust when samplesequal or large, > 204.Linearity^- Predictor^Yes^- violation reducesScatterplots powei, not Type Iincrease5.Multicol.^- BMDP7M^YesSingul. Protects172Appendix E: Testing For Homogeneity of Slopes The dependent variable was analogy word reading and thecovariates were letter-sound knowledge, phoneme identity, andrhyming ability. Failure of the homogeneity of slopes assumptionwould be indicated by a statistically significant interactionbetween treatment group membership and a covariate. As none ofthese interactions were significant, the ANCOVA assumption forhomogeneity of slopes is plausible and the analysis proceeded.Summary ANCOVA TableSource SS DF MS F pTreatment 6.02 2 3.01 1.50 .23Letter-sound 9.71 1 9.71 4.83 .03Phoneme id. 39.09 1 39.09 19.46 .00Rhyme .001 1 .001 .001 .98Treat*Letter 2.88 2 1.44 1.39 .26Treat*Phoneme 5.58 2 2.79 .72 .49Treat*Rhyme 6.66 2 3.33 1.66 .20Error 108.48 54 2.01Note. R1 = .71.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items