Transcript
British Journal of Educational Psychology (2013) © 2013 The British Psychological Society www.wileyonlinelibrary.com Theconnectionbetweenchildren’sknowledgeanduse of grapho-phonic and morphemic units inwritten text and their learning at school Peter Bryant*, Terezinha Nunes and Rossana Barros † Department of Education, University of Oxford, UK Background. Most psychologists who study children’s reading assume that theirhypotheses are relevant to children’s success at school. This assumption is rarely tested. Aims. The study’s aims were to see whether two successful measures of the processesunderlying children’s learning to read and write are related to their success in English,science, and mathematics as measured by school assessments. Sample. DatafromtheAvonLongitudinalStudyofParentsandChildrenwereavailableforbetween 2,500and 5,900children (indifferent analyses) ontheiruseof graphophonicand morphemic units in reading and writing and on their achievement in Key Stage 2 andKey Stage 3 assessments. Method. Hierarchicalmultipleregressionsassessedtherelationshipbetweenchildren’suseofgrapho-phonicandmorphemicunitsat8-and9-yearsandtheirperformanceintheKey Stage 2 (11-years) and Key Stage 3 (14-years) assessments in English, mathematics,and science. Results. The children’s grapho-phonic and morphemic skills predicted their achieve-ment in all three subjects at Key Stage 2, 3 years later, and at Key Stage 3, 5 years later,even after stringent controls for differences in age and IQ. The connection between thetwo types of orthographic skills and the children’s educational success was largelymediated by their reading ability as measured by standardised tests. Conclusions. Children’sknowledgeanduseofgrapho-phonicandmorphemicruleshasalastingeffectontheprogressthattheymakeatschool.Thisknowledgehasanimpactontheir reading ability which in turn affects their success in learning about English,mathematics and science. Learning about grapho-phonic and morphemic units Much of the teaching on reading and writing given to children when they begin to readand write is about grapheme – phoneme correspondences (GPCs). This concentration onGPCs is supported by research that has established, beyond any doubt, a connectionbetween young children’s awareness of phonological segments, particularly of pho-nemes, and their progress in learning to read (Badian, 1994; Bradley & Bryant, 1983;Cardoso-Martins & Pennington, 2004; de Jong & van der Leij, 1999; Ehri et al. , 2001;Muter & Snowling, 1998; Parrila, Kirby, & McQuarrie, 2004; Wimmer, Landerl, & *Correspondence should be addressed to Peter Bryant, Department of Education, University of Oxford, 15 Norham Gardens,Oxford OX2 6PY, UK (email:
[email protected]). † Note: We have corrected the author name Rossana Barros, instead of Rorrana Barros, in this version of the article. DOI:10.1111/bjep.12030 1 Schneider, 1994). Children’s progress in reading also benefits from extra help with GPCs(Bradley & Bryant, 1983; Ehri et al. , 2001; Hatcher, Hulme, & Ellis, 1994).The GPC system is economic and powerful, but in several alphabetic orthogra-phies, such as English, French, and Portuguese, it does not completely determine thespelling of all the words in the language. These orthographies also depend, broadly speaking, on two other sets of rules: (1) grapho-phonic spelling rules and (2)morphemic spelling rules.Grapho-phonic units in English were created over time to represent sounds that arebypassed in the GPC system. There are at least 20 vowel sounds in the English language(O’Connor,1982),forexample,butonlyfivevowellettersinwrittenEnglish.Thisiswhy the English orthography contains so many vowel digraphs, such as ae , ai , ea , and ee . For thesamereason,consonantdigraphs,suchass h , ch , th ,signifyconsonantphonemesthatno single alphabetic letter represents. Of course, it is possible to extend the grapheme – phoneme system to include digraphs. Two often juxtaposed letters, such as ea , form adistinctive visual pattern and thus a recognizable grapheme, but assimilation to the GPCsystemisoutofthequestionwhenitcomesto‘splitdigraphs’.Splitdigraphsaremadeby adding a final ‘e’ to the end of the written syllable, which, in English, has the effect of lengthening the vowel sound in that syllable. Consider, for example, the differencesbetween cap and cape , kit and kite , and hop and hope . The second word in each pair containsa‘long’vowelsound(tousethelayman’sperfectlyadequateterm),whilethefirstdoesnot,becausein thesecond wordthe finalconsonant isfollowedbyan e ,but there isno final e in the first word.There are many other conditional, grapho-phonic rules. The letter c , for example,signifiesthesound/k/whenitisfollowedbyanotherconsonant( climb ),andwhenitisthelast letter in a word ( picnic ), and when it immediately precedes a or o or u ( cup ), but itusuallysignifies/s/whenfollowedby e or i ( cease , cinema ),unlessthefollowing i ispartof a ian or ious sequence,inwhichcasetheletter c thatcomesjustbeforethesesequencesispronounced as/sh/( magician , delicious ). These grapho-phonic rules lie outside thesimple GPC framework.Morphemicspelling rules also go beyond GPCs. Morphemes are units of meaning andtheirimpact onthespellingofEnglish,French,andPortuguese words issogreatthatonecould reasonably propose that spelling in these three languages is determined almost asmuch by the meaning of the words as by their sounds. For example, education and mathematician end in exactly the same final syllable, which, however, is spelleddifferently in these two words. In both words this final syllable is a morpheme, but in education thismorpheme signifiesthat thewordisanabstractnoun formedfrom averb,and the ending is spelled in the same way as in other such nouns ( election, institution ). Mathematician isan‘agentive’wordwhichmeansthatittellsuswhataparticularpersondoes ( mathematics ) and the final syllable of all agentives which end in this sound (e.g., logician,electrician,magician )isalwaysspelledinthesameway(Nunes&Bryant,2006,2009). A longitudinal study by Nunes, Bryant, and Barros (2012) recently established thatchildren’s skills in using grapho-phonic and morphemic spelling units in reading and writing play an important part in their progress in learning to read. Children were givenmeasures of how good they were at using grapho-phonic units and morphemic units in various reading and spelling tasks when they were aged 9 and 10 years. Their scores inthese tasks were subjected to two principal component analyses, one for each group of tasks. Each analysis produced a latent variable, and these two latent variables providedseparate measures for the children’s use of grapho-phonic and morphemic orthographic 2 Peter Bryant et al. units. At the same time, the children were given a standardized reading comprehensionmeasure, and 4 years later they underwent a standardized reading fluency test. Thechildren’suseofmorphemicunitspredictedtheirreadingproficiencymorestronglythantheiruseofgrapho-phonicunitsdid,butbothvariablesmadesignificantandindependentcontributions to the children’s reading over the 5-year period, even after the effects of verbal IQ differences had been controlled. Grapho-phonic and morphemic skills and educational achievement Psychologists commonly assume that children’s success in reading has a significantimpact on their progress at school, but this assumption is rarely tested. The outcomemeasuresofnearlyallthelongitudinalandinterventionstudiesofphonologicalprocessesinlearning to read andwritearetestsofreading andwriting, not measuresofeducationalprogress. We know of only one study of the relationship between the processes underlyingchildren’sliteracylearningandtheirsuccessatschool.SavageandCarless (2004)foundastrongrelationbetweenEnglishchildren’sscoresinphonologicalawarenesstasksat4and5 years and their marks in school assessments in English, mathematics, and science at7 years. This is a valuable result, but, as the study was on children’s initial phonologicalskills, it did not deal with the educational significance of children’s subsequent learningabout grapho-phonic and morphemic units in reading and writing. Another point aboutthis study is that the participants’ IQ scores were not available to the researchers, whothemselves acknowledgethat some of the relationship between phonological scoresandeducational achievement might have been due to a common link between these twomeasures and IQ.Our hypothesis was that children’s use of grapho-phonic and morphemic units inreading and writing both play a part in children’s achievements at school. Thishypothesis was based on the evidence that these two orthographic skills affectchildren’s reading (Nunes et al. , 2012) and on the possibility that their reading ability might have a widespread effect on school learning. The hypothesis leads to twopredictions:(1) Measures of children’s use of grapho-phonic and morphemic units should predicthow well they do at school even after stringent controls for differences in IQ.(2) Any link between their grapho-phonic and morphemic skills and their subsequentprogress at school should be strongly mediated by their reading ability: Enteringreading ability into the equation should sharply reduce the direct links betweengrapho-phonic knowledge and their school achievements. We will describe a longitudinal study in which a large number of children in theUnited Kingdom were given measures of their use of grapho-phonic and morphemicunits, and these scores were then related to their performance in the nationwide (Key Stage) assessments of English, mathematics, and science, which they took when they were 11 and 14 years old, after controls for the effects of differences in IQ. Insubsequent analyses, scores for the children’s reading ability were entered into theequation to test whether this ability acted as a mediating link between the children’suse of grapho-phonic and morphemic units and their achievements in the English,mathematics, and science. Grapho-phonic and morphemic units in reading 3 Method Participants The data were collected as part of a large-scale epidemiological and longitudinal study,calledtheAvonLongitudinalStudyofParentsandChildren(ALSPAC).Theparticipantsin ALSPAC were a large proportion of the children born in Avon in the West of England in1991 – 1992. For details of how the children in the ALSPAC sample were recruited, seeGolding, Pembrey, Jones, and the ALSPAC team (2001) and Nunes et al. , (2012).The ALSPAC database contains full-scale WISC test (Wechsler Intelligence Test for Children; Wechsler, 1992) results for 7,354 children. The test was given to the children when their mean age was 8 years 6 months.The reading and spelling predictor measures of children’s use of grapho-phonic andmorphemic units were designed by Nunes and Bryant (2010; see also Nunes, Bryant, & Olsson, 2003; Nunes et al. , 2012) and were administered in a single session to a total of 7,648 children; their mean age on this testing occasion was 9 years 10 months. Theactual numbers of the children involved in the analyses which did not involvestandardized reading test scores varied from 5,967 to 5,940 with the Key Stage 2assessments at 11 years and from 4,055 to 4,021 in the Key Stage 3 assessments at14 years. The decline in numbers between 11 and 14 years was due partly to childrenmoving from the area between the two ages but mainly to the fact that many of the younger children in the srcinal sample took the Key Stage 3 assessments a year later than the rest of the sample, and for administrative reasons, their results were notavailable to the ALSPACresearch team. Theseyounger children were atthe sameschoolsas the rest of the sample and therefore shared the same background and the same kind of school experiences as the children whose Key Stage 3 results were included in the ALSPAC data set. A substantial subsample of this sample (6,921 children) was given a standardizedcomprehension test, the Neale Analysis of Reading Ability (NARA-II; Neale, 1997) at themeanageof9 years10 months.TheTestofWordReadingEfficiency(TOWRE;Torgesen, Wagner, & Rashotte, 1999) was given when the participants’ mean age was 13 years10 months to 5,521 participants.The level of the mothers’ education, the ethnic background, and the gender of thechildren in the total sample that was srcinally recruited for ALSPAC are reflected in thesubsample analysed here. Measures Predictors Theuseofgrapho-phonicandmorphemicunits. Themain predictors were two latent variables:Thefirstdefinedastheuseofgrapho-phonicunitsindecodingandspellingandthesecondastheuseofmorphemicunitsinreadingandspelling.Thetwolatentvariables were derived from the children’s scores in four tasks. These were as follows:(1) Ameasureoftheuseofgrapho-phonicunitsinreadingusingfiverealwordsandfivepseudo-words which required the use of phonological units that are larger thansingle letter-sound correspondences: words: s ite , fated , cuter , huge , taped ; andpseudo-words: s mape , sofe , duter , dape , loker . The Cronbach’s a reliability scorefor this measure was .756. 4 Peter Bryant et al. (2) Ameasureofmorphemicreadingskillsusingfiverealwordsandfivepseudo-words,inwhichthepronunciationdependsontheuseofmorphemicunits:Words: Union , unusual , native , misheard , dishonest ; pseudo-words: Uningest , disheat , uni- shaped , mishammer , amazive . In all these words and pseudo-words, the reader must decide how to parse the string and identify a suffix or a prefix to attain thecorrect pronunciation. For example, the opening vowel sounds in ‘unusual’ and‘unishaped’ are pronounced quite differently, because the prefix is ‘un’ in the first wordand‘uni’inthesecond.TheCronbach’s a reliabilityscoreforthismeasurewas.734.(3) A measure of grapho-phonic spelling skills using six real words ( smoke , white , baseball , slide , graceful , pavement ),inwhichthemainsourceofdifficultywastheuse of a grapho-phonic unit larger than a single letter. The Cronbach’s a reliability score for this measure was .744.(4) A measure of morphemic spelling skills using nine real words ( called , election , kissed , madness , emotion , magician , brought , richness , electrician ),inwhichthemain spelling difficulty was the suffix. The Cronbach’s a reliability score for thismeasure was .797.The Cronbach’s a reliability score was .818 for the 16 words used in the twographo-phonic tasks (Tasks1 and 3)combined and .842 forthe 19 words used in the twomorphemic tasks (Tasks 2 and 4) combined. The six a scores presented here show thatthe grapho-phonic and morphemic tasks did have a respectable degree of internalreliability.The 20 words in the two reading tasks were combined in a random order in one list.The children were shown each word in turn and asked to read it aloud. Their answers were scored as correct or incorrect. Refusals were marked as incorrect. To explain thepseudo-words to the children, the tester explained at the start that they would be shownsomemade-upwords,whichwerenotwordsthattheycouldrecognize.Theyshouldtrytoreadallthesemade-upwordsinthewaythattheythoughttheyshouldberead,evenifthey were guessing. The testers were thoroughly trained in the pronunciation of thepseudo-words that should be scored as correct, and they had models of real words to which they could compare the pseudo-word pronunciation. For example, duter shouldrhymewith cuter ,not cutter ,and mishammer shouldbereadbyanalogyto misheard ,intwo units, mis + hammer .The 16 words in the two spelling tasks were also combined in one mixed list and theorderofthewordswasrandom.Thetesterreadeachwordoutloudthreetimes:Firstonitsown,thenwithinaspecificsentenceincorporatingtheword,andfinallythewordbyitself again.Twoprincipalcomponentanalyseswerecarriedouttoobtainthemeasuresoftheuseof grapho-phonic and morphemic units in reading and spelling. Separate analyses arerequired when the constructs being measured differ, even if they are expected to behighly correlated. Weight and height, for example, are highly correlated (Kemsley,Billewicz, & Thomson (1962) report a value of .72 for women and .79 for men), but thishigh correlation does not mean that weight and height are the same construct.One of the two analyses included the three scores for the grapho-phonic measures which were reading real word, reading pseudo-words and spelling real words. The other included the equivalent scores for the morphemic measures.The principal components analysis of the grapho-phonic measures showed that thethree scores loaded on a single factor, which explained 74.287% of the variance. Each Grapho-phonic and morphemic units in reading 5
