You are on page 1of 36

SSR14

11/27/04

10:54 AM

Page 248

14 Children?s Reading Comprehension Difcultie


Kate Nation
Comprehension is the ultimate goal of reading. Everyone agrees that reading compre- hension is not a simple matter of recognizing individual words, or even of understand- ing each individual word as our eyes pass over it. All models of comprehension recognize the need for readers to build up a mental representation of text, a process that requires integration across a range of sources of information, from lexical features through to nowledge concerning events in the world !e.g., "arnham, #$$%& "ernsbacher, %''$& (intsch, %'')*. "iven the complex nature of reading comprehension, it is not surpris- ing that some individuals have dif+culties in this aea. ,ndivi dual differences in text comprehension have been observed in both developmental !e.g., -ation . /nowling, %''0. 1a hill, %''2* and college-aged populations !e.g., "ernsbacher . 3aust, %''%& 4ong, /eely, . 1ppy, %'''*. 5if+culty with eading comprehensio n has also been reported in a range of clinical disorders such as early onset hydrocephalus !5ennis . 6arnes, %''7*, autism !/nowling . 3rith, %')8*, nonverbal learning disorder !9elletier, Ahmad, . :our e, #$$%*, speci+c language impairment !ishop . A dams, %''$*, ;urner<s syndrome !;emple . Carney, %''8* and =illiams syndrome !4aing, >ulme, "rant, . (armiloff-/mith, #$$%*. ;hus, there is no shortage of evidence pointing to the fact that some individuals experience reading comprehension dif+culties ;he nature and origins of reading comprehension dif+culties, h wever, are not so clear. ;he aim of this chapter is to review what is nown about reading comprehension dif< culties in children, with a view to addressing two ma?or issues. 3irst, although individu- als who experience dif+culty with eading comprehension can be identi+ed, does it ma sense to tal about speci+c eading comprehension dif+culties< /econd, what are the causes of reading comprehension failure< ;he focus of the chapter will be on children who appear to show selective

impairments of reading comprehension. ;hat is, their reading accuracy is within the normal range for their age, but their comprehension of what is read is substantially below average. /tudies of such children allow us to identify cognitive systems that may be particularly crucial for the development of reading

SSR14

11/27/04

10:54 AM

Page 249

Children?s Reading Comprehension Difcultie #2' comprehension, and that are relatively independent of the processes underlying the devel- opment of word recognition s ills in reading.

peci? Decits in eading Comprehension?


Are there individuals who show speci+c eading comprehension de +cits<;he answer to this seemingly simple question is not straightforward. ;he starting place is to separate reading into two component parts, one concerned with recognizing printed words, and one concerned with understanding the message that the print conveys. Although the cor- relation between word recognition and reading comprehension is substantial !e.g., @uel, "rif+th, . ough !%')8* report correlations of .02 and .8' for +rst- and second-grad children*, it is not perfect and some individuals perform adequately on one component but poorly on the other. 1a hill and colleagues !1a hill, %''2& Auill . 1a hill, %''%* were the +rst to describe childen who obtained normal-forage text reading accuracy, but showed impaired reading comprehension. /tothard and >ulme !%''#, %''B* and -ation and /nowling !%''0* investigated populations of children selected in broadly similar ways. At a simple level of description level, these children !who will be referred to in this chapter as < poor comprehenders< read accurately but have speci+c dif ulty understanding what they read. ;ypically, poor comprehenders are rare in clinically referred samples of chil- dren with reading dif+culties !e.g., 4each, /carboough, . :esc orla, #$$7& /han weiler, 4undquist, (atz et al., %'''*. >owever, this is probably a reC ction of referral bias. ,ndeed, when populations of 0<$-year-old children have been sc reened in the D(, approximately %$E could be classi+ed as poor compehenders !-at ion . /nowling, %''0. /tothard . >ulme, %''#& Auill . 1a hill, %''%*. >ow might the <poor comprehender<pro+le be conceptualied< Acc ording to >oover and "ough<s !%''$* <simple view< of reading, reading comprehension comprises two sets of s ills, those concerned with decoding or recognizing printed words, and those involved in linguistic comprehension. ;he relationship between decoding and linguistic com- prehension is considered to be multiplicativeF there can be no reading comprehension without the ability to decipher or recognize words, and similarly, reading comprehension will fail if children lac the linguistic comprehension to understand what it is they have decoded. 9ut simply, both decoding and linguistic comprehension are necessary, and neither s ill on its own is suf+cient, if successful eading co mprehension is to follow. ;he essence of the simple model is captured beautifully by "ough, >oover, and 9eterson<s !%''8* account of the elderly @ohn Gilton, who due to failing sight was unable to reread the "ree and 4atin classics. >is solution

was to teach his daughters how to decode "ree and 4atin. >aving accomplished the basics of 4atin and "ree letter-sound correspon- dences, they were able to read the texts aloud while their father listened. ;he product was, for Gilton at least, successful reading comprehension. ;hus, according to the simple view, reading comprehension is the product of decoding and linguistic comprehension. ,t follows from this that children with poor reading comprehension must have de+cits either in decoding, linguisticcomprehension, or both. ;he logic of this view argues that reading comprehension de+ci s cannot be speci+c, bu

SSR14

11/27/04

10:54 AM

Page 250

#B$ Kate Nation instead must be related to wea nesses in one or both of its component parts. 3or the children described above as having speci+c eading comprehensio n impairments, which component of reading comprehension is at fault< Decoding difculties as a souce of poor reading comprehension According to the simple model, decoding s ill can place a constraint on reading com- prehension. A speci+c form of this hypothesis was poposed by 9erfetti !%')B* who claimed that when decoding is slow and effortful, resources are dedicated to word-level processing. 6y contrast, when decoding is automatic, resources are available for the tas of comprehension. ,n line with 9erfetti<s <verbal ef+cienc<hy pothesis, evidence demonstrates that reading comprehension is compromised when decoding is poor. =ord reading speed and reading comprehension correlate in child as well as adult populations !>ess . :adt e, %')%& @ac son . GcClelland, %'0'*, and 9erfetti and >ogaboam !%'0B* found that children with poor reading comprehension were slower at reading words and non- words than their classmates. Goreover, the relationship between decoding ef+ciency an reading comprehension is maintained over time, and measurements of nonword reading ta en in early childhood predict later variations in reading comprehension measured in secondary school years and adulthood !6ruc , %''$& 9erfetti, %')B*. As pointed out by 1a hill and colleagues, however, inef+cient ecoding is unli ely to be the only source of reading comprehension impairment. As noted above, some children have poor reading comprehension but show age-appropriate levels of text reading accu- racy, leading to the conclusion that inadequate decoding cannot be the source of poor comprehenders<dif+culties. owever, the demonstration of adequ ate text reading accuracy does not necessarily imply ef+cient wod-level processing !9erfetti %''2& 9erfetti, Garron, . 3oltz, %''8*. Even when reading accuracy is adequate, if it is slow or inef< cient, comprehension may be compromised. ;hus, 9erfetti argued it is necessary to show that poor comprehenders decode not ?ust as accurately as control children, but that they do so with equivalent ef+cienc, if their comprehension problem s are to be considered at all exceptional. /uch evidence was forthcoming from a study by -ation and /nowling !%'')a* who found that poor comprehenders read nonwords as quic ly as control children. ;his exper- imental +nding is con+rmedy observations that poor comprehend ers perform at ageappropriate levels on standardized tests of nonword reading accuracy

such as the Graded Nonword Reading Test !/nowling, /tothard, . Gc4ean, %''8* and nonword reading ef+ciency such as the Test for Word Reading Efcienc !;orgesen, =agner, . :ashotte, %'''. e.g., Garshall . -ation, #$$7& -ation, Garshall, . Altmann, #$$7*. ,mportantly, -ation and colleagues have used the strategy of matching poor comprehenders to control children on nonword reading, thereby eliminating the possibility that group differences in reading comprehension can be accounted for by differences in decoding s ill. ,t should be noted, however, that there are differences between poor comprehenders and typically developing children in some aspects of word reading. =e will return to this point later. >owever, if we ta e the central tenet of the theory to be that inaccurate or slow decod- ing leads to poor reading comprehension, then the children described by 1a hill and by

SSR14

11/27/04

10:54 AM

Page 251

:eadi ng 2 comprehension 7

Children?s Reading Comprehension Difcultie #B%

# % -onword reading $ << < $ % # 7 2 <

<

<

Figure 14.1 /catterplot showing the relationship between reading comprehension and nonword reading in 2%% 0<$-year-old children !z-scores*.

-ation and their colleagues !-ation . /nowling, %''0& 1a hill, %''2* are exceptions to the general pattern of association between these two factors. ;o illustrate this, +gue %2.% shows the relationship between nonword reading and reading comprehension in a sample of 2%% 0<$-year-old children& the two variables are plotted as z-scores, calculated across the whole sample of children. Children falling in the lower right quadrant show the poor comprehender pro+le of good nonwod reading s ills but poor rea ding comprehension. inguistic comprehension as a source of poor reading comprehension According to the logic of the simple model of reading, if poor comprehenders do not have de+cits in decoding, they should shw de+cits in linguist c comprehension. "enerally, the relationship between reading comprehension and listening comprehension is very close, especially as children get older and reading comprehension becomes more constrained by nowledge and understanding, rather than basic word-level decoding !/tanovich, Cunningham, . 3reeman, %')2*. ,n adults, listening and reading comprehension are strongly correlated ! r<s in the region of .'& 6ell . 9erfetti, %''2& "ernsbacher, Harner, . 3aust, %''$*. Although there are important differences between spo en lan- guage and written language !e.g., in the temporal characteristics of the two modalities*, evidence

suggests that listening and reading comprehension depend on very similar under- lying processes. As :ayner, 3oorman, 9erfetti, 9esets y, and /eidenberg !#$$%, p. 2#* put it, It can be reasonably argued that learning to read enables a person to comprehend written language to the same level that he or she comprehends spo en language.<

SSR14

11/27/04

10:54 AM

Page 252

#B# Kate Nation As would be predicted by the strong relationship between written and spo en language comprehension, children selected on the basis of their poor reading comprehension usually show poor listening comprehension. -ation and /nowling !%''0* as ed children to listen to stories, and at the end of each passage of text the children were as ed a series of questions. /ome questions tapped literal understanding of what they had heard, whereas others required inferences to be made. 9oor comprehenders performed less well than control children on this listening comprehension tas . Consistent with these +nd ings, -ation, Clar e, Garshall, and 5urand !#$$2* found that poor comprehenders also performed less well than control children !matched for age, nonverbal ability, and decod- ing ability* on a number of spo en language tas s, including the Comprehension subtest ta en from the Wechsler !ntelligence "cale for Children !=,/C-,,,u * !=echsler, %''#*. ;his test requires children to formulate a response to a variety of hypothetical situations presented orally !e.g., <what should you do if you cut your +n er<*. ;he poor compre- henders obtained scores well below those of the control children, and as a group their performance fell more than one standard deviation below age-expected levels on this standardized test. ,n summary, poor comprehenders do not have a comprehension impairment that is speci+c to eading. :ather, their dif+culties with eading comp rehension need to be seen in the context of dif+culties with language compehension more generally. /ome theo- rists have gone further and intimated that since poor comprehenders<performance is highly consistent across both written and spo en language, they should perhaps not qualify as having a reading impairment, so much as a more general language or cognitive de+cit. owever, the fact that poor comprehenders<dif+culties can be traced to more general dif+culties with spo en language does not negate the f ct that they have a reading dif+cult. 1ne can draw an analogy with developmental dyslexia. ;here is little doubt that dyslexic children have a reading problem. ,t is also the case however, that dyslexic children perform poorly on oral language tas s that involve phonological processing, such as phonological awareness, nonword repetition, rapid naming, name retrieval, and verbal short-term memory !e.g., /nowling, #$$$*. /ome of these dif+cuties may be causally lin ed to their reading dif+culties, others may be consequence , but the important point is that these dif+culties do not draw attention away fom the f act that children with dyslexia have <speci+< dif+culties with eading.

What Causes Poor Reading Comprehension?


As 9erfetti !%''2, p. ))B* ma es clear, <there is room for lots of

things to go wrong when comprehension fails.<Although it is the case that reading comp rehension de+cits ae often associated with word-level decoding dif+culties !e.g., erfetti , %')B*, discussion in this chapter continues to focus on children who have <speci+< readi ng comprehension dif< cultiesF speci+c in the sense that they ae able to read text, words, and nonwords at age- appropriate levels, but their reading comprehension is impaired. >owever, even restricting discussion in this way leaves a number of possible reasons for these children<s dif+cultie to be considered.

SSR14

11/27/04

10:54 AM

Page 253

Children?s Reading Comprehension Difcultie #B7 6efore reviewing these possible causes of reading comprehension failure, it is worth reCecting on some methodological issues surounding the study o f poor comprehenders. 1ne issue concerns the choice of tas s used to reveal the poor comprehender pro+le 1a hill and colleagues screen and select poor comprehenders from regular mainstream classrooms based on performance on the Neale #nalysis of Reading #$ility !-A:A-,,* !-eale, %''0*. ,n this reading test, children read aloud short passages of text !generating a score for reading accuracy* and are then as ed questions to assess their literal and infer- ential understanding of the text !generating a score for reading comprehension*. 9oor comprehenders are selected as children who show a signi+cant dscrepancy between their ageappropriate reading accuracy and their below-average reading comprehension. ;here are however, possible ob?ections to this approach, not least that in this particular reading test !the -A:A*, reading accuracy and reading comprehension are not measured independently from one another. =ith this limitation in mind, -ation and colleagues have selected poor comprehenders according to performance on tas s that assess the two components of reading !accuracy and comprehension* separately. ,n these studies, poor comprehenders are selected and de+ned as those childen who ach ieve poor reading comprehension scores on the -A:A, but achieve age-appropriate scores on a standard- ized test of <pure< decoding !nonword reading*. A second methodological issue concerns the nature of the comparison group of control children. ;o ensure that any differences between poor comprehenders and control children are not a consequence of group differences in basic decoding s ill, -ation and /nowling !%'')a* advocated matching the two groups for nonword reading ability. 3ollowing the same logic, -ation and colleagues also match poor comprehenders and control children for nonverbal cognitive ability. ;his approach is not followed by other research groups !e.g., Auill . 1a hill, %''%*. >owever, as a minority of children selected as poor comprehenders show rather low cognitive ability !-ation, Clar e, . /nowling, #$$#*, failing to control for cognitive ability could result in spurious conclusions. A +nal methodological note concerns the compehension-age match design. 3ollow- ing the logic of the reading-age match design !e.g., 6ryant . "oswami, %')8*, /tothard and >ulme !%''#* and Cain, 1a hill, and 6ryant !#$$$a* reasoned that in order to identify candidate causes of poor reading comprehension, poor comprehenders should be compared with younger, normally developing children whose comprehension s ills are at a similar level. ,f poor comprehenders show impairments in a particular cognitive or linguistic s ill relative to younger control children matched for comprehension age, that

s ill is unli ely to be a simple consequence of comprehension level. =ith these methodological issues in mind, we return to the question of what causes poor reading comprehension in children selected as poor comprehenders. 9erfetti and col- leagues !9erfetti, %')B, %''2& 9erfetti et al., %''8* have argued that poor comprehension may be a consequence of inadequate processing, lac of nowledge, or some combina- tion of both processing and nowledge-based wea nesses. ;wo sets of processes are con- sidered essential to the comprehension process, and are described as Iinevitabl<sources of comprehension dif+culty !erfetti et al., %''8, p. %2$*& the se are lexical processes and wor ing memory resources, which together form the central elements of the verbal ef< ciency hypothesis. =e begin by reviewing evidence concerning the performance of poor comprehenders on tas s tapping these s ills.

SSR14

11/27/04

10:54 AM

Page 254

#B2 Kate Nation e%ical processes =hat is meant by lexical processes in this context< =hile some authors use the term to refer to the ef+ciency of sublexical pocessing, that is, the a bility to ma e mappings between orthography and phonology, it is also used more broadly to capture, amongst other s ills, phonological processing and lexical access !e.g., 9erfetti, %''2*. :esearch on poor comprehenders has revealed a systematic pro+le of stength s and wea nesses across different aspects of lexical processing. ,t is thus important to consider different aspects of lexical processing separately. &honological s'ills( ,t is well established that children<s phonological s ills are intimately related to the development of literacy !e.g., "oswami . 6ryant, %''$* and a consider- able body of evidence points to core phonological de+cits char cterizing individuals with poor reading !e.g., /nowling, #$$$& /tanovich . /iegal, %''2*. /han weiler !%')'* pro- posed that reading comprehension dif+culties may be caused y a <phonological bottle- nec .<1n this view, comprehension problems are a consequence o f a child being unable to set up or sustain a phonological representation of verbal information when reading. Consistent with this, phonological s ills do account for signi<cant variance in reading comprehension performance !e.g., "ottardo, /tanovich, . /iegal, %''8*. >owever, as noted by Cain, 1a hill, and 6ryant !#$$$b*, the relationship between phonology and reading comprehension may not be direct. ,nstead, the relationship between phonological s ills and reading comprehension may be mediated by word recognition. ,n line with this view, a number of studies have demonstrated that phonological s ills are not impaired in children with speci+c compehension dif+cultiesF acoss a ra nge of different phono- logical processing tas s, including phoneme deletion, rhyme oddity, ?udgment and Cuenc, spoonerisms, and nonword repetition, poor comprehenders are indistinguishable from control children !e.g., Cain, et al., #$$$b& -ation et al., #$$2& -ation . /nowling, %'')a& /tothard . >ulme, %''B*. Hery clearly, a bottlenec in phonological processing cannot account for poor comprehenders<comprehension impairments. "emantic s'ills( 5espite adequate phonological s ills, poor comprehenders do show wea nesses in some aspects of oral language. ,n a series of studies, -ation and colleagues com- pared poor comprehenders with s illed comprehenders matched for chronological age, decoding level, and nonverbal ability. 9oor comprehenders were slower and less accurate at ma ing semantic ?udgments, and they produced fewer exemplars in a semantic Cuenc tas !-ation . /nowling, %'')a*& under some conditions, differences in

semantic priming !-ation . /nolwing, %'''* and relative wea nesses in picture naming !-ation, Garshall, . /nowling, #$$%* have also been observed. ,t is important to note, however, that the de+cits obseved in these experiments were not ?ust symptoms of generally poor language& for instance, de+cits in semantic ?udgment and semantic Cuenc were accompanied by normal levels of performance on parallel tas s tapping rhyme ?udgment and rhyme Cuenc. =hat seems to unite those aspects of lexical processing that poor comprehenders +n dif+cult is meaning.;o ?udge whether two words mean the same, or to produce exem- plars to a category label, clearly depends on an appreciation of word meaning !whereas,

SSR14

11/27/04

10:54 AM

Page 255

Children?s Reading Comprehension Difcultie #BB in contrast, commonly used measures of children<s phonological s ills, such as rhyme ?udgment, phoneme deletion, and nonword repetition are tas s that can be performed without access to semantics*. /uch semantic impairments are consistent with mildto- moderate de+cits in eceptive and expressive vocabulary that ha ve emerged in some, but not all, studies !e.g., -ation et al., #$$2& /tothard . >ulme, %''#*. ;hus, in line with 9erfetti<s verbal ef+ciency hypothesis, poor compehenders do have impairments in lexical processing, but only when semantic aspects of lexical processing are taxed. ,t is important to note that although -ation and /nowling characterized poor com- prehenders as having poor lexicalsemantic s ills, subsequent research has revealed oral language wea nesses that are not necessarily restricted to the semantic or lexical domain. 3or example, -ation et al. !#$$2* found that poor comprehenders scored lower than control children on tests tapping morphosyntax and the understanding of nonliteral aspects of language, as well as vocabulary. ;hese +ndings ae c onsistent with earlier wor by /tothard and >ulme !%''#* demonstrating group de+cits on a est of syntactic comprehension, the Test for the Reception of Grammar !;:1"* !6ishop, %')7*. ,nterestingly, not all studies +nd;:1"-de+cits in childen with poor text-lev el reading comprehen- sion !e.g., Auill . 1a hill, %''%*& however, inconsistent +ndi gs across studies are dif< cult to interpret as, typically, performance levels on the ;:1" have been close to ceiling. A new edition of the ;:1" !;:1"-#& 6ishop, #$$7* contains more items, and is stan- dardized through to adulthood. A recent study using this more sensitive test !Cragg . -ation, in press* provides clear evidence pointing to syntactic comprehension impair- ments in poor comprehenders !standard scores were )$ and '2 for the poor comprehen- ders and control children respectively*. ,n summary, there is considerable evidence supporting the view that poor compre- henders have oral language wea nesses. -ation et al. !#$$2* concluded that low-language characterized poor comprehenders as a group, and furthermore, a substantial minority of the sample met criteria for speci+c language impairment !/4,& ee 6ishop, %''0, for a review*. ,mportantly, however, and unli e the ma?ority of children with /4,, poor comprehenders showed no dif+culty with phonological pocessing. ,n stead, their oral lan- guage s ills were characterized by relative wea nesses in dealing with the nonphonological aspects of language, ranging from lexical-level wea nesses !vocabulary* through to dif< culties with interpreting nonliteral language. )isual word recognition( /o far discussion has focused on aspects of lexical processing captured by children<s oral language s ills. According to 9erfetti !%')B, %''2*, however, the ability to ma e mappings between

orthography and phonology is a lexical processing s ill that is vital to the reading comprehension process. 1n this view, the ability to decode and identify words accurately and ef+ciently allws resources t o be devoted to compre- hension processes. As discussed earlier, decoding ef+ciency is clearly related to reading comprehension in general terms. 6ut is there any evidence to suggest that poor compre- henders<poor comprehension is a consequence of ineffective, resource-demanding decod- ing or word identi+cation pocesses< ;he answer to this questio n seems to be noF as reviewed above, comprehension impairments remain even when care is ta en to match poor comprehenders and controls for basic decoding s ill !as measured by nonword reading accuracy and ef+ciency*. And, when goups are matched i n this way, poor

SSR14

11/27/04

10:54 AM

Page 256

#B8 Kate Nation comprehenders show normal phonological processing s ills, suggesting that their decod- ing is not underpinned by lowquality phonological nowledge. ,nterestingly, however, even when poor comprehenders are closely matched to control children for decoding ability, subtle group differences in visual word recognition have been observed. -ation and /nowling !%'')a* found that poor comprehenders were less accurate and ef+cient than contol children at reading irregula r words and low-frequency words& there were no group differences when reading regular words and high-frequency words. 5rawing on 9laut, GcClelland, /eidenberg, and 9atterson<s !%''8* connectionist model of word recognition, -ation and /nowling proposed that word recognition is compromised in poor comprehenders due to wea nesses in vocabulary and semantic nowl- edge, as reviewed in the previous section. ;o understand how these wea nesses in oral language may impact on the development of visual word recognition, we need to consider the role played by semantics in the word recognition process. According to 9laut et al.<s !%''8* connectionist model of word recognition, reading development is best characterized by a division of labour between a phonological pathway !consisting of connections between phonological and orthographic representations* and a semantic pathway !connections between semantic representations, phonology and orthography*. Although in the earliest stages of reading development, resources are devoted to establishing connections between orthography and phonology !a in to basic decoding or <sounding-out<, the semantic pathwa y becomes increasingly important later in development, especially for the ef+cient ea ding of exception or irreg- ular wordsF words that are not handled so well by the phonological pathway alone. =ith this framewor as a bac drop, it is possible to hypothesize how children<s spo en language ability inCuences the way in which their eading syste ms are established. 3or example, dyslexic children with impaired phonological s ill are thought to come to the tas of learning to read with poorly speci+ed phonological nw ledge in the spo en domain. As a result, they +nd it dif+cult to forge adequate c nections between orthog- raphy and phonology and consequently +nd decoding !especially onword reading* dif- +cult !arm . /eidenberg, %'''& /nowling, #$$$*. 9oor comprehen ders have no such dif+cultyF their stong phonological s ills allow them to devel op an ef+cient and ell- speci+ed phonological pathwa. ,n contrast however, relative we a nesses in vocabulary and semantic nowledge may constrain the development of the semantic pathway. A wea semantic pathway in 9laut et al.<s simulations lead to problems with irregular and low- frequency words <exactly the pro+le of wod recognition that h as been observed in poor comprehenders !-ation . /nowling, %'')a*. ,t is important to note, however, that group differences were very subtle. ;he poor comprehenders were reading words and

nonwords at age-appropriate levels as measured by standardized tests, and their phonological s ills were well developed. ,t seems unli ely that such children are devoting excessive resources to word identi+cation and decoding, or that their eading compr ehension is severely com- promised by inef+cient wodidenti+cation pocesses. ,n summary, a number of conclusions concerning the status of lexical processing in children selected as having <speci+< reading comprehension imp airments can be drawn. 3irst, there is very little evidence to suggest that they have dif+culty with phonologica processing, or that their comprehension impairment is a consequence of either a phono- logical processing or a basic decoding bottlenec . Although central to 9erfetti<s verbal ef<

SSR14

11/27/04

10:54 AM

Page 257

Children?s Reading Comprehension Difcultie #B0 ciency hypothesis, it is clear that these s ills are not compromised in children selected as having a speci+c eading comprehension problem. >owever, other aspects of lexical processing are wea in poor comprehenders. 5e+cits in semantic po cessing are apparent, and these may be related to more general wea nesses with linguistic comprehension. Wor'ing memory 4anguage comprehension places heavy demands on wor ing memory resources. =hether reading or listening, representations of words and sentences must be held in memory while other aspects of the text or discourse are processed and bac ground nowledge is activated and integrated !see e.g. (intsch . :awson, this volume*. /upport for the rela- tionship between comprehension and wor ing memory comes from a number of sources, including observations that college students selected on the basis of low wor ing memory span achieve lower comprehension scores than their Ihigh spa< peers, and perform less well on various components of comprehension such as pronoun resolution !5aneman . Carpenter, %')$, %')7*. Gost relevant for this chapter are investigations of wor ing memory in poor comprehenders. ;hree studies have addressed this issue directly. Auill, 1a hill, and 9ar in !%')'* as ed children to read aloud triplets of numbers and then to recall the +nal digit in each triple. oor comprehenders perfor med less well than control children, leading Auill et al. !%')'* to suggest that de+cits n nonlinguistic wor ing memory may underlie the reading comprehension problems seen in this group of chil- dren. >owever, as the counting span tas required children to read and recall digits, the data are more suggestive of a verbal memory de+cit than a nonlnguistic one. 1n a test of nonlinguistic spatial memory span, -ation, Adams, 6owyerCrane, and /nowling !%'''* found no differences between poor comprehenders and a control group. ;o investigate verbal wor ing memory in poor comprehenders further, /tothard and >ulme !%''#* adapted 5aneman and Carpenter<s !%')$* listening span tas . ;hey rea- soned that this tas would tap verbal wor ing memory, as its tas requirements !respond- ing to short sentences and then recalling the last word of each sentence* are similar to some of the simultaneous processing and storage demands of language comprehension itself. :ather surprisingly, however, they found no group differencesF poor comprehen- ders, age-matched controls and younger children with approximately the same level of comprehension as the poor comprehenders all performed at a similar level. 3rom these data, /tothard and >ulme concluded that wor ing memory de+cits are unli ely to be a common cause of reading comprehension dif+culties. ubsequent d ata, however, suggest that this conclusion was premature. ;he children in

/tothard and >ulme<s study were aged 8< years, and it is clear that they found the listening s pan tas demanding because performance was close to Coo. -ation et al. !%'''*, using the same materials as /tothard and >ulme, did observe substantial listening span de+cits in % -year-old poor compre- henders in line with the +ndings eported by Auill et al. !%')' *. ;a en together, these +ndings ae consistent with the general relationship between ve rbal !but not spatial* span and reading comprehension in children !e.g., /eigneuric, Erlich, 1a hill, . Auill, #$$$*. "iven the relative dif+culty poor compehenders have with the p rocessing and storage of verbal material, it is tempting to suggest that these verbal wor ing memory de+cit

SSR14

11/27/04

10:54 AM

Page 258

#B) Kate Nation may underpin their poor text comprehension. >owever, issues of causality are far from clear. ,t is generally accepted that in a complex memory span tas such as listening span, individuals< storage capacity is a function of how ef+cient th y are at the computational <or processing aspects <of the tas !e.g., 5aneman . ;ardif, %')0*. As, by de+nition poor comprehenders have poor language comprehension, verbal wor ing memory wea - ness may be a consequence of poor language comprehension, rather than a cause of it. 3or example, the listening span tas used by -ation et al. !%'''* required children to listen to sentences and to ma e a decision about their content !the processing compo- nent of the tas *, and then to remember the +nal wod in each s uccessive sentence !the storage component of the tas *. As poor comprehenders tend to perform less well than control children on measures of sentence comprehension !Cragg . -ation, in press& /tothard . >ulme, %''#*, differences in verbal memory are perhaps not surprising. ,n summary, although further research is needed to fully understand issues of causal- ity, it is clear that the relationship between language comprehension and verbal memory in poor comprehenders is an intimate one, as highlighted in a recent investigation of sen- tence repetition. Garshall and -ation !#$$7* as ed poor comprehenders and controls to repeat sentences !of increasing length and complexity* verbatim. ;wo +ndings ere clear. 3irst, poor comprehenders repeated fewer sentences correctly. ;hus, even though the tas required only straightforward verbatim repetition !rather than a complex span procedure*, verbal memory wea nesses were nevertheless evident. /econd, the nature of the errors made by poor comprehenders differed from those made by control children. =hile both groups of children were li ely to maintain the surface aspects of the sentences, poor com- prehenders were less li ely to maintain the meaning !or gist* of the target sentences. A possible interpretation of this +nding is that the childen sim ply did not understand the sentences as well as control children did, thus reducing the accuracy or reliability with which they were able to represent !and therefore remember* the content of sentences. *igher+order?discourse+le,el processes ,n addition to lexical processing and wor ing memory de+cits, range of discourse-level de+cits hae been implicated in poor reading comprehension. ;wo sets of processes that have attracted considerable research will be reviewed hereF inference ma ing and com- prehension monitoring processes. !nference ma'ing( ;o understand language, it is often necessary to ma e inferences <to go beyond what is stated explicitly in the text or discourse to

infer the intended message. Even very straightforward texts require inferences to be drawn. ;his point is nicely illus- trated by 1a hill !%''2* in her description of how the following story < can only be under- stood against a bac ground nowledge about birthday parties, the convention of ta ing presents to them, the need for money to buy presents, and so on<!p. )##*F
-ane was in,ited to -ac'?s $irthday( "he wondered if he would li'e a 'ite( "he went to her room and shoo' her piggy $an'( !t made no sound(

SSR14

11/27/04

10:54 AM

Page 259

Children?s Reading Comprehension Difcultie #B' As this example ma es clear, failure to draw inferences is li ely to seriously impede comprehension. -umerous studies have demonstrated that poor comprehenders have dif+culty drawing infeences when reading or listening, and it has been argued that such dif+culties ae causally implicated in children<s poor reading comprehension !Cain . 1a hill, %'''& 1a hill, %')#, %')2& for review, see 1a hill, %''2*. 6efore accepting that poor comprehenders<reading dif+culties re the consequence of problems with ma ing inferences, it is important to establish that poor inference ma ing is not an artefact of other factors. 3or example, to resolve an inference often requires the reader to hold information in memory across a number of sentences. 9otentially, there- fore, poor comprehenders may fail to ma e inferences not because they are unable to do so, but simply because of failure to remember premises presented earlier in the text. ;o test this idea, 1a hill !%')2* compared children<s ability to answer comprehension ques- tions requiring an inference under two conditions. ,n one condition, the text remained in full view, and children were allowed to loo bac at the text& in a second condition, the text was removed and comprehension questions had to be answered from memory. 9oor comprehenders<ability to draw inferences remained limited, even when the text was made available for the children to loo bac at. 3ollowing a similar procedure, Cain and 1a hill !%'''* replicated these +ndings.;hey also included a c ondition in which poor comprehenders were encouraged !with direct prompting* to search the text in order to +nd the information needed fom which to ma e an inference. ,nt erestingly, poor com- prehenders<performance increased, leading Cain and 1a hill to suggest that it is not so much the case that poor comprehenders cannot ma e inferences, but rather that they fail to do so spontaneously. ;he problems that poor comprehenders have ma ing inferences raises a number of issues that have been addressed by Cain, 1a hill, and colleagues. 3irst, linguists and psy- cholinguists distinguish between different types of inference !e.g., /inger, %''2*. 5o poor comprehenders have dif+culty with all types of infeence< ;wo t ypes of inference have been studied in detail in poor comprehendersF cohesive inferences and elaborative inferences. Cohesive inferences are needed to establish and maintain lin s between premises within the text and are necessary if adequate comprehension is to follow. Elaborative infer- ences are made when information external to the text is integrated with information con- tained in the text& these inferences are not always essential, but they are thought to enrich the readers< representation of the text. "enerally it appears t hat poor comprehenders perform less well than controls on both types of inference, although presenting the text for the children to refer bac to results in greater improvement for cohesive than elaborative inferences !Cain . 1a hill, %'''& Cain, 1a hill, 6arnes, .

6ryant, #$$%*. A different form of inferencing was assessed by 1a hill !%')7*. ,nstantiation refers to the process whereby a speci+c meaning of a wod is constructed, dep ending on context. 3or example, long-term recall of a sentence such as <the +sh attac ed the swimmer<is enhanced following a cue such as the noun shar', relative to the original !but less context speci+c* noun, s . Consistent with their tendency to draw fewer elaborative inferences, poor comprehenders are also less li ely to ma e instantiations than control children !1a hill, %')7*. A second issue concerns whether it the process of inference ma ing per se that is impaired for these children or, alternatively, is it that they lac the relevant nowledge needed to ma e the inference< ;o disentangle the effects of group differences in bac -

SSR14

11/27/04

10:54 AM

Page 260

#8$ Kate Nation ground nowledge from group differences in inference ma ing s ills, Cain et al. !#$$%* taught a novel nowledge base to groups of poor comprehenders and control children. /peci+call, %# pieces of information about the imaginary plane t Gan were read to the children !e.g., the ponds on Gan are lled with oange .uice/ $ears on Gan ha,e $right $lue fur*. Cain et al. then tested acquisition of the nowledge base using a forced-choice picture recognition tas , and a verbal recall tas . Any items that were not recognized or recalled correctly were retaught. ;he children then listened to a story tapping the nowledge base and were as ed questions that required either a cohesive or an elaborative inference to be made. 9oor comprehenders were poor at generating both types of inference, even though they did not differ in terms of world nowledge relevant to the text. ,mportantly, it is interesting to note that the poor comprehenders were also poor at answering literal ques- tions about the stories, suggesting that their lac of understanding is not ?ust a conse- quence of failure to ma e inferences. -evertheless, the results of this study are clear in showing that even when relevant bac ground is familiar, poor comprehenders draw fewer inferences than do s illed comprehenders. Arguably, a limitation of the research pointing to poor inference ma ing as a cause of poor reading comprehension is its circularityF since children in these studies were selected as poor comprehenders precisely because they performed poorly on a standardized test of discourse comprehension containing a high proportion of inference-based questions, it might be considered unsurprising that they performed less well than control children on experimental tas s tapping inference-ma ing ability. ;o address the issue of causality, Cain and 1a hill !%'''* utilized the comprehension-age match design described earlier. ;hey compared the inference-ma ing abilities of 0<-year-old poor co mprehenders with those of normally developing children aged 8< years. ;he two groups of children did not differ in terms of comprehension scores on the test used to screen participants, the Neale #nalysis of Reading #$ility . -evertheless, poor comprehenders made fewer inferences on the experimental tas s than the younger children, leading Cain and 1a hill to conclude that poor inference-ma ing ability is a candidate cause of poor reading and language comprehension. Comprehension monitoring( Comprehension monitoring refers to a set of metacognitive control processes that individuals can draw upon as they read or listen. ,n s illed reading, comprehension monitoring results in an assessment of whether comprehension has been successful, and repair strategies may be initiated if miscomprehension is

detected. ,n short, comprehension monitoring refers to a set of strategies that indicate that a reader is engaged with the text. 1a hill and Auill !%''8* described three strands of evidence pointing to de+cits in compehension monitoring in children selected as hav ing speci+c eading com- prehension dif+culties. irst, they are less li ely to resolve anomalies in text. 3or example, Auill et al. !%')'* presented scenarios containing an apparent anomaly !e.g., a mother is pleased that her son is not sharing sweets with his younger brother*. :esolving information !that the younger son is on a diet* was presented later in the scenario. Children were then as ed whether the mother had behaved appropriately, and why she too the action she did, the rationale being that these questions can only be answered correctly if the dif- ferent sources of information are integrated. =hile poor comprehenders were able to resolve anomalies well when the two pieces of information were ad?acent in the story,

SSR14

11/27/04

10:54 AM

Page 261

Children?s Reading Comprehension Difcultie #8% their performance fell dramatically when the two +ller sentenc s intervened between the two premises. ;hese +ndings suggest that poor compehenders are able to integrate infor- mation adequately, but fail to do so when the tas is made more demanding by increas- ing its memory load. A rather different second strand of evidence pointing to comprehension monitoring de+cits in poor compehenders come from an anomaly detection ta s in which children are as ed to underline any meaningless words or phrases !Auill . 1a hill, %''%*. Even when they were explicitly instructed that a text contained nonsense words and phrases, poor comprehenders were less li ely to detect them than control children. ,n addition, poor comprehenders were less li ely to detect inconsistencies in a text, particularly when the inconsistencies within the text were separated by a number of sentences. ;a en together, these +ndings suggest that poor compehenders are not engaged in constructive comprehension monitoringF they fail to notice when comprehension has gone astray and are thus not well placed to initiate repair strategies. 1bserving that poor comprehenders are less li ely to monitor their own comprehen- sion does not, on its own, establish comprehension monitoring as a cause of reading comprehension failure. ,nstead, faulty comprehension monitoring may well be a conse0uence of faulty comprehension rather than a cause of it. ,ndeed, comprehension monitoring ability is not a static or +ed variable& de /ousa and 1a hill !%''8* found that poor com- prehenders< comprehension monitoring ability increased substantially when they engaged in a more interesting tas . ,n contrast, comprehension monitoring levels were fairly con- stant for the s illed comprehenders, regardless of the tas <s interest level. ;hese +nding caution against the view that comprehension monitoring is a processing wea ness impli- cated in reading comprehension failure. ,n summary, studies by 1a hill and colleagues have provided numerous demonstra- tions of the dif+culties that poor compehenders have with disc ourse-level processes. Goreover, it has been proposed that poor inference s ills are a li ely cause of poor comprehension !e.g., Cain . 1a hill, %'''*. >owever, on a number of counts, 9erfetti has questioned the validity of the view that speci+c e+cits in higher-leel s ills are causally implicated in reading comprehension impairments !e.g., 9erfetti, %''2*. >is preferred account is that poor inference ma ing or failure to detect anomalies are not examples of structural or speci+c de+cits that cause a co rehension problem. :ather, they are the comprehension problem, a problem that stems from wea nesses in <the operation of basic processes that identify words, activate their meanings, con+gue phrases, assemble meanings and so forth< !9erfetti et al., %''8 , p. %B'*. 6efore consider- ing this

perspective further, it is necessary to move away from processing factors to consider the other potential cause of comprehension failure& that is, differences in nowledge. Knowledge (nowledge is essential to comprehension. =ithout an appreciation of the meanings of words, there can be no comprehension. Goving beyond the meaning of individual words, domain nowledge is also considered crucial for comprehension. Appreciation of the

SSR14

11/27/04

10:54 AM

Page 262

#8# Kate Nation domain that is being referred to in a text allows the reader to move from a word- or propositional-level representation of the text to one which integrates this nowledge with a broader body of bac ground nowledge, thus allowing the reader to build a potentially inference-rich mental model of the situation or event !see (intsch . :awson, this volume*. 9rior nowledge about a text predicts comprehension of it !/pilich, Hesonder, Chiesi, . Hoss, %'0'* and it is plain that complete lac of nowledge will result in a complete lac of comprehension !recall the example described earlier of Gilton<s daugh- ters reading aloud 4atin and "ree with no comprehension*. 6ut, is it the case that there are systematic de+ciencies in poor compehenders< nowledge bas e that can account for their faulty comprehension< As reviewed earlier, there is evidence suggesting that poor comprehenders have rela- tive wea nesses in expressive and receptive vocabulary !-ation et al., #$$2*, indicative of lac of nowledge at the word level. Although it seems li ely that lac of vocabulary nowledge may contribute to impaired comprehension, it is unli ely to be the whole storyF comprehension wea nesses are still apparent when care is ta en to include vocab- ulary that is familiar, and when domain nowledge is to some extent controlled by teach- ing the children a novel nowledge base from which comprehension is subsequently assessed !Cain et al., #$$%& although it should be noted that in this study poor compre- henders too longer to learn the nowledge base, and showed poorer retention of it over time. Although this was controlled statistically in their analyses, it cannot be ruled out that differences in nowledge base !perhaps in terms of the quality of its representation* may have existed between the two groups*. :ather than describe nowledge as being present or absent, a different approach is to as whether individuals differ in the extent to which they activate nowledge sponta- neously, or bring it to bear rapidly and ef+ciently at the app opriate time. 3or example, -ation and /nowling !%'')a* reported that poor comprehenders were slower to ma e semantic ?udgments than control children. ,n a similar vein, Cain and 1a hill !%'''* reported that poor comprehenders<ability to ma e inferences increased when they were assisted to +nd the elevant part of the text. ;hese two observ ations are both examples of instances when poor comprehenders had the required nowledge, but failed to deploy it either quic ly or spontaneously. Alternatively, however, these observations could be interpreted as indicative of lac of nowledge in that it is only when nowledge is thoroughly understood and properly integrated that it can be reCec ed on rapidly, or used to trigger inferences. ow+le,el ,ersus high+le,el processing1 and processing ,ersus 'nowledge As reCected in the abve review, the literature on speci+c ead ing

comprehension dif< culties has concerned itself with dichotomies. Are poor comprehenders<dif+culties bes understood in terms of processing de+cits or lac of nwledge< ,f poor comprehenders have processing wea nesses, are they Ilw level<or Ihigh leel < >owever, it is not clear whether these dichotomies are useful or psychologically valid. ;wo examples will be used to illustrate what is meant here, one concerning word meaning and vocabulary and one concerning verbal memory.

SSR14

11/27/04

10:54 AM

Page 263

Children?s Reading Comprehension Difcultie #87 ;here are now a number of studies demonstrating vocabulary wea nesses in children selected as having poor reading comprehension !e.g., -ation et al., #$$2& -ation . /nol- wing, %'')a& /tothard . >ulme, %''#*. ,n some ways, wea vocabulary is a clear index of lac of nowledge and, as noted above, comprehension will fail if children simply do not understand the words they read or hear. >owever, the question then arises as to why poor comprehenders have wea vocabulary nowledge. 4ac of vocabulary nowledge is associated with wea nesses in verbal ,J, and, consistent with this, poor comprehenders achieve lower verbal ,J scores than control children !-ation, et al., #$$#*. >owever, as a substantial component of verbal ,J is vocabulary nowledge, this observation does not help us understand the nature of poor comprehenders<dif+culti s. Gore interesting is the notion that an individual<s ability to learn new words or acquire new information from context is a vital s ill that mediates the high correlations observed between verbal ability, reading comprehension, and vocabulary nowledge !e.g., /ternberg . 9owell, %')7*. Cain, 1a hill, and Elbro !#$$2* examined the ability of poor comprehenders to learn new words from context by presenting stories containing a novel word !whose meaning was discernible from context* and as ing children to de+ne thenovel words, either before the context allowed word meaning to be inferred, or afterwards. 9oor comprehenders were less li ely to offer de+nitions for the nvel words, especially when the distance between the word and the information needed to infer its meaning was lengthened by inserting +ller sentences.;his study is interesting, as it demonstrates how Ihigher-leel< processes such as the ability to ma e inferences and integrate information within a text can inCuence the acquisition of basic Iwer-level< nowledge such as the meaning of a new word. ,n turn, nowledge of word meanings and their speedy activation during reading !or listening* may well assist children<s Ihigherleel<proces sing during language comprehension !-ation . /nowling, %'')b, %'''*. ;hus, it is perhaps not surprising to +nd that childen who are poor at ma ing inferences tend to hav e wea er vocabulary s ills relative to children who are s illed at ma ing inferences, and vice versa. ;he overlap between processing and nowledge is also demonstrated when we con- sider the possibility that long-term memory may contribute to poor comprehenders< de+cits on memoy span tas s. A well-replicated +nding is that poor comprehenders perform equivalently to control children on straightforward tas s of verbal short-term memory capacity !i.e., recall tas s such as forward digit span that do not require an addi- tional processing component& e.g., /tothard . >ulme, %''#*. /imilarly, poor compre- henders show normal effects of word length and phonological similarity in short-term memory !-ation et al., %'''& 1a hill, Auill, . 9ar in, %')8*. ;a en together, these +nd ings demonstrate that poor comprehenders

do not have de+cits i short-term verbal memory capacity. ,mportantly however, the extent to which poor comprehenders show normal short-term recall depends critically on the nature of the items to be recalled. -ation et al. !%''', Experiment #* compared short-term serial recall for lists of concrete words !e.g., tooth, plate, fruit* and abstract words !e.g., luc', pride, wise*. Although poor comprehenders and controls did not differ in memory span for concrete words, poor com- prehenders recalled fewer abstract words. ,t is established that the availability of seman- tic information inCuences shot-term recall, and that concrete words may receive more <semantic support< than abstract words during short-term recall !=al er . >ulme, %'''*. =ithin this theoretical framewor , -ation et al. !%'''* suggested that as poor

SSR14

11/27/04

10:54 AM

Page 264

#82 Kate Nation comprehenders have semantic wea nesses, they bene+t less fom s emantic support, espe- cially when the semantic contribution to recall is stressed by as ing them to recall abstract rather than concrete words, and under these circumstances short-term recall <usually reported as an area of strength <is compromised. ;hese two examples highlight the inherent dif+culty !and perha s even futility* of distinguishing between nowledge and processing as sources of poor comprehenders< dif+culties.=ithin an interactive !and developing* language sy stem, it seems li ely that dif+culties at one leel will inCuence peformance at anot her& similarly, long-term nowledge will inCuence pocessing ef+cienc, and individual di fferences in processing will lead to differences in long-term nowledge. -agy and Anderson !%')2* have argued that from the beginning of third grade, the amount of free reading children engage in is the ma?or determinant of vocabulary growth. 9reliminary data !Cain, %''2, cited in 1a hill . Auill, %''8* suggest that poor comprehenders have substantially less reading and reading-related experience than control children. Although Cain<s data need to be interpreted cautiously due to the sample size being very small, they are consistent with a view that sees individual differences in reading comprehension failure becoming compounded over time. ;hus, Gatthew effects !see the glossary at end of this volume* are li elyF poor comprehenders may read less, and learn less from their reading experi- ences than their peers& therefore impacting on subsequent reading and learning oppor- tunities over time and leading to the formation of wea Iintelectual habits<!9erfetti et al., %''8*.

Summary and Conclusions


Comprehension is complex and multifaceted, and it is thus no surprise that the popula- tion of children identi+ed as having eading comprehension dif< culties form a heteroge- neous group. Even when discussion is limited to those children who have welldeveloped decoding ability, as in this chapter, heterogeneity is still apparent !-ation et al., #$$#, #$$2*. >owever, it is possible to draw some clear conclusions. Children with <speci+< comprehension problems do exist, and they are not unusual !Auill . 1a hill, %''%*, although they are rare in clinically referred populations !e.g., 4each et al., #$$7& /han weiler, et al., %'''*. =hile it is clear that decoding inef+ciency will lead to eadin g comprehension dif+culties !e.g., erfetti, %')B*, not all children who have c omprehension dif+culties have impairments in basic decoding, nor do they experience a phonological bottlenec & the children described in the studies reviewed in this chapter decode well, and they have appropriate phonological processing s ills.

9oor comprehenders do, however, have wea - nesses in other aspects of language s ill with de+cits at bothlower !e.g., vocabulary nowl- edge* and higher levels !e.g., inference generation, understanding +guratie language* being reported !-ation et al., #$$2*. Although it is possible to describe different tas s according to whether they tap low-level or high-level processes, it is argued here that the distinction between different levels of processing may not be useful, at least until longer- term longitudinal data become available.

SSR14

11/27/04

10:54 AM

Page 265

Children?s Reading Comprehension Difcultie #8B 1ne of the dif+culties facing the esearcher interested in unde rstanding the nature of poor comprehenders<dif+culties is that typicall, the children are selected for study on the basis of their reading pro+le. Consequentl, we now very l ittle about the develop- ment of language in preschool or preliterate children who go on to become poor com- prehenders. 4ong-term longitudinal studies are needed if we are to understand better the precursors to, and consequences of, <speci+< dif+culties with reading comprehension.

SSR14

11/27/04

10:54 AM

Page 266

You might also like