Preschooler Components of Listening Comprehenison

A New Measure for Assessing the Contributions of Higher Level Processesto Language Comprehension Performance in Preschoolers

Brenda Hannon and Sarah FriasTexas A&M Kingsville

The present study reports the development of a theoretically motivated measure that provides estimatesof a preschooler’s ability to recall auditory text, to make text-based inferences, to access knowledge fromlong-term memory, and to integrate this accessed knowledge with new information from auditory text.This new preschooler component processes task is based on measures developed by Hannon andDaneman (2001) and August, Francis, Hsu, and Snow (2006), but it uses pictures and auditory text tomake it more suitable for children 4–6 years of age. The results show that the new task is suitable forunderstanding the contributions of higher level processes to performance on a measure of languagecomprehension. In fact, it appears to be a better predictor of language comprehension performance thanit is a measure of working memory. In addition, its medium-knowledge integration component is a goodpredictor of performance on a composite measure of fluid intelligence. Finally, a factor analysis reveals3 separate clusters of abilities: word decoding skills, higher level processes associated with text-basedprocessing, and higher level processes associated with accessing prior knowledge.

Keywords: language comprehension, preschoolers, higher level processes

A considerable amount of research has been directed towardunderstanding the contributions of a wide range of cognitiveskills/abilities to reading comprehension ability in beginning read-ers (Kendeou, van den Broek, White, & Lynch, 2009; McCardle,Scarborough, & Catts, 2001).1 A few studies have also targetedpreschoolers who are able to decode/identify only a few words(Kendeou, van den Broek, et al., 2009). This latter research isimportant because it increases our understanding of the develop-mental trajectories of the cognitive components in language com-prehension that are crucial to reading (Kendeou, van den Broek, etal., 2009). It also identifies the early language components (i.e.,phonological, orthographic, syntactic, lexical, inferential, etc.) thatare associated with the greatest risk of poor future reading achieve-ment (McCardle et al., 2001). Unfortunately, most studies havetargeted beginning readers rather than preschoolers who can, atbest, read only a few words (Kendeou, van den Broek, et al., 2009;for a review, see Storch & Whitehurst, 2002). Moreover, the fewstudies that have targeted preschoolers have tended to examine theinfluences of a single cognitive component rather than multipleones (Saarnio, Oka, & Paris, 1990). Thus, for preschoolers wehave limited knowledge about the interrelationships and depen-dencies of many cognitive components.

In this study we focused on higher level cognitive components/processes that are used to learn and infer text-based information, to

access prior knowledge from long-term memory, and to integrateprior knowledge with text-based information. Higher level pro-cesses are important for comprehension because they are used toconstruct an integrated and coherent representation of a text (e.g.,Kintsch, 1988, 1994, 1998). A number of developmental and adultstudies have examined the influences of higher level cognitivecomponents on reading comprehension, but few studies have ex-amined their influences on listening or language comprehension(i.e., listening comprehension, oral comprehension, and/or picturecomprehension) in preschoolers (van den Broek et al., 2005).Consequently, one of our goals was to determine whether higherlevel processes predict performance on a measure of languagecomprehension administered to preschoolers.

Our second goal was to develop a theoretically motivated mea-sure of higher level cognitive processes suitable for preschoolers.This new, language-based measure assesses a preschooler’s abilityto recall new information from auditory text and pictures, to drawtext-based inferences from auditory text and pictures, to accessprior knowledge from long-term memory, and to integrate thisaccessed knowledge with new auditory- and picture-based infor-

1 The literature includes many terms to describe language comprehen-sion ability (e.g., oral language skills, oral comprehension, reading com-prehension, and listening comprehension). In the present study we use theterms reading comprehension ability and reading skills to describe abili-ties/skills that are exclusively reading. In contrast, we use the termslanguage comprehension ability and oral language skills to describe abil-ities/skills that are either (a) exclusively orally based or (b) are primarilybut not exclusively oral language/listening. For instance, we describe ourstandardized measure of language comprehension as a measure of languagecomprehension because the passages and questions are presented aurally.However, this measure is not exclusively aural as the answer choices forthe multiple-choice questions are pictures.

Brenda Hannon and Sarah Frias, Department of Psychology, TexasA&M Kingsville.

I would like to thank Susan Goldman for her helpful comments. I wouldalso like to thank Stella Lopez and Stephanie Keller for all of theirencouragement. Participants were paid with a faculty research fund.

Correspondence concerning this article should be addressed to BrendaHannon, Department of Psychology, Texas A&M Kingsville, Kingsville,TX. E-mail: [email protected]

Journal of Educational Psychology © 2012 American Psychological Association2012, Vol. ●●, No. ●, 000–000 0022-0663/12/$12.00 DOI: 10.1037/a0029156

1

AQ: 1

Fn1

tapraid5/zcz-edu/zcz-edu/zcz00312/zcz2436d12z xppws S�1 7/2/12 19:35 Art: 2010-1170

bel

Text Box

no changes

mation. These four cognitive components were selected becausethey are highly predictive of performance on measures of readingand listening comprehension in beginning readers and adults (e.g.,August et al., 2006; Hannon & Daneman, 2001). They are also thesame components assessed in reading-based tasks from which thepreschooler task is modeled after, namely, the beginning readers’component processes measure called the Diagnostic Assessment ofReading Comprehension (DARC; August et al., 2006) and theadult component processes task (e.g., Hannon, 2012; Hannon &Daneman, 2001).

Our third goal was to validate the preschooler component pro-cesses task. For this goal we assessed its predictive, construct,discriminant, incremental, and nonverbal predictive validities (seethe last section of the introduction for more information about thisgoal).

In the following, we briefly review research assessing higherlevel processes, word decoding, working memory, fluid intelli-gence, and the relationship between reading and listening compre-hension. In the final section, we describe the adult and beginningreaders’ versions of the reading-based component processes taskas well as the present study.

Higher Level Cognitive Processes

There is considerable evidence in both the developmental andadult research literature that higher level processes—characterizedas those cognitive processes that are used for bridging or connect-ing ideas in a text, inferring themes from text, predicting futureoutcomes, inferring meanings of words from contextual clues, andcomputing the antecedent referents of pronouns—are all sources ofindividual differences in performance on measures of reading andlanguage comprehension (e.g., August et al., 2006; Barnes, Den-nis, & Haefele-Kalvaiti, 1996; Cain, Oakhill, & Bryant, 2004;Hannon, 2012; Hannon & Daneman, 1998, 2001, 2006, 2009;Long & De Ley, 2000; Long, Oppy, & Seely, 1994; Oakhill, 1982,1984). Indeed, August et al. (2006) showed that, in a group of 7-to 9-year-olds, higher level processes, measured by the reading-based DARC, were suitable for predicting performance on listen-ing comprehension and oral language measures (i.e., subtests fromthe Woodcock Language Proficiency Battery; rs � .46 and .53,respectively). Similarly, studies using multiple reading tasks inorder to assess the independence of knowledge of story structure,comprehension monitoring, and inferencing making in young read-ers, have shown that each measure accounts for as much as 26% ofthe variance in performance on a measure of global reading com-prehension (e.g., Cain et al., 2004).

Researchers also suggest that the higher level processes ofpreschoolers, although less sophisticated, mirror those of olderchildren and adults (e.g., Kendeou et al., 2005; van den Broek etal., 2005; van den Broek, Lorch, & Thurlow, 1996). Researchexamining vocabulary acquisition, for example, shows that chil-dren as young as 3 years old can acquire the definitions of newwords by listening to storybook readings (e.g., Senechal, 1997;Senechal & Cornell, 1993; see Senechal, Thomas, & Monker,1995, for 4- to 5-year-olds). In these studies, preschoolers infer thedefinitions of novel words by combining contextual clues that areinterspersed in pictures and short auditory stories (e.g., Senechal,1997; Senechal et al., 1995). Research also shows that 3- to5-year-olds correctly execute relational and causal inferences im-

plied in aural text 63% to 67% of the time (e.g., Daneman &Blennerhassett, 1984). Four- and 6-year-olds also tend to be betterat recalling events from television programs when the events havemany causal connections, rather than just a few, a finding thatsuggests they are sensitive to relational information (van denBroek et al., 1996). Indeed, research suggests that after the pre-sentation of aural or aural-pictorial narrative texts, preschoolerscan generate a variety of knowledge-based inferences, includingaction-related, goal-related, and causal antecedent (e.g., Kendeou,Bohn-Gettler, White, & van den Broek, 2008).

There are, however, some systematic age differences in thoseprocesses that are used to infer semantic relations (Kendeou et al.,2008; van den Broek et al., 2005). Younger children, such aspreschoolers and pre-readers, make fewer spontaneous inferencesthan older children (e.g., Oakhill, Cain, & Bryant, 2003), and theyare more likely to make inferences when they are questioned(Kendeou et al., 2008) or when the inferences are necessary forcoherence (e.g., Casteel, 1993; Casteel & Simpson, 1991). Incontrast to the inferences made by older children, the inferences ofpreschoolers are also generally less complex and require moresupport from prior knowledge, auditory text, or pictures (van denBroek et al., 2005). Whereas older children make inferences basedon both abstract and concrete objects, preschoolers are primarilylimited to concrete objects (e.g., van den Broek, 1989). Olderchildren make inferences based on both external and internalevents (e.g., feelings, goals of characters), whereas preschoolersare primarily limited to external ones (van den Broek et al., 2005).Finally, older children make inferences among multiple events/objects, whereas preschoolers are limited to single events/objects(van den Broek et al., 2005).

There are also limitations to the existing preschooler research.For example, although language studies show that preschoolers arecapable of executing higher level processes, very few studies havereported the simultaneous contributions of one or more higherlevel processes to language comprehension performance (see Flo-rit, Roch, & Levorato, 2011, and Kendeou et al., 2008, for excep-tions). Nor is there a psychometric measure of higher level pro-cesses suitable for preschoolers. Thus, although we know thatpreschoolers do execute many of the higher level processes impli-cated in language comprehension, we know little about their si-multaneous predictive powers for language comprehension ability.

Letter–Word Decoding/Identification

Most research examining reading development in pre-readers,preschoolers, and beginning readers has focused on basic literacyskills such as letter-word identification and phonological aware-ness (van den Broek et al., 2005). However, because letter-worddecoding/identification measures were included only as part of thevalidation of the preschooler task, we limit our discussion to (a) therelationship between letter-word decoding/identification (e.g.,letter-word identification, phonological awareness) and languagecomprehension and (b) the relationships between letter-word de-coding/identification and higher level processes.

It is generally accepted that letter-word decoding/identificationand listening comprehension are separate abilities. Indeed theseparation between letter-word decoding/identification and listen-ing comprehension in preschoolers and beginning readers is sup-ported by studies that use a number of different statistical methods,

2 HANNON AND FRIAS


Brenda

Sticky Note

The sentence beginning with Similarly is inaccurate because it reads like there were multiple tasks and each task assessed multiple components. Rather, it needs to read that each task is independent and therefore independently assesses a separate cognitive component. In other words it needs to read like I requested on the proof: Similarly, studies using multiple reading tasks that independently assess . . .

different tasks, and different populations. Preschooler studies us-ing principal component factor analysis, for example, show thatmeasures of letter-word decoding/identification (e.g., letter-soundknowledge, letter-word identification, phonological awareness,nonword repetition) and listening comprehension (e.g., the GroupReading Assessment and Diagnostic Evaluation, aural narrativetext) form separate factors (e.g., Aouad & Savage, 2010; Kendeou,Savage, & van den Broek, 2009; Protopapas, Simos, Sideridis, &Mouzaki, in press). Additionally, developmental studies examin-ing children who are poor readers show two distinct groups ofreaders: (a) children with good word decoding skills but poorlistening comprehension skills (e.g., Catts, Adlof, & Weismer,2006; Catts, Hogan, & Fey, 2003) and (b) children with poor worddecoding skills but good listening comprehension skills (e.g.,Adlof, Catts, & Little, 2006; Catts et al., 2006; Spooner, Baddeley,& Gathercole, 2004).

Developmental correlational and experimental studies examin-ing beginning readers also suggest that word/letter-word decoding/identification and higher level processes are largely independent(e.g., Aaron, Franz, & Manges, 1990; August et al., 2006; Carver,1998; Crain, 1989; de Jong & van der Leij, 2002; Frith & Snowl-ing, 1983; Oakhill et al., 2003). Studies examining beginningreaders, for instance, show that letter-word decoding/identificationmeasures (e.g., letter-word identification, phonological coding)and reading-based measures of higher level processes (e.g., textmemory, text inferencing and knowledge integration) are weaklyrelated (e.g., August et al., 2006; Oakhill et al., 2003; see Hannon,2012, for adults). However, although a few studies have examinedthe relationship between measures of letter-word decoding/identification and language comprehension in preschoolers (e.g.,Kendeou, Savage, & van den Broek, 2009; Kendeou, van denBroek, et al., 2009; Protopapas et al., in press) as well as therelationship between measures of higher level processes and lan-guage comprehension in preschoolers (e.g., Kendeou et al., 2008),to the best of our knowledge no study has examined the relation-ship between measures of letter-word decoding/identification andhigher level processes. Thus, for preschoolers it is unclear whetherletter-word decoding/identification and higher level processes aredependent or independent.

Working Memory and Fluid Intelligence

Because measures of working memory and fluid intelligencewere included only as part of the validation of the preschooler task,in this section we limit our discussion to research examining (a)the relationship between working memory and language compre-hension, (b) the relative power of higher level processes versusworking memory to predict reading and listening comprehension,and (c) the powers of working memory versus higher level pro-cesses to predict fluid intelligence.

There is evidence in both the preschooler and beginning readerresearch literature that reading- and listening-based measures ofcomplex working memory (i.e., measures that both process andstore information) are strongly related to measures of readingand/or language comprehension (Adams, Bourke, & Willis, 1999;Aouad & Savage, 2009; Daneman & Blennerhassett, 1984; Cain etal., 2004; Florit, Roch, Altoe, & Levorato, 2009; Florit et al., 2011;Leather & Henry, 1994). For preschoolers, typical correlationsbetween measures of working memory and listening comprehen-

sion range from .30 to .45 (e.g., Aouad & Savage, 2010; Florit etal., 2009; Leather & Henry, 1994), but they also can be as high as.80 (e.g., Daneman & Blennerhassett, 1984). These correlationsexist under diverse situations: (a) when the target population ispreschoolers or pre-readers; (b) when the working memory mea-sure is listening- or reading-based; and (c) when the workingmemory measure involves processing multisyllable nonwords, re-calling word lists backward, or recalling sentence final words.

Some issues, however, still remain. One issue is that the existingsentence-based, preschooler measures of working memory are notfree of problems. For instance, Daneman and Blennerhassett(1984) modeled their preschooler listening measure of workingmemory after the original Daneman and Carpenter (1980) adultlistening span; preschoolers listened to sets of sentences (e.g.,“Birds have wings”; “I like toys.”) and, at the end of each set, weresupposed to recall the final words of the sentences (e.g., wings,toys). However, because their preschoolers tended to recall fullsentences, Daneman and Blennerhassett based their working mem-ory scores on the number of sentences recalled rather than thesentence final words. As a result, critics contend that their studyonly shows that preschoolers who are better at recalling auditorysentences are also better at understanding language comprehension(Adams & Willis, 2001). Variants of Daneman and Blennerhas-sett’s listening span measure have also experienced similar prob-lems with preschoolers (e.g., Adams et al., 1999).

A second issue is whether working memory measures accountfor unique variance in language comprehension performance overand above measures of higher level processes. Recently, adultstudies examining the relative powers of measures of workingmemory (i.e., reading and operation span) versus higher levelprocesses (i.e., text memory, text inferencing, knowledge integra-tion) in adults to predict performance on measures of readingcomprehension show that working memory accounts for little ofthe variance in reading once the variance attributed to higher levelprocesses has been partialed out (e.g., Britton, Stimson, Stennett,& Gulgoz, 1998; Daneman & Hannon, 2007; Hannon & Daneman,2006, 2009). Recent developmental studies also show a similarpattern with 7- and 8-year-olds (e.g., Oakhill et al., 2003). How-ever, to date, no study has examined the power of working mem-ory versus higher level processes to predict language comprehen-sion performance in preschoolers. Thus, it is unclear whether thesame pattern of results will be observed when language compre-hension is the criterion measure and preschoolers are the targetpopulation.

A third issue is whether measures of working memory accountfor unique variance in fluid intelligence performance beyond mea-sures of higher level processes. A number of development andadult studies suggest that working memory is an excellent predic-tor of fluid intelligence (e.g., Ackerman, Beier, & Boyle, 2005; deRibaupierre & Lecerf, 2006; Engel de Abreu, Conway, & Gath-ercole, 2010; Engle, Tuholski, Laughlin, & Conway, 1999; Kyl-lonen & Christal, 1990; Swanson, 2008). Developmental studiesexamining 5- to 9-year-olds, for example, show that measures ofworking memory (e.g., counting and backward digits recall) typ-ically account for about 11.6% of the variance in performance onnonverbal measures of fluid intelligence (e.g., Raven’s ColoredProgressive Matrices; Engel de Abreu et al., 2010). Nevertheless,recently, some researchers have argued that the working memory–fluid intelligence relationship is the byproduct of a subset of

3PRESCHOOLER COMPONENT PROCESSES TASK


processes that are part of working memory, rather than workingmemory per se (e.g., Deary, 2002; Detterman, 2002). Studies, forexample, show that measures of knowledge integration, a higherlevel process that involves integrating text, are highly predictive ofa wide range of fluid intelligence measures, such as Raven’sMatrices, Cattell’s Culture Fair Test, and the Shipley (e.g., Bueh-ner, Krumm, & Pick, 2005; Hannon & Daneman, 2008; Oberauer,Su�, Wilhelm, & Wittman, 2003, 2008). However, this latterresearch has been limited to adults.

Relationship Between Reading and LanguageComprehension

Researchers tend to adopt one of two opposing views about therelationship between language and reading comprehension abili-ties during the early stages of reading development (Kendeou, vanden Broek, et al., 2009). One view suggests that language com-prehension skills are not important during early reading skillacquisition because language comprehension skills are not com-pletely functional until word decoding skills are acquired (Speece,Roth, Cooper, & de la Paz, 1999; Vellutino, Tunmer, Jaccard, &Chen, 2007). In contrast, the other view suggests that languagecomprehension skills are important during early reading skill ac-quisition even during the early stages of reading development(Bishop & Adams, 1990; Catts, Fey, Zhang, & Tomblin, 1999;Kendeou, van den Broek, et al., 2009; Paris & Paris, 2003; Pro-topapas et al., in press).

The lack of consensus in the literature is likely the result of anumber of limiting factors. For example, most studies have tar-geted children who can read, even though some language andletter-word decoding skills begin developing before reading acqui-sition (Kendeou, van den Broek, et al., 2009). In addition, manystudies have used cross-sectional designs, which are less accuratethan longitudinal designs for revealing developmental trajectoriesof skills (Kendeou, van den Broek, et al., 2009).

A recent study by Kendeou, van den Broek, et al. (2009)however, has provided considerable insight into the language-reading debate. In their study, Kendeou, van den Broek, et al. useda longitudinal design to assess the developmental trajectories andrelationships among letter-word decoding/identification, languagecomprehension, and reading comprehension in 6-year-olds whohad not entered Grade 1. At Time 1, the children completedmultiple measures of letter-word decoding/identification (subtestsfrom the Woodcock Reading Mastery Test—Revised; Woodcock,1987, and the Onset Recognition Fluency measure) and languagecomprehension (listening comprehension, television comprehen-sion, vocabulary). At Time 2, which was 2 years later, the samechildren completed additional measures of letter-word decoding/identification, language comprehension, and reading comprehen-sion.

Five of Kendeou, van den Broek, et al.’s (2009) results arerelevant to the present study. First, language comprehension per-formance at Times 1 and 2 were strongly related. Second, therewere nonsignificant relationships between letter-word decoding/identification skills and language comprehension at both Times 1and 2. These two nonsignificant relationships are consistent withthe typical finding that letter-word decoding/identification andlanguage comprehension skills are largely independent, even forpreschoolers (e.g., Kendeou, Savage, & van den Broek, 2009;

Protopapas et al., in press). Third, at Time 2 the relationshipsbetween (a) letter-word decoding/identification and reading com-prehension and (b) language and reading comprehension were bothstrongly related. These two significant relationships support theview that both letter-word decoding/identification and languagecomprehension skills contribute to early reading development. Inaddition, the latter finding suggests that for preschoolers, languagecomprehension ability, measured with tasks assessing listening andtelevision comprehension, is highly predictive of reading compre-hension ability. It also suggests the possibility that the componentprocesses of language comprehension are highly related to andpotentially predictive of reading comprehension ability.

The Preschooler Component Processes Task

The new preschooler task is modeled after the adult and chil-dren’s reading-based measures developed by Hannon and Dane-man (2001, 2006) and August et al. (2006). Below we first de-scribe these two measures and then the new preschooler task.

In the adult component processes task (e.g., Hannon & Dane-man, 2001), readers learn short three-sentence paragraphs thatdescribe relations among real and artificial terms. For example, “ANORT resembles a JET but is faster and weighs more. A BERLresembles a CAR but is slower and weighs more” and “A SAMPresembles a BERL but is slower and weighs more.” The relationsin a paragraph (e.g., faster than) form linear orders (e.g., nort �JET � CAR � berl � samp); however, in order to form theseorders, readers must access their prior knowledge for facts (e.g., “ajet is faster than a car”) and then integrate them with the paragraphinformation, because these facts are not explicitly stated in theparagraph. After studying a paragraph at their own pace, readersthen respond to four types of true–false test statements. Textmemory statements (e.g., “A NORT is faster than a JET”) assessthe ability to remember new information explicitly mentioned in aparagraph; no prior knowledge is required. Text inferencing state-ments (e.g., “A SAMP is slower than a CAR”) assess the ability tomake deductions or inferences about paragraph information (e.g.,from “A SAMP is slower than a BERL” and “A BERL is slowerthan a CAR,” a reader can logically deduce that “a SAMP is slowerthan a CAR”); again, no prior knowledge is required. On the otherhand, knowledge access statements (e.g., “a JET is faster than aCAR”) assess the ability to access facts from long-term memory;no paragraph information is required. Finally, knowledge integra-tion statements (e.g., “A NORT is faster than a CAR”) assess theability to access facts from prior knowledge (e.g., “a jet is fasterthan a car”) and integrate these facts with new paragraph infor-mation (e.g., “A NORT is faster than a CAR”).

Hannon and Daneman (2001) observed a pattern of correlationsamong the four components that matched the descriptions of thestatement types. Text memory and text inferencing, the two com-ponents that relied on text-based information but not informationfrom prior knowledge, were highly correlated with each other (r �.83), but neither component correlated well with the knowledgeaccess component, the component that relied only on informationfrom prior knowledge (average r � .26). On the other hand,knowledge integration, the component that depended on bothtext-based information and prior knowledge, correlated with thetwo text-based components (average r � .62) as well as theknowledge access components (average r � .35; e.g., Hannon &

4 HANNON AND FRIAS

AQ: 2


Daneman, 2001, Experiment 2). Taken together, these correlationssuggest that knowledge integration taps both new paragraph infor-mation and prior knowledge, but the text-based components aretapping different abilities than is the knowledge access component.

Hannon and Daneman (2001) also validated their measure mul-tiple ways. First, their adult task accounted for 50%–60% of theperformance on a standardized measure of global reading compre-hension (i.e., construct and predictive validity). It also accountedfor a greater amount of variance in reading performance than twoother well-known predictors, namely, working memory and vo-cabulary knowledge (i.e., incremental validity). Finally, Hannonand Daneman showed that the text memory component was thebest predictor of performance on a separate task that measured justtext memory, the text inferencing component was the best predic-tor of a separate task that measured just text inferencing, theknowledge access component was the best predictor of a separatetask that measured just knowledge access and that the knowledgeintegration component was the best predictor of a separate task thatmeasured just knowledge integration (i.e., construct validity).

Building on the research of Hannon and Daneman (2001),August et al. (2006) developed the DARC, a reading-based com-ponent processes task suitable for beginning readers. Like itspredecessor, the DARC assesses four higher level componentprocesses; however, unlike its predecessor, it includes only onemeasure of knowledge integration, rather than three (i.e., low-,medium-, and high-knowledge integration). Studies show that theDARC predicts performance on language comprehension mea-sures, such as both the reading and listening subtests of theWoodcock Language Proficiency Battery (e.g., August et al.,2006). Studies also show that the four components are, at best,weakly related to letter-word decoding/identification measures,such as the letter-word identification and letter-word attack sub-tests from the Woodcock Reading Mastery Test—Rrevised(Woodcock, 1987; rs � .28 and .22, respectively). This latterfinding provides some discriminant validity for the DARC becauseit suggests that higher level processes and letter-word decoding/identification are separate skills.

Like its predecessors, the preschooler task assesses four higherlevel processes. However, the preschooler task uses pictures andauditory text rather than printed text. The preschooler task alsoincludes an animated introduction that provides a contextualframework for facilitating learning, whereas the adult task and theDARC do not (see Fullerton, 1983, and Potts, 1977, for more aboutadditional context). Finally, the preschooler task includes threeknowledge integration measures that vary in complexity, whereasthe DARC includes only one. These additional knowledge inte-gration measures increase the potential of the preschooler task asa diagnostic tool.

Table 1 shows an example of the three parts of a paragraph: (a)an animated introduction, (b) a to-be-learned auditory paragraph,and (c) auditory test statements that assess four higher levelprocesses: text memory, text inferencing, knowledge access, andknowledge integration. The animated introduction consists of ashort video clip of a girl clicking a camera and the simultaneousauditory message, “Emma went to the zoo to take pictures of theanimals. While taking pictures of animals, Emma learned . . . .”The next two pictures accompany the auditory text for the two-sentence paragraph; the first picture accompanies the message, “AJIMP looks like a DOG,” and the second picture accompanies the

message, “A JIMP is larger than an ELEPHANT.” Yes–no teststatements assessing the four component processes follow thepresentation of the last picture.

We also assess the validity of the preschooler task in a numberof ways. To assess predictive validity, we examine the power ofthe preschooler task to predict an age-appropriate, standardizedmeasure of global language comprehension. To assess discrimi-nant validity, we examine the relationships between the pre-schooler task and two measures of letter-word decoding/identification. Letter-word decoding/identification was selectedbecause previous developmental research suggests that letter-worddecoding/identification and higher level processes are separateskills (e.g., August et al., 2006). For incremental validity wecompare the powers of the preschooler task versus a listening spanmeasure of working memory to predict performance on a stan-dardized test of global language comprehension. Working memorywas selected because previous research suggests that it is a verygood predictor of language comprehension performance, evenwhen the measures of working memory are listening-based and thetarget population is preschoolers (e.g., Daneman & Blenner-haussett, 1984). Finally, for nonverbal predictive validity we ex-amine the powers of the preschooler task to predict performanceon a nonverbal measure of fluid intelligence. Nonverbal fluidintelligence was selected because current adult research suggeststhat measures of higher level cognitive processes are predictive ofnonverbal fluid intelligence (e.g., Buehner et al., 2005; Hannon &Daneman, 2008; Oberauer et al., 2003, 2008).

Method

Participants and Design

Seventy-three preschoolers, 4 to 6 years of age, were recruitedvia flyers posted at the University of Texas at San Antonio anddaycares in the surrounding area. All preschoolers were pre-screened to ensure that they were monolingual English-speakingpreschoolers (as opposed to bilingual children or children in Grade1) and that they had no developmental, emotional, behavioral/learning disabilities. Each preschooler’s level of word decodingability was assessed using two criteria: (a) overall reading level,which was reported by the attending parent (i.e., not at all, someletters of the alphabet, a few sight words, etc.), and (b) phonolog-ical letter-word decoding skill, which was assessed with the letter-word attack subtest of the Woodcock Reading Mastery Test—Revised. Phonological decoding skill was selected as a criterionfor assessing word reading ability because research suggests thatphonemic awareness is strongly related to reading (e.g., Ehri,Nunes, Stahl, & Willow, 2001). To calculate a score for phono-logical word decoding skill, we subtracted three from a preschool-er’s letter-word attack score (i.e., phonological word decodingskill � letter-word attack score – 3) because the first three items onthe letter-word attack subtest are phonemes and not words. AsTable 2 shows, the average phonological word decoding score forthe 73 preschoolers was less than a quarter of a word (i.e., M �0.16). In other words, most of the preschoolers were unable todecode words. Indeed, 49 of the 73 preschoolers were unable todecode a single word, and of the remaining 21 preschoolers, 13


T1

T2


Table 1Example of Paragraph and Test Statements of the Preschooler Component Processes Task

Animated Introduction

Emma went to the zoo to take pictures of the animals. While takingpictures of animals, Emma learned:

Paragraph

A JIMP looks like a DOG.

A JIMP is larger than an ELEPHANT.

Relationship: JIMP � Elephant � DogText Memory

A JIMP looks like a DOG. A JIMP looks like an ELEPHANT. (n)A JIMP is larger than an ELEPHANT. An ELEPHANT is larger than a JIMP. (n)

Text Inferencing� not possible �

Low-knowledge integrationA JIMP is larger than a DOG. A DOG is larger than a JIMP. (n)

Medium-knowledge integrationA JIMP is larger than a CAT. A CAT is larger than a JIMP. (n)

High-knowledge integrationLike GIRAFFES, JIMPS have long legs. Like CATS, JIMPS have short legs. (n)

Low-knowledge access Type IAn ELEPHANT is larger than a DOG. A DOG is larger than an ELEPHANT. (n)

Low-knowledge access Type IIAn ELEPHANT is larger than a CAT. A CAT is larger than an ELEPHANT. (n)A GIRAFFE is larger than a DOG. A DOG is larger than a GIRAFFE. (n)

Note. See Appendix A for examples of paragraphs that include text inferencing statements.

6 HANNON AND FRIAS


could decode only one or two words.2 Because one of our goalswas to determine how well the preschooler task might predictlanguage comprehension ability in pre-readers, those preschoolerswho were unable to phonologically decode words (i.e., had ad-justed letter-word decoding scores � 0) and unable to identifywords by sight (i.e., had adjusted letter-word identificationscores � 0) were identified as pre-readers. This subset of 36pre-readers was included in the group data analyses as well as theanalysis that assessed how well the preschooler component pro-cesses task predicted language comprehension performance inpre-readers.

The average age of the 73 children was 5.31 years (SD � 0.55);13 were 4-year-olds, 52 were 5-year-olds, and eight were 6-year-olds. Thirty-eight were female and 35 were male. As well, 40 wereof European descent, 20 were of Hispanic descent, and the remain-ing 13 were either Asian, African American, or of mixed descent.The average family income was $36,000/year to $50,000/year (theattending parent selected one of seven choices; lowest option was�$12,000 and highest option was $100,000�/year). Finally par-ents had, on average, attended some classes at a 4-year college(range � some high school to post-graduate).

Each child was tested individually in one session that wasapproximately 2.0–2.5 hr in length. The session time included a10-min break. Although no parent was present in the testing room,if a child was too shy, the door was left ajar so that the attendingparent could hear, but not see, the session as it progressed. Besidesreviewing and signing a consent form, the parent completed a briefdemographics form as well as a form assessing the extent of theirchild’s prior knowledge for facts used in the preschooler compo-nent processes task. Information from this latter form was used toassess the prior knowledge assumption of the preschooler task.Each child was compensated with a $10.00 package of toys, andthe attending parent received a $10.00 gift card.

All children completed the following tasks in the followingorder: (a) the preschooler component processes task; (b) an objectassembly task (Wechsler Preschool and Primary Scale of Intelli-gence, 3rd ed. [WPPSI-III]; Wechsler, 2002); (c) the Gates-MacGinitie measure of comprehension, form PR (MacGinitie,MacGinitie, Maria, & Dreyer, 2000); (d) a block completion task(WPPSI-III; Wechsler, 2002); (e) mazes (WPPSI-III; Wechsler,2002); (f) animal pegs (WPPSI-III; Wechsler, 2002); (g) a measureof working memory; (h) a letter-word identification task (Wood-cock Reading Mastery Test—Revised; Woodcock, 1987); and (i)the letter-word attack task (Woodcock Reading Mastery Test—Revised; Woodcock, 1987). As an additional incentive to encour-age completion of the study, we told children that they wereplaying mental games and that for each game they completed theywould earn tickets that could be traded in for prizes, a practice ofa local children’s restaurant. All children completed all of thetasks.

Preschooler Component Processes Task (PR-CPT)

Each paragraph consisted of three parts: (a) a short animatedintroduction, (b) a two-sentence aural paragraph, and (c) test

2 By subtracting three from the letter-word attack score, the lowest valueof the letter-word attack score becomes –3.0. An alternative way of scoringthe data would be to subtract three from the letter-word attack score andthen replace each negative score with a score of 0.0. When negative scoresare replaced with 0.0, the average for the letter-word decoding task is stillless than one word (M � 0.82, SD � 1.69). However, whereas thedistributions was normal when three was subtracted from the raw score(i.e., skewness � 1.08 and kurtoisis � 1.24), when negative scores areadjusted to zero, the distribution becomes abnormal (skewness � 3.35 andkurtosis � 4.98).

Table 2Means and Standard Deviations for Measures of Higher Level Component Processes (i.e., Preschooler Task), LanguageComprehension, Working Memory, Word Decoding Tasks, and Composite Measure of Fluid Intelligence

Measure

Combined (n � 73) Pre-readers (n � 36)

M (SD) Min Max % (SD) M (SD) Min Max % (SD)

Preschooler Component Processes TaskText memory (max � 24) 16.1 (3.44) 10 24 67.0 (14.32) 16.4 (2.99) 11 23 68.3 (12.45)Text inferencing (max � 8) 5.6 (1.44) 2 8 69.9 (18.02) 5.7 (1.33) 3 8 70.8 (16.64)Low-k. integration (max � 4) 3.1 (0.84) 1 4 77.7 (21.07) 3.0 (0.91) 1 4 75.7 (22.75)Med-k. integration (max � 12) 8.4 (2.13) 3 12 70.1 (17.78) 8.2 (1.75) 5 12 70.6 (12.36)High-k. integration (max � 12) 8.5 (1.88) 5 12 70.9 (15.66) 8.5 (1.48) 6 11 68.1 (14.57)Low-know access-1 (max � 4) 3.0 (0.97) 1 4 74.6 (24.11) 2.9 (1.03) 1 4 72.9 (25.62)Low-know access-2 (max � 12) 9.3 (1.77) 5 12 77.6 (14.76) 8.9 (1.91) 6 12 74.5 (15.93)

Other tasksLanguage comprehension (max � 20) 11.3 (3.18) 5 18 10.8 (3.09) 5 18Working memory (max � 5) 3.8 (1.16) .25 5 3.7 (1.08) 1.5 5

Letter-word decoding/identification tasksLetter-word ident. (max � 63) 1.2 (5.62) �11 15 �2.9 (3.98) �11 0Letter-word attack (max � 29) 0.2 (2.19) �3 7 �1.1 (1.05) �3 0Fluid intelligence (max � 400) 75.1 (14.26) 44 97 67.3 (14.12) 44 90

Note. Fluid intelligence is a composite score of four measures. With the exception of the letter-word decoding tasks (i.e., letter-word identification andletter-word attack), all the measures have a minimum score of zero. The scores for the letter-word identification measure have been adjusted so that theyreport the number of sight words identified. For this reason the range of scores for the letter-word identification task is �13 to 63. The scores for theletter-word attack measure have been adjusted so that they report the number of words that were decoded. For this reason the range of scores for theletter-word attack measure is �3 to 29.


Fn2

AQ: 3


statements that assessed knowledge about the paragraph. All threeparts were presented via Microsoft PowerPoint using a 23-in. �14-in. (58.42-cm. � 35.56-cm) computer screen and computerspeakers. To synchronize the audio and visual information, theaudio information was embedded into the PowerPoint slides and,in many instances, began automatically once the slide changed.The total administration time was 15–20 min, and the Cronbachalphas indicated that all of the components were reliable for boththe 73 preschoolers (range � .83 to .85) and the 36 pre-readers(range � .78 to .81). See Table 1 and Appendix A for the completeset of stimuli.

PR-CPT materials. This section is divided into four parts:animated introduction, paragraphs, test statements, and validationof the prior knowledge assumption and pictures.

PR-CPT animated introduction. The animated introductionspresented the topics of the paragraphs. Each introduction included(a) a two-sentence pre-recorded auditory message and (b) ananimated clip-art image that were played simultaneously on acomputer. The tone of the voice in the auditory message waspositive and by supplying a name for each child depicted in ananimation, we believe the auditory message became more con-crete. The auditory messages were 13 to 18 words in length (M �16.6 words), and their durations ranged from 10 s to 11.25 s (M �10.8 s). The animated images were clip-art selected from Mi-crosoft Office Online. Each animation was in color and containeda single character that made small, subtle movements.

PR-CPT paragraphs. The preschooler task included onepractice paragraph and four critical paragraphs. Each paragraphwas presented aurally and described the relationships among a setof real and nonsense objects that could be used to construct a linearordering; for example, the size order of JIMP � ELEPHANT �DOG for the paragraph in Table 1. Each sentence in the paragraphwas accompanied by a picture of the real and nonsense objectsdescribed in the sentence.

The first two critical paragraphs included two real and onenonsense objects (e.g., the paragraph in Table 1), whereas thesecond two paragraphs included one real and two nonsense ob-jects. This increase in the number of nonsense objects presumablyincreased the difficulty of the paragraphs because two nonsenseobjects should be more difficult to learn than one. Each paragraphalso described one semantic feature among the objects (e.g., thesize feature for the paragraph in Table 1).

Two criteria were used to select the pictures of the objects. First,we tried to select prototypical pictures that preschoolers couldeasily recognize (see Validation section below for confirmation ofthis fact). Further, the pictures depicting the nonsense objectspreserved the basic properties of the original objects, even thoughthe pictures were modified slightly. For example, as shown inTable 1, the picture of a JIMP still looks like a dog.

We also included a number of other elements in the paragraphs:(a) Our selection of topics (i.e., fruit, animals, vehicles, sea life,and insects) was based on informal inquiries of parents, whoindicated that their young children were familiar with these topics.(b) The contents were pre-recorded to ensure consistent adminis-tration. (c) We tried to keep the language simple. For example, thephase “looks like” was used rather than “resembles,” which is theverb used in the adult task. (d) We tried to make the tone ofthe recordings similar to children’s television programs (i.e., pos-

itive). (e) Finally, each sentence and its corresponding picture werepresented simultaneously using PowerPoint.

Although the preschooler task was modeled after Hannon andDaneman’s (2001) adult task, there were a number of differences.For example, whereas the preschooler task included animatedintroductions, the adult task did not. Whereas the preschooler taskused one semantic feature and three objects, the adult task includedtwo to four features and five objects. As well, whereas the pre-schooler task included two sentences, the adult task included three.Finally, whereas the preschooler task contained four critical para-graphs, the adult task contained six.

PR-CPT test statements. After each paragraph, children an-swered pre-recorded test statements. Half of the statements weretrue and half were false. False statements were created by revers-ing the objects in the true statements.3 Although the test statementsassessed four main processes (text memory, text inferencing,knowledge access, and knowledge integration), there were twotypes of knowledge access statements (i.e., low-knowledgeaccess-1, low-knowledge access-2) and three types of knowledgeintegration statements (i.e., low-, medium-, and high-knowledgeintegration). These different types of knowledge access and knowl-edge integration statements allowed us to examine a preschooler’sability to manipulate different types of information. In total, therewere 76 test statements, and the dependent measure for each typeof statement was the total number correct. Table 3 describes eachstatement type in detail.

PR-CPT: Validating prior knowledge assumption and identi-fication of pictures. As in Hannon and Daneman (2001), datawere collected in order to verify that the assumption of priorknowledge was met. Although about 3% of the questions were leftunanswered by the attending parents, on average, parents believedthat their children knew approximately 92.1% of the facts pre-sented in the paragraphs. This percentage is similar to that ob-served by Hannon and Daneman.

We also assessed whether children could correctly identify theline drawings that depicted real and nonsense objects. Nine chil-dren (average age � 5.41, SD � 0.44, range � 4.5–5.92), whowere naive to the purpose of this study, were individually asked toidentify the object in each line drawing. The results showed that96% of the time the children correctly identified the objects.

PR-CPT procedure. The instructions were pre-recorded andplayed on computer speakers (see Appendix B for the full instruc-tions). Briefly, children were told that we were interested in seeinghow they used what they already knew about common things inorder to help them learn about new things. We also told them thatthey would be asked to answer some questions. The researchassistant presented the practice paragraph and encouraged eachchild to answer each question based on what they thought was true.The four critical paragraphs followed the practice paragraph.

Each paragraph was presented in a fixed order: the animatedintroduction, the first sentence and its corresponding picture, andthen the second sentence and its corresponding picture. As eachsentence played the research assistant pointed to the objects. Allparagraphs were presented two times in order to ensure some levelof encoding. After the second presentation of the paragraph, the

3 The only exceptions to this rule were the high-knowledge integrationstatements because they could not be created by reversing the objects.

8 HANNON AND FRIAS

Fn3

T3


test statements were played aurally in a fixed random order. Theresearch assistant manually recorded each “yes” or “no” responseon an answer sheet and then prompted the child with the word,“Ready?” before presenting the next test statement.

Measure of Language Comprehension

Our measure of language comprehension was the preschooler ver-sion (i.e., level PR) of the Gates-MacGinitie (MacGinitie et al., 2000).The PR version of the Gates-MacGinitie was selected over othermeasures, because the Gates-MacGinitie measures, in general, arevery popular measures that are widely used in schools systems as bothnorm-referenced and survey tests (e.g., MacGinitie & MacGinitie,1989). In addition, researchers have successfully administered the PRversion of the Gates-MacGinitie to both 4- and 5-year-olds (e.g.,Hresko, Peak, Herron, & Bridges, 2006; Stark et al., 1984). Thisversion of the Gates-MacGinitie included five short passages and 20questions that were read to the child. Each passage was broken intofive small segments, and each segment was followed by a singlemultiple-choice question that was composed of an auditory questionstem and three picture choices. The child’s task was to select thepicture that correctly answered the question.

Typical administrations of the Gates MacGinitie yield a Cron-bach alpha of .88. The actual Cronbach alphas for the 73 pre-schoolers and the 36 pre-readers were .75 and .75, respectively.The validity of the Gates-MacGinitie has been indirectly estab-lished from both field testing and equating studies (Fuhriman,2002). The Gates-MacGinitie also has some content validity, sincesome of its passages are adapted from previously published nar-rative and expository texts (MacGinitie & MacGinitie, 1989).

To verify that the PR version of the Gates-MacGinitie was suitablefor 4-year-olds, we compared the performance of the thirteen 4-year-olds to the performance of the fifty-two 5-year-olds. The results

revealed no age difference in language comprehension performance(t � 1.20, p � .24). Thus, it appears that the language comprehensionmeasure was equally suitable for 4- and 5-year-olds.

Measures of Letter-Word Decoding/Identification Skills

The measures of letter-word decoding/identification skills werethe letter-word identification and letter-word attack subtests fromthe Woodcock Reading Mastery Test—Revised (Woodcock,1987). We specifically chose these two letter-word decoding/identification measures because August et al. (2006) used thesemeasures to establish discriminant validity for the DARC, a pre-decessor of the preschooler task. Further, the letter-word andletter-word attack subtests are suitable for children as young as 2years of age (Woodcock, Mather, & McGrew, 2001). Finally, boththe letter-word and letter-word attack subtests have high reliabili-ties(Cronbach �s � .90 or higher), and their validities are consid-ered good (Woodcock et al., 2001). Indeed, the Cronbach alphasfor the two letter-word measures for the 73 preschoolers were .87and .93.

In the letter-word identification task, children identify letters/words, whereas in the letter-word attack task, they pronounceletters/nonwords. Thirteen of the first 14 items in the letter-wordidentification task are letters, and the first three items in theletter-word attack task are phonemes (i.e., p, k, and n). For thisreason, a sight word identification score was calculated by sub-tracting 13 from the total score of the letter-word identificationtask (i.e., sight word identification � letter-word identificationscore – 13). The scores range from –13 to 63. Similarly, a pho-nological word decoding score was calculated by subtracting threefrom the total score of the letter-word decoding task (i.e., phono-logical word decoding � letter-word attack score – 3). Scores for

Table 3Number, Descriptions, and Examples of Statement Types

Statement type Number Description Example

Text memory 24 Assessed a child’s ability to process explicit information stated ina paragraph; no prior knowledge was required.

“A JIMP is larger than a DOG.”

Text inferencing 8 Assessed a child’s ability to process information that was impliedin a paragraph; no prior knowledge was required.

Not possible for Paragraph 1 but possiblefor Paragraphs 2 and 4

Low knowledge access Assessed a child’s ability to access prior knowledge; no text-basedinformation was required. The statements varied in complexity;Type 2 presumably more complex

Type 1 4 Included 2 real terms and 1 feature mentioned in paragraph “An ELEPHANT is larger than a DOG.”Type 2 8 Included 1 real term and 1 feature mentioned in paragraph and 1

real term that was not mentioned“An ELEPHANT is larger than a CAT.”

Knowledge integration Assessed a child’s ability to integrate prior knowledge with text-based information. The statements varied in complexity,presumably low-know integration lowest in complexity.

Low 8 Example requires child to integrate the prior knowledge fact (“anelephant is larger than a dog”) with text fact (“A JIMP is largerthan an ELEPHANT”). Statements included 2 real terms and 1feature that were included in the paragraph.

“A JIMP is larger than a DOG.”

Medium 12 Included a nonsense term and feature mentioned in paragraph and1 real term that was not mentioned

“A JIMP is larger than a CAT.”

High 12 Included a nonsense term mentioned in paragraph and a real termand feature that were not mentioned

“Like GIRAFFES, JIMPS have legs.”

Note. Although the number of statements ranged from four to 24, their reliabilities were all high (Cronbach alphas ranged from .78 to .85). These highreliabilities suggest that the number of statements likely have a minimal impact on the final results of the study.



this measure ranged from –3 to 29. Both tasks ended when sixsuccessive responses are incorrect.

Measure of Working Memory

As mentioned earlier, the existing sentence-based, preschoolermeasures of working memory have been criticized because chil-dren recalled the sentences rather than the sentences’ final words(Adams & Willis, 2001). In our new measure we avoided thisproblem by requiring children to recall missing words from pre-viously presented sentences. This new cloze measure was identicalto an adult span task developed by Masson and Miller (1983), butit included simpler sentences. In this measure, children listened toa set of sentences and then, at the end of each set, heard twosentences from the set. However, both of these sentences weremissing words, which were denoted by beeps. At this point thechildren recalled the missing words. For example, the correctresponse for the three-sentence set, “I wear a coat. The girl used afork. The family is at the house,” and then the cloze sentences “Iwear a **beep**. The **beep** is at the house,” would be coat,family. This new cloze span task included four two-, three-, four-,and five-sentence sets (56 sentences total) and the two clozesentences that followed each set. One cloze sentence had a wordmissing near the beginning (e.g., “The **beep** is at the house”),while the other had a word missing at the end (e.g., “I wear a**beep**”). Children listened to increasingly longer sets of sen-tences until they responded incorrectly to three consecutive sets.At this point the task was over.

Scores were based on the set size for sets that had clozesentences answered correctly. For a child to get full credit for aparticular set size, the cloze sentences for at least two completesets needed to be answered correctly. If the cloze sentences foronly one complete set were answered correctly, then a part scoreof .25 was awarded (1 out of 4 � .25). For example, if the clozesentences for two of the four-sentence sets were answered cor-rectly and the cloze sentences for one five-sentence set were alsoanswered correctly, then the span score was 4.25 (i.e., 4.0 for thefour-sentence set and .25 for correctly completing one of the fourfive-sentence sets). All stimuli and instructions were pre-recordedto ensure consistent administration. The Cronbach alpha for thisnew span measure was .92.

Measures of Fluid Intelligence

The composite measure of fluid intelligence was derived fromthe object assembly, animal pegs, mazes, and block completionsubtests of the WPPSI-III (Wechsler, 2002). We selected thesefour nonverbal subtests for two reasons. First, when combinedthese four subtests become a composite measure of performanceintelligence (Goldstein & Hersen, 2000; Wechsler, 2002), whichcorrelates well with other measures of performance intelligence(Goldstein & Hersen, 2000) and working memory (e.g., Adams etal., 1999). Second, because these subtests are nonverbal theyprovided an intelligence measure without taxing the preschoolerswith additional verbal information. The object assembly, blockscompletion, mazes, and animal pegs subtests have high reliabilitywith Cronbach alphas � .89 or higher (Woodcock, Shrank,McGrew, & Mather, 2006), and a number of studies have shownconcurrent validity between the WPPSI-R (Wechsler, 1989) and

the Wechsler Intelligence Scale for Children (3rd ed.; Wechsler,1991) performance subtests (e.g., Goldstein & Hersen, 2000). TheCronbach alphas in the present study ranged from .77 to .84. Eachtask was administered using the standard instructions.

To confirm that all four subtests assessed the same construct, wesubmitted the accuracy scores to a factor analysis with a promaxrotation (i.e., correlated solution). The results revealed that all foursubtests loaded heavily on a single factor (eigenvalue � 1.89,accounting for 47.4% of the variance). Using the factor loadings ofeach subtest, we then created a composite fluid intelligence scoreby summing the products of each subtest’s factor loading with achild’s score for that same subtest: composite fluid intelligence �(object assembly factor loading � object assembly score) �(blocks factor loading � blocks score) � (mazes factor loading �mazes score) � (animal pegs factor loading � animal pegs score);see Hannon and McNaughton-Cassill, (2011) for procedure.

Results

There are four analyses sections: (a) data screening, (b) descriptivestatistics, correlations, and initial factor analysis, (c) validation of thepreschooler task using the group data (i.e., combined data for the 4-,5-, and 6-year-olds), and (d) regression analysis predicting languagecomprehension for a subset of children who can neither phonologi-cally decode nor identify words (i.e., pre-readers).

The descriptive statistics are reported for two groups of children:(a) the overall group (i.e., combined data for 4-, 5-, and 6-year-olds)and (b) just the 36 pre-readers. The separate descriptive statistics forthe pre-readers are included because researchers might be interested inthe separate performance of this group of children. For the correla-tions, a correlation greater than .50 was deemed to be a large corre-lation, a correlation between .36 and .49 a medium correlation, anda -correlation of .35 or less a small correlation.

Data Screening

Before completing the analyses we used SAS and regressionanalysis to screen the data for (a) outliers (DFITTS, DFBETAS,studentized residuals, and Cook’s D) and (b) data points exertingexcessive leverage (hat-values). The results revealed that no singledata point was an outlier exerting excessive leverage. The datawere also normally distributed with no evidence of skew or kur-tosis (Mardia’s PK � 0.986, a value well below the 1.96 limit;Tabachnick & Fidell, 2007). Indeed, even the skewness and kur-tosis for the letter-word measures, which were transformed so theyreported the number of words decoded/identified (i.e., scales of–13 to 63 and –3 to 29), were well below the critical threshold of3.0 (i.e., skewness and kurtosis of letter-word identification task �0.03 and 0.22, respectively, and skewness and kurtosis of letter-word decoding task � 1.08 and 1.24, respectively). The lack ofevidence for significant skewness and kurtosis in the two letter-word measures suggests that the distributions of the letter-wordmeasures are normal, even though the actual word reading levelsof the preschoolers are low.

Descriptive Statistics, Correlations, and Initial FactorAnalysis

Table 2 shows the descriptive statistics for both the 73 pre-schoolers and the 36 pre-readers (i.e., preschoolers who are unable

10 HANNON AND FRIAS

AQ: 4

AQ: 5


to decode or identify words), and Table 4 shows the correlationsfor the group data (i.e., all 73 preschoolers). In Table 2, the meansand standard deviations for the components of the preschooler taskare also reported in percentages so that direct comparisons can bemade between the preschooler task and the adult reading version ofthe component processes task. However, all of the analyses, in-cluding the correlations that are reported immediately below, usedjust the raw data. As Table 2 shows, all the measure captured awide range of individual differences.

Optimal level of difficulty. According to Kaplan and Sa-cuzzo (2001), when measures are at an optimal level of difficultythey capture a wide range of individual differences. For measureswith two choices, such as the yes–no components of the pre-schooler task, Kaplan and Sacuzzo recommend that the means beat or near 75%. As Table 2 shows, the range for performance onthe components was 67% to 77.7%. Thus, it appears that all of thecomponents are close to optimal levels of difficulty. Nevertheless,as Table 2 reports the lowest values for all of the components werebelow chance level. This finding is consistent with administrationsof other versions of the component processes task, such as thebeginning and adult reading versions of the component processestask (e.g., August et al., 2006; Hannon & Daneman, 2001).

Developmental trends. Three developmental trends wereevident in Table 4. The first developmental trend was that perfor-mance on all higher level component processes increased with age(rs � .23 to .44). The second developmental trend was thatperformance on the measure of language comprehension alsoincreased with age (r � .45). The third developmental trend wasthat fluid intelligence increased with age (r � .61). Because all ofour measures of fluid intelligence rewarded faster performance,this last developmental trend was probably a consequence ofcognitive speed differences between 4-, 5-, and 6-year-olds (e.g.,Fry & Hale, 2000).

Correlations among measures. As Table 4 shows, manycomponents of the preschooler task were significantly related toother cognitive measures. For instance, all of the components of

the preschooler task were significantly correlated with the measureof language comprehension (rs � .32 to .65). The components oftext memory, text inferencing, medium-knowledge integration,and high-knowledge integration were also significantly correlatedwith the measure of working memory (average r � .37), whereasthe two knowledge access components (i.e., low-knowledgeaccess-1 and low-knowledge access-2) were not (average r �.215). This latter finding is consistent with the adult readingcomprehension literature, which suggests that working memory ismore about having the capacity to process new information, ratherthan the ability to access knowledge from long-term memory (e.g.,Daneman & Hannon, 2007; Hannon & Daneman, 2001, 2009).Finally, consistent with the preschooler literature, the measures ofworking memory and language comprehension were significantlycorrelated (r � .53; e.g., Daneman & Blennerhassett, 1984).

Factor analysis of the components of preschooler task. Afactor analysis was completed to confirm that the pattern of rela-tionships among the components of the preschooler task wassimilar to the pattern of relationships observed by Hannon andDaneman (2001, 2006, 2009), who validated the adult readingversion of the component processes task. The expected pattern wasobserved. That is, as the first factor analysis in Table 5 shows, thetwo types of text-based statements, text memory and text infer-encing, loaded heavily on factor one (i.e., .771 and .843) butweakly on factor two (i.e., .297 and –.117). On the other hand, thetwo types of knowledge access statements, low-knowledgeaccess-1 and low-knowledge access-2, loaded heavily on factortwo (i.e., .843 and .760) but weakly on factor one (i.e., .024 and.219). Additionally, low-, medium-, and high-knowledge integra-tion loaded on both factors one and two (i.e., range of loadings �.334 to .747), although low-knowledge integration loaded more onfactor two (i.e., .641) than factor one (i.e., .338), whereas medium-and high-knowledge integration loaded more heavily on factor one(i.e., .747 and .701, respectively) than factor two (i.e., .843 and.760, respectively). This pattern of factor loadings for the knowl-edge integration components is different from the pattern observed

Table 4Correlations Among Components of Preschooler Task, Language Comprehension, Working Memory, Letter-Word Identification, WordAttack, Composite Measure of Fluid Intelligence, and Age of Child (n � 73)

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Text memory — .55� .53� .62� .51� .35� .29� .65� .47� .16 .34� .20 .30�

2. Text inferencing — .19 .45� .44� .08 .13 .47� .25� .13 .39� .19 .40�

3. Low-know integration — .38� .39� .43� .35� .47� .21 .09 .19 .24� .27�

4. Medium-know integration — .60� .23� .52� .54� .43� .32� .37� .44� .42�

5. High-know integration — .26� .42� .45� .34� .19 .29� .33� .39�

6. Low-know access-1 — .49� .32� .21 .12 .17 .28� .23�

7. Low-know access-2 — .43� .22 .25� .18 .32� .44�

8. Language comprehension — .53� .30� .39� .44� .45�

9. Working memory — .30� .27� .37� .28�

10. Letter-word identification — .69� .59� .53�

11. Letter-word attack — .36� .45�

12. Fluid intelligence — .61�

13. Age of child —

Note. Although the present study is based on the principal of componential analysis, we were curious about how much variance in languagecomprehension might be accounted for by a single composite measure of the pre-reader task. Therefore, we calculated a composite measure for thepre-reader task based on factor scores for the components. The composite preschooler measure had a .64 correlation with the measure of languagecomprehension.� p � .05.


T4

AQ: 6

T5


with the adult reading version of the component processes task.For the adult task, all three knowledge integration componentsloaded more heavily on factor one (i.e., text-based components oftext memory and text inferencing) than factor two (knowledgeaccess components).

Combined the two factors accounted for 65.6% of the overallvariance; factor one, with an eigenvalue of 3.39, accounted for48.45% of the total variance, while factor two, with an eigenvalueof 1.20, accounted for 17.19% of the total variance. Because thetwo types of text-based statements loaded heavily on factor one,we described this factor as a text-based processing factor.4 On theother hand, because the two types of knowledge access statementsloaded heavily on factor two, we described this factor as aknowledge-access factor.

Validating the Preschooler Task Using the Group Data

The preschooler component processes task as a predictor oflanguage comprehension. We completed a series of regressionanalyses in order to determine how much variance in performanceon a measure of language comprehension could be accounted forby the preschooler component processes task as a whole. Specif-ically, we first (a) used regressions to determine which complexcomponents (i.e., low-, medium-, and high-knowledge integration)and less-complex components (i.e., text memory, text inferencing,low-knowledge access-1, and low-knowledge access-2) accountedfor significant amounts of unique variance in language compre-hension performance and then (b) completed a single fixed-order

regression analysis that allowed the complex components to enterinto the regression model first and the less-complex componentssecond. In this latter model, complex components were allowed toenter first because they draw on the less-complex components.

To determine which complex and less-complex componentsaccounted for significant amounts of unique variance in languagecomprehension performance, we completed two preliminary step-wise regression analyses. The first regression allowed all thecomplex components to enter into the model freely, whereas thesecond regression allowed all the less-complex components toenter freely. The results of the first preliminary regression revealedthat the low- and medium-knowledge integration components ac-counted for the largest amounts of unique significant variance inlanguage comprehension performance, but the high-knowledgeintegration component did not contribute significantly after low-and medium-knowledge integration were controlled. The results ofthe second preliminary regression revealed that the text memoryand low-knowledge access-2 components accounted for the largestamounts of unique significant variance in language comprehensionperformance, but text inferencing and low-knowledge access-1component did not once the other less complex predictors werecontrolled. Because high-knowledge integration, text inferencing,

4 Note that the term text-based processing means processes that operateprimarily on the text; it should not be confused with the term textbaserepresentation, which is a component of the construction-integrationmodel.

Table 5Factor Loadings for Factor Analysis on Higher Level Cognitive Processes and Factor Analysison Higher Level Cognitive Processes and Lower Level Letter-Word Decoding/IdentificationSkills (n � 73)

Variable Factor 1 Factor 2 Factor 3

Factor loadings for measures of higher level cognitive processes (rotated pattern)

Text memory .77112 .29667Text inferencing .84309 �.11693Low-knowledge integration .33804 .64141Medium-knowledge integration .74725 .38360High-knowledge integration .70058 .35961Low-knowledge access .02422 .84260High-knowledge access .21883 .75973

Eigenvalue (% variance forfactor)

3.39 (48.45%) 1.20 (17.19%)

Factor loadings for measures of higher level cognitive processes and lower level letter-worddecoding/identification skills (rotated pattern)

Text memory .79332 .22142 .06280Text inferencing .83671 �.13858 .14598Low-knowledge integration .39586 .63037 �.05552Medium-knowledge integration .68531 .35575 .27871High-knowledge integration .68896 .33436 .11919Low-know access-1 .04991 .82587 .04973Low-know access-2 .16290 .76497 .20658Letter-word identification .03233 .12388 .93045Letter-word attack .21555 .04468 .84620

Eigenvalue (% variance forfactor)

3.76 (41.80%) 1.42 (15.73%) 1.17 (13.03%)

Note. Factor loadings that are italicized are significant. Factor loadings that are bolded and italicized representthe major components.

12 HANNON AND FRIAS

Fn4


and low-knowledge access-1 did not account for significantamounts of unique variance in language comprehension perfor-mance, they were excluded as predictors in the final regression.

For the final regression model medium-knowledge integration,low-knowledge integration, text memory, and low-knowledge-access-2 were entered into the model in a fixed order such that thetwo complex components, medium- and low-knowledge integra-tion, entered into the model first in order of the proportion ofunique variance explained in the preliminary regression analysisand the two less-complex components, text memory and low-knowledge access-2, entered second likewise in order of varianceexplained. This regression analysis is the first one depicted inTable 6.

As this regression analysis in Table 6 shows, three components(medium-knowledge integration, low-knowledge integration, andtext memory) accounted for a significant amount of variance inperformance on the measure of language comprehension. Thefourth component (low-knowledge access-2) was only marginallysignificant in the final model (p � .06). Combined, the threecomponents accounted for 46.6% of the total variance (r � .683).This finding suggests that multiple higher level processes accountfor performance on measures of language comprehension, namely,knowledge integration, text memory, knowledge access. This find-ing also suggests that the new component processes task is a verygood measure for predicting performance on measures of globallanguage comprehension performance in preschoolers.5

The preschooler component processes task versus age aspredictors of language comprehension. We completed tworegression analyses to determine the relative influence of ageversus higher level processes for predicting language comprehen-sion in preschoolers. In the first regression, age entered into theequation first and the components of the pre-reader task second. Inthe second regression the components entered first and age second.

As the second regression analysis in Table 6 shows, whenentered first, age accounted for 19.9% of the variance in languagecomprehension ability and the components of the preschooler taskaccounted for an additional 30.9% unique variance. However, asthe third regression analysis in Table 6 shows, when age wasentered as a predictor after the components of the preschooler taskwere entered, age accounted for only 4.2% unique variance inlanguage comprehension. The results of this latter regression sug-gest that higher level cognitive processes underlie age-relatedchanges in language comprehension in preschoolers. Looking atthe second and third regression analyses in Table 6, it appears thatage overlaps the most with the medium-knowledge integrationcomponent, indeed the amount of variance attributed to medium-knowledge integration reduced to 14.8% (see the second regres-sion analysis in Table 6) from 28.9% (see the third regressionanalysis in Table 6).

The preschooler task versus working memory as predictorsof language comprehension. We completed two regressionanalyses in order to compare the power of the components of thepreschooler task versus a measure of working memory to predictperformance on a measure of language comprehension. One re-gression allowed the components of the preschooler task (i.e., thesignificant predictors that were identified earlier) to enter into themodel first and the measure of working memory second. The otherregression allowed the measure of working memory to enter intothe model first and the components of the preschooler task second.

The results of the two regressions are the fourth and fifth onesdepicted in Table 6.

When combined, the preschooler task and working memorymeasure accounted for a considerable amount of variance in lan-guage comprehension performance. Indeed, as the fourth and fifthregression analyses in Table 6 show, together they accounted for52.3% of the total variance. Approximately 22.7% of this variancewas because of the shared overlap between the components of thepreschooler task and the measure of working memory (i.e.,28.4% � 5.7% � 22.7%), an additional 5.7% unique variance wasaccounted for by the working memory measure, and the remaining23.9% unique variance was accounted for by the components ofthe preschooler task. These latter two findings suggest that thepreschooler task accounts for more unique variance in languagecomprehension performance than does a measure of workingmemory (23.9% vs. 5.7%, respectively; i.e., incremental validity).

The preschooler task as a predictor of fluid intelligence. Inthis analysis we examined whether the knowledge integrationcomponents of the preschooler task might also account for perfor-mance on a composite measure of fluid intelligence. A second goalwas to compare the predictive powers of the knowledge integrationcomponents with that of working memory, another very goodpredictor of fluid intelligence in young children (e.g., de Ribaupi-erre & Lecerf, 2006; Engel de Abreu et al., 2010; Swanson, 2008).

We first examined the correlations among the measures of fluidintelligence, the components of the preschooler task, and workingmemory. As Table 4 shows, there were significant correlationsbetween fluid intelligence and working memory (r � .37) andbetween fluid intelligence and low-, medium-, and high-knowledge integration (rs � .24 to .44). However, of the threecorrelations between fluid intelligence and low-, medium-, andhigh-knowledge integration, the fluid intelligence–medium-knowledge integration correlation was the largest (r � .44). Be-cause medium-knowledge integration was the best predictor offluid intelligence, it was included in the subsequent regressionanalyses that assessed both the predictive power of the preschoolertask and the relative predictive powers of the preschooler taskversus working memory.

Two regression analyses were completed to compare the relativepowers of the measures of medium-knowledge integration versusworking memory to predict performance on the composite mea-sure of fluid intelligence. The first regression allowed medium-knowledge integration to enter into the model first and workingmemory second, while the second regression allowed workingmemory to enter first and medium-knowledge integration second.These two regression analyses are the sixth and seventh onesreported in Table 6.

As the sixth and seventh regression analyses show, when com-bined, medium-knowledge integration and working memory ac-

5 We also completed a similar regression analysis by using compositescores for text processing (sum of text memory, text inferencing), knowl-edge access (sum of two knowledge access measures) and knowledgeintegration (sum of three knowledge integration measures) as predictors.To create these composite scores we followed the same procedure that weused to create the composite measure of fluid intelligence. The resultsrevealed that when all three of these composite measures were entered intoa regression, they accounted for 41.6% of the variance in performance onour measure of language comprehension.


T6

Fn5


counted for 23.5% of the variance in fluid intelligence; approxi-mately 9.8% was accounted for by the shared overlap betweenmedium-knowledge integration and working memory (i.e., 14.0 –4.2 � 9.8%), an additional 4.2% unique variance was accountedfor by working memory, and an additional 9.5% unique variancewas accounted for by medium-knowledge integration. These find-ings support previous adult research that advocates working mem-ory and knowledge integration as predictors of fluid intelligence.The new finding here is that this is the first time these results havebeen observed with preschoolers.

Factor analysis of higher level processes and lower levelletter-word decoding/identification skills. According to devel-opmental reading studies, measures of higher level processes andletter-word decoding/identification are independent (e.g., August etal., 2006; Oakhill et al., 2003). The goal of this analysis was todetermine whether this same pattern exists when the target populationis preschoolers. To achieve this goal, we completed a factor analysisthat included the components of the preschooler task and the two

measures of letter-word decoding/identification skills. If higher levelprocesses are independent of letter-word decoding/identification skillsthen we should observe three factors: two factors representing thehigher level components (text-based processing and knowledge-access factors identical to the ones reported earlier in the descriptivestatistics) and a third factor representing lower level letter-word de-coding/identification skills. The second analysis in Table 5 shows theresults of this factor analysis.

As the second factor analysis in Table 5 shows, there were threefactors that accounted for 70.56% of the variance. Consistent withthe first factor analysis in Table 5, there were two factors thatrepresented the higher level components; one factor was stronglyrelated to text-based processing, while the second factor wasstrongly related to knowledge access. Specifically, the two types oftext-based statements, text memory and text inferencing, loadedheavily on factor one but not on factor two (loadings �.69000 vs.�.30000, respectively); the two types of knowledge-access state-ments, low-knowledge access-1 and low-knowledge access-2,

Table 6Regression Analyses of Language Comprehension and Fluid Intelligence With the Componentsof the Preschooler Task, Working Memory, and Age as Predictors (n � 73)

Variable R R2 R2 F

Components of preschooler task as predictors of language comprehension

1. Medium-knowledge integration .538 .289 .289 28.86�

2. Low-knowledge integration .608 .370 .081 8.94�

3. Text memory .683 .466 .096 12.50�

4. Low-knowledge access-2 .702 .493 .027 3.66��

Age then components of preschooler task as predictors of language comprehension

1. Age .446 .199 .199 17.58�



4. Text memory .713 .508 .095 13.22�

Components of preschooler task then age as predictors of language comprehension



3. Text memory .683 .466 .096 12.50�

4. Age .713 .508 .042 5.79�

Components of preschooler task then working memory as predictors of language comprehension



3. Text memory .683 .466 .096 12.50�

4. Working memory .723 .523 .057 8.14�

Working memory then components of preschooler task as predictors of language comprehension

1. Working Memory .533 .284 .284 28.21�

2. Medium-Knowledge Integration .633 .401 .117 13.64�

3. Low-Knowledge Integration .687 .472 .071 9.21�

4. Text Memory .723 .523 .051 7.38�

Medium-knowledge integration then working memory as predictors of fluid intelligence


2. Working Memory .485 .235 .042 3.86��

Working memory then medium-knowledge integration as predictors of fluid intelligence

1. Working Memory .374 .140 .140 11.55�


� p � .05. �� p � .05.

14 HANNON AND FRIAS


loaded heavily on factor two but not on factor one (loadings�.69000 vs. �.30000, respectively); and the three types ofknowledge-integration statements, low-, medium-, and high-knowledge integration, loaded on both the text-based factor (i.e.,factor one) and the knowledge access factor (i.e., Factor 2).

As the second factor analysis in Table 5 shows, however, noneof the components of the preschooler task loaded on the thirdfactor (all loadings �.30000). On the other hand, the measures ofletter-word decoding/identification skills, letter-word identifica-tion and letter-word attack, loaded heavily on factor three, but noton factors one and two (loadings �.69000 vs. �.30000, respec-tively). This latter finding suggests that lower level letter-worddecoding/identification skills are independent of higher level pro-cesses in preschoolers. In other words, this factor analysis providessome discriminant validity for the preschooler task.

The Preschooler Task as a Predictor of LanguageComprehension in Pre-Readers

This analysis determined how well the preschooler task ac-counted for language comprehension performance in a subset ofchildren who were unable to phonologically decode words oridentify words by sight (i.e., pre-readers). In order to achieve thisgoal we completed a series of regression analyses identical to aseries of regression analyses completed with the measure of lan-guage comprehension using the group data.

The results of the first preliminary regression, which assessedthe relative predictive powers of low-, medium-, and high-knowledge integration, revealed that the low- and medium-knowledge integration components accounted for significantamounts of unique variance in language comprehension perfor-mance, but the high-knowledge integration component did not.The results of the second preliminary regression, which assessedthe relative predictive powers of text memory, text inferencing,low-knowledge access-1, and low-knowledge access-2, revealedthat the text memory and low-knowledge access-2 componentsaccounted for significant amounts of unique variance in languagecomprehension performance, but text inferencing and low-knowledge access-1 component did not. Because high-knowledgeintegration, text inferencing, and low-knowledge access-1 did notaccount for significant amounts of unique variance in languagecomprehension performance, they were not included in the finalregression model.

In the final regression, medium-knowledge integration, low-knowledge integration, text memory, and low-knowledge-access-2were entered into the model in a fixed order such that the twocomplex components, medium- and low-knowledge integration,entered into the model first and the two less-complex components,text memory and low-knowledge access-2, entered second. Table7 shows the results of this regression analysis.

As Table 7 shows, all four components—medium-knowledgeintegration, low-knowledge integration, text memory, and low-knowledge access-2—accounted for a significant amount of vari-ance in performance on the measure of language comprehension.Combined, the four components in this regression analysis ac-counted for 64.1% of the total variance (r � .801). This finding isboth similar and different from the results of the regression anal-ysis predicting language comprehension for the group data. Thisfinding is similar inasmuch as the same four components (e.g.,

medium-knowledge integration, low-knowledge integration, textmemory, and low-knowledge access-2) emerged as significantpredictors of language comprehension in both sets of preliminaryregression analyses (i.e., the preliminary analysis for the groupdata and the preliminary analysis for the pre-readers). As well,both final regression analyses accounted for a significant amountof the total variance in performance on the measure of languagecomprehension. On the other hand, whereas three componentsaccounted for unique variance in language comprehension perfor-mance in the final regression analysis for the group data (i.e.,medium-knowledge integration, low-knowledge integration, andtext memory), all four components accounted for unique variancein language comprehension performance for the pre-readers. Al-though the source of this difference between the two regressionanalyses is beyond the scope of the present study, the results ofthese regression analyses do suggest that the new componentprocesses task is a suitable measure for predicting performance ona measure of language comprehension in pre-readers.

Discussion

The present study validates a new measure that provides esti-mates of a preschooler’s ability to recall text, to make text-basedinferences, to access knowledge from long-term memory, and tointegrate this accessed knowledge with new information from thetext. The results showed that the new preschooler task is suitablefor understanding the contributions of higher level processes toperformance on a measure of language comprehension. Indeed, theresults suggest that the preschooler task is a better predictor oflanguage comprehension than it is a measure of working memory.In addition, the medium-knowledge integration component was aspredictive of fluid intelligence as was a measure of workingmemory, one of the best predictors of fluid intelligence.

Skeptics might argue that there is nothing particularly novelabout our findings that the preschooler task predicts performanceon a standardized measure of language comprehension. Indeed, afew researchers have shown that higher level processes account forperformance on measures of language comprehension in pre-schoolers (e.g., Florit et al., 2011; Kendeou et al., 2008). From thisperspective, our findings are not that surprising. However, toconstrue the present findings as not theoretically informative is thesame as arguing against both componential analysis in general anda componential analysis interpretation of the contributions of mul-tiple, factorially separate higher level processes to language com-prehension, especially when this fact is acknowledged in the adult,adolescent, and developmental research (e.g., August et al., 2006;

Table 7Pre-Readers: Regression Analyses of Language ComprehensionWith the Components of the Preschooler Task as Predictors (n� 36)

Variable R R2 R2 F



3. Text memory .753 .567 .083 6.40�

4. Low-knowledge access-2 .801 .641 .074 6.10�

� p � .05.


T7


Barnes et al., 1996; Cain et al., 2004; Cromley & Azevedo, 2007;Hannon & Daneman, 2001). The present findings extend thisresearch by observing a similar pattern using a target population ofpreschoolers. The present findings also show, for the first time, therelative contributions of higher level processes versus workingmemory to language comprehension performance and fluid intel-ligence in preschoolers. Finally, the present findings show, for thefirst time, the relationships between higher level and lower levelprocesses in preschoolers, relationships that have theoretical im-plications for the construction-integration model of comprehension(Kintsch, 1988, 1998).

Predicting Language Comprehension Ability inPreschoolers and Pre-Readers

The present study revealed some novel findings with respect topredicting language comprehension ability. First, the preschoolertask accounted for a considerable amount of variance in languagecomprehension ability in both preschoolers and pre-readers. Spe-cifically, three components—medium-knowledge integration, low-knowledge integration, and text memory—accounted for morethan 45% of the variance in language comprehension ability inboth groups of children. From a practical perspective, this findingis important because it identifies the higher level processes thatcontribute to language comprehension before a child begins toread. Given that language comprehension is a strong predictor ofreading comprehension during early reading development (e.g.,Kendeou, van den Broek, et al., 2009; Protopapas et al., in press),and given that the present study shows that when combined knowl-edge integration, text memory, and knowledge access account fora large amount of variance in language comprehension perfor-mance in preschoolers (i.e., �40% of the total variance), it is quitepossible that these higher level processes are also highly predictiveof future reading ability. For this reason the preschooler taskshould be of interest to researchers, educators, and parents becauseit has the potential to be an assessment tool that will identify bothstrengths and weaknesses in higher level cognitive processes. Italso might lead to intervention programs that foster higher levelcognitive processes before a child actually begins to read (i.e.,early detection 3 early intervention).

Second, the results provide a plausible explanation for age-related changes in language comprehension in preschoolers. Spe-cifically, when the components of the preschooler task were en-tered into the regression model first and age second, the predictivepower of age reduced from 19.9% to 4.2% of the total variance inlanguage comprehension ability. This finding suggests that age-related differences in text memory, knowledge integration, andknowledge access account for slightly more than 78% of theage-related changes in language comprehension ability in pre-schoolers. It is also consistent with the results of Barnes et al.(1996), who showed similar results with knowledge integration.

Third, the results reveal the contributions of higher level com-ponents versus working memory to language comprehension abil-ity. Most researchers would agree that working memory is one ofthe best, if not the best, predictor of reading or language compre-hension in pre-readers, beginning readers, and adults (e.g., Dane-man & Merikle, 1996). Indeed, our results support this claiminasmuch as our new listening span measure of working memoryaccounted for more than 28.4% of the variance in preschooler

language comprehension ability. However the components of thepreschooler task also accounted for more than 48.0% of the vari-ance. In other words, the higher level components accounted forapproximately 65% more variance in language comprehensionability in preschoolers than did the listening span measure ofworking memory. Taken together, these findings suggest that thepreschooler component processes task is a better predictor oflanguage comprehension ability than a listening span measure ofworking memory. Fourth, consistent with previous developmentaland adult research (e.g., Britton et al., 1998; Daneman & Hannon,2007; Oakhill et al., 2003), our results showed that almost all of thevariance accounted for by the measure of working memory isshared with the measures of higher level component processes(i.e., 22.7% of the 28.4%). The new finding here is that this is thefirst time this result has been observed with a population ofpreschoolers and listening-/language-based measures of workingmemory and comprehension.

Finally, the present findings support theories of comprehensionthat suggest that higher level processes play an important role incomprehension. For example, according to the construction-integration model, the interplay between lower and higher levelprocesses results in the formation of a textbase representation(e.g., Kintsch, 1988). The construction-integration model also pos-its that the higher level process of knowledge integration is used tosupplement the text-based information with information from priorknowledge in order to achieve a personalized interpretation of atext that relates to other information held in our long-term memory(e.g., Kintsch, 1998). Consistent with both of these tenants, thepresent findings show higher level processes are an important partof language comprehension in preschoolers.

Relationships Between Reading and Listening

The findings of the present study are also analogous to a numberof previous findings assessing higher level processes and reading/language comprehension ability in preschoolers, beginning read-ers, and adults. For example, the present study extends previouspreschooler research by developing a preschooler language-basedpsychometric measure of higher level cognitive processes that ishighly predictive of language comprehension ability. Further, thepresent results replicate previous developmental research thatshows a reading-based measure of higher level processes that isanalogous to the preschooler task is predictive of listening com-prehension ability in beginning readers (e.g., August et al., 2006).Finally, although the preschooler task is language-based, our re-sults closely parallel those of earlier developmental and adultreading research, which shows that reading-based measures ofhigher level processes are highly predictive of reading comprehen-sion ability (e.g., August et al., 2006; Barnes et al., 1996; Cain etal., 2004; Hannon, 2012; Hannon & Daneman, 1998, 2001, 2006,2009; Oakhill, 1982, 1984).

Nevertheless, the results of previous and present studies alsoreveal at least one major difference between reading and listeningcomprehension in preschoolers. This difference is that measures oflanguage comprehension ability and letter-word decoding/identification skills are both highly predictive of reading compre-hension ability during the early stages of reading development(e.g., Kendeou, van den Broek, et al., 2009). Thus, even though thepresent study showed that measures of higher level processes are

16 HANNON AND FRIAS

AQ: 7


highly predictive of measures of language comprehension abilityand Kendeou, van den Broek, et al. (2009) showed that measuresof language comprehension were highly predictive of readingcomprehension ability, clearly measures of higher level processesare not the only predictors of reading comprehension ability.

Three Clusters of Abilities Rather Than Two

As noted by Kendeou, van den Broek, et al. (2009), mostresearchers agree that two separate clusters of abilities contributeto a considerable amount of variance in reading comprehension;one cluster of abilities translates symbols on a page into meaning-ful sounds and words (i.e., word decoding; e.g., Cain et al., 2004;Catts et al., 2003; Curtis, 1980; Oakhill et al., 2003; Perfetti, 1985),while the other cluster of abilities extracts meaningful informationfrom a text, accesses prior knowledge from long-term memory,and bridges text-based ideas (Perfetti, Landi, & Oakhill, 2005).Although we did not include a measure of reading comprehensionability in the present study, we did include measures that assessedthese two separate clusters of abilities. Therefore, we were able toassess the independent/interdependence of these two clusters ofabilities in preschoolers.

Consistent with previous preschooler and development research,our factor analysis revealed clusters of abilities. However, ratherthan the two expected clusters of abilities, we observed three. Onecluster was related to lower level letter-word decoding/identification skills, while the other two clusters were related tohigher level component processes, namely, one with text-basedprocessing and the other with accessing prior knowledge. The mostcogent evidence for this claim was the factor analysis, whichrevealed three distinct clusters. To the best of our knowledge thisis the first time research has revealed three distinct clusters ofabilities in preschoolers in a single study.

Although this finding was not predicted, there is some evidencein the literature that these measures would form three distinctclusters of abilities if they were simultaneously included in a factoranalysis. With respect to a separation between lower level letter-word decoding/identification skills and higher level components,August et al. (2006) showed that measures assessing letter-worddecoding/identification skills were, at best, weakly related toreading-based measures assessing the higher level processes oftext memory, text inferencing, knowledge access, and knowledgeintegration in a population of beginning readers. Similarly, Han-non (2012) observed, at best, weak correlations between measuresassessing lower level word processes and reading-based measuresof higher level processes in an adult population.

With respect to two factors representing higher level compre-hension processes as opposed to one, Barnes et al. (1996) showedthat reading-based measures of knowledge accessibility andknowledge-integration were separate higher level processes inchildren 6 to 15 years of age. Potts and Peterson (1985) showedthat for adults, reading-based measures of text memory and textinferencing formed a text-processing factor but were, at best,weakly related to a measure of knowledge access. Finally, usingstructural equation modeling, Hannon and Daneman (2006) con-firmed that reading-based measures of text memory and text in-ferencing formed a text processing factor that was separate from afactor formed from measures of knowledge access.

Nevertheless, one might argue that the reason why our measuresof lower level letter-word decoding/identification form a cluster ofabilities that is separate from the two clusters of abilities formed bythe measures of higher level processes is because the performanceson the lower level letter-word measures are at floor. However, ifthe letter-word measures were indeed at floor, then one would alsoexpect both letter-word tasks to be heavily skewed and, at best,weakly related to one another, which clearly did not happen in thepresent study. Thus, it appears that floor performance on theletter-word tasks cannot explain the results of our factor analysis.

Predicting Fluid Intelligence

Consistent with developmental studies (e.g., Engel de Abreu etal., 2010), the present study shows that a listening span measure ofworking memory accounted for 14.0% of the variance in a non-verbal measure of fluid intelligence. However, also consistent withthe adult research, the present study shows that the medium-knowledge integration component accounted for 19.3% of thevariance (e.g., Buehner, Krumm, & Pick, 2005; Hannon & Dane-man, 2008; Oberauer et al., 2003, 2008). This latter finding sug-gests that the verbal components of the preschooler task aresuitable for predicting nonverbal tasks, such as those used to assessfluid intelligence. In other words, the preschooler componentprocesses task has predictive validity for a nonverbal task.

Additional Findings and Other Contributions

There are also other findings that are not the primary focus ofthe present study but still warrant discussion. For example, howwere the preschoolers able to process the high-knowledge integra-tion statements adequately given the complexity of the syntax ofthe high-knowledge integration statements and given that pre-schoolers have difficulty processing complex syntax? Althoughthere are a number of possible explanations, one explanation is thatperhaps the concreteness of the pictures that accompanied theauditory text of the short paragraph helped the preschoolers over-come the complexity of the syntax of the high-knowledge integra-tion statements. For example, perhaps their memories of the pic-ture depicting “JIMPS with long legs” made it easier for them toreconcile this fact with the prior knowledge fact that “GIRAFFEShave long legs” in order to answer the high-knowledge integrationstatement, “Like GIRAFFES, JIMPS have long legs.” Thus, oneavenue for future research might be to determine the extent towhich pictures might scaffold difficult syntax for preschoolers andbeginning readers.

A second contribution of the present study was the developmentof a measure of working memory that is suitable for preschoolers.As mentioned in the introduction, the earlier versions of pre-schooler working memory measures were criticized because pre-schoolers recalled the full sentences, rather than the sentences’final words. However, by creating a cloze listening span task, thepresent study showed that a working memory measure that in-volved recalling only a single word can be developed for pre-schoolers and that such a measure can account for a significantamount of variance in language comprehension ability.

Limitations

Although the results reveal a number of novel findings, there areother predictors, which were not included in the present study, that



also might account for unique variance in language comprehen-sion. For instance, absent from the present study were measures ofprior knowledge, learning strategies, and vocabulary, three knownpredictors of reading and listening performance in adult readers(Britton et al., 1998). Although these absences do not invalidatethe present results, one avenue for future research might be toexamine the relationships among these constructs, higher levelprocesses, and language comprehension in preschoolers.

Another limitation of the present study is its methodology. Thepresent study used a psychometric approach to examine the con-tributions of higher level processes to language comprehensionability. However, other methodologies could have been employed.For example, Konold, Juel, McKinnon, and Deffes (2003) usedcluster analysis to identify different cognitive profiles that wereempirically linked to children’s reading acquisition. Yet anotherlimitation is our use of one language comprehension measure.Although we chose an age-appropriate version of the Gates-MacGinitie, which is a very good measure, future research mightwish to examine whether the present findings generalize to othermeasures of language comprehension, for example narrative com-prehension, television programs, or videos. Finally, the presentstudy examined only language comprehension ability. Future re-search might wish to examine the relative contributions of letter-word decoding/identification skills versus higher level processes topredict the future reading comprehension abilities of preschoolersand pre-readers.

Conclusions

In conclusion, the present study validated a multicomponentmeasure of higher level processes for preschoolers. The resultsshow that the new preschooler task is suitable for both predictinglanguage comprehension performance and understanding thehigher level processes that underlie language comprehension inboth a group of preschoolers who are, at best, capable of decoding/identifying a few words and a group of pre-readers who areincapable of decoding or identifying words. Indeed, the compo-nents of the preschooler task are better at explaining languagecomprehension ability than is a measure of working memory.Further, the present study revealed that the medium-knowledgeintegration component predicted performance on a composite mea-sure of fluid intelligence. Finally, although it was not a goal of thepresent study, the results revealed three clusters of abilities: onecluster that was related to letter-word decoding/identification skillsand two clusters that were related to higher level cognitive pro-cesses (i.e., text-based processing and knowledge-based process-ing). The present findings, however, are limited to the measuresused in the present study, and consequently, future research mightwish to determine whether the present findings generalize to othermeasures. In addition, future research might wish to explore therelationships among lower level letter-word processes, higher levelprocesses, and working memory using other methodology besidesa psychometric approach. Finally, future research should explorethe relations among the three clusters of abilities using differentpopulations, such as beginning readers, adolescents, and olderadults.

References

Aaron, P. G., Franz, S. S., & Manges, A. R. (1990). Dissociation betweencomprehension and pronunciation in dyslexic and hyperlexic children.Reading and Writing: An Interdisciplinary Journal, 2, 243–264. doi:10.1007/BF00257974

Ackerman, P. L., Beier, M. E., & Boyle, M. O. (2005). Working memoryand intelligence: Same or different constructs? Psychological Bulletin,131, 30–60. doi:10.1037/0033-2909.131.1.30

Adams, A., Bourke, L., & Willis, C. (1999). Working memory and spokenlanguage comprehension in young children. International Journal ofPsychology, 34, 364–373. doi:10.1080/002075999399701

Adams, A., & Willis, C. (2001). Language processing and working mem-ory: A developmental perspective. In J. Andrade (Ed.), Working memoryin perspective (pp. 79–100). New York, NY: Taylor & Francis.

Adlof, S. M., Catts, H. W., & Little, T. D. (2006). Should the simple viewof reading include a fluency component? Reading and Writing, 19,933–958. doi:10.1007/s11145-006-9024-z

Aouad, J., & Savage, R. (2009). Further evidence for the simple view ofreading. Canadian Journal of School Psychology, 24, 183–200. doi:10.1177/0829573509336280

August, D., Francis, D., Hsu, H., & Snow, C. (2006). Assessing readingcomprehension in bilinguals. The Elementary School Journal, 107, 221–238. doi:10.1086/510656

Barnes, M., Dennis, M., & Haefele-Kalvaitis, J. (1996). The effects of knowledgeavailability and knowledge accessibility on coherence and elaborative inferenc-ing in children from six to fifteen years of age. Journal of Experimental ChildPsychology, 61, 216–241. doi:10.1006/jecp.1996.0015

Bishop, D. V. M., & Adams, C. (1990). A prospective study of therelationship between specific language impairment, phonological disor-ders and reading retardation. Journal of Child Psychology and Psychi-atry, 31, 1027–1050. doi:10.1111/j.1469-7610.1990.tb00844.x

Britton, B., Stimson, M., Stennett, B., & Gulgoz, S. (1998). Learning frominstructional text: Test of an individual-differences model. Journal ofEducational Psychology, 90, 476 – 491. doi:10.1037/0022-0663.90.3.476

Buehner, M., Krumm, S., & Pick, M. (2005). Reasoning � workingmemory attention. Intelligence, 33, 251–272. doi:10.1016/j.intell.2005.01.002

Cain, K., Oakhill, J., & Bryant, P. (2004). Children’s reading comprehen-sion ability: Concurrent prediction by working memory, verbal ability,and component skills. Journal of Educational Psychology, 96, 31–42.doi:10.1037/0022-0663.96.1.31

Carver, R. P. (1998). Predicting reading level in Grades 1 to 6 fromlistening level and decoding level: Testing theory relevant to simpleview of reading. Reading and Writing, 10, 121–154. doi:10.1023/A:1007923124312

Casteel, M. (1993). Effects of inference necessity and reading goal onchildren’s inferential generation skills. Developmental Psychology, 29,346–357. doi:10.1037/0012-1649.29.2.346

Casteel, M., & Simpson, G. B. (1991). Textual coherence and the devel-opment of inferential generation skills. Journal of Research in Reading,14, 116–129. doi:10.1111/j.1467-9817.1991.tb00013.x

Catts, H. W., Adlof, S. M., & Weismer, S. E. (2006). Language deficits inpoor comprehenders: A case for the simple view of reading. Journal ofSpeech, Language, and Hearing Research, 49, 278–293. doi:10.1044/1092-4388(2006/023)

Catts, H. W., Fey, M. E., Zhang, X., & Tomblin, B. J. (1999). Languagebasis of reading and reading disabilities: Evidence from a longitudinalinvestigation. Scientific Studies of Reading, 3, 331–361. doi:10.1207/s1532799xssr0304_2

Catts, H. W., Hogan, T. P., & Fey, M. E. (2003). Subgrouping poor readers on thebasis of individual differences in reading-related abilities. Journal of LearningDisabilities, 36, 151–164. doi:10.1177/002221940303600208

Crain, S. (1989). Why poor readers misunderstand spoken sentences. In D.

18 HANNON AND FRIAS

AQ: 8


http://dx.doi.org/10.1007/BF00257974

http://dx.doi.org/10.1007/BF00257974

http://dx.doi.org/10.1037/0033-2909.131.1.30

http://dx.doi.org/10.1080/002075999399701

http://dx.doi.org/10.1007/s11145-006-9024-z

http://dx.doi.org/10.1177/0829573509336280

http://dx.doi.org/10.1177/0829573509336280

http://dx.doi.org/10.1086/510656

http://dx.doi.org/10.1006/jecp.1996.0015

http://dx.doi.org/10.1111/j.1469-7610.1990.tb00844.x

http://dx.doi.org/10.1037/0022-0663.90.3.476

http://dx.doi.org/10.1037/0022-0663.90.3.476

http://dx.doi.org/10.1016/j.intell.2005.01.002


http://dx.doi.org/10.1037/0022-0663.96.1.31

http://dx.doi.org/10.1023/A:1007923124312

http://dx.doi.org/10.1023/A:1007923124312

http://dx.doi.org/10.1037/0012-1649.29.2.346


http://dx.doi.org/10.1044/1092-4388%282006/023

http://dx.doi.org/10.1044/1092-4388%282006/023

http://dx.doi.org/10.1207/s1532799xssr0304_2

http://dx.doi.org/10.1207/s1532799xssr0304_2

http://dx.doi.org/10.1177/002221940303600208

Shankweiler & I. Y. Liberman (Eds.), Phonology and reading disability:Solving the real puzzle (International Academy for Research in LearningDisabilities Monograph Series, No. 6, pp. 133–165). Ann Arbor, MI:University of Michigan Press.

Cromley, J. G., & Azevedo, R. (2007). Testing and refining the direct andinferential mediation model of reading comprehension. Journal of Ed-ucational Psychology, 99, 311–325. doi:10.1037/0022-0663.99.2.311

Curtis, M. E. (1980). Development of components of reading skill. Journalof Educational Psychology, 72, 656 – 669. doi:10.1037/0022-0663.72.5.656

Daneman, M., & Blennerhassett, A. (1984). How to assess the listeningcomprehension skills of pre-readers. Journal of Educational Psychology,76, 1372–1381. doi:10.1037/0022-0663.76.6.1372

Daneman, M., & Carpenter, P. A. (1980). Individual differences in workingmemory and reading. Journal of Verbal Learning & Verbal Behaviour,19, 450–466. doi:10.1016/S0022-5371(80)90312-6

Daneman, M., & Hannon, B. (2007). What do memory spans like readingspan really measure? In N. Osaka, R. H. Logie, & M. D’Espisito (Eds.),The cognitive neuroscience of working memory: Behavioral and neuralcorrelates (pp. 21–42). New York, NY: Oxford University Press. doi:10.1093/acprof:oso/9780198570394.003.0002

Daneman, M., & Merikle, P. M. (1996). Working memory and languagecomprehension: A meta-analysis. Psychonomic Bulletin & Review, 3,422–433. doi:10.3758/BF03214546

Deary, I. J. (2002). g and cognitive elements of information processing: Anagnostic view. In R. J. Sternberg & E. L. Grigorenko (Eds.), The general factorof intelligence: How general is it? (pp. 151–181). Mahwah, NJ: Erlbaum.

de Jong, P. F., & van der Leij, A. (2002). Effects of phonological abilitiesand linguistic comprehension on the development of reading. ScientificStudies of Reading, 6, 51–77. doi:10.1207/S1532799XSSR0601_03

de Ribaupierre, A., & Lecerf, T. (2006). Relationships between workingmemory and intelligence from a developmental perspective: Convergentevidence from a neo-Piagetian and a psychometric approach. EuropeanJournal of Cognitive Psychology, 18, 109 –137. doi:10.1080/09541440500216127

Detterman, D. K. (2002). General intelligence: Cognitive and biological ex-planations. In R. J. Sternberg & E. L. Grigorenko (Eds.), The general factorof intelligence: How general is it? (pp. 223–243). Mahwah, NJ: Erlbaum.

Ehri, L. C., Nunes, S. R., Stahl, S. A., & Willows, D. M. (2001). System-atic phonics instruction helps students learn to read: Evidence from theNational Reading Panel’s meta-analysis. Review of Educational Re-search, 71, 393–447. doi:10.3102/00346543071003393

Engel de Abreu, P. M. J., Conway, A. R. A., & Gathercole, S. E. (2010).Working memory and fluid intelligence in young children. Intelligence,38, 552–561. doi:10.1016/j.intell.2010.07.003

Engle, R. W., Tuholski, S. W., Laughlin, J. E., & Conway, A. R. A. (1999).Working memory, short-term memory, and general fluid intelligence: Alatent-variable approach. Journal of Experimental Psychology: General,128, 309–331. doi:10.1037/0096-3445.128.3.309

Florit, E., Roch, M., Altoe, G., & Levorato, M. C. (2009). Listeningcomprehension in preschoolers: The role of memory. British Journal ofDevelopmental Psychology, 27, 935–951. doi:10.1348/026151008X397189

Florit, E., Roch, M., & Levorato, M. C. (2011). Listening text comprehen-sion of explicit and implicit information in preschoolers: The role ofverbal and inferential skills. Discourse Processes, 48, 119–138. doi:10.1080/0163853X.2010.494244

Frith, U., & Snowling, M. (1983). Reading for meaning and reading forsound in autistic and dyslexic children. British Journal of DevelopmentalPsychology, 1, 329–342. doi:10.1111/j.2044-835X.1983.tb00906.x

Fry, A. F., & Hale, S. (2000). Relationships among processing speed,working memory, and fluid intelligence in children. Biological Psychol-ogy, 54, 1–34. doi:10.1016/S0301-0511(00)00051-X

Fuhriman, D. O. (2002). Literacy synthesis and professional development.Retrieved from http://www.fuhriman.us/la.assess/gates.html

Fullerton, A. M. (1983). Age differences in the use of imagery in integrating newand old information in memory. Journal of Gerontology, 38, 326–332.

Goldstein, G., & Hersen, M. (2000). Handbook of psychological assess-ment (3rd ed.). Oxford, England: Oxford University Press.

Hannon, B. (2012). Understanding the relative contributions of lower-levelword processes, higher-level processes, and working memory to readingcomprehension performance in proficient adult readers. Reading Re-search Quarterly, 47, 125–152.

Hannon, B., & Daneman, M. (1998). Facilitating knowledge-based infer-ences in less-skilled readers. Contemporary Educational Psychology,23, 149–172. doi:10.1006/ceps.1997.0968

Hannon, B., & Daneman, M. (2001). A new tool for measuring andunderstanding individual differences in the component processes ofreading comprehension. Journal of Educational Psychology, 93, 103–128. doi:10.1037/0022-0663.93.1.103

Hannon, B., & Daneman, M. (2006). What do tests of reading comprehen-sion ability such as the VSAT really measure? A componential analysis.In A. Mittel (Ed.), Focus on educational psychology (pp. 105–146).Hauppauge, NY: Nova Science.

Hannon, B., & Daneman, M. (2008, July). Individual differences in cog-nitive abilities: A component processes account. Poster presented atInternational Congress, Berlin, Germany.

Hannon, B., & Daneman, M. (2009). Age-related changes in readingcomprehension: An individual-difference perspective. Experimental Ag-ing Research, 35, 432–456. doi:10.1080/03610730903175808

Hannon, B., & McNaughton-Cassill, M. (2011). SAT performance: Un-derstanding the contributions of cognitive/learning and social/personality factors. Applied Cognitive Psychology, 25, 528–535. doi:10.1002/acp.1725

Hresko, W. P., Peak, P. K., Herron, S. R., & Bridges, D. L. (2006). Testreview: Young children’s achievement test. Journal of Psychoeduca-tional Assessment, 24, 272–277. doi:10.1177/0734282906286340

Kaplan, R. M., & Saccuzzo, D. P. (2001). Psychological testing: Princi-ples, applications, and issues. Belmont, CA: Thomson Wadsworth.

Kendeou, P., Bohn-Gettler, C., White, M. J., & van den Broek, P. (2008).Children’s inference generation across different media. Journal of Re-search in Reading, 31, 259–272. doi:10.1111/j.1467-9817.2008.00370.x

Kendeou, P., Lynch, J. S., van den Broek, P., Espin, C. A., White, M. J., &Kremer, K. E. (2005). Developing successful readers: Building early com-prehension skills through television viewing and listening. Early ChildhoodEducation Journal, 33, 91–98. doi:10.1007/s10643-005-0030-6

Kendeou, P., Savage, R., & van den Broek, P. (2009). Revisiting the simpleview of reading. British Journal of Educational Psychology, 79, 353–370. doi:10.1348/978185408X369020

Kendeou, P., van den Broek, P., White, M. J., & Lynch, J. S. (2009).Predicting reading comprehension in early elementary school: The in-dependent contributions of oral language and decoding skills. Journal ofEducational Psychology, 101, 765–778. doi:10.1037/a0015956

Kintsch, W. (1988). The role of knowledge in discourse comprehension: Aconstruction-integration model. Psychological Review, 1988, 95, 163–182. doi:10.1037/0033-295X.95.2.163

Kintsch, W. (1994). Text comprehension, memory, and learning. AmericanPsychologist, 49, 294–303. doi:10.1037/0003-066X.49.4.294

Kintsch, W. (1998). Comprehension: A paradigm for cognition. New York,NY: Cambridge University Press.

Konold, T. R., Juel, C., McKinnon, M., & Deffes, R. (2003). A multivariatemodel of early reading acquisition. Applied Psycholinguistics, 24, 89–112.

Kyllonen, P., & Christal, R. E. (1990). Reasoning ability is (little morethan) working memory capacity? Intelligence, 14, 389 – 433. doi:10.1016/S0160-2896(05)80012-1

Leather, C., & Henry, L. (1994). Working memory span and phonological


AQ: 10


http://dx.doi.org/10.1037/0022-0663.99.2.311

http://dx.doi.org/10.1037/0022-0663.72.5.656

http://dx.doi.org/10.1037/0022-0663.72.5.656

http://dx.doi.org/10.1037/0022-0663.76.6.1372

http://dx.doi.org/10.1016/S0022-5371%2880%2990312-6

http://dx.doi.org/10.1093/acprof:oso/9780198570394.003.0002

http://dx.doi.org/10.1093/acprof:oso/9780198570394.003.0002

http://dx.doi.org/10.3758/BF03214546

http://dx.doi.org/10.1207/S1532799XSSR0601_03

http://dx.doi.org/10.1080/09541440500216127

http://dx.doi.org/10.1080/09541440500216127

http://dx.doi.org/10.3102/00346543071003393


http://dx.doi.org/10.1037/0096-3445.128.3.309

http://dx.doi.org/10.1348/026151008X397189

http://dx.doi.org/10.1348/026151008X397189

http://dx.doi.org/10.1080/0163853X.2010.494244

http://dx.doi.org/10.1080/0163853X.2010.494244

http://dx.doi.org/10.1111/j.2044-835X.1983.tb00906.x

http://dx.doi.org/10.1016/S0301-0511%2800%2900051-X

http://www.fuhriman.us/la.assess/gates.html

http://dx.doi.org/10.1006/ceps.1997.0968

http://dx.doi.org/10.1037/0022-0663.93.1.103

http://dx.doi.org/10.1080/03610730903175808

http://dx.doi.org/10.1002/acp.1725

http://dx.doi.org/10.1002/acp.1725

http://dx.doi.org/10.1177/0734282906286340

http://dx.doi.org/10.1111/j.1467-9817.2008.00370.x

http://dx.doi.org/10.1007/s10643-005-0030-6

http://dx.doi.org/10.1348/978185408X369020

http://dx.doi.org/10.1037/a0015956

http://dx.doi.org/10.1037/0033-295X.95.2.163

http://dx.doi.org/10.1037/0003-066X.49.4.294

http://dx.doi.org/10.1016/S0160-2896%2805%2980012-1

http://dx.doi.org/10.1016/S0160-2896%2805%2980012-1

awareness tasks as predictors of early reading ability. Journal of Exper-imental Child Psychology, 58, 88–111. doi:10.1006/jecp.1994.1027

Long, D. L., & De Ley, L. (2000). Implicit causality and discourse focus:The interaction of text and reading characteristics in pronoun resolution.Journal of Memory and Language, 42, 545–570. doi:10.1006/jmla.1999.2695

Long, D. L., Oppy, B. J., & Seely, M. R. (1994). Individual differences inthe time course of inferential processing. Journal of Experimental Psy-chology: Learning, Memory, and Cognition, 20, 1456 –1470. doi:10.1037/0278-7393.20.6.1456

MacGinitie, W., & MacGinitie, R. (1989). Gates-MacGinitie ReadingTests (3rd ed.). The Reading Teacher, 43, 256–258.

MacGinitie, W., MacGinitie, R., Maria, K., & Dreyer, L. (2000). Gates-MacGinitie Reading Tests. Chicago, IL: Riverside.

Masson, M., & Miller, J. (1983). Working memory and individual differ-ences in comprehension and memory of text. Journal of EducationalPsychology, 75, 314–318. doi:10.1037/0022-0663.75.2.314

McCardle, P., Scarborough, H. S., & Catts, H. W. (2001). Predicting,explaining, and preventing children’s reading difficulties. Learning Dis-abilities Research & Practice, 16, 230 –239. doi:10.1111/0938-8982.00023

Oakhill, J. V. (1982). Constructive processes in skilled and less-skilledcomprehenders. British Journal of Psychology, 73, 13–20. doi:10.1111/j.2044-8295.1982.tb01785.x

Oakhill, J. V. (1984). Inferential and memory skills in children’s compre-hension of stories. British Journal of Educational Psychology, 54, 31–39. doi:10.1111/j.2044-8279.1984.tb00842.x

Oakhill, J. V., Cain, K., & Bryant, P. E. (2003). The dissociation of wordreading and text comprehension: Evidence from component skills. Lan-guage and Cognitive Processes, 18, 443– 468. doi:10.1080/01690960344000008

Oberauer, K., Suß, H.-M., Wilhelm, O., & Wittman, W. W. (2003). Themultiple faces of working memory: Storage, processing, supervision,and coordination. Intelligence, 31, 167–193. doi:10.1016/S0160-2896(02)00115-0

Oberauer, K., Suß, H.-M., Wilhelm, O., & Wittman, W. W. (2008). Whichworking memory functions predict intelligence? Intelligence, 36, 641–652. doi:10.1016/j.intell.2008.01.007

Paris, A. H., & Paris, S. G. (2003). Assessing narrative comprehension inyoung children. Reading Research Quarterly, 38, 36–76.

Perfetti, C. A. (1985). Reading ability. New York, NY: Oxford UniversityPress.

Perfetti, C. A., Landi, N., & Oakhill, J. (2005). The acquisition of readingcomprehension skill. In M. J. Snowling & C. Hulme (Eds.), The scienceof reading: A handbook (pp. 227–247). Malden, MA: Blackwell. doi:10.1002/9780470757642.ch13

Potts, G. R. (1977). Integrating new and old information. Journal of VerbalLearning & Verbal Behavior, 16, 305–320. doi:10.1016/S0022-5371(77)80053-4

Potts, G. R., & Peterson, S. B. (1985). Incorporation versus compartmen-talization in memory for discourse. Journal of Memory and Language,24, 107–118. doi:10.1016/0749-596X(85)90018-X

Protopapas, A., Simos, P. G., Sideridis, G. D., & Mouzaki, A. (in press).The components of the simple view of reading: A confirmatory factoranalysis. Reading Psychology.

Saarnio, D. A., Oka, E. R., & Paris, S. G. (1990). Developmental predictorsof children’s reading comprehension. In T. H. Carr & B. A. Levy (Eds.),Reading and its development: Component process approaches (pp.57–79). New York, NY: Academic Press.

Senechal, M. (1997). The differential effect of storybook reading on

preschoolers’ acquisition of expressive and receptive vocabulary. Jour-nal of Child Language, 24, 123–138. doi:10.1017/S0305000996003005

Senechal, M., & Cornell, E. H. (1993). Vocabulary acquisition throughshared reading experiences. Reading Research Quarterly, 28, 360–374.doi:10.2307/747933

Senechal, M., Thomas, E., & Monker, J. (1995). Individual differences in4-year-old children’s acquisition of vocabulary during storybook read-ing. Journal of Educational Psychology, 87, 218–229. doi:10.1037/0022-0663.87.2.218

Speece, D. L., Roth, F. P., Cooper, D. H., & de la Paz, S. (1999). Therelevance of oral language skills to early literacy: A multivariate anal-ysis. Applied Psycholinguistics, 20, 167–190. doi:10.1017/S0142716499002015

Spooner, A. L. R., Baddeley, A. D., & Gathercole, S. E. (2004). Canreading and comprehension be separated in the Neale Analysis ofReading Ability? British Journal of Educational Psychology, 74, 187–204. doi:10.1348/000709904773839833

Stark, R. E., Bernstein, L. E., Condino, R., Bender, M., Tallal, P., & Catts,H. (1984). Four-year follow-up study of language impaired children.Annals of Dyslexia, 34, 49–68. doi:10.1007/BF02663613

Storch, S., & Whitehurst, G. (2002). Oral language and code-relatedprecursors to reading: Evidence from a longitudinal structural model.Developmental Psychology, 38, 934 –947. doi:10.1037/0012-1649.38.6.934

Swanson, H. L. (2008). Working memory and intelligence in children:What develops? Journal of Educational Psychology, 100, 581–602.doi:10.1037/0022-0663.100.3.581

Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5thed.). New York, NY: Allyn & Bacon.

van den Broek, P. (1989). Causal reasoning and inference making injudging the importance of story statements. Child Development, 60,286–297. doi:10.2307/1130976

van den Broek, P., Kendeou, P., Kremer, K., Lynch, J., Butler, J., White,M. J., & Lorch, E. P. (2005). Assessment of comprehension abilities inyoung children. In S. Paris & S. Stahl (Eds.), Children’s readingcomprehension and assessment (pp. 107–130). Mahwah, NJ: Erlbaum.

van den Broek, P., Lorch, E. P., & Thurlow, R. (1996). Children’s andadults’ memory for television stories: The role of causal factors, story-grammar categories, and hierarchical level. Child Development, 67,3010–3028. doi:10.2307/1131764

Vellutino, F. R., Tunmer, W. E., Jaccard, J. J., & Chen, R. (2007).Components of reading ability: Multivariate evidence for a convergentskill model of reading development. Scientific Studies of Reading, 11,3–32.

Wechsler, D. (1989). Wechsler Preschool and Primary Scale ofIntelligence–Revised. San Antonio, TX: Psychological Corporation.

Wechsler, D. (1991). The Wechsler Intelligence Scale for Children (3rded.). San Antonio, TX: Psychological Corporation.

Wechsler, D. (2002). WPPSI-III administration and scoring manual. SanAntonio, TX: Psychological Corporation.

Woodcock, R. W. (1987). Woodcock Reading Mastery Test—Revised.Circle Pines, MN: American Guidance Services.

Woodcock, R. W., Mather, N., & McGrew, K. S. (2001). Woodcock-Johnson III Tests of Cognitive Abilities examiner’s manual. Itasca, MN:Riverside.

Woodcock, R. W., Shrank, F. A., McGrew, K. S., & Mather, N. (2006).Woodcock-Johnson III Normative Update (NU) Tests of Achievement,Forms A and B. Retrieved from http://riversidepublishing.com/products/wjIIINormativeUpdate/details.html

(Appendices follow)

20 HANNON AND FRIAS

AQ: 16


http://dx.doi.org/10.1006/jecp.1994.1027

http://dx.doi.org/10.1006/jmla.1999.2695

http://dx.doi.org/10.1006/jmla.1999.2695

http://dx.doi.org/10.1037/0278-7393.20.6.1456

http://dx.doi.org/10.1037/0278-7393.20.6.1456

http://dx.doi.org/10.1037/0022-0663.75.2.314

http://dx.doi.org/10.1111/0938-8982.00023

http://dx.doi.org/10.1111/0938-8982.00023




http://dx.doi.org/10.1080/01690960344000008

http://dx.doi.org/10.1080/01690960344000008

http://dx.doi.org/10.1016/S0160-2896%2802%2900115-0

http://dx.doi.org/10.1016/S0160-2896%2802%2900115-0


http://dx.doi.org/10.1002/9780470757642.ch13

http://dx.doi.org/10.1002/9780470757642.ch13

http://dx.doi.org/10.1016/S0022-5371%2877%2980053-4

http://dx.doi.org/10.1016/S0022-5371%2877%2980053-4

http://dx.doi.org/10.1016/0749-596X%2885%2990018-X

http://dx.doi.org/10.1017/S0305000996003005

http://dx.doi.org/10.2307/747933

http://dx.doi.org/10.1037/0022-0663.87.2.218

http://dx.doi.org/10.1037/0022-0663.87.2.218

http://dx.doi.org/10.1017/S0142716499002015

http://dx.doi.org/10.1017/S0142716499002015

http://dx.doi.org/10.1348/000709904773839833

http://dx.doi.org/10.1007/BF02663613

http://dx.doi.org/10.1037/0012-1649.38.6.934

http://dx.doi.org/10.1037/0012-1649.38.6.934

http://dx.doi.org/10.1037/0022-0663.100.3.581

http://dx.doi.org/10.2307/1130976

http://dx.doi.org/10.2307/1131764

http://riversidepublishing.com/products/wjIIINormativeUpdate/details.html

http://riversidepublishing.com/products/wjIIINormativeUpdate/details.html

Appendix A

Stimuli for the Preschooler Component Processes Task

Note that some types of test statement were not possible for someparagraphs. For instance, text inferencing statements are not possiblefor paragraphs with two real and one artificial object (Paragraphs 2and 3) because no inference can be made from the presented infor-

mation. As well, neither low-knowledge access nor low-knowledgeintegration statements are possible for paragraphs with one real andtwo artificial objects (Paragraphs 1, 4, and 5) because both types ofstatements require two real terms to be mentioned in the paragraph.

Paragraph 1: PracticeParagraph 2: Animal Paragraph in Table 1Paragraph 3 and Test Statements (Vehicles)

James was looking at pictures of vehicles. While looking at vehicles, James learned:

A NORT looks like a BIKE.

A NORT is larger than a PLANE.

Relationship: NORT � Plane � Bike

(Appendices continue)


AQ: 17


Appendix A (continued)

Text MemoryA NORT looks like a BIKE. A NORT looks like a PLANE. (n)A NORT is larger than a PLANE. A PLANE is larger than a NORT. (n)

Text Inferencing� not possible �

Low-knowledge integrationA NORT is larger than a BIKE. A BIKE is larger than a NORT. (n)

Medium-knowledge integrationA NORT is larger than a CAR. A CAR is larger than a NORT. (n)

High-knowledge integrationLike CARS, NORTS travel on the ground.Like ROCKETS, NORTS can fly in the air. (n)

Low-knowledge access Type IA PLANE is larger than a BIKE. A BIKE is larger than a PLANE. (n)

Low-knowledge access Type IIA PLANE is larger than a CAR. A CAR is larger than a PLANE. (n)A ROCKET is larger than a BIKE. A BIKE is larger than a PLANE. (n)

Paragraph 4 and Test Statements (Sea Life)

Mark was looking at sea life while swimming in the ocean.While looking at the sea life, Mark learned:

A LORK looks like a SHARK but a LORK is smaller.


22 HANNON AND FRIAS



A PARM looks like a LORK but a PARM is smaller.

Relationship: Shark � LORK � PARMText Memory

A LORK looks like a SHARK. A LORK does not look like a SHARK. (n)A PARM looks like a LORK. A PARM does not look like a LORK. (n)A LORK is smaller than a SHARK. A SHARK is smaller than a LORK. (n)A PARM is smaller than a LORK. A LORK is smaller than a PARM. (n)

Text InferencingA PARM looks like a SHARK. A PARM does not look like a SHARK. (n)A PARM is smaller than a SHARK. A SHARK is smaller than a PARM. (n)

Low-knowledge integration� not possible �

Medium-knowledge integrationA LORK is smaller than a WHALE. A WHALE is smaller than a LORK. (n)A PARM is smaller than a WHALE. A WHALE is smaller than a PARM. (n)

High-knowledge integrationLike WHALES, LORKS have fins. Like TURTLES, LORKS have legs. (n)Like WHALES, PARMS have fins. Like TURTLES, PARMS have legs. (n)

Low-knowledge access Type IIA TURTLE is smaller than a SHARK. A SHARK is smaller than a TURTLE. (n)

Paragraph 5 and Test Statements (Insects)

Judy went to look at bugs at the park yesterday. While looking at the bugs, Judy learned:





A SHET looks like a BUTTERFLY but a SHET is more colorful.

A MARB looks like a SHET but a MARB is more colorful.

Relationship: MARB � SHET � ButterflyText Memory

A SHET looks like a BUTTERFLY.A SHET does not look like a BUTTERFLY. (n)A MARB looks like a SHET.A MARB does not look like a SHET. (n)A SHET is more colorful than a BUTTERFLY.A BUTTERFLY is more colorful than a SHET (n).A MARB is more colorful than a SHET.A SHET is more colorful than a MARB. (n)

Text InferencingA MARB looks like a BUTTERFLY.A MARB does not look like a BUTTERFLY. (n)A MARB is more colorful than a BUTTERFLY.A BUTTERFLY is more colorful than a MARB. (n)

Low-knowledge integration� not possible �

Medium-knowledge integrationA SHET is more colorful than a LADYBUG.A LADYBUG is more colorful than a SHET. (n)A MARB is more colorful than a LADYBUG.A LADYBUG is more colorful than a MARB. (n)


24 HANNON AND FRIAS



High-knowledge integrationLike a LADYBUG, a SHET can fly through the air.Like a LADYBUG, a MARB can fly through the air.Like a WORM, a SHET can travel on the ground. (n)Like a WORM, a MARB can travel on the ground. (n)

Low knowledge access Type IIA BUTTERFLY is more colorful than a WORM.A WORM is more colorful than a BUTTERFLY. (n)

Appendix B

Instructions

We are going to play a game. This game is to see how you usewhat you already know about common things in order to help youlearn about some new things.

In this game, you will hear a story about what a child learned.You will also see pictures that describe what the child learned.Some of the pictures in the story are common things that youalready know, like HORSES and TOYS, but other things are new,like ZILLS and LEATS.

YOUR JOB is to try to learn all about these new things, likeZILLS and LEATS. You should also think about how these newthings are similar and different from the things you alreadyknow. Once you learn all about these new and common things,I will be asking you some questions. Ok, do you have anyquestions so far? (If they do, then play instructions again.)

Ok, so are you ready to play the game? So, you will listen tothe story and see the pictures two times. Then there will besome questions. When you hear a question, be sure to answerwith a Yes or a No. It is important that you understand thatsome of the questions may ask you about things you alreadyknow (like HORSES and TOYS). Some of the questions mayask you about the new things you learned about in the story(like ZILLS and LEATS). Other questions may ask you abouthow the new and common things are the same or different.

Received October 18, 2010Revision received May 30, 2012

Accepted June 1, 2012 �



JOBNAME: AUTHOR QUERIES PAGE: 1 SESS: 1 OUTPUT: Mon Jul 2 19:35:52 2012/tapraid5/zcz-edu/zcz-edu/zcz00312/zcz2436d12z

AQ1: Author: Please supply three to five keywords for your article.

AQ2: Author: Here you cite Paris & Paris, 2003, but there is no corresponding reference-list entry.Please provide a reference-list entry for this citation.

AQ3: Author: What does WPPSI-III stand for?

AQ4: Author: Is the WPPSI-R the same as the WPPSI-III? If not, please provide a citation andreference-list entry for the WPPSI-R.

AQ5: Author: Please provide a citation and reference-list entry for the WISC-III.

AQ6: Author: Here you cite Kaplan and Sacuzzo (2001), but there is no corresponding reference-list entry. Please provide a reference-list entry for this citation.

AQ7: Author: To which Cain et al., 2004 reference does this citation refer?

AQ8: Author: Here you cite Konold, Juel, McKinnon, and Deffes (2003), but there is nocorresponding reference-list entry. Please provide a reference-list entry for this citation.

AQ9: Author: The Cain, K., Oakhill, J., & Lemmon, K. (2004) reference is not cited in the text. Okayto delete from reference list? If not, please provide an exact location for citation within thetext.

AQ10: Author: Please provide volume and page numbers for the Engel de Abreu, P. M. J.,Conway, A. R. A., & Gathercole, S. E. (2010) reference.

AQ11: Author: The Landi, N. (2010) reference is not cited in the text. Okay to delete from referencelist? If not, please provide an exact location for citation within the text.

AQ12: Author: The Lonigan, C. J., Burgess, S. R., & Anthony, J. L. (2000) reference is not cited inthe text. Okay to delete from reference list? If not, please provide an exact location for citationwithin the text.

AQ13: Author: The Perfetti, C. A., & Lesgold, A. M. (1979) reference is not cited in the text. Okayto delete from reference list? If not, please provide an exact location for citation within thetext, and please provide a page range.

AQ14: Author: The Pressley, M., Wharton-McDonald, R., Allington, R., Block, C. C., Morrow, L., &

AUTHOR QUERIES

AUTHOR PLEASE ANSWER ALL QUERIES 1

JOBNAME: AUTHOR QUERIES PAGE: 2 SESS: 1 OUTPUT: Mon Jul 2 19:35:52 2012/tapraid5/zcz-edu/zcz-edu/zcz00312/zcz2436d12z

Tracy, D., & Woo, D. (2001) reference is not cited in the text. Okay to delete from referencelist? If not, please provide an exact location for citation within the text.

AQ15: Author: Is Riverside Publishing the author of this test, or just the copyrightholder/publisher? If there is a different author, please provide the name(s).

AQ16: Author: Please provide a city of publication for the Wechsler, D. (2002) reference.

AQ17: Author: Please provide a title for Appendix A.

AUTHOR QUERIES

AUTHOR PLEASE ANSWER ALL QUERIES 2

Preschooler Components of Listening Comprehenison

Documents