kuhn-et-al.-rrq-2010

7/29/2019 Kuhn-et-al.-RRQ-2010

1/22

230 Reading Research Quarterly 45(2) pp. 230251 dx.doi.org/10.1598/RRQ.45.2.4 2010 International Reading Association

Over the past decade, the field of literacy educa-tion has seen a major shift in fluencys role inthe literacy curriculum, moving from a rarely

encountered instructional component to one that isoften responsible for driving major instructional deci-sions (e.g., Riedel, 2007; Schilling, Carlisle, Scott, &

Zeng, 2007). This shift is due, in part, to the identifi-cation of fluency as one of the areas reviewed by theNational Reading Panel (National Institute of ChildHealth and Human Development [NICHD], 2000). Italso results from a broader reconsideration of the roleof oral reading in the development of skilled reading(e.g., Rasinski, 2006; Reutzel, Fawson, & Smith, 2008).The recognition of the importance of fluency that hasemerged as part of our developing understanding ofthe construct has led to a corresponding emphasis on

fluency assessment and instruction within the literacycurriculum (e.g., Pikulski & Chard, 2005).

Most literacy educators consider fluency to be a crit-ical component of reading development (e.g., Rasinski,Blachowicz, & Lems, 2006; Samuels & Farstrup, 2006).However, the current implementation of fluency in-

struction in many classrooms is often driven by assess-ments that build upon an incomplete conceptualizationof the construct and can lead to both inappropriate in-struction and a serious misconception of this essentialcharacteristic of skilled reading. Further, despite thesignificant amount of attention the construct of readingfluency has received recently, there are still a numberof questions surrounding our understanding of whatconstitutes fluency, its role in the reading process, andhow its assessment and instruction fit into the literacy

Review of Research

Aligning Theory and Assessmentof Reading Fluency: Automaticity,

Prosody, and Definitions of FluencyMelanie R. KuhnBoston University

Paula J. SchwanenflugelThe University of Georgia

Elizabeth B. MeisingerThe University of Memphis

Consulting Editors:Betty Ann Levy, McMaster University, Hamilton, Ontario, Canada

Timothy V. Rasinski, Kent State University, Ohio, USA

Over the past decade, fluent reading has come to be seen as a central component of skilled reading and a driving forcein the literacy curriculum. However, much of this focus has centered on a relatively narrow definition of reading fluency,

one that emphasizes automatic word recognition. This article attempts to expand this understanding by synthesizing sev-eral key aspects of research on reading fluency, including theoretical perspectives surrounding automaticity and prosody.It examines four major definitions of reading fluency and their relationship to accuracy, automaticity, and prosody. Aproposed definition is presented. Finally, the implications of these definitions for current assessment and instruction areconsidered along with suggestions for reenvisioning fluencys role within the literacy curriculum.

A B S T R A C T

7/29/2019 Kuhn-et-al.-RRQ-2010

2/22

Aligning Theory and Assessment of Reading Fluency: Automaticity, Prosody, and Definitions of Fluency 231

curriculum. We plan to use the opportunity presentedin this article to synthesize several key aspects of theresearch surrounding reading f luency, from theoreticalperspectives to the role evaluation plays in determin-ing practice, with an emphasis on the work that hasoccurred since the National Reading Panels (NICHD,2000) report.

Although there are a number of definitions of read-ing fluency, each of which places varying emphasis onits components, there seems to be a growing consen-sus that accuracy, automaticity, and prosody all makea contribution to the construct (e.g., Hudson, Pullen,Lane, & Torgesen, 2009; Rasinski, Reutzel, Chard, &Linan-Thompson, in press). Yet the way in which thesecomponents are conceptualized, their role in readingdevelopment, and their function in reading compre-hension all have significant influence on how they aretaught and assessed. In particular, we plan to considerautomaticity and prosody in greater detail. And, whilewe do not focus on the development of accurate word

recognition, per se, we address accuracy as part of thebroader discussion of both automaticity and fluency as-sessment. This decision was made in part to move awayfrom the view that reading fluency results from an im-provement in the ability of students to recognize wordsand their component elements with increasing rapidity.Instead, we refer the reader to elegant discussions of thedevelopment of reading accuracy by Chall (1996), Ehri(1995), and Perfetti (1992).

Our discussion is divided into several parts; thefirst provides theoretical perspectives on reading flu-ency, particularly the role of automaticity and prosody

in fluency. We then consider four definitions of fluentreading; each definition places differing emphasis onfluencys component parts as well as on the role flu-ency plays in the reading curriculum. As part of thesedefinitions, we also present our own conception of whatconstitutes fluent reading. Next, we explore the rela-tionship between certain conceptualizations of readingfluency, dominant assessments, and current practice.Finally, we consider the implications of these defini-tions for assessment and instruction and make sugges-tions for incorporating a broader understanding of thegoals and purposes of reading fluency within a reenvi-sioned literacy curriculum.

Theoretical Perspectives on ReadingFluency

AutomaticityAutomatic word recognition is central to the constructof fluency and fluencys role in the comprehensionof text (e.g., Samuels, 2004, 2006). But what are the

qualities that make for automaticity as it relates to read-ing fluency? According to Logan (1997; see also Moorsand DeHouwer, 2006), processes are considered to beautomatic when they possess four properties: speed,effortlessness, autonomy, and lack of conscious aware-ness. These properties can be considered together orseparately when determining whether a skill is automa-tized (Moors & DeHouwer, 2006).

The first of these properties is speed, which isthought to emerge concurrently with accuracy as learn-ers engage in practice (Logan, 1988). As automaticitydevelops, whether in terms of reading, perceptual-mo-tor activities, or another skilled task, the learners per-formance not only becomes accurate, it gets faster.However, this increase in speed is not limitless. Rather,the learning curve for these tasks follows what is knownas thepower law; this states that reaction time decreas-es as a function of practice until some irreducible limitis reached. Speed increases throughout practice, butthe gains are largest early on and diminish with further

practice (Logan, 1997, p. 123).In terms of connected text, the power law can be

seen in Hasbrouck and Tindals (2006) oral reading flu-ency norms; for example, between winter and springof the first-grade year students at the 50th percentileincrease their reading rate approximately 30 correctwords per minute, whereas their peers in the eighthgrade gain only 18 correct words per minute over theentire school year and the gains for adult skilled read-ers, who have reached asymptote, are infinitesimal.

The second attribute of automaticity is effortless-ness (Logan, 1997). This refers to the sense of ease with

which a task is performed and to the ability to carryout a second task while carrying out the first, automaticone. If a person is able to accomplish two tasks at once,then at least one of those tasks is, by necessity, auto-matic. In terms of fluency, effortlessness can be seen intwo ways. First, fluent readers lack a sense of strugglein recognizing most of the words they encounter in text.This effortlessness in word recognition is derived, inpart, from unitization, a process that involves collaps-ing some of the sequential steps used to identify words(Cunningham, Healy, Kanengiser, Chizzick, & Willitts,1988). Slow, algorithmic sequential word identificationprocesses are seemingly replaced by a shift toward di-

rect single-step retrieval of larger units (such as wordsand phrases) in long-term memory. These retrievedskills essentially outpace the slower algorithmic wordidentification processes and can be completed morequickly (Logan, 1988). Second, most fluent readers notonly decode text, but also simultaneously comprehendwhat they are reading. Inefficient word recognitionhampers comprehension and takes up precious cogni-tive resources that should be used for understanding.

With automatization of lower level processes, readers

7/29/2019 Kuhn-et-al.-RRQ-2010

3/22

Reading Research Quarterly 45(2)232

can shift their attention from lower level skills to higherlevel, integrative aspects of reading such as reading f lu-ently with comprehension. Disfluent readers, on theother hand, are unable to integrate these lower levelskills with higher level ones, primarily because of theeffort they need to expend on word recognition (e.g.,LaBerge & Samuels, 1974; Samuels, 2006).

In addition to rate and effortlessness, automatic pro-cesses are also autonomous; that is, they occur with-out intention, beginning and running to completionindependent of the direction or intent of the personundertaking the act (Logan, 1997). In contrast, a non-autonomous process is deliberate, allowing an individu-al to maintain control over the act and deciding whetherit occurs. In the case of reading, fluent readers havelittle choice but to recognize words as they encounterthem whereas beginning readers do not find readingto be an obligatory act. For example, fluent readers of-ten find themselves inadvertently reading the text thatruns along the bottom of a news program, although

they are eventually able to use their available cognitiveresources to inhibit it. Disf luent readers, on the otherhand, are either unable to process the text at all or mayfind their attentional resources excessively preoccupiedby it (Schwanenflugel & Ruston, 2008). However, au-tonomous processing of words comes early in the de-velopment of reading, perhaps even before children aretruly f luent readers (Schwanenflugel, Morris, Kuhn,Strauss, & Sieczko, 2008; Stanovich, Cunningham,& West, 1981). Indeed, continued lack of autonomyof lexical processing is an indicator that the child (oradult) is not yet a fluent reader (Protopapas, Archonti,

& Skaloumbakas, 2007; Schwanenflugel et al., 2006).The final characteristic of automaticity is a lack ofconscious awareness (Logan, 1997). Once lower levelword recognition skills become automatic, the con-scious awareness of the subskills that comprise themdisappears. This lack of conscious awareness in wordrecognition differentiates fluent from disfluent read-ers. Disfluent readers tend to be keenly aware of thesteps they need to undertake to determine the wordsin a text and find the process to be slow and deliberate(e.g., Chall, 1996). However, because word recognitionhas become automatic for fluent readers, they are ableto identify nearly every word they encounter without

conscious effort.Although each of these four properties can be ap-

plied to automatic word recognition, it is important toremember that these attributes develop on a continu-um, as well as at different rates, so that readers whohave had an intermediate amount of practice may besomewhat fast, somewhat effortful, somewhat autono-mous, and partially unconscious (Logan, 1997, p. 128).Further, as readers gain skill and are exposed to moretexts, automaticity may expand not just at the sublexical

(i.e., phoneme and rime level) and word level, but alsoat the phrasal and perhaps even the sentence level.

Developing Automatic Word RecognitionAlthough the aforementioned discussion indicates thecomplexities of automaticity, the issues surroundingthe development of automatic word recognition are still

critical to reading fluency and therefore deserving ofour attention. So how does automatic word recognitiondevelop? The basic answer is that it occurs through con-sistent practice (Logan, 1997; Samuels, 2004). However,what that practice consists of and what it results in arecentral determinants in our current understanding ofboth reading fluency and its implementation in theclassroom.

When discussing word recognition automatic-ity, we are talking about comparatively instantaneousidentification. Such rapid word recognition is impor-tant because readers need to integrate informationfrom multiple sourcesphonemic, semantic, phrasal,textual, and so on. However, because of the cognitiveresources used by word recognition, beginning readersmust switch between these multiple sources rather thanprocess them in a unified manner. To move beyondthis serial processing and toward the autonomous wordrecognition entailed by fluent reading, learners requirethe opportunity for extensive practice in the reading ofconnected text (Kuhn et al., 2006; Schwanenf lugel etal., 2009).

According to Logan (1997), every encounter witha task lays down a trace, or instance representation, inmemory. As the number of encounters, or instances,

increase, learners begin to build their knowledge base.When individuals first encounter a representation, theirperformance is based on an algorithmic computationthat involves thinking or reasoning. However, as theirencounters with a particular task increase, their knowl-edge base becomes more extensive and their retrievalsbegin to be based on past instances, or memories of pastsolutions, rather than on the need to formulate a solutionbased on slow algorithmic processes (see also Rawson,2007; Rawson & Middleton, 2009). When this knowl-edge base is substantial enough, learners performancecan be based entirely on memory retrieval. However, al-though it is most likely that automaticity will occur after

numerous exposures to a task, it is conceivable that itcould occur after only one encounter. And, adding onetrace to the initial encounter, or even the first 10 en-counters, will have greater impact on the readers abil-ity to retrieve that trace quickly, or from memory, thandoes adding one trace to the one-hundredth encounter(Logan, 1992; Logan, Taylor, & Etherton, 1999). Thisnotion has important implications for reading practice.

When reading, learners encounter letters, words,and phrases and construct higher order propositional

7/29/2019 Kuhn-et-al.-RRQ-2010

4/22


structures; and each reading leaves a trace at each levelof representation (Logan, 1997). Although it is true thatthe number of times individuals encounter instancesat these different levels of representation varies fairlydramatically, for example, readers encounter letters andeven high-frequency words far more often than they doa particular higher order structure, there will still be

some benefit for readers from each encounter at everylevel.We contend that this argument has important im-

plications for practice. To begin with, readers can ben-efit from both repetition (e.g., Levy, 2001; Logan, 1997;Samuels, 2006) and the wide reading of texts (e.g.,Schwanenflugel & Ruston, 2008; Stanovich, 1986).Repetition of text allows for the kind of consistent prac-tice that is important to readers. And, drawing fromboth the Samuels and the Logan theories of automatic-ity, repetition allows for the deepening of traces (Logan,1997) and the freeing up of attention (Samuels, 2006).Further, Logan pointed out that, in addition to devel-oping automatic word recognition, repeated readingsallow learners to establish prosody, identify appropri-ate phrasing, and determine meaning. Thus difficultiesencountered in a text can be successfully solved as thetext is read repeatedly and, as a result, similar difficul-ties are likely to be more readily solved when encoun-tered in another text.

Another important implication concerns the pow-er law. Because most of the gains made with repeatedreadings, both in terms of accuracy and automatic-ity, occur between the third and the fifth repetition(e.g., OShea, Sindelar, & OShea, 1987; Reutzel, 2003;

Rawson & Middleton, 2009), the power law mentionedearlier provides a reasonable explanation for decreasinggains across continued repetitions. Indeed, after someminimal amount of practice, readers seem to rely on di-rect retrieval of text meanings rather than on slow algo-rithmic processing of each word (Rawson & Middleton,2009).

However, it is important to note that Logan (1997)also argued that some variability in practice can benefitlearners: Automaticity transfers to similar stimuli, sothere should be some benefit in exposing readers to dif-ferent materials (p. 139). Wide reading provides oppor-

tunities for just such transfer, and research conducted onstudents who were asked to read a wide variety of mate-rials with adequate support (e.g., Kuhn, 2005; Kuhn etal., 2006; Schwanenflugel et al., 2009; Schwebel, 2007)indicates that their automaticity does improve. Becausethere is a great deal of word overlap in the materialsused for beginning readers (e.g., Adams, 1990), it seemslikely that seeing words in multiple contexts improvesstudents recognition of those words (Mostow & Beck,2005; Rashotte & Torgesen, 1985). However, the ex-

act degree of similar versus unique words needs to bedetermined (e.g., Allington, 2009; Hiebert, 2006).

In addition to the sheer number of words that oc-cur in multiple contexts, it might also help to have stu-dents read across themes, so that when a new word isencountered, there is a greater likelihood of it beingseen within a different but supportive context (e.g.,Logan, 1997). In this way, students are more likely tobuild upon and have the opportunity to expand theirconceptual, as well as their orthographic, knowledge.

We consider this understanding to be a complementto the arguments presented by Stanovich (1986) in hisarticle describing the Matthew Effect in reading anddemonstrated in research we recently conducted withseveral colleagues (Kuhn et al., 2006; Schwanenflugelet al., 2009). Not only do readers who read widely havemore accurate and automatic word recognition, but theyalso have a more extensive vocabulary and encounter abroader range of concepts than do their peers who readin a more limited way (Stanovich, 1986). Continued

practice on the same words, or same texts, beyond acertain point may not only be redundant, it may havethe perverse effect of fixing students attentional focuson the lower level aspects of text rather than shiftingtheir focus toward practicing the integration of higherlevel skills. Practice through wide reading would trans-late into greater fluency, leading to further increases inreaders ease and comfort with texts.

We wish to conclude this discussion with a re-iteration of what we consider to be a central tenet ofautomaticity; it is important to stress that, whether de-veloped through repetition or the wide reading of texts,automaticity occurs on multiple levels and connects tocomprehension in multiple ways (e.g., Samuels, 2004;Logan, 1997). We also want to stress that it is this inter-action, occurring between various levels of processing,rather than simple speeded word recognition, which iscentral to a readers construction of meaning from text(Bredekamp & Pikulski, 2008; Fuchs, Fuchs, Hosp,& Jenkins, 2001; Hudson et al., 2009; Wolf & Katzir-Cohn, 2001).

ProsodyAlthough automaticity is central to childrens develop-ment as fluent readers, it does not account for all as-

pects of the construct. A second critical componentof reading fluency is the ability to read with prosody;that is, with appropriate expression or intonation cou-pled with phrasing that allows for the maintenance ofmeaning (Cowie, Douglas-Cowie, & Wichmann, 2002;Miller & Schwanenflugel, 2006, 2008; Schwanenf lugel,Hamilton, Kuhn, Wisenbaker, & Stahl, 2004).

However, the import of developing expressivenessin reading, as children proceed from reading in a stac-cato, flat, word-by-word manner to something that

7/29/2019 Kuhn-et-al.-RRQ-2010

5/22


sounds more or less like everyday speech, is not en-tirely clear. Is expressiveness merely an epiphenomenonwhich proceeds of its own accord with little impact onother aspects of reading, or is it some essential ingre-dient that benefits (or perhaps enables) other readingprocesses? Our question is the following: If the devel-opment of expressiveness is important, what about it isimportant and what is it important for? If it is essentialto reading, we may, indeed, wish to prioritize prosodyin our instruction. If it is inessential or emerges with-out instruction, then we might decide not to. In recentyears, the evaluation of expressiveness in fluent readinghas become the focus of empirical research to addressthese questions. In what we present here, we equatereading with expression with reading prosody.

Prosody is the music of language. Indeed, some an-thropologists have claimed that speech prosody servedas the protolinguistic base from which music itself mayhave emerged (Simpson, Oliver, & Fragaszy, 2008).Prosody captures the rise and falls of pitch, rhythm,

and stressthe pausing, lengthening, and elision sur-rounding certain words and phrases that is found in thepull of linguistic communication (Hirschberg, 2002).However, there are clear developments in childrensunderstanding and use of prosody in their own speechthat are ongoing during the period in which they arelearning to read.

In this section, we begin by considering the spec-trographic features measured to discern the qualities ofprosody and their import in the development of read-ing prosody. We outline the psycholinguistic functionsof prosody. We consider the costs and benefits of vari-ous ways of measuring prosody for reading fluency. Wethen describe what we know about where prosody fitsin our conceptions of the development of reading skill.

Prosody FeaturesThe first of these features is fundamental frequency (F

o)

or, more simply, pitch. Pitch needs to be consideredrelative to a speakers voice range and native language.For example, young children with their high-pitchedvoices may not have prosodic room to regulate pitch.Language features such as tones in tone-bearing lan-guages such as Chinese will affect measured pitch.

Declarative sentences or statements are usually sig-

naled by an initial rising and then falling pitch (calledpitch declination or, simply, declination). As sentenc-es become longer, there is a general flattening out ofpitch (Ladd, 1984), so we can expect chi ldren to dis-play smaller sentence-final declinations as they readcomplex texts (Benjamin, Schwanenflugel, & Kuhn,2009). Yesno questions are usually marked by sus-tained rising pitch, but this rising pitch is not obliga-tory for all question types (Miller & Schwanenflugel,2006). Further, childrens understanding of declarative

question prosody (e.g., He ate a bologna sandwich?) is stillunder development until around age 11 (Patel & Grigos,2006). Consequently, we should not tell children to useascending pitch at each and every question mark as issometimes the advice given to teachers (Hudson, Lane,& Pullen, 2005).

Pitch can convey pragmatic information as well. Aplateau contour can convey a sense of boredom or reci-tation effect. A continuation rise can indicate continua-tion or uncertainty (Hirschberg, 2002). Neither patternin childrens readings should necessarily be taken asindicating a lack of f luency. As children learn to readwith good prosody, they come to display an intonation-al pitch contour increasingly similar to the one used byadults when they read. In our studies, this has been avery consistent pattern associated with good fluency(Schwanenflugel et al., 2004; Miller & Schwanenflugel,2006, 2008).

Another prosodic feature is duration. Vowels instressed words are usually longer than in unstressed

words (Temperley, 2009) and even longer in phrase-final position. Stressed syllables tend to also havegreater intensity, or volume (Cooper & Paccia-Cooper,1980). Duration has to be taken in context with thespeakers overall speaking rate. Thus, faster readerswill have shorter segment durations than slower read-ers. However, syllable duration will become shorter asspeakers proceed over long sentences (Ladd, 1984).This means that a child who has been told to readquickly will show less evidence of stress marking andphrase-final lengthening. Children will not be able toread both very quickly and with proper prosody, so di-

recting them to read passages quickly and accuratelywill have the perverse effect of having them read lessexpressively.

Stress is a property in speaking that makes onesyllable in a word more prominent than its neighbors(Himmelmann & Ladd, 2008, p. 248). Knowledge ofa words stress seems to be retrieved automaticallywhen a word is read (Gutirrez-Palma & Palma-Reyes,2008). Function or closed class words tend to be un-stressed. However, English favors a regular distributionof stressed and unstressed syllables, and this will causeEnglish speakers to add or move stress to keep up aregular stress pattern (e.g., I gave it to the postman) and

to avoid stress clashes (e.g., She turned thirteen versusthirteen donuts; Temperley, 2009). Stress can be usedto distinguish grammatical form class in English (e.g.,permit [noun] versus permit [verb]) with nouns beingmore likely to be stressed on the first syllable than areverbs (Kelly & Bock, 1988). Each language, however,follows its own rhythmic pattern. Sensitivity to stresspatterns is related to the development of skilled read-ing (de Bree, Wijnen, & Zonneveld, 2006; Goswami etal., 2002; Jarmulowicz, Taran, & Hay, 2007; Orsolini,

7/29/2019 Kuhn-et-al.-RRQ-2010

6/22


Fanari, Tosi, de Nigris, & Carrier, 2006; Thomson,Fryer, Maltby, & Goswami, 2006; Whalley & Hansen,2006; Wood, 2006). So, in monitoring for prosody inchildrens reading, we should look for the familiar stresspatterns associated with the language that the readersspeak, keeping in mind that nonnative speakers are un-likely to show nativelike use of stress (Guion, Harada,

& Clark, 2004).Pausing is noted by a spectrographic silence in oralreading beyond that invoked by some consonant com-binations. Slow speakers make more pauses, and peoplediffer considerably as to whether they make sentence-internal pauses in speech (Eisler, 1968; Krivokapic,2007). Regardless, intrasentential pauses tend to beshorter than intersentential ones (Cooper & Paccia-Cooper, 1980). Pauses tend to be larger both preceed-ing and following syntactically complex phrases and asinformation load increases (Cooper & Paccia-Cooper,1980; Ferreira, 1991; Zvonik & Cummins, 2003). Still,

we should not expect readers to pause midsentencesimply because they have completed a complex nounphrase. Neither should we consider a pause in mid-sentence a reading error in long, complex sentences.Our work has suggested that most midsentence pausesamong young readers are related to decoding abilities(Miller & Schwanenflugel, 2008).

What Are the Psycholinguistic Functionsof Prosody?Prosody provides a variety of natural breakpoints incontinuous speech. These intonational units provide

distributional edges that allow the listener, includ-ing children, to break up continuous speech for pars-ing (Ramus, Hauser, Miller, Morris, & Mehler, 2000).

Words at the right edges of these units are likely to pos-sess boundary tones that indicate the end of the partic-ular unit, typically word-final lengthening, declination,or pausing. If speech has these boundary markers in-serted incorrectly, it is difficult both to understand andto parse (Sanderman & Collier, 1997; Shukla, Nespor,& Mehler, 2007); it is possible that the intermittentpausing found in the disf luent reading of young chil-dren may have this effect also, but this has yet to be

determined.As indicated by Wheeldon and Lahiri (1997), pro-

sodic constituents are derived from syntactic constitu-ents but are not necessarily isomorphic to them (p.357). Thus, syntactic bracketing (e.g., [[[The girl]

NP[[I]

NP

[[danced with]V

[at the party]PP

]VP

]S]

NP[tripped]

VP]

S) is

considerably richer than the bracketing that prosodyimposes. So one cannot assume that the positive effectsof syntactic bracketing of text and greater syntacticawareness on readers comprehension (e.g., Mokhtari &

Thompson, 2006; Young & Bowers, 1995) will be thesame as those found for reading prosody.

One of the essential functions of prosody is to pro-vide a basic cognitive skeleton that allows one to holdan auditory sequence in working memory (Frazier,Carlson, & Clifton, 2006; Swets, Desmet, Hambrick,& Ferreira, 2007). By cognitively bracketing key in-formational units such as phrases, prosody assists inmaintaining an utterance in working memory until amore complete semantic analysis can be carried out(Koriat, Greenberg, & Kreiner, 2002). Although thereis no evidence currently that the development of ap-propriate reading prosody allows this to occur, it hasbeen shown that people have better memory for poeticversions of texts that have enhanced prosodic features(Goldman, Meyerson, & Cot, 2006). It is possiblethat the construction of a good prosodic reading (com-pared with an inappropriate rendering) might improvecomprehension.

Prosody can also serve to disambiguate semantically

and syntactically ambiguous sentences. Because speak-ers rarely recognize their own ambiguity, they dont useprosody reliably to disambiguate their own utterances(Allbritton, McKoon, & Ratcliff, 1996; Beach, 1991;Snedeker & Trueswell, 2003), but listeners use it whenits available (Snedeker & Trueswell, 2003). Childrenhave a fragile awareness of how prosody relates to dis-ambiguation (Snedeker & Yuan, 2008). Consequently,we should not expect children to use this type of disam-biguating prosody in their oral readings.

Prosody carries more than just syntactic phras-ing, however. Different prosodic patterns convey dif-

ferent emotions (Banse & Scherer, 1996; Juslin &Laukka, 2003). For example, happiness is characterizedby fast speech rate, high, rising pitch and variability,and fast voice onsets; and sadness nearly the opposite.Uncertainty is signaled by a sustained rise in pitch(Hirschberg, 2002). However, during the period wherechildren are developing fluency, their concomitant un-derstanding of emotional prosody is still not fully adult-like (Fujiki, Spackman, Brinton, & Illig, 2008; Wells &Peppe, 2003), so we should not expect them to conveythese attitudes fully in their readings.

Prosody also carries discourse information. Higher,more variable pitch tones and longer pauses are typi-

cally seen at higher levels in the discourse hierarchy,for example, at topic shifts and the initial position ina paragraph (Noordman, Dassen, Swerts, & Terken,1999; Smith, 2004). High pitch tones are used to intro-duce new topics and low pitch tones are used to indi-cate that the topical anaphor is in short-term memory(Wennerstrom, 2001). Pitch and punctuated stress isalso used to dictate informational focus and contrast(Carlson, Dickey, Frazier, & Clifton, 2009; Couper-Kuhlen & Selting, 1996). Informationally related

7/29/2019 Kuhn-et-al.-RRQ-2010

7/22


utterances are distinguished by short pause durationsbetween them and faster rates (den Ouden, Noordman,& Terken, 2009). However, again, children do not havefull understanding of the import of these discourse el-ements of prosody possibly until they are adolescents(Chen, 1998; Wells & Peppe, 2003), so it is unclearwhether they will know to convey this information intheir oral readings. To date, discourse features havebeen largely ignored in the study of the development ofreading prosody. We currently do not know whether orwhen children come to use these features in their oralreadings as they become fluent readers.

In sum, we see that a tremendous amount of infor-mation is available for communication in the prosody ofsophisticated readers. However, during the same periodwhen children are learning to read fluently they are alsodeveloping a general understanding of the various usesof prosody. At this point, research is unclear on whichattributes could serve as valid, reliable assessments ofchildrens ability to read fluently. Further, most of the

studies described in this article regarding prosody fo-cus on English speakers. (Only 26% of the studies usedlanguages other than English and most of these wereGermanic languages.) Prosody is not identical acrosslanguages, so it is important to understand the limita-tions of current research with regard to linguistic diver-sity, including bilingual children.

Measuring Prosody: Direct MeasuresVersus RatingsThe sine qua non of reading fluency is that childrenread in a manner that approximates speech. Yet readingprosody is not identical to speech prosody. Even in f lu-ent readers, reading prosody has fewer end-of-sentencerises, fewer very low pitch ranges (as for parentheticalspeech), and possesses generally less variability thanspontaneous speech (Esser & Polomski, 1988). Overall,adults pause less in read speech, show more consistentstress placement and generally cleaner speech (Howell& Kadi-Hanifi, 1991) than in spontaneous speech.Minor syntactic boundaries are less likely to be markedin read speech than spontaneous speech (Blaauw,1994). Indeed, it may be that only professionals, suchas television newscasters, truly read in a way that ap-

proximates speech (Esser & Polomski, 1988); so thisexpectation is likely a bar set too high for determiningthe achievement of reading fluency. However, where toset the bar and how to set it empirically is the issue inquestion.

There are two basic ways to measure reading pros-ody: rating scales and spectrographic measures. In theclassroom, rating scales are relied upon for evaluation,and the NAEP Oral Reading Fluency Scale is the mostcommon measure (Pinnell et al., 1995). This 4-point

scale distinguishes reading that sounds primarily word-by-word from reading that occurs in larger, meaning-ful phrase groups (Pinnell et al., 1995, p. 15). Anotherpopular rating scale that focuses more on the prosodiccharacteristics of oral reading is the MultidimensionalFluency Scale (Rasinski, Rikli, & Johnston, 2009; Zutell& Rasinski, 1991). This scale consists of four separate4-point subscales that distinguish phrasing and ex-pression, smoothness and accuracy, and pacing. Thesescales are then summed to represent childrens overallratings of fluency. Rasinski et al. (2009) have reportedinter-rater agreement within 2 points to be 86%. Morerecently, however, Klauda and Guthrie (2008) added tothis scale by including a 4-point rating scale that dis-tinguished passage-level expressiveness with a 1 indi-cating that the child read with no mood or tone to a 4indicating that the child read the whole or nearly thewhole passage in an expressive manner that created amood or tone that seemed in accord with the authorsintention (p. 314).

Unfortunately, even after collapsing two points onthe scale, the researchers were only able to achieve 79%agreement so it is unclear whether this revised scalewill have sufficient reliability to add a degree of preci-sion to fluency ratings. Whether rating scales will everhave the precision necessary for them to add meaning-fully to our measurement of reading fluency beyondtext reading speed and accuracy (see also Fuchs et al.,2001) is a concern, but it is an avenue that needs tobe pursued. Still, we believe that these more complexscales are the general direction in which rating scales ofprosody need to go.

In our own work on reading prosody, we have al-ways employed spectrographic measures. We do recog-nize that the technical skills related to spectrographicmeasurement will be beyond the needs of most teach-ers, and perhaps reading specialists, although we notethat these tools are becoming increasingly easy to use.

What is needed right now is research that allows us torelate spectrographic measures directly to various rat-ing schemes so that informationally valid and reliableratings of reading prosody having curricular utility canbe created. We are certain that these measures will needto include some notation of the complexity of the textbeing read (Benjamin et al., 2009) as well as their gen-

eral discourse features because it is tempting to assigna child a NAEP rating of 4, say, on a simple passagewhile the same child might receive only a 1 on a morecomplex one. Benjamin and Schwanenflugel (2009)have shown that prosody measured from simple pas-sages is simply less predictive of reading skill than isprosody measured from passages that press the upwardlimits of childrens skills. We are also encouraged thatthe need to revise prosody rating scales might be pre-empted by recent advances in artif icial intelligence tools

7/29/2019 Kuhn-et-al.-RRQ-2010

8/22


that may allow us to automate the process of identifyingthe adultlike extent of childrens reading expressiveness(Mostow & Duong, 2009).

Where Does Reading Prosody Fit in OurConceptions of Development of ReadingSkill?Prosody is at the heart of the development of readingskill. Prosody is likely another aspect of the fundamen-tal phonological representations that drive much of thedevelopment of early reading skill (de Bree et al., 2006;Goswami et al., 2002; Surnyi et al., 2009). However,because there are distinct prosody features at lexical,phrasal, sentence, and discourse levels, these may onlybe partially related to phonological codes that connectto basic phonological (segmental) awareness (Whalley& Hansen, 2006).

Prosody is most certainly related to the developmentof reading f luency. As children become more fluent

readers, they also make shorter and less variable in-tersentential pauses, shorter and less frequent intrasen-tential pauses, and larger pitch declinations and displaya more adultlike intonation contour (Clay & Imlach,1971; Cowie et al., 2002; Miller & Schwanenflugel,2006, 2008). These changes in reading prosody be-tween first and second grade are predictive longitu-dinally of later reading fluency, beyond measures ofword reading efficiency and text reading rate (Miller &Schwanenflugel, 2008). Pauses seem to be more con-nected to word reading skill than fluency itself, but aschildren read more complex passages, pauses in fluent

readers will more systematically mark the greater syn-tactic complexity and sheer length of the sentences thataccompany such texts (Benjamin et al., 2009).

Reading prosody also seems to be related to read-ing comprehension. Our own work has found varyingpatterns regarding the relationship between readingprosody and reading comprehension, some of whichseem to be attributable to passage characteristics.Measurements of prosody from simple texts (relative tothe absolute levels of reading skills of the children) donot seem to contribute much to our ability to predictreading comprehension skill (Schwanenflugel et al.,2004). Measurements of prosody from more complex

texts do predict reading comprehension skills beyondthose accounted for by word reading efficiency or textreading rate measures (Benjamin et al., 2009; Klauda& Guthrie, 2008; Miller & Schwanenflugel, 2006).Indeed, we have found that the reading prosody of sim-ple texts in first grade predicts childrens reading com-prehension skills of more complex texts two years later(Miller & Schwanenf lugel, 2008). Thus, it appears thathaving appropriate reading prosody is independentlyrelated to good reading comprehension.

At present we do not know the directionality of thisrelationship. That is, does reading with good prosodyhelp the reader comprehend what is being read or doescomprehending while reading simply promote goodreading prosody? Or is the relationship between read-ing prosody and reading comprehension reciprocal?Currently, we know of two studies that have addressedthe directionality issue, and they have come to differ-ent conclusions using different methods. Using secondand third graders as participants, Schwanenflugel et al.(2004) evaluated two structural equation models im-plying different directionality. In one, reading prosodyserved as a partial mediator with word reading efficien-cy to predict reading comprehension score outcomes. Inthe other, reading comprehension and word reading ef-ficiency predicted reading prosody as outcomes. In thatstudy, only the first model (i.e., that reading prosodypredicted reading comprehension) fit the data. Klaudaand Guthrie (2008) examined the issue of whetherchanges in ratings of syntactic prosody were recipro-

cally related to changes in reading comprehension lon-gitudinally beyond word reading speed over the courseof the fifth-grade year. They found evidence for reci-procity. Whether the differences in outcomes betweenthese studies could be attributed to differences to theage of the children or the particulars of the methods isnot clear. Directionality and causality between readingprosody and comprehension remain to be determined.

Finally, we hypothesize that the development oforal reading prosody will be related to the movementtoward what psycholinguists have called implicitprosody (Fodor, 2002), which may develop as children

make the transition from oral reading to silent read-ing (McCallum, Sharp, Bell, & George, 2004; Prior &Welling, 2001). According to Fodor (2002), a defaultprosodic contour is projected onto the reading materi-als during silent reading. Several findings support theexistence of this implicit prosody during silent reading.

Among them, in silent reading, adults read words withmultiple stressed syllables more slowly than words witha single stressed syllable, even though syllable structureis not actually needed (Ashby, 2006; Ashby & Clifton,2005). Further, adult readers appear to dwell on com-mas during silent reading (Hirotani, Frazier, & Rayner,2006), particularly when they are needed to disambigu-

ate syntactically ambiguous sentences (Kerkhofs, Vonk,Schriefers, & Chwilla, 2008). Event-related brain po-tentials seem to be linked to focus during silent read-ing in adults (Stolterfoht, Friederici, Alter, & Steube,2007). Whether there will be a one-to-one relationshipbetween implicit prosody and all the various featuresfound in oral reading prosody remains to be seen.

We can already identify some places where theremay be differences between skilled and novice oralreaders. Among them, most adults do not pause on each

7/29/2019 Kuhn-et-al.-RRQ-2010

9/22


and every comma when they read aloud (Chafe, 1988);pausing on commas is a feature of younger, less gener-ally skilled readers (Miller & Schwanenflugel, 2006).Further, we need to ensure that our theories regard-ing the development of implicit prosody do not exceedwhat we know about the developing status of childrensunderstanding of prosody, particularly during the pe-riod that they are learning to read, which for somefeatures extends until age 18 or so (Plante, Holland, &Schmithorst, 2006).

Still, the implicit prosody hypothesis is an intrigu-ing one that needs further research. Moreover, it willbe important to understand more about how prosodicreading is acquired so we can determine how it maypromote, or perhaps enable, the development of implicitprosody in the silent reading prosody of children. Onestudy is particularly intriguing with respect to this.Kleiman, Winograd, and Humphrey (1979) showedthat below-average fourth-grade readers had difficultymarking phrase boundaries in silent reading compared

with sentences that were presented in both spokenand written form. This is suggestive, at least, that poorreaders may be having difficulties generating implicitprosody during silent reading to support their compre-hension. Of course, this is not the only explanation forthese findings, but they fit the pattern anticipated bythis view. Similarly, Rasinski et al. (2009) have foundthat oral reading prosody ratings using the multidimen-sional fluency scoring rubric bear a substantial relation-ship to silent reading comprehension scores.

In the beginning of our discussion of prosody, weasked what the role of reading prosody was in the devel-

opment of reading fluency. We asked whether expres-siveness is merely an epiphenomenon which proceedswith little impact on other aspects of reading and, ifnot, then for what is prosody used. We believe we cansay rather conclusively at this point that good readingprosody emerges as children develop efficient word andtext oral reading skills. To connect to our earlier discus-sion regarding automaticity, we can say that childrenwho develop efficient word and text reading skills seemto use the newly freed up resources gained from theseautomated skills and shift their attention to the integra-tion of speech prosody with integrative reading skills.Thus, prosody seems to be related to the development of

good oral reading fluency and, indeed, may be a markerof it. If so, prosody should be measured whenever read-ing fluency is measured.

We also suggested that good reading prosody maysupport reading comprehension, but the directionalityof this has yet to be determined. The directionality issueis important so that we can determine whether a par-ticular instructional emphasis on prosody is necessary.If acquiring good reading prosody supports improvedcomprehension (as we think evidence is beginning to

support), then we should emphasize prosody in our in-struction along with these other skills. If, instead, ac-quiring good reading prosody is a ref lection of efficientdecoding and comprehension skills alone, it may notmake much sense to focus childrens attention instruc-tionally on developing newscaster-like oral reading be-cause this by itself would have limited utility. Once weare certain that developing good reading prosody hascausal value for improved reading comprehension, thenwe should shift our research to considering better (andworse) ways of integrating such instruction in our lit-eracy practice.

DefinitionsHaving discussed the automaticity and prosody con-structs surrounding reading fluency, we turn to themultiple ways in which fluency is defined. This is notan exclusively theoretical issue or simply a matter of

semantics. Because classroom instruction developsaround teachers perceived understanding of a con-struct, the way in which they view certain aspects ofthe reading process has a decisive role in their teachingand assessment of those aspects. Further, these concep-tualizations strongly affect learners understanding ofwhat reading is as well as what it means to be a reader.It is also important to highlight the commonalities anddifferences in these definitions while working toward amore cohesive understanding of what fluency is, as wellas of what it is not. So, while many definitions of f lu-ency highlight the importance of accuracy, automaticity,and prosody in relation to the comprehension of text,

(e.g., Fuchs et al., 2001; NICHD, 2000; Rasinski et al.,in press; Samuels, 2006; Torgesen & Hudson, 2006),which of these elements they emphasize and the rolesthey are assigned in the development of skilled readingvary widely.

Fluency as Accuracy and AutomaticityThe first definition emphasizes accurate and automaticword recognition and those components, such as pho-nemic awareness and lettersound correspondences,which allow students to rapidly, and correctly, identifywords (Fletcher, Lyon, Fuchs, & Barnes, 2007; Good,

Kaminski, Simmons, & Kameenui, 2001). As can beseen in the earlier discussion of automaticity, there islittle dispute that accurate, automatic word recognitionis a critical component of fluent reading, or that pho-nemic awareness, letter naming, or other componentscontribute to the development and consolidation ofstudents word recognition (e.g., Ehri, 1995; NICHD,2000). In fact, most fluency researchers (e.g., Rasinskiet al., 2006; Samuels & Farstrup, 2006) agree that ac-curate and automatic word identification plays a central

7/29/2019 Kuhn-et-al.-RRQ-2010

10/22


role in fluent reading, and that components, such asphonemic awareness and letter naming, are importantin the process of developing accuracy and automaticityin their turn (e.g., Chall, 1996; Ehri, 1995).

What needs to be challenged, however, is the em-phasis that is placed on accuracy and automaticity, tosome extent, simply because they are the most quan-tifiable elements of fluency (Paris, 2008; Torgesen &Hudson, 2006) and often at the expense of other as-pects of fluent reading, such as phrasing, appropri-ate pacing, stress, and emphasis (e.g., Kuhn & Stahl,2003). Although these elements are central to fluentreading, they are by no means the only elements criticalto the process. By focusing on these elements over thepast decade, to a large extent through the dominanceof Dynamic Indicators of Basic Early Literacy Skills(DIBELS; Good & Kaminski, 2002), and similar assess-ments such as AIMSweb (Shinn & Shinn, 2002) andcurriculum-based measurements (CBMs; Deno, 1985)in the classroom, rate measures such as the DIBELS

Oral Reading Fluency have become privileged, driv-ing the literacy curriculum (e.g., Riedel, 2007; Samuels,2007). Given that this perspective presents a limitedview of fluency, it is essential that reading educatorsconsider a broader definition of the construct, one thatplaces weight on its less quantifiable elements.

Fluency as ProsodyThe National Assessment of Educational Progress(NAEP; Daane, Campbell, Grigg, Goodman, & Oranje,2005; Pinnell et al., 1995), on the other hand, viewsoral reading performance as an important indicator of

skilled reading. However, while it includes measure-ments of accuracy and rate as part of its evaluation, itparcels out fluency as a distinct component, definingit as phrasing, adherence to the authors syntax, andexpressiveness (Daane et al., 2005, p. v). The result ofthis wording is that f luency becomes equated with mostworking definitions of prosody (e.g., Kuhn & Stahl,2003; Schreiber, 1991; Torgesen & Hudson, 2006).

At first, it was unclear why the authors of the NAEPassessment would make this distinction, but an expla-nation may be found in the historical context surround-ing the measure. The 1992 NAEP was one of the firstlarge-scale evaluations of oral reading performance, un-

dertaken at a time when f luency was largely a neglectedcomponent in the reading process (e.g., Allington, 1983;Dowhower, 1991). In those few cases where fluency wasconsidered, it was primarily in terms of rate and accu-racy and was generally measured as the number of cor-rect words read in a minute (e.g., Kuhn & Stahl, 2003).One of the goals of the original NAEP evaluation of oralreading performance was to describe those aspects oforal reading that go beyond accuracy and rate whichthe authors felt may have wide applicability for reading

educators (Pinnell et al., 1995, p. 2). By designing theOral Reading Fluency Scale, the authors hoped to coun-terbalance some of the overemphasis on rate and accu-racy and to integrate oral language elements into thediscussions that surround oral reading performance.

What is interesting is that, when looking across arange of discussions that have taken place around theconstruct of fluency, both before the initial NAEP pub-lication (Pinnell et al., 1995) and increasingly since(e.g., Kuhn & Stahl, 2003; Zutell & Rasinski, 1991),it becomes apparent that there is a recognition of theimportance of prosodic elements in most definitionsof fluency (e.g., Hudson et al., 2009; NICHD, 2000).

Whether that acknowledgment is as robust as we mightwant is something we are attempting to address here.However, we would argue that a definition that sepa-rates rate and accuracy from prosody reinforces the po-sition that correct words per minute can be treated asan isolated measure of oral reading performance. As aresult, we consider it important to keep an integrated

definition that includes accuracy, speed, and prosody.

Fluency as Skilled ReadingA third definition of reading fluency equates it withskilled reading. According to Samuels (2006), the mostimportant characteristic of the f luent reader is the abil-ity to decode and to comprehend the text at the sametime (p. 9) and other characteristics of f luency suchas accuracy of word recognition, speed of reading, andthe ability to read orally with expression (p. 9)simplyserve as indicators that f luency has been achieved. Thisdefinition initially holds great appeal; by including text

comprehension within the definition of fluent reading,it becomes possible to differentiate two groups of stu-dents; word callers, who simply read words, or bark atprint (Samuels, 2007), without attending to the mean-ing, and fluent readers who construct meaning from thetext as they read. Although word callers are not ubi-quitous (Meisinger, Bradley, Schwanenf lugel, Kuhn, &Morris, 2009; Meisinger, Bradley, Schwanenflugel, &Kuhn, in press), their numbers do increase across theelementary grades. As such, it seems reasonable to pre-suppose that instruction that focuses on speed and ac-curacy of word identification with little or no regard tounderstanding will further serve to inflate their num-

bers (Applegate, Applegate, & Modla, 2009; Pressley,Hilden, & Shankland, 2006).

However, this broad definition of fluency gives uspause. Skilled reading is a complicated act that requiresthe coordination of input from multiple sources, in-cluding syntactic knowledge, background knowledge,vocabulary knowledge, orthographic knowledge, andaffective factors, among others (e.g., McKenna & Stahl,2003; RAND Reading Study Group, 2002), that al-lows the reader to construct meaning from text. Rather

7/29/2019 Kuhn-et-al.-RRQ-2010

11/22


than defining fluency as simultaneously decoding andcomprehending (Samuels, 2007), it can be argued thatfluent reading merely allows comprehension to occur(e.g., Levy, 2001). Just as readers fluency can vary withvarious texts (e.g., Allington, 2009; Hiebert, 2006), thatis, readers may be able to read independent level textswith good fluency yet be disfluent when reading textsthat are challenging in terms of vocabulary or content,it is also possible for readers comprehension of difficulttexts to vary despite their reading of these texts withadequate fluency.

For example, lets consider what happens when youread a complicated text, such as a theoretical paper. Asa reader with a strong background in the subject, youare likely to read that text fluently, that is accurately, ata good rate, and with appropriate parsing and cadence.But it is also likely that you will have only surface-level comprehension on the initial reading. However,by rereading that text and grappling with its mean-ing, you will deepen your understanding of the mate-

rial (Pressley, 2000). Similarly, even with relativelyeasy texts, say C.S. Lewiss The Lion, the Witch and theWardrobe, you and another reader with similar levels offluency may develop highly differing interpretations ofthe text depending on your varying background knowl-edge. At the same time, we do agree that fluent read-ing without any concomitant comprehension would bemerely word calling. So how do we rectify these po-tentially disparate understandings? Although it is rea-sonable to expect a basic level of comprehension beforeconsidering an individuals reading fluent, if only toprevent the term from being equated with surface-levelfeatures (accuracy, speed, and expression), it is criticalnot to confound the two constructs given the complexi-ties of both.

Fluency as a Bridge to ComprehensionThe final definition we consider here views fluency asa bridge between decoding and comprehension (Chard,Pikulski, & McDonagh, 2006;Pikulski & Chard, 2005).This position indicates that fluency likely has a recipro-cal relationship with comprehension, both contributingto and possibly resulting from readers understandingof text (e.g., Klauda & Guthrie, 2008; Stecker, Roser,& Martinez, 1998). It also accounts for the theoretical

discussions surrounding automaticity and prosody, in-dicating that both aspects of the construct facilitate, andbenefit from, comprehension. Further, this definitionmoves away from what the authors call a surface con-ceptualization of fluency; such an understanding seesthe construct primarily as an oral reading phenomenonand, as a result, has a tendency to stress its more con-crete elements of accuracy, rate, and prosody throughboth assessment and instruction (Chard et al., 2006;Pikulski & Chard, 2005). Given most reading is silent,

rather than oral, this insight is particularly important.Indeed, children are thought to make the transition topredominantly silent reading during late elementaryschool (Prior & Welling, 2001), generally around fourthgrade.

Our Definition

Having reviewed multiple ways of conceptualizingreading f luency, we propose the following definition tosynthesize the information presented thus far:

Fluency combines accuracy, automaticity, and oral reading

prosody, which, taken together, facilitate the readers con-

struction of meaning. It is demonstrated during oral read-

ing through ease of word recognition, appropriate pacing,

phrasing, and intonation. It is a factor in both oral and silent

reading that can limit or support comprehension.

Although this definition is clearly influenced by thosepresented elsewhere (e.g., Harris & Hodges, 1995;

Pikulski & Chard, 2005; Reutzel, 1996), it attempts toincorporate several critical points. First, it highlightsthe relationship between fluency and comprehension.Next, it emphasizes prosody along with accurate andautomatic word recognition without privileging any ofthese components. Third, it begins to address the un-derstanding that fluency plays a role in silent as wellas oral reading. Finally, it attempts to reconceptualizetwo aspects of the construct that have the potential tobe problematic when taken in isolation from the rest ofthe components: rate and expression. We also recognizethat there may be a reciprocal relationship between flu-ency and comprehension; however, this issue requiresfurther research. As such, we have chosen not to in-clude reciprocity in our definition.

When discussing oral reading fluency in terms ofassessment and instruction (e.g., Mathson, Allington, &Solic, 2006; Samuels, 2007), there has been a tendencyto focus on decoding speed at the expense of prosody.This results in students being encouraged to read as fastas possible rather than at a rate that replicates oral lan-guage; the hope is that the use of the term appropriatepacing over rate has the potential to begin addressingthis misconception. The second aspect that can leadto problems of interpretation is the use of the term ex-

pressive reading as an equivalent of prosody. Althoughexpression is an accurate term for many types of texts(e.g., narratives, plays, poetry), it has been argued that itis inappropriate for informational text. However, thesetexts have their own prosodic indicators (e.g., Carlsonet al., 2009; den Ouden et al., 2009), so it may be thatthe term intonation is more precise than is expressionfor describing the suprasegmental features that occuras part of the reading of informational text. It may bethat the use of these alternative terms to describe f luent

7/29/2019 Kuhn-et-al.-RRQ-2010

12/22


reading will help to counter certain misunderstandingsthat have developed around the construct.

Current Assessment of FluencyGiven the aforementioned review, it is important to ex-pand the discussion to the assessmentand practice

of fluency as it is currently being implemented in manyschool districts across the United States (e.g., Riedel,2007). Since the introduction of No Child Left Behindand Reading First (government mandated educationalreforms), the instructional landscape in the UnitedStates has undergone a major shift (e.g., Cervetti,

Jaynes, & Hiebert, 2009; Garcia & Bauer, 2009). Therehas been an attempt to refocus literacy education on fiveareas of l iteracy development reviewed by the NationalReading Panel (NICHD, 2000): phonemic awareness,phonics, fluency, vocabulary, and comprehension. Thishas been coupled with a new emphasis on regular as-

sessment and scientifically based reading research. Ourgoal here is not to critique Reading First, per se (seeConnor, Jakobsons, Crowe, & Meadows, 2009; Dubin,2008; Gamse, Bloom, Kemple, & Jacob, 2008; Teale,Paciga, & Hoffman, 2007 among others for discussionsof the impact of Reading First), but to instead look at theways in which fluency assessment and instruction havebeen affected by conceptualizations that have domi-nated educational practice since the inception of thislegislation.

While Reading First focused on five areas of readingas critical to skilled reading development, two of those

five, vocabulary and comprehension, are significantlymore complex and, therefore, more difficult to measure.As a result, designing assessments that readily demon-strate student growth in these areas has been somewhatproblematic (e.g., McKenna & Stahl, 2003; Paris, 2008).One result of this difficulty has been a greater emphasison those areas that are easy to measure (Duffy, 2007;Paris, 2008): phonological awareness, the alphabeticprinciple, and oral reading fluency, or what have beenreferred to as the big ideas (Good et al., 2001, p. 7) ofbeginning reading. In fact, when discussing these con-cepts, Good and his colleagues title the section of theirpaper that focuses on these components Measuring

whats important: The foundational skills of beginningreading (p. 6).

Although these factors are among the crit ical un-derstandings that students must establish if they areto become successful readers, this list needs to be pur-posefully expanded to provide a better sense of thecomplexities of beginning reading; as such, factors thatemphasize oral language, motivation, extensive oppor-tunities to read and interact with connected text, and arange of other skills that contribute to vocabulary and

comprehension development should also be included aspart of a balanced reading curriculum (Bredekamp &Pikulski, 2008; Pikulski, 2005; Schwanenflugel et al.,in press; Shanahan, 2005).

So is the reason for the emphasis on three of the fivecomponents simply the result of their ease of assess-ment? In our opinion, to some degree, yes. Accordingto Paris (2005, 2008), reading skills can be classifiedalong a continuum of constrained and unconstrainedskills. Constrained skills develop over a relatively briefperiod of time, incorporate a limited set of knowledgeand skills, can be taught directly, and can be readily as-sessed quantitatively. Further, these skills are importantbecause they enable the development of unconstrainedskills to occur in relation to text. When placing skillsalong this continuum, Paris (2008) argued that phono-logical awareness, phonics, and oral reading fluency areconstrained, or in the case of oral reading fluency some-what constrained, and that vocabulary and compre-hension are unconstrained. And it is also the case that

the testing of constrained skills is both uncomplicatedand inexpensive, allowing students to show significantgains over short periods of time. As a result, it is easy forthem to become the focus of attention.

However, ease of measurement is only one reasonthat the big ideas (Good et al., 2001, p. 7) of begin-ning reading have gained dominance in many schoolsreading curriculum. A more fundamental reason isthat a number of researchers (e.g., Fletcher et al., 2007;Kameenui, Simmons, Good, & Harn, 2001) considerthese skills to be integral to later reading success. Inthis line of thought, if learners encounter difficultieswith these skills early on, they are increasing theirlikelihood of developing later reading difficulties dra-matically. According to Good and his colleagues (2001),differences in developmental reading trajectories canbe explained, in part, by a predictable and consequen-tial series of reading-related activities that begin withdifficulty in foundational skills (p. 6). To circumventthis problem, it is important to identify any weaknessesthat students are experiencing with these skills earlyand provide intensive instruction in the correspondingareas. And the best way to determine whether studentsare making appropriate progress is through regularassessments. CBM (Deno, 1985), along with its com-

mercially available variants (e.g., DIBELS [Good &Kaminski, 2002], AIMSweb [Shinn & Shinn, 2002]),have come to the fore as a means of accomplishing thisgoal. Further, these measures have become highly in-fluential in informing early reading instruction in gen-eral and fluency instruction in particular.

Curriculum-Based Measurements (CBMs)CBM was originally designed to evaluate studentsgeneral reading progress by measuring the number of

7/29/2019 Kuhn-et-al.-RRQ-2010

13/22


correctand incorrectwords read aloud in one min-ute (e.g., Deno & Marston, 2006; Madelaine & Wheldall,1999, 2004; Samuels, 2007). However, they have sincebeen adapted for use as a measure of oral reading f lu-ency as well. The initial drive behind these assess-ments was to provide teachers with a quick alternativeto norm-referenced standardized tests (Madelaine &

Wheldall, 1999). A number of reasons have been citedfor this decision, including standardized measures lackof technical adequacy (e.g., issues surrounding contentvalidity), their insensitivity to small changes in learn-ers development, their inappropriateness as a meansof tracking students progress or as a basis for instruc-tional decision making, and a tendency toward the mis-use of the data that norm-referenced standardized testsprovide.

CBMs, on the other hand, are meant as an alter-native that incorporates standardized procedures, butprovides teachers with information that is reliableand valid, quick and easy to administer repeatedly, in-

expensive, unobtrusive, sensitive to small changes inprogress, and able to be used to make instructional de-cisions (Madelaine & Wheldall, 1999, p. 74). Studieshave indicated that the use of these measures as a meansof tracking learners reading development can lead toimprovements in reading achievement (Deno, 2003;Fuchs, Deno, & Mirkin, 1984; Stecker & Fuchs, 2000;

Wayman, Wallace, Wiley, Tichdt, & Espin, 2007), andthere is evidence that they correlate highly with stan-dardized tests of reading comprehension as well (e.g.,Deno & Marston, 2006; Fuchs, Fuchs, & Maxwell,1988).

Initially, CBMs were developed as a means of evalu-ating learners progress on passages drawn directly fromtheir curriculum, a procedure that allowed teachers di-rect insight into their students ability with the mate-rial they were using in the classroom (Deno & Marston,2006). While researchers note clear advantages to thisapproach, they discuss disadvantages as well, includingthe amount of time required to identify reading pas-sages and the variability in difficulty across, and evenwithin, texts. To rectify these issues, passages identifiedat a given reading level but selected from material out-side the curriculum have been used for these measuresas well (Powell-Smith & Bradley-Klug, 2001). This shiftaway from specific, classroom-based literacy curriculaalso laid the groundwork for commercial versions oforal reading fluency assessments, the best known ofwhich is DIBELS (Good & Kaminski, 2002).

Dynamic Indicators of Basic EarlyLiteracy Skills (DIBELS)

As with CBMs, the DIBELS oral reading fluency mea-sures the number of correct words students can read in

one minute. According to the DIBELS website (dibels.uoregon.edu/samples/index.php), the oral reading flu-ency measure, along with other measures that are partof the DIBELS data system (Good & Kaminski, 2002),is in use at over 15,000 schools, making it likely themost frequently used single assessment of connected-text reading fluency in the United States today. TheDIBELS tests provide a developmental timeline andcorresponding benchmarks for skills acquisition thatallows teachers to determine a developmental trajectoryfor each student. The measures include initial-soundfluency, letter-naming fluency, phoneme segmentationfluency, nonsense word fluency, oral reading fluency,retell fluency, and word use fluency. Taken together,these assessments are meant to be easy and inexpen-sive to administer, effective at identifying students whoare likely to experience later reading difficulty basedon their progress on a series of constrained skills, anddesigned to provide data that can serve as the basis forinstructional decision making.

At this point, it is useful to note that the authors ofDIBELS (Kameenui et al., 2001) define fluency differ-ently from the way we have discussed it thus far. Ratherthan employ what they term a traditional definition offluency, that of proficient word recognition in the read-ing of connected text, they modify the definition to in-clude fluency in the component skills and lower-levelprocesses (p. 308). This understanding translates intoautomaticity in phonemic awareness, letter recognition,and decoding and accounts for the term being used inconnection with all the DIBELS measures, not just oralreading fluency of connected text. However, this un-derstanding of fluency as automaticity is also integral tothe DIBELS oral reading fluency, which is described asa measure of the accuracy and f luency of connected-textreading. This results in the DIBELS actually narrowing,rather than expanding, the understanding of fluencyso that the term becomes a synonym for automaticity,even as it is applied to a broader range of concepts thanconnected-text reading. As Hudson and her colleagues(2009) succinctly argued, the concept of automaticityactually implies more about a response than does theconcept of f luency; accordingly, they retain the termautomaticity, rather than fluency, to describe a responsethat requires few processing resources, is obligatory,

and outside of conscious control (p. 9).

The Use of CBMs and DIBELS in PracticeA key premise of both CBMs and the DIBELS oral read-ing fluency is that students reading rate and accuracyare effective proxies for general reading ability (e.g.,Deno & Marston, 2006; Fuchs et al., 2001; Samuels,2007). As such, they are seen as a means of trackingstudents reading development and as the basis fordetermining whether students are receiving effective

7/29/2019 Kuhn-et-al.-RRQ-2010

14/22


instruction. In addition, the DIBELS oral reading flu-ency, along with other versions of CBMs, is seen as anindicator of connected-text fluency (Good et al. 2001).These scales have established benchmarks designed todetermine learners risk level in relation to reading de-velopment (Good & Kaminski, 2002; Shapiro, 2004).

And the use of these measures is seen as a valuablemeans of helping students avoid later reading difficul-ties and the negative cycle that develops as a result ofunsuccessful early experiences with print. As Good andhis colleagues (2001) stated, few would argue with theconcept of prevention and the need for formative assess-ment to inform instruction (p. 9).

Indeed, there is little arguing with the desire to pre-vent reading difficulties (Snow, Burns, & Griffin, 1998)or of the importance of using appropriate assessmentsto inform instruction (Duffy, 2007). Nor is there anydoubt that ensuring students have extensive experienc-es with text and appropriate forms of instruction, someof which focus on the development of constrained skills,

will prevent many students from experiencing laterreading difficulties (e.g., Cunningham & Stanovich,1998; Shanahan, 2005). However, whether the assess-ments discussed earlier are the best means for helpinglearners meet these goals is significantly more problem-atic. The answer depends, to a large extent, on how theassessments measure and define fluency and on formsof instruction that are used as a result. So, for example,if the emphasis is on automaticity (e.g., Kameenui etal., 2001) or rapid decoding (Shinn et al., 1992 citedin Madelaine & Wheldall, 1999, p. 76), either as part offluencys working definition (e.g., oral reading fluency

or the oral translation of text with speed and accuracyFuchs et al., 2001, p. 239) or as part of its measurement(e.g., Fletcher et al., 2007; Torgesen & Hudson, 2006),there is, almost inevitably, a corresponding privilegingof speeded decoding in its instruction (Applegate et al.,2009; Pressley et al., 2006). Further, the importancethat administrators, teachers, and other constituentscurrently assign to these measures, coupled with theirrepeated use over the course of the elementary schoolyears, has intensified these issues (Paris, 2005, 2008).This excessive focus on rate can lead to fast, staccatoreading rather than reading with appropriate pacingand may actually interfere with, rather than promote,

comprehension (Samuels, 2007).Because excessive rate impedes comprehension, ei-

ther by shifting the focus away from understanding orby actually interfering with the construction of mean-ing, most researchers (e.g., Fletcher et al., 2007; Hudsonet al., 2009; Rasinski et al., in press) consider appropri-ate or conversational pacing, along with other prosodicfeatures, as central to their definition of fluency. So whydoes this understanding not translate to assessment?It appears that measuring prosody is considered to be

somewhat problematic (e.g., Fuchs et al., 2001). Threeprimary concerns underlie this perception (Torgesen& Hudson, 2006). The first involves prosodys am-biguous relationship with comprehension; estimates ofprosodys contribution to comprehension beyond thataccounted for by rate measures have ranged from smallto moderate (Schwanenflugel et al., 2004; Miller &Schwanenflugel, 2006, 2008). Second, those measuresof prosody that are readily implemented in classrooms,such as the NAEP fluency scale (Pinnell et al., 1995)or the multidimensional f luency scale (Rasinski et al.,2009), are far less precise than are measures of correctwords per minute. This means the results from thesemeasures are less sensitive to small, ongoing changesin fluency (e.g., Klauda & Guthrie, 2008). Finally, mea-sures of prosody have the highest levels of reliabilitywhen they include a measure of reading rate as well,making their implementation somewhat redundant(Torgesen & Hudson, 2006).

Despite their shortcomings, compelling arguments

can be made for the use of f luency scales. According tothe NAEP analysis of oral reading (Daane et al., 2005;Pinnell et al., 1995), all three elements of fluencyaccuracy, rate, and prosodyare related not only toone another, but also to overall reading comprehension.That is, students with higher NAEP ratings (levels 3 or 4on the NAEP oral reading f luency scale) not only tendedto read texts with a higher degree of accuracy and at afaster rate, but they also had a higher score in terms oftheir overall reading proficiency. Their peers with lowerNAEP ratings (levels 1 or 2 on the NAEP oral readingfluency scale), on the other hand, read fewer words per

minute, had a higher percentage of miscues, and hada lower overall reading proficiency score. In addition,the appropriate use of prosodic elements appears to re-flect a readers comprehension of a text (Mathson et al.,2006). And, although fluency scales do involve qualita-tive judgments, several researchers (e.g., Kuhn, 2005;McKenna & Stahl, 2003) found high levels of inter-raterreliability after brief training on these measures; in fact,they have established levels as high as 100% using theNAEP scale, perhaps because the NAEP has the broad-est descriptive categories, increasing the likelihood ofagreement. Although current fluency scales are not asprecise as we might wish, they do provide additional

insight into students reading development.Ultimately, it is essential to expand the way fluency

is measured so that it encompasses more than rate andaccuracy. According to Deno and Marston (2006) thedefinition of fluency should not be limited to correctwords per minute, because this understanding leavesout important features of the construct, such as pros-ody. We would argue that such an emphasis leaveschildren in danger of focusing on speed at the expenseof comprehension (see also Samuels, 2007; Wixson &

7/29/2019 Kuhn-et-al.-RRQ-2010

15/22


Lipson, 2009). In fact, Samuels (2007) correctly arguedthat curriculum-based measures were conceived of as ameans of monitoring students reading progress, broad-ly considered, and that their use as a fluency measureleads to an overemphasis on speed at the expense ofunderstanding (p. 565). Wixson and Lipson (2009)concurred, arguing that it is the use of these assess-ments as both screening and progress monitoring mea-sures that leads to such an instructional focus.

Further, Deno and Marston (2006) found the notionof benchmarks for reading rates to be highly problematicbecause it relies on an oversimplification of the relation-ship between word recognition and comprehension andimplies there is a point beyond which text comprehen-sion is guaranteed. In fact, Hudson and her colleagues(2009) noted that there are times when a slower readingrate is necessary to ensure the construction of meaning.

While it is true that exceedingly slow word recognitionhinders comprehension and that skilled readers wordrecognition is automatic, it is also the case that skilled

readers vary their reading pace depending upon the dif-ficulty of the text and the complexity of the ideas theyare encountering. Given this, if learners are to becomeskilled readers, it is important that they learn to be flex-ible, rather than simply fast, oral readers. By includingmeasures of prosody as part of the evaluation process,the likelihood that learners develop the mistaken no-tion that fluent reading and fast reading are one andthe same decreases. As such, it seems the positives ofincluding measures of prosody along with a measure ofrate and accuracy outweigh the negatives.

Implications for Assessmentand Instruction

Implications for AssessmentGiven what we know about fluency assessment andwhat we hope to see in fluency instruction, what dowe propose? First, it is essential that f luency be seen asmore than simply correct words per minute. Withoutthe addition of some measure of prosody, there contin-ues to be too high a risk that oral reading fluency willbe seen only as a measure of quickly decoding a pas-

sage (e.g., Samuels, 2007) and that instruction will con-tinue to follow suit (Wixson & Lipson, 2009). For now,prosodic measures such as the NAEP oral reading flu-ency scale (Pinnell et al., 1995) or the multidimensionalfluency scoring guide (Rasinski et al. 2009; Zutell &Rasinski, 1991) can serve as a rough gauge of how wellstudents are integrating the suprasegmental features oflanguage into their oral reading. We further think thatimprovements will be made in such rating measures aswe gain a more specific understanding of the linkages

between identifiable spectrographic elements of proso-dy and comprehension. We believe that research shouldgo in the direction of creating prosody rating schemesthat preserve the ease and general utility of ratings withthe refinement of spectrographic measures.

Second, it remains critical that students are not fo-cusing on rate at the expense of meaning; to preventoveremphasizing rapid decoding, a measure of compre-hension should be used in conjunction with any evalu-ation of reading fluency (Samuels, 2006). This can beundertaken in several ways, from brief discussions ofthe passage being read to answering a range of ques-tions, from factual to inferential, which are related tothe material to student retellings of the text (McKenna& Stahl, 2003). However, although there are a range ofpossibilities available for evaluating comprehension,simply asking students to reiterate as many words asthey can remember after the oral reading of a passagefails to reflect any evidence of the processes they may beusing to construct meaning (e.g., Pressley et al., 2006;

Samuels, 2007). As such, it is important to encouragethe use of a comprehension measure that allows learn-ers to demonstrate understanding rather than simplyrecite words.

Third, although it is important to evaluate stu-dents oral reading (e.g., Daane et al., 2005; McKenna& Stahl, 2003), this is only one piece of information ina readers profile. Despite evidence that there are oftenhigh correlations between f luency measures and stan-dardized comprehension measures (Daane et al., 2005;Deno & Marston, 2006; Fuchs et al., 2001; Madelaine&Wheldall, 1999, 2004), the correlations are not per-

fect. In fact, looking at research conducted across arange of populations and a variety of standardized com-prehension assessments, we noted correlations rangingfrom a low of 0.61 (Sibley, Biwer, & Hesch, 2001) to ahigh of 0.91 (Fuchs et al., 1988) for curriculum-basedmeasures and from 0.45 (Pressley et al., 2006) to 0.80(Riedel, 2007) for DIBELS. This can cause a number ofstudents to be misidentified, either as having readingdifficulties when they do not or as making sufficientprogress in their reading development when, in fact,they are struggling (e.g., Riedel, 2007).

To decrease the likelihood of misidentifying studentachievement levels, students fluency results should be

considered as part of a broader range of assessments(e.g., McKenna & Stahl, 2003) and classroom-baseddata (Afflerbach, 2004; Glasswell & Teale, 2007). Italso makes sense to look at these measures more quali-tatively; that is, what types of miscues are the readersmaking and in what context (e.g., McKenna & Picard,2006), how does the readers rate vary with the type oftext and its instructional level (Kuhn, 2007), and howappropriate is their prosody for the text they are reading(Rasinski, 2004)?

7/29/2019 Kuhn-et-al.-RRQ-2010

16/22


Implications for InstructionIn line with our suggestions that assessment of read-ing fluency should be multifaceted, we present alterna-tives to lessons that stress automaticity as the end goalof fluent reading (e.g., Mathson et al., 2006; Wixson &Lipson, 2009). This article began by highlighting theshift that has taken place around fluency and its role

within the literacy curriculum over the past decade.During this period, fluency, seen primarily in termsof rate measures, has become a driving force in read-ing instruction. Although fluent reading is critical tolater reading success (Kuhn & Stahl, 2003; Rasinskiet al., in press), it is only one component of literacylearning. Effective instructional approaches for flu-ency development such as fluency-oriented reading in-struction (FORI; Stahl, Heubach, & Holcomb, 2005),wide fluency-oriented reading instruction (Wide FORIor Wide Reading; Kuhn et al., 2006) and the fluencydevelopment lesson (FDL; Rasinski, Padak, Linek, &

Sturtevant, 1994) view the comprehension of texts,rather than an increase in reading rate, as the primarygoal. These approaches all recognize that the develop-ment of automaticity, prosody, and reading comprehen-sion occur through the scaffolded reading of a range oftexts. It is the various forms of supported reading (forexample, echo, choral, partner, and repeated reading)that allow learners to engage with and learn from thematerial they are reading.

The types of supported reading that comprise effec-tive fluency practices, such as FORI, Wide FORI, andFDL, also integrate and further develop the componentskills that lead to automatic word recognition (e.g.,Kuhn et al., 2006; Samuels, 2006). Although instruc-tion in constrained skills, such as phonemic aware-ness and word recognition, provide a critical base onwhich to develop automaticity (e.g., Bear & Templeton,1998; Levy, 2001; Paris, 2008), the practice with con-nected text provided by these approaches allows learn-ers to consolidate these components. Further, there isevidence that an overemphasis on word instruction inisolation (Allington, 1983, 2009; Chomsky, 1976) canactually work against students development as skilledreaders. It is equally critical to remember that the rela-tion between childrens basic reading skills (e.g., word

reading and reading fluency) and reading compre-hension diminishes as children age (Meisinger et al.,2009; Schwanenflugel et al., 2006; Vellutino, Fletcher,Snowling, & Scanlon, 2004). As young readers developautomatic word reading skills, attentional resources arefreed for comprehension processes (LaBerge & Samuels,1974; Perfetti, 1985). However, at some point the issueshifts from managing freed resources to using contentknowledge, accessing the meanings of sophisticatedvocabulary, drawing appropriate inferences, and moni-

toring comprehension progress (Chall, 1996; Sweet &Snow, 2003).

Finally, it is essential that students read substan-tial amounts of connected text if they are going to be-come fluent readers (Logan, 1997; Stanovich, 1986).

Although an effective literacy curriculum will includea wide range of materials, including poetry and otherrelatively brief texts, if these are the only texts that stu-dents are reading, they will not provide learners withsufficient practice to develop their fluency, regardless ofhow repeatedly they are read (Schwanenflugel, Kuhn etal., 2008). Nor is it sufficient to read a single longer text,say a narrative or expository trade book or a selectionfrom a basal reader or literature anthology designed forsecond or third graders, if that text is read only onceover the course of a school week (Hiebert, 2004). Andalthough independent reading is central to developingreading fluency, students who are averse to reading areunlikely to benefit from DEAR or SSSR (Hasbrouck,2006) unless they are provided with a range of options

such as scaffolded silent reading (Reutzel, Fawson, &Smith, 2008), partner reading (Meisinger & Bradley,2008), reading-while-listening (Chomsky, 1976; Pluck,2006) or other forms of assisted reading along with tra-ditional independent silent reading (for a more compre-hensive review of fluency instruction, see Rasinski etal., in press). What is critical here is that learners areprovided with extensive opportunities to engage withconnected texts, whether they are reading repeatedly orwidely, and that sufficient support is provided to allowstudents to succeed given the level of challenge that ispresented by various texts.

ConclusionsThe title of this article implies that we need to alignour assessment practices with our theories of readingfluency. At the basic level, we know that f luency incor-porates automaticity and prosody (e.g., Erekson, 2003;Logan, 1997; Samuels, 2004). We also know that flu-ent reading facilitates comprehension and that compre-hension may mediate aspects of fluency such as pacing(e.g., H

kuhn-et-al.-rrq-2010

Documents