Beyond the bridge: an integrative developmental framework for first-language reading fluency

1 Introduction

Reading fluency is readily observable but conceptually unstable. Hesitant, fragmented reading contrasts clearly with continuous, well-organized text processing, yet fluency has been defined as rapid processing, accurate and automatic reading, expressive oral performance, and a multidimensional capacity (Kuhn et al., 2010; Wolf and Katzir-Cohen, 2001). These definitions assign different weight to observable performance and imply different relations between fluency and the wider reading system. The central problem is therefore not only how fluency should be measured but also what the construct comprises and which structural position it occupies as reading develops.

The bridge metaphor connects increasingly efficient word recognition with successful meaning construction. Automaticity theory proposes that effortful word identification can disrupt continuity and constrain the processing available for understanding text, whereas reliable recognition permits more sustained engagement beyond the individual word (LaBerge and Samuels, 1974; Samuels, 2006). The bridge formulation preserves this insight while emphasizing that fluent reading cannot be reduced to speed alone (Pikulski and Chard, 2005). It is especially informative during the transition from effortful word identification to sustained text processing.

A fixed intermediate position does not capture the full empirical pattern. Longitudinal comparisons of outcome and mediator models have supported mediation, with developmental qualifications (Kim et al., 2021; Kim and Wagner, 2015). Silent reading fluency predicted comprehension earlier in a long-term Finnish study, whereas a smaller reverse relation emerged later (Psyridou et al., 2023). A recent scoping review identified four recurring structural positions: outcome, direct contributor, mediator, and part of reciprocal or developmentally variable relations (Steklács and Vuong, 2026). The present question concerns the processing conditions under which each position becomes informative.

The instability also concerns modality and measurement. Oral and silent reading share substantial processes, but their overt demands and observable indicators are not equivalent. Direct comparisons show that their relations and relative fluency depend on reader group and the tasks used (Denton et al., 2011; van den Boer et al., 2022). Timed oral-reading scores are useful indicators, yet a rate–accuracy composite does not represent fluency in the same way as a broader component profile (Valencia et al., 2010). Apparent stability in a score may therefore conceal changes in both the processes represented and the manifestation through which they are observed.

This article proposes that fluency development changes how processing time is organized across connected text. As word identification becomes more reliable, timing can become increasingly responsive to linguistic integration demands and reading goals. Three configurations organize this hypothesis: fluency as an emerging outcome of word-recognition development, a mediator and coordinator of connected-text processing, and part of reciprocal, task-dependent relations with meaning construction. These configurations overlap, and their relative prominence may differ across readers, languages, modalities, texts, and tasks. Their proposed ordering concerns changing constraints rather than fixed stages.

The proposed framework characterizes fluency as multidimensional but bounded. Its core dimensions are accuracy, temporal continuity and pace regulation, and, in oral reading, prosodic organization. Connected-text coordination is proposed as an explanatory account of how successive word-identification events relate temporally to linguistic grouping and local semantic updating. The dimensions describing performance are distinguished from this explanation, from supporting resources, and from broader comprehension outcomes.

Efficient word recognition is foundational to connected-text processing, while broader comprehension additionally involves inferencing, knowledge integration, evaluation, and situation-model construction. The framework predicts that, as identification consolidates under comparable conditions, global rate becomes less informative about comprehension while demand-sensitive local timing retains or increases its informativeness. This joint prediction connects changes in processing constraints with changes in what fluency measures explain.

The article clarifies the boundaries of reading fluency, specifies candidate processes underlying its developmental reorganization, and derives a prioritized research agenda. Its proposed contribution links reduced disruption from word identification to selective timing adjustments and their changing relationship with comprehension. Developmental variability, component interaction, and regulation provide established foundations for this more specific explanatory and predictive commitment.

The account integrates automaticity, synchronization, word-to-text integration, and developmental component models (Breznitz, 2006; Kim, 2015, 2020; Perfetti and Stafura, 2014). Sections 2–6 develop the argument from published developmental, cross-orthographic, modality-sensitive, and measurement-focused evidence. Section 7 presents the architecture and a prioritized research agenda. Sections 8–9 compare the framework with existing explanations and consider its implications and limits. The focus is first-language (L1) reading development, with compensatory profiles providing relevant constraints on the proposed relations.

2 From the bridge role to a developmental problem

2.1 Four structural positions of reading fluency

A scoping review of recent L1 research identified four recurrent structural positions: fluency as an outcome of foundational word-reading development, a direct contributor to comprehension, a mediator between word-level processes and comprehension, and part of reciprocal or developmentally variable relations (Steklács and Vuong, 2026). These positions describe the function assigned to fluency within theoretical or statistical models. They are not competing definitions of the construct.

In the outcome position, increasingly accurate and efficient word recognition makes text reading less fragmented, and fluency emerges as its connected-text consequence. In the direct-contributor position, fluency relates to comprehension beyond isolated word-reading efficiency. Consistent with this position, oral text-reading fluency has served as an indicator of broader reading competence and a predictor of comprehension, although prediction alone does not establish a causal mechanism (Fuchs et al., 2001; Jenkins et al., 2003).

The bridge role is most explicit in mediation models, where efficient word reading supports comprehension partly through the capacity to sustain and coordinate processing across connected text. Developmental findings indicate that this mediating relation changes as the contributions of word reading and language comprehension are reorganized (Kim, 2015; Kim et al., 2014). Statistical mediation nevertheless identifies a relational pattern rather than the psychological process that produces it.

The reciprocal or developmentally variable position allows direction and relative strength to change. Fluency may predict subsequent comprehension during one period, while comprehension or interpretive processing later contributes to prosodic organization or fluent performance (Veenendaal et al., 2016; Zhao et al., 2025). Such findings do not imply universally balanced two-way effects, but they challenge a permanently fixed unidirectional relation.

The developmental problem is therefore to explain when and under what conditions each position becomes most informative, rather than to select one as universally correct. Figure 1 summarizes the four structural positions.

The four structural positions and the three configurations proposed later operate at different analytic levels. The former summarize how studies locate fluency within a model. The latter organize hypotheses about changes in the relative constraints on connected-text processing. A direct contribution to comprehension can coexist with mediation or reciprocity and therefore does not require a fourth configuration.

2.2 Why the bridge metaphor remains incomplete

The bridge remains valuable because it captures a consequential transition: efficient word recognition must be organized across connected text to support comprehension. Its limitation lies in the structural stability it implies. Reading development changes not only performance within word recognition, fluency, and comprehension but also the weight, direction, and informational meaning of their relations. The traditional formulation also privileges forward directionality, although active self-regulation allows developing interpretations to influence ongoing reading (Duke and Cartwright, 2021).

A single intermediate position also obscures modality differentiation. Oral reading adds articulation, eye–voice coordination, and audible prosodic organization. Silent reading permits greater local variation in pace and rereading. Cross-sectional comparisons show substantial overlap and age-group differences, including a larger silent-reading advantage in adolescents than in children (van den Boer et al., 2022). The same bridge representation may therefore refer to non-equivalent oral and silent manifestations.

Finally, stable measurement can create the appearance of a stable construct. The observed fluency–comprehension relation varies with developmental phase, orthographic depth, and measurement characteristics (Leachman et al., 2025). Repeated use of the same timed oral-reading indicator may conceal a changing psychological meaning. The bridge is therefore retained as one important predominant configuration, particularly when word-recognition efficiency is being transformed into sustained connected-text processing, but not as a complete developmental architecture.

A broader conceptual framework should therefore specify the construct's boundaries, distinguish the resources that support coordination from coordination itself, and represent feedback from meaning construction without incorporating comprehension into fluency. It should also explain why oral and silent manifestations—and their measures—have different informational value at different points in development.

3 Reframing reading fluency as a bounded and multidimensional construct

3.1 Foundational resources, core dimensions, observable manifestations, and structural positions

Fluency must remain distinguishable from reading competence as a whole. Multidimensional accounts emphasize accuracy, automaticity or appropriate rate, and expressive organization (Hudson et al., 2009; Kuhn et al., 2010; Wolf and Katzir-Cohen, 2001). Incorporating all supporting resources and comprehension outcomes into fluency would dissolve its conceptual boundaries.

The framework distinguishes four analytic categories. Foundational resources enable fluent text reading, including efficient word recognition, orthographic learning, lexical quality, phonological and morphological knowledge, and broader linguistic and cognitive resources. Core dimensions characterize fluent performance: accuracy, temporal continuity and pace regulation, and—in oral reading—prosodic organization. Observable manifestations are modality- and task-specific behaviors such as rate, errors, pauses, phrasing, acoustic prosody, fixations, regressions, and rereading. Structural position denotes the function assigned to fluency in a model, such as outcome, direct contributor, mediator, or part of a reciprocal relation.

Resources support performance, dimensions are inferred from manifestations, and structural position depends partly on operationalization. A rate score represents only part of fluency. Neither a resource's causal contribution nor an association with comprehension makes it a core dimension.

These analytic distinctions do not imply modular independence. Vocabulary and syntax can support grouping and local updating, attention can sustain continuity, and comprehension-related expectations can regulate pace. Researchers must distinguish the resources supporting performance from the dimensions elicited and measured and the structural relations observed.

3.2 From word-level automaticity to connected-text coordination

Efficient word recognition is foundational but not equivalent to text-level fluency. Precise, integrated orthographic, phonological, and semantic representations support reliable identification, while orthographic learning and repeated print encounters consolidate lexical representations (Perfetti, 2007; Share, 1995). Synchronization theory emphasizes compatible timing among component processes, and the Reading Systems Framework explains how lexical information is incorporated into an evolving text representation (Breznitz, 2006; Perfetti and Stafura, 2014). These established accounts supply mechanisms on which the present framework draws.

Connected-text coordination is an explanatory hypothesis about temporal relationships among word identification, linguistic grouping, and local semantic updating. It must be distinguished from the descriptive dimensions above. Indicator covariance cannot establish coordination. Tests require predicted responses to independently specified processing demands, independently assessed outcomes, and comparisons with alternative explanations.

Coordination organizes the account of temporal relations among established processes without introducing an additional faculty or core dimension. Reliable identification may coexist with difficulties in local integration, and continuous processing does not ensure the knowledge or reasoning needed for full comprehension. Sections 4.2 and 7.2 specify tests of the proposed explanation.

Repeated reading experience may strengthen lexical representations and sensitivity to orthographic, lexical, and syntactic regularities. In sufficiently informative contexts, accumulated linguistic knowledge can support graded expectations about upcoming input, including its semantic and syntactic features (Wong et al., 2025). Anticipatory processing is proposed as one mechanism supporting adaptive coordination. Informative expectations may facilitate ongoing processing, while discrepancies between expected and encountered input may prompt additional local processing, selective slowing, or rereading. Repeated encounters may improve the calibration of these expectations and adjustments, allowing time and attention to be allocated in ways that support accurate identification and local meaning construction.

This account must accommodate differences among languages, writing systems, and orthographies. Psycholinguistic grain size theory explains how the consistency of orthographic–phonological mappings shapes the units and strategies used in learning to read. Within alphabetic systems, relatively consistent grapheme–phoneme mappings favor reliance on smaller decoding units, whereas less consistent mappings encourage the use of multiple grain sizes (Ziegler and Goswami, 2005). Morphological structure also requires consideration alongside orthographic consistency (Goswami and Ziegler, 2006). Word identification remains fundamental, but its supporting units and processes can carry different weights across languages and developmental points. Differences in decoding grain size do not, by themselves, establish corresponding differences in contextual prediction.

The developmental hypothesis is that reading experience can support better calibrated coordination among word identification, linguistic expectations, and local updating. Prediction is a supporting process whose contribution depends on context, available resources, and task demands (Huettig and Mani, 2016). It is not an additional defining dimension or a necessary condition of fluency. Empirical evaluation must distinguish anticipation from facilitation attributable to integration after input is encountered (Wong et al., 2025). Neither local slowing nor rereading alone identifies which process produced an adjustment.

3.3 Accuracy, temporal continuity and pace regulation, and—in oral reading—prosodic organization

Accuracy is necessary because unreliable word identification constrains fluent processing. It may dominate performance while decoding is unstable but become less discriminating near ceiling. Errors and self-corrections can still reveal lexical-access or monitoring difficulties. High accuracy satisfies only one condition of fluency and must be interpreted within a broader performance profile (Valencia et al., 2010).

Identical accuracy scores can reflect scattered substitutions, recurring difficulty with a pattern, or errors clustered around demanding material. Self-correction may indicate effective monitoring, unresolved instability, or both. Interpretation therefore requires attention to error type, location, recurrence, and repair.

Appropriate rate is more informative than maximal speed (Kuhn et al., 2010). Both fragmented processing and appropriate slowing for difficult text or close interpretation can increase reading time. The framework treats selective local slowing as potentially part of fluent performance when it supports ongoing integration and a return to continuity. Overall rate must therefore be interpreted alongside the location, duration, and subsequent course of processing adjustments.

The distinction between pace and pace regulation becomes increasingly important as reading develops. A beginner's long pauses may mark unresolved word identification, whereas an advanced reader's local slowing may preserve coherence or resolve ambiguity. Total time collapses these behaviors into one value. Their interpretation requires independently specified processing demands and comprehension outcomes, so that neither slowing nor successful comprehension alone defines an adjustment as fluent.

In oral reading, prosodic organization provides evidence about phrasing, stress, intonation, and pausing. Associations with comprehension are supported by individual studies and a meta-analysis (Benjamin and Schwanenflugel, 2010; Wolters et al., 2022), but evidence that prosody mediates decoding–comprehension relations was limited in Schwanenflugel et al. (2004). Prosodic features develop with reading proficiency (Godde et al., 2020). The framework hypothesizes changing interpretive emphasis, from constraints imposed by decoding and articulation toward linguistic grouping and task-sensitive expression.

Prosody is thus a candidate indicator whose interpretation may shift with development. Acoustic prosody is specific to oral reading, whereas silent reading may involve implicit prosodic organization without an acoustic signal. Its possible changing role motivates the configurations examined next.

4 The developmental reorganization of reading fluency

Reading-fluency development involves hypothesized changes in processing constraints alongside efficiency gains. Three overlapping configurations distinguish fluency as an emerging outcome of word-recognition development, a mediator between foundational skills and comprehension, and part of reciprocal, task-dependent relations with meaning construction (Figure 2). This theory-guided synthesis draws on published findings. Its developmental ordering remains hypothetical.

Longitudinal observations provide qualified grounds for this organization. Kim and Wagner (2015) found that text-reading fluency was not independently associated with comprehension in Grade 1 after word-reading fluency and listening comprehension were considered. In Grades 2–4, it statistically mediated the word-fluency–comprehension association. Psyridou et al. (2023) found selective within-person predictive relations, with silent fluency predicting comprehension from Grade 1 to Grade 2 and comprehension weakly predicting silent fluency from Grade 7 to Grade 9. These findings motivate changing relations without establishing a universal sequence. Statistical mediation does not identify coordination as a mechanism, and longitudinal prediction does not establish regulation during a reading episode.

Outcome, mediation, and reciprocity describe relations among constructs. The processing explanation concerns how lexical constraints, integration demands, and regulation shape temporal organization. Their predominance requires independently assessed component abilities and responses to specified demands. It may vary with linguistic knowledge, orthography, text, and task, including within a reader.

4.1 Fluency as an emerging outcome of word recognition

Word-identification limitations are hypothesized to predominate here. Orthographic mapping explains how connections among spellings, pronunciations, and meanings support familiar-word recognition (Ehri, 2005, 2014). Share (1999) found that second graders who encountered pseudowords in short texts subsequently identified the encountered spellings more accurately and named them more quickly than alternative spellings with identical pronunciations. This supports experience-dependent orthographic learning. Consequences for connected-text continuity require further evaluation.

More reliable word identification is proposed to reduce recognition-related interruptions and permit sustained processing across words. Accuracy may improve before such continuity is established. Oral hesitations, repetitions, and pauses are possible indicators, but recognition difficulty must be distinguished from linguistic grouping and speech planning. Relevant processing units and demands vary across orthographies (Ziegler and Goswami, 2005). This configuration is therefore defined relative to the reader and text, including unfamiliar written words encountered by otherwise proficient readers.

A foundational test could strengthen recognition of selected words and assess transfer to those words in unfamiliar passages against a matched control condition. Spoken-word knowledge, meanings, and text demands would be controlled, with independent word-reading tasks verifying learning. Benefits confined to isolated words would limit the proposed transfer to continuity. This commitment is shared with word-learning accounts.

4.2 Connected-text coordination and the mediating role of fluency

As word identification becomes more reliable, connected-text performance may be increasingly constrained by integrating successive words into phrases and sentences. Statistical mediation by text fluency provides relational evidence compatible with a coordinating function (Cadime et al., 2024; Kim and Wagner, 2015). Establishing that interpretation requires evidence about processing under varying integration demands.

Word-to-text integration connects lexical information with an evolving interpretation through activation and selection of contextually relevant meanings (Perfetti and Stafura, 2014). The framework proposes that reduced recognition-related disruption allows time to be allocated more selectively to unresolved linguistic relations. Local slowing or rereading can support updating before processing continues, while linguistic expectations help determine where adjustment is needed. Automatic responses and deliberate attention may both contribute. The proposed coordinating function concerns this pattern of adjustment and recovery across text. Its effectiveness must be evaluated through independent comprehension outcomes.

An initial silent-reading test could vary whether a sentence interpretation requires revision while holding critical words constant across counterbalanced versions. Adjustments concentrated around predefined integration regions, followed by recovery toward control-condition timing, are predicted to be associated with more accurate interpretation than delays persisting into subsequent regions. Regions and recovery criteria would be specified before examining comprehension. Analyses would assess information beyond word-reading efficiency, oral-language comprehension, and overall reading time. A single pause or regression would not establish coordination.

This processing account motivates the central developmental prediction. When word identification strongly constrains reading, global rate can reflect interruptions that also limit comprehension. As these constraints diminish, differences in understanding are hypothesized to depend increasingly on how processing time is distributed around integration demands. Under comparable conditions, the rate–comprehension association is therefore predicted to weaken while the association between demand-sensitive local timing and comprehension remains stable or strengthens. This joint pattern depends on a change in the processes limiting comprehension and is not entailed by faster word identification alone. Section 7.2 specifies its longitudinal evaluation.

4.3 Reciprocal relations and adaptive regulation

This configuration foregrounds how the developing interpretation and reading purpose guide subsequent processing. Contextual expectations and discrepancies between an interpretation and incoming information provide candidate mechanisms for redirecting attention, adjusting pace, or revisiting text. Adjustments may involve automatic responses and deliberate strategies, drawing on established accounts of active regulation (Duke and Cartwright, 2021). Experience is hypothesized to improve their calibration to linguistic demands and goals. Meaning-based regulation can operate before word recognition is fully consolidated.

Regulation during reading must be distinguished from prediction across developmental assessments. Psyridou et al. (2023) found selective and generally weak within-person associations, including a later association from comprehension to silent fluency. Zhao et al. (2025) assessed separate grade cohorts twice over 6 months and found prediction from text fluency to comprehension in Grade 4 and the reverse direction in Grade 6. These findings motivate changing directional relations without establishing continuous bidirectionality or the mechanisms of online regulation.

Experimental evidence provides a complementary basis for considering rereading. Schotter et al. (2014) found that masking previously read words, thereby preventing effective rereading, reduced comprehension. This supports the contribution of access to earlier text without implying that every regression is adaptive. The relevant question is whether returning to particular information helps resolve a difficulty that matters for the reader's goal.

The regulatory hypothesis predicts that contextual inconsistencies relevant to an assigned goal will elicit more targeted rereading of information needed to resolve them than matched inconsistencies incidental to that goal. The predicted selectivity should be associated with independently assessed understanding. Task effects would support a shared regulatory principle. The developmental extension requires repeated assessment of whether selectivity becomes more informative as word identification consolidates, alongside the changing contribution of global rate. Section 7.2 identifies this as a secondary hypothesis.

4.4 Developmental reorganization as a testable hypothesis

The configurations overlap because their constituent processes remain relevant across development. Familiar language may permit selective regulation, whereas unfamiliar written forms can reinstate identification demands without reversing developmental progress. Linguistic and orthographic conditions shape how each configuration is expressed.

Longitudinal evaluation must distinguish within-reader change from between-reader differences while addressing text difficulty, measurement comparability, and restricted ranges. Existing accounts already accommodate developmental and task-dependent relations (Kim, 2020). Section 7.2 specifies comparisons that test the proposed joint pattern against these accounts and determine whether the configurations add explanatory constraints or reproducible predictive information.

5 Oral and silent reading as overlapping but non-equivalent manifestations

5.1 A shared processing core and modality-specific demands

At the level of connected-text processing, oral and silent reading both require reliable word recognition, lexical access, linguistic grouping, local semantic updating, and continuity. These shared processing requirements help explain their association and why oral reading fluency can provide evidence about broader reading development. Shared foundations, however, do not make the two manifestations interchangeable.

Implementation and observable output nevertheless differ. Oral reading couples text processing with continuous articulation, eye–voice coordination, and overt timing and prosody. Its pauses, errors, repairs, and phrasing reflect both reading and production. Silent reading permits flexible pacing, regressions, and rereading without audible disruption. Similar total rates can therefore reflect different constraints, and direct comparisons reveal common variance alongside modality-specific demands (Kim et al., 2011; van den Boer et al., 2014).

Neither modality transparently reveals a single mechanism: an oral pause or silent regression can reflect word-recognition difficulty, integration, monitoring, or strategic adaptation. The framework hypothesizes that oral and silent fluency involve overlapping but non-equivalent forms of connected-text coordination.

5.2 Developmental change in oral and silent manifestations

The relative informational value of the two manifestations changes across development. Early oral performance makes instability visible through errors, pauses, repetitions, and disrupted phrasing. Completion time alone cannot establish that a silently read text was processed. Even in first-grade research, silent fluency has therefore been operationalized through a timed sentence-verification task (Kim et al., 2011). As word recognition stabilizes, silent reading becomes increasingly prominent in independent reading and permits internally regulated allocation of time.

Verification nevertheless alters the task. Adding sentence judgments or comprehension questions introduces linguistic, decision, and memory demands beyond silent passage timing. Modality and verification procedure therefore jointly shape the observable evidence of fluent processing.

Oral and silent performance can be strongly related without being interchangeable. Cross-sectional evidence shows age- and reader-group differences in relative oral and silent fluency (van den Boer et al., 2022), while oral and silent word-reading fluency show overlapping but partly distinct predictor profiles across grades (Georgiou and Rothou, 2024). In middle school, associations with comprehension also depend on how fluency and comprehension are measured (Denton et al., 2011). Together, these findings support conditional, developmentally variable overlap rather than a universal claim that silent fluency is closer to comprehension.

The framework expects the informativeness of oral and silent measures to depend on proficiency and task. Oral reading remains informative about expressive interpretation and linguistic organization in advanced readers, while well-designed silent measures can reveal differences earlier. Both involve reliable word identification. Whether their timing patterns reflect shared coordination requires independent evaluation.

Relations with comprehension are likewise conditional. Oral fluency can be especially informative when production makes disruptions in word recognition and grouping audible. Silent fluency may become more closely associated with comprehension when pacing and rereading respond to meaning. The relevant pattern depends on developmental level, language, task, and measurement.

5.3 Prosody as a developmentally shifting indicator of integration

Prosody is distinctive because its overt form belongs to oral reading, whereas the linguistic organization it may reflect extends beyond speech production. Phrasing, stress, intonation, and pausing can indicate grouping compatible with syntax and developing meaning. A meta-analysis found an overall prosody–comprehension association, moderated by the prosodic feature measured but not significantly by grade or orthographic depth (Wolters et al., 2022). Developmental changes in interpretation therefore require evidence beyond this aggregate association.

The framework proposes that early prosody is especially constrained by word identification and articulation, so flat or fragmented reading need not imply weak syntactic or semantic understanding. Prosodic organization may become more informative about linguistic grouping as identification improves. Longitudinal links between early prosodic development and later comprehension are compatible with this possibility (Miller and Schwanenflugel, 2008), but do not establish the proposed shift. In advanced reading, phrasing and emphasis may also reflect discourse, genre, communicative intention, and interpretive variation.

Such hypothesized shifts would limit comparison through a single prosodic rubric. Early scoring may mainly capture whether recognition is stable enough to sustain phrase-level output. Transitional scoring can become more sensitive to alignment with syntactic boundaries. Advanced performance may permit several coherent realizations, so restrained expression or an unconventional contour should not automatically be classified as disfluent without reference to genre and communicative purpose.

Veenendaal et al. (2016) found that Grade 4 comprehension predicted Grade 5 prosody, whereas Grade 5 prosody predicted Grade 6 comprehension. These grade-specific associations permit reciprocal interpretations without establishing a monotonic shift toward comprehension-driven prosody. Prosody's proposed changing role therefore remains conditional. Acoustic measures cannot be generalized directly to silent reading. Implicit phonological or prosodic organization may occur during silent reading, but its structure and indicators cannot be assumed to match overt prosody.

Prosody therefore illustrates the distinction between shared processing and modality-specific manifestation. Organizing linguistic units is common to both modalities, but only oral reading renders that organization acoustically. Audible prosody can thus support hypotheses about integration without directly measuring silent prosodic organization.

6 Contextual conditions and the measurement-dependent visibility of fluency

6.1 Orthographic and linguistic conditions

Fluency develops within a writing system and language. The framework proposes shared requirements for word identification and connected-text coordination, with varying developmental timetables and indicators. Children learning more consistent orthographies generally acquire accurate decoding earlier, while differences in fluency can remain salient after accuracy becomes high (Seymour et al., 2003). Orthographic depth can therefore change the relative diagnostic value of fluency dimensions without necessarily implying a different reading architecture.

Foundational skills can carry different weights across orthographies. Evidence from Finnish, Hungarian, Dutch, Portuguese, and French shows shared predictors with varying relative influence (Ziegler et al., 2010). Orthographic consistency can alter which aspects of accuracy and timing remain discriminating, but it does not by itself establish more advanced integration or stronger contextual prediction. Linguistic structure adds further demands. In Greek, morphological awareness has been linked through different statistical pathways to oral text fluency and silent word-reading measures (Giazitzidou et al., 2024). The developmental direction of morphological contributions is considered in Section 7.1.

Transparency should therefore not be equated with processing simplicity. The dimensions that discriminate among readers, and the point at which fluency is most informative as an outcome, direct contributor, or mediator, may shift without requiring language-specific constructs.

Linguistic conditions are thus conceptualized as modifiers of developmental timing and pathway strength rather than deviations from an English-based sequence (Share, 2008). Cross-orthographic studies should distinguish differences in timetable or observable profile from genuinely different structural organization.

6.2 Reader, text, task, and instructional conditions

Fluency is instantiated by a reader processing a particular text for a purpose. Vocabulary, syntax, attention, working memory, background knowledge, and experience can lessen or increase demands on connected-text coordination, but remain supporting resources rather than internal dimensions. An efficient word reader may slow in an unfamiliar domain or syntactically demanding passage, whereas relevant knowledge may stabilize processing.

Text difficulty at word, sentence, and discourse levels shapes local processing and overall rate. Slower reading may therefore be an appropriate adaptation to a demanding passage rather than evidence of reduced fluency. In a large primary-grade sample, oral-reading rates in informational text varied across Grades 1–3 and text levels (Tortorelli, 2019).

A reading task is part of the conditions under which the system operates. Reading aloud expressively, scanning for a fact, answering inferential questions, and studying an argument require different allocations of time and attention. Speed-focused instructions may discourage rereading, whereas meaning-focused directions may encourage local slowing. Task duration also matters: performance can change within a 3-min oral-reading passage even when 1 and 3-min scores predict comprehension similarly (Cadime et al., 2023).

Instruction and the wider educational and social context shape what readers practice and what they treat as fluent performance. Repeated oral reading emphasizes accuracy, continuity, and phrasing, whereas extensive silent-reading experience emphasizes self-regulated pacing. These conditions alter the demands on coordination and the strategies adopted. They do not become fluency dimensions. Similar performances may therefore arise from different reader-text-task configurations.

These conditions interact rather than contribute independently. A demanding text may expose a vocabulary limitation, speed-focused instructions may suppress rereading, and prior knowledge may compensate for lexical or syntactic difficulty. Consequently, the same reader may display different but adaptive fluency profiles across purposes, while similar observed profiles may arise from different combinations of resources and demands.

6.3 Measurement-dependent visibility

Measurement selects indicators and defines their scoring or aggregation. In this sense it is an observational lens. Procedures that change reading conditions, however, also alter the task and may influence processing, as when comprehension questions introduce additional demands. The framework separates measurement as representation from these task effects.

A words-correct-per-min score is an efficient and generally reliable indicator for screening and progress monitoring (Fuchs et al., 2001), but it is not exhaustive. The same score can conceal the distribution of errors, linguistic placement of pauses, endurance, prosodic organization, or adaptation to difficulty. A rate–accuracy composite and a broader profile incorporating rate, accuracy, prosody, and comprehension do not necessarily represent oral fluency in the same way (Valencia et al., 2010).

The linguistic level sampled by an assessment is particularly important for the proposed construct. Sublexical and isolated-word measures can reveal foundational efficiency but do not require sustained coordination across discourse. Sentence and passage measures add sequential, syntactic, and local integration demands, although they differ in length and opportunities for adaptation. Passage-based fluency should therefore not be treated as interchangeable with word-list efficiency, and connected-text measures need not sample coordination in the same way (Jenkins et al., 2003; Klauda and Guthrie, 2008).

Process-sensitive measures provide complementary evidence: error and repair profiles index reliability and monitoring. Acoustic analyses separate pausing, phrasing, stress, and timing. Eye movements reveal fixation, regression, rereading, and local adaptation during silent reading. None directly reveals an internal fluency mechanism. Their value lies in triangulating processing configurations that similar total scores may conceal.

Developmental comparability cannot be inferred from an unchanged task and score. Accuracy may dominate early variance and later approach ceiling. Rate may shift from decoding effort toward text difficulty or regulation. Prosody may become increasingly interpretive. Measurement invariance should therefore be tested after considering task artifacts, restricted variance, and floor or ceiling effects. Developmental phase, orthographic depth, and measurement characteristics moderate the observed fluency–comprehension relation (Leachman et al., 2025). A narrow measure can consequently make the construct appear more stable, unitary, or speed-centered than it is.

Non-invariance may reflect changed psychological contributions, but artifacts and range restrictions must first be examined. It neither proves a configuration change nor licenses direct comparison of unlinked latent scores or relations. Developmental tests require comparable indicators or models that explicitly accommodate measurement differences, followed by independent evaluation of the proposed process interpretation.

7 A proposed integrative developmental framework for reading fluency

7.1 Architecture of the proposed framework

The framework places fluency within an interacting reading system (Figure 3). Foundational resources include decoding, orthographic learning, lexical quality, linguistic knowledge, attention, and working memory. This list is illustrative. Resources support fluent performance without becoming its internal dimensions. Accuracy, temporal continuity and pace regulation, and oral prosodic organization describe performance, while connected-text coordination is a provisional account of its temporal organization. Broader comprehension remains conceptually distinct (Duke and Cartwright, 2021; Perfetti and Stafura, 2014).

Oral-language abilities provide a fundamental linguistic basis for reading comprehension. Spoken and written language draw on overlapping phonological, lexical, syntactic, semantic, inferential, and discourse-level resources (Perfetti and Stafura, 2014). Phonological and lexical knowledge support learning to decode and reliable word recognition. Syntax and semantics support grouping and local integration, while inferential and discourse processes contribute to broader understanding. These resources can influence comprehension through word identification and fluent processing, as well as through pathways not fully mediated by fluency. Their contribution should be assessed independently when evaluating the additional explanatory value of fluency indicators (Duke and Cartwright, 2021).

The layered architecture does not imply sequential or independent processing. Foundational resources and meaning construction can influence ongoing reading, and repeated encounters with print can support further learning (Ehri, 2014; Share, 1995). Morphological knowledge illustrates such reciprocal possibilities. The Morphological Pathways Framework distinguishes morphological awareness from the use of morphological information in identifying and interpreting written words (Levesque et al., 2021). Recurring morphemes may support lexical access by linking written forms with phonological, semantic, and grammatical information. Repeated encounters with related words are hypothesized to strengthen these connections, while increasingly reliable identification may support this learning by enabling more successful encounters.

Deacon et al. (2013) found that morphological awareness predicted growth in reading accuracy from Grades 2 to 3, while accuracy also predicted growth in awareness after prior performance and relevant abilities were considered. This concerns awareness and accuracy. Extension to online morphological processing and multidimensional connected-text fluency requires separate evaluation. Tests should distinguish morphological awareness, morphological processing, word-reading efficiency, and reading experience.

In a previous collaborative study involving 4,134 Hungarian children in Grades 2–4, we found moderate positive associations between morphological awareness and reading comprehension across all three grades (Varga et al., 2022). These findings support the relevance of morphological knowledge in an agglutinative language with relatively transparent orthography. The cross-sectional design establishes associations and differences between grade cohorts, while leaving developmental direction unresolved. The Greek findings of Giazitzidou et al. (2024) likewise concern cross-sectional pathways across different morphological and reading measures. Together, these studies constrain claims about direction and generalization. Consistent grapheme–phoneme mappings can coexist with substantial morphological demands, and transparency alone cannot determine morphology's contribution.

Oral and silent fluency are overlapping manifestations whose tasks and observable demands differ (Kim et al., 2011; van den Boer et al., 2022). Articulation, eye–voice coordination, and overt prosody characterize oral output, while silent reading permits different pacing and rereading opportunities. The developing interpretation and reading purpose may regulate subsequent processing in either modality. These contributions depend on resources and task demands, and their changing relative prominence is a developmental hypothesis.

Compensatory profiles constrain the relations represented in the architecture. Brèthes et al. (2022) found that university students with dyslexia had weaker word- and text-reading fluency while achieving good untimed comprehension. Listening comprehension, vocabulary, and general knowledge contributed to its statistical explanation. Cavalli et al. (2019) found preserved comprehension when university students with dyslexia read a printed book without time pressure, whereas some outcomes were poorer with an e-book. These findings concern selected adult samples and particular assessment conditions.

Within the framework, oral-language abilities and background knowledge can support meaning construction through pathways not fully mediated by fluency. Additional time and selective rereading are candidate means of sustaining understanding when identification remains effortful. Executive functions and strategic control may contribute, but their compensatory role requires independent assessment. Preserved comprehension establishes neither efficient processing nor successful coordination. Fluency, comprehension, and the time and effort required must therefore be distinguished.

A specific rereading-based explanation predicts a greater comprehension cost from restricting access to earlier text in readers with weak word-reading efficiency but preserved untimed comprehension than in readers matched on baseline comprehension who have more efficient word reading. Reader profiles would be established on independent baseline measures. Process measures would distinguish targeted interpretive repair from repeated identification attempts. Failure to observe the interaction would weaken this particular explanation. Persistent lexical constraints can coexist with effective meaning-based regulation, so fully consolidated recognition cannot be assumed to be a prerequisite for regulatory contributions.

Context and measurement have different roles. Language, orthography, text, task, instruction, reader characteristics, and educational conditions can influence development and processing. Measurement selects observable indicators, while procedures that alter reading conditions also affect the task. Figure 3 therefore represents multiple resource-to-comprehension pathways and conditional feedback within the reading system, while keeping measurement outside it as an observational lens.

7.2 Prioritized hypotheses and empirical evaluation

The research agenda prioritizes one developmental hypothesis, followed by secondary tests of regulation and modality. Construct boundaries and distinctions among supporting resources, fluency indicators, and comprehension are measurement requirements across all tests. Illustrative tests in Section 4 concern shared processing commitments. The primary hypothesis evaluates their proposed developmental organization. A longitudinal study could address it, with targeted extensions using overlapping measures. Initial studies should specify a manageable language context, modality, and range of text demands. Cross-linguistic generalization requires subsequent evaluation.

7.2.1 Primary hypothesis H1: developmental divergence in the informativeness of global and local timing

Under comparable reading conditions, consolidation of word identification is expected to coincide with a weakening association between global reading rate and comprehension, while the association between independently specified, demand-sensitive local timing patterns and comprehension remains stable or strengthens. This joint prediction concerns changing explanatory relations beyond gains in average performance.

Longitudinal tests should repeatedly assess word-reading efficiency, oral-language comprehension, global rate, local timing, and independent comprehension outcomes. Tasks can vary interpretation-revision demands using counterbalanced versions with matched critical words. Critical regions and recovery criteria must be pre-specified. Preregistered multilevel models should test whether word-reading consolidation moderates each timing–comprehension association, separating within-reader change from between-reader differences. Analyses must address text difficulty, reliability, measurement comparability, floor and ceiling effects, and the mathematical dependence of local and total time. Language and orthography should inform processing units and task construction.

Evaluation should include competing accounts that allow developmental variation in component contributions and explanations based on word-to-text integration. Support requires the predicted joint developmental pattern. Claims of added predictive value additionally require reproducible improvement in independent data against appropriate comparison models. A benefit from local measures at one assessment supports their measurement utility without establishing developmental reorganization. Failure to observe the joint pattern in adequately sensitive studies would require revision. Equivalent explanation by existing accounts would limit theoretical added value, and improved prediction would not by itself identify a distinct causal coordination mechanism.

7.2.2 Secondary hypothesis H2: goal-dependent allocation of processing time

When word-identification demands are comparable, a contextual inconsistency is expected to elicit more targeted rereading when resolving it is relevant to the assigned goal. This selective response should also be associated with more accurate understanding of goal-relevant content.

An initial experiment could cross contextual consistency with reading goals in counterbalanced passage versions, holding critical words constant. Regions containing information needed to resolve an inconsistency would be specified independently of responses. Eye movements would measure the probability and duration of returns to those regions, while separate questions would assess interpretation. Support requires the consistency-by-goal interaction and its relationship with comprehension, assessed beyond overall time and independently measured component abilities. This examines a regulatory principle shared with existing accounts. Its developmental extension requires repeated assessment of whether selectivity becomes more informative as identification consolidates.

7.2.3 Secondary hypothesis H3: generalization across oral and silent reading

Local timing patterns responsive to independently specified integration demands are expected to provide information about comprehension in both modalities beyond word-reading efficiency, oral-language comprehension, and global rate. The predicted convergence concerns their functional relationship with understanding, while observable expression and relative contribution may differ.

A within-reader design could counterbalance oral and silent reading of parallel passages with comparable lexical, syntactic, and comprehension demands. Oral measures could include pause location and duration and phrase-boundary timing, with articulation demands assessed separately. Silent measures could capture fixation time, rereading, and recovery in corresponding regions. Criteria would be specified in advance, and comprehension assessed independently in both modalities. Support requires the predicted contribution in each modality. Evidence confined to one would narrow the account's scope, with reliability and modality-specific demands considered when interpreting differences.

H2 and H3 can be investigated through targeted extensions of the longitudinal study or separate experiments with compatible measures. Their findings would clarify the processing interpretation and scope of the framework. Evidence for developmental reorganization would still depend on H1.

7.2.4 Exploratory extensions

Language, writing system, orthographic consistency, and morphological structure should inform all tests, including the selection of reading units, passages, and timing indicators. Exploratory extensions can examine how these conditions, text characteristics, and reader profiles alter the magnitude or developmental course of the predicted relations. Directional moderation hypotheses supported by existing evidence should be specified before analysis. Other analyses should be labeled exploratory. Unexpected patterns may motivate revisions, with new predictions evaluated in independent data. H1 should be judged against its pre-specified conditions and criteria.

8 Discussion: from a static bridge to a developmentally reorganized reading system

8.1 Reframing the bridge

The framework explains fluency's changing structural position through proposed changes in the constraints organizing processing time. A bridge-like configuration is especially informative when word-identification efficiency supports sustained integration across text. Outcome, direct-contributor, mediator, and reciprocal positions can coexist (Steklács and Vuong, 2026). The three configurations connect these positions with hypotheses about lexical disruption, integration, and regulation under different reading conditions.

Definitions, samples, languages, texts, tasks, measures, and analyses also create heterogeneity. Moderation by development, orthography, and measurement supports a conditional account (Leachman et al., 2025). The question is which processes and conditions make a bridge-like configuration informative.

8.2 Relation to existing theories

Mechanisms from established accounts provide the foundations of the framework. Automaticity theory explains how efficient word identification reduces attentional demands (LaBerge and Samuels, 1974). Synchronization theory emphasizes temporal compatibility among processes (Breznitz, 2006), and word-to-text integration explains how lexical information connects with an evolving text representation (Perfetti and Stafura, 2014). Active regulation is explicit in the Active View of Reading (Duke and Cartwright, 2021). Component interaction and regulation are therefore theoretical antecedents.

The Dual Route Cascaded model provides a computationally implemented account of visual word recognition and reading aloud, specifying interactions between lexical processing and grapheme–phoneme conversion (Coltheart et al., 2001). Cascaded activation allows one level to influence subsequent levels before its processing is complete. This provides an explicit precedent for temporally overlapping processes. The present framework draws on such word-identification accounts when examining lexical constraints in relation to connected-text integration and regulation. Its hypotheses concern developmental relations among word-reading efficiency, local timing, and comprehension beyond the word-level tasks simulated in the 2001 model. Broader scope alone does not establish added explanatory value.

(Van Dijk and Kintsch 1983) describe comprehension as the strategic construction of a coherent textbase and the activation and updating of a situation model, informed by knowledge and goals. Their account includes interactions across levels, revision of provisional interpretations, and the learning and automatization of comprehension strategies. These principles underpin the present account of meaning-sensitive regulation. Goal-sensitive processing and local slowing are already compatible with the earlier account, so their observation alone would provide limited evidence of added value.

The closest developmental antecedents are Kim's component-based fluency model and the Direct and Indirect Effects Model of Reading (Kim, 2015, 2020). These already distinguish word- and text-reading fluency, accommodate changing mediation, and specify relations that vary with development, text, and task. Developmental variability itself is not a novel contribution. Different predictor profiles for oral and silent word reading (Georgiou and Rothou, 2024) and orthography-dependent processing units (Ziegler and Goswami, 2005) provide further constraints.

The proposed extension specifies how changes in processing constraints alter the relationship between timing and comprehension. As recognition-related interruptions diminish, linguistic grouping and meaning updating are hypothesized to account for a greater share of consequential timing differences. Accumulated experience can support expectations that help locate processing demands and calibrate responses to them. Selective slowing followed by recovery of continuity can then support comprehension without increasing passage-wide speed. This account links a proposed change in temporal organization to the joint developmental prediction in H1. Knowledge, orthography, and task conditions constrain that change.

H1 commits the framework to a weakening rate–comprehension association alongside a stable or strengthening relationship between demand-sensitive local timing and comprehension. Developmental variability alone does not entail this joint pattern. It must survive controls for component abilities, measurement comparability, difficulty, and restricted ranges, and explain or predict outcomes beyond appropriate developmental and integration accounts. Equivalent explanation by those accounts would limit the proposed extension. Support for H1 would establish neither a distinct causal coordination mechanism nor the necessity of exactly three configurations.

8.3 Interpreting fluency across development and modality

Fluency indicators provide probabilistic evidence with multiple possible explanations. Inaccuracy may signal recognition difficulty, while atypical phrasing may reflect integration, monitoring, interpretation, or production demands. Identical rates can reflect effortful decoding or selective adjustment. Early prosodic patterns predict later fluency and comprehension (Miller and Schwanenflugel, 2008), but shifts in their explanatory meaning require independent evaluation. No single behavior directly establishes coordination.

Interpretation requires patterns across dimensions and conditions. Strategic rereading and decoding pauses can produce similar overall rates with different implications. Longitudinal observations help establish whether the sources and organization of performance change within readers.

Oral reading reveals errors, pausing, phrasing, and repair, whereas silent reading permits internally regulated pacing and rereading. Comparisons across reader groups indicate conditional overlap between modalities (van den Boer et al., 2014, 2022). These cross-sectional findings do not establish within-reader developmental change. Oral prosody remains potentially informative across proficiency levels, with its interpretation requiring independent evaluation.

8.4 Implications for assessment and instruction

Assessment should match the developmental question. Unstable recognition warrants attention to accuracy and interruptions. Integration demands call for timing, phrasing, and error profiles. Advanced reading may require silent measures of pacing and adaptation. Words-correct-per-min scores remain useful for screening and monitoring but require complementary evidence to distinguish underlying profiles (Valencia et al., 2010).

Indicator selection should match purpose: brief composites for screening, followed by component or process measures for explanation and evaluation of change. Documenting text and task demands helps distinguish difficulty from adaptation when interpreting lower rates.

Instructional implications remain hypotheses. Persistent inaccuracy directs attention to word recognition and orthographic learning. Accurate but fragmented reading calls for continuity and phrasing, while integration difficulty requires linking fluent processing with meaning. Advanced fluency involves regulating pace and rereading for purpose, genre, and difficulty. These principles establish neither a universal instructional sequence nor intervention efficacy.

8.5 Scope and limitations

The framework addresses L1 fluency development. Its distinctions among resources, dimensions, contextual conditions, and consequences remain open to refinement. Motivation, strategies, instruction, experience, and social conditions may shape fluency without becoming its dimensions.

Second-language and multilingual reading may involve differently organized relations among proficiency, exposure, cross-language transfer, and linguistic knowledge. Their exclusion limits the present claims without implying that the proposed processes are irrelevant beyond L1 reading.

The configurations are overlapping heuristics whose predominance may vary with orthography, morphology, instruction, experience, text, task, and reader profile. Compensatory profiles in dyslexia provide relevant constraints on assumptions about word identification, fluency, and comprehension. Their interpretation requires direct evaluation in heterogeneous samples, with attention to linguistic resources, processing time, and task conditions.

Coordination requires evidence beyond covariance among fluency indicators. Tests must specify independent demands and outcomes, assess component abilities, and compare alternatives. Failure to observe the predicted differentiation and joint developmental pattern under adequately sensitive conditions would warrant revision.

Heterogeneity and contextual moderation may support fewer, additional, or differently organized configurations. The proposed number and boundaries remain revisable, subject to explicit empirical comparisons.

9 Conclusion

This framework proposes that reading-fluency development changes the organization of processing time across connected text. As word identification becomes more reliable, selective adjustment to integration demands and reading goals is hypothesized to become more informative about comprehension. Three overlapping configurations connect this proposal with fluency's roles as an emerging outcome, mediator and coordinator, and participant in reciprocal relations. Their number and ordering remain open to empirical comparison.

Accuracy, temporal continuity and pace regulation, and oral prosodic organization characterize fluent performance. Coordination is proposed as an explanation of its temporal organization, with linguistic and cognitive resources and broader comprehension kept conceptually distinct. Reading experience, morphological learning, orthography, modality, and task demands constrain the possible relations. Compensatory profiles show that lexical difficulty can coexist with preserved comprehension.

The central empirical commitment is a joint developmental divergence: under comparable conditions, global rate becomes less informative about comprehension while demand-sensitive local timing retains or increases its informativeness. Secondary hypotheses examine goal-sensitive rereading and cross-modal generalization. Independent outcomes, comparable longitudinal measurement, and competing explanations provide the criteria for evaluating these claims. The framework advances a specific connection between changing processing constraints, the organization of reading time, and the interpretation of fluency measures, with revisions required when its predicted relations fail.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. During the preparation and revision of this work, the author used ChatGPT (OpenAI, https://chatgpt.com; accessed August and September 2026) to support literature organization, language editing, structural refinement, and document assembly. The author critically evaluated, revised, and verified all AI-assisted content, citations, references, and figures and takes full responsibility for the content of the publication.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Pedro Guijarro-Fuentes, University of the Balearic Islands, Spain

Reviewed by: Halima Sahraoui, University of Toulouse, France

Claudia Marzi, Istituto di Linguistica Computazionale Antonio Zampolli Consiglio Nazionale delle Ricerche, Italy