What Do Theories of Dyslexia Explain?
Dyslexia theories explain how persistent differences in word reading and spelling develop, why they appear in multiple cognitive profiles, and how experiences, genes, language, brain development, and instruction jointly influence outcomes. A theory is useful when it specifies mechanisms, predicts results across developmental stages and languages, survives competing explanations, and improves the interpretation of evidence. The best-supported account recognizes the central role of speech-sound knowledge in acquiring alphabetic reading while also recognizing that the pathways leading to persistent literacy difficulties are heterogeneous. Pennington (2006); Peterson and Pennington (2015).
The phonological deficit theory places difficulty representing, accessing, or manipulating speech sounds near the center of many dyslexia pathways. The multiple-deficit model treats reading disability as a probabilistic outcome of interacting risks and protections operating at genetic, neural, cognitive, linguistic, and environmental levels. These are explanatory frameworks with overlapping predictions. Their most productive comparison asks which proposed factors account for individual variation, which factors are causally upstream, and which mechanisms can be changed through instruction. Snowling and Hulme (2021); Pennington (2006).
The International Dyslexia Association revised its definition in October 2025. It locates dyslexia in persistent difficulty with word reading and/or spelling, involving accuracy, speed, or both, while identifying phonological and morphological difficulties as common but not universal. The revision explicitly recognizes variation across writing systems and complex, interacting influences during development. This description is compatible with strong phonological evidence and requires theoretical models to explain meaningful exceptions and additional risk pathways. International Dyslexia Association (2025); Catts, Haynes, and Joshi (2026).
The Explanandum: Word Reading, Spelling, Fluency, and Development
Why the level of explanation matters
Word-level reading requires a learner to connect visual symbols to spoken language, acquire reliable correspondences between graphemes and phonemes where applicable, develop efficient access to stored word forms, and coordinate these processes with meaning. Spelling recruits related but partly distinctive retrieval and transcription processes. A student may read unfamiliar words inaccurately, recognize familiar words slowly, spell inconsistently, or display combinations of these difficulties. The same observed reading score can arise from different sequences of developmental constraints and learning experiences. Ziegler and Goswami (2005); Snowling and Hulme (2021).
An explanatory account needs to separate the behavior being explained, such as inaccurate decoding, from the cognitive measures correlated with it and from earlier causes of those cognitive differences. Slow naming, poor phoneme manipulation, and reduced print exposure may be predictors, consequences, correlates, or contributors, depending on age and design. Group-average associations alone cannot establish a chain of causation for every child. Longitudinal studies, experimental training, family-risk designs, natural variation in instruction, and carefully tested mechanistic models provide different parts of the causal picture. Wagner and Torgesen (1987); Peterson and Pennington (2015).
Dimensional difficulties and diagnostic categories
Reading accuracy, fluency, and spelling are distributed continuously in the population. Diagnostic and educational categories draw boundaries for decisions about assessment, access, and support. A threshold does not imply that every person who crosses it possesses the same cognitive mechanism. A child just above a test-score cutoff may resemble a child just below it more closely than two children with the same formal label resemble each other. Multiple-deficit frameworks are particularly useful for analyzing that continuous variation and the accumulation of risks. Pennington (2006); International Dyslexia Association (2025).
DSM-5-TR places persistent academic-skill difficulties within specific learning disorder, with impairment in reading among the possible specifiers. Dyslexia remains a commonly used term for word-reading and related spelling problems. US schools determine special-education eligibility through the Individuals with Disabilities Education Act and applicable state procedures, a process distinct from a clinical diagnosis. Cognitive theories inform assessment questions; they do not replace measures of actual reading performance, developmental history, and the educational context. American Psychiatric Association, Specific Learning Disorder; IDEA, 34 CFR 300.8.
The Phonological Deficit Theory of Dyslexia
Its central causal claim
The phonological theory holds that difficulty with aspects of the speech-sound system makes it harder to learn how print represents language. In alphabetic writing systems, readers must connect letters and letter combinations with phonemes, blend sequences to decode unfamiliar words, and consolidate successful decoding episodes into increasingly automatic word recognition. Phonological representations, awareness, access, and short-term storage are related but separable components of this process. The theory therefore predicts characteristic weaknesses in tasks that require analyzing or manipulating sound structure, particularly at early stages of literacy acquisition. Wagner and Torgesen (1987); Melby-Lervåg, Lyster, and Hulme (2012).
Phonemic awareness is the ability to consciously identify and manipulate individual speech sounds. A student can understand a spoken word while having difficulty deleting its first phoneme, blending sounds into a new word, or connecting each sound to a written symbol. Phoneme awareness becomes more explicit through alphabetic instruction, so a low score is best interpreted within the student’s literacy and language experience. Phonological deficits are neither equivalent to low intelligence nor a statement about motivation. They concern specific aspects of linguistic processing relevant to mapping sound and print. Wagner and Torgesen (1987); International Dyslexia Association, explanation of 2025 definition.
Representations, access, awareness, and memory
The phrase phonological processing covers more than one proposed mechanism. Some theories emphasize the precision and stability of speech-sound representations. Others focus on access to otherwise adequate representations, phonological working memory, or metalinguistic operations demanded by laboratory tasks. Poor performance on a phoneme-deletion test can reflect several component processes, including comprehension of instructions and executive demands. Strong studies use multiple measures, distinguish latent constructs from single tasks, and ask whether observed differences remain after accounting for reading experience. Wagner and Torgesen (1987); Melby-Lervåg, Lyster, and Hulme (2012).
The evidence for phonological involvement is unusually broad: early phonological skills predict later word reading; many children with dyslexia perform below peers on phonological-awareness tasks; and teaching the alphabetic principle and phoneme–grapheme connections can improve reading outcomes. The meta-analytic literature supports an important relationship while also showing that the strength and interpretation of individual indicators vary. A causal contribution is plausible and supported for particular skills and learning conditions, but one test of phonology cannot fully identify the developmental origins of an individual’s literacy difficulty. Melby-Lervåg, Lyster, and Hulme (2012); What Works Clearinghouse (2016, revised 2019).
Predictions and testable consequences
A strong phonological account predicts that children who find sound segmentation and print–sound integration difficult will, on average, have more trouble learning unfamiliar written words. It also predicts that instruction enabling reliable decoding should produce measurable benefits, especially when baseline decoding is weak. A more stringent version would predict that phonological deficits are sufficient and necessary for every case of dyslexia. Findings of reading difficulties with relatively preserved performance on common phonological tasks challenge that universal version, while measurement limitations complicate claims of a genuinely intact phonological system. Ramus et al. (2003); Peterson and Pennington (2015).
The phonological theory also predicts reciprocal development. Reading instruction can increase explicit phonemic awareness, and improved phonological analysis can improve decoding, creating feedback loops. Causality should therefore be investigated at multiple ages instead of inferred from a single cross-sectional contrast. When a reader has received substantial instruction, current phonological scores may look different from the original risk profile because of compensation, remediation, or repeated literacy exposure. Wagner and Torgesen (1987); Snowling and Hulme (2021).
The Evidence Supporting Phonological Explanations
Longitudinal prediction and family-risk research
Prospective studies measure precursor skills before formal reading acquisition and examine whether these predict subsequent literacy. Such designs help establish temporal order, but they cannot completely separate familial, linguistic, educational, and cognitive mechanisms. Children with family histories of dyslexia can show different combinations of early language weaknesses; oral-language risk is especially important for later reading and comprehension outcomes. Developmental risk is better understood as a trajectory that can be shaped by early language and print experiences than as a fixed destiny identified by a single precursor. Snowling and Hulme (2025); Snowling and Hulme (2021).
Intervention and experimental evidence
Experimental intervention evidence tests whether changes in instruction produce changes in reading. Systematic phonics and explicit connections between spoken and written units have a stronger educational evidence base than attempts to improve reading solely through general sensory or cognitive exercises. Still, treatment effects vary by students, outcome measures, intensity, developmental stage, and implementation quality. Interventions often contain several ingredients, so successful instruction supports the importance of particular learning processes without uniquely proving a single biological theory of dyslexia. Galuschka et al. (2014); What Works Clearinghouse (2016, revised 2019).
One frequent reasoning error is to assume that the success of phonics instruction proves that every person with dyslexia has one identical phonological impairment. A reading intervention can compensate for multiple upstream weaknesses by directly strengthening an important learning pathway. Conversely, an intervention with weak results does not show that phonology is irrelevant if the intervention was insufficiently intensive, poorly aligned to the deficit, or delivered after years of reading failure. Mechanism identification and outcome evaluation are related scientific questions that require distinct designs. Pennington (2006); Galuschka et al. (2014).
Evidence that narrows the strongest single-deficit claim
In a detailed multiple-case investigation, Ramus and colleagues assessed phonological, magnocellular, and cerebellar theories and found that explanatory claims involving broad sensory and motor abnormalities were harder to generalize than phonological findings. Such studies are especially informative because individual-case variation can be hidden in group means. Yet selected small samples cannot establish prevalence or settle every alternative mechanism. The most defensible conclusion is that phonological differences are major, highly replicated correlates and contributors, while their exact form and universality remain objects of research. Ramus et al. (2003); Ramus (2003).
A separate multifactorial analysis comparing phonological performance with reading-disability classification found substantial misclassification when phonological measures were used alone. That result does not refute the involvement of phonology; it shows why one cognitive index performs poorly as a stand-alone classification rule. Sampling, thresholds, age, language, task demands, and the outcome definition all affect calculated sensitivity and specificity. These considerations make it essential to test explanatory theories against independent reading measures rather than assuming that a theory-congruent test score is itself a diagnosis. Bridging sensory and language theories of dyslexia (2021); International Dyslexia Association, 2025 definition explanation.
The Double-Deficit Hypothesis: Phonology and Naming Speed
Why rapid automatized naming became important
Rapid automatized naming, commonly called RAN, measures how quickly a person names familiar visual stimuli such as letters, digits, colors, or objects in serial arrays. RAN performance predicts aspects of reading development, particularly automaticity and fluency, and can reveal difficulties that are not fully captured by phonemic-awareness measures. The double-deficit hypothesis proposes that phonological-awareness weakness and slow naming contribute partially distinct risks. Its strongest prediction is that readers with both weaknesses may show particularly persistent or severe problems. Vukovic and Siegel (2006); Wolf and colleagues, double-deficit perspective.
RAN tasks recruit coordinated visual identification, lexical retrieval, attention, articulation, sequencing, and processing speed. That breadth gives naming measures predictive value while making the causal interpretation difficult. A slow naming score does not reveal which subcomponent created the difficulty or prove a special naming-speed disorder. Naming and phonological measures also correlate, and investigators have used different definitions and cutoffs to produce double-deficit subgroups. The double-deficit hypothesis is therefore best treated as a testable account of partly distinct predictors rather than a fixed taxonomy that maps neatly onto every person with dyslexia. Vukovic and Siegel (2006).
Do the proposed subtypes remain stable?
A clinically useful subtype should be replicable across samples, ages, orthographies, and measurement batteries; it should forecast outcomes beyond ordinary continuous scores; and ideally it should alter intervention decisions. Reviews have found mixed evidence about the independence and stability of naming-speed deficits, and methodological differences complicate cross-study comparisons. Profiling both phonological awareness and RAN can be informative for research and educational planning, but declaring one of three immutable subtypes from brief tests would overstate the present evidence. Vukovic and Siegel (2006); Pennington (2006).
Rapid naming is particularly relevant when a reader can decode accurately yet remains very slow with connected text. This does not mean that a specific exercise targeting raw naming speed will necessarily produce fluent reading. Transfer from cognitive drill to real literacy needs experimental demonstration, including meaningful improvements in word reading, connected-text accuracy and speed, and retention. Educational interventions should follow measured learning needs and published outcome evidence rather than the superficial resemblance between a training game and a laboratory risk marker. What Works Clearinghouse (2016, revised 2019); Peterson and Pennington (2015).
Pennington’s Multiple-Deficit Model
Probabilistic risk instead of a single necessary mechanism
Pennington’s influential 2006 model explained why neurodevelopmental disorders often share risk factors and why a single-cognitive-deficit account struggles with comorbidity. In a multiple-deficit framework, numerous genetic and environmental influences contribute to intermediate neurocognitive processes, each of which affects reading outcomes with varying strength. Several risks may converge on low decoding accuracy, while protective skills, instruction, and experience can buffer the same vulnerability. The model is probabilistic: possession of one risk increases the probability of a reading difficulty without determining any individual outcome. Pennington (2006).
This approach gives a formal reason to expect different profiles within the same diagnostic category. One child may have strong spoken vocabulary, pronounced phonemic weaknesses, and adequate naming speed; another may have language-history difficulties, slow RAN, and weaker orthographic consolidation. Both may meet the same word-reading criterion, yet the pathways contributing to their difficulties differ. More complex models should earn their explanatory value through better prediction and replication. Listing many associated characteristics without specifying their unique or shared contributions is description, not a validated causal model. Peterson and Pennington (2015); Snowling and Hulme (2021).
Shared and disorder-specific risks
Dyslexia frequently co-occurs with attention-deficit/hyperactivity disorder, developmental language disorder, speech-sound difficulties, and other learning difficulties. Shared genetic or cognitive risk factors can contribute to these patterns, while other influences may be more closely tied to word reading or particular forms of oral language. The multiple-deficit model predicts comorbidity without assuming that all co-occurring conditions are manifestations of one disorder. Attention difficulties may affect learning opportunity, classroom engagement, and testing performance, but they do not automatically explain persistent decoding weakness. Pennington (2006); Snowling and Hulme (2021).
A model that treats overlapping traits as dimensions also prompts researchers to study controls beyond a simple dyslexia-versus-typical-reader comparison. Comparing children with dyslexia only, ADHD only, both, and neither can clarify shared and specific factors, although group assignment itself can create statistical artifacts. Continuous modeling of reading and attention may avoid some diagnostic-cutoff distortions. Multiple-deficit theory therefore changes research design: it encourages analyses of correlated liabilities, developmental mediation, heterogeneity, and the possibility that several paths lead to similar reading outcomes. Pennington (2006).
Risks, protection, and reciprocal feedback
Genetic liability does not specify a fixed developmental outcome. Educational access, instructional quality, vocabulary, language input, early screening, and targeted intervention interact with individual learning characteristics. A student with early decoding problems may read less voluntarily, gaining fewer chances to strengthen orthographic knowledge, vocabulary, and automatic word recognition. This feedback can widen differences over time. At the same time, a supportive teaching environment and accessible formats may protect participation and learning even when some word-level difficulties persist. International Dyslexia Association (2025); Snowling and Hulme (2025).
Theoretical models should specify how early risks operate over time. A factor that is causally central at kindergarten entry may lose predictive weight after years of instruction; conversely, accumulated print experience and language knowledge may become increasingly important. The same observed weakness can be an early cause in one trajectory and a downstream effect in another. Developmental multiple-deficit approaches accommodate these dynamic processes, although demonstrating any particular causal pathway still requires longitudinal evidence and experimental or quasi-experimental leverage. Snowling and Hulme (2021); Peterson and Pennington (2015).
Comparing the Main Models: What Each Predicts
Phonological account
A phonological account predicts relatively strong links among speech-sound processing, alphabetic decoding, nonword reading, and early literacy growth. It is particularly compelling when a child’s oral phonological weaknesses precede and predict later problems mastering letter–sound relations. Its limits appear when a single measured phonological variable is expected to explain all spelling, speed, language, and comorbidity outcomes. The claim should be decomposed into distinct submechanisms and tested for developmental timing rather than reduced to a universal one-test explanation. Melby-Lervåg, Lyster, and Hulme (2012); Ramus et al. (2003).
Double-deficit hypothesis
The double-deficit account predicts additive or interacting effects of phonological weakness and naming-speed weakness, particularly on reading fluency. Its research value lies in identifying cases where phonemic-awareness measures alone underestimate risk. The critical tests concern whether RAN adds prediction after accounting for other predictors, whether proposed subgroups replicate, and whether distinct intervention responses follow. A correlation between RAN and reading is a useful starting observation rather than sufficient evidence that naming speed independently causes dyslexia. Vukovic and Siegel (2006).
Multiple-deficit and developmental models
The multiple-deficit framework predicts heterogeneous and overlapping risk profiles, nonidentical pathways to similar reading outcomes, and co-occurrence with developmental conditions sharing some liabilities. It is a broad explanatory architecture rather than a single mechanistic algorithm. Its main challenge is operational precision: researchers must specify which risks, how they interact, the strength and developmental timing of their effects, and how the model would be falsified. Specific, measured predictions distinguish scientific multicausal explanation from a general claim that everything influences reading. Pennington (2006); Peterson and Pennington (2015).
A useful synthesis
These accounts can coexist at different levels. Phoneme-processing difficulties may be one common cognitive mediator, rapid naming may reveal another predictor or a partly shared processing pathway, and a multifactorial developmental model can explain how these arise and combine in specific environments. No single framework automatically establishes the treatment needs of a learner. The synthesis becomes meaningful only if it generates more accurate out-of-sample prediction, stronger accounts of developmental change, and better selection of educational supports than competing models. Snowling and Hulme (2021); Bridging sensory and language theories (2021).
Language, Morphology, and Orthographic Learning
Dyslexia is a language-and-literacy phenomenon
Phonology operates within a larger linguistic system that includes vocabulary, grammar, semantic knowledge, morphology, and the ability to understand connected discourse. Early oral-language weaknesses can forecast later literacy difficulty, especially when children must learn new sound–symbol mappings with limited language support. Morphological knowledge can assist word identification and spelling by revealing meaningful word parts and relationships. The 2025 dyslexia definition specifically recognizes phonological and morphological processing as common but nonuniversal areas of difficulty, consistent with multidimensional oral-language risk. International Dyslexia Association (2025); Snowling and Hulme (2025).
The Reading Is Language model emphasizes the developmental foundation provided by language to word reading, comprehension, and written expression. A student may develop relatively accurate decoding yet have persistent comprehension difficulty because oral-language knowledge remains weak; another may understand spoken stories very well but have severe word-decoding difficulty. These contrasting profiles show why word-level dyslexia, language disorder, and reading comprehension difficulty require distinct assessment questions. They also illuminate why a theory of word decoding should not be expected to explain every reading comprehension outcome. Snowling and Hulme (2025), Reading Is Language.
Orthographic learning as a developmental mechanism
As a learner successfully decodes unfamiliar words, repeated encounters help form representations that support rapid recognition and spelling. Developmental connectionist models describe learning as changes in distributed mappings among orthographic, phonological, and related information rather than as memorization of isolated rules. Such models show how different initial capacities and exposure patterns could create nonuniform literacy trajectories. Computational simulations can demonstrate that a mechanism is sufficient to generate certain patterns; establishing that the human brain uses exactly that mechanism requires convergent behavioral, developmental, and neurobiological tests. Seidenberg and McClelland (1989); Peterson and Pennington (2015).
Orthographic learning depends on both access to language and the distribution of print experiences. A student may know many spoken words yet struggle to identify their written counterparts quickly, or may gradually compensate through repeated word exposure while remaining slower with new strings. This interaction explains why decoding, spelling, and fluent recognition are related dimensions whose relative severity changes with instruction and age. It also supports studying word-level learning in the context of ordinary literacy, rather than inferring the entire process from a single visual-memory test. Ziegler and Goswami (2005); Snowling and Hulme (2021).
Theories of Auditory Timing, Sensory Processing, and Visual Attention
Temporal sampling and auditory theories
Auditory temporal theories investigate whether atypical perception or neural tracking of speech rhythms contributes to subsequent phonological difficulty. In the temporal-sampling framework, neural oscillatory alignment to speech at different timescales may help establish representations used for syllables, stress, rhythm, and phonemes. This is a mechanistic proposal linking sensory processing to language learning. The existence of group differences on auditory measures does not establish that auditory timing is the primary cause in every person with dyslexia, and specific neural-to-reading causal chains remain under investigation. Goswami (2011); Ramus (2003).
A useful test of an auditory theory asks whether a sensory measure predicts future phonological and literacy change independently of language, attention, hearing, and reading experience, and whether modifying the proposed sensory mechanism improves actual reading. Some sensory differences may be present in subgroups or arise alongside other developmental features. The presence of several pathways is compatible with a multiple-deficit account, but evidence quality must be judged separately for each proposed pathway. Theoretical plausibility by itself cannot justify replacing reading instruction with auditory exercises. Ramus et al. (2003); Bridging sensory and language theories (2021).
Magnocellular, visual, and cerebellar proposals
Other historical accounts emphasize visual motion processing, magnocellular pathways, binocular/oculomotor systems, or cerebellar skill automatization. These proposals have generated measurable hypotheses and important comparisons. However, heterogeneity, task confounds, and limited evidence of direct causal mediation restrict universal claims. Some children with dyslexia also have eye-health or binocular vision conditions requiring independent care, but such conditions do not explain the core persistent word-reading difficulty in most cases. Visual discomfort and learning disability should be investigated as distinct questions when appropriate. Ramus et al. (2003); Ramus (2003).
Theoretical critique is strongest when it avoids dismissing a finding merely because it is not universal. A visual-attention difference may be important for a subgroup, may be secondary to reduced reading fluency, or may reflect an overlapping but independent difficulty. Distinguishing these options requires longitudinal tests, well-matched comparison groups, and interventions with objective literacy outcomes. Claims that colored overlays, eye-training, or broad sensory stimulation correct dyslexia should be evaluated on reading-specific controlled evidence rather than testimonials or pre–post changes in visual comfort. Peterson and Pennington (2015); Galuschka et al. (2014).
Neurobiology and Genetics: Levels of Explanation
Neural networks are mechanisms and developmental outcomes
Reading acquisition gradually coordinates visual word processing with phonological, semantic, and attentional networks, often involving left-hemisphere language and occipitotemporal regions. Neuroimaging studies have documented group differences associated with dyslexia, but these differences can reflect risk, learning history, compensatory strategies, and changes produced through instruction. Neural observations become explanatory when tied to specific tested processes and developmental timing. A classification algorithm trained on a limited imaging dataset does not replace individual clinical or educational assessment. Norton, Beach, and Gabrieli (2015).
A brain-level account can coexist with cognitive theories. A neural difference might contribute to phonological learning, might reflect repeated reading difficulty, or might participate in several pathways at once. The strongest mechanistic studies connect imaging findings with behavior and repeated developmental measurements; intervention designs can help test change but often cannot isolate one causal structure. Inferring that dyslexia has a single brain location or that one image could distinguish all affected readers from their peers would go beyond the established evidence. Norton, Beach, and Gabrieli (2015); Peterson and Pennington (2015).
Polygenic risk and gene–environment processes
Familial aggregation and heritability research demonstrate substantial genetic contributions to individual differences in reading, while genetic influences operate in interaction and correlation with developmental environments. Heritability describes population-level variation under studied conditions; it is not an individual’s percentage of genetic causation. Shared genetic liabilities may help explain associations among dyslexia, language difficulties, and ADHD. An individual genetic test cannot currently establish a clinical diagnosis of dyslexia or replace the observation of reading skills and educational response. Peterson and Pennington (2015); Snowling and Hulme (2021).
Multiple-deficit models fit this genetic complexity because common variants and environmental influences can contribute through overlapping cognitive pathways rather than through one deterministic lesion. Nevertheless, multilevel diagrams can create an impression of precision that empirical data have not achieved. Genetic correlations do not identify particular developmental mediators, and statistical associations may reflect complex gene–environment relationships. Mechanistic claims must be evaluated at the correct level, with careful separation of association, prediction, mediation, and experimental causation. Pennington (2006); Peterson and Pennington (2015).
Dyslexia Across Languages and Writing Systems
The way dyslexia manifests depends partly on orthographic depth and the writing system being learned. In relatively consistent alphabetic orthographies, accuracy can improve earlier while slow, effortful reading remains prominent; in less consistent systems, decoding accuracy and spelling may remain especially demanding. Syllabic, morphosyllabic, and other writing systems recruit partly different grain sizes and linguistic information. The 2025 IDA definition explicitly includes variation in accuracy, speed, and spelling across orthographies, creating a cross-linguistic challenge for theories built exclusively on English-speaking samples. Ziegler and Goswami (2005); International Dyslexia Association (2025).
The psycholinguistic grain-size theory predicts that the relative importance of phonemes, syllables, larger orthographic units, and morphological information changes with the statistical structure of a written language. That does not imply that dyslexia is a different disorder in every language. It implies that the expression and measurement of underlying vulnerabilities depend on the learning problem posed by a particular orthography. Translating English phoneme tests or reading-rate thresholds without validation may produce misleading conclusions about a multilingual student. Ziegler and Goswami (2005).
Multiple-deficit models can be evaluated cross-linguistically by asking which predictors generalize, which are orthography-dependent, and how early instruction mediates relationships. Testing equivalent latent constructs across languages is difficult; matched-word tests may differ in frequency, length, morphology, and script transparency. Longitudinal multilingual designs and locally normed measures are particularly informative. A broad model gains credibility when its predicted interactions survive these challenges, rather than when findings from English are assumed universally applicable. Snowling and Hulme (2021); International Dyslexia Association, 2025 definition explanation.
Why Research Findings Sometimes Disagree
Different samples and operational definitions
Some studies classify dyslexia by word-reading accuracy, others by fluency, spelling, composite achievement, prior diagnosis, school services, or a cutoff relative to age expectations. Case–control comparisons can exaggerate apparent discontinuities when groups are drawn from the extremes of a continuous distribution. Studies using one phonological task may disagree with studies measuring several phonological dimensions because the tasks differ in reliability and executive or linguistic demands. These design variations influence claims about whether a cognitive deficit is present, necessary, or independent of other risks. Pennington (2006); International Dyslexia Association (2025).
Correlation, prediction, mediation, and intervention
A variable can distinguish affected readers without causing their reading difficulty. A predictor measured before reading instruction offers stronger temporal information, but an unmeasured common cause may still explain its association with literacy. Mediation models depend on assumptions about confounding and timing; cross-sectional mediation is particularly vulnerable. Randomized training improves causal inference about the intervention but cannot necessarily establish that one isolated cognitive mechanism explains all improvement. When reading outcomes themselves are measured unreliably or selectively reported, impressive mechanistic effects may not transfer to ordinary learning. Wagner and Torgesen (1987); Peterson and Pennington (2015).
Measurement invariance and individual differences
Measures of phonological awareness, RAN, working memory, and oral language do not have identical meaning across ages, languages, literacy experiences, or task formats. Floor effects can obscure performance among beginning readers, while ceiling effects can hide residual weaknesses in older students. A true individual difference requires a reliable profile across tasks and occasions rather than interpretation of an isolated discrepancy. Strong models specify measurement error, test generalization to independent samples, and report whether latent factors offer information beyond observed word-reading, spelling, and fluency scores. Melby-Lervåg, Lyster, and Hulme (2012); Vukovic and Siegel (2006).
What the Theories Mean for Screening and Assessment
Screening estimates risk; assessment identifies a learning profile
Universal early literacy screening can identify children who may need supplemental instruction or further evaluation. Measures often include letter–sound knowledge, phonemic awareness, rapid naming, decoding, and oral-language indicators, selected according to age and local validity evidence. Screening provides a risk estimate rather than a clinical diagnosis. A low phoneme score alone does not establish dyslexia; neither does a typical phoneme score exclude it when reading fluency or spelling is persistently impaired. The theory should direct the choice of hypotheses to test, with observed academic performance remaining central. International Dyslexia Association, 2025 definition explanation; American Psychiatric Association.
A comprehensive evaluation examines standardized word-reading accuracy, decoding of unfamiliar words, reading rate and fluency, spelling, comprehension as relevant, instructional history, oral-language skills, developmental and family history, functional effects, and opportunities for effective instruction. The evaluator considers hearing, vision, language proficiency, intellectual and broader developmental factors, attention, and other possible contributors as indicated. Such assessment can identify co-occurring disorders and support appropriate accommodations without requiring an unvalidated phonological, RAN, or neuroimaging subtype. International Dyslexia Association (2025); American Psychiatric Association.
Can a cognitive profile distinguish one theory from another?
A profile showing very low phonemic awareness and relatively preserved naming is compatible with a phonological pathway; a profile dominated by naming-speed difficulty prompts more questions about automaticity, processing demands, and independent literacy evidence. Neither pattern is a clean causal experiment. Differences among standardized cognitive scores may reflect measurement error, developmental compensation, or environmental opportunity. Clinicians and educators should document what a profile describes, whether competing interpretations remain viable, and which documented reading needs warrant instruction. Vukovic and Siegel (2006); Pennington (2006).
IDEA eligibility decisions in the United States apply educational definitions and rules governing evaluation and services. A learner can qualify for additional instructional support based on documented needs even when investigators disagree about the most complete mechanism of those needs. Conversely, theoretical claims do not establish eligibility by themselves. Early intervention should proceed when meaningful literacy risk is evident instead of being postponed until a subtype label or exhaustive etiological explanation is available. Individuals with Disabilities Education Act regulations; International Dyslexia Association, 2025 definition explanation.
What the Theories Mean for Instruction and Intervention
Teach the literacy skill that is demonstrably difficult
Evidence-based teaching connects language with print through explicit, systematic instruction in phonemic awareness where needed, letter–sound correspondences, decoding, word recognition, spelling patterns, reading connected text, and relevant language skills. The balance changes with age and observed performance. A learner with inaccurate unfamiliar-word decoding may need a different emphasis from a learner whose decoding is accurate but markedly slow. Educational planning should monitor transferable word-reading, fluency, and spelling outcomes instead of focusing exclusively on a hypothesized cognitive deficit. What Works Clearinghouse (2016, revised 2019); Galuschka et al. (2014).
A multiple-deficit perspective encourages the practitioner to measure progress across related domains rather than assume that improving one score will normalize every outcome. Persistent language weaknesses may call for complementary vocabulary and oral-language instruction; attention difficulties may require access and engagement supports; slow reading may call for targeted guided practice and accommodations. This is a needs-based educational principle, not evidence that a highly individualized cognitive-profile-matching algorithm outperforms established instruction. Comparative intervention studies are needed to establish that stronger claim. Snowling and Hulme (2025); Pennington (2006).
What models cannot justify by themselves
A theory about auditory oscillations, visual attention, cerebellar automatization, or rapid naming does not by itself validate a marketed program targeting that process. Brain-training tasks, colored overlays, sensory stimulation, and branded dyslexia packages require controlled evidence of durable, meaningful improvements in literacy over suitable comparison conditions. Mechanistic change, student satisfaction, and improvements in a trained task are weaker endpoints than generalizable word-level reading, spelling, and functional academic progress. Claims of unique superiority should be tested independently against high-quality alternatives. Galuschka et al. (2014); Peterson and Pennington (2015).
When a learner receives targeted instruction, accommodations such as accessible text, text-to-speech, additional time when functionally indicated, or spelling assistance can support access to content. These adjustments address participation while explicit instruction addresses teachable literacy processes. Evidence for the most effective intervention depends on age, baseline skill, outcome, fidelity, and dose. The multiple-deficit framework encourages attention to the whole developmental profile, but direct reading and writing measures remain the indispensable tests of educational benefit. What Works Clearinghouse (2016, revised 2019); International Dyslexia Association (2025).
Evidence Status: Established Findings and Open Questions
Findings with substantial support
There is strong support for the role of phonological skills in learning alphabetic reading; persistent word-reading and spelling difficulties can be observed across the lifespan; reading outcomes emerge from genetic, developmental, language, instructional, and environmental influences; and explicit, research-supported literacy teaching can improve important outcomes. Independent variations in severity, fluency, spelling, and co-occurring developmental needs are well documented. These findings justify substantial phonological emphasis within a wider developmental account. Melby-Lervåg, Lyster, and Hulme (2012); Snowling and Hulme (2021); International Dyslexia Association (2025).
Plausible mechanisms with developing evidence
Specific pathways from auditory temporal processing to phonological representations, the precise causal independence of rapid naming, and the detailed mapping from genomic liabilities to neural and cognitive reading mechanisms remain subjects of active research. Multiple-deficit models provide a useful framework, but individual factor weights, interactions, and developmental transitions remain incompletely specified. Computational reading models generate testable accounts of learning but require empirical validation against diverse human trajectories. The strength of evidence should be assigned separately to each proposed mechanism, not inherited from the general plausibility of multifactorial causation. Goswami (2011); Vukovic and Siegel (2006); Seidenberg and McClelland (1989).
Claims that remain contested or insufficiently demonstrated
A universal single cognitive marker for all dyslexia cases, stable RAN-defined diagnostic subtypes, clinical diagnosis from a brain scan, and reliable selection of specialized branded interventions from a brief cognitive profile have not been established. Observations about subgroups can motivate further studies without functioning as general diagnostic rules. Results from small samples, poorly matched controls, or training confined to the test task warrant particular caution. A responsible evidence hierarchy favors replicated longitudinal and experimental outcomes over appealing neuroscientific imagery or anecdotal success. Ramus et al. (2003); Vukovic and Siegel (2006); Norton, Beach, and Gabrieli (2015).
Research Priorities for a More Precise Theory
Future empirical work can strengthen the field by measuring reading, spelling, phonology, naming speed, oral language, orthographic learning, attention, and educational input repeatedly from before school into later development. Models should preregister clear predictions, quantify uncertainty, test alternative causal structures, and report results for independent samples. Combining dimension-based analyses with diagnostic comparisons may reveal nonlinear thresholds while reducing artifacts caused by arbitrary cutoffs. Cross-linguistic replication is essential for discovering which mechanisms generalize across different learning systems. Pennington (2006); Ziegler and Goswami (2005).
Intervention studies can compare instructional components directly, measure transfer to meaningful literacy outcomes, track durability, and examine whether baseline characteristics predict differential benefit. Even when a model predicts treatment moderators, researchers must test the interaction rather than simply show that both groups improved. Studies of bilingual learners, older readers, and individuals with overlapping language or attention difficulties can determine where current explanations fail or need refinement. Such research makes theories accountable to the developmental realities and educational outcomes they are intended to explain. Galuschka et al. (2014); Snowling and Hulme (2021).
Frequently Asked Questions
Is dyslexia caused by a phonological deficit?
Phonological processing difficulties are a common and well-supported contributor to dyslexia, especially during the acquisition of alphabetic reading. The evidence does not support treating one measured phonological weakness as a universal necessary condition or sufficient diagnostic rule. A complete explanation considers individual variability in language, fluency, spelling, development, and instruction. International Dyslexia Association (2025); Melby-Lervåg, Lyster, and Hulme (2012).
What is the difference between a phonological theory and a multiple-deficit theory?
The phonological theory identifies a key candidate mechanism in the sound-to-print learning process. A multiple-deficit theory explains why different combinations of risks and protective factors can produce related reading outcomes and why developmental conditions overlap. Phonological difficulties can be one important pathway within a multiple-deficit model. Pennington (2006).
Does the double-deficit hypothesis establish distinct dyslexia types?
It proposes partially distinct phonological and naming-speed risk dimensions and predicts that their combination may be associated with greater difficulty. Subtype boundaries and the independence of RAN remain methodologically contested, so the hypothesis is useful for research and assessment questions without supplying an automatic diagnosis or separate treatment prescription. Vukovic and Siegel (2006).
Can brain scans identify which theory applies to a person?
Neuroimaging can investigate group-level neural mechanisms and developmental change. Current imaging results do not establish an individual diagnostic test or a validated method for assigning a learner to one cognitive theory. Literacy performance, history, and comprehensive professional evaluation remain central. Norton, Beach, and Gabrieli (2015).
Do dyslexia theories apply equally across languages?
Theories must explain both shared linguistic mechanisms and the different demands of alphabetic, morphosyllabic, and other writing systems. Orthographic consistency and morphological structure affect whether accuracy, speed, or spelling difficulties become most visible. Valid assessment requires suitable language-specific instruments and developmental comparisons. Ziegler and Goswami (2005); International Dyslexia Association (2025).
Which theory should guide a child’s reading intervention?
Instruction should be selected according to documented literacy and language needs and supported by intervention evidence. Phonological theory strongly informs explicit decoding instruction, while multiple-deficit perspectives encourage attention to co-occurring needs, risk, and contextual supports. No theory automatically licenses replacing demonstrated reading instruction with unvalidated cognitive training. What Works Clearinghouse (2016, revised 2019); Galuschka et al. (2014).
Conclusion: From Core Mechanisms to Developmental Explanation
A mature theory of dyslexia must account for the strong phonological foundations of word reading, the incomplete fit of any single risk marker, individual differences in accuracy and fluency, and the effects of language, instruction, and developmental experience. Phonological accounts identify processes of major explanatory importance; multiple-deficit models organize the interacting risks that shape different trajectories. Their integration is scientifically productive when it yields explicit, falsifiable predictions and educationally meaningful improvements. The practical center remains accurate measurement of literacy needs and access to effective teaching at every age. Catts, Haynes, and Joshi (2026); Snowling and Hulme (2021).
Related Articles
Specific Learning Disorders and Dyslexia — Scientific research chapter on developmental learning disorders, diagnostic frameworks, and reading, writing, and mathematics.
Phonological Awareness and Dyslexia: What Research Shows — Focused review of phoneme processing, prediction, causal evidence, and reading instruction.
Dyslexia and the Brain: Reading Networks and Neuroplasticity — Neural-network evidence, neuroplasticity, and the limits of brain-based diagnosis.
Dyslexia and Vision: Eye Movements, Colored Overlays, and Evidence — Appraisal of sensory and visual theories and proposed vision interventions.
ADHD and Executive Function — Attention and executive-function mechanisms relevant to developmental co-occurrence.
Autism and Adult Neurodiversity — Distinctions among neurodevelopmental profiles and individual developmental differences.
References
American Psychiatric Association. (n.d.). What are specific learning disorders? https://www.psychiatry.org/patients-families/specific-learning-disorder/what-is-specific-learning-disorder
Catts, H. W., Haynes, C. W., and Joshi, R. M. (2026). Defining dyslexia: 2025 revision. Annals of Dyslexia. https://doi.org/10.1007/s11881-026-00363-4
Galuschka, K., Ise, E., Krick, K., and Schulte-Körne, G. (2014). Effectiveness of treatment approaches for children and adolescents with reading disabilities: A meta-analysis of randomized controlled trials. PLOS ONE, 9(2), e89900. https://doi.org/10.1371/journal.pone.0089900
Goswami, U. (2011). A temporal sampling framework for developmental dyslexia. Trends in Cognitive Sciences, 15(1), 3–10. https://doi.org/10.1016/j.tics.2010.10.001
Individuals with Disabilities Education Act. (n.d.). 34 CFR § 300.8, Child with a disability. https://sites.ed.gov/idea/regs/b/a/300.8
International Dyslexia Association. (2025). Definition of dyslexia. Adopted October 22, 2025. https://eida.org/definition-of-dyslexia/
International Dyslexia Association. (2025–2026). Explanation of the 2025 updated definition of dyslexia. https://dyslexiaida.org/explanation-2025-dyslexia-definition-explanation/
Lovett, M. W., Steinbach, K. A., and Frijters, J. C. (2000). Remediating the core deficits of developmental reading disability: A double-deficit perspective. Journal of Learning Disabilities, 33(4), 334–358. https://pubmed.ncbi.nlm.nih.gov/15493096/
Melby-Lervåg, M., Lyster, S.-A. H., and Hulme, C. (2012). Phonological skills and their role in learning to read: A meta-analytic review. Psychological Bulletin, 138(2), 322–352. https://doi.org/10.1037/a0026744
Norton, E. S., Beach, S. D., and Gabrieli, J. D. E. (2015). Neurobiology of dyslexia. Current Opinion in Neurobiology, 30, 73–78. https://doi.org/10.1016/j.conb.2014.09.007
O’Brien, G., and Yeatman, J. D. (2021). Bridging sensory and language theories of dyslexia: Toward a multifactorial model. Developmental Science, 24(3), e13039. https://doi.org/10.1111/desc.13039
Pennington, B. F. (2006). From single to multiple deficit models of developmental disorders. Cognition, 101(2), 385–413. https://doi.org/10.1016/j.cognition.2006.04.008
Peterson, R. L., and Pennington, B. F. (2015). Developmental dyslexia. Annual Review of Clinical Psychology, 11, 283–307. https://doi.org/10.1146/annurev-clinpsy-032814-112842
Ramus, F. (2003). Developmental dyslexia: Specific phonological deficit or general sensorimotor dysfunction? Current Opinion in Neurobiology, 13(2), 212–218. https://doi.org/10.1016/S0959-4388(03)00035-7
Ramus, F., Rosen, S., Dakin, S. C., Day, B. L., Castellote, J. M., White, S., and Frith, U. (2003). Theories of developmental dyslexia: Insights from a multiple case study of dyslexic adults. Brain, 126(4), 841–865. https://doi.org/10.1093/brain/awg076
Seidenberg, M. S., and McClelland, J. L. (1989). A distributed, developmental model of word recognition and naming. Psychological Review, 96(4), 523–568. https://doi.org/10.1037/0033-295X.96.4.523
Snowling, M. J., and Hulme, C. (2021). Annual research review: Reading disorders revisited — The critical importance of oral language. Journal of Child Psychology and Psychiatry, 62(5), 635–653. https://doi.org/10.1111/jcpp.13324
Snowling, M. J., and Hulme, C. (2025). Risk factors for dyslexia: Addressing oral language deficits. Mind, Brain, and Education, 19(3), 168–175. https://doi.org/10.1111/mbe.70009
Snowling, M. J., and Hulme, C. (2025). The Reading Is Language model: A theoretical framework for language and reading development and intervention. Annual Review of Developmental Psychology, 7, 195–218. https://doi.org/10.1146/annurev-devpsych-111323-084821
Vukovic, R. K., and Siegel, L. S. (2006). The double-deficit hypothesis: A comprehensive analysis of the evidence. Journal of Learning Disabilities, 39(1), 25–47. https://doi.org/10.1177/00222194060390010401
Wagner, R. K., and Torgesen, J. K. (1987). The nature of phonological processing and its causal role in the acquisition of reading skills. Psychological Bulletin, 101(2), 192–212. https://doi.org/10.1037/0033-2909.101.2.192
What Works Clearinghouse. (2016; revised 2019). Foundational skills to support reading for understanding in kindergarten through 3rd grade. Institute of Education Sciences, U.S. Department of Education. https://ies.ed.gov/ncee/WWC/PracticeGuide/21/Published
Ziegler, J. C., and Goswami, U. (2005). Reading acquisition, developmental dyslexia, and skilled reading across languages: A psycholinguistic grain size theory. Psychological Bulletin, 131(1), 3–29. https://doi.org/10.1037/0033-2909.131.1.3