The normal-variability Evo-Devo conceptualization of dyslexia is fundamentally at odds with an alternative, more traditional Evo-Devo model (Jimenez-Bravo et al., 2017; Murphy and Benitez-Burraco, 2018; Benitez-Burroco and Murphy, 2019). At the core of the traditional orientation is the assertion that dyslexia results from early onset structural aberrations of neuronal migration across multiple cortical subregions (e.g., Jimenez-Bravo et al., 2017). From this array of regional developmental anomalies, the traditionally oriented Evo-Devo model of dyslexia places special emphasis on epigenetic by environmental interactions in brain evolution, which adversely affect the species-specific, evolving genes that regulate cortical migration and patterns of brain rhythms (oscillations). The model portrays evolutionary and developmental selection of phase and phase/amplitude couplings in speech as the origin of human language, and the genes involved as the primary source of deficits associated with various categorical clinical language disorders, including dyslexia. This leads to a diagnosis of dyslexia as an “oscillopathic disease.” In contrast, the normal-variability approach, with its focus on the stress/dyslexia connection, argues that evolution in child development prioritizes selection for the genes that regulate the range and timing of neuroplasticity. Neuroplasticity of specific neuronal circuitry, then, determines the successful entrainment of patterns of brain rhythms: patterns compromised in poorer readers and suspected to be stress-linked. Also see Soloduchin and Shamir (2018) for an evolutionary account of brain rhythms as secondary to the evolutionary selection of neuroplasticity.

Both models are premised on the powerful role of the environment in the epigenetic regulation of gene expression. Numerous studies support the primary role of the environment in shaping cis-regulatory regions of the neuro-genome, which conserve the gene’s nucleotide sequence during phylogenetic and ontogenetic development (Brun et al., 2009; Petanjek et al., 2011; Somel et al., 2011; Zhang et al., 2011). However, in our model, such gene regulation typically produces high, but normal interindividual variability in the behaviors and brain regions undergoing evolutionary expansion (i.e., reading) (Mueller et al., 2013). And, even at the lowest levels of ability, such regulatory variability does not reflect oscillopathic disease. Rather, it is the unexceptional outcome of an emerging evolutionary skill under positive selection pressure. Thus, dyslexia in the normal-variability model is a dimensional disability shaped by selection for neuroplasticity, as opposed to a categorical abnormality associated with selection for patterns of brain oscillations.

The more traditional view of dyslexia as a medical condition has come under increasing criticism from scholars in the field (Peters and Ansari, 2019; Protopappas, 2019; Protopappas and Parrila, 2019; Guidi et al., 2018), and from researchers who design their studies in dyslexia with the understanding that reading ability varies along a normal continuum (e.g., White et al., 2019; Zuk et al., 2019). Detailed consideration of the scientific merits of the brain abnormality claim is beyond the scope of this paper. Nonetheless, some background and several counter-intuitive examples, should raise questions about the assumption of abnormality and may serve as an entry point for future more comprehensive discussions.

Research support for approaching dyslexia within the framework of a traditional medical model invariably dates back several decades to examinations of eight postmortem brains and clinical studies of adult neuropsychological patients. There can be no doubt that rare genetic variants producing neurological abnormalities may be associated with developmental reading difficulties (Czamara et al., 2011). However, such limited evidence provides no justification for inferring that these select cases are representative of the large numbers of individuals with reading difficulties, across languages and in every literate society. A national testing program of grades 4–12 in over 8,000 schools in the U.S. revealed that only 35% of the students were grade-efficient in reading (National assessment of educational progress [NAEP],, 2015). The sheer number of school-age children and young adults with reading difficulties argues against an underlying medical condition. A second line of argument calling for reconsideration of the assumption of underlying pathology comes from the critical research review by Guidi et al. (2018). Their analysis revealed that recent experimental and gene function studies failed to substantiate the link between dyslexia and a deficit in the migration of neurons in the developing neocortex, which is a major claim undergirding the abnormality assumption and the Evo-Devo model of dyslexia as a developmental aberration (Jimenez-Bravo et al., 2017). Guidi et al. (2018) concluded their review by recommending a “fresh start” in our endeavors to understand the true nature and underlying neurobiology of dyslexia.

Predictions from the two theoretical platforms differ substantially. For example, in the traditional approach, reduced patterns of neuronal activation revealed by functional imaging experiments are expected to reflect poorer performance, while increased levels reflect better performance. In the normal-variability model, evolution seeks to maintain stress/growth homeostasis to meet task demands. The release of stress hormones, interacting with local glutamate levels, are predicted to dynamically modulate neuronal response in a classic inverted U-shaped Yerkes–Dodson, arousal response curve (e.g., Devilbiss, 2019). As a result, task performance will be degraded at points far-left and far-right on the arousal-performance curve. Hypoactivation and hyperactivation can detract equally from performance, and in any event the range of response is seen as adaptive variability. The normal-variability model predicts that task performance will be maximized at a moderate level of neuronal activation.

In addition to patterns of brain activation, neuroimaging research in dyslexia has reported more or less white or gray matter volume, more or less cortical gyrification, or differences in the efficiency of signal transmission in white matter tracts (e.g., Demonet et al., 2004). Such quantitative variations are all influenced by experience and expected in the course of normal development. Moreover, during reading and on tests of phonological awareness, good and poor readers engage the same brain areas but at different levels of intensity (Ligges et al., 2010; Kovelman et al., 2012). Indeed, Mascheretti et al. (2017) tallied 26 studies in dyslexia showing hyperactive brain areas and 42 studies showing hypoactivity. These studies represent a diversity of lines of investigation and are important for their theoretical significance. However, a valid inference of brain impairment requires evidence of structural, and behaviorally significant abnormalities in neurological organization observed in large studies with representative samples. Finally, the widely recognized candidate genetic variants for dyslexia (i.e., DYXIC1, K1AA0319, DCDC2, ROBO1) may only become risk genes if the time course of their expression is altered by experience, including stressful events, again, a part of normal development (Kershner, 2019).

In summary, the belief that dyslexia is caused by a brain abnormality is clouded by counter evidence and a lack of supporting research. Nonetheless, the two Evo-Devo models, which stand in marked contrast on this issue, need to be evaluated more specifically on their own terms. The two models differ primarily on two fundamental points. In the current model, dyslexia is a normal developmental variation resulting from selection for the genes that regulate neuroplasticity which, in turn, is essential for maintaining a balance between stress protection and cognitive growth. In dyslexia, reduced stress-related plasticity of the cognitive growth side of the equation upsets this balance. In contrast, the alternative model portrays dyslexia as a neuropathological condition originating from selection during primate brain evolution of the genes regulating the brain rhythms specific to human speech and language. Dyslexia results from a deleterious gene mutation or abnormal genetic patterns of rhythmic gene expression. Both models have identified poorer entrainment of specific oscillations in the speech envelope as the final outcome of these developmental pathways, and as an etiological factor in dyslexia. However, in the normal-variability model, poor speech tracking and entrainment are a secondary manifestation of compromised neuroplasticity. And, finally, the distinguishing feature of the normal-variability model hypothesizes that poorer oscillatory entrainment stems from adaptations to stress system dysregulation.

Oscillations in the flow of speech output occur simultaneously over multiple time scales in fractions of a second. The auditory receptive system has to attend, select, segment, and parse relevant sensory and linguistic features from this vocal stream for assimilation, which primarily involves frequency oscillations in the delta (1–3 Hz), theta (4–8 Hz), and low gamma (typically 35–45 but can go to 90 Hz) ranges. In the normal flow of speech expression, this amplitude envelope of modulated frequencies, varying in intensity, carries linguistic information that must be phase-locked or entrained by the receiving brain and continuously recalibrated, not only for comprehension but to support attention, memory, and the phonological processes in reading (Gross et al., 2013; Murphy, 2015; Goswami, 2019). For instance, tracking, fine-tuning and resetting by the brain (entrainment) to the rhythms in the phonological elements of speech involve the delta band for consolidating the mental representations for tonal and prosodic perception, theta for syllables, and gamma for phonemes (Goswami, 2011).

Postnatally, when brain oscillations entrain to the speech envelope, neuronal excitability facilitates the structural organization of the brain circuitry that subserves these three components of phonology. Firmly establishing the phonological lexicon in memory storage in early childhood provides the neuronal basis for making rhyme and rhythm judgments, sensitivity to syllabic stress and syllables, and phonemic awareness: all required subsequently for recruiting the word/sound/meaning pathways essential for fluent reading (Goswami, 2019). Decoding written words and sentences for meaning depends on the quality of the hierarchical structure in memory of these oscillations and their phonological units. Simply stated, early entrainment sets the stage, via attentional controls, for the visual word form in beginning reading to activate the distributed representations of the oscillatory and phonological hierarchies. During this protracted period of learning, phrasal tonality comes before syllabic stress and awareness of syllables, and syllabication comes before learning the phonemic constituents of the syllables. Learning to read and read well depends on the sequential, top–down acquisition of this hierarchy, and the ability to rapidly access from memory the entire package when decoding connected text. Thus, the degree of successful phase synchronization of speech oscillations in early childhood may have a direct bearing on poor reading in light of the generally held view that poor phonological coding is a signature feature of dyslexia (e.g., Guidi et al., 2018).

We can reasonably assume an embryonic automatic period of auditory entrainment (e.g., Telkemeyer et al., 2011), but postnatally, entrainment is not stimulus driven (Gross et al., 2013; Poeppel and Assaneo, 2020). Such speech tracking or entrainment is a core ingredient of endogenous selective attention (e.g., Lakatos et al., 2008; Obleser and Kayser, 2019), and is thought to be modulated by regional cortical areas that work together (Power et al., 2016). These topologically linked and co-activated neuronal assemblies are strongly right hemisphere lateralized (Golumbic et al., 2012; Daitch et al., 2013; Szczepanski et al., 2014; Marshall et al., 2015b; Spagna et al., 2015, 2016, 2018; Poeppel and Assaneo, 2020). Of particular interest, is that the low frequency oscillations (i.e., delta, theta), which entrain the slower wide-band fundamental rhythmic components of the phonological dictionary, are under control of the right lateralized, ventral attention (VAN) and bilateral dorsal attention (DAN) networks (Gross et al., 2013).

Pronounced hemispheric asymmetries in speech tracking are supported further by the Asymmetric Sampling in Time (AST) model of speech processing (Poeppel, 2003; Giraud and Poeppel, 2012). After initial registration bilaterally in primary auditory areas, speech entrains the phase of delta/theta in the right hemisphere, and the amplitude of gamma in the left, where gamma is down-sampled by theta to match the optimum frequency for selective extraction and coding of phonemic information. Generally, gamma is committed to local left hemisphere processing in the 35–45 Hz range, while delta and theta (1–8 Hz) in the right may also drive cross-hemispheric long-distance communication. Moreover, low frequency entrainment and cross-hemisphere signaling depend on the initiation of synaptic plasticity (Hahn et al., 2019). These traveling oscillations are flexibly coupled, with the phase of the slower frequencies powering the amplitude of the faster frequencies (Zhang et al., 2018; Benitez-Burroco and Murphy, 2019). For instance, in such phase/amplitude couplings (PACS), the phase of delta oscillations can modulate the amplitude of theta, and the phase of theta, as in processing phonemes, can modulate the amplitude of gamma. Theta/gamma PACs may have from 4 to 8 cycles of gamma for every cycle of theta, and the precision of the ratio is thought to facilitate computations for specific linguistic operations (Murphy, 2015). It follows, that gamma under- or over-sampling, tied to poor quality low frequency entrainment, would result in an unusual and faulty temporal format for phonemic categories. Finally, In reading there is some evidence that theta and gamma have their origins in the hippocampus, are coupled by the thalamic reticular nucleus, and arguably forwarded by the right attentional networks across the corpus callosum (the main conduit carrying signals between hemispheres) to instantiate the excitability needed for phonemic processing by the left hemisphere dorsal reading network (Gross et al., 2013; Marshall et al., 2015a; Murphy, 2015; Molinaro et al., 2016; Meyer, 2017).

To summarize, converging evidence supports the view that oscillations dedicated to enhancing the brain’s phonological lexicon and required for reading, are powered by low frequencies, and are phase and PAC modulated by the predominantly right ventral (VAN) and dorsal (DAN) attention networks for (1) early establishment, (2) later accessibility, and (3) routing of information between networks and hemispheres.

The coordinated functions of the right hemisphere ventral (VAN) and bilateral dorsal (DAN) attention networks (Corbetta et al., 2008; Petersen and Posner, 2012; Daitch et al., 2013; Gross et al., 2013; Duecker and Sack, 2015; Frarrant and Uddin, 2015; Spagna et al., 2016), which play a key role in synchronizing brain oscillations (cf. Thiele and Bellgrove, 2018), evolved as a survival mechanism to bring perceptions to consciousness, and to regulate content processing networks (Petersen and Posner, 2012). VAN’s primary hubs are the right inferior frontal cortex (IFC) and right temporal parietal junction (TPJ). DAN’s primary hubs are the right frontal eye field (FEF) and bilateral intra-parietal sulci (IPS). DAN’s right hemisphere frontoparietal circuit controls multimodal orienting across the span of attention, while VAN modulates DAN’s interhemispheric rivalry between its bilateral posterior hubs interconnecting the right IPS with the left IPS (Duecker and Sack, 2015). VAN and DAN receive modality-specific inputs, and are coordinated top–down in hierarchical control by right hemisphere hubs of the salience (SN) and frontoparietal (FPN) networks, which together, form a supramodal cognitive control network (CCN) (Menon and Uddin, 2010; Spagna et al., 2015, 2018; Wu et al., 2018, 2019). Controlling hubs of the SN are the right frontal insular cortex (FIC) and the right frontal cingulate cortex (FCC). Main hubs of the FPN, which more directly modulate the attentional networks, are the right dorsolateral prefrontal cortex (DLPFC) and right posterior parietal cortex (PPC).

Thus, VAN and DAN, with delta and theta as main carrier frequencies (e.g., Gross et al., 2013) and oversight from the FPN, may modulate the entrainment and processing of the oscillations that make up the repertory of the brain’s phonological information. VAN and DAN are allied closely in their functions and richly interconnected by: (1) right posterior parietal nodes in common with the FPN (Petersen and Posner, 2012); (2) the second branch of the right superior longitudinal fasciculus (Chica et al., 2018); and (3) the right posterior middle frontal gyrus (Corbetta et al., 2008). DAN is activated by alerting signals and expectations, and motivated by current goals, directs the focus of attention, orienting, and response selection. VAN responds with DAN by gating behaviorally relevant inputs, inhibiting distractions when DAN is focused, and is a circuit breaker for reorienting and resetting attentional focus (Corbetta et al., 2008; Daitch et al., 2013).

It is a reasonable conjecture that VAN/DAN’s survival functions, nurtured by evolution in the pursuit of attentional controls, have been adaptively appropriated by selection pressure to subserve the low frequency rapid focusing, reorienting, and resetting of attention needed for (1) the entrainment of human speech and (2) literacy. In human evolution, VAN and DAN have been repurposed, but retain their primitive survival functions. A frequent view of the behavioral systems involved in VAN/DAN cooperative interaction are alerting and orienting (Fan et al., 2009; Petersen and Posner, 2012; Spagna et al., 2015). The alerting function sustains a state of arousal and readiness, while orienting selects the most relevant endogenous and exogenous events. The supramodal cognitive control network (CCN) coordinates the alerting and orienting functions, resolves conflicts among competing mental events, and organizes response selection. Hence, the VAN/DAN complex of coactivated controls is well-suited in its genetic predisposition for (1) the selective alignment of neuronal oscillations with the rhythms of incoming speech, (2) maintaining the focus of attention for sampling, and (3) continuously resetting the entrainment process to accommodate the rapid flow of the amplitude envelope.

Behavioral research has shown that individuals with dyslexia are poorer in aspects of both VAN/DAN functions (Goldfarb and Shaul, 2013; Gabay et al., 2020), and imaging studies with children and adults with dyslexia have reported weaker encoding of the delta band (Power et al., 2013, 2016; Soltesz et al., 2013). Testing for oscillations by hemisphere interactions, multiple studies have converged in identifying the right hemisphere as the source of the atypical encoding of the low auditory frequencies in dyslexia (Hamalainen et al., 2012; Lizarazu et al., 2015; Cutini et al., 2016; Molinaro et al., 2016; Di Liberato et al., 2018) (for a review, see Kershner, 2020b). The studies by Cutini et al. (2016); Power et al. (2016), and Di Liberato et al. (2018) compared children with dyslexia to age and reading-level matched groups of good readers, effectively eliminating the possibility that the low frequency anomaly may result from their poorer reading. The study by Cutini et al. (2016) localized the right hemisphere’s poorer entrainment to the supramarginal gyrus (SMG) and angular gyrus (AG), key nodes of the TPJ with common posterior parietal connectivity with both VAN and DAN. Finally, there is some evidence that individuals with dyslexia may fail to show left hemisphere dominance for phonemic processing of gamma, and may oversample gamma in the right (Lehongre et al., 2011; De Vos et al., 2017); both likely caused by poorly specified delta/theta entrainment, with delta the first priority as delta provides the fundamental underlying power within the phonological hierarchy (Meyer, 2017).

Goswami (2011, 2019) was the first to hypothesize atypical encoding by the right hemisphere of low frequency delta/theta oscillations as an etiological factor in dyslexia. Faulty VAN/DAN attentional controls as the source of the entrainment deficiency adds a new dimension of support for Goswami’s temporal sampling framework (TSF). When the brain cannot detect temporal order in the speech stream, a continuous processing mode suppresses low frequency oscillations (Lakatos et al., 2008). According to the TSF, the right hemisphere’s atypical entrainment and insensitivity to amplitude changes or resets in the low frequency range (<10 Hz) prevents consolidation of the phonological building blocks that are called upon subsequently for efficient phonemic processing by the left hemisphere. It is generally assumed that phonemic processing by the left is carried out by PAC interhemispheric transfer powered by the right. However, Di Liberato et al. (2018) has presented evidence that low frequencies also entrain phonemic features directly in the right hemisphere. Thus, the VAN/DAN right hemisphere constellation of integrative attentional controls may entrain delta, theta, and to some extent, also gamma.

In summary, a considerable pool of experimental evidence supports the hypothesis that poor modulation of the entrainment and synchronized processing of auditory low frequency oscillations by VAN and DAN may be an etiological factor in dyslexia.

Assuming this to be relatively well-established, a salient issue in the dyslexia puzzle becomes the identification of the experiential and heritability circumstances that undermine the developmental integrity of the brain’s attentional networks. The current Evo-Devo model proposes that the evolutionary selection for the timing of neuroplasticity in early childhood is a key organizing biological principle in dyslexia (Kershner, 2020a). Plasticity, defined as reorganizational capabilities in neuronal circuits and behavior in response to patterns of sensory inputs, is an essential requirement for learning to read (Vandermosten et al., 2016). The Evo-Devo model suggests that neuroplasticity may be curtailed by adversity, which destabilizes stress/growth homeostasis, leading in infancy or early childhood to suppressed low frequency entrainment by the attentional networks.

Of fundamental importance to understanding the merits of an adversity/plasticity/attention networks and reading connection, the human brain is highly neotenous (Somel et al., 2009; Bufill et al., 2011; Petanjek et al., 2011; Somel et al., 2011; Liu et al., 2012; Miller et al., 2012; Somel et al., 2014; Matsuda et al., 2020). Neoteny, or developmental allochrony, refers to the prolongation of high cortical metabolism and synaptic plasticity in regional neocortical areas in humans compared to other primates. This period of extended neuroplasticity is rooted in evolutionary cortical expansion, reflecting the regions and networks that are under positive selection for ontogenetic change and less influenced by genetic factors. Precise calculations of neuroplasticity can be made regionally and at the level of individual voxels by using PET-based measurements of aerobic glycolysis (AG, non-oxidative metabolism of glucose). For this calculation, AG is inferred whenever less than 6 molecules of oxygen are consumed for each molecule of glucose. From an early peak of maximum plasticity, the neuronal metabolism of AG demonstrates a curvilinear decrease, extending into the seventh decade of life (Goyal et al., 2017). And, significantly, AG corresponds to the energy-expensive gene expression programs supporting cognitive growth (Goyal et al., 2014, 2017; Magistretti, 2016).

In addition, repeated-measurement resting state fMRI can reveal intersubject variability in connectivity within and between specific brain networks, which may reflect points of current evolutionary selection (Mueller et al., 2013). The upshot of this is that both measurements across the brain’s networks converge in finding (1) superior AG ratings (Vaishnavi et al., 2010; Blazey et al., 2018), and (2) topmost interindividual variability (Mueller et al., 2013) in the frontoparietal control and attentional networks, i.e., FPN, VAN and DAN. From this, we can conclude that in typical development the right hemisphere attentional networks, which are dominant over the left (e.g., Gross et al., 2013; Spagna et al., 2018), are poised for growth with an elevated level of neuroplasticity.

In dyslexia, however, this plasticity appears to be compromised (e.g., Cutini et al., 2016). Two major stress-reactive neuromodulatory systems for arousal and attention are mediating candidates in dyslexia for a linkage between stress and the loss of attentional network plasticity. One is the HPA system (McGowen and Mathews, 2018; Raymond et al., 2018) and the other is the Locus coeruleus-norepinephrine (LC/NE) system (Glennon et al., 2019). Under mild stress, negative feedback loops from the HPA to the brain serve to moderate homeostatic balance, simultaneously buffering against the potential adverse effects of strong emotions while facilitating neuroplasticity in higher cortical regions (Kershner, 2020a). However, as disturbing events accumulate in number, intensity, and duration, the HPA system may go into allostatic overload (Burns et al., 2018). When that happens, stress protection becomes paramount. The HPA releases a supraoptimal flood of glucocorticoids (cortisol in humans), which can potentially have toxic neurological effects (Peters et al., 2017; McGowen and Mathews, 2018). Such stress effects have been shown to influence synaptic number, dendritic spine formation, and arbor shaping, in the hippocampus, amygdala, and prefrontal cortex (PFC). To mitigate extensive cellular metabolic damage, individuals have the innate adaptive capacity to accelerate maturation of the brain’s emotional circuits (Bath et al., 2016; Callaghan and Tottenham, 2016) and areas subserving higher cognitive processes (Gur et al., 2019). Faster development is an evolutionary strategy which counters neoteny, but may increase survival and fitness in uncertain and threatening environments (Krugers et al., 2017; Ellis and Del Giudice, 2019). Precocious maturation dampens the HPA production of cortisol, but at a cost of neuroplasticity in the expanding brain regions subserving recently evolved and evolving skills such as reading (Wagner et al., 1997; Peters et al., 2017; Gollo et al., 2018; Benitez-Burroco and Murphy, 2019). The adaptive effect on the brain’s hierarchical organization is a loss of top–down control, which offsets the stress/growth balance by reallocating neural resources from higher cognitive functions to stress guardianship (Elzinga and Roelofs, 2005; Arnsten, 2009; Qin et al., 2009; Roozendaal et al., 2009; van Marie et al., 2009; Zhang et al., 2020). At its core, the HPA system acts as a master switch regulating stress vs. growth genetic programs (e.g., de Kloet et al., 2005; Peters et al., 2017). Thus, stress-induced HPA dysregulation compromising neuroplasticity to favor stress management is one potential pathway for a stress connection to the attentional networks.

The second stress system, the LC/NE system, has a more direct effect on the attention networks. It has its origin in the noradrenergic locus coeruleus, a small nucleus located in the brainstem (Glennon et al., 2019). The LC/NE system, which evolved to support defensive behaviors and integrate autonomous functions with higher cognitive functions, releases norepinephrine (NE) diffusely throughout the brainstem, midbrain, cerebellum, and neocortex (Totah et al., 2019). NE interacts with local glutamate levels, the brain’s primary excitatory neurotransmitter, to modulate arousal and states of alertness in a Yerkes–Dodson inverted U-shaped arousal-response curve. Glutamate induces synaptic plasticity, but if excessive can lead to neurodegeneration (Yan et al., 2020). On the one hand, moderate LC/NE stimulation via thalamocortical relays, increases the magnitude of neuronal gain (responsiveness) in primary visual and auditory pathways, improving the fidelity of signal transmission by increasing the signal/noise ratio (Waterhouse and Navarra, 2019). But, of equal significance, subsets of LC neurons have evolved to target and modulate the efficiency of higher-order multisensory signal transmission in the brain circuits controlling arousal, orienting, and attending to behaviorally relevant stimuli (Aston-Jones and Cohen, 2005; Totah et al., 2019).

Of particular relevance, the LC/NE system has efferent pathways that feed NE selectively to the right hemisphere VAN and DAN via right thalamocortical relays (Aston-Jones and Cohen, 2005; Corbetta et al., 2008; Grefkes et al., 2010; Petersen and Posner, 2012; Thiele and Bellgrove, 2018). Indeed, evidence suggests that the LC/NE system coactivates VAN and DAN, which become phase modulated at delta/theta rhythms in controlling the focus of attention and switching between networks (Daitch et al., 2013; Gross et al., 2013). Conditions of unexpected uncertainty can drive LC/NE tonic output to high levels, having the potential to suppress slow-wave synchrony and block network resets (Aston-Jones et al., 1997; Corbetta et al., 2008; Mather et al., 2016). Thus, stress-induced excessive LC/NE release has the potential to directly down-regulate plasticity to favor stress management and alter the efficiency of the attention networks.

Consideration of the combined activation of both stress systems provides a cohesive theoretical basis for causal linkages between stress axis dysregulation and reduced neuroplasticity of the attentional networks in dyslexia. But, aside from research showing that dyslexia is associated with stress-related genes (Zakopoulou et al., 2019), with dysregulation of the HPA (Espin et al., 2019; Huang et al., 2020), and with reduced plasticity of the attentional networks (Cutini et al., 2016), appreciation of the proposition is too new to have motivated the systematic research needed to test its validity. However, a recent study offers optimistic support for one of the model’s predictions (Elhadidy et al., 2019). BDNF is a brain-derived neurotropic factor that can influence the expression of sets of genes regulating synaptic neuroplasticity in response to stress (Gray et al., 2013; Gregorenko et al., 2016; Peters et al., 2017). BDNF is widely distributed, with concentration in areas of cortical expansion (i.e., VAN/DAN), where metabolism by aerobic glycolysis (AG) incites the expression of BDNF and plasticity genes (Magistretti, 2016). Excessive Stress reduces levels of BDNF and AG plasticity, and is associated with impaired memory consolidation (e.g., Menezes et al., 2020). Elhadidy et al. (2019) confirmed lower levels of plasma BDNF in a Canadian sample of 28 boys and 14 girls (6–12 years) with dyslexia compared to age-matched good readers. The results were strongly supportive of the hypothesis, with no overlap between groups in BDNF levels. In the dyslexic group, BDNF ranged from 0.86 to 1.34 ng/ml with a mean of 1.10. In the control group, BDNF ranged from 1.60 to 2.40 ng/ml with a mean of 2.00. The authors recommended BDNF testing as a biomarker for dyslexia. More cautiously, the results call for replications with reading-level controls, but are a significant endorsement of a relationship between stress-related, diminished neuroplasticity and dyslexia.

Finally, a remarkable characteristic of children’s developmental stress system reprogramming is that it can also result from epigenetic patterns of maternal or paternal inheritance (Roth et al., 2009; Kolb et al., 2012; Burns et al., 2018; Posner et al., 2019). Based on animal models, stress experienced by parents prior to mating has been shown to alter the behavioral and cortico-limbic functions of their offspring. Research has only skimmed the surface of the multiple factors involved. However, the epigenetic methylation status of HPA system promoter genes and BDNF have been identified as candidate pathways for mediating such transgenerational effects (Roth et al., 2009; Burns et al., 2018). A study that deserves some scrutiny in this context, shows apparent support for the inheritance of a right hemisphere attentional deficit linked to dyslexia (Thiede et al., 2019). In a speech-sound discrimination study, 44 at-risk, 8 to 9 day-old newborn infants were compared to no-risk infants, matched on gender, gestational and measurement age, and parental educational level. Hemispheric EEG recordings of mismatched responses (MMR) revealed that the at-risk infants failed to show typical right hemisphere vowel change, neural speech-sound discrimination. However, the study did not test for oscillations, and does not confirm an epigenetic pattern of inheritance, or whether parental stress was an issue. Nonetheless, it does suggest a hereditary basis for one aspect of underperformance of the attentional networks.

In summary, VAN and DAN, modulate low frequency entrainment, are brain networks characterized by protracted developmental features (i.e., neoteny), and are integral in day-to -day functions with the cortico-limbic stress systems. Stress that exceeds an individuals range of resilience has the potential to (1) suspend neuroplasticity, responding to cortisol outflows by the HPA system, in areas of high cortical expansion, i.e., the attentional networks and (2) suppress VAN/DAN neuroplasticity directly by overabundant NE release from the LC/NE system. Both systems are candidates for the stress-induced, compromised neuroplasticity that may underly the disrupted delta/theta entrainment in dyslexia. Such epigenetic, transcriptional and functional reprogramming of the stress axis may occur prenatally or in early childhood, and we have to allow for the possibility of inheritance from the preconceptual stress experienced by parents. Still absent, however, are more direct tests confirming the specific stress system(s) pathways between stress and the attentional networks.

The research reviewed supports the significance of VAN and DAN, and the potential role of stress, in children’s acquisition of the discrete phonological fundamentals required to easily learn how to read. However, for a comprehensive understanding of dyslexia, how this plays out in the broader context of collaboration among the brain’s major networks may be essential (Bailey et al., 2018). Evo-Devo’s core concept of fitness as a stress/cognitive growth balance is imbedded in the brain’s network architecture, where affective reactions to stress are focused internally and processed by the default mode network (DMN), while cognitive requirements are focused externally and controlled by the frontoparietal network (FPN) and attentional networks (Dixon et al., 2017; Schultz et al., 2019).

In coordinating network adjustments to stress, a network stress/growth tradeoff is keyed by activation of the central core of the DMN. The central core is made up of the medial prefrontal cortex (MPFC), precuneus, and posterior cingulate cortex (PCC), and is an extension of the brain’s stress axis (Andrews-Hanna et al., 2010; Wu et al., 2018; Satpute and Lindquist, 2019). The DMN central core, largely independent of sensory input, is thought to engage in self-referential processing, high-level emotion, and autobiographical memory (Dixon et al., 2017).

Stress that challenges the stress/growth balance can be defined at three levels: (1) the cellular metabolic level (Halliday and Mallucci, 2019; (2) a psychological level, as uncertainty (Peters et al., 2017); and (3) in terms of energy transfer in evolution, as entropy (Pryluk et al., 2019). It is a fundamental principle in evolution, and organic playout of the 2nd law of thermodynamics, that evolution seeks to maximize the production of entropy in non-linear and non-eqilibrium biological systems (Bazarov, 1964; Jeffrey and Rovelli, 2020). A balance between stress and growth during development is such a system. Greater entropy promotes increased cognitive complexity and the flexibility of permitting selection among competing thoughts and actions (cf. Fan et al., 2014). In this context, a stress/growth balance is orchestrated within a homeostatic range by dynamic fluctuations between the lower entropy DMN (stress biased) and high entropy FPN and attentional networks (growth biased). In effect, in stress management, evolution strives to maximize plasticity by maintaining a progressive state of a system’s non-equilibrium within a range favorable to cognitive growth (Martyushev, 2013). The salient point is that stress usually functions to keep the system in relative balance, but if excessive may circumscribe entropy at all three levels, reflected in reduced plasticity of potential arrangements of the stress/growth system. Stress, depending on the source, duration and intensity, may (1) lower plasma BDNF, (2) limit the strategies available in coping with uncertainty, and (3) reduce the complexity of high-level cognitive reasoning. Supraoptimal stress reallocates metabolic, behavioral, and energy resources from growth to the more primitive, stable side of the equation.

The CCN (i.e., cognitive control network which combines the salience and frontoparietal networks) and the attentional networks, which work together in reading (Horowitz-Kraus et al., 2015; Ihnen et al., 2015; Freedman et al., 2020), are typically anti-correlated with the default mode network (DMN) (Qin et al., 2009; Spreng et al., 2012; Dwyer et al., 2014; Fan et al., 2014; Utevsky et al., 2014; Dixon et al., 2017, 2018; Wu et al., 2018; Hugdahl et al., 2019; Zhang et al., 2020). An inverse correlation between the CNN-attentional network hierarchy and the DMN is exemplary of a stress/growth dynamic balance. For example, Fan et al. (2014) found that activation of the CCN including the DAN, in parallel with deactivation of the DMN, varied parametrically as a function of uncertainty. Establishing the same point, Wu et al. (2018) reported that activation of the CCN increased and activation of the DMN decreased as a function of entropy.

Research using dynamic causal modeling with a large sample of adults (n = 404) helps to put this tradeoff in an operational framework (Zhou et al., 2018). Dynamic causal modeling, an at-rest imaging procedure which provides measures of efferent vs. afferent connectivity, was used to test the causal hierarchical structure of the anticorrelation between the core default mode network (DMN) and the dorsal attention network, DAN. The study revealed that the salience network (SN), was at the top of the hierarchy, directing the anticorrelation between the DMN and DAN. Under SN top–down surveilance, the stress/growth tradeoff was modulated by excitatory signals from the DMN to DAN, simultaneously with inhibitory influences from DAN to the DMN (cf. Menon and Uddin, 2010; Critchley and Harrison, 2013; Udden, 2015; Mai et al., 2019). This reciprocity creates feedforward and backward loops that serve to stabilize the low entropy DMN under stress, while nurturing and giving an edge to the entropy needed for continued cognitive growth. Thus, moderate levels of stress serve to balance the stress/growth system and entropic balance between networks. However, excessive stress has the potential to partially reverse this dynamic by subduing DAN’s bias toward increasing entropy and inhibitory controlling pathways to the self-centered DMN.

When the DMN was first recognized as a content processing rather than control network (Power et al., 2011), its negative correlations with regional cortical areas actively engaged in task performance were puzzling and not readily interpretable. However, an Evo-Devo perspective predicts an inverse relation between the stress-activated DMN and the cognitive growth networks, and characterizes the relationship as a reflection of favorable behavioral variability, and a potential index of optimally balanced competition (cf. Koyoma et al., 2010). Thus, a negative correlation between the stress and growth networks may be a manifestation of the stress-growth genetic program trade-off, signaling the fluidity of an ongoing evolutionary process among networks.

Research by Bailey et al. (2018) aimed to determine the relative patterns of resting-state activation across the brain’s reading networks. The researchers applied the 7-cortical networks identified by Yeo et al. (2011) to a meta-analytical data base of 11,406 imaging studies. The results demonstrated that the FPN, VAN, and DAN combined to make up 56% of the brain’s reading network activations. Thus, a large proportion of the reading brain’s activation at rest is attributable to executive control and attentional networks. These core controlling networks are thought to be anatomically separate from the traditional reading regions recruited for processing cognitive content (Petersen and Posner, 2012; Ihnen et al., 2015). Therefore, substandard development of the brain’s attentional controls would necessarily deflate the growth side of the stress/growth balance, with deleterious consequences for both reading acquisition and later reading fluency. More specifically, a stress-induced allostatic overload favoring the DMN side of the equation has the potential for negative effects at every phase of reading skill development by overriding the processing control and growth functions of the CCN, VAN, and DAN.

The VAN/DAN complex as an origin of reading difficulties is consistent with the causal link between atypical low frequency speech encoding by the right hemisphere in dyslexia (e.g., Goswami, 2019), but it is also consistent with the acknowledged role that selective attention plays in the superior visuospatial processing ability of better readers (Zhao et al., 2011; Franceschini et al., 2012; Gabrieli and Norton, 2012; Vogel et al., 2012; White et al., 2019).

The visuospatial processes with relevance to dyslexia involve the magnocellular-dorsal stream (MDS). Multiple studies have found evidence linking dyslexia to the MDS, visuospatial orienting, and the right posterior parietal cortex (e.g., Facoetti et al., 2001; Franceschini et al., 2018; Fu et al., 2018; Vidyasagar, 2019; Archer et al., 2020). The MDS or “where” stream relays retinal signals, via the lateral geniculate nucleus and along secondary visual pathways, to the right hemisphere dorsal attention network (DAN), which leads in top-down feedback control of visuospatial processing (Underleider and Mishkin, 1982; Grefkes et al., 2010). Projecting downstream, DAN (right frontal eye field and bilateral intraparietal sulci) is the interhemispheric controlling network in the MDS, with its posterior parietal axons capable of modulating gamma in the left hemisphere (Marshall et al., 2015a, b). Evidence suggests that early stages of reading are associated with increased connectivity from the left posterior parietal cortex to the visual word form area (VWFA) for both graphemic and phonological processing (Desroaches et al., 2010; Centanni et al., 2019; Moulton et al., 2019). The VWFA is a region in the ventral occipital cortex of the left hemisphere, thought to become finely tuned for reading whole words. Thus, in reading, the temporal flow of signals from the MDS and auditory pathways may be regulated by VAN/DAN, and feedforward by phase amplitude couplings (PACs) to the VWFA for word recognition and letter to sound correspondence. This invites the hypothesis that the hypoactivation of the VWFA documented in poor readers (Centanni et al., 2018a, 2019) may result from an impairment in the right attentional networks. More specifically, such a deficit, possibly resulting from stress-linked NE release to VAN/DAN, would disrupt downstream auditory and visual signal transmission, suggesting VAN and DAN as a common multisensory origin of the auditory-phonological and visuospatial deficits in dyslexia (cf. Facoetti et al., 2008; Gori and Facoetti, 2015).

The putative central role of the attention networks in dyslexia provides a blueprint for a more comprehensive model of reading and dyslexia composed of widely separated patterns of connectivity across distributed brain networks, involving the CCN, VAN/DAN, and the DMN. For instance, such a model suggests that when reading connected text for comprehension, top–down control by the CCN may modulate continuous information transactions in thought between the DMN and the attentional networks. The internally focused DMN would be tasked with extracting the emotional tone of events as they interact with our autobiographical memory and self-image, while the externally focused attentional networks assimilate the narrative flow of information in the text. However, excessive stress or a dysregulated stress axis would reset this processing balance, curtailing VAN/DAN’s negative inhibitory controls over the DMN and overengaging the DMN. Indeed, research has shown that subclinical stress may cause the DMN to disengage from the executive networks (Schultz et al., 2019). Similarly, acute stress in healthy individuals has been shown to produce large-scale network reconfigurations resulting in reduced activation of the FPN coupled with greater within network connectivity of the DMN (Qin et al., 2009; Zhang et al., 2020). When such an imbalance in processing favoring the DMN happens, individuals may become self-absorbed and unable to access the attentional and cognitive resources needed to sustain a favorable balance of the stress/growth equation. Some support for the applicability of this scenario to dyslexia comes from studies showing that better readers exhibit a negative correlation between the DMN and bilateral reading areas (Koyoma et al., 2010, 2011), while poorer readers have shown stronger connectivity between the DMN and reading-related areas (Schurz et al., 2015). Unfortunately, there is a paucity of network research in dyslexia, and none of the studies to date have included controls over reading-level. As a result, the proposed network model stands as a theoretical proposal, pending future research to examine its validity.

To summarize, the evolutionarily conserved balance between the stress/cognitive growth genetic programs is embedded in the architecture of the brain’s major attentional control and emotional networks. Maintaining a favorable range in the homeostatic competition between these networks, i.e., VAN, DAN, and the DMN, appears to be critical to both the phonological and visuospatial processing requirements of reading. Stress, defined in physiological, psychological or entropy terms, can challenge the growth side of this equation by overengaging the DMN at the expense of the functional integrity of VAN and DAN. When network homeostasis fails, evolution favors stress management. Finally, while a stress/growth balance impacting neuroplasticity is a well-established principle in Evo-Devo, firming up its applicabilty to an interactional network understanding of dyslexia will require more direct experimental evidence.