Towards a Brain-Computer Interface (BCI) for Improving Phonological Processing in Developmental Dyslexia: An Exploratory Study

51 Brain-computer interfaces (BCIs) have immense potential regarding the provision of 52 therapies for disorders of development, but to date have typically been created for non-53 linguistic disorders such as ADHD (attention deficit hyperactivity disorder). Here we 54 present a BCI that aims to improve linguistic phonological processing in developmental 55 dyslexia. Phonological ‘deficits’ are considered a core feature of dyslexia across 56 languages. A non-invasive EEG-BCI relying on auditory inputs and visual feedback was 57 developed to optimise brain patterns related to phonology (speech-sound processing). 58 These patterns were identified using Temporal Sampling (TS) theory, which proposes that 59 phonological difficulties in dyslexia are related to impaired auditory processing of 60 amplitude envelope rise times and low-frequency speech envelope information <10 Hz. 61 These impairments are thought to affect automatic features of speech processing from 62 birth, impairing the development of a phonological system. Adults with and without a 63 diagnosis of developmental dyslexia played the BCI for 16 sessions, and received pre- 64 and post-testing regarding phonological awareness and single word and nonword 65 reading skills. Significant associations between their BCI scores (a measure of BCI 66 learning) and improvements in syllable stress discrimination, nonword reading and 67 amplitude rise time discrimination were found. The data are interpreted with respect to TS 68 theory. 69 70 Key words: Brain–Computer Interface, Dyslexia, Phonological Processing, Temporal 71 Sampling Theory, EEG 72 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 4 1. Introduction 73 Theories of developmental dyslexia attempt to provide a systematic causal 74 framework for understanding this specific learning difficulty (e.g., Magnocellular theory, 75 Stein & Walsh, 1997; Visual Attention Span theory, Valdois et al., 2004; Sluggish 76 Attentional Shifting theory, Facoetti et al., 2010, Temporal Sampling [TS] theory, 77 Goswami, 2011). A focus on development is absolutely critical to identifying core factor/s 78 for effective remediation, accordingly here the focus is on the phonological ‘core deficits’ 79 that pre-date learning to read (Stanovich, 1998), and on TS theory. Theories focused on 80 the visual system are not considered, as typically the theorised deficts can only be 81 detected once reading instruction commences (see Goswami, 2022a, for a recent survey 82 of dyslexia theories). Regarding the phonological ‘core deficit’, studies in many 83 languages have demonstrated that a key developmental factor in the etiology of dyslexia 84 is phonological learning. Via the natural acquisition of spoken language, infants and 85 children implicitly learn a phonological system comprising the sounds and combinations 86 of sounds that are permissible in their language/s, long before reading instruction 87 commences (Kuhl, 2004). In effect, their brains develop phonological representations of 88 the sound structures of individual words, via automatic sensory-motor learning, and TS 89 theory proposes that this automatic learning is impaired in dyslexia. The current BCI 90 focuses on phonological learning at the level of speech rhythm patterns, the factor that 91 governs infant language acquisition across all languages studied to date (Mehler et al. 92 1988; Nazzi et al., 1998). 93 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 5 The development of ‘phonological awareness’ (PA) in children is typically measured 94 by behavioural performance in PA tasks, simple oral tasks that explore a child’s ability to 95 consciously detect and manipulate the component sounds in words at all linguistic levels 96 (speech rhythm and prosody, syllables, rhyme, phonemes, see Ziegler & Goswami, 97 2005). These phonological impairments persist into adulthood, although in consistent 98 orthographies like Italian, German or Spanish, in adulthood they are indexed by 99 significantly impaired speed in PA tasks (Landerl & Wimmer, 2000; Ziegler et al., 2010). 100 In inconsistent orthographies like English, phonological difficulties in adulthood can be 101 indexed by impairments in both speed and accuracy in PA tasks (Snowling, 2000). PA 102 follows a similar developmental sequence across languages, predicts reading acquisition 103 in all languages so far studied, and is impaired in children with dyslexia across languages 104 (Ziegler & Goswami, 2005). Training phonological skills, particularly in the pre-school and 105 earliest school years, can significantly mitigate the impact of a family risk for dyslexia 106 (Schneider et al., 2000). Accordingly, the current BCI for dyslexia was developed to 107 remediate the unconscious neural factors associated with inefficient phonological 108 processing. 109 As phonological learning in infants begins with speech rhythm, recent infant EEG 110 (Electroencephalography) studies of neural speech processing also informed the design 111 of the BCI. When infants listen to sung infant-directed speech, which is highly rhythmic, 112 cortical tracking of low-frequency speech envelopes appears to come online first 113 (measurable from 2 months of age), notably in the delta and theta electrophysiological 114 bands (0.5 – 4 Hz, 4 – 8 Hz, see Attaheri et al., 2022; Ni Choisdealbha et al., 2023). This 115 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 6 low-frequency cortical tracking also underpins the learning of phonetic information (Di 116 Liberto et al., 2023), which begins to emerge around 7 months. Individual differences in 117 both delta-band cortical tracking at 11 months and in the ratio of theta-delta PSD (power 118 spectral density) predict individual differences in language outcomes at age 2 years 119 (measured by vocabulary tests and a nonword repetition task, Attaheri et al., 2024). More 120 accurate delta band cortical tracking and a lower theta-delta ratio predicted better 121 language outcomes. These infant studies were informed by TS theory, an auditory theory 122 of dyslexia, which also informed the current study (Goswami, 2011, 2015, 2022b). 123 With respect to dyslexia, TS theory proposes that sensory/neural processing 124 differences regarding speech prosody (speech rhythm patterns) lead affected children to 125 develop atypical phonological representations of spoken language, from infancy onwards 126 (Goswami, 2022a). Neurally, adult studies suggest that speech is encoded by 127 neuroelectric oscillations (rhythmic changes in electrical brain potentials in large cell 128 networks) which respond to different temporal levels of speech information (such as 129 phrases, syllables and phonemes, Giraud & Poeppel, 2012; Gross et al., 2013). TS 130 theory suggests that in developmental dyslexia, encoding of low-frequency envelope 131 information <10 Hz (delta and theta band information) is impaired, in part because of 132 poorer auditory discrimination of amplitude ‘rise times’ in the speech envelope. ‘Rise 133 times’ in amplitude (the rates of change between sound onset and sound peak in a given 134 amplitude modulation, AM) provide sensory landmarks that automatically trigger brain 135 rhythms and speech rhythms into temporal alignment, via phase-resetting ongoing neural 136 activity (Doelling et al., 2014). This phase-resetting process is known to be impaired in 137 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 7 dyslexia (Lizarazu et al., 2021). The amplitude envelope is the slow-varying energy 138 contour of speech that determines the perception of speech rhythm (Greenberg, 2006), 139 and it contains a range of AM patterns at different temporal rates which broadly match 140 EEG rates such as delta, theta and beta/low gamma. Further, speech modelling studies 141 of infant- and child-directed speech show that the phase relations between these different 142 AM rates provide systematic statistical cues to phonological units such as stressed vs 143 unstressed syllables, syllables, and onset-rimes (‘acoustic-emergent phonology’, Leong 144 & Goswami, 2015; Leong et al., 2017). Accordingly, a nascent phonological system can 145 be extracted from the speech signal via the automatic alignment of neuroelectric 146 oscillations to the AM information in speech via efficient phase-resetting driven by 147 amplitude rise time (ART) discrimination. 148 Children with dyslexia in a range of languages exhibit impaired ART discrimination 149 compared to chronological age matched-controls (English, Spanish, French, Finnish, 150 Chinese, Hungarian, and Dutch; Goswami, 2015, for review). Children with dyslexia 151 learning English, Spanish and French also show impaired neural encoding of low-152 frequency speech envelope information in the delta and theta neurophysiological bands 153 during natural speech listening (DiLiberto et al., 2018; Molinaro et al., 2016; Destoky et 154 al., 2020; other languages not yet tested). A BCI for dyslexia could therefore target neural 155 encoding directly, for example via improving phase locking values (see Arias, Molinaro & 156 Lizarazu, 2021). However, TS-driven studies have shown that one neural marker of 157 impaired phonological processing appears to be the theta-delta oscillatory ratio during 158 natural speech listening (Attaheri et al., 2024; Araújo et al., 2024). During continuous 159 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 8 speech listening, English-speaking children with dyslexia show a higher theta-delta ratio 160 than control children, which is significantly related to their offline performance in PA tasks 161 (a higher ratio is associated with worse performance, Araújo et al., 2024). Further, 162 English-learning infants aged 4 – 11 months with a higher theta-delta ratio during 163 continuous speech listening go on to exhibit poorer language skills at 24 months (poorer 164 vocabulary and nonword repetition, see Attaheri et al., 2024). The recent developmental 165 research base thus suggests that the theta-delta ratio during natural speech listening 166 could also be an effective target for a BCI for dyslexia. 167 These TS-driven developmental data informed the current BCI. The aim of the BCI 168 was to change the ratio of the neural oscillations that (by TS theory) underpin statistical 169 learning of the AM hierarchy, thereby ameliorating the ‘phonological deficit’ in dyslexia. A 170 non-invasive BCI targeting the self-regulation of low-frequency (delta and theta) neural 171 oscillations during natural speech listening was developed by the second author as part 172 of his PhD and piloted with 15 adult participants, 7 of whom had a statement of dyslexia. 173 Araújo (2023) designed an engaging interface based on a space ship rocketing up into 174 space, aimed at teaching learners of the BCI how to self-regulate their own theta-delta 175 ratio by controlling the space ship’s position using their brains. A closed-loop operant 176 learning BCI was created, in which learners aimed to make the space ship ascend as far 177 as possible on the gaming window in each of 16 BCI sessions (described in detail in 178 Araújo et al., 2023). Participants received a stronger visual reinforcement (the screen 179 glowed greener) the higher they made the spaceship go. No visual reinforcement was 180 given if the spaceship’s position remained below a threshold line located across the 181 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 9 middle of the gaming window. Listening to the audio signal of a story as input, the 182 participant was then encouraged to try out cognitive strategies focused on auditory 183 processing to modulate their oscillatory patterns that controlled the spaceship. The 184 spaceship’s position was estimated via real-time classification of time-series EEG data 185 using a pre-trained signal processing and machine learning model (described below). 186 This feedforward model shows minimal computational overhead, allowing for smooth 187 online control of the BCI with minimal lags. 188 In the original paradigm, successful BCI learning was indexed by whether the 189 spaceship position distribution of session 1 had a significantly lower mean than session 190 16 (using a t-test, see Araújo, 2023). The value of the t statistic became the participant’s 191 ‘BCI Score’, the magnitude of which reflected the degree to which the participant had 192 reduced their theta-delta ratio. Inspection of the BCI scores suggested that 12 of the 15 193 participants had learned the BCI successfully (2 controls and 1 dyslexic did not learn). 194 The baseline-normalized band frequency magnitude across the learners’ delta and theta 195 rhythms was then used to compare their distributions in session 1 with distributions from 196 session 16. The data showed that the BCI helped participants to reduce their theta-delta 197 ratio by significantly increasing neural signal magnitude for the slower delta rhythm and 198 significantly decreasing it for the faster theta rhythm. Further, individual BCI scores were 199 associated with significant improvement in the speed of syllable stress discrimination 200 judgements (r= 0.59, p< .05) and showed a trend in improvement for single word reading 201 as measured by the TOWRE (Test of Word Reading Efficiency, Torgesen, Wagner & 202 Rashotte, 1999; r= 0.48, p= .07). The cortical dynamics targeted by the BCI should (by 203 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 10 TS theory) only improve phonology and reading. Adult participants also received a test of 204 arithmetic reasoning (WRAT, Wide Range Achievement Test, Snelbaker et al., 2001) 205 during the study, and as expected BCI scores were not associated with changes in 206 arithmetical reasoning from pre-test to post-test (Araújo, 2023). 207 The original paradigm was developed during the Pandemic, therefore the BCI was 208 based on a g-tec hardware set-up which was not suitable for taking to schools and using 209 with children. Accordingly, as a further pilot, in the current study the second author 210 adapted the closed-loop operant learning system to work with mobile EEG headcaps 211 specifically the CGX Quick-20m wireless headset. This system employs dry electrodes 212 recorded positioned at 19 scalp locations following the International 10-20 system (P1, 213 FP2, F3, F4, Fz, F7, F8, C3, C4, T3, T4, T5, T6, P3, P4, Pz, O1, and O2), with A1 and A2 214 serving as linked-ear references. Signals were digitized at 24-bit resolution and sampled 215 at 500 Hz. The portability and ease of setup of this dry electrode system make it 216 particularly well suited for use in schools with children, where the application of traditional 217 EEG systems would be impractical. A new group of adults with and without dyslexia were 218 recruited by the first author, and received a similar protocol to that used in Araújo (2023), 219 which is described fully below. Participants were pre- and post-tested on a range of 220 phonological, reading and control tasks (detailed below) before and after 16 gaming 221 sessions with the BCI. The hypothesis was that learning the BCI would improve their 222 neural theta-delta ratios during natural language listening, and that this improvement 223 (indexed by their BCI scores based on spaceship position, the measure of real-time 224 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 11 theta-delta learning) would be significantly associated with improvements in reading and 225 phonological processing. 226 2. Materials and Methods 227 2.1. Participants 228 Twelve control (typically-developing) adults (mean age of 24.21 ± 6.31 years; 7 female and 229 5 male) and twenty adults with a current or childhood diagnosis of dyslexia (mean age of 230 23.65 ± 5.52 years; 16 female and 4 male) participated in the study. All participants were 231 native English speakers with normal or corrected-to-normal vision and no reported hearing 232 impairments. Typically developing participants were included if their efficiency index (EI) 233 on the Test of Word Reading Efficiency (TOWRE, Torgesen et al., 1999) exceeded 95 (the 234 EI mean is 100, S.D. 15, see Section 2.2 for further detail) . Participants in the dyslexia 235 group were included only if they could provide formal documentation of a dyslexia 236 diagnosis from a qualified professional , such as a Health and Care Professions Council 237 (HCPC) registered assessor, the Accessibility and Disability Resource Centre at University 238 of Cambridge, or a specialist teacher with a current Specific Learning Difficulties (SpLD) 239 Assessment Practicing Certificate. All participants provided informed consent for the study 240 in accordance with the Declaration of Helsinki, and the study was reviewed by the 241 Psychology Research Ethics Committee of the University of Cambridge who gave it a 242 favourable opinion. 243 2.2. Experimental Protocol 244 The experimental protocol spanned ten days and included two assessment sessions, one 245 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 12 at the beginning (Day 1) and one at the end (Day 10), to evaluate participants ’ cognitive 246 and linguistic profiles. These pre - and post -intervention sessions comprised tasks 247 measuring phonological and reading skills, acoustic processing, and skills that were not 248 expected to be improved by the BCI (non-verbal reasoning and arithmetic ability). The eight 249 days in between were dedicated to the BCI training intervention, which will be described in 250 Section 2.3. The experimental protocol can be found in Figure 2.1(a). The measures used 251 in the pre- and post-test sessions are described below. 252 1) Phonology and Reading Measures 253 Phonology and reading skill s were measured using an experimental phoneme deletion 254 task, an experimental syllable stress recognition task, an experimental Rapid Automatized 255 Naming (RAN), and the standardized word and nonword item lists from the TOWRE. 256 257 The phoneme deletion task was adapted from McDougall et al. (1994). This task required 258 participants to listen to a spoken item and delete a target phoneme (e.g., “BICE” without 259 the /b/ becomes “ICE”). The target consonant phoneme appeared in initial, medial, or final 260 positions, and all correct responses formed real English words. The task comprised 18 261 trials (3 practice and 15 experimental items), presented through sound files recorded by a 262 female speaker of standard Southern British English. The same test was administered both 263 before and after the BCI training. 264 265 In the syllable stress discrimination task, participants heard pairs of different four-syllable 266 words and made a same-different judgement regarding whether the pair of words shared 267 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 13 the same stress pattern (e.g., difficulty - voluntary = yes). The items were from Leong et al. 268 (2011), and comprised 80 randomized pairs, some of which were deliberately mis-stressed 269 (e.g., di-FFI-cul-ty – VO-lun-ta-ry = NO). This task was administered before and after the 270 BCI training. This task was a variation to the protocol used in Araújo (2023), in which pairs 271 of identical words were used as stimuli (also drawn from Leong et al., 2011). Here we used 272 comparisons between different words to increase task difficulty, with the aim of reducing 273 the ceiling effect observed with original design. 274 275 In the RAN task, participants named pictures of familiar items (e.g., cup, book, tree) aloud 276 as fast as possible. Four pages of pictures (two pages of target words with low phonological 277 neighborhoods and two with high phonological neighborhoods) were administered at both 278 the pre- and post-sessions. Both the time taken to complete the task and accuracy were 279 recorded. 280 281 Reading was assessed using the TOWRE. The participant received a list of single words 282 to read aloud in 45 seconds, and a list of nonword items to read aloud in 45 seconds. 283 Version A of this task was given at the pre-training stage while version B was used at the 284 post-training stage. The highest available age bracket for calculating scaled scores ranges 285 from 17 years 0 months to 24 years 11 months. Since some participants in the study were 286 older than this range , raw scores were used for statistical analyses. However, for 287 participant recruitment, an EI was calculated using scaled scores from 17 -24 age group, 288 as all participants were over 18 years of age. These scaled scores were used for typically 289 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 14 developing group screening purposes and were not included in further analysis 290 2) Non-verbal I.Q. 291 All participants completed the matrix reasoning subtest of the Wechsler Intelligence Scale 292 for Adults (WAIS; Wechsler, 1955), a widely used measure of non-verbal intelligence. This 293 is a nonverbal reasoning task in which individuals are asked to identify patterns in designs. 294 This pattern recognition task was administered at both pre- and post-intervention sessions. 295 3) Arithmetic Task 296 Participants completed the standardized arithmetic subscale of the Wide Range 297 Achievement Test (WRAT) (Snelbaker et al., 2001), which includes basic math problems 298 requiring written responses. Version TAN was administered before the intervention, and 299 Version BLUE after. This task was included to test whether the BCI would affect any 300 academic skill, rather than specifically affect word reading. 301 4) Acoustic Threshold for Amplitude Rise Time (ART): 1 Rise Task 302 Participants also completed a sine tone rise time task (labelled the 1 Rise task in our prior 303 publications with children, e.g. Flanagan et al., 202 4) to assess sensitivity to ART. Each 304 trial presented three 500 -Hz tones, with one (the target) having a slower onset rise time 305 than the two standard tones. Using an AXB format displayed as cartoon dinosaurs, 306 participants were asked to identify which of the first or third sounds differed from the middle 307 tone. The task used 39 stimuli with rise times ranging from 300 ms to 15 ms in 7.3 ms 308 steps. Verbal instructions and five practice trials with feedback were provided before the 309 main task. Participants performed the same test both before and after the BCI training. 310 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 15 2.3. BCI Training 311 Between the pre- and post-intervention assessment sessions, participants completed eight 312 days of BCI training. The protocol was structured to span ten days in total, allowing at most 313 two rest days (typically the weekend) to accommodate participant schedules. Most 314 participants completed the intervention over two consecutive weeks. Each daily session 315 included two BCI runs. Prior to each session, an EEG cap was fitted, and electrode 316 impedances were checked and maintained below 100 Ω. Participants were encouraged to 317 listen carefully to the words in the story and try to identify listening strategies to keep the 318 spaceship ascending on the screen. However, no explicit suggestions regarding how to 319 achieve this goal were given. 320 321 Each BCI run consisted of two distinct phases: a baseline stage and a BCI control stage. 322 The interface of the BCI and the timeline of the experiment are depicted in Figure 2.1 (a). 323 In the baseline stage (lasting four minutes), participants viewed a vertically moving 324 spaceship displayed on the screen. During this phase, they had no neural control over the 325 spaceship’s position. Instead, the spaceship moved randomly, with positions sampled from 326 a Gaussian distribution (mean = 0.5, SD = 0.15) and mapped onto a vertical scale ranging 327 from 0 (top) to 1 (bottom). The position updated at a refresh rate of 4 Hz. This random 328 movement served two purposes: it provided data to estimate individualized decoder 329 thresholds based on each participant’s typical neural activity, and it avoided any neural 330 entrainment that might occur with fixed or repetitive visual patterns. A semi -transparent 331 white overlay and the message “Good luck! Please wait...” were displayed to indicate the 332 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 16 system was in passive mode. No auditory input was presented during this stage. 333 334 After the baseline, participants entered the BCI stage, which lasted for the duration of a 335 ten-minute auditory story. At this point, the semi -transparent overlay and the baseline 336 message were removed, and participants began listening to a narrated version of Winnie-337 the-Pooh through headphones. Simultaneously, they gained neural control over the on -338 screen spaceship, which moved vertically based on real-time EEG activity. Specifically, the 339 spaceship’s position was determined by the log -transformed theta/delta power ratio 340 measured from centrally located electrodes (F3, F4, C3, Cz, C4, P3, P4) . To personalize 341 control sensitivity, decoder boundaries were set using each participant’s baseline 342 distribution: the median of their log-transformed theta/delta ratio defined the vertical midline 343 of the screen, while the upper and lower boundaries were set at three standard deviations 344 above and below the median. Participants were instructed to raise the spaceship as high 345 as possible and to keep it stable during the story. In terms of neural dynamics, this 346 corresponded to decreasing the theta/delta ratio and reducing its variance. 347 348 The BCI was designed to provide feedback based on neural patterns previously associated 349 with continuous speech processing and phonological awareness in children with and 350 without dyslexia. The same decoder was used across participants, with calibration derived 351 from each session's baseline. This allowed for continuous control based on dyslexia -352 relevant neural dynamics, specifically those shown to relate to phonological awareness in 353 previous studies. 354 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 17 355 The neurofeedback display was designed to be intuitive and engaging. To enhance 356 motivation and user experience, the traditional cursor was replaced with a spaceship 357 graphic, and the background featured a subtle space-themed design. A visual midline was 358 drawn on the screen to indicate the target region for upward control. In addition, a five -359 timestep history trace was implemented, appearing as a contrail behind the spaceship, 360 allowing participants to visually track their recent performance. To enhance participant 361 motivation, a cumulative score related to the real -time theta/delta ratio was presented in 362 the top-left corner of the screen. The score was updated at each refresh of the spaceship 363 position. The instantaneous score was derived from the log transform of the ratio 364 standardized to each participant's baseline, which also determined the spaceship's position. 365 It was multiplied by 10 when the spaceship occupied the lower half of the screen and by 366 20 when it occupied the upper half to reinforce positive feedback. Each instantaneous 367 score was continually added to the total score displayed. The story audio was not 368 influenced by task performance and remained constant throughout the session. 369 370 To further reinforce successful BCI control, a visual reward system was implemented. A 371 semi-transparent green overlay appeared on the screen, with its intensity varying 372 according to the spaceship’s vertical position. The screen was scaled from 0 (top) to 1 373 (bottom), and the green glow was calculated using the formula: 374 𝐺𝑙𝑜𝑤 𝑖𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 = 𝑚𝑎𝑥(𝑚𝑖𝑛(255 − 510𝑥𝑡 ,255),0) 375 Where 𝑥𝑡 is the scaled spaceship position. This meant that the glow reached full intensity 376 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 18 when the spaceship was at the top of the screen, gradually faded when it approached the 377 midline, and disappeared entirely when below the midline. This continuous visual 378 reinforcement served as an intuitive feedback signal to encourage better control of the 379 spaceship. 380 381 To prepare the signal for use in the BCI decoder, the theta/delta ratio was log-transformed. 382 This transformation was necessary because the raw ratio data exhibited a skewed, non -383 Gaussian distribution across and within participants, along with a wide and variable 384 dynamic range. Applying a log transformation reduced skewness and the influence of 385 outliers by compressing the range of values. Crucially, because the log function is 386 monotonically increasing, it preserved the relative ordering of values in the original signal. 387 This ensured that the neurofeedback interface remained stable, interpretable, and 388 sensitive to individual neural dynamics. The BCI neural feedback is shown in Figure 2.1(b). 389 390 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 19 Figure 2.1 Panel (A) shows the whole Experimental Protocol, including the time line of 391 each BCI session and the spaceship interface. Panel (B) depicts the decoder logic behind 392 the neural feedback. 393 2.4. EEG preprocessing 394 The EEG signal was acquired in real time using a CGX wireless headset and continuously 395 processed throughout the neurofeedback task. Raw data were initially converted from a 396 24-bit compressed format to microvolts and streamed at a sampling rate of 500 Hz. EEG 397 preprocessing followed two distinct strategies: real-time processing for neurofeedback 398 delivery, and offline preprocessing for subsequent data analysis. 399 2.4.1 Real-Time Processing 400 During the BCI intervention, real -time processing prioritized low computational demand 401 and effective noise suppression to ensure smooth feedback. A zero -phase, fourth-order 402 Butterworth bandpass filter (0.5–10 Hz) was applied to selected central channels to reduce 403 noise and isolate relevant neural signals. A 3 -second sliding window was used for 404 continuous feature extraction. During pilot testing, we observed that the spaceship position 405 could be influenced by abnormal eye movements. To manage transient artifacts (e.g., eye 406 movement, muscle activity or movement), an online threshold -based artifact rejection 407

was employed. Samples exceeding a channel -specific threshold —determined 408 from the 95th percentile of the participant’s baseline amplitude distribution—were replaced 409 with random clean segments drawn from the individual’s baseline data, preserving inter -410 channel relationships . This procedure ensured that the spaceship position was less 411 affected by abnormal EEG segments during BCI training . The resulting preprocessed 412 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 20 signal was then used for real-time spectral analysis and neurofeedback computation. 413 2.4.2. Offline Preprocessing 414 Despite the use of an online artifact rejection method, the influence of spaceship position 415 could not be completely eliminated during online processing. To address this limitation, a 416 more comprehensive offline preprocessing pipeline was applied to obtain cleaner data. 417 Power line noise was removed using a notch filter, and the data were bandpass -filtered 418 between 0.5 and 48 Hz using an 8th -order Butterworth filter with zero -phase filtering to 419 avoid phase distortion. The signal was then downsampled to 250 Hz to reduce 420 computational load. Given that dry EEG systems tend to produce noisier recordings than 421 gel-based systems, Artifact Subspace Reconstruction (ASR) was used to suppress high -422 amplitude transients such as muscle bursts and cable movements. Channels were marked 423 as noisy if their voltages exceeded ±100 μV or if their power spectra deviated more than 3 424 standard deviations from the mean. The EEG data were then re-referenced to the average 425 of all channels. Independent Component Analysis (ICA) was performed, and components 426 associated with ocular, muscular, or blink artifacts (e.g., EOG, EMG) were identified and 427 removed. The cleaned data were segmented into consecutive, non-overlapping 3-second 428 epochs. Finally, previously identified noisy channels were interpolated using a spline 429 interpolation method. 430 2.5. Statistical Analysis 431 Statistical analyses were conducted to evaluate the effectiveness of the BCI 432 neurofeedback intervention and its relationship with behavioral performance. First, t o 433 assess whether participants exhibited neurophysiological changes across the training 434 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 21 period, we tested whether there was a significant change in participants’ neural responses 435 over time. Specifically, we compared the distribution of the log-transformed theta/delta ratio 436 between the first session and the final session. This was done separately for both (i) the 437 real-time ratio used for intervention feedback (derived from the online preprocessing 438 pipeline), and (ii) the ratio extracted from the offline preprocessed data. T-tests were used 439 to assess differences in the ratio values across sessions for each participant. The resulting 440 t-statistic served as a summary measure of change in BCI performance over time. 441 Participants who showed a significantly lower ratio by the final session were considered to 442 have demonstrated learning. 443 Second, to evaluate whether participants improved on relevant behavioral skills following 444 the BCI intervention, pre- and post -intervention behavioral scores (e.g. phonological 445 awareness, reading ability) were compared using paired -sample t -tests. A significant 446 increase in post-test scores was interpreted as evidence of behavioral improvement. 447 Third, we investigated whether changes in neural measures were associated with changes 448 in behavioral performance. Pearson correlations were conducted to assess the relationship 449 between the t-statistics derived from the BCI measures (both real-time and offline) and the 450 differences in behavioral performance between pre -test and post -test. For all analyses 451 described above, p -values were corrected for multiple comparisons using the false 452 discovery rate (FDR), and statistical significance was defined as corrected p < 0.05. 453 3. Results 454 3.1. Neurophysiological Changes Across Sessions 455 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 22 To evaluate whether participants exhibited neurophysiological changes across the BCI 456 training period, we compared the log-transformed theta/delta power ratios between the first 457 and final sessions. This analysis was conducted twice, separately for data processed using 458 the real-time processing pipeline (following Araújo, 2023) and for data preprocessed offline 459 (analysis added here). 460 For each participant, a t-test was conducted to assess whether the theta/delta ratio 461 significantly decreased from the first to the last session. The resulting t-statistic served as 462 an individual -level summary of neural change and was used as the participant’s BCI 463 learning score (hereafter BCI score) in subsequent analyses. Participants were considered 464 to have learned the BCI control task successfully if their t-statistic was greater than zero 465 and the corresponding p-value was less than 0.05. This criterion indicates a statistically 466 significant reduction in the theta/delta ratio across BCI sessions. 467 Based on the real-time EEG data, 9 out of 12 participants in the control group and 13 out 468 of 20 participants in the dyslexia group met this learning criterion. Following offline 469 preprocessing, which involved the removal of ocular, muscular, and movement -related 470 artifacts (e.g., EMG and EOG signals), the resulting theta/delta ratios exhibited generally 471 lower t-statistics. Under this more stringent preprocessing, 6 of 12 participants in the 472 control group and 10 of 20 in the dyslexia group showed significant improvement according 473 to the same criterion. 474 Figure 3.1 presents a visual summary of the t-statistics for each participant under both 475 processing pipelines. The figure includes two subplots: the left subplot (in blue) represents 476 the control group, and the right subplot (in red) represents the dyslexia group. In each 477 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 23 subplot, individual participant data are shown using paired line plots that connect the t -478 statistics obtained from the real -time and offline pipelines, illustrating the direction and 479 magnitude of change after artifact correction. As the offline preprocessing pipeline is more 480 stringent, it would be expected that the BCI performance scores are lower for the offline 481 preprocessing, which was the case for both groups. Overall, a higher T-score indicates 482 better learning of the BCI. Overlaid on each set of lines, boxplots depict the overall 483 distribution of t-statistics for each preprocessing method within each group. 484 485 Figure 3.1. The t-statistics used as a measure of BCI performance. Data are shown for 486 each participant, under real-time feedback versus offline preprocessing. 487 3.2. Behavioral Improvements Following BCI Intervention 488 To investigate whether participants showed behavioral improvement over the course of 489 training, we examined performance across the acoustic, cognitive and linguistic tasks. As 490 will be recalled, two measures were not expected a priori to show improvement following 491 BCI training, nonverbal IQ (WAIS Matrices) and Arithmetic. 492 3.2.1. Pre-Intervention Scores 493 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 24 As a first step, we compared pre -intervention scores between the control and dyslexia 494 groups to assess baseline differences in task performance (Table 3.1). As expected, there 495 were no significant group differences in non-verbal cognitive tasks such as Arithmetic and 496 Matrices Reasoning, indicating that both groups were matched on non-reading academic 497 performance and general reasoning ability. In contrast, significant group differences were 498 observed in pre-test measures of reading and phonological processing. These included 499 syllable stress discrimination accuracy, RAN completion time, and both word and non-word 500 raw scores on the TOWRE. These results are consistent with the known language and 501 literacy difficulties associated with dyslexia. An exception was the phoneme deletion task, 502 which showed no significant group difference . This was likely due to a ceiling effect. The 503 task may have been too easy for both groups, limiting its sensitivity to detect individual 504 differences. Contrary to our prior adult studies, no significant group difference in sensitivity 505 to ART was observed, although the dyslexic group showed worse performance. 506 507 Table 3.1. Group performance on pre-test measures, 2-tailed t-tests 508 Behavioral Test DYS Mean (S.D.) CTRL Mean (S.D.) t-score p-value WRAT Arithmetic (scaled score) 101.2 (14.0) 106.1 (15.7) 0.905 0.266 WAIS Matrices (T-Score) 56.6 (5.9) 57.3 (7.6) 0.307 0.423 Phoneme Deletion (n correct) 13.4 (2.2) 13.8 (1.7) 0.583 0.353 Syllable Stress Recognition (ACC) 62.6 (10.7) 76.6 (16.9) 2.868 0.009 Syllable Stress Recognition (RT, s) 3.4 (1.4) 2.8 (0.9) -1.457 0.145 RAN (mean time, s) 32.0 (4.9) 27.4 (3.2) -2.880 0.009 TOWRE SWE (raw score) 81.5 (6.8) 99.3 (10.4) 5.885 <0.001 TOWRE PDE (raw score) 45.8 (7.3) 59.9 (7.1) 5.348 <0.001 1 Rise Task (time threshold, ms) 136.5 (93.2) 92.2 (75.4) -1.391 0.145 509 3.2.2. Pre-Intervention versus Post-BCI Behavioural Scores 510 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 25 We next compared pre - and post -intervention scores within each group to examine 511 behavioral changes over time (Table 3.2). In the control group, significant improvements 512 were observed in Arithmetic performance, syllable stress accuracy, RAN completion time, 513 and TOWRE non-word reading. In the dyslexia group, significant improvements were found 514 in Matrices Reasoning, phoneme deletion accuracy, syllable stress discrimination accuracy, 515 RAN completion time, and both TOWRE word and non-word reading. With the exception 516 of the improvement in Arithmetic (controls only) and Matrices Reasoning (dyslexics only), 517 these improvements were in line with our a priori expectations. However, both groups in 518 these analyses included participants who did not meet criterion for learning the BCI. 519 520 Table 3.2. Pre- versus post-intervention scores for the behavioral tasks, 2-tailed t-tests 521 Behavioral Test Group Pre-test mean (S.D.) Post-test mean (S.D.) t-score p-value WRAT Arithmetic (scaled score) CTRL 106.1 (15.7) 110.4 (17.0) -2.738 0.051 DYS 101.2 (14.0) 102.8 (14.6) -1.281 0.108 WAIS Matrices (T-Score) CTRL 57.3 (7.6) 59.2 (7.1) -1.675 0.164 DYS 56.6 (5.9) 59.5 (4.9) -4.681 <0.001 Phoneme Deletion (n correct) CTRL 13.8 (1.7) 14.0 (1.3) -0.561 0.521 DYS 13.4 (2.2) 13.9 (2.1) -2.127 0.031 Syllable Stress Recognition (ACC) CTRL 76.6 (16.9) 80.7 (18.2) -2.865 0.051 DYS 62.6 (10.7) 67.4 (12.6) -2.623 0.013 Syllable Stress Recognition (RT, s) CTRL 2.8 (0.9) 2.4 (0.9) 1.670 0.164 DYS 3.4 (1.4) 3.4 (1.9) 0.062 0.423 RAN (mean time, s) CTRL 27.4 (3.2) 25.3 (3.2) 4.891 0.004 DYS 32.0 (4.9) 28.0 (4.3) 6.631 <0.001 TOWRE SWE (raw score) CTRL 99.3 (10.4) 98.6 (8.9) 0.581 0.521 DYS 81.5 (6.8) 87.2 (7.3) -4.467 <0.001 TOWRE PDE (raw score) CTRL 59.9 (7.1) 61.7 (5.0) -2.399 0.071 DYS 45.8 (7.3) 48.4 (7.4) -2.931 0.009 1 Rise Task (time threshold, ms) CTRL 92.2 (75.4) 75.2 (57.2) 1.073 0.350 DYS 136.5 (93.2) 119.3 (93.9) 1.440 0.095 522 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 26 3.2.3 Pre-Intervention versus Post-BCI Scores, BCI Learners only 523 We next assessed the same relationships for the BCI learners only (Table 3.3). In the 524 control group, significant improvements were observed in RAN completion time only, for 525 both real-time and offline preprocessing. In the dyslexia group, significant improvements 526 were found in Matrices reasoning, RAN completion time , syllable stress discrimination 527 accuracy and TOWRE non-word reading, for both real-time and offline preprocessing. With 528 the exception of the improvement in Matrices reasoning (dyslexics only), these 529 improvements were in line with our a priori expectations based on TS theory. 530 531 Table 3.3. Pre- and post-intervention behavioral improvement in BCI learners only by group, 532 2-tailed t-tests 533 Preprocessing Behavioral Test Group Pre-test mean (S.D.) Post-test mean (S.D.) t-score p-value realtime (CTRL: n=9, DYS: n=13) WRAT Arithmetic (scaled score) CTRL 108.1 (15.7) 112.0 (15.8) -2.135 0.110 DYS 101.3 (11.4) 101.9 (13.5) -0.409 0.431 WAIS Matrices (T-Score) CTRL 56.7 (7.5) 59.1 (8.0) -2.137 0.110 DYS 56.5 (4.1) 59.9 (3.1) -4.137 0.002 Phoneme Deletion (n correct) CTRL 13.6 (1.9) 13.9 (1.5) -0.894 0.397 DYS 14.1 (1.4) 14.5 (0.7) -1.585 0.116 Syllable Stress Recognition (ACC) CTRL 77.1 (15.6) 80.8 (18.4) -2.025 0.110 DYS 64.7 (12.1) 70.4 (12.9) -2.312 0.039 Syllable Stress Recognition (RT, s) CTRL 2.7 (0.9) 2.2 (0.6) 1.986 0.110 DYS 3.3 (1.3) 3.3 (1.5) -0.121 0.503 RAN (mean time, s) CTRL 28.0 (3.3) 25.7 (3.6) 6.984 0.001 DYS 33.1 (5.3) 28.4 (4.7) 6.047 <0.001 TOWRE SWE (raw score) CTRL 97.3 (11.4) 97.3 (9.8) 0.000 0.889 DYS 80.5 (7.2) 87.2 (8.6) -4.624 0.001 TOWRE PDE (raw score) CTRL 59.0 (7.7) 61.2 (5.3) -2.443 0.110 DYS 46.4 (7.5) 50.0 (6.3) -3.065 0.012 1 Rise Task (time threshold, ms) CTRL 75.8 (57.2) 55.7 (27.0) 1.072 0.360 DYS 104.7 (77.0) 93.1 (72.7) 0.768 0.327 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 27 offline (CTRL: n=6, DYS: n=10) WRAT Arithmetic (scaled score) CTRL 110.5 (17.7) 116.3 (14.7) -2.956 0.127 DYS 101.1 (15.1) 104.2 (15.2) -1.835 0.083 WAIS Matrices (T-Score) CTRL 57.8 (8.1) 59.5 (9.7) -1.185 0.379 DYS 56.8 (3.6) 59.6 (3.8) -5.250 0.002 Phoneme Deletion (n correct) CTRL 12.8 (1.9) 13.3 (1.5) -0.889 0.415 DYS 13.7 (1.6) 14.0 (1.2) -0.896 0.246 Syllable Stress Recognition (ACC) CTRL 73.8 (18.4) 76.7 (21.5) -1.075 0.379 DYS 61.5 (9.5) 68.6 (12.0) -2.468 0.036 Syllable Stress Recognition (RT, s) CTRL 3.0 (0.9) 2.5 (0.4) 1.343 0.379 DYS 3.2 (1.1) 3.2 (1.4) 0.036 0.540 RAN (mean time, s) CTRL 28.1 (4.0) 25.8 (4.1) 4.834 0.038 DYS 33.1 (5.9) 29.5 (5.1) 4.936 0.002 TOWRE SWE (raw score) CTRL 98.2 (10.8) 96.2 (10.3) 1.369 0.379 DYS 80.1 (6.0) 83.5 (6.5) -3.511 0.011 TOWRE PDE (raw score) CTRL 56.8 (8.8) 59.8 (6.1) -2.423 0.160 DYS 44.6 (5.2) 48.5 (5.0) -2.830 0.025 1 Rise Task (time threshold, ms) CTRL 80.1 (70.3) 58.0 (31.7) 0.769 0.424 DYS 131.7 (78.5) 107.7 (71.1) 1.055 0.228 534 3.3. Associations Between Neural and Behavioral Changes 535 To explore whether the observed improvements in performance were systematically 536 related to learning the BCI, we calculated Pearson correlations between participants’ BCI 537 learning scores and their changes in the behavioral tasks (computed as post-intervention 538 behavioural score minus pre-intervention score in each case ). The results are shown in 539 Table 3.4. BCI learning scores were quantified using the t-statistics from session -wise 540 comparisons of the theta/delta ratio. T-scores were considered separately for the real-time 541 and offline EEG pipelines. Higher t-statistics indicated greater success in reducing 542 theta/delta ratios across training, and hence should be positively related to improvements 543 in accuracy in the behavioural tasks, and negatively related to improvements in processing 544 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 28 time. Given that TS theory would predict improvements in the phonological and reading 545 tasks, we tested these correlations using one-tailed tests. 546 3.3.1. Real-Time BCI Improvement; Correlations by Group 547 In the real -time pipeline, no significant correlations were found between BCI learning 548 scores and improvements in Arithmetic or Matrices Reasoning tasks (Table 3.4). This 549 aligns with expectations, as these tasks reflect non-verbal cognitive abilities that were not 550 targeted by the BCI. In contrast, significant associations were observed between neural 551 improvement and phonological recoding for both groups . Specifically, participants with 552 higher BCI learning scores showed greater gains in TOWRE non -word reading, r= .61, 553 p< .05 (DYS) and r= .72, p< .05 (CTRL). In the control group, better BCI performance was 554 also associated with a greater decrease in response time on the syllable stress 555 discrimination task (r= .79, p< .01), suggesting faster phonological processing. 556 3.3.2. Offline Preprocessing BCI Improvement; Correlations by Group 557 In the offline preprocessing pipeline, only the participants with dyslexia showed significant 558 changes in phonology and reading. As with the real-time data, no significant relationships 559 were detected between BCI scores , Arithmetic and Matrices reasoning tasks. In the 560 dyslexic group, positive correlations were found between BCI learning scores and 561 improvements in syllable stress discrimination accuracy (r= .51, p< .05), and TOWRE non-562 word reading (r= .56, p< .05). Additionally, a significant negative correlation was observed 563 between BCI performance and rise time threshold (r= -.50, p< .05) , indicating that 564 participants with greater neural adaptation were more sensitive to amplitude rise time 565 following training. 566 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 29 567 In each case, Figure 3.2 (real-time processing) and Figure 3.3 (offline preprocessing) 568 display the relevant scatter plots and regression lines for the behavioural tasks with 569 significant correlations. All correlation coefficients for each behavio ural task, across both 570 preprocessing strategies and participant groups, are provided in Table 3.4. 571 572 Figure 3.2. The correlation between BCI intervention training improvement and behavioral 573 task improvement (real-time preprocessing). 574 575 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 30 Figure 3.3. The correlation between BCI intervention training improvement and behavioral 576 task improvement (offline preprocessing). 577 Table 3.4 Pearson correlations between BCI score and behavioral tests by group, 1-tailed 578 tests 579 BCI score (t-value pre/post) behavioral test (Post - Pre) Control Dyslexia r p-value r p-value realtime (CTRL: n=12, DYS: n=20) WRAT Arithmetic (scaled score) 0.133 0.628 -0.208 0.432 WAIS Matrices (T-Score) 0.348 0.375 0.300 0.365 Phoneme Deletion (n correct) 0.247 0.375 -0.134 0.634 Syllable Stress Recognition (ACC) -0.431 0.715 0.339 0.286 Syllable Stress Recognition (RT, s) -0.786 0.008 -0.013 0.478 RAN (mean time, s) 0.038 0.628 -0.224 0.365 TOWRE SWE (raw score) -0.185 0.628 0.143 0.365 TOWRE PDE (raw score) 0.722 0.014 0.611 0.017 1 Rise Task (time threshold, ms) -0.231 0.375 -0.150 0.365 offline (CTRL: n=12, DYS: n=20) WRAT Arithmetic (scaled score) 0.176 0.681 -0.037 0.584 WAIS Matrices (T-Score) 0.278 0.681 -0.124 0.568 Phoneme Deletion (n correct) 0.180 0.681 -0.191 0.584 Syllable Stress Recognition (ACC) -0.411 0.908 0.513 0.024 Syllable Stress Recognition (RT, s) -0.436 0.440 -0.094 0.443 RAN (mean time, s) 0.078 0.681 -0.079 0.443 TOWRE SWE (raw score) -0.086 0.681 -0.100 0.568 TOWRE PDE (raw score) 0.402 0.440 0.560 0.024 1 Rise Task (time threshold, ms) -0.101 0.681 -0.502 0.024 580 3.3.3. All BCI Learners: Correlations 581 Finally, given that many control participants also demonstrated learning of the BCI, we 582 considered the BCI learners only as a single pooled group. We computed Pearson 583 correlations between participants’ BCI learning scores and their changes in the 584 behavioral tasks, adding the groups to achieve reasonable power (N = 22 for real time 585 data, 13 dyslexics and 9 controls; N = 16 for offline data, 6 dyslexics and 6 controls). The 586 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 31

are shown in Table 3.5. In both the real time data and the offline data, BCI 587 learners showed significant improvements in nonword reading (r= .70, p< .01; r= .66, 588 p< .01, respectively). For the real time data (N = 22), BCI learners also showed 589 significant improvement in amplitude rise time discrimination (r= -.53, p< .05) and in the 590 speed of making syllable stress pattern judgements (r= -.47, p< .05). 591 592 Table 3.5 Pearson correlations between BCI score and behavioral test in all BCI learners, 593 1-tailed tests 594 BCI score (t-value pre/post) behavioral test (Post - Pre) r p-value realtime (n=22) WRAT Arithmetic (scaled score) 0.138 0.463 WAIS Matrices (T-Score) 0.128 0.463 Phoneme Deletion (n correct) -0.067 0.463 Syllable Stress Recognition (ACC) 0.167 0.343 Syllable Stress Recognition (RT, s) -0.467 0.029 RAN (mean time, s) 0.067 0.463 TOWRE SWE (raw score) -0.239 0.572 TOWRE PDE (raw score) 0.696 0.001 1 Rise Task (time threshold, ms) -0.527 0.018 offline (n=16) WRAT Arithmetic (scaled score) 0.006 0.869 WAIS Matrices (T-Score) -0.112 0.790 Phoneme Deletion (n correct) -0.159 0.790 Syllable Stress Recognition (ACC) 0.242 0.371 Syllable Stress Recognition (RT, s) -0.039 0.691 RAN (mean time, s) -0.073 0.691 TOWRE SWE (raw score) -0.181 0.790 TOWRE PDE (raw score) 0.661 0.003 1 Rise Task (time threshold, ms) -0.271 0.371 595 4. Discussion. 596 The current report suggests that a BCI aimed at normalising the low-frequency 597 oscillatory neural patterns associated with continuous speech processing in 598 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 32 developmental dyslexia can improve linguistic phonological processing of syllable stress 599 patterns and phonological recoding of print to sound (nonword reading) for adults both 600 with and without dyslexia. The BCI developed here also improved ART sensitivity for all 601 BCI learners. ART is an important acoustic cue used for automatic oscillatory phase-602 resetting during speech-brain alignment (Doelling et al., 2014). These improvements are 603 in line with TS theory. 604 TS theory is based on atypical encoding of the low-frequency envelope information 605 thought to govern prosodic perception in dyslexia (Goswami, 2011). Prior neuroimaging 606 studies have shown that children with dyslexia learning English, Spanish and French 607 show impaired neural encoding of low-frequency speech envelope information <10 Hz 608 during natural speech listening (DiLiberto et al., 2018; Molinaro et al., 2016; Destoky et 609 al., 2020), and that English-speaking children with dyslexia show a higher theta-delta 610 ratio during natural speech listening, which is significantly related to their performance in 611 phonological awareness tasks (a higher ratio is associated with worse performance, 612 Araújo et al., 2024). Accordingly, the theta-delta ratio was targeted by the current BCI. 613 Improvements in syllable stress processing following BCI training were expected on the 614 basis of related TS-driven speech modelling work, which indicated that sensory 615 discrimination of the phase relations between AMs at the delta (0.5 – 4 Hz) and theta (4 – 616 8 Hz) rates govern whether a strong or a weak syllable is perceived (Leong et al., 2014; 617 Leong & Goswami, 2015). Here, significant correlations between participants’ BCI scores 618 and syllable stress processing were demonstrated for both control adults (for real-time 619 processing and response time) and for adults with dyslexia (for offline processing and 620 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 33 response accuracy). Significantly faster syllable stress processing was also exhibited by 621 the pooled group of all BCI learners (Table 3.5, real time data). No associations were 622 found for phoneme-level linguistic tasks in any analyses, however this could reflect 623 ceiling effects on the phoneme deletion task that was selected for this study. 624 For both adults with dyslexia (Table 3.4) and all pooled BCI learners (Table 3.5), 625 there was also a significant correlation between BCI scores and enhanced ART 626 discrimination. BCI learners showed better discrimination of ART following BCI training. 627 This could be promising therapeutically for children, as by TS theory it is impaired ART 628 discrimination which affects neural speech encoding via oscillatory speech-brain 629 alignment. Both impaired ART discrimination and associated impaired neural speech 630 encoding of low-frequency speech information compromise the efficient development of a 631 phonological lexicon. Indeed, experimental work with dyslexic adults has demonstrated 632 such a relationship regarding impaired ART discrimination and impaired speech encoding 633 (Lizarazu et al., 2021), while a series of studies across languages (summarized in 634 Goswami, 2015) demonstrate that impaired ART discrimination is significantly related to 635 impairments in phonological awareness at many linguistic levels. Accordingly, if learning 636 the BCI leads to enhanced ART discrimination, this should have positive effects on 637 developmental trajectories for phonological development. 638 Most promising of all regarding the compromised reading skills that ensue from the 639 phonological processing difficulties that characterize developmental dyslexia, control 640 adults, dyslexic adults and all BCI learners showed enhanced nonword reading after 641 learning the BCI. For dyslexic participants, both real-time BCI scores and offline BCI 642 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 34 scores showed significant correlations with nonword reading (Table 3.4), while real-time 643 BCI scores showed a significant correlation with nonword reading for control adults 644 (Table 3.4). When all BCI learners were considered as a pooled group, both real-time and 645 offline BCI scores were significantly correlated with improvement in nonword reading 646 (Table 3.5). As impaired nonword reading is a hallmark of childhood dyslexia across 647 languages, further development of the current BCI for children may thus offer significant 648 therapeutic benefits. 649 To our knowledge, this is the first BCI for dyslexia that targets the pre-reading 650 ‘phonological deficit’ (see Christodoulides et al., 2022, for an EEG classifier study 651 intended to inform a dyslexia BCI based on magnocellular theory, Ortiz et al., 2020, for 652 EEG classifiers for dyslexia based on AM-noise; Arias et al., 2021, for an in-principle BCI 653 to enhance neural entrainment; and Günet, 2020, for a dyslexia BCI based on 654 multisensory training). The BCI developed here was informed by the TS theory of 655 dyslexia, an auditory theory that proposes that the auditory organization of speech 656 information by a child (assigning acoustic elements of speech perception to the 657 groupings comprising words in a particular language) is impaired at the prosodic level, 658 leading to developmental differences in the accuracy of phonological representations at 659 the level of syllable stress patterning. As the prosodic or rhythmic level is the foundational 660 perceptual (AM) level regarding the rest of the linguistic hierarchy (syllables, onset-rimes 661 and phonemes, see Leong & Goswami, 2015), these inaccurate prosodic representations 662 affect all levels of phonological representation for affected children, making learning to 663 read difficult and effortful in every language (see Goswami, 2022a, for a detailed 664 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 35 explanation). Accordingly, if the current BCI is able to improve syllable stress processing, 665 children’s access to all levels of phonology in the linguistic hierarchy should improve. 666 It is important to note that neural data suggest that the phonological representations 667 developed by individuals with dyslexia are not noisy, as previously believed, rather they 668 are subtly different in organization from those developed by non-dyslexic individuals 669 (Keshavarzi et al., 2022, children; Tan et al., 2022, adults). According to TS theory, the 670 main difference regarding phonological representations lies in encoding accurately the 671 low-frequency amplitude envelope information (see Keshavarzi et al., 2023, for 672 experimental evidence that amplitude envelopes for multi-syllabic words are also 673 produced inaccurately by children with dyslexia). This difference in phonological 674 representations for words means that when print is encountered and visual codes for 675 representing spoken language are acquired (culturally-specific codes that are taught and 676 learned using symbol-sound correspondences), the dyslexic child is at a disadvantage 677 from the outset. If the current BCI can be applied with children prior to learning to read, 678 this disadvantage could potentially be eliminated before school entry. Indeed, brain 679 imaging studies across languages show that visual symbol learning, whether of the 680 alphabet or of characters such as Kanji, is linked to sound from the very beginning of 681 acquiring reading (Blau et al., 2010; Froyen et al., 2009; Maurer et al., 2005, 2011; Yang 682 et al. 2020). Accordingly, by targeting neural features of the dyslexic brain’s response to 683 acoustic linguistic input (natural speech) before reading instruction commences, the 684 current BCI may be able to facilitate visual symbol learning in any language. 685 The current study has a number of limitations. Firstly, the training sessions were given 686 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 36 over a relatively short period of time, and some participants did not learn the BCI (real-time 687 processing, 7/20 dyslexics, 3/12 controls; offline preprocessing, 10/20 dyslexics, 6/12 688 controls). One explanation could be insufficient gaming experience, a ccordingly a longer 689 gaming period may be beneficial in studies which involve child participants. Secondly, the 690 sample size was relatively small. However, it is comparable to prior studies attempting to 691 create BCIs for dyslexia (Günet, 2020; Christodoulides et al., 2022). Thirdly, EEG data is 692 prone to exhibiting highly variable day-to-day variations. To mitigate this problem, baseline 693 data was collected before each BCI run and these recorded EEG patterns were used to 694 define the upper and lower limits of the spaceship on the screen on each run. Fourth, a 695 single story (Winnie the Pooh) was used throughout the whole training protocol. While this 696 was helpful in allowing direct comparison of performance across sessions, it also made the 697 protocol quite tedious, which may have led to de -motivation – an especially pertinent 698

if the participants were to be children. Accordingly, it would be best if future work 699 could devise an operant learning protocol that could handle any story input in any language. 700 Finally, while the provided instructions were quite clear regarding the gaming objective (i.e. 701 making the spaceship go upwards as consistently as possible on the screen while listening 702 to the words in the story carefully), the instructions were also kept purposefully vague so 703 that participants could decide by themselves which strategy to employ. Some participants 704 spontaneously made remarks indicating their chosen strategies, for example “letting your 705 brain flow up and down with the syllables in speech in a new way”. To optimize children’s 706 BCI learning, it may be useful to give them explicit suggestions about utilizing strategies of 707 this nature. 708 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 37 In conclusion, the exploratory data presented here suggest that a simple and engaging 709 BCI for improving phonological processing during natural speech listening can be created 710 using EEG data informed by the TS theory of developmental dyslexia. Participants who 711 learned the BCI showed improved processing of syllable stress patterns in words, 712 improved phonological recoding skills (nonword reading), and improved ART discrimination. 713 These improvements occurred even though no direct training of phonology, nonword 714 reading nor ART discrimination occurred during the study. This is particularly interesting 715 theoretically, as it suggests that the therapeutic benefits resulted from improving the neural 716 theta-delta ratio during natural speech listening. Therapeutic interventions which filter 717 speech to enhance ARTs have also been shown to improve speech processing in 718 participants with dyslexia via changing the theta-delta ratio (Mandke et al., 2023; see also 719 Van Herck et al., 2022, for a related envelope -enhanced method that did not explore the 720 theta-delta ratio). Accordingly, further investigation of neural speech processing in dyslexia 721 guided by TS theory may identify other, possibly more effective, neural targets for BCI 722 development. 723 724 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 38

725 The authors would like to thank all the participants who volunteered for the study. This 726 research was funded by a donation to U.G. from the Yidan Prize Foundation. The sponsor 727 played no role in the study design, data interpretation, nor writing of the report. 728 729 DATA AND CODE AVAILABILITY 730 Data and code will be made available on request. 731 732 DECLARATION OF COMPETING INTEREST 733 The authors declare no conflicts of interest. 734 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 39 Arias, F.J.C., Molinaro, N. & Lizarazu, M. (2021). Real time EEG neurofeedback as a tool 735 to improve neural entrainment to speech. 736 https://www.biorxiv.org/content/10.1101/2021.04.19.440176v1.full.pdf 737 Araújo, J. (2023). Computational framework enabling an EEG-based BCI for 738 neurofeedback in language disorders: The case of dyslexia. Ph.D. dissertation, University 739 of Cambridge. 740 Araújo, J., Simons, B.D. & Goswami, U. (2023). Remediating phonological deficits in 741 dyslexia with brain-computer interfaces. In (Ed.) C. Guger, Brain-Computer Interface 742 Research, A State of the Art Summary 12. New York: Springer. 743 Araújo, J., Simons, B.D., Peter, V., Mandke, K., Kalashnikova, M., Macfarlane, A., 744 Gabrielczyk, F., Wilson, A.M., Di Liberto, G.M., Burnham, D., & Goswami, U. (2024). 745 Atypical low-frequency cortical encoding of speech identifies children with developmental 746 dyslexia. Frontiers in Human Neuroscience, 18, 1403667. 747 Attaheri, A., Ní Choisdealbha, A., Di Liberto, G., Rocha, S., Brusini, P ., Mead, 748 N., Olawole-Scott, H., Boutris, P ., Gibbon, S., Williams, I., Grey, C., Flanagan, S., M., 749 Goswami, U. (2022). Delta- and theta-band cortical tracking and phase-amplitude 750 coupling to sung speech by infants. Neuroimage, 247, 751 118698. https://doi.org/10.1016/j.neuroimage.2021.118698 752 Attaheri, A., Ní Choisdealbha, A., Rocha, S., Brusini, P ., Di Liberto, G., Mead, N., 753 Olawole-Scott, H., Boutris, P ., Gibbon, S., Williams, I., Grey, C., Alfaro e Oliveira, M., 754 Brough, C., Flanagan, S., and Goswami, U. (2024). Infant low-frequency EEG, cortical 755 power, cortical tracking and phase-amplitude coupling predicts language a year 756 later. PLoS ONE 19 (12): e0313274. https://doi.org/10.1371/journal.pone.0313274 757 Blau, V., Reithler, J., van Atteveldt, N., Seitz, J., Gerretsen, P ., Goebel, R., Blomert, L. 758 (2010). Deviant processing of letters and speech sounds as proximate cause of reading 759 failure: a functional magnetic resonance imaging study of dyslexic children. Brain, 133(3), 760 868–879. DOI: 10.1093/brain/awp308 761 Christodoulides, P ., Miltiadous, A., Tzimourta, K. D., Peschos, D., Ntritsos, G., 762 Zakopoulou, V., Giannakeas, N., Astrakas, L. G., Tsipouras, M. G., Tsamis, K. I., Glavas, 763 E., & Tzallas, A. T. (2022). Classification of EEG signals from young adults with dyslexia 764 combining a Brain Computer Interface device and an Interactive Linguistic Software Tool. 765 Biomedical Signal Processing & Control, 76, 1036-46. 766 Destoky F, Bertels J, Niesen M, Wens V, Vander Ghinst M, Leybaert J, et al. (2020) 767 Cortical tracking of speech in noise accounts for reading strategies in children. PLoS Biol 768 18(8): e3000840. https://doi.org/10.1371/journal.pbio.3000840 769 Di Liberto, G. M., Peter, V., Kalashnikova, M., Goswami, U., Burnham, D., & Lalor, E. C., 770 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 40 2018. Atypical cortical entrainment to speech in the right hemisphere underpins 771 phonemic deficits in dyslexia. NeuroImage, 175, 70-79. 772 https://doi.org/10.1016/j.neuroimage.2018.03.072 773 DiLiberto, G.M., Attaheri, A., Cantisani, G., Reilly, R.B., Ní Choisdealbha, A., Rocha, S., 774 Brusini, P ., & Goswami, U. (2023). Emergence of the cortical encoding of phonetic 775 features in the first year of life. Nature Communications, 14, 7789. 776 https://doi.org/10.1038/s41467-023-43490-x 777 Doelling, K. B., Arnal, L. H., Ghitza, O., Poeppel, D. (2014). Acoustic landmarks drive 778 delta-theta oscillations to enable speech comprehension by facilitating perceptual 779 parsing. Neuroimage, 85, 761-768. DOI: 10.1016/j.neuroimage.2013.06.035 780 Facoetti, A., Trussardi, A. N., Ruffino, M., Lorusso, M. L., Cattaneo, C., Galli, R., Molteni, 781 M., & Zorzi, M. (2010). Multisensory spatial attention deficits are predictive of 782 phonological decoding skills in developmental dyslexia. Journal of Cognitive 783 Neuroscience, 22, 1011-1025. https://doi.org/10.1162/jocn.2009.21232. 784 Flanagan, S., Wilson, A. M., Gabrielczyk, F. C., Macfarlane, A., Mandke, K., Goswami, U. 785 (2024). Amplitude rise time sensitivity in children with and without dyslexia: differential task 786 effects and longitudinal relations to phonology and literacy. Frontiers in Psychology, 787 15. https://doi.org/10.3389/fpsyg.2024.1245589 788 Froyen, D., Bonte, M.L., van Atteveldt, N., Blomert, L., (2009). The long road to 789 automation: neurocognitive development of letter-speech sound processing. J. Cogn. 790 Neurosci. 21, 567–580. https://doi.org/10.1162/jocn.2009.21061 791 Giraud, A.L., and Poeppel, D. (2012). Cortical oscillations and speech processing: 792 emerging computational principles and operations. Nature Neuroscience, 15, 511-517. 793 doi:10.1038/nn.3063 794 Goswami, U. (2011). A temporal sampling framework for developmental dyslexia. Trends 795 in Cognitive Sciences, 15, 3-10. https://doi.org/10.1016/j.tics.2010.10.001 796 Goswami, U. (2015). Sensory theories of developmental dyslexia: Three challenges for 797 research. Nature Reviews Neuroscience, 16, 43-54. http://dx.doi.org/10.1038/nrn3836 798 Goswami, U. (2022a). Theories of Dyslexia. In M. Skeide (Ed.), The Cambridge 799 Handbook of Dyslexia and Dyscalculia (Cambridge Handbooks in Psychology, pp. 5-24). 800 Cambridge: Cambridge University Press. doi:10.1017/9781108973595.002 801 Goswami, U. (2022b). Language acquisition and speech rhythm patterns: an auditory 802 neuroscience perspective. R. Soc. Open Sci. 9: 803 211855. https://doi.org/10.1098/rsos.211855 804 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 41 Gross J., Hoogenboom N., Thut G., Schyns P., Panzeri S., Belin P., & Garrod, S. (2013). 805 Speech rhythms and multiplexed oscillatory sensory coding in the human brain. PLoS 806 Biology, 11(12), e1001752. https://doi.org/10.1371/journal.pbio.1001752. 807 Günet, E. (2020). Improving reading abilities in dyslexia with neurofeedback and 808 multisensory learning. Unpublished PhD thesis, Sabanci University, Turkey. 809 https://research.sabanciuniv.edu/id/eprint/41210/ 810 Kalashnikova, M., Goswami, U., & Burnham, D. (2020). Novel word learning deficits in 811 infants at family risk for dyslexia. Dyslexia, 26(1), 3-17. https://doi.org/10.1002/dys.1649 812 Keshavarzi, M., Mandke, K., Macfarlane, A., Parvez, L., Gabrielczyk, F.C., Wilson, A.M., 813 Flanagan, S. & Goswami, U. (2022). Decoding of speech information using EEG in children 814 with dyslexia: Less accurate low -frequency representations of speech, not ‘noisy’ 815 representations. Brain & Language, 235, 105198. 816 817 Keshavarzi, M., Di Liberto, G., Gabrielczyk, F.C., Wilson, A.M., Macfarlane, A. & Goswami, 818 U. (2023). Atypical speech production of multi-syllabic words and phrases by children with 819 developmental dyslexia. Developmental Science, e13428. 820 Kuhl, P . K. (2004). Early language acquisition: Cracking the speech code. Nature 821 Reviews Neuroscience, 5, 831-843. https://doi.org/10.1038/nrn1533 822 Landerl, K., Wimmer, H. (2000). Deficits in phoneme segmentation are not the core 823 problem of dyslexia: Evidence from German and English children. Applied 824 Psycholinguistics, 21, 2 , 243 – 262. DOI: https://doi.org/10.1017/S0142716400002058 825 Leong, V. & Goswami, U. (2015). Acoustic-emergent phonology in the amplitude 826 envelope of child-directed speech. PLoS One, 10(12), e0144411. 827 https://doi.org/10.1371/journal .pone.0144411 828 Leong, V., Hämäläinen, J., Soltész, F., & Goswami, U. (2011). Rise Time Perception and 829 Detection of Syllable Stress in Adults with Developmental Dyslexia. Journal of Memory 830 and Language, 64, 59-73. https://doi.org/10.1016/j.jml.2010.09.003 831 Leong, V., Stone, M., Turner, R.E., & Goswami, U. (2014). A role for amplitude 832 modulation phase relationships in speech rhythm perception. Journal of the Acoustical 833 Society of America, 136, 366-381. https://doi.org/10.1121/1.4883366 834 Leong, V., Kalashnikova, M., Burnham, D. & Goswami, U. (2017). The temporal 835 modulation structure of infant-directed speech. Open Mind, 1(2), 78-90. 836 https://doi.org/10.1162/OPMI_a_00008 837 Lizarazu, M., Lallier, M., Bourguignon, M., Carreiras, M., Molinaro, N. (2021). Impaired 838 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 42 neural response to speech edges in dyslexia. Cortex, 135, 207-218. 839 https://doi.org/10.1016/j.cortex.2020.09.033 840 Mandke, K.M., Flanagan, S.A., Macfarlane, A., Feltham, G., Gabrielczyk, Wilson, A.M., 841 Gross, J., & Goswami, U. (2023). Neural responses to natural and enhanced speech edges 842 in children with and without dyslexia. Frontiers in Human Neuroscience, 17, 1200950. 843 Maurer U., Brem S., Bucher K., & Brandeis D. (2005). Emerging neurophysiological 844 specialization for letter strings. Journal of Cognitive Neuroscience, 17(10), 1532-1552. 845 DOI: 10.1162/089892905774597218 846 Maurer, U., Schulz, E., Brem, S., der Mark, S. van, Bucher, K., Martin, E., Brandeis, D. 847 (2011). The development of print tuning in children with dyslexia: Evidence from 848 longitudinal ERP data supported by fMRI. Neuroimage 57, 714–722. 849 https://doi.org/10.1016/j.neuroimage.2010.10.05 850 Molinaro, N., Lizarazu, M., Lallier, M., Bourguignon, M., & Carreiras, M. (2016). Out-of-851 synchrony speech entrainment in developmental dyslexia. Human Brain Mapping, 37, 852 2767–2783. https://doi.org/10.1002/hbm.23206 853 Ní Choisdealbha, A., Attaheri, A., Rocha, S., Mead, N., Olawole -Scott, H., Brusini, P ., 854 Gibbon, S., Boutris, P ., Grey, C., Hines, D., Williams, I., Flanagan, S.A., & Goswami, U. 855 (2023). Neural phase angle from 2 months when tracking speech and non-speech rhythm 856 linked to language performance from 12 to 24 months. Brain & Language, 243, 105301. 857 Ortiz, A., Martinez-Murcia, F. J., Luque, J. L., Giménez, A., Morales-Ortega, R., & 858 Ortega, J. (2020). Dyslexia diagnosis by EEG temporal and spectral descriptors: An 859 anomaly detection approach. International Journal of Neural Systems, 30 (07), 2050029. 860 https://doi.org/10.1142/S012906572050029X 861 Schneider, W., Roth, E., & Ennemoser, M. (2000). Training phonological skills and letter 862 knowledge in children at risk for dyslexia: A comparison of three kindergarten intervention 863 programs. Journal of Educational Psychology, 92, 284–295. 864 http://dx.doi.org/10.1037/0022-0663.92.2 .284 865 Snowling, M. J. (2000). Dyslexia (2nd ed.). Oxford: Blackwell Publishers.. 866 Stanovich, K. E. (1998). Refining the Phonological Core Deficit Model. Child Psychology 867 and Psychiatry Review, 3, 17-21. https://doi.org/10.1111/1475-3588.00203 868 Stein, J., & Walsh, V. (1997). To see but not to read: The magnocellular theory of 869 dyslexia. Trends in Neuroscience, 20, 147–152. https://doi.org/10.1016/S0166-870 2236(96)01005-3 871 Tan Y , Chanoine V, Cavalli E, Anton J-L and Ziegler JC (2022) Is there evidence for a noisy 872 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 43 computation deficit in developmental dyslexia? Front. Hum. Neurosci. 16:919465. doi: 873 10.3389/fnhum.2022.919465 874 Valdois, S., Bosse, M. L., & Tainturier, M. J. (2004). The cognitive deficits responsible for 875 developmental dyslexia: Review of evidence for a selective visual attentional disorder. 876 Dyslexia, 10, 4, 339-363. https://doi.org/10.1002/dys.284. 877 Van Herck, S., Vanden Bempt, F., Economou, M., Vanderauwera, J., Glatz, T., Dieudonné, 878 B., Vandermosten, M., Ghesquière, P . and Wouters, J., (2022). Ahead of maturation: 879 Enhanced speech envelope training boosts rise time discrimination in pre ‐readers at 880 cognitive risk for dyslexia. Developmental Science , 25(3), p.e13186. DOI: 881 10.1111/desc.13186 882 Yang, Y ., Yang, Y . H., Li, J. J., Xu, M., & Bi, H. Y . (2020). An audiovisual integration deficit 883 underlies reading failure in nontransparent writing systems: An fMRI study of Chinese 884 children with dyslexia. Journal of Neurolinguistics, 54, 100884. 885 https://doi.org/10.1016/j.jneuroling.2019.100884 886 Ziegler, J. C., & Goswami, U. (2005). Reading acquisition, developmental dyslexia, and 887 skilled reading across languages: A psycholinguistic grain size theory. Psychological 888 Bulletin, 131, 3-29. https://doi.org/10.1037/0033-2909.131.1.3 889 Ziegler, J. C., Bertrand, D., Tóth, D., Csépe, V., Reis, A., Faísca, L., Saine, N., Lyytinen, 890 H., Vaessen, A., Blomert, L. (2010). Orthographic Depth and Its Impact on Universal 891 Predictors of Reading: A Cross-Language Investigation. Psychological Science. 21 (4), 892 551-559. https://doi.org/10.1177/0956797610363406 893 894 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint 44 FIGURE LEGENDS 895 Figure 2.1 Panel (A) shows the whole Experimental Protocol, including the time line of each 896 BCI session and the spaceship interface. Panel (B) depicts the decoder logic behind the neural 897 feedback. 898 Figure 3.1. The t-statistics used as a measure of BCI performance. Data are shown for each 899 participant, under real-time feedback versus offline preprocessing. 900 Figure 3.2. The correlation between BCI intervention training improvement and behavioral task 901 improvement (real-time preprocessing). 902 Figure 3.3. The correlation between BCI intervention training improvement and behavioral task 903 improvement (offline preprocessing). 904 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint Daily Session: BCI task: 2 runs Rest: ~ 5mins Experimental Protocol: BCI training: 8 daily sessions Cognitive testing: 2 days (pre, post) Rest: 2 days Baseline Stage: 4 mins BCI Stage: 10 mins 𝑥 𝑡 = 𝜃(𝑡) 𝛿(𝑡) log[𝑥(𝑡)] Ratio distribution log distribution BCI task (a) (b) EEG data .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint Syllable Stress RT TOWRE PDE .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint TOWRE PDE Rise time threshold Stress syllable accuracy .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.700941doi: bioRxiv preprint