---

KEYS :

• EEG spikes can affect language and attention—even without obvious seizures—but impact varies. The location (e.g., left perisylvian/temporal “language” areas) raises plausibility, yet the burden (how often spikes happen, especially during non-REM sleep) matters most. (PMC)
• High, sleep-activated spike burden (e.g., the epilepsy–aphasia spectrum, LKS/ESES) clearly impairs language; brief or low-burden spikes usually act as a modifier, not the sole cause of a child being non-verbal. (PMC)
• Treatment can help some children, but results are mixed. Levetiracetam reduced spikes and improved language/behavior in a pediatric RCT with subclinical discharges; sulthiame can suppress spikes, yet one small study reported worsened reading/memory despite EEG improvement. Decisions should be individualized and closely monitored. (PMC, PubMed)

---

Key definitions (super simple)

• Interictal epileptiform discharges (IEDs): brief “spikes” or spike-and-wave bursts on EEG between seizures. They can occur in kids who never have noticeable seizures. (PMC)
• Spike-Wave Index (SWI): the percentage of non-REM sleep occupied by spike/spike-wave activity. High SWI (historically ≥80–85%) is characteristic of CSWS/ESES; lower levels may still matter but are less clearly harmful. (PMC)
• Perisylvian/temporal language network: brain areas that process speech sounds and language meaning. Spikes here are more likely to disturb listening/understanding in the moment. (NCBI)

---

What the research actually shows

1) Spikes can hurt thinking and school skills—even without seizures

Multiple studies link IEDs to weaker cognitive and academic performance in children, independent of obvious seizures. Effects include language, memory, and attention differences compared with peers. (PMC, PubMed)

• In rolandic/centrotemporal patterns, children (with or without seizures) show language and memory weaknessesversus controls. (PubMed)
• A 2024 study of new-onset rolandic discharges (with and without seizures) again suggested higher risk of cognitive impairment. (epilepsybehavior.com)

2) Burden beats location for predicting impact

Where spikes appear matters (language cortex makes language disruption more plausible), but how often and when they happen matters more. High, sleep-activated spike loads (the epilepsy–aphasia spectrum) are most strongly tied to major language problems and regression. (PMC)

Practical takeaway: A child can have spikes “in the language area” and still not have a “spike-driven aphasia” if the spike burden is low and not continuous in NREM.

3) Awake spikes vs. sleep spikes

Frequent awake IEDs (e.g., present in >10% of the recording) can measurably impair performance on attention/processing tasks. Sleep spikes are important too, especially when persistent through NREM; extremely high SWI defines ESES. (PubMed, PMC)

4) The contrast case: LKS/ESES (why your child’s EEG probably isn’t that)

Landau–Kleffner syndrome (LKS) and CSWS/ESES show near-continuous sleep spikes over perisylvian/posterior temporal regions and cause auditory verbal agnosia (“word deafness”) and marked language regression—often even without obvious seizures. These are high-burden, sleep-dominated patterns. (PMC)

---

“So can brief spikes alone cause a child to be non-verbal?”

• Sometimes a contributor, rarely the sole cause. Brief/intermittent spikes—especially around sleep onset—can create transient cognitive impairment (“missed moments” of processing), making learning language harder, but they don’t usually explain a child being fully non-verbal by themselves. Expect modest contributions unless the sleep spike burden is high. (PMC)

Think of spikes as static on the line: if static is loud and constant (high SWI), communication breaks down; if static is brief and occasional, it adds friction but usually isn’t the whole story.

---

Will treating the spikes make a child “normal”?

What looks promising

• Levetiracetam: In a randomized controlled trial in young children with subclinical discharges (no classic seizures), levetiracetam reduced spikes and was associated with better language, social, and cognitive scoresover ~6 months. (PMC)

Why results can be mixed

• Sulthiame (sultiame): Effective at suppressing rolandic spikes, but in a small prospective study all six children showed worsened reading, and most had declines in memory/attention despite large EEG improvements. This doesn’t mean sulthiame is “bad” for every child; it does mean spike suppression ≠ automatic language gains, and medication choice/side-effect profile matter. (PubMed)

Bottom line: Treating spikes can help some kids—especially if spikes are a big driver—but outcomes vary. Monitor function, not just the EEG.

---

How to figure out whether spikes really matter for your child

1. Quantify sleep spike burden (SWI).
   Ask for an overnight (12–24 h) video-EEG that reports:
   • SWI for N2/N3 (% of 1-s epochs with ≥1 spike)
   • Spikes/min in wake and N1
   • Topography (e.g., left posterior-temporal predominance)
     Daytime “nap” EEGs often miss the true overnight burden; an overnight study is more informative. (PMC)
2. Run a time-boxed treatment trial (if agreed).
   Over 4–6 weeks, track functional changes—not just the EEG:
   • Response to name; following 1-step commands
   • Number of new sounds/syllables/words
   • Joint attention (look-and-share time)
   • Sleep quality (bedtime, night-wakings)
3. Optimize everything that amplifies spike effects.
   Good, regular non-REM sleep; screen/treat snoring/OSA; check iron/ferritin and vitamin D if advised—sleep and health influence spike load and learning.

---

Don’t ignore the non-EEG pieces of the puzzle

Even when spikes are present, many children’s main challenges stem from developmental language disorder (DLD), childhood apraxia of speech (CAS), autism spectrum, or global developmental delay. These require therapy-firstapproaches regardless of EEG:

• Speech–language therapy with CAS-savvy methods (e.g., DTTC, ReST), 3–5 sessions/week when possible.
• Parent-implemented programs (Hanen, JASPER/PRT-style play coaching).
• AAC now (picture exchange or a robust speech-generating app). AAC supports speech; it doesn’t “block” it.
• Occupational therapy for sensory/self-regulation; behavioral supports as needed.

Medication that truly helps will often show up first as steadier listening, less “on/off” hearing, better attention, and more vocal play—creating a stronger foundation for therapy to land.

---

Frequently asked questions:

Q: My child’s spikes are in the “hearing/language area.” Doesn’t that prove they caused the delay?
A: It increases plausibility—but without a high spike burden (especially in NREM sleep), spikes are usually a piece of the puzzle, not the whole picture. Measure the SWI to move from “possible” to known. (PMC)

Q: Are awake spikes important?
A: Yes—if they’re frequent. Studies show that frequent awake IEDs (e.g., >10% of the record) can impair attention and processing speed. (PubMed)

Q: If we suppress the spikes, how fast should we see change?
A: If spikes are a major driver, families often notice better listening/engagement within weeks, and language inches forward as therapy takes advantage of that. If no functional gains appear over 4–6 weeks at a therapeutic dose, spikes may be secondary and the plan should be revisited. (PMC)

Q: Should every child with spikes be medicated?
A: No. Many neurologists “treat the child, not just the EEG.” Consider a trial when language delay is severe and IEDs are frequent/sleep-activated, while weighing side-effects and tracking real-world gains. (PMC)

---

A quick, copy-paste note you can bring to your neurologist

Request: “We’d like an overnight video-EEG with a Spike-Wave Index (SWI)—the % of 1-s NREM epochs containing spike/spike-wave—plus spikes/min in wake and N1, and a map of lateralization/topography. If we try medication, we’ll run a 4–6-week functional checklist (listening to name, 1-step commands, new syllables/words, joint attention, sleep). We’ll continue intensive SLT (CAS-aware) and AAC in parallel.”

---

References (selected)

• Cheng D, et al. The effect of interictal epileptiform discharges on cognitive and academic performance in children.2020. (IEDs linked to lower cognitive/academic performance.) (PMC)
• García-Ramos C, et al. Cognition and Brain Development in BECTS. 2015. (Subtle but real cognitive effects in “benign” rolandic patterns.) (PMC)
• Danielsson J, et al. Cognitive deficits in children with benign rolandic epilepsy. 2009. (Language/memory differences vs. controls.) (PubMed)
• Ebus S, et al. Cognitive effects of IEDs in children. 2012. (Frequent awake IEDs >10% impair performance.) (PubMed)
• Singhal NS, et al. Continuous Spike-Wave during Slow Wave Sleep and related conditions. 2014. (SWI thresholds; why high sleep burden matters.) (PMC)
• Shah S, et al. 2023. (Daytime sleep vs. overnight EEG for SWI; overnight often needed.) (PMC)
• Pearl PL. The Landau–Kleffner Syndrome. 2001. (Auditory verbal agnosia; posterior temporal focus in LKS.) (PMC)
• Wang M, et al. Levetiracetam in children with ASD + subclinical discharges (RCT). 2017. (Reduced spikes with language/behavior gains.) (PMC)
• Wirrell E, et al. Sulthiame in BECTS. 2008. (Spike suppression but worse reading/memory in a small cohort.) (PubMed)
• Neumann H, et al. Cognitive development in children with new-onset rolandic discharges. 2024. (Higher risk of cognitive impairment with/without seizures.) (epilepsybehavior.com)

---

KEY Point:

If your child’s EEG shows brief, localized spikes near language areas and their biggest challenge is communication, it’s reasonable to ask: “Are these spikes part of the problem?” The fairest answer is: possibly—and we can measure how much. Quantify the sleep spike burden, run a time-boxed, function-tracked treatment plan if indicated, and double-down on therapy. That combination gives your child the best chance to grow their voice. Therefore, we suggest always take a 24 hours EEG for your child and try to remove those big spikes for a month or two month with medication and check the improvement of the child.

---

If your child’s EEG shows brief, localized spikes near language areas and their biggest challenge is communication, it’s reasonable to ask: “Are these spikes part of the problem?” The fairest answer is: possibly—and we can measure how much. Quantify the sleep spike burden, run a time-boxed, function-tracked treatment plan if indicated, and double-down on therapy. That combination gives your child the best chance to grow their voice.

  WE SUGGEST ALWAYS MAKE EEG FOR CHILDREN WITH LANGUAGE DELAY AND DEVELOPMENT DELAY AND TRY TO REMOVE HIGH SPIKES WHICH ARE ON IMPORTANT LOCATIONS OF SUCH DELAY IN THE CHILD.

PART 2 – HOW AI AND LLM CAN HELP

---

---

AI-Assisted Exploration of EEG Spikes and Language Delay in Children

So far on this article, we examined the phenomenon of children with abnormal EEG spikes who present with severe language delay or nonverbal behavior despite not meeting criteria for Landau‐Kleffner syndrome (LKS) or Electrical Status Epilepticus in Sleep (ESES).

In this second part, we discuss how advanced AI tools – especially large language models (LLMs) – can aid researchers and clinicians in uncovering relevant scientific literature to investigate this issue. We then review key findings from recent studies to evaluate the hypothesis that reducing or eliminating such EEG spikes could improve speech and language comprehension in these children, even outside of classic epileptic aphasia syndromes.

---

Leveraging LLMs and AI to Find Relevant Research

Research into the cognitive effects of subclinical epileptiform EEG discharges spans neurology, pediatrics, and psychiatry, making it challenging for clinicians to gather all pertinent information. LLMs (like ChatGPT and related AI assistants) can greatly streamline this process by quickly sifting through vast bodies of text and extracting key insights:

• Rapid Literature Search & Summarization
  AI language models can scan thousands of journal articles, conference papers, and case reports for specific patterns (e.g. “EEG spikes in children with language delay”). They generate concise summaries of findings, saving researchers the time of reading each paper in full. For example, LLMs are capable of efficiently extracting data and producing quality summaries from large, unstructured medical text, helping experts keep up with rapidly evolving evidence (Spence & Schneider, 2009).
• Uncovering Connections
  Beyond summarizing individual papers, LLMs can highlight connections across studies that might not be obvious. By synthesizing information from neurology, developmental pediatrics, and neuropsychology literature, an AI assistant might reveal a pattern – for instance, noting that several independent case reports across decades all observed language gains after a particular anti-seizure therapy.
• Identifying Effective Therapies and Gaps
  An AI system can quickly enumerate what therapies have been tried (e.g. valproate, levetiracetam, corticosteroids) and which appeared most effective at reducing EEG spikes or improving language. At the same time, it can highlight gaps in knowledge – such as the lack of large randomized controlled trials – alerting researchers to areas in need of further study.
• Customized Queries and Hypothesis Generation
  AI tools allow clinicians to ask very specific questions (e.g. “Do subclinical spike waves in the left temporal lobe correlate with expressive language delays?”). By aggregating data across studies, the model might suggest that left-hemisphere spikes indeed disrupt verbal processing, or that controlling such activity early in development could yield language improvements.

⚠️ Of course, LLMs are not infallible – they may produce errors or “hallucinate” information. Researchers must use these tools as a starting point and cross-check critical facts against original sources. But when used responsibly, AI and LLMs can dramatically accelerate the literature review process and enable a more data-driven approach to complex clinical questions.

---

For example on our topic : Lets quickly Try again with LLM, What the Literature Reveals – Can Removing EEG Spikes Improve Language?

Using AI-assisted literature review, we compiled and analyzed scientific studies related to children who exhibit significant language impairment alongside epileptiform EEG abnormalities, but who are not diagnosed with LKS or ESES.

This group can include children with autism spectrum disorder (ASD) or developmental language disorder who have epileptic discharges on EEG without obvious seizures.

The central question is: Do these EEG spikes contribute causally to their speech/language deficits – and if so, can treating them lead to improvement in communication?

Why This Hypothesis Matters

• In LKS and related encephalopathies like CSWS/ESES, aggressive treatment of epileptiform activity often leads to recovery of language function, implying a direct link between EEG spikes and language ability.
• The children in this context differ in that they may have never fully developed speech, and their EEG shows intermittent spikes (often during sleep) that do not meet the full criteria of LKS/ESES.
• Some autistic children show LKS-like EEG patterns but regress earlier and do not consistently respond to standard LKS treatments.

This raises the scientific debate: Are the EEG spikes a cause of the language impairment, or merely co-occurring with an underlying developmental disorder?

---

Evidence from the Literature

• Case Reports of Improvement
  • Example: Three autistic children (ages 3–5) with frequent spikes but no seizures improved dramatically in language and social skills within one month of starting valproic acid. They no longer met autism criteria at 7–11 months follow-up (Chez et al., 2004).
• Observational Studies
  • A clinical series of 176 children with epileptiform EEG discharges treated with AEDs (mostly valproate) showed 46.6% normalized EEGs after ~10 months, ~17% improved, and none worsened (Galanopoulou et al., 2009).
• ASD-Specific Findings
  • Divalproex sodium: beneficial in ASD children with mood instability/EEG abnormalities (Hughes & Melyn, 2005).
  • Levetiracetam: reduced hyperactivity but not language.
• Contrasting Results
  • An RCT (Hirota et al., 2018) found no significant effect of anticonvulsants on autism/language outcomes.
  • Tharp (2004) concluded: “no justification” for anticonvulsants in ASD without seizures.

---

Mechanistic Rationale

• Spikes = network disruption
  • Abnormal synchronous firing → brief disruption of brain processing.
• Location matters
  • Left hemisphere spikes affect verbal/language function.
  • Right hemisphere spikes affect visuospatial skills.
• Frequency and duration
  • Frequent or long discharges (>3s) impair cognition more than brief/infrequent spikes.

Thus, repeated spikes in language areas could act like “tiny short-circuits” in the speech network, preventing normal perception/production.

---

Therapy Approaches

• Antiepileptic drugs (AEDs): valproate, levetiracetam, benzodiazepines, ethosuximide.
• Steroids/ACTH: dramatic in LKS, less proven in non-LKS.
• Other approaches: IVIG, rare epilepsy surgeries.
• Always combined with speech-language therapy.

Timing is critical: Early treatment (ages 1–3) may allow normal language development; late treatment often shows limited benefit.

---

Closing Thoughts

The evidence is mixed but promising:

• Some children clearly improve when spikes are suppressed.
• Others show no benefit, suggesting developmental pathology is primary.

Likely, two groups exist:

1. Spike-dependent children – language delay caused/exacerbated by EEG spikes.
2. Developmental-origin children – language delay independent of spikes.

The challenge: distinguish subgroups early.

LLMs and AI can accelerate progress by:

• Continuously scanning new studies.
• Linking EEG patterns with language outcomes.
• Helping identify biomarkers to predict responders.

---

Sources

1. Spence, S.J., & Schneider, M.T. (2009). The role of epilepsy and epileptiform EEGs in autism spectrum disorders.Pediatric Research, 65(6), 599–606.
2. Millichap, J.G. (1994). Abnormal EEG in Autism: Valproate Response. Pediatric Neurology Briefs, 8(3):20. Link
3. Great Ormond Street Hospital. Landau-Kleffner Syndrome (LKS). Link
4. Galanopoulou, A.S., et al. (2009). Should epileptiform discharges be treated? Epilepsia, 50(Suppl 8), 18–24.
5. Foliaki, S.T., et al. (2019). Association between Interictal Epileptiform Discharges and Autism Spectrum Disorder.Brain Sciences, 9(10):283. PMC
6. Chez, M.G., et al. (2004). Improvement in Language with Antiepileptic Therapy in Autism. Ann Neurol, 55(Suppl 1): S153.
7. Hughes, J.R., & Melyn, M. (2005). Correlation between EEG abnormalities and autistic regression. Epilepsia, 46(1): 18–24.
8. Tharp, R.I. (2004). Epileptic encephalopathies and their relationship to developmental outcome. Epilepsia, 45(Suppl 2): 134–137.
9. Clusmann, J. et al. (2023). The future landscape of large language models in medicine. Communications Medicine, 3:141. Nature
10. ASHA Leader. Identification and Treatment of Landau-Kleffner Syndrome. Link

---