Criticism and open problems

The human eye detects single photons. ↗iDirect detection of a single photon by humansNature Communications · 2016 · journalMetadata matched Sharks detect 0.5 µV/m electric fields. Migratory birds' compass is disrupted by 15 nT RF noise ↗iResonance effects indicate a radical-pair mechanism for avian magnetic compassNature · 2004 · journalMetadata matched — 0.03% of the geomagnetic background. Biology operates at the quantum limit of electromagnetic sensitivity because evolution optimized detection, not tolerance.

Therapeutic stimulation and field-based devices demonstrate effects under specified clinical protocols. Their usefulness is a positive control for physical intervention. Transfer to ambient exposure requires local dose, frequency-dependent tissue transfer, waveform and the same endpoint; regulatory approval and patent claims are separate evidence classes.

The Ion Forced Oscillation mechanism (Panagopoulos 2025, Frontiers in Public Health ↗iHuman-made electromagnetic fields: Ion forced-oscillation and voltage-gated ion channel dysfunction, oxidative stress and DNA damage (Review)International Journal of Oncology · 2021 · reportMetadata matched) demonstrates a biological response threshold of 10⁻⁵ V/m for polarized, coherent fields — five orders of magnitude below typical environmental levels. There is no intensity gap. The gap existed only in models that assumed thermal effects were the only mechanism.

Kim 2026 ↗iElectromagnetic field-inducible in vivo gene switch for remote spatiotemporal control of gene expressionCell · 2026 · journalMetadata matched identifies CYB5B dependence in an engineered gene-switch response using 60 Hz burst repetition and a 4 kHz intraburst structure. A required component is a candidate sensor or mediator; proximal readouts and rescue/bypass experiments distinguish those roles. The result does not transfer directly to an ambient 60 Hz sinusoid.

In 2013, honeybee researcher Uwe Greggers ↗iReception and learning of electric fields in beesProceedings of the Royal Society B: Biological Sciences · 2013 · journalMetadata matched stated that anthropogenic electric fields are 'much lower in energy than those produced by the bees themselves' and that bees 'should be naturally protected.' In 2025, Mallinson et al. ↗iWeak anthropogenic electric fields affect honeybee foragingiScience · 2025 · journalMetadata matched tested this empirically in field experiments — and found that anthropogenic AC electric fields reduce bee foraging by 71% (iScience / Cell Press). The intuitive assessment that 'the field is too weak' failed the empirical test. This is exactly the error ICNIRP makes for human health: assuming without testing that environmental field strengths are below biological thresholds.

TTFields uses dielectrophoresis (DEP) — a quadratic mechanism that requires field strengths of 100–300 V/m. Environmental IF exposure operates via a fundamentally different mechanism: Ion Forced Oscillation (IFO-VGIC), which is linear in field strength with a demonstrated threshold of 10⁻⁵ V/m (Panagopoulos 2025 ↗iHuman-made electromagnetic fields: Ion forced-oscillation and voltage-gated ion channel dysfunction, oxidative stress and DNA damage (Review)International Journal of Oncology · 2021 · reportMetadata matched). Different mechanisms have different intensity requirements.

Pulsed LED-driver fields are more biologically active per watt than TTFields' pure sine wave (Campisi 2010iReactive oxygen species levels and DNA fragmentation in astrocytes after low-intensity microwave exposureNeuroscience Letters · 2010 · journalNo verified source link: amplitude-modulated 900 MHz produced more ROS and DNA fragmentation than continuous wave). LED drivers produce pulsed, harmonically rich waveforms with 50,000+ switching transitions per second — each transition a potential Ca²⁺ spike via VGCC gating asymmetry.

The TTFields clinical program validates that IF frequencies disrupt cell division. It does not define the minimum intensity at which disruption begins — that question is answered by IFO biophysics, not by the DEP mechanism TTFields employs.

Critically, Neuhaus et al. (Nature 2020) showed that TTFields' frequency-dependence differs between cancer and normal cells: cancer cells are most affected at 150–200 kHz, while normal cell mitosis is disrupted at ~50 kHz. LED driver switching frequencies (20–100 kHz) span precisely this normal-cell sensitivity range — a coincidence that the intensity-only argument cannot address.

Yes. Education, contraception, housing, labour markets, partnership formation, migration, policy, desired family size, tempo and ART all affect observed fertility. A period TFR is a five-year-age-group sum, not a direct assay of gametes or conception.

V2 therefore models ASFR before TFR and keeps demand/opportunity, tempo and ART/live-birth delivery explicit. It does not assign their residual variation to a biological field pathway.

However: sleep deprivation alone reduces testosterone by 10–15% and sperm count by 29% in controlled laboratory experiments on healthy young men — without any cultural, behavioral, or chemical exposure change (Leproult & Van Cauter 2011 ↗iEffect of 1 Week of Sleep Restriction on Testosterone Levels in Young Healthy MenJAMA · 2011 · journalMetadata matched; Walker 2017iWhy We Sleep: Unlocking the Power of Sleep and DreamsScribner · 2017 · bookNo verified source link). The question is not whether sleep affects fertility (it does, conclusively) but whether EMF disrupts sleep (the CRY/RPM mechanism and LED melatonin suppression data say yes).

Lindgren's 2025 metric is BERM's theory premise. BERM then proposes background, orientation and spectral dependence as discriminating hypotheses. The reported 87.5% algebraic correspondence with RPM is a structural comparison, not a derived geometry-to-RPM operator, so it does not give the CRY pathway a direct geometric foundation.

Each downstream biological link — from field geometry to chromophore response to organ-level endpoints — requires its own experimental validation. The relevant tests are discriminating physical and biological experiments: pre-specified angle, background or PSD dependence with calibrated fields and appropriate sham/thermal controls.

The registry includes bounded findings such as in-vitro human sperm endpoints, animal blood–testis-barrier and ovarian studies, and mechanistic redox/tight-junction work. These can motivate organ-specific states rather than a single generic biological-capacity curve.

Their systems, frequencies, amplitudes, durations and endpoints differ. Animal and cell results cannot simply be converted into a population dose, fertility probability or country forecast.

Mobile subscription density is a composite proxy for the overall electromagnetic environment — it tracks base station deployment, Wi-Fi proliferation, IoT density and indoor electronics adoption, not just RF exposure. The current N = 163 cohort result uses this composite proxy for descriptive cohort analysis.

Likewise, eDRX is device reception/paging scheduling metadata, not by itself a known downlink RF field signature. Any envelope or beat feature must be measured in the actual field before it is tested biologically.

The evidence base is heterogeneous. Study quality, exposure characterisation, thermal control, endpoint selection and replication vary substantially. Reviews can establish that findings exist across systems, but their certainty assessments and sensitivity analyses need to be reported rather than replaced with a single headline.

For example, the WHO-commissioned reproductive reviewiIncomplete referenceEnvironment International · 2024 · meta-analysisNo verified source link reported adverse findings in several analyses while rating much of the certainty low or very low and requiring sensitivity to high-SAR studies. V2 treats that as context, not a settled population effect.

Yes. These exposures may affect reproductive biology and may co-vary with technology, urbanisation and socioeconomic change. They are competing explanations and potential interactions, not nuisance variables that can be dismissed by a simple correlation.

A useful test measures or designs around plausible co-exposures, compares alternative causal models and reports which inference changes when they are included.

BERM’s metabolic synthesis permits BMI to mediate part of a pathway, but this is a causal hypothesis rather than a universal adjustment rule. Specify the graph and target estimand before choosing covariates; BMI may also proxy confounding or change over time.

Mazur et al. 2013 ↗iIs Rising Obesity Causing a Secular (Age-Independent) Decline in Testosterone among American Men?PLoS ONE · 2013 · journalMetadata matched (PLOS ONE, n = 991 US Air Force veterans, 20-year follow-up) provides the critical test: men who MAINTAINED THEIR WEIGHT still lost 117 ng/dL (19%) of their testosterone over 20 years. Obesity cannot explain this decline. The direct pathway accounts for approximately two-thirds of the total effect; the mediated pathway (via BMI) accounts for approximately one-third.

Santi et al. 2025 ↗iTemporal trends in serum testosterone and luteinizing hormone levels indicate an ongoing resetting of hypothalamic-pituitary-gonadal function in healthy men: a systematic reviewJournal of Endocrinological Investigation · 2025 · journalMetadata matched (n = 1,064,891, the largest meta-analysis ever conducted) found no BMI temporal trend in their study population, yet testosterone declined significantly. Rising obesity is not the driver in this dataset.

Klimentidis 2010 ↗iCanaries in the coal mine: a cross-species analysis of the plurality of obesity epidemicsProceedings of the Royal Society B: Biological Sciences · 2010 · journalMetadata matched motivates a shared metabolic comparison across species. Liver glucose output, insulin dynamics, repair and meal/tissue phase provide measurable links. The trends do not by themselves identify EMF or eliminate differences in nutrition, activity, breeding, chemicals and disease.

A measurement-ready FieldState needs documented calibration, B₀ vector, organ transfer, PSD, circadian context, phase/coherence and provenance. It must then be joined to a pre-specified organ or couple endpoint with evidence- and parameter-linked mappings.

Calibration should use a training period only, followed by an independent laboratory replication and a held-out ASFR/TFR period. Both null and non-null results should update the causal registry.

LED lights are certified for electromagnetic compatibility (EMC) — meaning their emissions don't interfere with other electronic devices beyond regulatory limits. They are not certified for biological safety of their intermediate-frequency emissions. CISPR 15, the standard that governs LED lighting EMF, was designed to protect radio reception, not biological systems.

No regulatory body has evaluated the biological effects of continuous 20–200 kHz fields from LED drivers. The Panagopoulos 2025 ↗iHuman-made electromagnetic fields: Ion forced-oscillation and voltage-gated ion channel dysfunction, oxidative stress and DNA damage (Review)International Journal of Oncology · 2021 · reportMetadata matched IFO-VGIC threshold (10⁻⁵ V/m) is orders of magnitude below any EMC limit. EMC compliance and biological safety are entirely different standards measuring entirely different things.

Zeghoudi et al. 2025 ↗iExperimental measurement of the near electric emission from LED driver as function of power variation and lighting modeOptik · 2025 · journalMetadata matched (Optics & Laser Technology) directly measured LED driver near-field emissions and confirmed E-field components at centimeter distances. A typical home has 15–30 LED bulbs, each containing a switch-mode power supply. The EU incandescent ban (2009–2012) replaced zero-IF sources with continuous-IF sources for ~450 million people without any assessment of the electromagnetic change.

A 2022 IJRB systematic review of IF-EMF ↗iEffects of intermediate frequency electromagnetic fields: a review of animal studiesInternational Journal of Radiation Biology · 2023 · journalMetadata matched (300 Hz–10 MHz) animal studies confirmed that this frequency band has received minimal health research compared to ELF and RF — the regulatory gap is now documented in the peer-reviewed literature itself.

The Adey-Blackman calcium window, documented since 1976iEffects of weak amplitude-modulated microwave fields on calcium efflux from awake cat cerebral cortexBioelectromagnetics · 1976 · reportNo verified source link, shows that EMF biological effects do not follow a linear dose-response. Calcium efflux from cat brain tissue occurred at specific intensity windows (0.1–1.0 mW/cm² at 450 MHz amplitude-modulated at 16 Hz) but NOT at higher or lower levels. This 'window effect' means the ICNIRP approach — setting a threshold above which effects occur — is structurally wrong. Effects occur in windows, not above thresholds.

This also explains why replication studies that use different intensities may fail to find effects that the original study found: they may be testing outside the window. The failure to replicate is not evidence of absence — it is evidence of non-linearity. The window effect has been confirmed across multiple laboratories and biological endpoints since 1976.

The window phenomenon is consistent with the physics of resonance: biological systems respond maximally at specific frequencies and intensities where energy transfer to target molecules is optimized. Above and below these windows, the coupling is less efficient. This is the same physics that makes a radio tuner work — it detects signals only at the tuned frequency, not above or below it.

In linear models, EMF and GDP are collinear (r = 0.87) and neither is significant after controlling the other. This is a symmetric identification problem, not evidence against EMF.

Three structural differences distinguish EMF from GDP. First, electricity access is a binary threshold. Adjusting for the fraction of the population with electricity access improves TFR prediction (r: −0.864 → −0.885). GDP has no equivalent threshold — everyone participates in economic activity at some level, but only some people have electricity.

Second, mobile phones (the information device) are the WEAKEST EMF proxy (RMSE 1.053). Residential electricity (the infrastructure) is the BEST (univariate RMSE 0.533). If 'information → choices' were the mechanism, the information device should predict best. It doesn't.

Third, sentinel species respond to electric fields (Mallinson 2025 ↗iWeak anthropogenic electric fields affect honeybee foragingiScience · 2025 · journalMetadata matched: bees −71%) but not to GDP. Dogs, bees, and frogs do not make economic choices.

Electricity access correlates with development, but it has a property that no other development indicator has: a PHYSICAL THRESHOLD. The IFO-VGIC activation threshold (10⁻⁵ V/m) is exceeded by every household appliance at operating distance. This means electrification is not just correlated with biological effects — it is the MECHANISM of exposure.

In partially electrified countries (Nigeria 55%, Ethiopia 51%, Uganda 42%), the national TFR is a mixture of the electrified population (exposed to EMF, lower TFR) and the unelectrified population (not exposed, TFR near biological maximum ~6.5). No other development indicator has this binary mixture structure. You cannot say '55% of the population has education' in the same binary way — education is a continuous gradient.

The binary threshold produces a testable prediction: DHS (Demographic and Health Survey) micro-data comparing TFR of electrified vs unelectrified households within the same country, controlling for income and education, should show lower TFR in electrified households. This test has not yet been conducted.

Sousouri et al. 2025 ↗iCACNA1C genotype determines sleep EEG response to 5G exposure in double-blind studyNeuroImage · 2025 · journalMetadata matched (NeuroImage, ETH Zurich) conducted a double-blind randomized controlled trial with CACNA1C genotyping. Participants carrying the rs7304986 T/C variant showed altered sleep spindle dynamics under 3.6 GHz RF exposure below ICNIRP limits, while CC homozygotes did not. The subjects did not know their genotype, the exposure condition, or the hypothesis. Nocebo cannot produce genotype-dependent neurophysiological changes in a double-blind design.

This transforms the EHS question from 'do self-reporters feel something?' (where nocebo is a valid concern) to 'does ion channel genotype predict measurable brain response to RF?' (where nocebo is mechanistically excluded). The relevant variable is not self-diagnosis but VGCC polymorphism — a biological property the subject cannot influence by belief.

Belpomme et al. 2022 ↗iWhy electrohypersensitivity and related symptoms are caused by non-ionizing man-made electromagnetic fields: An overview and medical assessmentEnvironmental Research · 2022 · reportMetadata matched characterized ~1,000 EHS patients with objective biomarkers (histamine, S100B, nitrotyrosine). Combined with Sousouri's genotype data ↗iCACNA1C genotype determines sleep EEG response to 5G exposure in double-blind studyNeuroImage · 2025 · journalMetadata matched, the evidence points to a continuous distribution of EMF sensitivity across the population, with the clinical EHS phenotype representing the tail — not a psychosomatic category.

This argument has been applied to every non-thermal EMF finding for 50 years — and it has been wrong every time.

In 1976, AdeyiEffects of weak amplitude-modulated microwave fields on calcium efflux from awake cat cerebral cortexBioelectromagnetics · 1976 · reportNo verified source link and Blackman documented calcium efflux from brain tissue at specific field intensities. Dismissed as 'artifact' because the non-linear dose-response was 'implausible.' In 2026, Kim et al. (Cell) ↗iElectromagnetic field-inducible in vivo gene switch for remote spatiotemporal control of gene expressionCell · 2026 · journalMetadata matched showed that rhythmic calcium oscillations — not linear calcium increase — drive gene expression, explaining the window effect.

In 1995, Lai and Singh ↗iAcute low‐intensity microwave exposure increases DNA single‐strand breaks in rat brain cellsBioelectromagnetics · 1995 · journalMetadata matched found DNA strand breaks from 2450 MHz radiation. Industry funded counter-studies and pressured their university. The mechanism was later identified: oxidative stress via VGCC-mediated calcium influx, confirmed by melatonin's protective effect.

In 2026, Kim et al.'s Cell paper ↗iElectromagnetic field-inducible in vivo gene switch for remote spatiotemporal control of gene expressionCell · 2026 · journalMetadata matched was called 'incredibly implausible' by physicist Andrew YorkiCommentary on the Kim et al. electromagnetic-field sensor studyNew Scientist · 2026 · commentaryNo verified source link. The paper used CRISPR screening to identify Cyb5b as an EMF sensor and demonstrated reversible gene expression control in transgenic mice.

The pattern is consistent: data is strong, mechanism is unknown, critics declare impossibility. Then the mechanism is found. The 'implausibility argument' is not science — it is an argument from ignorance.

BERM offers a testable physical proposal rather than a completed solution. For a normalized positive-norm mode, χ_geo follows from the inverse rank-one metric; under explicit matter–metric and linear-response assumptions, a conditional response-operator form also follows. These derivations do not supply the ion-channel or tissue kernel, its sign, lag or calibration. The photon-sensor analogy motivates sensitivity experiments but cannot substitute for those measurements.

Blue light does suppress melatonin — this is established. But it is not the whole story.

Duraccio et al. (2019)iBlue light filtering does not improve sleep outcomes in adolescentsJ Adolesc Health · 2019 · reportNo verified source link found that blue-light-filtering glasses did NOT significantly improve adolescent sleep quality. If blue light were the only mechanism, filtering it should work. It doesn't.

BERM proposes that LED lamps produce TWO biologically relevant outputs: blue light (optical, retinal) and IF emissions (electromagnetic, systemic). Blue-light filtering addresses the first but not the second.

This generates a testable prediction (SLEEP-1): a Faraday-shielded LED lamp that blocks IF emissions while preserving identical light spectrum should produce less biological disruption than an unshielded lamp. If shielding makes no difference, the IF channel hypothesis is weakened.

Iversen 2025 ↗iMagnetically Stimulated Myogenesis Recruits a CRY2-TRPC1 Photosensitive Signaling AxisCells 14(3):231 · 2025 · journalMetadata matched identifies a CRY2/RFK/FAD/TRPC1-dependent response to a 1.5 mT pulsed field in myoblasts. This expands the candidate receiving machinery. Counting its downstream functions does not determine a 25% contribution to reproductive effects; existing pathway weights remain scenario parameters.

Use sham and field exposure at every intervention level: no blocker, L-type blockade, TRPC1 intervention and both. Record the first calcium response, ER-store change, later repair and tissue function. L-type blockade does not remove every VGCC subtype, and the remaining response is not automatically TRPC1-mediated.

Genetic rescue, a downstream bypass and timed intervention distinguish a receiver from a mediator or repair enabler. The stronger prediction is that the same measured receiving state explains the intervention contrast in a specified gonadal system.

The current reanalysis starts from 530 unique genotoxicity publications, not 530 fertility experiments. After excluding 19 ambiguous experimental outcome classifications, the main experimental analysis has 460 publications; 130 meet the quality screen. Studies sharing a donor group are not independent replications.

The GSM–UMTS difference concerns reported DNA-effect classifications: 73/119 versus 7/30, with an adjusted UMTS/GSM OR of 0.205. The quality-restricted interval includes 1. An exploratory broader biological-outcome analysis did not preserve that ordering. A technology label therefore cannot serve as a universal biological effect coefficient.

Background reporting was recorded in 46/460 publications. Missing reporting identifies a measurement gap, not a measured contaminated control. The analysis did not find a general historical attenuation of DNA-effect reporting. Publication year cannot substitute for cell lineage, passage, differentiation or actual prior fields.

BERM now tests moderators as measured receiving states: waveform and local vector fields; receptor/ER/redox state; light order and biological phase; repair and prior exposure. Iversen 2025 ↗iMagnetically Stimulated Myogenesis Recruits a CRY2-TRPC1 Photosensitive Signaling AxisCells 14(3):231 · 2025 · journalMetadata matched found weaker PEMF response after the stated dark pretreatment, so “less light always increases CRY sensitivity” is not a general rule.

Compare competing predictions before examining the outcome. Saturation predicts impaired absolute baseline with a reduced increment; protective adaptation predicts better challenge tolerance; sensitization predicts a larger increment; selection changes the response distribution. Sannino 2024 ↗iProtective effect of radiofrequency exposure against menadione-induced oxidative DNA damage in human neuroblastoma cells: The role of exposure duration and investigation on key molecular targetsBioelectromagnetics 45(8):365-374 · 2024 · journalMetadata matched and Sannino 2022 ↗iInhibition of Autophagy Negates Radiofrequency-Induced Adaptive Response in SH-SY5Y Neuroblastoma CellsInternational Journal of Molecular Sciences 23(15):8414 · 2022 · journalMetadata matched provide repair-linked tests of history, while Luukkonen 2009 ↗iEnhancement of chemically induced reactive oxygen species production and DNA damage in human SH-SY5Y neuroblastoma cells by 872 MHz radiofrequency radiationMutation Research 662:54–58 · 2009 · journalMetadata matched supplies a demanding CW/GSM contrast.

No. The direct evidence establishes a bounded proposition: a verbal explanation can be coherent while the speaker lacks access to all causal antecedents (Delgado 1969iPhysical Control of the Mind: Toward a Psychocivilized SocietyHarper & Row · 1969 · bookRegistered identifier; Gazzaniga 2000iCerebral specialization and interhemispheric communication: Does the corpus callosum enable the human condition?Brain · 2000 · reviewRegistered identifier; Nisbett & Wilson 1977iTelling more than we can know: Verbal reports on mental processesPsychological Review · 1977 · reviewRegistered identifier). It does not establish that a particular fertility explanation is biologically generated. [E]

BERM adds a composed and falsifiable application. It predicts that preregistered biological states will improve temporal and out-of-sample prediction of later behaviour and later stated reasons beyond prior reports and socioeconomic variables. It also predicts a framing-dependent evidence threshold in matched-review experiments and an ordered biological → behavioural → report response to a valid upstream intervention. [L*]

Disagreement with BERM is not evidence of cognitive immunology. If those discriminating predictions fail in adequately powered independent replications, the fertility application and Epistapege extension are rejected.