Czech Education: Why Language Skills Shouldn’t Block Future Scientists & Engineers

A significant portion of the ability to learn a foreign language is independent of overall intelligence, raising questions about current educational practices that prioritize language proficiency as a universal requirement for academic advancement.

The issue came to light in a case study at Univerzita J. E. Purkyně (UJEP) in Ústí nad Labem, where a student named “Adam” excelled in a technically focused field of study but repeatedly failed a mandatory foreign language exam. Further assessment revealed he had specific learning differences impacting his language acquisition.

This experience, highlighted by UJEP’s counseling center, isn’t isolated. The university notes that students with additional support needs often struggle more with language requirements than with their core subject matter. This finding underscores a growing debate about the fairness and effectiveness of standardized language testing in higher education.

Neuroscientific research suggests that language aptitude is largely genetically determined, separate from general intelligence, and shaped by brain structures individuals cannot control. A person with an IQ of 130 may repeatedly fail language tests, while someone with an IQ below 60 can communicate fluently in 15 to 20 languages.

This raises a critical question: how many potentially brilliant scientists, engineers, and programmers are being filtered out of the educational system due to language barriers?

The relationship between intelligence and language learning has been a long-standing area of study in cognitive psychology. A key study conducted by Fred Genesee at McGill University in 1976 found that IQ correlated with reading and written tests, but not with listening comprehension or interpersonal communication. Students with below-average IQs in an immersion environment communicated as effectively as those with above-average IQs.

A 2016 meta-analysis by Shaofeng Li, encompassing 66 studies, established a correlation of r = 0.50 between language aptitude and intelligence. Which means that IQ explains approximately 25% of the variance in language aptitude, with the remaining 75% being independent of intelligence. For the Modern Language Aptitude Test (MLAT), the correlation is higher at r = 0.64, but still far from identical.

Neuroscientific research reinforces this dissociation at the brain structure level. Evelina Fedorenko at MIT identified a domain-specific fronto-temporal language network that is separate from a domain-general fronto-parietal network responsible for mathematics, logic, and working memory. These networks do not mutually activate. Patients with Williams syndrome exhibit low IQs but preserved language abilities, while those with specific language impairment have normal IQs but impaired language skills. This is known as a double dissociation, considered strong evidence for the separability of two cognitive systems.

Perhaps the most dramatic illustration is the case of “Christopher,” a linguistic savant studied by Neil Smith and Ianthi-Maria Tsimpli at University College London. Despite a nonverbal IQ in the range of 40–60, Christopher could communicate in 15 to 20 languages. He was unable to care for himself in daily life but could translate between a variety of typologically distinct languages, and his English skills indicated an IQ over 120. His case represents a striking double dissociation: language ability can exist almost independently of general intelligence.

Breakthroughs in understanding language talent came from functional magnetic resonance imaging (fMRI) studies from the EvLab project at MIT and Harvard. Jouravlev, Mineroff, Blank, and Fedorenko (2021) scanned 17 polyglots—nine of whom were hyperpolyglots proficient in 10 to 55 languages—and compared them to 217 control subjects. The results were surprising: polyglots showed less and less extensive activation in the language network when processing their native language. Their brains worked more efficiently, with lower neural activity costs per unit of language processing. This difference was functionally selective, manifesting only in the language network, not in networks for cognitive control or rest.

Follow-up studies by Malik-Moraleda et al. (2024) with 34 polyglots showed that activity in the language network increased with proficiency in a given language, with one exception: the native language evoked approximately 25% less activation than other well-mastered languages. This suggests the highest cognitive efficiency for the language acquired earliest.

Structural brain changes were documented in a pioneering study by Mechelli et al. (2004) in *Nature*: learning a second language increases gray matter density in the left inferior parietal cortex, with the degree of remodeling depending on the level achieved and age of onset. White matter, particularly the arcuate fasciculus connecting Broca’s and Wernicke’s areas, plays a key role. López-Barroso et al. (2013) in *PNAS* demonstrated that learning new words is mediated by the left arcuate fasciculus and that individual differences in its microstructural integrity explain differences in the ability to learn new words.

Working memory is another crucial factor. A meta-analysis by Linck et al. (2014) in *Psychonomic Bulletin &amp. Review*, including 79 samples and 3,707 participants, found a reliable positive correlation between working memory and second language performance (ρ = 0.255). Working memory is largely heritable, and its capacity varies significantly between individuals—independently of IQ.

Genetic research into language abilities began with the discovery of the FOXP2 gene in the KE family. Lai, Fisher, Hurst, Vargha-Khadem, and Monaco (2001) identified a point mutation in the FOXP2 gene on chromosome 7q31 that perfectly matched all fifteen affected family members with severe speech and language disorder. Enard et al. (2002) subsequently demonstrated that human FOXP2 protein differs at two amino acid positions from chimpanzee FOXP2 and that these changes were subject to positive selection during the last approximately 200,000 years of human evolution.

The CNTNAP2 gene on chromosome 7q35 is a direct regulatory target of FOXP2. Vernes et al. (2008) in the *New England Journal of Medicine* demonstrated that FOXP2 binds to CNTNAP2 and dramatically reduces its expression, and that multiple variants in CNTNAP2 are associated with specific language impairment. Whitehouse et al. (2011) confirmed that CNTNAP2 variants influence early language development in 1,149 children from the general population. Other significant genes include KIAA0319 and DCDC2 (neural migration, reading) and CMIP and ATP2C2 (phonological short-term memory).

Twin studies provide the strongest evidence for the heritability of language abilities. Rimfeld, Dale, and Plomin (2015) analyzed 6,263 pairs of twins in *Translational Psychiatry* and found heritability of 53–62% for different second languages. Crucially, the genetic influence is distributed: one-third is shared with intelligence, one-third is independent of intelligence and related to the native language, and one-third is unique to second language learning. Dale, Harlaar, Haworth, and Plomin (2010) found a heritability of 0.67 in an earlier study of 604 twin pairs. The largest genome-wide association study of language abilities (Eising et al., 2022, *PNAS*) estimated heritability at the single-nucleotide polymorphism level at 13–26% and found molecular links to the neural structure of language areas of the brain.

In other words, there are specific genes for the ability to learn foreign languages that are unrelated to IQ. A student may inherit exceptional spatial reasoning, analytical skills, and working memory for mathematics—and simultaneously inherit weak phonological memory, low phonemic coding ability, and unfavorable variants of the CNTNAP2 or CMIP genes.

Paradoxically, traits typically associated with highly intelligent individuals—carefulness and analytical thinking—can actively hinder achieving language fluency.

Ullman’s declarative/procedural model (2001, *Nature Reviews Neuroscience*) explains why. Mental lexicon relies on declarative memory (temporal lobe), while mental grammar relies on procedural memory (frontal cortex, basal ganglia). When learning a second language, adults initially rely more on declarative memory—conscious, rule-based learning. Fluency, however, requires a transition to the procedural system. Highly intelligent individuals may excel in the declarative phase (memorizing rules) but get stuck in it and never achieve automatic fluency.

Gregersen and Horwitz (2002, *Modern Language Journal*) found that anxious language learners exhibit significantly more conscientious traits—higher personal scales, more procrastination, greater fear of evaluation, and greater concern over mistakes. Flett et al. (2016) proposed a multifaceted model of conscientiousness in language anxiety: maladaptive conscientiousness (unrealistic standards coupled with fear of failure) creates a vicious cycle—fear of errors leads to avoidance of speaking, which leads to less practice and slower learning. Intelligent individuals with high standards are more susceptible to this cycle.

Guiora et al. (1972, 1980) introduced the concept of language ego permeability. The degree to which an individual is able to “permeate” their native language ego and adopt a new language identity correlates with better pronunciation and willingness to adopt a new sound system. Guiora’s experiments even demonstrated that alcohol and Valium improved pronunciation in a second language—interpreted as a reduction in defensive mechanisms of the ego. Highly intelligent, analytically oriented people may have “thick ego boundaries” that make them resistant to “sounding foolish” in a new language.

Approximately 60% of the world’s population uses two or more languages in daily life. According to available estimates, roughly 40% of people are monolingual, 43% are bilingual, 13% are trilingual, and around 3% speak four or more languages.

In Europe, according to Eurobarometer 540 (2024, 26,523 respondents in 27 EU member states), 59% of Europeans can converse in at least one foreign language, 28% in two, and 11% in three or more. Among young Europeans (15–24 years), this figure is 79% for one foreign language. The most linguistically proficient EU countries are Luxembourg, the Netherlands, Sweden, and Latvia (over 95%), while the least linguistically proficient are Hungary, Romania, and Poland (under 40%).

In Africa, multilingualism is the norm—estimated at 50% of the population, with many individuals speaking three to five languages. In the US, approximately 75% of the population is monolingual. There are approximately 7,168 living languages in the world, according to the Ethnologue database (28th edition, 2025), of which 44% are endangered.

A study by Hartshorne, Tenenbaum, and Pinker (2018, *Cognition*) with 669,498 participants—the largest study of its kind to date—brought a critical revision to the sensitive period hypothesis: the ability to learn a language remains high until age 17–18, much later than previously thought by Lenneberg in the 1960s. To achieve native-like grammar, however, one needs to start around age ten. After 17–18, there is a significant decline, and it is unclear whether the cause is biological or social.

The Czech education system places language requirements at every step. In elementary school, the first foreign language is compulsory from the 3rd grade (target level A2), and a second foreign language was added from the 2013/2014 school year, no later than the 8th grade (level A1). The Czech School Inspectorate stated in 2023 that approximately 60% of foreign language teachers in elementary schools would agree to abolish the compulsory second foreign language and that less than half of students would choose it voluntarily.

The new Framework Education Programme for Basic Education approved at the end of 2024 further increases this burden: English will be compulsory from the 1st grade with a target level of B1, and the second language will move to the 7th grade.

In secondary school, the foreign language maturity exam is not absolutely compulsory—students can choose mathematics instead of a foreign language as another subject in the common part. In practice, however, approximately 79% of graduates choose English. The English didactic test corresponds to level B1, lasts 110 minutes, and the passing threshold is 44%.

There is no nationwide legal requirement for foreign languages at the university level, but in reality, a foreign language (typically English at levels B1–B2) is a compulsory part of the study plans for the vast majority of faculties and a prerequisite for admission to the state final exam.

The overall failure rate for the English didactic test is relatively low—3.5% in the spring term of 2025 (approximately 2,100 students out of 60,800 taking the exam). However, it is necessary to distinguish the overall failure rate in English from the overall failure rate in the state maturity exam as a whole, where the differences between types of schools are vast: gymnasiums show 1–2%, secondary vocational schools 24%, and extension studies 37%. Failure rates for German on the didactic test are 22%, exceeding 40% in vocational schools. Technically oriented students in vocational schools are disproportionately affected by language requirements.

The system of adjusting the conditions of the maturity exam (PUP MZ) offers students with specific learning disabilities three levels of support: from simply increasing time by 25% (group 1) through formal test adjustments and 50% extra time (group 2) to content adjustments and doubling the time (group 3).

Critically, however, what PUP MZ does *not* allow is exemption from the exam or omission of content. CERMAT explicitly states that the granted accommodations in no way include exemption from certain tasks, easing of exams, or omission of certain content. Experts and parents consider group 1 insufficient—some students fail due to this and realize too late that they needed a higher level of accommodation.

At the elementary school level, the situation is seemingly more flexible: within the individual educational plan, the second foreign language can be replaced with other support. At maturity and at university, this option does not exist.

International comparisons show a wide range of approaches.

The United Kingdom represents the most liberal model: a foreign language is compulsory only until age 14, optional at GCSE (16), and no British university requires a foreign language for admission. Only about 45% of students choose a language at GCSE. UCL historically had the only language graduation requirement, but suspended it in 2021.

The United States offers the most developed system of alternatives for students with learning disabilities. Federal laws ADA and Section 504 require reasonable accommodations. Universities like Columbia, University of Washington, and University of Georgia allow students with documented learning disabilities to meet the language requirement with substitute courses—linguistics, foreign culture, or international studies. Students still must complete the same number of credits, but do not have to demonstrate language proficiency. Some schools additionally accept American Sign Language as a full substitute.

Germany offers alternative educational pathways: vocational maturity (Fachabitur) requires only one foreign language and allows study at applied universities, while classical maturity (Abitur) requires two languages. The system of compensation for disadvantages (Nachteilsausgleich) provides adjustments to exam conditions, but not exemption from the subject.

Finland has some of the strictest requirements (two compulsory languages in addition to the mother tongue), but addresses student difficulties with a three-tiered support system rather than exceptions.

Ireland is unique in offering a formal system of exemptions from Irish for students with dyslexia—students must score at or below the 10th percentile on a reading test.

One of the main arguments for compulsory language education—the cognitive benefits of bilingualism—rests on shaky ground. Meta-analyses in recent years have shown that the so-called “bilingual advantage” in executive functions approaches zero after correcting for publication bias. Lehtonen et al. (2018, *Psychological Bulletin*) found no strong or consistent evidence. Gunnerud et al. (2020) found an overall effect of only g = 0.06 in children. Paap (2024) labeled the entire hypothesis a prime example of the replicability crisis in cognitive science.

Bilingualism has undeniable communicative, cultural, and social benefits. Claims about cognitive benefits such as “improved working memory” or “delayed dementia,” however, are not reliably supported.

The economic benefits of language skills are real, but modest in the context of the problem. In the US, the wage premium for bilingualism is only 2–3% after controlling for cognitive abilities. In Germany, it is significantly higher—around 9%, up to 24% for advanced levels.

The argument that machine translation will gradually reduce the necessitate for language skills is supported by data. A study by CEPR (Frey and Llanos-Paredes, 2025) showed that areas with higher leverage of machine translation have seen a decline in demand for foreign language skills in the labor market.

Objections to easing language requirements have a valid core.

First, English is *de facto* the language of international science and technology. A scientist or engineer who cannot read professional literature in English is professionally disadvantaged. This objection is valid, but overlooks a key difference: the ability to understand professional text is a different skill than the ability to pass an English language test. Many technically proficient people read English manuals and forums daily without achieving B1 in spoken communication.

Second, language education has cultural and social value beyond economic benefit. Learning languages opens perspectives, builds empathy, and allows access to other cultures. This is true, but so are history of art, philosophy, or music—and these are not prerequisites for obtaining a degree in computer science.

Third, there is a concern about a “slippery slope”: if we allow exceptions from languages, where will it end? This objection has procedural merit, but the American and Irish systems show that clearly defined conditions for substitutions and exceptions have worked for decades without abuse.

Based on international practice and neuroscientific findings, specific reform proposals can be formulated.

A system of substitutions after the American model. Students with documented low language aptitude or specific learning disabilities could replace the language requirement with substitute courses—linguistics, cultural studies, or programming. This is not an exemption, but a substitution.

Restoring multi-level maturity tests. The Czech maturity exam had two levels of difficulty from 2011–2013. Returning to this model would allow differentiation: grammar school students would grab the higher level, vocational school students the lower. The current uniform level of B1 is trivial for grammar schools and unattainable for some vocational schools.

A formal system of exceptions after the Irish model. For students with diagnosed specific learning disorders affecting language learning, there should be the possibility of exemption conditional on a standardized diagnostic process.

Separating the language requirement from access to higher education. In technical and scientific fields, the language requirement should not be a condition for obtaining a degree, but an optional part of the study plan. Students should have the opportunity to demonstrate the ability to work with English-written professional literature in an alternative way.

Systematic data collection. The Czech Republic urgently needs research quantifying the impact of language requirements on the educational paths of students with different cognitive profiles. CERMAT should publish maturity results differentiated by students with PUP MZ, and MŠMT should track the reasons for incomplete studies in relation to language subjects.

The Czech educational system treats language proficiency as a universally required and uniformly testable skill. Neuroscientific research, however, shows that language aptitude is 53–67% heritable, independent of intelligence in one-third of cases, and determined by brain structures that individuals cannot control. Turker and Reiterer (2021) proposed the first neuro-cognitive model of language aptitude, according to which individual differences in the morphology of the auditory cortex—largely genetically and prenatally predetermined—limit individual neural plasticity for language learning.

A system that does not allow highly intelligent students with low language aptitude an alternative path to education is, in effect, selecting based on neurological profile, not on intellectual potential.

Among the students who repeatedly fail the English maturity exam or do not complete university due to a language exam, there may be future top scientists, engineers, and programmers—people whose brains are predisposed to other types of cognitive work than phonemic coding and memorizing grammatical patterns. But we don’t know. Because we don’t measure it.

The article is based on primary scientific sources (peer-reviewed studies and meta-analyses) and official data from Czech institutions (CERMAT, MŠMT, ČŠI, UJEP). A systematic study quantifying the “lost talent for science and technology fields” due to language barriers in the Czech academic environment was not found—this is an identified gap in the literature. Data on maturity failures are for the spring term of 2025. The bilingual advantage is presented with both sides of the debate. Reform proposals are derived from existing international practice, not hypothetical models.

[1] UJEP Counseling Center. Students with specific learning disabilities—basic terminology, impact on education, specific needs. Poradenske-centrum.ujep.cz.

[2] Genesee, F. (1976). The role of intelligence in second language learning. *Language Learning*, 26(2), 267–280.

[3] Li, S. (2016). The construct validity of language aptitude: A meta-analysis. *Studies in Second Language Acquisition*, 38(4), 801–842.

[4] Fedorenko, E. Et al. (2011). Functional specificity for high-level linguistic processing in the human brain. *PNAS*, 108(39), 16428–16433. See also: Amatuni, A., Diez, I., Kharas, N. Et al. (2024). Functional dissociation of the language network and other cognition in early childhood. PMC11184366.

[5] Smith, N. V., Tsimpli, I.-M. (1995). *The Mind of a Savant: Language Learning and Modularity*. Blackwell. Updated in Smith, N. V. Et al. (2011). *Signs of a Savant*. Cambridge University Press.

[6] Jouravlev, O., Mineroff, Z., Blank, I. A., Fedorenko, E. (2021). The little and efficient language network of polyglots and hyperpolyglots. *Cerebral Cortex*, 31(1), 62–76. PubMed: 32820332.

[7] Malik-Moraleda, S., Jouravlev, O. Et al. (2024). Functional characterization of the language network of polyglots and hyperpolyglots with precision fMRI. *Cerebral Cortex*, PMC9882290.

[8] Mechelli, A. Et al. (2004). Structural plasticity in the bilingual brain. *Nature*, 431, 757.

[9] López-Barroso, D. Et al. (2013). Word learning is mediated by the left arcuate fasciculus. *PNAS*, 110(32), 13168–13173.

[10] Linck, J. A. Et al. (2014). Working memory and second language comprehension and production: A meta-analysis. *Psychonomic Bulletin & Review*, 21(4), 861–883.

[11] Lai, C. S. L., Fisher, S. E., Hurst, J. A.,