The Hardest Parts of Learning English: A Research-Backed Guide to ESL/EFL Difficulties

TL;DR

English is hard for learners across every linguistic level at once: an opaque spelling system, an unusually large vowel and vocabulary inventory, a tense/aspect and article system that maps poorly onto many other languages, and a stress-timed rhythm that compresses speech — plus psychological and cultural barriers that slow progress even when the grammar is “known.”
The single biggest predictor of which difficulties hit hardest is the learner’s first language (L1): article-less L1s (Russian, Chinese, Japanese, Korean, Polish) struggle most with a/an/the; syllable-timed L1s (Spanish, French, Italian) struggle with rhythm and reduction; L1s lacking /θ/ and /ð/ struggle with “th.”
The most authoritative consensus across applied linguistics, SLA, and ELT sources is that the hardest features are not the “rules” but the probabilistic, context-dependent, exception-riddled systems: articles, prepositions, phrasal verbs, collocations, the present perfect, connected speech, and pragmatic appropriateness — areas where there is no reliable rule to memorize.

Key Findings

Pronunciation and phonology are the difficulties learners most often cannot self-correct, and where breakdowns in being understood are most severe.
Spelling/orthography is genuinely irregular for historical reasons — English is one of the least transparent (“deepest”) alphabetic writing systems.
Grammar difficulties cluster around features that are probabilistic or absent in the L1: articles, the present perfect, prepositions, phrasal verbs.
Vocabulary is hard because of sheer size, polysemy, collocation, false friends, and the Germanic/Latinate layering that creates near-synonyms with different registers.
Listening and speaking are hard because of connected speech, accent variety, and the cognitive load of real-time processing.
Psychological and sociolinguistic factors — anxiety, fossilization, plateau, pragmatic failure — determine whether learners keep improving or stall.

The “th” sounds /θ/ and /ð/. The voiceless /θ/ (think) and voiced /ð/ (this) are interdental fricatives absent from most of the world’s languages, including French, German, Japanese, Mandarin, and many others. Learners substitute /t/, /d/, /s/, /z/, or /f/ depending on their L1. The sound is mechanically simple (tongue tip between the teeth, blow air) but consistent use takes weeks, and the same digraph “th” maps to two different phonemes plus /t/ in some loanwords, compounding the difficulty.

The /r/–/l/ distinction. Famously hard for Japanese, Korean, and Chinese learners, whose languages do not contrast these as separate phonemes; “rice” becomes “lice,” “very” becomes “wewy.” English also has two allophones of /l/ — “clear l” (leaf) and “dark l” (feel) — adding a further layer.

The large vowel inventory. English has roughly 20 vowel phonemes (around 12–20 depending on dialect and analysis), versus just five vowel phonemes in Spanish, Japanese, and many other languages. Learners from five-vowel systems cannot hear or produce contrasts like /iː/ vs /ɪ/ (sheep vs ship) or /æ/ vs /ɛ/ (cat vs bet), causing both comprehension and production errors. This is transfer, not deficiency: the L1 simply lacks the categories.

Short vs long vowels and the schwa. Length and quality distinctions (book vs moon) are hard to perceive. The schwa /ə/ — English’s most frequent vowel — appears in virtually every unstressed syllable (banana /bəˈnænə/, computer /kəmˈpjuːtər/). Learners who pronounce every vowel “as written” sound over-articulated and unnatural, because they miss the vowel reduction that drives English rhythm.

Word stress (lexical stress). English uses stress contrastively: REcord (noun) vs reCORD (verb), DESert vs deSERT. Stress placement is only partly rule-governed and must often be memorized per word. Misplaced stress is a frequent cause of being misunderstood — listeners identify words by stress pattern as well as by segments.

Sentence stress, rhythm, and the stress-timed nature of English. English is stress-timed: stressed syllables recur at roughly regular intervals and unstressed syllables are compressed between them. Speakers of syllable-timed languages (Spanish, French, Italian, Mandarin, Hindi, Korean) give every syllable equal weight, which sounds “flat” or “staccato” and, more importantly, makes their own listening comprehension harder because they don’t expect the compression. This distinction was first systematically described by Kenneth Pike in 1945.

Connected speech: linking, elision, assimilation, weak forms. In natural speech, words run together and sounds change or disappear: “next week” → “nex’ week” (elision), “handbag” → “hambag” (assimilation), “ten people” → “tem people,” and function words reduce to weak forms (“of” → /əv/, “and” → /ən/). Research finds assimilation is the most difficult connected-speech feature for learners to handle, and that recognizing elided forms is weakly correlated with listening proficiency. Learners do not need to produce all of these, but they must recognize them to understand native speech.

Intonation. Pitch movement carries grammatical and attitudinal meaning (questions, sarcasm, politeness, contrast). Learners from languages with flatter or differently-patterned intonation (e.g., Spanish has fewer intonation variations than English) may fail to convey or perceive these meanings, leading to misunderstanding even when words are correct.

Spelling–pronunciation mismatch. Because the same spelling can be pronounced many ways and the same sound spelled many ways (see §2), learners cannot reliably pronounce a new written word or spell a new spoken one.

2. Spelling & Orthography

English is a deep (opaque) orthography — one of the least transparent of any alphabetic language. It violates the ideal “one symbol = one sound” principle more than almost any other language. The classic example is the trigraph “ough,” which represents at least seven distinct pronunciations (through /uː/, though /oʊ/, thought /ɔː/, cough /ɒf/, bough /aʊ/, rough /ʌf/, hiccough /ʌp/).

Why it is so irregular — historical reasons:

Layered borrowing. English absorbed vocabulary from Old English (Germanic), Old Norse, Norman French, Latin, and Greek, importing each source’s spelling conventions.
The printing press froze spelling around 1475–1630, locking in spellings just as the Great Vowel Shift and other sound changes continued, creating permanent mismatches between writing and speech.
Dutch/Flemish printers introduced foreign spelling patterns (the silent “h” in ghost, aghast, ghastly).
Etymological “corrections” by Renaissance scholars added silent letters to show Latin/Greek roots: the “b” in debt and doubt, the “s” in island (mistakenly, since “island” is actually Old English).
No regulating academy. Unlike Spanish (Real Academia Española), French (Académie française), or German, English never had an authority to rationalize spelling.

Consequences for learners: silent letters (knight, subtle, honest, Wednesday), the same sound spelled many ways (the /iː/ in see, sea, receive, key, machine), the same letters pronounced many ways (the “c” in cat /k/ vs city /s/), and homophones (flour/flower, hear/here) and homographs (lead the metal vs lead the verb).

3. Grammar

The article system (a/an/the/zero) — the single hardest grammatical feature for many learners. Articles are notoriously difficult, especially for speakers of article-less languages (Russian, Chinese, Japanese, Korean, Polish). Peter Master, the leading authority on the topic, calls article selection “a universally acknowledged difficulty for nonnative speakers of English” (Master 1990, TESOL Quarterly 24(3)) and identifies two compounding factors: articles are the most frequent function words in English — “the” is the single most frequent word in written English, accounting for roughly 6–7% of all words in most texts — yet they are unstressed and phonologically reduced, making them hard to perceive (Master 2002, System 30(3)).

The difficulty is structural and semantic, not merely a matter of rules. Ionin, Ko & Wexler (2004, Language Acquisition 12(1)) showed in their Fluctuation Hypothesis that learners from article-less L1s “fluctuate between the meanings of definiteness and specificity” — because English grammaticalizes definiteness (the/a) while some languages grammaticalize specificity. Their data on L1-Russian and L1-Korean learners found errors were systematic: “overuse of the in specific indefinite contexts, and overuse of a in non-specific definite contexts.” Liu & Gleason (2002, Studies in Second Language Acquisition 24(1)) further showed that the four nongeneric uses of “the” (cultural, situational, structural, textual) are acquired at different rates.

Quantitatively, articles are acquired late and never fully mastered by many. In the classic morpheme-order studies (Dulay & Burt 1973/74; Bailey, Madden & Krashen 1974), articles fall in the middle of the acquisition sequence, but for article-less L1s they lag far behind. Corpus data on Japanese learners (Longman Learners’ Corpus) found articles had the lowest accuracy of all eight grammatical morphemes at every proficiency level — roughly 60–80% accuracy, never reaching the conventional 80–90% “acquisition” threshold even among proficient learners, while morphemes like copula and progressive -ing did. Even highly advanced learners typically reach only about a 90% success rate on forced-choice article tasks. Because article choice depends on discourse, shared speaker–hearer knowledge, countability, number, definiteness, specificity, and genericity simultaneously, no fixed rule set captures it — some researchers (Butler 2002) have argued the system is so resistant to instruction that learners “use articles almost randomly.”

The tense and aspect system, especially the present perfect. English has a rich tense/aspect system. The present perfect (I have lived here for ten years) is widely cited as the hardest single tense, because it expresses a relationship between past events and present relevance — a “reference time” concept that many languages (Chinese, Urdu, Bengali) do not grammaticalize. Learners whose L1 lacks perfect or perfect-continuous forms translate them as simple past or simple present, producing characteristic errors. The choice of form (simple vs continuous vs perfect) depends on subtle factors and speaker intention rather than rigid rules, and the present perfect overlaps heavily with the present perfect continuous.

Prepositions. English has around 60–70 prepositions — more than most languages — and their use is largely arbitrary and idiomatic rather than logical (arrive at a place but in a country; on Monday but in July but at 3 o’clock). Their meanings extend from concrete spatial senses to abstract ones with no L1 equivalent, and a single L1 preposition (Spanish “por”) may map to several English ones. They are a top source of L1-transfer errors and are easily fossilized.

Phrasal verbs. Combinations of verb + particle (give up, put off, look after, come up with) whose meaning is often non-compositional (you cannot deduce “give up” = quit from “give” and “up”). They are also grammatically complex: separable vs inseparable, and pronoun placement (pick it up, not pick up it). The same particle (“up”) carries many meanings, and the same verb (“put”) forms many unrelated phrasal verbs. Learners from languages without phrasal verbs often avoid them, sounding less natural.

Other persistently hard areas:

Conditionals and modals. The conditional system (zero, first, second, third, mixed) and the subtle, overlapping meanings of modal verbs (must/have to, may/might/could for probability, should/ought to) express degrees of probability, obligation, and hypotheticality that rarely map cleanly onto the L1.
Countable vs uncountable nouns. Whether a noun is countable determines article, quantifier (many vs much), and verb agreement. The category is partly arbitrary and L1-dependent: nouns uncountable in English (information, advice, furniture, luggage, news) are countable in many languages, producing errors like “informations” and “many furnitures.”
Word order. English is relatively fixed Subject-Verb-Object, which constrains learners from freer-word-order L1s.
Gerunds vs infinitives (enjoy doing vs want to do), the subjunctive, subject–verb agreement, and reported speech (backshifting tenses) are all commonly cited.

4. Vocabulary & Lexis

Sheer size. English has one of the largest vocabularies of any language. As of January 2026 the Oxford English Dictionary contained 520,779 entries and 888,251 meanings; the 20-volume Second Edition lists 171,476 words in current use (plus 47,156 obsolete words); and total estimates run toward a million when scientific and technical terms are included. This is a consequence of the layered borrowing history. The encouraging counterpoint from vocabulary research (Nation & Waring 1997; Webb & Rodgers): the first 1,000 word families cover roughly 75% of written English and ~80% of speech, the first 2,000 reach about 87% of speech, and the most frequent 3,000 word families are enough to understand about 95% of movies and TV — so learners need a manageable core, not the whole dictionary.

Germanic vs Latinate layers and near-synonyms. Because of the Norman Conquest and later Latin/Greek borrowing, English often has two or three words for one concept (ask/inquire/interrogate; kingly/royal/regal) with different registers and connotations. Choosing the right one for the context is a subtle, advanced skill.

Collocations. Which words “go together” is often unpredictable: “heavy rain” not “strong rain,” “make friends” not “get friends,” “do homework” but “make the bed.” Granger’s research showed L2 learners have a “misguided sense of collocation” (producing “blissfully ignorant” correctly but generalizing wrongly). Collocational knowledge is one of the main things separating intermediate from advanced learners.

Polysemy and homonymy. Common words carry enormous numbers of meanings: lexicographer Peter Gilliver calculated that the verb “run” alone has 645 distinct senses (taking the OED record from “put” in 2007), and the verb “set” had 580 senses requiring some 60,000 words to describe in OED2. Learners may “know” a word but misinterpret it in a specific context.

False friends. Words that look/sound like an L1 word but mean something different (Spanish “embarazada” = pregnant, not embarrassed; “actually” ≠ “actualmente”) cause confident errors, and partial false friends (deceptive only in some contexts) are especially treacherous.

Idioms, slang, register, and connotation. Idioms (“bite the bullet,” “break the ice”) are non-literal and culturally loaded; slang and colloquialisms shift quickly and vary by region; and learners must learn to shift between formal and informal register appropriately.

5. Listening & Speaking

Understanding fast natural speech. Beyond connected speech (§1), the speed and density of native speech overload learners’ real-time processing. They may know every word on the page but fail to segment them in the stream of speech.

Accent and dialect variety. Learners are usually trained on General British or General American English but must cope with a wide range of “Englishes” — Australian, Indian, Irish, Scottish, Singaporean, Nigerian, plus regional accents within each. Research shows learners (especially East Asian learners trained on prestige accents) find unfamiliar varieties significantly harder to understand and often hold negative attitudes toward non-native accents, which reduces both comprehension and their own confidence.

Fluency, automaticity, and “thinking in English.” Early learners translate mentally from the L1, which is slow and error-prone. Building automaticity — retrieving words and structures without conscious effort — is the central challenge of becoming fluent, and requires large amounts of practice that classroom settings often cannot provide.

6. Writing

Register and tone; punctuation. Learners must master conventions of formality, plus English punctuation rules (comma usage, etc.) that differ from L1 conventions.

Paragraph and essay structure; coherence and cohesion. English academic writing favors a direct, linear, thesis-first structure with explicit topic sentences and cohesive devices. Robert Kaplan’s contrastive rhetoric (1966) argued that “each language and each culture has a paragraph order unique to itself” and that L1 rhetorical patterns transfer into and interfere with L2 writing. Coherence (logical organization of ideas) and cohesion (the linguistic ties — conjunctions, reference, lexical repetition — that bind a text, per Halliday & Hasan 1976) are repeatedly identified as the weakest aspects of ESL academic writing.

7. Sociolinguistic & Pragmatic Issues

Pragmatic competence — using language appropriately in context — is widely held to be one of the hardest things to acquire, because socio-pragmatic rules are rarely taught explicitly and are hard to learn without immersion. Misunderstandings in cross-cultural communication “often arise not from grammar or vocabulary errors, but from pragmatic failures.”

Specific challenges include:

Politeness and indirectness. Knowing that “Could you possibly…?” is more appropriate than “Give me…”; how much to soften requests, refusals, complaints, and apologies. Norms differ sharply by culture (the relative directness or indirectness valued in the L1 transfers into English), and speech acts like refusals are especially hard because they require mastery of target-culture norms.
Knowing what is appropriate in a given setting (formal vs casual; email conventions — students who omit greetings/closings are perceived as impolite).
Cultural references and humor, which depend on shared background knowledge.

8. Psychological & Motivational Factors

Foreign Language Anxiety (FLA). Defined by Horwitz, Horwitz & Cope (1986), who developed the Foreign Language Classroom Anxiety Scale (FLCAS), FLA is a distinct anxiety tied to language learning, comprising communication apprehension, fear of negative evaluation, and test anxiety. It is robustly negatively correlated with willingness to communicate and with achievement — anxious learners speak less, withdraw from oral activities, and “get so nervous they forget things they know.”

Fear of making mistakes feeds anxiety and reduces the speaking practice learners need.

Fossilization and the plateau effect. Larry Selinker (1972) coined “interlanguage” for the learner’s evolving, idiosyncratic system, and “fossilization” for the point at which it stops developing — errors become permanently established despite further exposure and instruction. Selinker estimated that about 95% of L2 learners fail to reach native-like competence. The related “plateau effect” describes the common stalling of progress at intermediate-to-advanced levels.

First-language interference (L1 transfer). The unifying thread of this entire report: difficulties are largely predictable from the L1. Where the L1 and English differ — in sounds, rhythm, articles, tense, word order, or pragmatic norms — learners transfer L1 patterns, producing systematic errors. This is why pronunciation, grammar, and even rhetorical problems differ so consistently by native-language background.

Lack of speaking opportunities in EFL (as opposed to ESL) contexts, where English is studied but not used in daily life, limits the practice needed to build fluency and overcome the other difficulties.

Recommendations

For learners — staged priorities:

Beginner: Prioritize high-frequency vocabulary (the first 1,000–2,000 word families, which cover ~80–87% of speech) and the most communicatively important sounds and contrasts for your L1. Don’t aim to eliminate your accent — aim for intelligibility. Benchmark to advance: you can understand most everyday speech and make yourself understood.
Intermediate: Tackle the probabilistic systems through exposure, not rules — learn vocabulary in chunks/collocations, learn phrasal verbs and prepositions in context, and do focused listening on connected speech (transcribe short clips, slow audio to 0.75×). Start consciously using the present perfect and articles. Benchmark: ~3,000 word families (enough for ~95% of films/TV).
Advanced: Target the things that separate you from native-like use — collocation, register, polysemy, pragmatic appropriateness, and academic writing structure. Actively seek out varied accents. Benchmark: ~8,000+ words; comfortable independent reading.

To avoid fossilization and plateau: seek corrective feedback (a teacher, tutor, or AI tool), record and analyze your own speech, and keep pushing into uncomfortable new contexts rather than recycling what you already know. The threshold that should change your strategy: if the same errors recur despite exposure, switch from passive input to focused, corrected output practice.

To manage anxiety: low-stakes speaking practice (with AI partners, tutors, or sympathetic peers) builds willingness to communicate; reframe mistakes as necessary and accent as identity, not failure.

For teachers: anticipate L1-specific difficulties (contrastive analysis); teach pronunciation, pragmatics, and collocation explicitly and communicatively, not just grammar; expose learners to multiple English varieties; and prioritize recognition of connected speech for listening.

Caveats

Difficulty is L1-relative. Almost every “hard feature” listed is hard for some learners more than others depending on their first language; there is no single universal difficulty ranking.
Some claims are contested. The strict “natural order” of morpheme acquisition has been challenged by L1-dependent corpus findings; “fossilization” is debated (some researchers question whether it is truly permanent); and Kaplan’s contrastive rhetoric has been criticized as overly deterministic about L1 thought patterns. The 95%-fail-to-reach-native-competence figure originates with Selinker and is an estimate, not a precise measurement.
Source quality varies. Many illustrative examples come from teaching blogs and commercial ELT sites, useful for showing common difficulties but not peer-reviewed; the strongest claims here (articles, anxiety, fossilization, contrastive rhetoric, connected speech, vocabulary coverage) rest on named academic sources (Master; Ionin/Ko/Wexler; Liu & Gleason; Horwitz et al.; Selinker; Kaplan; Halliday & Hasan; Nation & Waring).
Vocabulary-size and dictionary figures vary by counting method (what counts as a “word”), so treat them as orders of magnitude, not exact counts.
This report describes difficulties, not impossibilities — every one of these is routinely overcome by motivated learners with sufficient, well-directed practice.