Research

The Irregularities of English: A Catalogue of What Must Be Memorized

TL;DR

• This document catalogues the systematic irregularities of English — the places where the language forces you to memorize a list instead of applying a rule. It is the what-and-how-much companion to PAIN-POINTS, which covers the broader why-it’s-hard (psychology, listening, pragmatics, L1 transfer). Where they overlap, this file links rather than repeats.
• An irregularity here means: knowing the rule does not let you produce the form. You must be told that the past of go is went, that the plural of child is children, that ough is pronounced eight different ways. None of it is derivable.
• The damage is concentrated in a handful of systems: spelling↔sound, verb conjugation, noun plurals, comparatives, pronoun case, articles, prepositions, and question/negation formation. Each chapter below ends with a World English target — a pointer to the docs/ specification that regularizes it.
• Scale, at a glance: ~200 irregular verbs in everyday use (400–600 counting variants), ~20 vowel phonemes written with 5 vowel letters, ough with 8–9 sounds, dozens of irregular plurals, and a closed but heavily-used set of suppletive comparatives and pronouns.

How to read this file. Each chapter states what is irregular, gives concrete examples, estimates the scale of the problem, and names the World English target that fixes it. The fixes themselves live in ../docs/.

1. Spelling & Orthography

English is a deep (opaque) orthography: the written form is a poor predictor of the spoken form, and vice-versa. This is historical residue — layered borrowing from Old English, Norse, French, Latin, and Greek; the printing press freezing spellings around 1475–1630 just as pronunciation kept shifting; etymological “corrections” that inserted silent letters; and the absence of any regulating academy (see PAIN-POINTS.md §2).

The irregularities:

• One sound, many spellings. The vowel /iː/ is spelled at least eight ways: see, sea, scene, receive, key, machine, field, people.
• One spelling, many sounds. The digraph ough has 8–9 pronunciations with no governing rule: through /uː/, though /oʊ/, thought /ɔː/, cough /ɒf/, rough /ʌf/, bough /aʊ/, thorough /ə/, hiccough /ʌp/, lough /ɒx/. The letter c is /k/ in cat but /s/ in city; g is /g/ in get but /dʒ/ in gem.
• Silent letters. knight, gnome, write, comb, debt, doubt, subtle, island, honest, Wednesday, colonel, receipt. Several were never pronounced — they were added to show Latin/Greek roots (the b in debt), and one (island) is a plain mistake.
• Unpredictable consonant doubling. travelling vs traveling, referred vs offered; fulfil vs fulfill. The rule (double after a stressed short vowel) has many exceptions and splits by dialect.
• British/American divergence. colour/color, centre/center, realise/realize, catalogue/catalog, defence/defense, travelled/traveled. A learner must pick a camp and still read the other.
• Homophones & homographs. flour/flower, hear/here, their/there/they’re sound alike but differ in writing; lead (metal) vs lead (verb), read (present) vs read (past) are written alike but differ in sound.

Scale. English is routinely ranked among the least phonetically transparent of all alphabetic writing systems. Roughly 25% of common words contain a spelling that cannot be predicted from their sound.

World English target → docs/orthography (light, legibility-preserving regularization) and docs/pronunciation (the grapheme↔phoneme mapping that makes spelling predict sound).

2. Pronunciation

Even setting spelling aside, the spoken system has irregularities a learner cannot derive.

• Unpredictable word stress. Stress is contrastive and must often be memorized per word: REcord (noun) vs reCORD (verb), DESert vs deSERT, PREsent vs preSENT. There is no reliable rule for where stress falls.
• The schwa and vowel reduction. /ə/ is the most frequent vowel in English and appears in nearly every unstressed syllable (banana /bəˈnænə/, computer /kəmˈpjuːtər/), yet it is never written as itself. The same letter a is a full vowel in cat and a schwa in about.
• A large vowel inventory. ~20 vowel phonemes written with five vowel letters, forcing contrasts (sheep/ship, cat/cot/caught) that the spelling does not mark.
• th is two sounds. The single digraph spells both /θ/ (think) and /ð/ (this), and the choice is not predictable from spelling.
• Connected speech. Linking, elision, assimilation, and weak forms reshape words in running speech (next week → nex’ week; of → /əv/), so the careful “dictionary” pronunciation is not what is heard.

Scale. Pronunciation is the area learners can least self-correct (see PAIN-POINTS.md §1); stress errors alone are a leading cause of being misunderstood.

World English target → docs/pronunciation (a regular sound↔spelling mapping, a learner respelling key, and stress that is always marked rather than moved to one predictable position — see P4, which deliberately keeps each word’s native stress instead of regularizing its placement, since a single fixed-position rule would make many words unrecognizable).

3. Verb Conjugation

The verb system is the densest pocket of pure memorization in English.

• Irregular past and past participle. go → went → gone, take → took → taken, bring → brought → brought, go → went (suppletion — a completely unrelated stem). The three “principal parts” follow no single pattern; Brehe’s Grammar Anatomy devotes Chapter 3 entirely to listing them.
• be has eight forms. be, am, is, are, was, were, being, been — the most irregular word in the language, and the most frequent.
• Third-person singular -s. I go / he goes. A single agreement ending that exists only in one cell of the present tense, with its own spelling sub-rules (go → goes, try → tries).
• Pattern families that look regular but aren’t. sing/sang/sung but bring/brought; speak/spoke but sneak/sneaked-or-snuck. Apparent patterns invite wrong guesses.

Scale. ~200 irregular verbs in everyday use (400–600 counting prefixed/variant forms), of which ~50 carry the bulk of daily usage — but those 50 are the most common verbs in the language, so there is no avoiding them.

World English target → docs/morphology (every verb takes regular -ed; the be paradigm and third-person -s are regularized).

4. Noun Plurals & Countability

Pluralization should be the simplest rule in any language. In English it is a rule plus a list of exceptions.

• Irregular plurals. child → children, foot → feet, tooth → teeth, mouse → mice, man → men, woman → women, person → people.
• Zero plurals. sheep → sheep, fish → fish, deer → deer, series → series.
• -f → -ves. leaf → leaves, wife → wives, knife → knives — but roof → roofs, chief → chiefs. The exceptions have exceptions.
• Imported Latin/Greek plurals. cactus → cacti, fungus → fungi, analysis → analyses, criterion → criteria, phenomenon → phenomena, index → indices.
• Uncountable nouns. information, advice, furniture, luggage, news, equipment take no plural and no a/an, even though their equivalents are countable in many languages — producing the classic errors informations, an advice, many furnitures.

Scale. A few dozen high-frequency irregular plurals plus an open-ended tail of borrowed forms, plus a fuzzy, partly-arbitrary countable/uncountable boundary that must be learned noun by noun.

World English target → docs/morphology (all nouns pluralize with -s/-es) and docs/grammar (the countable/uncountable distinction is regularized).

5. Adjectives & Adverbs

• Suppletive comparatives. good → better → best, bad → worse → worst, far → further/farther → furthest, little → less → least, many/much → more → most. The comparative is an unrelated word, not a derived form.
• The -er/-est vs more/most boundary. big → bigger but beautiful → more beautiful (not beautifuller). The cutoff depends on syllable count with a grey zone (clever → cleverer or more clever?) that even natives disagree on.
• Irregular adverb formation. Most adverbs add -ly (quick → quickly), but good → well (suppletive), fast → fast (zero), hard → hard (and hardly means something else entirely).

Scale. A small closed set of suppletive forms — but they are among the most frequent adjectives in the language (good, bad, much, many, little), so they cannot be skipped.

World English target → docs/morphology (a single regular comparative rule; regular adverb formation).

6. Pronouns

Pronouns are the one place modern English still inflects for grammatical case, irregularly.

• Case forms. I / me / my / mine, he / him / his, they / them / their / theirs, who / whom / whose. The forms are suppletive and must be memorized as a grid.
• who vs whom. A case distinction that confuses native speakers and is dying in speech, yet still tested in formal writing.
• Reflexives are inconsistent. myself, yourself, himself (possessive + self) but himself/themselves (object + self) — two different formation patterns in one paradigm.

Scale. A small closed class, but extremely high frequency, and the case grid is a common error source for speakers of caseless or differently-cased L1s.

World English target → docs/grammar (regularized pronoun case and who/whom, and a single object-based reflexive pattern).

7. Grammar Systems (Articles, Prepositions, Tense, Phrasal Verbs)

These are not single-word exceptions but whole systems that resist rule-statement — the hardest grammar for many learners (see PAIN-POINTS.md §3).

• Articles (a/an/the/zero). The single hardest grammatical feature for many learners. Choice depends simultaneously on definiteness, specificity, countability, number, and shared knowledge, with no fixed rule — the alone has four distinct nongeneric uses.
• Prepositions (~60–70, idiomatic). Largely arbitrary: arrive at a place but arrive in a country; on Monday but in July but at 3 o’clock. One L1 preposition often maps to several English ones and vice-versa.
• Tense/aspect overlap. A rich system whose forms overlap in use; the present perfect (I have lived here for ten years) is the most-cited single difficulty, because it encodes a past-with-present-relevance relationship many languages don’t grammaticalize.
• Phrasal verbs. give up, put off, look after, come up with — meaning is often non-compositional (give up ≠ give + up), and the grammar is irregular too (separable pick it up vs inseparable, with pronoun-placement rules).
• Modals (can/could/may/might/must/shall/should/will/would/ought to). Defective verbs (no past -ed, no -ing) whose meanings overlap on two axes — probability (may/might/could/must all grade certainty) and obligation (must/have to/should/ought to) — with no clean rule for which to pick, and suppletive past forms (can→could).
• Conditionals (zero / first / second / third / mixed). Five shapes graded by tense-backshift plus would: if it rains, I will go (real) vs if I had money, I would buy it (present-unreal, marked by a past tense) vs if I had known, I would have told you (past-unreal, marked by had + participle). The “remoteness via backshift” device is opaque, and the third conditional rides on the present-perfect machinery English encodes inconsistently.
• Gerund vs infinitive complements. Which verbal complement a verb takes is lexically fixed with no rule: enjoy doing (gerund only) but want to do (infinitive only), begin takes either, and a handful (stop, remember, try) change meaning with the choice (stop smoking = quit vs stop to smoke = pause in order to). Learned verb by verb.

Scale. Articles are acquired late and often never fully mastered (corpus accuracy ~60–80%); prepositions and phrasal verbs are open-ended, idiom-by-idiom learning with no rule to fall back on.

World English target → docs/grammar (a leaner tense/aspect set, a simplified article rule, regularized preposition choices, countable nouns, a single to-infinitive complement) and docs/style (prefer plain regular verbs over idiomatic phrasal verbs).

8. Sentence Structure

Some irregularity lives not in the words but in how sentences are assembled.

• Do-support. English questions and negatives require a “dummy” auxiliary do that carries no meaning: You like it → Do you like it? / I do not like it. No other everyday operation needs an empty helper word, and learners routinely produce You like it? / I not like it.
• Subject–auxiliary inversion. Questions invert subject and auxiliary (She is here → Is she here?), but only the auxiliary, and only sometimes — interacting awkwardly with do-support.
• Tag questions. You’re coming, aren’t you? / He left, didn’t he? The tag must mirror the auxiliary, polarity, and subject of the main clause — a small agreement computation for every tag.
• that vs which (and the comma). Restrictive that vs non-restrictive which, with a punctuation rule attached, is a distinction many natives ignore and few can state.

Scale. Do-support and inversion affect every question and negation in the language, so the cost is paid constantly even though the rule set is small.

World English target → docs/grammar (regular question and negation formation) and docs/style (fixed, unambiguous word order).

Catalogue → Fix map

Sources & references

• PAIN-POINTS — the project’s research-backed survey of learner difficulty (carries the underlying academic citations: Master; Ionin/Ko/Wexler; Liu & Gleason; Selinker; and others).
• Brehe’s Grammar Anatomy — the standard-English grammar baseline; Ch. 3 (irregular verbs), Ch. 6 (prepositions), Ch. 20 (nouns).
• Irregular-verb count (~200 in regular use; 400–600 with variants): English irregular verbs — Wikipedia.
• ough pronunciations (8–9): Ough (orthography) — Wikipedia.

Caveat. Counts are order-of-magnitude, not exact — they depend on how one defines “irregular” and what counts as a distinct word. Irregularity is also relative to the learner’s first language: a feature that is hard for one L1 may be easy for another. This catalogue describes structural irregularity, not impossibility — every item here is routinely mastered by motivated learners; the point of World English is to make that effort unnecessary.