Possibly the first person to publish a list of minimal pairs, homophones and homographs, certainly the earliest I have been able to find. He has about 140 items in his homograph list, including several that are no longer distinct such as champaign, chap, char and rarity.
Homographs are those words which have one spelling but two pronunciations and two distinct meanings or usages. A classic case would be a word like wound, which as a noun or present tense verb means injury or injure and with a different pronunciation is the past tense of the verb wind, itself a homograph. The term is contrasted with homophones, words with two spellings and two meanings but only one pronunciation such as fair/fare, and with homonyms, words with one spelling, one pronunciation, but two unrelated meanings, such as bear or just or left. The fact that the meanings are unrelated is what distinguishes homonyms from polysemes, words with varied meanings or usages, such as course or table or paper, where all the meanings can be traced back to the same root. English has an enormous number of polysemes, but only a relatively small set of true homonyms.
Not everybody uses this set of definitions, though they would be accepted by the majority of trained linguistic scholars. In wider usage (reflected in many dictionaries) the term homograph includes what I have here called homonyms and polysemes, i.e. words of different meaning but the same spelling and pronunciation, such as right and fly. For those people the term for what I am calling homographs is heteronyms, a term not much used by professional linguists. (Heterophone would be a much more appropriate label since what is different is the sound, not the name.)
Homographs are a minor problem for anyone learning English as a foreign language, but a much greater problem for anyone trying to design a foolproof text-to-speech algorithm for a computer. If one pronunciation is far commoner than the other (as with the word second, for instance) the programmer will probably ignore the exceptional case. Where both pronunciations occur frequently, as with object or read or row, the programmer must try to find contextual markers on which to base a rule.
The source of this list was the Roger Mitton machine-readable version of the 1974 Oxford Advanced Learners Dictionary, incorporating Mitton's 1991 additions. The dictionary contained 537 words which had more than one pronunciation listed. Some of these were simply words with varying pronunciations and no shift of meaning, such as breeches, dowsing, garage and piano or varying stress patterns such as bow-wow, bye-bye, and fricassee, and these were discarded. There were also four strong-form/weak-form pairs, a, an, to and 'cos (it is not clear why these were the only such pairs to emerge), three cases of abbreviations matching ordinary words, am (before noon), in (inch), and no (number), and one case of a loan word overlapping with an established English word, real (probably referring to the old Spanish coin rather than the football team).
The remaining 488 words plus about 140 more which were either not in the dictionary or not given two pronunciations in the dictionary have been classified into relevant groupings and are listed below. The spellings and phonetic transcriptions are mainly as they appear in the dictionary, though I have used Gimsonian IPA symbols rather than the Alvey transcription that Mitton had to use so that he could store his dictionary as a text file. There is also a complete alphabetical checklist of homographs which you can consult if looking for a particular word.
A number of double-stress words showed up. These are words whose pronunciation varies with their position in the phrase, front-stressed before a noun and end-stressed when final in the phrase, though without substantial change of meaning. (Compare "an overnight bag" with "Are you staying overnight?")
There are a number of other English words which behave in the same way, such as afternoon, bamboo, downhill, downstairs, inside, overseas, princess, routine, sardine, underground, upstairs, together with many compound adjectives (easy-going, home-made), all nationality adjectives ending in -ese, numbers from 13 to 99 (apart from multiples of 10), and many place names such as Bombay, Hong Kong, New York, and Torquay. In these other cases the dictionary did not record both stress patterns. Probably only the word overall (with its secondary meaning of an item of clothing) should be counted as a homograph, since in the other cases the change of pronunciation signals only a syntactic feature rather than a shift of meaning.
The next distinct group, which was by far the largest, was the set of nouns (or adjectives) with front-stress against verbs with end-stress with 295 words altogether. One suspects that in a good many cases the distinction is unnecessary for intelligibility; the set of -port words (export, import, transport) for instance are often heard with front stress even when being used as verbs, and I have heard on air the word increase stressed both ways as noun and both ways as verb. Similarly combat is always front-stressed as a noun, but may have either stress pattern when used as a verb. All these words are more likely to have stress on the second syllable when an inflection adds a third syllable, thus trans'porting, in'creasing or com'bating. In thirty-five cases (listed in the table below) there is a large difference in meaning and use between the two spoken forms while in other cases the difference is more syntactic than lexical.
In most cases any adjective senses ally themselves with the noun and exhibit front stress, but in one case, content, the adjective sense is end-stressed and relates more closely to the verb than to the noun. The adjective compact seems to occur both front-stressed and end-stressed with no change of meaning, although the noun is always front-stressed and the verb always end-stressed. Similarly, the noun complex is always front-stressed, though the adjective seems to vary freely between front stress and end stress. Perhaps the most memorable example of its occurrence with end stress is in the line from the Tom Lehrer song about Oedipus Rex, "You must have heard of his complex complex." (The force of the metre here tends to give both words end-stress.) All the other adjectives in the full list, absent, abstract, compound, converse, frequent, perfect, present, quadruple and second, were front-stressed.
Several problems occur with the prefix re-, for instance in the word recall where usage is inconsistent and is probably changing, and with refresh which is coming into the language as a front-stressed noun meaning a second cup of coffee in a restaurant. There are four other cases of a large shift of meaning between the front-stressed noun and end-stressed verb, namely record, recount, refuse and relay, and a further seventeen cases are in the long list where there is no significant difference in meaning. Other cases of re- verbs not involving a change of stress are listed in a separate section below.
There are four cases needing further comment.
The word entrance, while looking like a stress homograph, should perhaps be counted as a true homograph, since the noun sense derives from the verb enter while the verb sense derives from the noun trance.
The word deserts exists as two different nouns, one front-stressed meaning 'dry places', and the other end-stressed meaning 'what one deserves' and occurring usually in the fixed phrase 'get one's just deserts'. This second use has a homophone in the word desserts meaning 'sweet courses', which gives rise to many spelling errors and headline puns.
The word process exists as a noun with front-stress and as two different verbs, one with front-stress with a meaning linked to the noun process and one with end-stress with a meaning linked to the noun procession.
Similarly the word second exists as two separate verbs, one with front stress meaning to support a proposal at a formal meeting, and one with end stress meaning to send somebody away on temporary duty. The second use is fairly unusual, being mainly confined to military and civil service contexts.
The meanings of the thirty-five special cases are as follows:
prayer of the day
container for face-powder
occupying a small space
set of psychological symptoms
(1) substance combining chemical elements (2) enclosed group of buildings
not linear in progression
to make more complex
to direct an orchestra or choir to allow electrons to pass along a wire
what is contained
(to make) happy
to become smaller
to talk to another person
path between cliffs
to make dirty
to run away from
(1) dry places (front-stressed) (2) what one deserves (end-stressed)
runs away from
to give delight
piece of writing
to take advantage
to visit regularly
US trainee doctor
to be against
gift, time now
in this place
to hand over
(1) to modify (front-stressed) (2) to move in procession (end-stressed)
money earned by selling something
what is grown or made on a farm
to stick out
request to return faulty goods to a factory
(1) music on disc or a written log of events (2) best ever performance or result
to write down
second counting of votes in a close election (often spelled re-count)
to tell (a story, experience, etc)
second (free) cup of tea or coffee
to make something seem new
not to agree
race in which a baton is passed between team members
to pass information/lay a carpet again
part of a minute
number two in sequence
(1) to support a proposal at a formal meeting (front-stressed)
(2) to send away on temporary duty (end-stressed)
to force a person to accept
This is the full list of 295 words:
There were two interesting words which reversed the trend of this set, words where the front-stressed form was the (3rd person singular) verb and a form with stress later in the word was the (plural) noun:
Another large group was the set of words ending with -ate where the noun/adjective sense uses a schwa while the verb sense uses a full /eɪ/ diphthong. There were 45 of these (or 74 counting the inflectional variants). All of them retain the same stress pattern whether noun/adjective or verb except for alternate and consummate which, like analyses and diagnoses and unlike the other stress homographs, puts the stress at the front for the verb and later for the noun/adjective. I have omitted discriminate, which is very common as a verb but extremely rare as an adjective, although the negative of the adjective, indiscriminate, is common. One that is a relatively recent development (first attested 1909) is the verb sense of curate, meaning to be in charge of an art exhibition, a back-formation from curator. The noun sense, assistant to a minister of religion, is much older.
A similar but smaller group was the set of words ending with -ment where the noun sense uses a schwa while the verb sense uses a full vowel.
The dictionary listed five of these (ten including the inflectional variants), though in present day usage this set is no longer consistently differentiated;
most speakers nowadays will use a schwa for the verbs to implement or to supplement, and probably for the whole set.
A fifth set was that in which the noun/verb or adjective/verb distinction was made by voicing a final consonant. There were eighteen of these. One of them, the word close, exists as a verb with /z/, as an adjective with /s/, and as two different nouns, with /z/ meaning 'conclusion' and with /s/ meaning 'street with no exit'. In the form closer it may be a comparative adjective with /s/ or an agent noun with /z/. It is also worth noting the way used occurs with /s/ in the colligation 'used to' to make a past tense while the pronunciation with /z/ has the meaning 'employed'. In the nineteenth century a distinction was made between 'a mouse' with final /s/ and 'to mouse' (catch mice) with final /z/, but that is not heard nowadays.
Apart from the words beginning with re- which occur as front-stressed nouns or end-stressed verbs such as rebel, record, refill and reprint, listed under stress homographs above, there are a number of others which are not distinguished by main stress but by the presence of secondary stress on the prefix and a consequent change of pronunciation from /rɪ/ to /ri/. The first pronunciation tends to be associated with the meaning 'back' and the second with 'again'. The prefix re- with the second pronunciation often has the potential to enter homographs as one returns to earlier meanings of the base form. Thus we talk of reform /rɪ'fɔm/ in the church or the legal system, but a sculptor might reform /,ri'fɔm/ a clay model; you can make a remark /rɪ'mɑk/ in speech, while a teacher may remark /,ri'mɑk/ a contested exam paper. You may resent /rɪ'zent/ unfair criticism, while a garbled message may be resent/,ri'sent/. This distinction is sometimes signalled by including a hyphen in the spelling for the second meanings, re-form, re-mark and re-sent. Notice that the main stress does not move, and that where the base begins with <s> the pronuciation shifts between /z/ and /s/ as in resent, reserve, resign and resort. More of these forms may be added to the language, though those listed below are the only ones I know of so far that are well attested. There are two other special cases, the verb represent and the noun recreation in which the stress is on the prefix and the distinction is between the pronunciations /re/ and /ri/.
(get better after illness/put new covers on soft furniture)
Two more small groups could also be identified and separated. The first was the nine -ed adjectives with matching verb past tenses:
Not in the dictionary with both pronunciations but behaving similarly is the word beloved, which has three syllables as a noun or attributive adjective, but only two as a passive participle. ("I loved and was beloved again": Byron.) The adjective supposed has a rare archaic three-syllable pronunciation which is preserved in the derived adverb, supposedly. Winged was listed only as a one-syllable verb, but is also heard often as a two-syllable adjective in the phrase "Winged Victory". This dictionary listed wicked only as a two-syllable adjective, but the full OED also lists a one-syllable pronunciation, meaning "having a wick". Similarly this dictionary has rugged only as a two-syllable adjective, but the full OED has rug as a verb, which you might use in "We don't need another rug; the room is well rugged already."
The other group was the set of French loan words whose Anglicised plural is not represented in the spelling. Only five were recorded in the lists, but there are also many proper names, especially those associated with products, books or works of art, that behave in the same way, such as Beaujolais, Chablis, Charollais, Rabelais, Beaumarchais or Millais. There are also phrases such as faux pas and pince-nez which can be given a /z/ plural. A special case would be Grand Prix. Several such races might be given a /z/ plural, but a careful editor would write Grands Prix for the plural, making this not a homograph.
That left the following set of 119 homographs (179 if all the inflectional variants are counted) arising from a variety of causes. This section of the list includes those words like moped and wound which are most typical of what we have in mind when we think of homographs. These include a number which were not in the original dictionary list.
I have included several which contrast an ordinary word with a place or personal name: the words nice, angers, lens and tours all contrasting with French cities, liege with a Belgian city, trier with a German city, hue with a Vietnamese city, reading with an English or American town, abridge with an English village, scone with a Scottish village and palace, wear with an English river, tangier with a Moroccan port, mobile with a city in Alabama, natal with a South African province, rainier with a prince of Monaco, munch with a Norwegian artist, pears with an English singer, dieter with a German first-name, ravel with a French composer, mime with a character in Wagner's Ring cycle, and job with an Old-Testament prophet. Like august/August, and polish/Polish they are distinguished by capitalisation and would only be homographs in sentence-initial position or in all-upper-case writing. The word romance as a verb meaning to pay court to would most likely have the schwa in the first syllable, while romance meaning a piece of fiction could have either schwa or a full diphthong, and Romance as a description of languages derived from Latin would almost certainly have the full diphthong. The surname Brazil is usually front-stressed (as in Angela Brazil the author) while the name of the country is end-stressed. The girl's name Nancy is another homograph in contrast with a French city. Sue as an abbreviation for Susan is never palatalised, while the verb sue often is in British English.
The word dove is a homograph for speakers of Amercian English but not for the British for whom the past tense of dive is dived. The verb sundry (dry in the sun) is not well attested, though the participle sundried is common on food labels, sometimes with a hyphen, sun-dried. The adjective pasty (pale) is always so spelled, but the noun (a meat pie) has an alternative spelling of pastie, common in Ireland. I am also unsure how well the verb predate is attested in the meaning "devour prey", though it is common in its other meaning of "come before". In six cases of recent loan words, expose, charge, pate, rose, resume and attaches, the homograph is disambiguated by retaining the French accents in the English spelling.
The contrast between the two-syllable and three-syllable pronunciations of evening may not be consistently made by all RP speakers, though all would recognise it, I believe.
There are several cases involving Latin or Greek terms or names: ate versus the Goddess Ate, dives versus Dives (rich man), agape, manes and pace. Two cases, axes and bases, contrast a Latin plural with an English plural.
Many of these words are particularly popular with crossword setters since they allow for the creation of misleading cryptic clues. (See below for examples.)
(shorten/village in Essex)
(makes angry, city in France)
(noun, adjective as in 'arithmetic mean')
eɪt or et
(consumed, Goddess of mischief)
(diplomats/fastens together; not a homograph if one retains the accent)
(plural of axe/plural of axis)
(South African boss/makes a sheep-like noise)
(plural of base/plural of basis)
(kind of fish/low voice)
(army officer's servant/comic book hero)
(play violin/bend from the waist)
(archer/rower at bow of boat)
(English surname/South American country)
(a garment/with a larger bosom)
(rock dwelling/schoolboys' warning of an approaching master)
(attack/short for chargé d'affaires, temporary ambassador)
(silk measure, one who denies)
(one who eats little to lose weight, German first name)
(plunges into water, rich man in St Luke's gospel story)
(musical note/auxiliary verb)
(female deer/auxiliary verb)
(past of dive (US)/bird of peace)
(sliding container in a desk/one who draws a cheque, as in "refer to drawer")
One other word worth drawing attention to is quite. This has two distinct meanings, completely as in "You're quite right"
and to some extent as in "He's quite clever". Although both senses are pronounced the same at the segmental level, i.e. the same sequence
of vowels and consonants, they are consistently differentiated by intonation, the first sense occurring in phrases with a falling tone and the
second in phrases with a fall-rise. This makes it apparently the only example of a toneme distinction in English, i.e. a case where a distinction
in the meaning of a word is indicated by a tone difference. I would argue that this meets the definition of a homograph, namely two meanings,
two pronunciations, but only one spelling.
There is one homograph I know of (there may be more) which arises from abbreviating two different words.
Reg is pronounced /reʤ/ when it is short for Reginald
or registration, as in "a T-reg car". It is pronounced /reg/ when it is short
for regulation as in "Queen's Regs" (the British Army's rule book). Luckily the second form is almost always plural
and the first almost always singular, so there is little chance of confusion. There are at least five cases where a homograph arises
between a full word and an abbreviation. One is the homograph of path, which is
pronounced /pɑƟ/ by RP speakers in its ordinary meaning as a place to walk,
and /pæƟ/ when it is an abbreviation for pathology, as in "we are waiting
for the path reports"; another is the abbreviation Staffs /stæfs/
for Staffordshire against the word staffs /stɑfs/,
which would be homographs in upper-case writing; a third is the word coop meaning enclosure as opposed to Coop,
sometimes written Co-op or Coöp, the abbreviation for the Coöperative Stores; a fourth is the
word coax pronounced /kəʊks/ meaning 'persuade' versus the abbreviation
for coaxial cable coax pronounced /kəʊ'æks/, and a fifth is the word thou, pronounced /ðaʊ/ when it is the old fashioned word for 'you' and /Ɵaʊ/ when it is an abbreviation for 'thousand'. A possible sixth example
is brat, pronounced /bræt/ to mean an ill-behaved child,
and /brɑt/ as an abbreviation for Bratwurst in the USA. (I am grateful to Peter Reitan for sending me
recordings of TV commercials which attest this pronunciation.)
Another special case is place names. In England there are three towns called Gillingham. The one in Kent is pronounced /'ʤɪlɪŋəm/ while the ones in Dorset and Norfolk are pronounced /'gɪlɪŋəm/. Unfortunately only people who live locally to one of these towns (as I do) would be likely to know this and maintain the distinction consistently. There are four communities called Plaistow, pronounced /'pleɪstəʊ/ in Derby, /'plɑstəʊ/ in Kent or Essex, and /'plæstəʊ/ in Sussex. Alford is pronounced /'ɑfəd if you are referring to the Scottish one or /'ɔlfəd if you mean the English one. Lagos in Nigeria is pronounced /'leɪgɒs/ while Lagos in Portugal is usually / 'lɑgɒs/. A similar case is Berkeley in California, pronounced /'bɜklɪ/ (by RP speakers) or /'brklɪ/, contrasted with Berkeley in the west of England, pronounced /'bɑklɪ/. There is also the case of the Kentucky Derby, pronounced /'dɜrbɪ/, and the British horse race, the Derby at Epsom, pronounced /'dɑbɪ/. I suspect that one could find more cases of pairs of places in different countries with the same spelling and different pronunciations.
Return to contents
Example of cryptic crossword clues which draw on homographs, all taken from The Times:
Number of elks, perhaps, surrounding a wild animal's lair. (8) Answer: DEADENER (something which numbs, DEER—elks perhaps— round A DEN—wild animal's lair). (26 April 2006).
Stagger, describing Birmingham area. (8) Answer: ASTONISH (Aston-ish, Aston being a part of Birmingham.) (8 June 2006).
One very reliable tugboat (5, 2, 8) Answer: TOWER OF STRENGTH (19 June 2006).
Difference in sequence (3) Answer: ROW (21 June 2006).
Energy to get closer (3) Answer: ZIP (something which closes, as well as energy) (2 July 2005).
Nice fellow councillor blocking attempt to obtain novel (10, 5) Answer: FRENCHMAN'S CREEK (title of a novel by Daphne du Maurier, a Nice fellow is a FRENCHMAN, CR abbreviation for Councillor in the word SEEK,attempt). (27 July 2006).
Polish brigade involve one in runners on board. (10) Answer: CHAIRWOMEN (Women who run boards, CHARWOMEN are the brigade of polishers, surrounding 1). (25 January 2007).
Less sensitive issue of magazine. (10) Answer: NUMBER (22 October 2008).
Object to having ice-cream not guacamole for starters. (5) Answer: THING (Object, first letters of the next five words.) (31 July 2009)
Champ who put one screaming on the canvas? (5) Answer: MUNCH (Chew, Norwegian artist who painted The Scream.) (13 March 2010)
Shot or shut (5-2) Answer: CLOSE-UP (22 January 2011)
References and acknowledgements:
Higgins, J (1984). "It or ate; a note on the pronunciation of words ending in -ate.", ELT Journal 38, 1, p. 50-51.
My thanks to Clay Stabler for suggesting affect, arithmetic, batman, bowman, construct, coop, curate, hue, legitimate, mime, pace, preposition, and rainier, to Eduardo Parra for agape, Peter Reitan for bustier and brat, to Andrew Leventis
for supposed, and to Daniel Asimov and Charles Jeffery for the extremely specialised homographs of periodic and unionised.
Always nice to hear complimentary things about one’s work. I don’t earn more than a few pennies a week from the site, so I don’t expect any payment. By all means redistribute it in the way you suggest. Just as long as I am credited as the author, that is all I care about. If you are likely to make serious money with it, then a donation to a cancer charity would be appreciated. Our daughter died of breast cancer in 1997.