By Fred W. Riggs
Note: This essay explains how the terminological (lexicographic or semantic) analysis of concepts contrasts with the conceptual (or onomantic) approach. The two models complement each other and should be used together. The onomantic model produces a two-by-two table which identifies a logically necessary fourth concept (here called allonym ) which completes the scheme containing homonyms and two types of heteronyms (homographs and homophones). Conventionally, however, only the word-to-meanings paradigm is followed, producing the illogical results generated by this example derived from WordNet. Some wordgame enthusiasts enjoy listing homonyms and heteronyms -- see the links given below. Their lexicological focus is reflected in virtually all glossaries for special languages -- they also follow the orthographic (semantic) paradigm, with only an occasional step towards the logical classification of concepts, as one may see by viewing specialized glossaries.
WordNet is a Lexical Database for English prepared by the Cognitive Science Laboratory at Princeton University. It contains linked terms in a
Full details and the illustrative material involving homonyms and heteronyms can be found at: WordNet To activate the SEARCH engine, use the link to HTML FORMS INTERFACE. In the following analysis I look at the lexical meanings of homonym and heteronym and develop a two-by-two table setting forth the logical relations to be derived from their definitions. Anyone wishing to see the results before reading the explanation that precedes it can jump to table .
I searched for word at this site on the premise that it represents the most important term and concept in the WordNet project. The record for word identifies ten senses (i.e., ten difference concepts) starting with the following:
A follow-up search produced a long string of "hyponyms," described as various kinds of WORD(1) -- i.e., word in its first sense -- since all the other meanings of word are irrelevant here, there is no reason to talk about them in this discourse. However, let me note in passing that normally the different senses of any word can be understood without ambiguity -- consider, for example, the phrases: "to have a word (discourse) with someone; to keep one's word (promise); to send word (news); to give the word (order); the word (revelation) of God".
In the WordNet lists, hyponyms are arranged hierarchically: some items in the list are recursive in the sense that they include the hyponyms of each hyponym. The text explains that hyponym is the generic term used to designate a member of a class: X is a hyponym of Y if X is a (kind of) Y. In WordNet, the set of all hyponyms of a word, called a synset, is based on just one of the word's senses -- there are other synsets for the other meanings of word. Each entry in a synset is preceded by "=>", indented from the left according to its level of hyponymy in the hierarchy.
This approach is strictly semantic and not onomantic: it starts with words and their definitions, as taken from dictionaries. This means that logical problems involving relationships between concepts and the terms used to represent them are not considered, except in the determination of hyponym hierarchies. In short, the "net" in WordNet applies only to lexical definitions, not to the underlying logic of relations between concepts. (Parenthetically, note that others have puzzled about the illogic of dictionary definitions of homonym/ heteronym -- see, for example, Keith C. Ivey's Figuring Out the Definition) -- and various homonym lists entered below at links )
By reversing this semantic model, the onomantic (ana-semantic) paradigm permits us to analyze sets of related concepts and ask whether or not the existing terminology adequately represents them. Where deficiencies are identified, it provokes inquiry into the possibility of adding new terms that would permit clear representation of useful but hitherto undesignated concepts. When this analysis reveals that a word is being used in more than one sense, parenthesized sense numbers are added to clarify this point, and to justify suggested neologisms that could be added. A table is used to identify conceptual relationships, and one bold face word in each cell indicates the word I prefer to use for that concept. Neologisms are underlined to identify them as suggestions not supported by dictionary entries.
To illustrate this approach I append a set of terms from the WordNet data, using homonym/hypernym, and homograph/homophone as a starting point. The analysis that follows uses the onomantic approach.
Excerpts from WordNet: some Hyponyms of WORD(1)
The treatment of homonym and heteronym illustrates well the bind generated by a purely semantic approach. The supplementary onomantic approach (reversing the semantic paradigm) can clarify some issues posed by these definitions, as demonstrated in the two-by-two table given below, using spelling and pronunciation as the two variables.
Meaning is not included in this table because we may assume that the meanings of any one word differ from those of any other word -- even the most exact synonyms typically have different connotations. The definition of synonym offered by WordNet makes this clear. It reads:
Since any word can have more than one meaning, it is clear that even though, for example, book, work, and text could be used as synonyms in some contexts, each word has other meanings which would make them not synonymus in other contexts. If two words can be found that could be interchanged in all contexts, that would be remarkable indeed.
SYNONYMS vs. EQUIVALENTS
To underline this point let me make a distinction between synonym and equivalent. In the quotation above, they mean the same thing, but here I'd like to distinguish between them. The WordNet definition is semantic -- it talks about the established meanings of words. By contrast, In an onomantic context, we can say that any term that can replace another to designate a defined concept is an equivalent. This just reverses the semantic perspective: instead of defining the established meaning of a word, it describes a concept and presents one or more terms that can be used to represent it. To illustrate, consider first the following WordNet definition:
We can invert this text to create this onomantic record:
the phonological or orthographic sound or appearance of a word: FORM, WORD-FORM , WF
What we find here is that both form and word-form can be used as equivalents to represent the concept described in the text that precedes it. This is not a claim that the text defines a meaning of the term. Indeed, if we accept WF as an abbreviation or acronym to represent this concept, we will see that these two letters, although they have no intrinsic lexical meaning, can be used as equivalents of form and word-form to designate concepts described in this text. The word acronym itself is thought by some -- whimsically -- to be an acronym -- see ACRONYM . Formally speaking, the text does not define a word. Instead, it identifies a concept, with or without a name, The terms that follow are presented as designators of the concept regardless of what other meanings they may or may not have. If they happen to have other meanings, anyone using them for this concept needs to make certain their intended meaning is understood. -- thus form, which has many possible meanings, will be unambiguous only if anyone using it to mean a WF makes certain that it can be so understood, in context.
WORDS AND WORD-FORMS
Every word normally has several forms (word and words; or run, ran, running, for example) and a word-form may well represent several words, such as fair (impartial) and fair (an event).
As the WordNet definition specifies, word-forms can be represented by a sound (phonologically) and by appearance (orthographically). I shall use two other words for these concepts that more more clearly put them in a language context: pronunciation and spelling. I prefer these words because many sounds do not represent speech, but we only pronounce words and many things, like furniture and houses, have appearances, but we only spell words.
This is an unimportant quibble, however. The important point is that a two-by-two table can represent all possible combinations of pronunciation (sound) and spelling (appearance) of words. To put it the other way around, different words may have the same spelling, and/or the same pronunciation, or they may be spelled and pronounced differently. The semantic point may also be inverted -- different words always have different meanings (including synonyms!) but one concept may be represented by different words -- thus word, word-form and WF are different ways (spelling and pronunciation) of representing one idea (notion, concept) -- this makes them equivalents but not synonyms, as noted above. To visualize this basic distinction graphically, consider the following table:
|SAME||A homonym||B homograph (or heteronym)|
|DIFFERENT||C homophone (or heteronym)||D allonym,|
Most words differ from each other in both spelling and pronunciation -- therefore they belong to cell D in this table -- I shall call them allonyms for reasons explained below. Different words that are spelled and pronounced the same way are classed in cell A and are correctly called homonyms -- but some writers, confusingly, call them heteronyms
When different words are spelled the same way but pronounced differently, they belong to category B. It is precise to call them homographs and they are sometimes (misleadingly) called heteronyms . By contrast, when different words are pronunced the same way but spelled differently, we may properly call them homophones -- rarely, they have also been called heteronyms . In the rest of this paper, I shall use heteronym (rarely) as a broad term to include both homographs and homophones, but not homonyms.
Fortunately, two of the terms identified above by WordNet
have only one meaning and they may be used to represent concepts
B and C, as follows:
HETERONYMS: HOMOGRAPHS AND HOMOPHONES
B: HOMOGRAPH Two or more words spelled the same way but pronounced differently (example: to bow [bau] low, and bow [bo] and arrow) Unfortunately, homonym and heteronym are both used, in some contexts, to refer to B. In this paper, I shall use homonym to refer only to A, and heteronym only for a broader concept that includes both B and C.
Wordgamesters, who enjoy puzzles based on the appearance of words, have given us some Web Sites that focus on homographs -- see Homographs by John Higgins and Homograph Page by Jon and Sandra Vahsholtz
C: HOMOPHONE. Two or more words pronounced the same way but spelled differently (example: to, too, two, 2). Note that WordNet does not include the symbol, (2), although it can be pronounced the same way -- perhaps that is because symbols, although written, are not "spelled"! Extended lists of homophones can be found in the English Homophone Dictionary, by Peter Suber, the Homonym List by Alan Cooper, and the Homonym/Homophone Page by Tracy Finifter. Actually, all three of these sites contain homophone lists -- but both Cooper and Finifter use homonym to refer to homophones. Strangely, I have not been able to find any true homonym lists -- words that are pronounced and spelled the same way.
When different words are spelled and pronounced the same way, we call them homonyms(1) as in A. (Parentheses are used here to represent different senses of a word) Sad to say, homonym(2) is equivocal when used to mean (A, B, or C). Because equivocal is used in Onomantics to make an important point, let me digress to say a few words about it, and to distinguish it from polysemy and ambiguity.
POLYSEMY, AMBIGUITY AND EQUIVOCATION
Polysemy characterizes words that have more than one meaning -- any dictionary search will reveal that most words are polysemes -- word itself has 12 significant senses, according to WordNet. This means that the word, word, is used in texts scanned by lexicographers to represent twelve different concepts. The point is that words are not meanings, although they can have many meanings.
Lexicographers make a clear distinction between different words by writing separate entries for each of them, whether or not they are spelled the same way. My dictionary has 5 entries for the form, bow -- this shows that lexicographers recognize this form (spelling) as a way of representing five different words. Three of them are pronounced bo and two bau which identifies two homophones in this set of five homographs, each of which is a polyseme, capable of representing more than one concept. To summarize: bow is a word-form that stands for two different homophones and, as a homograph, represents five different words.
Moreover, the form bow is polysemic and can represent more than 20 concepts (its various meanings or senses). By gratuitously putting meaning in its definition of a homograph, WordNet can mislead readers who might think that a word is a homonym because it has several meanings -- but having one word represent more than one concept is normal -- just consider term as an example: it can not only refer to the designator of a concept, but also the duration of something, like the school year or a politician's hold on office, a legal stipulation, one's standing in a relationship (on good terms) and many other notions -- more than 17 are identified in my dictionary. By contrast, homonyms are different words and each of them (as a polyseme) can have multiple meanings.
To make their definitions precise, lexicographers need criteria to distinguish different words from each other even though they are spelled the same way. This usually hinges on etymology and, sometimes, parts of speech. One might, for example, think that that firm (steadfast) and firm (business unit) are two senses of one word (polyseme). Not so! Lexicographers class them as different words because the first evolved from a Latin stem meaning throne or chair, and the latter from a different root in Italian meaning signature.
Dictionaries are not uniform in their treatment of the different grammatical forms of a word. In some of them, the adjective firm (securely) is handled as a different word from the noun firm (to settle) even though they have the same etymology. I am not persuaded such differences justify treating grammatical classes (adjectives, nouns, and verbs) of a word-form that belongs to a single lexeme as different words -- the precise meaning of lexeme is explained below. The relevant point here is that deciding whether or not a form identifies one or more than one lexeme does not hinge on meanings. There is agreement that a word-form represents different words when they evolved from separate roots, and some lexicographers treat each grammatical use of a lexeme (noun, verb, adjective) as though it were a different word.
Ambiguity refers to a cognitive process -- when someone fails to understand the meaning of a word in context, then the word is being used ambiguously. Words are not ambiguous because they are polysemes, homographs, or homophones. For example bow (bau), as a polyseme, can mean a head gesture, submission, crushing, or ushering in. Normally, in context, one's intention when using this word is quite unambiguous. Even more clearly, when speech is involved, the distinction between bow (bau) to gesture, and bow (bo) -- as in rainbow or bow and arrow -- is well understood. Thus ambiguity need not arise just because words are homonyms or polysemes.
Equivocation characterizes words that are polysemic in a single context of use. Because it is often difficult to determine which of a word's possible meanings is intended, ambiguity is likely to arise. Word itself is a good example. One meaning of word is implicit in the WordNet definition of word-form quoted above:
Consider that the dictionary entry for sing includes sang, sung, singing as different forms of a single word. However, in the WordNet definition of word as it appears in the quotation heading this note, we read:
Here we can see that the meaning of word is the same as word-form. When one speaks of singing, one can only use one form of the word -- in "she sang beautifully" the word used is only one of the forms of SING. Interestingly, WordNet offers no term for the sense of word that includes all its inflected forms. However, as I understand it, lexeme is used to represent this concept: it is defined in my dictionary as a meaningful linguistic unit... in the vocabulary of a language. In this sense SING as a lexeme includes all its forms. To make this distinction here, I will write sing to represent a word-form found in the lexeme, SING. The form of a word used to head a dictionary entry is called its citation form. Normally, the citation form of any lexeme is the form used to represent it -- the full lexeme however, includes all of its various forms.
In the first quotation offered above, WordNet clearly means SING, a lexeme, but in the second, it means sing, the form.. No doubt this is a subtle distinction and in most contexts, it makes no difference which meaning is intended. However, in some contexts it is an important distinction and not having terms to represent these concepts unequivocally can lead to ambiguity. WordNet gives us word-form (to identify sing) but not lexeme or some equivalent (to categorize SING). The goal of the Onomantic approach is to identify an unequivocal term for every concept needed within a given field of discourse. Such terms are often needed if we want to avoid ambiguity in our professional work. A major cause of ambiguity in scholarly work is equivocation, the lack of adequate terms to distinguish between different senses of a word that occur within the same field of discourse. By contrast, polysemy does not lead to ambiguity so long as each sense of a word can be distinguished from others by its context of use.
Not surprisingly, many words used in everyday language are equivocal because, in the same context, they can represent more than one concept. In everyday language, this poses no big problem because having a rough idea about what one wants to say is generally quite o.k. However, in specialized languages, it becomes more important to avoid equivocation, and we may suppose that most technical technical glossaries will offer terms that have only one meaning within their special domains. A collection of such works on the INTERNET can be found at Glossaries. Readers might try running through one in a field with which they are familiar to see whether all their terms have unequivocal meanings.
Here is a simple test case: You will find the following definition of confidence -- Support for the government by the majority of the members of parliament -- in the HTP Glossary for Political Science. Canadian readers will understand this term in the context of their own parliamentary system of government, but Americans who are familiar only with their own separation of powers regime might be puzzled and assume the word should refer to a leader's "self confidence," or the degree of mutual trust manifest in one's "confidence in a colleague."
Returning now to our analysis of the four cells identified in the table above, consider that homonym is used equivocally for two different concepts, represented by cell A, and by (A,B & C).
More formally, we can represent these two concepts as follows:
The definition of homonym offered in WordNet clusters (A, B & C) by writing::
This means that "homonym" is defined by WordNet in sense #2. My dictionary defines homonym as meaning A: i.e., homonym(1)
Clearly the difference between OR and AND is fundamental, giving us two different concepts, as represented in box A or boxes A, B, and C. Equivocation is underlined by the admission, in the same entry, that homophone and homograph represent two related concepts that may also be spoken of as homonym. WordNet also combines all three concepts under one term. I have entered these equivocal concepts in the table as homonym(1) in cell A, and homonym(2) in all the cells except D.
In cell A homonym(1) is written in bold face to show my preference for this term as an unambiguous way to represent concept A, but homonym(2) is shown in cells A, B and C to show that the same word can be used equivocally as a generic term for all three concepts. However, to avoid possible ambiguity, we could use modifiers to create unequivocal terms that can easily distinguish between these equivocal meanings of homonym. Perhaps strict homonym could represent the narrower sense (1) in A, and loose homonym the broader sense (2) in (A, B and C).
A more complex and baffling problem involves the use of heteronym. As defined by WordNet, it means:
This word, therefore, is equivalent to homograph, C. The definition of homograph in the WordNet text adds a meaningless characteristic by referring to "meaning" as a defining feature. I ignore it here for reasons explained above. Since most words are polysemes -- word, for example, has ten senses identified in WordNet -- to say that different words have different meanings does not distinguish between polysemes (one word with several meanings) and homonyms (set of words having the same form).
Since we already have homograph to represent this concept unambiguously, it seems redundant to waste heteronym as an equivalent term -- but that seems to be the recognized usage. However, if we wanted to make better use of the word, we could stipulate a different meaning for it. For example, since homonyms by definition have the same spelling and pronunciation, we might use heteronym as an antonym to represent words that differ from each other in spelling or pronunciation, but not in both -- this would make it a broader term for words that are not homonyms but are either spelled or pronounced the same. We could then say quite clearly that there are two kinds of heteronyms: homographs and homophones. Such a usage would enable us to distinguish easily between cell A (homonym) and cells B and C, (heteronyms), and it would also create a useful contrast between words that are both spelled and pronounced alike, and those that are only pronounced or spelled alike.
None of these terms takes into account the situation represented by cell D. Keith C. Ivey, in his discussion of homonyms, recognizes this fact and writes:
Unfortunately, this seemingly neat solution doesn't work because all heteronyms are different words as Ivey's examples show. He illustrates homophones with board/bored, clearly two different words though pronounced alike, and his example of homographs (the verb desert/the noun desert) again shows, by their pronunciation, that they are different words. Even his example of a homonym -- words having both the same sound and spelling, as illustrated by "to quail and a quail" -- clearly shows they are different words. Lexicographers underline this point by writing separate entries for different words, whether or not they have the same spelling and pronunciation.
One could stipulate a phrase, like uniquely different words to represent category D, but this expedient is cumbersome and not transparent. A simpler solution, I believe, can be found by means of a neologism. It is not difficult to think of a suitable term. I shall propose one based on the way linguists talk about lexemes.
But first let me explain that a lexeme can be either a word or a phrase. All common words and phrases that can be viewed as suitable for entry in a dictionary are lexemes -- proper names, by contrast, are not lexemes -- thus hope is a lexeme but Hope is not. Some, but not all, dictionaries do enter proper names, but their treatment involves identifying an object, not defining concepts. Because lexicographers treat certain phrases as appropriate for use as dictionary entries, they need the concept of a lexeme -- also sometimes called a lexical unit. For convenience, they often use word to include phrases, as in head word , a technical term that includes any phrase or word that heads a dictionary entry. More precisely, all head words are lexemes, even if they are phrases. However, lexemes are sometimes entered several times in those dictionaries that view the nominal, verbal and adjectival forms of a lexical unit as different words.
This brings me to the suggestion I have for a term to represent box D. Technically, some linguists refer to each of the different forms of a single lexeme as an allolex. The same stem, allo-, is used in other linguistic terms such as allomorph, allophone, and allograph. A definition of allomorph can be found in Linguistic Glossary
My Webster's Collegiate enters allograph in the sense of a letter or combination of letters that can represent the same sound -- thus "f" (find) and "ph" (phone) are two allographs for one phoneme. We could, using this model, coin allonym as a neologism to represent any word or phrase that is spelled and pronounced differently from all other words and phrases. We cannot use allolex because this word has already been preempted to represent the different forms of a single word -- for example, program and programme are different ways of spelling one word, and there are many such variations between British and American usage that create allolexes. Also, dialect differences produce different ways of pronouncing the same word.
Even if this were not so, I would prefer allonym because it rhymes with homonym and heteronym. This makes it easier to remember that an allonym is a word that differs in spelling and pronunciation from all other words, whereas both homonyms and heteronyms identify words that are the same, in some ways, as other words. This term also enables us to name the blank cell D, in our table, as follows:
D. ALLONYM: Any lexeme (word or phrase) that is both spelled and pronounced differently from all others
No doubt in ordinary usage, we will have little need for this term, although it would simplify lexical explanation if one could start by making the claim that most words in English are allonyms. The clear exceptions could then be discussed under the heading of homonymy -- different words pronounced and spelled the same way. Moreover, this would also facilitate the explanation of heteronymy, not as an antonym for homonyms but rather as two kinds of quasi-allonyms: these are words that do differ from each other
Just because most English words are allonyms, we need not dismiss the term as unnecessary. In fact, the prevalence of allonymy may be a distinctive feature of some languages. In others, allonyms are rare. Consider Chinese, for example. Almost no Chinese words are allonyms -- most of them are homophones which can represent different words, each of which is signified by a unique character (ideogram). However, Chinese lacks homographs and therefore homonymy is also rare in that language, and written words are unambiguous. Ambiguity is reduced in spoken Chinese by the generous use of phrases: many lexemes are composed of two words -- thus a compound, east-west, represents the concept of a thing in Chinese, just as hard rock in English refers to a kind of music, not a solid object. Such phrases are allonymic in Chinese where each of them has only one meaning. We do not think of them as a single word because they cannot be written together but must retain their form as separate characters. It's as though we could not write understand as a single word, but had to see it as a phrase composed of under and stand .
A strikingly different situation exists in Japanese where two orthographic systems are used: there is the Katakana system for representing speech syllables, and the borrowed Chinese ideograms that represent concepts more than sounds. Thus Japanese words may be both homonyms and allonyms at the same time: the former applies to the Katakana version which can represent different words having the same pronunciation and written form -- but they are allonyms when represented by ideograms each of which is unique.
Japanese texts ofen mix the Katakana and Chinese symbols -- one reason is to avoid ambiguity where the syllabic representation would be ambiguous. There are analogies in English involving the use of Arabic symbols for numbers. Thus the spoken word (TOO) might be ambiguous when used to mean "2", but using the symbol would make one's intentions clear. Of course, "2" can also be viewed as a homograph insofar as "two" and "twice" are different words that can be represented by this symbol. In English, allonymy is normal and therefore we ignore it, while focussing attention on the homonyms and heteronyms that are exceptional -- as illustrated by the links listed below. By contrast, in Chinese allonymy is rare and would, no doubt, attract attention if it did occur. In Japanese, the situation is more complex: allonymic when Chinese characters are used, but sometimes homonymic in Katakana.
My point is that just because most words in English are allonyms, we need not ignore the concept and reject a technical term for it. The concept may prove helpful in other contexts -- e.g., in non-Western though not in Western societies. It could also help us explain contrasts -- for example, it would be much easier to understand heteronymy (including homophones and homographs) if we started by explaining allonymy as the English norm.
Perhaps if Chinese linguists had been the first to explain the English language, they would have started from that end of the spectrum. It is, after all, ethnocentric to take the normal for granted -- like the proverbial fish who doesn't know what water is. Our terminology and concepts focus on the atypical because we notice the unfamiliar: we comment about poetry but not about the prose we speak daily.
The purpose of this note is not to focus on homonymy/allonymy -- these are marginal curiosities of language, although they have spurred the interest of word lovers who enjoy compiling lists of them -- see Links . I discuss them, instead, as a concrete example used to clarify the far more important contrast between a purely semantic approach which limits itself to the cognitive analysis of words as established and defined in dictionaries, and a supplementary onomantic approach that can augment (not replace) semantic analysis by investigating the logical structure of related concepts, evaluate the established terminology, and propose additional terms for missing items: for example, allonym for D.
Sometimes a useful distinction can be made simply by borrowing a technical term already in use by specialists, or just by writing a word in a different way. For example, the use of lexeme by contrast with word-form enables us to clarify the important distinction between SING ( a word-set which includes sing, sang, sung, and singing) and sing as just one item in this list. I have also tried to show that although word, as a polyseme, need not cause confusion,WORD(1) is used equivocally by lexicographers and terminologists, leading sometimes to ambiguity. The solution to such ambiguities caused by using words that have several relevant meanings in the same sphere of disccourse can be found by anyone using an onomantic approach to make useful conceptual distinctions and find suitable terms to represent them, thereby supplementing the familiar semantic methodology employed by WordNet..
See linked pages:  Onomantics  and SITES for Terminology 
Go to TOP for explanation of the onomantic analysis of homonyms and heteronyms, illustrated in the TABLE.
See illustrative wordgame comments and data at:
 Figuring Out the Definition Keith C. Ivey || Homographs John Higgins || Homograph Page Jon/Sandra Vahsholtz
|| English Homophone Dictionary, Peter Suber || Homonym Alan Cooper || Homonym/Homophone Tracy Finifter 
click here for Home Page links:
|Personal Autobio||Prism GRD||Globalization Concepts||Ethnicity ETHNIC-L||COCTA Onoma.||COVICO Choices||Impeach