Semantics is the study of meaning in language.  The term is taken from the Greek seme, meaning sign.   The word meaning can be defined in many ways, but the definition most pertinent to linguistics and the one we will use is that meaning is "the function of signs in language."  This understanding of meaning corresponds to German philosopher Ludwig Wittgenstein's definition: 'the meaning of a word is its use in the language' (in other words, the role a word plays in the language).  

      The term semantics was only invented in the 19th century, but the subject of meaning has interested philosophers for thousands of years.  The Greek philosophers were the first people known to have debated the nature of meaning.  They held two opposing views on the subject.

      The naturalist view, held by Plato and his followers, maintained that there was an intrinsic motivation between a word and its meaning.  The meaning of a word flows directly from its sound.  The Greek word thalassa, sea, in its classical pronunciation, supposedly sounded like the waves rushing up onto the beach.  If the naturalist view were entirely correct for all words, we would be able to tell the meaning of any word just by hearing it.  In reality only a few onomotopoeic words in each language actually sound something like what they mean:swoosh, splash, bow wow, meow.  Poets can skillfully use words with sound features that heighten the meaning intended:

      a.) Shevchenko (Ot topota kopyt pyl po polu idyot.)

      b.) Lermontov (... a on, myatezhny, prosit buri, kak budto v buryakh yest pokoi.)

But poetic sound imagery represents a rare, highly clever use of language, so the naturalist approach is applicable to only a tiny portion of any language.

      The conventionalist view of Aristotle and his followers holds that the connection between sound and meaning is completely arbitrary, a matter of social convention and prior agreement between speakers.  It is true that the form of most words is arbitrary from an extra-linguistic point of view. This position is much nearer the truth. 

      However, the form of a word may be motivated by the forms of other words in a language.  That is, although a word's meaning is arbitrary from the point of view of the real world, is is often somehow motivated by the system of the language it is a part of. In studying morphology we saw that the meaning of a word can often be deduced from knowing the meaning of its parts.  Since words often originate from other words, a word very often has some historical reason for being the shape it is.  Sometimes the origin (or etymology) of a word is completely transparent, as in the case of unknown from known, or discomfort from comfort.  At other times the origin of a word is less immediately obvious but nevertheless present in the form of a word, as in the case of acorn < oak + orn.

      Philologists (this is a broader term for people who study language as well as anything created with language) often make a distinction between meaning and conceptConcept is the totality of real world knowledge about an item, while meaning is a category of language.  It is possible to know the meaning of the word without knowing everything about the concept referred to by that meaning.  For example, one can know the meaning of a word like diamond without knowing the chemical composition of the stone or that carbon and pencil lead are, chemically speaking, composed of the same substance.  In other words, one can know the word diamond means a type of gemstone without understanding the full concept associated with that gemstone in the real world. 

      Sometimes, however, meaning and concept cannot be so easily differentiated.  For instance, the meaning of many abstract words completely parallels the concept they refer to, as with the word tradition and the concept "tradition."  It is arguable that one cannot know the meaning of the word "tradition" without understanding the concept "tradition." 

      Linguists have a second way of looking at the distinction between linguistic and real-world knowledge.  They often discuss the difference between a word's sense and its referenceA word's sense is how the word relates to other words in a language (Wittgenstein's "meaning"); it's reference is how it relates to real world concepts.  The French word mouton refers to a sheep as well as to the meat of the animal as used for food (the sense of the word combines two references). In English we have two separate words for each extra-linguistic reference.  The sense of the English word sheep is limited by the presence of the word mutton in English.  There are many such examples when comparing languages:

      a.) Cherokee nvda means the concept sun as well as moon

      b.) Russian ruka, hand or arm --kist' ruki specifically means hand.

      c.) English hand vs. Cherokee atisa (right hand), akskani (left hand)

      d.) English uncle vs. stry´ko(father's brother) vs. ujo (mother's brother).

      Thus, the sense of a word concerns its linguistic boundaries in a particular language.  The reference of a word concerns which concepts it refers to in the real world. 

      The distinction between a word's sense and its reference, or between linguistic meaning and real-world concept--difficult though this distinction may be to draw in many cases--is useful in comparing semantic categories across languages. Languages may divide the same set of real-world concepts in very different ways. 

      The concept of blood relations offers a good example.  Each language has its own set of kinship terms to refer to one's parents' generation (mothers, fathers, aunts, uncles).  Discussing kinship terms from the point of view of real world concepts allows comparison across languages without bias in favor of the meaning categories of any particular language.  Here is a comparison of kinship terms for aunts and uncles in English and Pitjantjatrara, an Australian language:

kamuru --bio. brother of female parent

ngunytju-- bio. sister of female parent or female parent

kurntili --bio. sister of male parent

mama-- bio. brother of male parent or male parent

      Another good example is provided by color terms.  Every language has a set of basic color terms.  But these color terms do not divide the color spectrum in the same place.  In other words, the extra-linguistic concept "color" is reflected in each language idiosyncratically.  For instance, literary Welsh divides the green/brown part of the spectrum quite differently than English (Welsh as it is used in everyday speech today has conformed to the English divisions of color):

English                       green  !           blue       !           grey               !           brown 

Welsh           gwyrdd !     glas (Engl. blue + color of plants)   !   llwydd

      The linguistic division of concepts is in part arbitrary and idiosyncratic to each language, in part motivated by factors actually present in the real world.  Extra-linguistic (real world) factors may result in universal tendencies in how languages divide up concepts.  A second look at color terms illustrates this point.  

      Berlin and Kay (1969) in Basic Color Systems, noted that the number of basic color terms may differ across languages.  A basic color term is defined as one which cannot be said to be a part of the meaning of another basic term.  Yellow is a basic color term in English because it is not a type of red, green, blue, etc.; but turquoise is not because it is a type of blue.  Berlin and Kay found that some languages have only two basic color terms while others have as many as 11.  Berlin and Kay discovered that the number of color terms seems to be systematically related to the core color represented by the terms:

   1                2.             3                  4        5                       6

black/white    +  red    +  yellow/green   +      blue +   brown        + purple, pink, orange, grey


A few languages, including Russian, have 12 terms because in place of one generic term for blue, there are two words: Russ. goluboi/sinii.   Subsequent research has shown that a minority of languages violate this scheme.

      The question arizes as to why color terms in languages should conform so universally to such a general pattern.  Geoffrey Sampson, who wrote Schools of Linguistics, offers what might be a partial explanation.  Physiologically, red is the most noticeable color to the eye; therefore, it is the first most likely to be represented after light and dark.  Distinctions involving yellow and green occur frequently in nature, whereas blue only naturally occurs in a few things: the color of some flowers, of bodies of water, the sky; the distinction between blue and other types of dark, therefore, is relatively unimportant to many aboriginal cultures and not as likely to be included in the language under specific terms.

      Linguists who study meaning (semanticists) often divide the meaning of a word into semantic components based on real world concepts, such as human/ live/ dead/ animal/ plant/ thing/ etc.  Discussing the meaning of words by breaking it down into smaller semantic components such as is called componential analysis

      Noting how semantics is based on extra-linguistic categories, a group of linguists (including the Polish-born Australian linguist Anna Wierzbicka) have tried to reduce all meaning in language to a set of universal core concepts, such as tall, short, male, female, etc.  This finite set of concepts are then used universally, to describe the meanings of all words in all languages.  This semantic approach to language structure has problems.

      The first problem is in deciding which concepts are basic and which are derived.  Whatever language is used to label the concepts in the first place biases the semantic analysis in favor of the semantic structure of that lang. 

      A second problem is the old difficulty of distinguishing between sense and reference.  The linguistic boundaries between conceptual features vary across languages.  This is especially true with grammatical categories of meaning.

    a.) Masculine and feminine in English and Russian. 

    b.) Animacy in Russian and Polish (objects), and Cherokee (one plural ending from human, another for inanimate, and a different one for plants and animals). 

      Attempts to reduce meaning in all languages to a limited set of conceptual categories existing outside of language have been unsuccessful.

      Third, part of the reason the semantic universalists have been unsuccessful is that meaning is more than simply a reflection of real world categories.  Meaning is a linguistic category rather than a real world category reducible to pure logic and perception.  The role of semantics in language is often highly idiosyncratic.  We have seen that semantic factors often serve as constraints on morphology and syntax. Here are some more examples:

      a.) English locative adverbs with toponyms (This is my bed; I sleep here/in it; This is Fairhaven; I live here/*in it)  Note the distinction between an idiosyncratic semantic constraint and a logical constraint. Idiosyncratic semantic constraints in the grammar result in reference being made using one form instead of another. 

      Logical constraints result in reference not being made at all.  Compare the illogical sentence: Here is my thoroughness--I sleep *here/*in it.  If a sentence is illogical, than all paraphrases are equally illogical.  Other examples: the Russian -ovat suffix;  the plural of fishes in English. 

      Thus, meaning is not merely a reference to concepts in the real world.  It depends on linguistic factors in part unique to each individual language; meaning depends not only on the logical combination of real world concepts.  The system of language cannot be described only in terms of extra-linguistic logic.


How meaning affects word associations in language

      The purely linguistic side of meaning is equally evident when examining how words combine with one another to produce phrases.  The set of restrictions on how a word may combine with other words of a single syntactic category is referred to as the word's collocability.  Two words may have the same referent, and yet differ in their ability to combine with particular words.

      In English, the word flock collocates with sheep ; and school with fish, although both flock and school mean group.

      Also, addled combines only with brains or eggs (one must steam rice and boil water), blond collocates with hair, while red may collocate with hair as well as other objects. 

      Idiosyncratic restrictions on the collocability of words (and by idiosyncratic here I mean that part of meaning which is peculiar to language structure and not deriving purely from logic) result in set phrases: green with jealousy; white table vs. white lie.  On can get or grow old, but only get drunk, get ready, not *grow drunk, *grow ready

      Every language has its own peculiar stock of set phrases.  In English we face problems and interpret dreams, but in modern Hebrew we stand in front of problems and solve dreams.  In English we drink water but eat soup.  In Japanese the verb for drink collocates not only with water and soup, but also with tablets and cigarettes

      From the point of view of etymology, set phrases are of two types. 

      1.) The first type of set phrase, the collocation, may be defined as "a set phrase which still makes sense": make noise, make haste.  One simply doesn't say to produce noise or make swiftness, even though such phrases would be perfectly understandable. Since collocations still may be taken literally, they can be paraphrased using regular syntactic transformations: Haste was made by me, noise was made by the children

      2.) Phrases whose words no longer make sense when taken literally are called idioms.  The semantic relations between words in idiomatic set phrases may be illogical to varying degrees: white elephant sale, soap opera, to see red, break a leg, small voice, loud tie, wee hours of the night. 

      Also, true idioms cannot be paraphrased by regular means, because they do not participate in the regular syntactic relations of the language: John kicked the table--The table was kicked by John.  vs. John kicked the bucket.   A bearded sailor walked by.-- A sailor who was bearded passed by. vs.  An occasional sailor walked by.

      Thus, meaning involves real-world concepts and logic but it is at the same time a linguistic category.  The semantic structure of a language is the language's special system of conveying extra linguistic relations by idiosyncratic linguistic means. 

Semantic relationships between words

      Modern studies of semantics are interested in meaning primarily in terms of word and sentence relationships.  Let's examine some semantic relationships between words:

      Synonyms are words with similar meanings.  They are listed in a special type of dictionary called a thesaurus.  A regular dictionary lists words according to form, usually in alphabetical order; a thesaurus lists words  according to meaning.  Synonyms usually differ in at least one semantic feature.  Sometimes the feature is objective (denotative), referring to some actual, real world difference in the referents: walk, lumber, stroll, meander, lurch, stagger, stride, mince.  Sometimes the feature is subjective (connotative), referring to how the speaker feels about the referent rather than any real difference in the referent itself: die, pass away, give up the ghost, kick the bucket, croak.  There tend to be very few absolute synonyms in a language.  Example: sofa and couch are nearly complete synonyms, yet they differ in their collocability in at least one way: one may say couch potato, but not *sofa potato.  

      One type of synonym is called a paronym.  Paronyms are words with associated meanings which also have great similarities in form: proscribe/ prescribe,  industrial/ industrious,  except/accept,  affect/effect.  Many errors in speech and writing are due to mixups involving paronyms.

      Antonyms are words that have the opposite meaning.  Oppositeness is a logical category.  There are three types:

      Complementary pairs are antonyms in which the presence of one quality or state signifies the absence of the other and vice versa.  single/ married, not pregnant/pregnant  There are no intermediate states.

      Gradable pairs are antonyms which allow for a gradual transition between two poles, the possibility of making a comparison--a little/a lot  good/bad,    hot/ cold    cf. the complementary pair: pregnant/not pregnant

      Relational opposites are antonyms which share the same semantic features, only the focus, or direction, is reversed: tie/untie, buy/sell, give/receive, teacher/pupil, father/son.

      Some concepts lack logical opposites that can be described in terms of any special word; colors are a good example: the logical opposite of red is not red.   Such concepts may form relational antonyms, however, through symbolic systems of thinking.  For instance, in Cold War thinking, the relational opposite of American is Russian; in current US politics, the relational opposite of Democrat is Republican.

      Homonyms are words that have the same form but different meanings.  There are two major types of homonyms, based upon whether the meanings of the word are historically connected or result from coincidence.

      Coincidental homonyms are the result of such historical accidents as phonetic convergence of two formerly different forms or the borrowing of a new word which happens to be identical to an old word.  There is usually no natural link between the two meanings: the bill of a bird vs the bill one has to pay; or the bark of a dog vs the bark of a tree. 

      The second type of homonym, polysemous homonyms, results when multiple meanings develop historically from the same word.  The process by which a word acquires new meanings is called polysemy.  Unlike coincidental homonyms, polysemous homonyms usually preserve some perceptible semantic link marking the development of one meaning out of the other, as in the leg  of chair and the leg of person; or the face  of a person vs. the face of a clock.

      Sometimes it is impossible to tell whether two words of identical form are true homonyms (historically unrelated) or polysemous homonyms (historically related), such as ice scate vs. skate the fish: skate--fish (from Old English skata'ice skate (from Dutch schaat');  deer/dear are historically related (cf. darling, German Tier, animal.)

      Since polysemy is so difficult to separate from true homonymy, dictionaries usually order entries according to 1) the first recorded appearance of word or 2) frequency of meaning use.  This is a problem for lexicographers, the people who study words and write dictionaries.  

      There are universal tendencies in the directionality of polysemy.  studies of polysemy in a wide variety of languages generally find the following directions in meaning shift:

   1) part of body to part of object (hands, face, lip, elbow, belly, veins of gold or leaf); but: appendix.

   2) animal to human for personality traits (shrew, bear, wolf, fox, quiet as a fish); but: my cat is a real Einstein.

   3) space to time (long, short, plural),

   4) spatial to sound (melt, rush,)

   5) sound to color (loud, clashing, mellow)  

   6) Physical, visible attribute to emotional or mental, invisible quality (crushed, big head, green with envy, yellow coward, sharp/dull, spark)  

   Directionality in polysemy seems to be logically motivated: concrete meanings give rise to abstract ones (sharp knife-> sharp mind); mundane gives rise to the technical (chip of wood-> computer chip).

A note about spelling and semantics

      In a language like English where spelling often diverges widely from pronunciation, There is a special type of homonym called the homophone. Homophones have the same pronunciation but different spellings: meet/meat,  peace/piece, whether/weather, you, ewe, through/threw, to, two, too. cot/caught. flour/flower.  Homophones are usually are true homonyms in that they derive from completely unrelated sources. There are also occasional polysemous homophones: draft (into the army), draught (of beer), or the Russian voskresenie (Resurrection) --> voskresenye  (Sunday). 

      A language like English also has homographs, words spelled alike but pronounced differently in each of their meanings.  In English, most homographs are polysemous homographs: use (the noun vs. the verb), record (the noun vs. the verb).  But there are a few true homonyms that are homographs: wind (a noun meaning moving air vs. a verb meaning what is done to a watch or clock).

      Polysemy may involve conversion of one part of speech to another.  Many verbs in English, especially monosyllabic verbs, can be nouns: bend, drink, kill.  Many verbs can be transitive or intransitive: walk, fly, burn, return .  So polysemy can result in new grammatical as well as lexical meaning. 

      In certain instances, polysemy acts as a regular, productive pattern which affects entire classes of words rather than single, isolated words.  In English, words with certain functions systematically have a secondary function.  For example: the + noun in English may mean a) a single specific example (Thanks for letting me ride the horse (this specific horse) or b) a general type (example: The zebra is a relative of the horse.).  A new noun in English will combined with the automatically in these two different meanings: the wug (may mean one particular wug already mentioned or a general category of beings.)   Also, any verb in English can mean either single or multiple action: He hit the table (maybe once or more than once).  If a new verb is created it automatically inherits these two meanings: He burbled the table (maybe once, maybe more than once).   Every language has examples of such regular, grammatical polysemy. 

      There are a few other minor semantic relations that may pertain between words.  The first involves the distinction between a category vs. a particular type or example of that category.  For example, a tiger is a type of feline, so feline is a category containing lion, tiger, etc.; color is a category containing red, green, etc, red, green are types of colors. Thus, feline and color are hyponyms, or cover words, and red, green, lion, tiger are their taxonyms

      The second involves a whole vs. part of the whole. A finger is a part of a hand, thus hand is the holonym of finger; and finger is a meronym of hand.  Similarly, family is the holonym of child, mother or father

      Remember semantic terms such as synonym, hyponym, meronym, etc.  are relational rather than absolute: red is a taxonym in relation to the word color;  it is a relational antonym of the word green; finally red is a synonym of ruddy, etc.

      Now that we have discussed various semantic relations between words, let us try to categorize these relations in a more concise way.  Psychologically, it has been shown, the meanings of words or phrases are always related in one of two possible ways.

      The first way involves some relation based on similarity: the two referents actually physically resemble one another in some way (iconicity): face and hands of a clock, chip (of wood vs. potato chip).

      Certain figures of speech, or tropes, that is, words used in other than their literal meaning, are based on similarity relations:

      A metaphor is an implied comparison using a word to mean something similar to its literal meaning.  A contradiction arises between the literal meaning and the referent.  Metaphors can be fresh and creative or hackneyed (the eye of night for moon).  Metaphors that cease to tickle listeners with their creativity are called dead metaphors: they simply become secondary meanings of words, polysemous homonyms.  We don't even sense the original creativity that went into the first usages of such historical metaphors as: leg, handle.  Most compliments or insults contain metaphors: calling someone a pig, a worm, a big ox or a monster; or an angel.  

      A simile is a direct comparison using like or as: Examples: quiet as a mouse, as mad as a hatter.  New similes can be created, but each language has its own particular store of accepted similes that function as collocations.  English: healthy as a horse, quiet as a mouse.  Other languages have their own stock of well-0established similes: Russian: healthy as an ox, Mongol: quiet as a fish. 

      The second type of meaning relation between polysemous words is based on contiguity.  Here the two referents do not resemble one another; rather they occur in the real world in some spatial proximity to one another (either as parts of a whole or as one item located next to another).  For instance, the mother has many mouths to feed, all hands on deck, to boot someone out of a place, London issued a statement (London here means the people governing England).

      A few figures of speech are based on contiguity relations.

      Metonymy: Use of word to mean something existing in close physical proximity: Saying London to mean the people who govern England. Also: TheWhite House said meaning The president said.

      Synecdoche: Using a part to describe a whole: all hands on deck,  he has x mouths to feed. 

      All semantic relationships in all languages can be described based on similarity or contiguity.  This seems to stem directly from the structure of the human brain.  People who suffer brain damage affecting language usually experience impairment of either their similarity relations or their contiguity relations.