Semantics is "the study of the linguistic meaning of words, phrases, and sentences" [Fromkin and Rodman, 124]. As discussed in syntax and morphology, all parts of the language have some meaning (even if they are only "markers" like "wa" and "o" in Japanese). Words and morphemes have meanings as defined in our internal lexicon. If we know the word "assassin," we know that it is someone who is a human murderer, and is a killer of prominent people [Fromkin and Rodman, 125]. All words contain semantic properties such as that, sometimes more, sometimes less. A "tempter," for example, is a man who tempts (someone, usually a woman), while a "temptress" is a woman who tempts (someone, usually a man). Both words imply humans.
Homonyms (including that class of words that have similar pronunciations and different spellings and that class of words that are spelled the same way but have completely different meanings) add to the ambiguity of the language and often require additional semantic context to determine the proper meaning of a given word. ("He grabbed a bat" would be a good example, because we do not know if the "he" we are talking about is a baseball player or a veterinarian.) Ambiguity can also be caused by the use of synonyms (words that have similar meanings); minor differences can be found and identified with synonyms. Antonyms are words that have properties that are mutually exclusive (such as fast/slow). Fromkin and Rodman note that some antonyms are complementary pairs (fast/slow), while others are gradeable (they need specification, like hot/cold). In gradeable pairs, one of them is marked and the other is unmarked; the unmarked one is used to determine the degree of one or the other (such as, "how tall is it" being answered with "three hundred feet"). Other antonyms are relational opposites (employer/employee). (There will, of course, be times when otherwise antonymical words will mean the same thing such as good/bad in reference to the "quality" of a scare) [124-134].
Names, the final class of words in semantics, always refer to a unique object. In English, names rarely have articles preceding them unless there is a need to clarify, such as "the Mississippi River." Other languages, such as Greek, require articles before names. Proper names, unlike other words, cannot be pluralized and remain definite (such as "the Smiths" implying the Smith family). Proper names, may, however, be plural in their "natural" state, such as "the Pleiades." Rarely will adjectives be used with a proper name, but they can be used to further specify or to emphasize a quality [Fromkin and Rodman, 135-137].
Linguists indicate the semantic properties of words with semantic features and semantic redundancy rules. For example, if a word has the feature [+human], it is automatically [-abstract]. This means that antonyms almost always have reciprocal features, such that:
fast [+fast] [-slow] slow [+slow] [-fast]
There are literally hundreds of semantic redundancy rules for every word, most of which need not be stated. When we look at a dictionary, we are looking at the most common usage of the word (the properties are refined to the most common present and not-present properties) [Fromkin and Rodman, 124-137].
But as we have seen, knowledge of words does not constitute knowledge of the language. Just as we must know the syntax (how to put words together into sentences) of a language, we must know the semantic meaning that the syntax enforces. We must know how adjectives affect nouns, to what degree, and in what order. Although "large" in English implies "large for the modified object," such as "large balloon" or "large house," in another language, there might be two words for the concept "large" - one for smaller items, like a balloon, and one for larger items, like a horse. The order is important as well: "brick red" is not the same thing as "red brick." As Fromkin and Rodman say, "meanings build on meanings" [138].
Noun phrases take on different roles within sentences depending upon location. For example, in one sentence, there may be an agent (a "doer"), a theme (a "recipient"), a location ("where"), a goal ("where the action is directed"), a source ("where the action originated"), and an instrument ("the object used to accomplish the action") [Fromkin and Rodman, 139]. In all cases, the verb indicates the action. While some languages, such as English, allow various themes - even non-themes - to be the subject of the sentence, others, such as German, are much stricter. Languages which are stricter in what theme a subject may represent generally have a strong "case" system, or what morphological shape the noun takes according to its thematic role in the sentence. English has the genitive/possessive case (the possessive form of a noun). The universal principle of "theta-criterion" has been proposed, which states in part that a particular thematic role may occur only once in a sentence [Fromkin and Rodman, 140-142].
Semantics and syntax closely interact in that something must generally be syntactically correct to be semantically correct (in turn, the words must also be morphologically correct). The semantics of the language tell us when the syntax is incorrect because meaning is not present. However, one will often find syntactically and semantically correct sentences that are still "false" in actuality. But we have the knowledge in principle of how to discover the truth of a sentence, even if we do not have the direct means to do so, because we know the language. If the language is unknown to the reader, then we do not have any way to determine its truth. Often, however, the truth of one sentence entails the truth of another (in much the same way that semantic redundancy rules imply properties of words). As we know how to determine the truth of sentences, so do we know how to find the reference of noun phrases objects Fromkin and Rodman, 142-148].
Rules of language, however, can be broken often. Although a sentence may be syntactically correct, it is often devoid of proper semantic meaning (such as Noam Chomsky's famous phrase, "Colorless green ideas sleep furiously") [Noam Chomsky, Syntactic Structures, The Hague: Mouton, 1957 quoted in Fromkin and Rodman, 149]. Although anomalous in nature, semantic meaning is often broken for poetic imagery. More often, anomalous phrases are used metaphorically, where the reader is required to stretch his or her imagination to derive the proper meaning (such as Cervantes' statement, "Walls have ears"). Idioms, phrases which have specific meaning when in fixed forms, cannot be combined with other phrases using "normal" semantic rules. Specifically, they cannot be reformed ("bite your tongue" cannot be reformed into "bite the tongue which is yours" without losing meaning) - they are frozen in meaning [Fromkin and Rodman, 150-153].
While it is all well and fine to be able to create meaningful sentences, it is vital that they be meaningful in the context of a given discourse. Spoken communication is most often "telegraphic" in nature with verb phrases not being specifically mentioned, pronouns abounding, clauses dropped, and other "breaking" of semantic rules. In nearly all cases, there is contextual knowledge to fill in the missing gaps in communication to make the discourse cohere. Articles such as the and a determine whether a specific instance or just some instance of the referenced noun is being used ("a contract" or "the contract"). Yet, there are also rules of spoken conversation to ensure that the elisions of language do not interfere with the meaning of the words. The "maxims of conversation" (first discussed by H. Paul Grice in 1967) include the cooperative principle (a speaker's contribution to the discourse should be as informative as required, neither more nor less) and the maxim of relevance (a speaker's contribution should be relevant to the conversation) [Fromkin and Rodman, 154-158].