Lexical semantics

What is lexical semantics?

We introduced semantics as the study of meaning, but what does it mean to do lexical semantics? The word lexical refers to morphemes that are part of the lexicon and are not part of the grammar (though this distinction is not so easy to uphold in all cases). A simple phrase like the house consists of one word that is grammatical in nature, namely the, and another word that is lexical, namely house. We can assign a logical meaning to the, and it combines with words of a particular lexical category, namely nouns. The word house, on the other hand, does not have a logical meaning; it refers to a set of entities that share certain features, but these features might be hard to pin down, and they are certainly difficult to describe with some logical notation. In contrast to the which combines with nouns, house is a noun, i.e. it is a member of a lexical category. Lexical semantics, then, is the study of the meaning of words in the lexical categories – nouns, verbs, adjectives, and adverbs.

If lexical words cannot be demarcated with logical tools, how should we analyze them? Dirk Geeraerts says that “there can be no semantic description without some sort of decompositional analysis” (1989:588). When we seek to describe the meaning of house, we must decompose it at some level. The trouble is that there are all sorts of features we might decompose house into. Linguistic features are normally written in small caps in brackets and refer to a component of a linguistic item. By restricting the features to particular components of linguistic items, features that would only apply to the referent of a word in the real world become only tangentially related. For example, the features [+wall] or [+door] might be features of houses in the real world, but they are not necessarily features of the linguistic item house. The word house does, however, have certain linguistic features, such as [+count] referring to the possibility of counting houses. This allows us to say something like Simeon owns forty houses where the number forty requires a plural noun that can be counted. Using a [-count] noun results in ungrammaticality, e.g. * Simeon owns forty muds. Thus, decomposition of a lexical item is unfruitful unless certain meaning components are decomposed, namely grammatically-significant features of words.

Meaning components that are grammatically significant are what they sound like – they are components of meaning that affect the grammar, or how the lexical item combines with grammatical elements. We have already described grammatical morphemes as having logical meanings, so if a meaning component in a lexical item affects a grammatical item, it must be a logical property of the lexical item. This suggests that even though lexical items cannot be exhaustively described with logical properties, they nevertheless may be associated with certain logical properties. The mass vs. count distinction has already been mentioned in this respect; it refers to whether a noun can be counted or not (a logical property). Besides the [count] feature affecting whether the noun combines with numbers or not, it also affects grammatical items like the indefinite article, e.g. Josiah owns a house vs. * Josiah owns a mud. Whereas count nouns like house combine easily with the indefinite article a, mass nouns like mud do not. Thus, this is a component of meaning that the grammar can “see” in several ways, i.e. the grammar is affected by the [+count] feature, because it is itself a logical feature. The following sections are devoted to high-level overviews of the lexical categories and some of the basic components of meaning one should be familiar with to engage in serious analysis of the lexical items in that category. The lists are by no means exhaustive. Throughout, I provide references to important works on the various topics in footnotes.

How to study nouns

To study nouns, one should at least be familiar with the distinctions between mass and count nouns, proper and common nouns, and kind vs. entity interpretations. Contrasts are given in the following examples:

• (72) a. *Josiah owns a mud. (mass)
• b. Josiah owns a house. (count)

1 Work has been done to try to narrow the primitive features of lexical items to a manageable set, but it is difficult to judge whether these features are truly comprehensive or not and/or which are truly grammatically-significant. For the most valiant attempt at such a classification scheme, see the work of Anna Wierzbicka (1992, 1996, 1999); Wierzbicka and Goddard (2002).

• (73) a. Hosannah touched Hope. (proper)
• b. *Hosannah touched dog. (common)
• (74) a. Emmett fears a turtle (at the zoo). (entity)
• b. Emmett fears turtles (but not zebras). (kind)

We have already said that the mass-count distinction is a grammatically significant component of meaning. The basic distinction is that the indefinite article and plural morphemes do not combine easily with mass nouns, but count nouns are fine with either. The phrases a mud or muds are either ungrammatical or require very special contexts to be grammatical, so (72) is ungrammatical unless it is understood as something like a mud pile in context.

Proper and common nouns represent another basic distinction. Proper nouns in English can refer to entities without a definite or indefinite article, so a transitive verb can have a proper noun as an argument without an article of any kind, as in (73). Common nouns in English require an article to refer to an entity, so the same environment leads to ungrammaticality in (73). Because different languages have grammatical elements that behave slightly differently, we would not expect every language to have this same rule (Greek famously allows for the article with proper nouns). However, we would expect every language to have a distinction between these kinds of concepts that nouns refer to – namely whether a noun can itself refer to a specific entity (proper nouns) or whether it refers to a description of a class of entities (common nouns).

The kind-entity distinction is between reference to a class described by a noun (kind) and an individual member of that class (entity).4 In English, we use the bare plural (e.g. turtles) and the definite article (e.g. the turtle) to refer to kinds, though these are not the only interpretations of these constructions. In (74), a statement is made about a particular turtle that Emmett fears, namely one at the zoo (maybe a giant snapping turtle). He may fear that turtle but not fear turtles more generally. The kind interpretation in (74) in contrast refers to the entire class of animals we call turtle. Emmett generally fears members of this class, though there may be exceptions. As we can see, these two sentences have different truth-conditions – one may be false and the other true and vice-versa.

How to study verbs

Two main categories of grammatically-significant meaning components within verbs have dominated many discussions on verbal semantics: aktionsart and argument structure. The following examples represent just a few of the basic distinctions often noted in the literature:

• (75) a. Olivia ran *in/for three minutes. ((a)telicity)
• b. Maverick knows/*is knowing French. (stativity)
• c. Ruth reached/*was reaching the summit. (durativity)
• (76) a. The pot broke. vs. Elizabeth broke the pot. (unaccusative)
• b. The engine coughed. vs. *Ezra coughed the engine. (unergative)
• c. Esther gave the ball to Hannah/gave Hannah the ball. (ditransitive)

The examples in (75) refer to situations that differ with respect to their temporal makeup and the kind of eventuality referred to, and these affect how the verbs combine with other temporal elements in the clause, specifically adverbials and tense and aspect markers. These categories typically fall under the rubric of aktionsart, which is also sometimes called lexical aspect (see Semantics C for further discussion).5

One of the classic distinctions is between atelic predicates and telic predicates, which differ based on whether they have a natural endpoint or not (the former lacking an endpoint and the latter having an endpoint). In general, for -adverbials combine with atelic predicates, as in Olivia ran for three minutes, while telic predicates combine with in -adverbials, as in Olivia ran the race in three minutes. As these examples show, the verb run does not itself determine telicity – it allows for both readings depending on whether it has a specific type of argument or not. In other words, run is atelic by default, but since run the race patterns with telic predicates, the argument the race is creating an endpoint to the running event. Other verbs may be necessarily telic or necessarily atelic.

Stativity refers to whether the eventuality referred to by the verb contains a change or not. Stative verbs are homogeneous, i.e. they are the same throughout and lack change. The progressive be + -ing in English requires verbs that refer to changes over time, so it is generally disallowed with stative verbs. While Maverick knows French is fine, using the stative verb know with the progressive (i.e. * Maverick is knowing French) would reveal that the speaker does not really know English.

Because the progressive requires change over time, it also only combines straightforwardly with predicates that have time spans, i.e. predicates that are durative. A predicate like reach the summit happens in an instant. As soon as the individual gets to the top, the event happens and is over in the same moment. These are called achievement verbs, and they either do not combine with the progressive or receive a special interpretation (see also 9.4.3.2 and Semantics C). Although we can conclude from Ruth reached the summit that Ruth did arrive at the top of the mountain, we cannot conclude the same from Ruth was reaching the summit, since it can be followed with but she turned back near the end when her fingers turned blue. In each of these examples, the verb’s temporal characteristics affect how it interacts with the grammar of a language.

Verbs have arguments that complete the thought of the verb, and the structure of these arguments varies between different classes of verbs. Although The pot broke and The engine coughed are both intransitive verbs, one (break) allows for a causer to be added, while the other (cough) disallows the additional causer (see Hale and Keyser 2002). Elizabeth can break a pot, but Ezra cannot cough an engine. Verbs like break have been termed unaccusative, and verbs like cough unergative. These classes of intransitive verbs have several distinct grammatical properties.

Another much-discussed class is ditransitive verbs, such as give. These take two arguments and often participate in the “dative alternation” where the recipient can either come right after the verb (gave Mary …) or at the end with a to -phrase (gave … to Mary). Although many different classes of verbs participate in this alternation, they do not all do so in the same way. The verb give requires an animate recipient, but the verb send does not, as in Esther sent a ball to New York, but # Esther gave a ball to New York. Despite this contrast, send can be used identically to give in (76): Esther sent the ball to Hannah/sent Hannah the ball. All these variations in the representation of a verb’s arguments reflect grammatically-significant differences in verbal meaning.

How to study adjectives

One of the syntactic hallmarks of adjectives is that they can modify nouns, but just as nouns may refer to different things, adjectives may also modify nouns in different ways, as illustrated in (77). Semantically, many adjectives make reference to scales, graded values of a certain property, such as height, and these scales are also grammatically significant as demonstrated in (78). Finally, some adjectives like tall and short have different interpretations when used with measure phrases like five feet:

• (77) a. Rachel is a beautiful girl. (intersective)
• b. Rachel is a beautiful dancer. (subsective)
• (78) a. Sarah is completely full.
• b. *Sarah is completely tall.
• (79) a. Leah is five feet tall.
• b. ?Leah is five feet short.

In (77a), the adjective beautiful is interpreted intersectively. This means that the statement allows us to conclude that Rachel is both a girl and she is beautiful. In contrast, (77b) can have either the intersective reading where Rachel is both beautiful and a dancer (though her dancing may be atrocious), or a subsective reading where Rachel is someone who dances beautifully (though she may not be a beautiful person). The subsective interpretation arises because dancer can be interpreted as a description of a person or someone engaging in an activity, and beautiful can modify a person or how someone engages in the activity of dancing. A very similar adjective like pretty does not have the subsective reading because it cannot describe how someone dances, i.e. we would not say * Rachel dances prettily. The basic takeaway from these facts is that adjectives may modify nouns in different ways depending on the nature of both the noun and adjective and how they interact.

A scale is an ordered set of values. For example, we might have a scale from 1 to 5. In this set of numbers, there is an order. 2 is more than 1, and 3 is more than 2. There is also a minimum of the scale, namely 1, as well as a maximum, namely 5. Thus, this scale has certain properties: it is ordered, it has a set of values, and it has a minimum and maximum. Although not all adjectives are scalar, many of them are (an example of a non-scalar adjective is something like wooden). Scalar adjectives have different grammatical properties depending on the nature of the scale they refer to. While an adjective like full has a closed scale, tall has an open scale. A closed scale means that there is an ending value on the scale, i.e. a minimum or maximum; an open scale means there is no such value. For full, the endpoint on the scale is when some container has reached its capacity to hold something. Once this has been reached, nothing else will fit into the container. When a person like Sarah is the container, her being completely full normally means she cannot consume anything else. Since completely would modify the end of the scale and tall has no such end, * Sarah is completely tall is ungrammatical.

The contrast between (79a) and (79b) is between positive and negative adjectives. Positive adjectives combine easily with measure phrases, but crucially, they do not entail that the bare form of the adjective holds. We cannot conclude from Leah is five feet tall that Leah is tall. In fact, most would conclude the opposite, namely that she is short. In contrast, negative adjectives like short make reference to the same exact scale, but they usually do not combine with measure phrases. When they do, they do entail that the bare adjective holds. Leah is five feet short (which we would only say in very specific contexts) can only mean that she is five feet tall and that she is also short. Although we have only scratched the surface of the issues involved, such contrasts demonstrate how the scale structure of adjectives affects their grammatical behavior.

Doing lexical semantics

We began this section with a question that we have not adequately answered: what does it mean to do lexical semantics? We know what we are trying to do – determine grammatically-significant components of meaning – but we have not yet shown how to do this with a real word in the Greek NT. This is illustrated in Exegesis A, but the following steps can serve as a guide to engaging in lexical semantic analysis:

1. Gather all the data into something like a spreadsheet, so that notes can easily be made for each example.

2. Note the syntactic context of each example at an appropriate level of granularity, e.g. if a noun, what modifiers does it have, is it in a prepositional phrase, is it serving as an argument of a verb, etc.

• a. Are there any obvious grammatical patterns that are known to be reflected in the semantics? e.g. the noun never takes the plural (so it is probably a mass noun).

3. Determine the interpretation of the easy cases in context to come up with a list of possible senses.

4. Do the examples with one sense have any grammatical similarities?

5. Based on 4, go back and look at the more difficult cases in 3.

6. Repeat steps 3–5 over and over again until you know how to interpret the word in all contexts of use.