Show Summary Details

Page of

PRINTED FROM the OXFORD RESEARCH ENCYCLOPEDIA, LINGUISTICS (linguistics.oxfordre.com). (c) Oxford University Press USA, 2016. All Rights Reserved. Personal use only; commercial use is strictly prohibited. Please see applicable Privacy Policy and Legal Notice (for details see Privacy Policy).

date: 23 September 2017

Lexical Semantic Framework for Morphology

Summary and Keywords

The central goal of the Lexical Semantic Framework (LSF) is to characterize the meaning of simple lexemes and affixes and to show how these meanings can be integrated in the creation of complex words. LSF offers a systematic treatment of issues that figure prominently in the study of word formation, such as the polysemy question, the multiple-affix question, the zero-derivation question, and the form and meaning mismatches question.

LSF has its source in a confluence of research approaches that follow a decompositional approach to meaning and, thus, defines simple lexemes and affixes by way of a systematic representation that is achieved via a constrained formal language that enforces consistency of annotation. Lexical-semantic representations in LSF consist of two parts: the Semantic/Grammatical Skeleton and the Semantic/Pragmatic Body (henceforth ‘skeleton’ and ‘body’ respectively). The skeleton is comprised of features that are of relevance to the syntax. These features act as functions and may take arguments. Functions and arguments of a skeleton are hierarchically arranged. The body encodes all those aspects of meaning that are perceptual, cultural, and encyclopedic.

Features in LSF are used in (a) a cross-categorial, (b) an equipollent, and (c) a privative way. This means that they are used to account for the distinction between the major ontological categories, may have a binary (i.e., positive or negative) value, and may or may not form part of the skeleton of a given lexeme. In order to account for the fact that several distinct parts integrate into a single referential unit that projects its arguments to the syntax, LSF makes use of the Principle of Co-indexation. Co-indexation is a device needed in order to tie together the arguments that come with different parts of a complex word to yield only those arguments that are syntactically active.

LSF has an important impact on the study of the morphology-lexical semantics interface and provides a unitary theory of meaning in word formation.

Keywords: morphology, lexical semantics, semantic features, decomposition, co-indexation

1. Goals, Assumptions, and Priorities1

The Lexical Semantic Framework (LSF, Lieber, 2004) is concerned with the study of the semantics of word-formation processes. Its central goal is to characterize the meaning of simple lexemes and affixes and to show how these meanings can be integrated in the creation of complex words.

LSF offers a systematic treatment of issues that figure prominently in the study of word formation:

  1. (a) The polysemy question: why do derivational affixes often exhibit polysemy? Affixation of -er, for instance, creates nouns that have several interpretations: (i) agent (driver), (ii) instrument (opener), (iii) experiencer (hearer), (iv) stimulus (pleaser), (v) measure (fiver), (vi) denominal noun (villager), (vii) patient/theme (keeper), and (viii) location (diner).

  2. (b) The multiple-affix question: why are there affixes that create the same kind of derived words (both -er and -ant create agent nouns, e.g. writer, driver, servant, accountant)?

  3. (c) The zero-derivation question: how do we account for zero-affixation, that is, semantic change with no (overt) formal change (i.e., “conversion”)?

  4. (d) The form and meaning mismatches question: why are there instances where the form and meaning correlation is often not one-to-one?

In order to tackle these issues, LSF is characterized by four general features. First, it is decompositional. Thus, it makes use of primitives (atoms) of the right “grain size” (Lieber, 2004, p. 4). These primitives are features which make up the lexical-semantic representation of both simplex and complex lexemes. Second, it is lexical by nature. Thus, it is designed in a way that allows a treatment of the semantic properties of words (lexical-semantic properties) as opposed to semantic properties of other levels (e.g., phrases and discourses). Third, it is cross-categorial enough, in order to allow an in-depth analysis of all categories, such as nominals and verbs. Finally, given that word-formation processes serve to extend the simplex lexicon, LSF covers both simplex and complex words. Thus, LSF allows one to deal with the semantics of simplex and complex words in a parallel way.

In the following, we delve more deeply into the way LSF accounts for the semantic properties of both simplex and complex words. Section 2 “Semantic features: Skeleton and body.” presents the basic architecture of LSF. It introduces the semantic features used in LSF, and the way these features are distributed amongst the Grammatical Skeleton and the Pragmatic Body. Section 3 “Major ontological classes.” focuses on the two major ontological classes, namely substances/things/essences and situations. The remaining sections present the lexical-semantic properties of morphologically complex formations and show how the principle of co-indexation ties together the arguments of morphological building blocks. Section 4 “Derivation in LSF.” focuses on derivation, and Section 5 “Compounding in LSF.” on compounding. Section 6 “Form and meaning mismatches.” offers a treatment of form and meaning mismatches.

2. Semantic Features: Skeleton and Body

LSF has its source in a confluence of research approaches that follow a decompositional approach to meaning (Jackendoff, 1990; Pustejovsky, 1995; Szymanek, 1988; Wierzbicka, 1996). In this respect, LSF defines simple lexemes and affixes by way of a systematic representation that is achieved via a constrained formal language that enforces consistency of annotation.

Given that LSF applies a decompositional approach to meaning, it makes use of a repository of universal semantic features to which every particular language has access. The following 18 features form part of the set of universal semantic features (Lieber, 2009, pp. 85–86): material, dynamic, I(nferable) E(ventual) P(osition of) S(tate), C(omposed of) I(ndividuals), B(ounded), Loc(ation), scalar, animate, human, female, age, artifact, n dimension, orientation, consistency, function, contact, and motion with respect to focal point. These features make up the lexical-semantic representation of lexemes in a compositional manner.

Lexical-semantic representations in LSF consist of two parts: the Semantic/Grammatical Skeleton and the Semantic/Pragmatic Body (for similar claims in the literature, see Rappaport Hovav and Levin (1996, 1998) and Mohanan and Mohanan (1999)). Depending on their syntactic relevance, the universal semantic features can be part of the skeleton or the body of a lexeme in a given language.

The distinction between skeleton and body is a key aspect of the organization of lexical- semantic representations in LSF. The skeleton is fully formalizable, decompositional, hierarchically arranged, contains those aspects of meaning relevant to syntax, and is stable from speaker to speaker. All speakers, for example, are expected to share the same skeletal information for particular lexemes.

The body is partially formalizable and systematic, and consists of two parts (Lieber, 2009, p. 83). The first featural part of the body comprises the universal features that are semantically but not syntactically active in a given language. The second part of the body encodes all those aspects of meaning that are perceptual, cultural, and encyclopedic, such as color, function, and dimension. The first part of the body is largely stable from one speaker to the other, whereas the information encoded in the second part diverges.

In order to better understand the makeup of lexical-semantic representations in LSF, consider, for example, the lexeme piston. All speakers share the skeletal information that piston is a common noun (i.e., a substance/thing/essence). Speakers also share the featural bodily information that piston is human-made (i.e., an artifact) and that it has a function. The second part of the body of the lexical-semantic representation of piston, however, is the locus of specialized user knowledge and, thus, it diverges from one speaker to another. Some speakers, for instance, may know that pistons have something to do with cars, others may know that pistons are parts of engines, and others may even know that pistons are parts of reciprocating engines, are located in cylinders, etc.

As far as English is concerned, there are at least seven semantic features relevant to syntax and necessary for word formation. As illustrated in (1), features that are encoded into the skeleton are presented in square brackets. Features that are part of the body are enclosed in angle brackets (from Lieber, 2009, p. 85):

(1)

Lexical Semantic Framework for Morphology

The seven syntactically active features of English which are needed for the analysis of lexical meaning are defined as follows (from Lieber, 2009, p. 80):

(2)

Lexical Semantic Framework for Morphology

Features in LSF are used in (a) a cross-categorial, (b) an equipollent, and (c) a privative way. This means that they are used to account for the distinction between the major ontological categories, may have a positive or a negative value (binary value), and may or may not form part of the skeleton of a given lexeme. With respect to the privative use of these features, consider for example the feature [material]. This feature may be used in the lexical-semantic representation of nouns, but it is not relevant to the discussion of the semantics of verbs or adjectives. Thus, verbs and adjectives should not be characterized by this feature in their lexical-semantic representation.

3. Major Ontological Classes

3.1. substances/things/essences and situations

The features [material] and [dynamic] classify items into the two major ontological classes as substances/things/essences and as situations. Items with the feature [material] as the outermost function of their skeleton belong to the class of substances/things/essences, and items with the feature [dynamic] as their outermost function are situations.

The binary use of these features captures further distinctions that manifest themselves in these two classes. The presence of a positive or a negative value of the feature [material] derives the distinction between concrete and abstract nouns as shown in Figure 1:

Lexical Semantic Framework for MorphologyClick to view larger

Figure 1: substances/things/essences in LSF.

The positive value of the feature [dynamic] characterizes the subclass of events or processes and the negative value flags the subclass of states, as illustrated in Figure 2:

Lexical Semantic Framework for MorphologyClick to view larger

Figure 2: situations in LSF.

Figure 2 informs us that, in LSF, happy belongs to the class of situations. In particular, adjectives in LSF are characterized by the features [−dynamic] and [scalar]. Thus, they are instantiations of the major ontological class of situations and, more specifically, of the subclass of states, since they are characterized as [−dynamic]. They differ from stative verbs in that they bear the feature [scalar]. As shown in Figure 3, gradable adjectives such as red are characterized as [+scalar] and nongradable adjectives, such as dead, are [−scalar].

Lexical Semantic Framework for MorphologyClick to view larger

Figure 3: [+scalar] and [−scalar] states in LSF.

The features [material] and [dynamic] are not mutually exclusive, since there are substances/things/essences that are “processual in flavor, denoting states, events, actions, or even relations of some sort, and also those which lack a processual flavor” (Lieber, 2004, p. 26). The difference between the processual mother, which involves having or caring for a child, and money, which lacks this processual aspect, is given in Figure 4 (adapted from Lieber, 2004, p. 27):

Lexical Semantic Framework for MorphologyClick to view larger

Figure 4: Processual substances/things/essences in LSF.

3.2. Formalizing Lexical-Semantic Representations

In this section, we focus on the inner structure of the lexical-semantic representation of lexemes in LSF and present some of the major characteristics (i.e., aspectual and quantitative characteristics) of substances/things/essences and situations.

Features in LSF act as functions and may take arguments. In particular, the skeleton comprises a function and one or more arguments predicated of that function (3a). Like Jackendoff’s (1990) Lexical Conceptual Structures, functions and arguments of a skeleton are hierarchically arranged as illustrated in (3b):

(3)

Lexical Semantic Framework for Morphology

Let us apply this general format to substances/things/essences and situations.

3.2.1. Substances/Things/Essences

The skeletal part of lexemes denoting substances/things/essences is characterized by at least one feature, namely [material], and the so-called “R” argument. This argument suggests “referential” and is involved in referential uses of NPs (Williams, 1981, p.86). It is an external argument and it is called referential because it is at the same time an argument of the predicate and its referent. In this respect, the concrete table and the abstract time are treated as one-place predicates that carry the “R” argument. Their skeletal representations are given in (4a) and (4b) respectively:

(4)

Lexical Semantic Framework for Morphology

Relational substances/things/essences are treated as two-place predicates.3 Thus, leg (e.g., the leg of X) and friend (e.g., the friend of X) in (5) carry two arguments in their skeleton. In these cases, the highest argument of a relational noun is the “R” argument, and the second argument is the possessor argument:

(5)

Lexical Semantic Framework for Morphology

Consider now the lexical-semantic representation of the substance/thing/essence lexeme mother in (6):

(6)

Lexical Semantic Framework for Morphology

The skeleton in (6) reads as follows: mother is a concrete (i.e., [+material]) processual (i.e., [dynamic]) substance/thing/essence. In addition, it is a functional two-place predicate (i.e., mother of X). Thus, it carries two arguments (an “R” argument and a possessor argument).

So far, we have seen how the semantic features of the skeleton can be described in this framework, but as we said, the representation of a lexical item consists of two parts, the skeleton and the body. In the following schemata, I give the partial lexical-semantic representations of the words author and bed (adapted from Lieber, 2009, p. 86):

(7)

Lexical Semantic Framework for Morphology

The examples in (7) illustrate that the complete lexical-semantic representation of a lexeme in LSF consists of three parts. The first part is the skeleton. The skeleton of author consists of the features [+material] and [dynamic] and two arguments (e.g., author of a book). Thus, author is a concrete processual substance/thing/essence. The skeleton of bed is characterized by the feature [+material] and one argument, the “R” argument. The second part of the representation is the systematic part of the body and consists of those semantically active features which are, nevertheless, syntactically inactive. This part informs us that author is <+animate> and <+human> and has a <function>. In a similar vein, bed is <−animate>, <+artifact>, <3 dimension>, and <horizontal> and has a <function>. The third part of the representation is the part of the body in which encyclopedic information about the lexical item is provided. Thus author {writes for publication, …} and bed is {for sleeping, …}.

3.2.2. Situations

situations are those items that carry the feature [dynamic] as the outermost function of their skeleton. (8a) provides the skeleton of the event snore and (8b) gives the skeleton of the state know:

(8)

Lexical Semantic Framework for Morphology

Observe that snore has one argument in its skeleton, since it is a one-place predicate. The skeleton for know comes with two arguments, since this verb is a two-place predicate. It has both an internal and an external argument (e.g., I know the answer.).

The addition of the feature [I(nferable) E(ventual) P(osition of) S(tate)]4 to the skeleton of a verb captures further distinctions, as illustrated in Table 1:

Table 1: Aspectual classification of verbs.

[+dynamic ([ ], [ ])]

simple activity verb

do

[+dynamic, +IEPS ([ ], [ ])]

change of state

break

[+dynamic, +IEPS ([ ], [ ])]

change of place

ascend

[+dynamic, −IEPS ([ ])]

manner of change

walk

[IEPS] accounts for the major aspectual classes of verbs and signals the addition of a sequence of Places or States. That is, it adds a Path component of meaning. In particular, the presence of this feature with a positive value means that the path is directed. Consider, for example, that the [+IEPS] verbs brake and ascend imply a Path in which there is progression from one point to the other but the initial and final points are not the same. A negative value signals a random path. This state of affairs is exemplified by walk, in which we can make no inference with respect to the relationship between the initial and final points (they might be the same).

As presented in the previous section, adjectives belong to the class of situations, and, in particular, to the class of states. (9) provides the skeletal part of the adjectives red, dead, and fond. Gradable adjectives are characterized as [+scalar], and nongradable adjectives are flagged by the feature [−scalar].

(9)

Lexical Semantic Framework for Morphology

Like verbs and nouns, adjectives can be either one-place predicates or two-place predicates. The adjectives red (9a) and dead (9b) are one-place predicates, and fond (9c) is a two-place predicate (e.g., fond of music).

3.3. Quantity

An important aspect of the semantics of both substances/things/essences and situations is quantity. In LSF, quantitative semantics—that is, duration, internal individuation, and boundaries of substances/things/essences and situations—can be characterized by the same features, namely, [B(ounded)] and [C(omposed of) I(ndividuals)]. The use of the same set of features for both classes follows from the cross-categorial characteristic of LSF, since semantic features can be used for the discussion of various categories. It should be mentioned that [B] and [CI] are meant to handle only those quantitative characteristics of meaning that manifest themselves at the lexical level (as opposed to quantitative characteristics of other levels, such as phrases and discourses).

In particular, as far as substances/things/essences are concerned, the feature [B] can be used to distinguish between count [+B] and mass [−B] nouns. In addition, the feature [CI] is used for the distinction between nouns that are not composed of discernible replicable parts (i.e., [−CI]) and aggregates, which are characterized as [+CI]. Consider the following from Lieber (2004, p. 137):

Table 2: Classification of nouns based on quantitative semantics.

[+B, −CI]

singular count nouns

apple

[−B, −CI]

mass nouns

luggage

[+B, +CI]

group nouns

flock

[−B, +CI]

plural nouns

sheep

The noun apple, for example, is characterized as [+B, −CI] because it is an individual noun (count noun) and at the same time is not composed of discernible replicable parts. The word luggage bears the specification [−B, −CI] in its skeleton, since it is a mass noun composed of discernible replicable parts. As a group noun, flock is characterized by the features [+B, +CI]. Finally, the plural noun sheep is a conglomeration of similar individuals and is, therefore, characterized by the features [−B, +CI] in its skeleton.

With respect to situations, [B] and [CI] capture quantitative and aspectual characteristics. [+B] events are those verbs which may have duration, but their duration is not linguistically significant, whereas [−B] events are those verbs whose duration is linguistically significant. A temporally punctual situation (i.e., a [+B] event) such as explode and a temporally durative one (i.e., a [−B] event) such as walk behave differently with adverbials, as in (10) (from Lieber, 2004, p. 138):

(10)

Lexical Semantic Framework for Morphology

The examples in (10) show that language makes a distinction between those events that are punctual and those events that are durative; the durative adverbial “for an hour” can be used with the [−B] walk but not with the [+B] explode.

The use of the feature [CI] with respect to events captures the distinction between iterativity and homogeneity. In particular, the use of [CI] with a positive value is the equivalent of plurality in nouns. Just as plural nouns are composed of discernible replicable parts, [+CI] events such as pummel denote repeated actions of the same sort; pummel means ‘to produce repeated blows’. The use of [CI] with a negative value with respect to events corresponds to nonplural nouns (single individuals or mass nouns). Therefore, [−CI] situations are those events which are not composed of multiple, repeated actions of the same sort. The following summarizes the various aspectual event classes:

Table 3: Classification of verbs based on quantitative semantics.

[+B, −CI]

nonrepetitive punctuals

explode

[−B, −CI]

nonrepetitive duratives

ascend

[+B, +CI]

<logically impossible>

[−B, +CI]

repetitive duratives

pummel

A [+B, +CI] situation is not possible, since this would mean that an event could be both punctual and composed of multiple, repeated, and identical actions.

4. Derivation in LSF

The previous section was devoted to the study of some salient characteristics of simplex lexical items. In the following sections, we delve more deeply into the study of the way the theoretical apparatus of LSF accounts for the semantics of complex words.

The first issue which arises in derivation is how we should treat elements below the level of word, that is, affixes. In LSF, affixes have a skeleton, and their semantic contribution can be accounted for by the same semantic features which are needed for the description of the semantics of simplex words. Consider, for example, the affix -er:

(11)

Lexical Semantic Framework for Morphology

The semantic contribution of the affix -er can be described as the addition of the features [+material] and [dynamic] to a base. More specifically, -er creates concrete and processual substances/things/essences. Notice, however, that although affixes have a skeletal part, the semantic content of an affix is abstract and underdetermined, since affixes have no (or little) body. This is a source for polysemy in word formation. That is, the semantic contribution of the affix can be spelled out in several ways when the affix is combined with the more semantically robust base and deployed in context.

4.1. S-selection

The lexical-semantic representation of an affix contains information about the type of lexemes that could serve as bases for the derivation. A basic tenet of LSF is that selection is primarily semantic (s-selection) and not categorial (c-selection) (Lieber, 2007). That is, affixes select for the semantic category rather than the syntactic category of their bases (see Plag, 2004 for a similar proposal).

A comparison between the traditional c-selection hypothesis and the hypothesis that selection is primarily semantic shows that s-selection allows us to capture subtle selectional properties of affixes that are difficult to model in terms of c-selection. In order to best account for the selectional properties of -ship, for instance, it does not suffice to merely mention that this affix attaches to nouns (e.g., stewardship, accountantship). One must specify that -ship shows a strong preference for concrete processual nouns, that is, substances/things/essences that are [+material, dynamic].5

LSF allows us to model even more subtle selectional properties as in the lexical-semantic representation of a prefix such as re-. In terms of c-selection, re- attaches to verbs, but an analysis that takes into consideration the semantic properties of the base can account for the fact that re- does not attach to all verbs (e.g., *reknow, *reexplode). This prefix only attaches to a well-defined class of situations with the skeleton in (12):

(12)

Lexical Semantic Framework for Morphology

The representation in (12) informs us that re- attaches to [+dynamic] situations, that is, events, that are not inherently iterative (i.e., [−CI]), may be inchoative or unaccusative (i.e., [+IEPS]), and have an argument which can be interpreted as a Path or Result that is reversible (i.e., [Path. State: stage level]). The three dots indicate that there may be other arguments in the base as well. With respect to the Path argument, Lieber (2004, 2007) argues that the distinction between stage- and individual-level items (Carlson, 1977) captures the difference between permanent and nonpermanent results in situations. In particular, stage-level results are temporary and, consequently, reversible, whereas individual-level results cannot be reversed.

The foregoing discussion shows that the theoretical apparatus of LSF allows one to model the selectional properties of affixes in a detailed manner (e.g., prefixation of re-) and to take into consideration that even within the same class, an affix may show a strong preference for a specific subclass of bases to attach to (e.g., -ship).

4.2. The Principle of Co-indexation

The creation of a morphologically complex word involves not only the combination of two (or more) morphemes on a structural level, but also the integration of distinct morphemes on a semantic level. In order to account for the fact that several distinct parts, an affix and a lexical base in the case of derivation and two lexemes in the case of compounding, integrate into a single referential unit, Lieber (2004) introduces the Principle of Co-indexation. This is “a device we need in order to tie together the arguments that come with different parts of a complex word to yield only those arguments that are syntactically active” (Lieber, 2004, p. 45). Co-indexation reads as:

(13)

Lexical Semantic Framework for Morphology

The highest argument of the skeleton is the argument of the outermost lexical function of the head. In order to identify the highest argument of the skeleton in complex formations, consider the schemata below:

(14)

Lexical Semantic Framework for Morphology

(15)

Lexical Semantic Framework for Morphology

The schema in (14) illustrates the mechanism of concatenation of functions which accounts for compounding, and the schemata in (15) show the subordination of functions which is evident in affixation. As far as compounding is concerned, the highest argument in (14) is the argument of the lexeme [β F2].

Observe that affixation in (15) is accounted for by two subschemata. As will be seen in section 4.3 “Co-indexation in derivation,” (15a) accounts for affixes that carry an argument in their skeleton. In these cases, the highest argument of the skeleton in (15a) is the argument of the affix, that is, the argument of [αF1]. As proposed by Andreou (2014, 2015), affixation that involves subordination of functions without indexing of arguments is regulated by (15b). In this schema, the highest argument of the skeleton is the argument of the base, that is, the argument of [αF2].

4.3. Co-indexation in Derivation

In this section, we examine the way the mechanism of co-indexation ties together the arguments of the base and the affix in derived words. Given that affixes have their own skeleton, affixation involves the addition of this skeletal material as the outermost function to the skeleton of a base, thereby subordinating the base in question, as schematically shown in (15).

In (15a), both the base [β F2] and the affix [αF1] come with open argument positions in their skeletal part. Thus, the principle of co-indexation is needed in order to regulate the interaction between the arguments of the base and the affix.

The derivation of driver, for example, involves the co-indexation of the highest argument of the nonhead, which is the verb drive, with the only argument of the head, which in this particular case is the affix -er. The skeletons of -er and drive, as well as the application of the principle of co-indexation, are illustrated below:

(16)

Lexical Semantic Framework for Morphology

Since there are no semantic conditions on the head argument, the highest argument of the nonhead, in this particular case the external argument of the verb drive, is co-indexed with the highest (and only) unindexed argument of the head, that is, the “R” argument of -er. Given that co-indexation accounts for the referential properties of complex items, the result of the co-indexation process is that the derived word is interpreted as bearing the role of the external argument of the verb; in this case, it is an agent.

As noted in the definition of the principle of co-indexation in (13), indexing must be consistent with semantic conditions on the head argument, if any. Although the combinatorial properties of -er do not impose such semantic conditions, other affixes such as -ee come with specific semantic requirements in their skeleton. Consider the lexical-semantic representation of -ee:

(17)

Lexical Semantic Framework for Morphology

It follows from the representation in (17) that -ee places specific semantic conditions on its co-indexed argument. In particular, it requires to be co-indexed with a sentient and nonvolitional argument (Barker, 1998). Consider now the derivation of the word employee (Lieber, 2004, p. 63):

(18)

Lexical Semantic Framework for Morphology

Given that the first argument of the verb employ is always volitional, the only argument of the affix must be co-indexed with the second argument of the base, namely the patient argument, which is always nonvolitional. From this follows the theme reading of the word employee.

The mechanism of subordination of functions, which regulates affixation, does not always involve indexing of arguments. Consider, for example, the negative prefix in- on adjectives (Andreou, 2015). This prefix does not carry an argument in its skeleton, as shown in (19). Thus, the contribution of in- is characterized by the feature [−Loc], which flags negation. The use of the [−Loc] feature in LSF is equivalent to the negation operator “¬”.

(19)

Lexical Semantic Framework for Morphology

Given that the prefix does not carry an argument in its skeleton, the only argument present in the schemata in (20a) and (20b) is the argument of the base, that is, the adjective. Thus, these cases of derivation are accounted for by the schema in (15b), which regulates subordination of functions without indexing of arguments.

(20)

Lexical Semantic Framework for Morphology

The schemata in (20) can inform the discussion on a key aspect of word formation, that is, “constructional polysemy” (Copestake & Briscoe, 1996; Pustejovsky & Boguraev, 1996). This is the kind of “polysemy that follows from a single skeleton which is interpreted in a number of ways depending upon the bases with which it combines” (Lieber, 2004, p. 89). In other words, the different readings of in- in (20) do not follow from distinct representations, but from the interaction of the general negative meaning of this prefix with the properties of the bases it attaches to. In particular, the schemata in (20) capture the generalization that in- delivers contrary readings with gradable adjectives (e.g., inappropriate) and contradictory readings with nongradable adjectives (e.g., inanimate). Whether in- results in a contrary or contradictory reading is a matter of the gradability of the base it attaches to. The adjective appropriate is gradable (i.e., [+scalar]), and thus between appropriate and inappropriate there can be intermediate states. In addition, both appropriate and inappropriate can be false at the same time. The adjective animate, however, is nongradable (i.e., [−scalar]), and, as a result, there can be no middle state between animate and inanimate. From this follows the contradictory reading of inanimate.

To sum up, co-indexation is a mechanism that operates on the arguments of the base and the affix and determines the referential properties of the derived word. The fact that co-indexation must be consistent with semantic conditions on the head argument accounts for the agentive reading of driver and the object-oriented meaning of employee. The affix -er most productively selects for verbs whose external argument is volitional, and -ee has scope over the sentient and nonvolitional argument of the verb it combines with. There are affixational processes, nevertheless, that do not involve co-indexation of arguments. As illustrated by prefixation of in-, in these cases, the affix does not carry an argument in its skeleton. Finally, the underdetermined semantic content of affixes is a source for polysemy in word formation. The different readings of words derived with the same affix do not follow from distinct representations of the affix, but from the interaction of the general meaning of the affix with the properties of the bases it attaches to.

5. Compounding in LSF

Compounds in LSF are formed by concatenation of skeletons with concomitant co-indexing. Thus, compounding involves the composition of bases. In what follows, we present the way co-indexation works in the three compound types, namely, argumental, coordinate, and attributive compounds (Bauer, Lieber, & Plag, 2013).

5.1. Argumental Compounds

Argumental compounds in LSF are those compounds in which there is an argumental relation between the head and the nonhead. In synthetic compounds, such as bus driver, the nonhead is interpreted as an argument of the verbal base (i.e., the head). Consider for example the compound burrito assembler (from Lieber, 2010, p. 135):

(21)

Lexical Semantic Framework for Morphology

This compound is formed as follows. The first step involves the derivation of the word assembler. The affix -er has no semantic conditions with respect to the argument it co- indexes with, and, as a result, the “R” argument of the affix co-indexes with the highest argument of the verb, the external argument. The second step is the co-indexing of the “R” argument of the nonhead burrito with the only unindexed argument of the verb, the internal argument. From this follows the object-oriented reading of the compound.

5.2. Coordinate Compounds

In coordinate compounds, the compound members bear equal semantic weight (Bauer et al., 2013, p. 479). In order to account for the semantics of coordinate compounds, one must take into account the complete lexical-semantic representations of the compound members, since the nearly identical features of the compound members allow for the complete identification of reference. Consider, for example, the compound actor author:

(22)

Lexical Semantic Framework for Morphology

Consider, first, the skeletal part of the two lexemes. In (22), the lexemes actor and author have identical skeletons. Both are concrete processual substances/things/essences. In accordance with the principle of co-indexation, the “R” arguments of the two lexemes are co- indexed. Consider now the body part of the two lexemes. Observe that the first featural part of the two lexemes is identical. That is, both lexemes are <+animate>, <+human>, and have a <function>. The information in the second encyclopedic part, however, diverges; actor {performs in a film, …}, whereas, author {writes for publication, …}.

The analysis shows that the coordinative interpretation is the result of compounding two lexemes that have nearly identical lexical-semantic representations. Only encyclopedic knowledge differs from one lexical item to the other. In particular, co-indexation of the “R” arguments of actor and author, as well as the compatibility of the features of the skeleton and the formal part of the body, allows for the complete identification of reference.

5.2. Attributive Compounds

In attributive compounds, there is a modification relation between the head and the nonhead. This category of compounding in LSF is the default category, since attributive compounds involve neither argumental relation between the head and the nonhead (subordinate type) nor identification of reference (coordinate type). Consider, for example, the compound bamboo bed:

(23)

Lexical Semantic Framework for Morphology

Both bamboo and bed in (23) carry an “R” argument. Thus, the principle of co-indexation applies in order to tie together the “R” arguments of the two lexemes. Although the skeletons of bamboo and bed are identical, the body parts of the two lexemes are not compatible.

Thus, a coordinative interpretation is ruled out. In addition, there is no relation of subordination between the head and the nonhead. This rules out a subordinative interpretation. Therefore, the interpretation of bamboo bed is a matter of the speaker finding a plausible relationship between the two lexemes. In this example, the relationship between the head and the nonhead depends on the encyclopedic knowledge that bamboo has a <function>, namely, it can be used as a material for <+artifacts>. From this follows the modification relation “made of” in bamboo bed.

5.3. Exocentricity

Exocentric compounds do not introduce any extension to the framework. In particular, the lexical-semantic representation of exocentric compounds is not different in principle from the representation of endocentric compounds. Thus, exocentricity is attributed to general grammatical principles and mechanisms that are not specific to compounding. The exocentric birdbrain and the endocentric bamboo bed, for example, exhibit the same relationship between their members. That is, the relationship between their members is one of attribution. They differ only with respect to the fact that the exocentric birdbrain is interpreted metonymically (Bauer, 2008, 2010). In particular, birdbrain is based on the part for whole metonymy, and thus a part of a person (i.e., brain) is used to denote the person in question.

6. Form and Meaning Mismatches

Mismatches between form and meaning are evident in conversion (e.g., hammerN hammerV), derivational redundancy (e.g., re-rewrite), semantic subtraction (e.g., ritualistic), and in cases of empty morphs (e.g., orient-at-ion).

Conversion is defined as semantic change without concomitant formal change. In English, there is Noun-to-Verb conversion (e.g., hammerN hammerV), Verb-to-Noun conversion (e.g., throwV throwN), and Adjective-to-Verb conversion (e.g., coolA coolV). English conversion in LSF is best analyzed as “relisting in the lexicon” (Lieber, 1992, 159) rather than the addition of phonologically null affixal material (Don, 1993; Hale & Keyser, 2002). A comparison between the products of the highly productive -ize and conversion reveals that converted items show a greater semantic diversity. Consider, for example, hammer and bark (from Plag, 1999, p. 220). The verb to hammer belongs to the instrumental category, and the verb to bark belongs to the privative category. These categories, crucially, are never expressed by the affix -ize and its neologisms. Given the broad range of meanings expressed by conversion in English, we should not collapse conversion and affixation. Thus, a conversion verb such as hammer in LSF is analyzed as a noun that gets relisted in the mental lexicon as verb.

Let us turn to derivational redundancy. In LSF there is no principled constraint on the recursive attachment of affixes with the same meaning. Expressing the same content more than once in the same word is possible when useful and meaningful. Thus, derivational redundancy is both useful and meaningful in re-rewrite, for example, in which the iterative meaning of rewrite is intensified by the “redundant” prefixation of the same prefix, re-. In cases of semantic subtraction, a derivational affix attaches to an already affixed word, but the meaning of the affix closest to the base does not contribute to the meaning of the whole (Lieber, 2004, p. 174). Consider, for instance, [[[ritual] ist] ic]. Although the affix -ic creates relational nouns, the meaning of ritualistic is not ‘pertaining to a ritualist’ but rather ‘pertaining to ritual’. The meaning of -ist is in a way canceled as a result of affixation of -ic. In LSF, cases of semantic subtraction are accounted for by allomorphy. Thus, -istic in ritualistic is considered an allomorph of -ic.

In a similar vein, empty morphs, that is, forms that lack semantic content, are also amenable to treatment as cases of allomorphy. Consider, for example, that -at- in orientation adds no meaning to the whole. In LSF, forms such as -at- are analyzed as cases of allomorphic variation of either the base or the affix (for more on this issue, see Bauer et al., 2013, p. 181).

7. Looking Back, Looking Forward

The focus of generative morphological theory has long been on examining the formal rather than the semantic properties of lexical items. LSF is a framework of meaning in word formation and offers a systematic treatment of the way meanings can be integrated in the creation of complex words.

LSF is decompositional, lexical by nature, cross-categorial, and focuses on both morphologically simplex and complex words. Thus, LSF allows one to analyze in a parallel and decompositional manner the meaning properties of both morphologically simplex and complex words.

A key aspect of the organization of lexical-semantic representations in LSF is the distinction between skeleton and body. Semantic features that are syntactically active belong to the skeleton, and the remaining semantic features are relegated to the body.

The integration of distinct representations on a semantic level is regulated by the Principle of Co-indexation. This principle accounts for the way the distinct parts of complex words (e.g., derived and compound words) integrate into a single referential unit.

Affixation is LSF is accounted for by subordination of functions. The apparatus of LSF allows one to offer a detailed account of the selectional properties of affixes and to take into consideration that even within the same class, an affix may show a strong preference for a specific subclass of bases to attach to.

As far as polysemy in affixation is concerned, the different readings of words derived with the same affix do not follow from distinct representations of the affix, but from the interaction of the general meaning of the affix with the properties of the bases it attaches to. Thus, the underdetermined semantic content of affixes is a source for polysemy in word formation.

Compounding in LSF is accounted for by concatenation of functions with concomitant co-indexing. The interpretation of subordinate compounds (e.g., burrito assembler) follows from the argumental relation between the head and the nonhead. The coordinate interpretation of a compound such as actor author is the result of compounding two lexemes that have nearly identical lexical-semantic representations. Only encyclopedic knowledge differs from one lexical item to the other. The category of attributive compounds is the default category in LSF, since attributive compounds involve neither argumental relation between the head and the nonhead (subordinate type) nor identification of reference (coordinate type). The interpretation of attributive compounds, such as bamboo bed, depends on establishing a plausible relationship between the two lexemes. This relationship follows from encyclopedic knowledge. Finally, exocentricity is attributed to general mechanisms such as metonymy that are not specific to compounding.

LSF offers a treatment of issues pertaining to form and meaning mismatches. Semantic change with no (overt) formal change (i.e., conversion) is analyzed as relisting in the lexicon. In LSF there is no principled constraint on the recursive attachment of affixes with the same meaning (i.e., derivational redundancy). Thus, expressing the same content more than once in the same word is possible when useful and meaningful. Finally, empty morphs and cases of semantic subtraction are amenable to treatment as cases of allomorphy.

Work in LSF has inaugurated a new research program in the morphology-lexical semantics interface and has shown that meaning is essential for a proper treatment of word formation. Following Lieber’s (2004) seminal monograph, a number of publications that deal with the interface between morphology and lexical-semantics have appeared. With respect to affixation, these include studies on event and result nominals (Lieber, in press-a; Melloni, 2011), transposition and conversion (Lieber, 2015), historical examination of affixes (Trips, 2009), analysis of the selectional properties of affixes in Optimality Theory (Lieber, 2010b), and evaluative affixation (Andreou, 2015). As far as compounding is concerned, Lieber (2010) discusses nonaffixal (de)verbal compounds (e.g., dog attack, attack dog), Lieber (2016c) examines synthetic compounds, and Lieber (2016b) and Andreou (2014) offer a more detailed survey of exocentricity. Lieber (2016a) introduces further mechanisms that account for semantic properties of derived words in context.

Further Reading

Andreou, M. (2015). On the headship status of Greek diminutive suffixes: A view from lexical semantics. SKASE Journal of Theoretical Linguistics, 12(1), 39–56.Find this resource:

Jackendoff, R. (2009). Compounding in the parallel architecture and conceptual semantics. In R. Lieber & P. Štekauer (Eds.), The Oxford handbook of compounding (pp. 105–128). Oxford: Oxford University Press.Find this resource:

Kawaletz, L., & Plag, I. (2015). Predicting the semantics of English nominalizations: A frame-based analysis of -ment suffixation. In L. Bauer, L. Körtvélyessy, & P. Štekauer (Eds.), Semantics of complex words (Vol. 3, pp. 289–319). Cham: Springer.Find this resource:

Lieber, R. (2010a). On the lexical semantics of compounds: Non-affixal (de)verbal compounds. In S. Scalise & I. Vogel (Eds.), Cross-disciplinary issues in compounding (pp. 127–144). Amsterdam: John Benjamins.Find this resource:

Lieber, R. (2010b). Toward an OT morphosemantics: The case of -hood, -dom, and -ship. In S. Olsen (Ed.), New impulses in word formation (pp. 173–232). Hamburg: Buske.Find this resource:

Lieber, R. (2015). The semantics of transposition. Morphology, 25, 353–369.Find this resource:

Lieber, R. (2016a). English nouns: The ecology of nominalization. Cambridge: Cambridge University Press.Find this resource:

Lieber, R. (2016b). The lexical semantic approach to compounding. In P. ten Hacken (Ed.), The semantics of compounding (pp. 38–53). Cambridge: Cambridge University Press.Find this resource:

Lieber, R. (2016c). On the interplay of facts and theory: Revisiting synthetic compounds in English. In H. Harley & D. Siddiqi (Eds.), Morphological metatheory (pp. 513–536). Amsterdam: John Benjamins.Find this resource:

Melloni, C. (2011). Event and result nominals: A morpho-semantic approach. Bern: Peter Lang.Find this resource:

Trips, C. (2009). Lexical semantics and diachronic morphology: The development of -hood, -dom and -ship in the history of English. Tübingen: Niemeyer.Find this resource:

References

Andreou, M. (2014). Headedness in word formation and lexical semantics: Evidence from Italiot and Cypriot (PhD diss.). University of Patras.Find this resource:

Andreou, M. (2015). Lexical negation in lexical semantics: The prefixes in- and dis-. Morphology, 25(4), 391–410.Find this resource:

Aronoff, M., & Cho, S. (2001). The semantics of -ship suffixation. Linguistic Inquiry, 32(1), 167–173.Find this resource:

Barker, C. (1998). Episodic -ee in English: A thematic role constraint on new word formation. Language, 74, 695–727.Find this resource:

Bauer, L. (2008). Exocentric compounds. Morphology, 18, 51–74.Find this resource:

Bauer, L. (2010). The typology of exocentric compounding. In S. Scalise & I. Vogel (Eds.), Cross-disciplinary issues in compounding (pp. 167–176). Amsterdam: John Benjamins.Find this resource:

Bauer, L., Lieber, R., & Plag, I. (2013). The Oxford reference guide to English morphology. Oxford: Oxford University Press.Find this resource:

Carlson, G. (1977). Reference to kinds in English (Unpublished PhD diss.). University of Massachusetts, Amherst.Find this resource:

Copestake, A., & Briscoe, T. (1996). Semi-productive polysemy and sense extension. In J. Pustejovsky & B. Boguraev (Eds.), Lexical semantics: The problem of polysemy (pp. 15–68). Oxford: Clarendon Press.Find this resource:

Don, J. (1993). Morphological conversion (PhD diss.). University of Utrecht.Find this resource:

Hale, K., & Keyser, S. J. (2002). Prolegomena to a theory of argument structure. Cambridge, MA: MIT Press.Find this resource:

Jackendoff, R. (1990). Semantic structures. Cambridge, MA: MIT Press.Find this resource:

Lieber, R. (1992). Deconstructing morphology. Chicago: Chicago University Press.Find this resource:

Lieber, R. (2004). Morphology and lexical semantics. Cambridge: Cambridge University Press.Find this resource:

Lieber, R. (2007). The category of roots and the roots of categories: What we learn from selection in derivation. Morphology, 16(2), 247–272.Find this resource:

Lieber, R. (2009). A lexical semantic approach to compounding. In R. Lieber & P. Štekauer (Eds.), The Oxford handbook of compounding (pp. 78–104). Oxford: Oxford University Press.Find this resource:

Lieber, R. (2010). On the lexical semantics of compounds: Non-affixal (de)verbal compounds. In S. Scalise & I. Vogel (Eds.), Cross-disciplinary issues in compounding (pp. 127–144). Amsterdam: John Benjamins.Find this resource:

Lieber, R., & Baayen, H. (1999). Nominalizations in a calculus of lexical semantic representations. In G. Booij & J. van Marle (Eds.), Yearbook of morphology 1998 (pp. 175–198). Dordrecht: Springer.Find this resource:

Löbner, S. (1985). Definites. Journal of Semantics, 4(4), 279–326.Find this resource:

Löbner, S. (2011). Concept types and determination. Journal of Semantics, 28, 279–333.Find this resource:

Mohanan, T., & Mohanan, K. (1999). On representations in grammatical semantics. In T. Mohanan & L. Wee (Eds.), Grammatical semantics: Evidence for structure in meaning (pp. 23–76). Stanford, CA: CSLI Publications.Find this resource:

Plag, I. (1999). Morphological productivity: Structural constraints in English derivation. Berlin: Mouton de Gruyter.Find this resource:

Plag, I. (2004). Syntactic category information and the semantics of derivational morphological rules. Folia Linguistica, 38(3–4), 193–225.Find this resource:

Pustejovsky, J. (1995). The generative lexicon. Cambridge, MA: MIT Press.Find this resource:

Pustejovsky, J., & Boguraev, B. (1996). Introduction: Lexical semantics in context. In J. Pustejovsky & B. Boguraev (Eds.), Lexical semantics: The problem of polysemy (pp. 1–14). Oxford: Clarendon Press.Find this resource:

Rappaport Hovav, M., & Levin, B. (1996). Two types of derived accomplishments. In M. Butt & T. H. King (Eds.), Proceedings of the first LFG conference (pp. 375–388). Grenoble: RANK Xerox.Find this resource:

Rappaport Hovav, M., & Levin, B. (1998). Building verb meanings. In M. Butt & W. Geuder (Eds.), The projection of arguments: Lexical and compositional factors (pp. 97–134). Stanford, CA: CSLI Publications.Find this resource:

Szymanek, B. (1988). Categories and categorization in morphology. Lublin: Catholic University Press.Find this resource:

Wierzbicka, A. (1996). Semantics: Primes and universals. Oxford: Oxford University Press.Find this resource:

Williams, E. (1981). On the notions of ‘lexically related’ and ‘head of the word’. Linguistic Inquiry, 12, 245–274.Find this resource:

Notes:

(1.) Work on this article was partially funded by the Deutsche Forschungsgemeinschaft SFB 991 “The Structure of Representations in Language, Cognition, and Science” (Project: C08).

(3.) On relational nouns, see Löbner (1985, 2011).

(4.) For more on IEPS, the interested reader is referred to Lieber and Baayen (1999).

(5.) For a treatment of -ship, also see Aronoff and Cho (2001).