Foundations of Lexical Semantics with Malka Rappaport Hovav (Part 1 of Lexical Semantics)

September 8, 2021 // The Biblical Languages Podcast

Image

Share this Post

In this episode of the Biblical Languages Podcast, Kevin Grasso interviews Malka Rappaport Hovav on the foundations of lexical semantics.

Malka Rappaport Hovav holds the Henya Sharef chair in Humanities at the Hebrew University of Jerusalem, and is a member of the Department of Linguistics as well as the Center for Logic, Language, and Cognition. Her research focuses on lexical representation and its relation to conceptual structure and morpho-syntactic realization.

Given the complexity of this episode's topic, we've also written this blog post to accompany it. Below you'll find further explanation, diagrams, and a glossary of terms used.

There is a glossary at the end of this post to help readers who might be unfamiliar with some of the linguistic jargon. Terms found in the glossary are those that have all caps on their first use in the post.

Here's a snapshot of what you can expect in this conversation:

  • What exactly lexical semantics is, and how it relates to formal semantics
  • The concept of lexicalized meaning and how it helps us determine what words really mean
  • The Manner/Result Complementarity Hypothesis and how it can help us categorize verbs
  • Argument Realization, verb classes, and how syntax relates to the meaning of words
  • And much, much more!

Show notes


Research referenced in this episode:

Levin, B. and M. Rappaport Hovav (2013) "Lexicalized Meaning and Manner/Result Complementarity'', in B. Arsenijević, B. Gehrke, and R. Marín, eds., Subatomic Semantics of Event Predicates, Springer, Dordrecht, 49-70.

This is Part 1 of our series on Lexical Semantics. Check out the other episodes in this series:

  • Part 2: Greek and Septuagint Lexicography with William Ross
  • Part 3: Reinier de Blois on Hebrew lexicography (coming 9/22)
  • Part 4: Nijay Gupta on Greek word studies (coming 9/29)
  • Part 5: Keren Dubnov on Hebrew word studies (coming 10/6)
  • Part 6: Wrapup with Kevin and Nick on Hebrew (coming 10/13)
  • Part 7: Wrapup with Kevin and Nick on Greek (coming 10/20)
 

What is Lexical Semantics?


Lexical semantics is the study of meaning of lexical items. Lexical items are those morphemes that bear the main conceptual content in language. For example, in a sentence like The dog entered the room, the lexical items are dog, enter, and room. The other morphemes, like the and the past tense marker -ed, are called functional/grammatical morphemes.

These functional morphemes can change without the main conceptual content changing, so if we have The dog has entered the room with the addition of the perfect marker has, the same event is being described in this sentence as in the sentence with the simple past, i.e. The dog entered the room.

Even though lexical semantics primarily deals with the words that give conceptual content, these words also interact with grammatical morphemes and syntactic structure in particular ways to produce the meaning of the sentence, so the lexical semanticist must also know something about grammatical morphemes and syntax as well.

 

Lexicalized Meaning and Manner/Result Complementarity


There are many terms and concepts that can be useful as we seek to study lexical items. The study in Levin and Rappaport Hovav 2013 introduces several foundational terms that can help us. To begin with, we have lexicalized meaning. By “lexicalized meaning”, we refer to meaning encoded in the conceptual content of the word and not derived either as inferences from the arguments of a word, from other aspects of the context, or even from functional categories associated with words.

For example, when we hear a sentence like I cracked the egg, we usually infer that there was some intentional act of cracking that took place that I engaged in, but this “intentional” component of meaning is not actually a part of the meaning of the word crack. We can see this in other uses of the word crack, such as in The egg cracked where we can, and often do, interpret the sentence as not including someone that intentionally cracked the egg. Thus, even though we have an intentional component of meaning in a sentence like I cracked the egg, it seems that this component may be coming from the combination of the words I and crack together along with the fact that cracking eggs is such a normal event, and the intentional component is actually not coming from the word crack itself.

To illustrate this further, we can talk about the manner/result complementarity hypothesis. This hypothesis states that a verb may lexicalize either the manner in which something changes or the result of something changing, but not both at the same time. For example, verbs of motion may specify manner. These are verbs like run, jump, skip, hop, waltz, etc. that tell us nothing about the path of the agent, only that the agent is engaged in that activity. There does not need to be any change in location (i.e. a result) from any of these verbs, so it is perfectly possible to say that I ran in place, and I didn’t move.

On other other hand, there are some verbs that specify that a result has taken place, namely that the entity engaged in the action has moved locations. These are verbs like enter, arrive, go, come, descend, etc. where the one who goes must actually change location. In this case, the verb itself tells us nothing about the manner in which the going is done, so I am going to the neighbor’s house means that I am traversing a path from my current location to wherever my neighbor’s house is, but I can do that by means of running, jumping, skipping, or waltzing. The verb go does not tell us anything about how we got there.

This same dichotomy can be found throughout the lexicon in (presumably) all the world’s languages. For example, change of state verbs like break, crack, melt, freeze, heat, widen, etc. specify a result, while verbs of contact like hit, pound, batter, wipe, rub, etc. only specify the manner of how two entities interact with each other, and there is no necessary change of state that takes place.

This is related to lexicalized meaning because we can say that manner verbs lexicalize a manner meaning (the manner is encoded in the verb in every use), while result verbs lexicalize a result meaning. Still, we may have inferences that arise outside of the lexicalized meaning, i.e. an understanding of properties of a situation in the world that may arise from the words or other extra-linguistic factors.

For example, when I use a manner verb like wipe in The husband wiped the table, there may be an inference that the table is now cleaner (the result) than when it first started, but this is not necessarily true. We can follow up such a sentence with but the table is no cleaner than when he first started without a contradiction. On the other hand, The husband cleaned the table must mean that a result took place such that the table is cleaner now than when he first started, and saying otherwise is a contradiction.

Thus, even though the verb wipe does not lexicalize a result, it may be associated with certain results in certain contexts. Such meanings are not lexicalized in the verb, but they may arise from other factors in the context that lead us to these kinds of inferences. Part of the task of lexical semantics is to determine what contexts gives rise to such inferences, and this, in turn, may allow us to determine how or when such a meaning is present in other contexts.

The manner/result complementarity hypothesis was originally formulated to be about roots and not just words. All languages have roots which make up words. In some languages (such as Hebrew), these roots are easier to see, but even in a language like English, we also have roots. For example, crack may be considered a root, since it can actually be a noun (There is a crack in the vase), a verb (I cracked the vase), and an adjective (This vase is uncrackable). The idea is that crack is a result root, so its result meaning actually carries through to all uses of the root, regardless of whether the root is instantiated as a noun, adjective, or verb. When we talk about lexicalized meaning, there are times when a meaning may be lexicalized in a word after it has been categorized (i.e. put into a noun, verb, or adjective form) or before (i.e. in the root itself).

 

Argument Realization


Verbs not only specify events in the world, but they also specify how their arguments (roughly the participants in an event) are structured. This is called a verb’s argument realization, or how the verb’s arguments are realized in syntactic structures.

For example, a verb like eat not only describes an event of someone consuming something, but it also specifies that the eater is the grammatical subject of the verb, and the thing eaten is the grammatical object of the verb. This may seem incredibly obvious to us, but this is only because every language that we know of lexicalizes this verb with the same argument realization, i.e. the eater is the subject, and the thing eaten is the object. The question is why Language would be so consistent in this respect and what might this tell us about verb meaning and the nature of subjecthood and objecthood.

Verbs may differ subtly from language to language in how their arguments are structured, even when you compare “translation equivalents” in the two languages. So when we look at a verb like give in English, we see that it typically has three arguments: a giver, something given, and a recipient. These three arguments can be realized in two distinct ways. We can say Mary gave Bill a cookie or Mary gave a cookie to Bill. This is called an argument alternation.

When we go to another language and analyze the translation equivalent of give, we may not find all the same argument realization patterns that we have in English, or we may find more. The Biblical Hebrew word נָתַן, for example, has the same argument alternation we find in English with the double object construction or the recipient marked with the preposition ל, but it also has other argument realizations that we do not have in English, such as a participant being marked with בְּ and the closest meaning in English being something like place. In this respect, the Biblical Hebrew translation equivalent of give is broader than the English word give, even though in many cases the best translation of נָתַן is still give in English.

Crucially, it is syntax that triggers or reflects this meaning change that we see in נָתַן, and the fact that give is incompatible with a similar syntactic frame shows either that the meaning of that argument realization pattern (with the meaning of the particular prepositions that go along with it) in Biblical Hebrew is different or the meaning give and נָתַן differ in crucial ways.

 

Methodology


The concepts we have discussed can help us to develop a methodology for studying lexical items in dead languages like Ancient Hebrew and Greek. 

First, we must determine what we are looking for. When we are determining the meaning of a word, we are first and foremost trying to figure out what meaning that word is contributing to the context. In order to do this, we have to look at how the word is used in a variety of different contexts to see what meanings the context is contributing vs. what meaning the word itself contributes. In other words, we are trying to strip away all the inferences that the context might be giving us to see just that meaning that the word gives.

Second, when we are trying to discover the meaning of a word, syntactic distribution helps us to see how the lexical item interacts with meanings of grammatical morphemes and syntactic structures. Because we do not have access to native speaker judgments, we cannot know what structures were not possible, but we can know how native speakers used the words in texts, and this at least gives us a range of possible syntactic structures that are compatible with the word.

Third, we can categorize words into well-known cross-linguistic patterns. For verbs of motion, we may begin by comparing how a particular verb compares syntactically and semantically with known result verbs to see if a result is encoded or a manner. For nouns, we may begin by looking at how the word interacts with the plural morpheme, i.e. whether it is a mass or count noun. These are well-known categories that hold across languages and have certain semantic and syntactic properties.

Fourth, we can determine the lexicalized meaning by seeing what kind of result or manner or other kind of meaning is encoded in every occurrence of the word. Sometimes, we may legitimately have more than one meaning for a particular word, but many times, inferences that we think are associated with the word are actually associated with other words found in a particular context that are triggered only when those words are present. Again, crack does not imply intentionality, even though first or second person subjects often trigger that inference when the word crack is used. After looking at how the word is used in a variety of contexts, we can strip away the lexicalized meaning of the word to see what the word is contributing to each context.

 

Where to listen

You can listen and subscribe for updates here: https://biblingo.org/podcast/

You can also listen and subscribe on Apple PodcastsGoogle PodcastsSpotify, and Overcast.
 
You can also watch on YouTube.

Listen to the full episode, share your feedback, and let us know topics you'd like to hear on future episodes!


September 8, 2021 // The Biblical Languages Podcast

Image

Share this Post

In this episode of the Biblical Languages Podcast, Kevin Grasso interviews Malka Rappaport Hovav on the foundations of lexical semantics.

Malka Rappaport Hovav holds the Henya Sharef chair in Humanities at the Hebrew University of Jerusalem, and is a member of the Department of Linguistics as well as the Center for Logic, Language, and Cognition. Her research focuses on lexical representation and its relation to conceptual structure and morpho-syntactic realization.

Given the complexity of this episode's topic, we've also written a blog post to accompany it for further explanation (see below).

Here's a snapshot of what you can expect in this conversation:

  • What exactly lexical semantics is, and how it relates to formal semantics
  • The concept of lexicalized meaning and how it helps us determine what words really mean
  • The Manner/Result Complementarity Hypothesis and how it can help us categorize verbs
  • Argument Realization, verb classes, and how syntax relates to the meaning of words
  • And much, much more!

Show notes

Research referenced in this episode:

Levin, B. and M. Rappaport Hovav (2013) "Lexicalized Meaning and Manner/Result Complementarity'', in B. Arsenijević, B. Gehrke, and R. Marín, eds., Subatomic Semantics of Event Predicates, Springer, Dordrecht, 49-70.

This is Part 1 of our series on Lexical Semantics. Check out the other episodes in this series:

  • Part 2: Greek and Septuagint Lexicography with William Ross
  • Part 3: Reinier de Blois on Hebrew lexicography (coming 9/22)
  • Part 4: Nijay Gupta on Greek word studies (coming 9/29)
  • Part 5: Keren Dubnov on Hebrew word studies (coming 10/6)
  • Part 6: Wrapup with Kevin and Nick on Hebrew (coming 10/13)
  • Part 7: Wrapup with Kevin and Nick on Greek (coming 10/20)

What is Lexical Semantics?

Lexical semantics is the study of meaning of lexical items. Lexical items are those morphemes that bear the main conceptual content in language. For example, in a sentence like The dog entered the room, the lexical items are dog, enter, and room. The other morphemes, like the and the past tense marker -ed, are called functional/grammatical morphemes.

These functional morphemes can change without the main conceptual content changing, so if we have The dog has entered the room with the addition of the perfect marker has, the same event is being described in this sentence as in the sentence with the simple past, i.e. The dog entered the room.

Even though lexical semantics primarily deals with the words that give conceptual content, these words also interact with grammatical morphemes and syntactic structure in particular ways to produce the meaning of the sentence, so the lexical semanticist must also know something about grammatical morphemes and syntax as well.

Lexicalized Meaning and Manner/Result Complementarity

There are many terms and concepts that can be useful as we seek to study lexical items. The study in Levin and Rappaport Hovav 2013 introduces several foundational terms that can help us. To begin with, we have lexicalized meaning. By “lexicalized meaning”, we refer to meaning encoded in the conceptual content of the word and not derived either as inferences from the arguments of a word, from other aspects of the context, or even from functional categories associated with words.

For example, when we hear a sentence like I cracked the egg, we usually infer that there was some intentional act of cracking that took place that I engaged in, but this “intentional” component of meaning is not actually a part of the meaning of the word crack. We can see this in other uses of the word crack, such as in The egg cracked where we can, and often do, interpret the sentence as not including someone that intentionally cracked the egg. Thus, even though we have an intentional component of meaning in a sentence like I cracked the egg, it seems that this component may be coming from the combination of the words I and crack together along with the fact that cracking eggs is such a normal event, and the intentional component is actually not coming from the word crack itself.

To illustrate this further, we can talk about the manner/result complementarity hypothesis. This hypothesis states that a verb may lexicalize either the manner in which something changes or the result of something changing, but not both at the same time. For example, verbs of motion may specify manner. These are verbs like run, jump, skip, hop, waltz, etc. that tell us nothing about the path of the agent, only that the agent is engaged in that activity. There does not need to be any change in location (i.e. a result) from any of these verbs, so it is perfectly possible to say that I ran in place, and I didn’t move.

On other other hand, there are some verbs that specify that a result has taken place, namely that the entity engaged in the action has moved locations. These are verbs like enter, arrive, go, come, descend, etc. where the one who goes must actually change location. In this case, the verb itself tells us nothing about the manner in which the going is done, so I am going to the neighbor’s house means that I am traversing a path from my current location to wherever my neighbor’s house is, but I can do that by means of running, jumping, skipping, or waltzing. The verb go does not tell us anything about how we got there.

This same dichotomy can be found throughout the lexicon in (presumably) all the world’s languages. For example, change of state verbs like break, crack, melt, freeze, heat, widen, etc. specify a result, while verbs of contact like hit, pound, batter, wipe, rub, etc. only specify the manner of how two entities interact with each other, and there is no necessary change of state that takes place.

This is related to lexicalized meaning because we can say that manner verbs lexicalize a manner meaning (the manner is encoded in the verb in every use), while result verbs lexicalize a result meaning. Still, we may have inferences that arise outside of the lexicalized meaning, i.e. an understanding of properties of a situation in the world that may arise from the words or other extra-linguistic factors.

For example, when I use a manner verb like wipe in The husband wiped the table, there may be an inference that the table is now cleaner (the result) than when it first started, but this is not necessarily true. We can follow up such a sentence with but the table is no cleaner than when he first started without a contradiction. On the other hand, The husband cleaned the table must mean that a result took place such that the table is cleaner now than when he first started, and saying otherwise is a contradiction.

Thus, even though the verb wipe does not lexicalize a result, it may be associated with certain results in certain contexts. Such meanings are not lexicalized in the verb, but they may arise from other factors in the context that lead us to these kinds of inferences. Part of the task of lexical semantics is to determine what contexts gives rise to such inferences, and this, in turn, may allow us to determine how or when such a meaning is present in other contexts.

The manner/result complementarity hypothesis was originally formulated to be about roots and not just words. All languages have roots which make up words. In some languages (such as Hebrew), these roots are easier to see, but even in a language like English, we also have roots. For example, crack may be considered a root, since it can actually be a noun (There is a crack in the vase), a verb (I cracked the vase), and an adjective (This vase is uncrackable). The idea is that crack is a result root, so its result meaning actually carries through to all uses of the root, regardless of whether the root is instantiated as a noun, adjective, or verb. When we talk about lexicalized meaning, there are times when a meaning may be lexicalized in a word after it has been categorized (i.e. put into a noun, verb, or adjective form) or before (i.e. in the root itself).

Argument Realization

Verbs not only specify events in the world, but they also specify how their arguments (roughly the participants in an event) are structured. This is called a verb’s argument realization, or how the verb’s arguments are realized in syntactic structures.

For example, a verb like eat not only describes an event of someone consuming something, but it also specifies that the eater is the grammatical subject of the verb, and the thing eaten is the grammatical object of the verb. This may seem incredibly obvious to us, but this is only because every language that we know of lexicalizes this verb with the same argument realization, i.e. the eater is the subject, and the thing eaten is the object. The question is why Language would be so consistent in this respect and what might this tell us about verb meaning and the nature of subjecthood and objecthood.

Verbs may differ subtly from language to language in how their arguments are structured, even when you compare “translation equivalents” in the two languages. So when we look at a verb like give in English, we see that it typically has three arguments: a giver, something given, and a recipient. These three arguments can be realized in two distinct ways. We can say Mary gave Bill a cookie or Mary gave a cookie to Bill. This is called an argument alternation.

When we go to another language and analyze the translation equivalent of give, we may not find all the same argument realization patterns that we have in English, or we may find more. The Biblical Hebrew word נָתַן, for example, has the same argument alternation we find in English with the double object construction or the recipient marked with the preposition ל, but it also has other argument realizations that we do not have in English, such as a participant being marked with בְּ and the closest meaning in English being something like place. In this respect, the Biblical Hebrew translation equivalent of give is broader than the English word give, even though in many cases the best translation of נָתַן is still give in English.

Crucially, it is syntax that triggers or reflects this meaning change that we see in נָתַן, and the fact that give is incompatible with a similar syntactic frame shows either that the meaning of that argument realization pattern (with the meaning of the particular prepositions that go along with it) in Biblical Hebrew is different or the meaning give and נָתַן differ in crucial ways.

Methodology

The concepts we have discussed can help us to develop a methodology for studying lexical items in dead languages like Ancient Hebrew and Greek. 

First, we must determine what we are looking for. When we are determining the meaning of a word, we are first and foremost trying to figure out what meaning that word is contributing to the context. In order to do this, we have to look at how the word is used in a variety of different contexts to see what meanings the context is contributing vs. what meaning the word itself contributes. In other words, we are trying to strip away all the inferences that the context might be giving us to see just that meaning that the word gives.

Second, when we are trying to discover the meaning of a word, syntactic distribution helps us to see how the lexical item interacts with meanings of grammatical morphemes and syntactic structures. Because we do not have access to native speaker judgments, we cannot know what structures were not possible, but we can know how native speakers used the words in texts, and this at least gives us a range of possible syntactic structures that are compatible with the word.

Third, we can categorize words into well-known cross-linguistic patterns. For verbs of motion, we may begin by comparing how a particular verb compares syntactically and semantically with known result verbs to see if a result is encoded or a manner. For nouns, we may begin by looking at how the word interacts with the plural morpheme, i.e. whether it is a mass or count noun. These are well-known categories that hold across languages and have certain semantic and syntactic properties.

Fourth, we can determine the lexicalized meaning by seeing what kind of result or manner or other kind of meaning is encoded in every occurrence of the word. Sometimes, we may legitimately have more than one meaning for a particular word, but many times, inferences that we think are associated with the word are actually associated with other words found in a particular context that are triggered only when those words are present. Again, crack does not imply intentionality, even though first or second person subjects often trigger that inference when the word crack is used. After looking at how the word is used in a variety of contexts, we can strip away the lexicalized meaning of the word to see what the word is contributing to each context.

Where to listen

You can listen and subscribe for updates here: https://biblingo.org/podcast/

You can also listen and subscribe on Apple PodcastsGoogle PodcastsSpotify, and Overcast.
 
You can also watch on YouTube.

Listen to the full episode, share your feedback, and let us know topics you'd like to hear on future episodes!

Image

The Biblical Languages Podcast hosts discussions and interviews related to learning the biblical languages and issues relevant to biblical exegesis. Episodes cover topics in 4 major categories: language acquisition, linguistics, cultural backgrounds, and exegesis.

Listen & subscribe here: https://biblingo.org/podcast/

 

Share this Post