52
Running Head: CONTEXT IN LEXICAL ACCESS OF POLYSEMES The Integration of Contextual Information in the Lexical Access of Polysemes Joel Fishbein Pomona College Lexical Access of Polysemes 1

The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

  • Upload
    others

  • View
    15

  • Download
    0

Embed Size (px)

Citation preview

Page 1: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Running Head: CONTEXT IN LEXICAL ACCESS OF POLYSEMES

The Integration of Contextual Information in the Lexical Access of Polysemes

Joel Fishbein

Pomona College

Lexical Access of Polysemes 1

Page 2: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Abstract

How the language processor uses linguistic information to interpret words in a sentence is

not fully understood (Federmeier, 2007). This self-paced reading experiment investigated

whether the language processor uses contextual information at the point of lexical access to

immediately commit to one sense of a polyseme, or whether context is used later in

disambiguation. Participants read context-setting sentences which either biased for the figurative

sense interpretation of a polyseme or did not bias towards any specific sense interpretation,

followed by sentences containing the polyseme which required the language processor to select

the polyseme’s figurative or literal sense. While a facilitatory effect of context was observed,

participants encountered greater difficulty when accessing the figurative sense of the polyseme

as compared to the literal sense, irrespective of whether the context biased towards the figurative

sense interpretation of the polyseme or not. These results constitute evidence that the language

processor uses structural information in polyseme disambiguation, and that contextual

information has a late effect on sense selection in comparison to structural information.

Lexical Access of Polysemes 2

Page 3: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

The Integration of Contextual Information in the Lexical Access of Polysemes

In comprehending sentences during natural speech, the human language sentence

processor must rapidly activate a representation of each word that it encounters. The activation of

the representation of a word is known as lexical access. Lexical access is made complex by the

fact that many words, and especially the most frequently-used words in language, have several

interpretations (Zipf, 1945), only one of which may be appropriate for a given discourse context.

Therefore, in order to fully model speech comprehension, we must be able to model the

strategies that the language processor uses to select one interpretation of a word over the others.

Context is often the factor which determines what interpretation of a word a language

processor must ultimately use to understand a sentence; in the context of talking about a river, for

example, the processor will know to interpret bank as the land alongside the river, whereas in the

context of talking about stock markets, a processor will know to interpret that same word to

mean a financial institution. While context clearly facilitates the interpretation of a word with

multiple interpretations in some fashion, the time course for integrating context into lexical

access is not transparent; does the processor activate all possible interpretations of a word, and

then use context to immediately select one? Or does it use contextual information later in

interpretation? In the present paper, we investigated this question by observing language

processors’ processing of metonymies, which are figures of speech that require language

processors to interpret polysemes using their figurative sense in order to arrive at the correct

reading of the sentence. Importantly, metonymy operates upon the fact that some polysemous

words have both a literal (primary) and figurative (secondary and non-literal) sense, and the

Lexical Access of Polysemes 3

Page 4: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

language processor can rapidly commit to one these senses during interpretation of a polyseme

based on available linguistic information. We found that the language processor uses some

source of information before using contextual information in the lexical access of polysemes. We

interpreted our results to suggest that the language processor uses structural information to make

an immediate assumption about the thematic role of the polyseme (Clifton, 1993) and

consequently its animacy (Traxler, 2012) in the processing of metonymies. According to our

interpretation, contextual information is integrated after structural information. Our findings can

be made consistent with the Underspecified Model of metonymy processing proposed by Frisson

and Pickering (1999), so long as the Underspecified Model can be expanded to allow the use of

structural and contextual information in the selection of a polyseme’s sense.

In the rest of §1, we review the literature to date on lexical access and metonymy. In §2,

we present two experiments, a norming study and a self-paced reading study, whose results bear

implications on lexical access and metonymy processing. Finally, in §3, we provide general

discussion of the results from §2, relating our findings to existing models of language processing.

Lexical Access of Homophones and Polysemes

Swinney (1979) examined the processing of homophones, words with two distinct

meanings, under varying contexts to determine the time course of their processing. The

experiment was a lexical decision task, wherein participants listened to a sentence, and at some

point while listening to the sentence, they had to decide whether a string of letters presented

visually was a word or not. In the experiment, participants heard sentences that contained

homophones, and the words in the sentences that preceded the homophone either biased for one

of the two possible meanings of the homophone, or did not bias for either meaning. The lexical

Lexical Access of Polysemes 4

Page 5: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

decision words were either related to the meaning facilitated by context, were related to the

meaning not facilitated by context, or were related to neither meaning. For example, participants

listened to sentences such as:

(1) Rumor had it that, for years, the government building had been plagued with problems. The man was not surprised when he found several bugs in the corner of his room.

While listening to sentences such as (1), participants were presented a word visually either

immediately or three syllables after the target, bug. The lexical decision word for (1) was either

ant (related to the contextually salient meaning of the target), spy (related to the non-salient

meaning of the target), or sew (not related to either meaning).

Crucially, Swinney (1979) found that when the target was presented immediately after the

homophone, participants responded equally fast to the congruent and incongruent lexical

decision items, regardless of whether there was a strong biasing context for one of the meanings

of the homophone or not. However, when the lexical decision items were presented three

syllables after the end of the homophone, participants responded faster to the lexical decision

items congruent to the salient meaning of the homophone. This result suggests that in the

processing of homophones, context does not appear to affect lexical access directly; rather, both

meanings of the homophone are initially activated, and context has a relatively late effect on

meaning selection. More recent work has replicated this result (see Carroll, 2007 for review).

Context is not the only factor that affects the selection of a homophone meaning,

however. Hogaboam and Perfetti (1975) demonstrated that when context does not bias the

language processor towards one meaning of a homophone over the others, the language

processor will select the dominant, or more frequently used, meaning of a word over the

subordinate meanings. Likewise, Simpson (1981) found that context and dominance exert

Lexical Access of Polysemes 5

Page 6: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

independent effects on the lexical access of homophones. He concluded that in the absence of

any biasing context, the language processor will always access the meanings of a homophone

from highest to lowest relative frequency, and that the processor will only access the subordinate

meanings first when there is a context which biases towards interpretations using those meanings

of the homophone. These findings and others help to explicate the processing of homophones,

but later research (e.g. Frazier and Rayner, 1990; Pylkkänen et al., 2006) found that polysemes

patterned differently from homophones, and so required an alternative explanation.

Unlike homophones, polysemes are words which have one meaning but more than one

sense. The multiple senses of a polyseme all relate to the central meaning, and are often

distributed in systematic ways; for example, a place name like Vietnam has a location sense,

which corresponds to the physical location, and an event sense, which corresponds to an event

that happened in that location (Nunberg, 1979). Experimental findings have shown that the

language processor does not process polysemes in the same way that it does homophones. For

example, Frazier and Rayner (1990) found that whereas the language processor will select the

dominant meaning of a homophone and suppress the subordinate meanings in the absence of a

context that biases the selection of one meaning, it does not select one sense of a polyseme over

the others in the absence of a context that biases for the selection of one sense. Rather, all senses

of the polyseme are equally available to the processor until sufficient contextual information is

available for them to choose one. The language processor accomplishes this by activating an

underspecified representation of the polyseme, which allows the processor to maintain a very

Lexical Access of Polysemes 6

Page 7: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

general representation of the polyseme, but to retain access to all senses of the polyseme until

sufficient information is available to select one of the senses.1

Frazier and Rayner (1990) and others (e.g. Pickering & Frisson, 2001) have suggested

that the differences in processing between meaning ambiguities and sense ambiguities reflect

differences in representation between homophones and polysemes. Since all senses of a

polyseme are available to language processors in the absence of a biasing context, it is widely

believed that all the senses of a polyseme are accessed from one core meaning of the word.2

When a polyseme is encountered, then, processors may ‘minimally commit’ by activating an

underspecified meaning of the polyseme, and delaying the selection of one sense of the word

until sufficient contextual information becomes available (Frazier & Rayner, 1990; Frisson &

Pickering, 1999; Frisson & Pickering, 2001).

Neuroimaging research has provided further evidence that the multiple senses of a

polyseme are stored on a single lexical node. Pylkkänen, Llinas and Murphy (2006) compared

the processing of polysemes and homophones using MEG. In their experiment, participants read

two phrases, each containing the same ambiguous word (a homophone or a polyseme), presented

word-by-word in a rapid serial visual presentation (RSVP) design. For example, participants read

homophone phrase pairs such as river bank - savings bank and polyseme pairs such as lined

paper - liberal paper. By observing latency of the M350 component, which has been correlated

Lexical Access of Polysemes 7

1 Linguists have employed underspecification to explain many phenomena in language, such as the language processor’s ability to realize a single phoneme in many different ways (Wheeldon & Waksler, 2004) and to process ambiguous sentences (Swets et al., 2008).

2 Psycholinguistic evidence (e.g. Klein & Murphy, 2002; Foraker & Murphy, 2009) has been presented which suggests that at least in some cases, different senses of a word are represented separately in the lexicon. For example, Foraker and Murphy (2009) replicated the dominance effects using polysemes that Simpson (1981) found with homophones, suggesting that at least some polysemes are represented as separate lexical entries, like homophones. However, the causes for representing two senses of a word in separate lexical entries remains unclear, and neuroimaging work (Pylkkänen et al., 2006) has disconfirmed this result. For our purposes here, we will leave this debate aside.

Page 8: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

with lexical activation (Pylkkänen & Marantz, 2003), the authors found that lexical activation of

the second homophone in each homophone pair was delayed, but the second polyseme in each

polyseme pair was not. They concluded that the latency in the activation of the homophones was

correlated with competition between the separate meanings, which is consistent with models that

posit separate meanings of a homophone being stored in separate lexical entries. Since

polysemes show no such competition effects, it is theorized, the senses of the polyseme must

have be on the same lexical entry.

One aspect of polysemy processing that remains largely unexplored is the time course

that a processor takes in using context to select one sense of a polyseme over the others. On trials

in which participants read biasing contextual information before encountering the polyseme,

Frazier and Rayner (1990) found evidence that participants selected a sense upon or shortly after

fixating on polyseme. Yet it remains unclear as to when a processor will commit to one sense if

the disambiguating contextual information is extra-lexical (meaning outside the sentence

containing the polyseme), or if the contextual information is generally assumed as shared

common knowledge, and is never specifically stated within the discourse. One possible way to

elucidate what role context plays in polysemy disambiguation is to determine how the language

processor handles metonymies, as described below.

Metonymy

Metonymy is a device of figurative language wherein a speaker uses the name of one

entity to refer to some related or encompassing entity (Nunberg, 1995; Frisson & Pickering,

2001). For example, if a person were to say that she will, “read Dickens,” it is understood in

most contexts that she intends to read one of Dickens’s books. Importantly, the mechanism upon

Lexical Access of Polysemes 8

Page 9: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

which metonymy operates is polysemy; the language processor must access the figurative sense

of the polyseme to felicitously interpret the metonymic expression (Nunberg, 2004; Frisson &

Pickering, 2001). Therefore, the conclusions drawn about the metonymy processing can be

directly applied to models of polysemy processing.

In an eye-tracking study, Frisson and Pickering (1999) produced further evidence for an

underspecified representation of polysemes in the lexicon. In their experiment, participants read

sentences like those in the following pair (italics added):

(2) a. The two businessmen tried to purchase the convent at the end of last April, which upset quite a lot of people.

b. That blasphemous woman had to answer to the convent at the end of last March, but did not get a lot of support.

Their findings indicated that comprehenders had no greater difficulty processing the polyseme

(in the above sentences, convent) when the language processor was required to interpret it using

its figurative sense (2b) as compared to its literal sense (2a). These results support an

underspecified model of polysemous words in the lexicon; since language processors were able

to access the literal and figurative senses of the polysemes with equal ease, it is theorized, then it

is likely that both senses were equally accesible to the processor upon first encountering the

polyseme. The underspecified model of polysemy specifies that all the senses of the polyseme

are equally accessible to the processor at the point of lexical access, so Frisson and Pickering’s

(1999) findings are predicted by the underspecified model.

In a separate eye-tracking experiment, Frisson and Pickering (2007) examined the effects

of context on facilitating new metonymic interpretations, such as facilitating a metonymy from a

producer(e.g. the name ‘Needham’) to a product (e.g. the works of Needham). They found that

Lexical Access of Polysemes 9

Page 10: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

when processors were given contextual information that facilitated a metonymic interpretation

(e.g. that Needham is an author) before encountering the metonym, they had significantly less

difficulty in processing the metonymy. This finding suggests that extra-lexical information can

be rapidly integrated to facilitate the access of new, never-before-accessed senses of a word. Yet

it is presently unclear whether the result found on the influence of context on new polysemous

senses can be broadened to make claims about the lexical access of existing senses of a

polyseme.

The picture of the on-line processing of metonymy created by Frisson and Pickering

(1999; 2001) is this: polysemous representations of a word are represented in the lexicon by one

underspecified meaning of that word, such that activating that underspecified meaning makes all

the senses of the polyseme accessible to the processor. Once the senses are all active, processors

can integrate more general linguistic information, such as context, to select one sense. What

remains unclear, however, is the effect that prior context can have on the processing of an

existing polyseme at the point of lexical access of that polyseme. Based on the Underspecified

Model proposed by Frisson and Pickering, (2001), we propose that there are two contrastive

alternatives for modeling prior context’s influence on sense selection:

(3) a. Underspecified - Early Selection Model:When a polyseme is encountered, all senses of the word are accessible to the processor since an underspecified representation of the polyseme has been activated. However, when a strong biasing context has already been established which selects for one of the senses over the others, the processor will make an immediate selection of the salient sense, and will suppress the non-salient senses.

b. Underspecified - Late Selection Model:When a polyseme is encountered, all senses of the word are accessible to the processor since an underspecified representation of the polyseme has been activated. Those senses remain accessible even when a strong biasing context has already been established. All senses of the word remain accessible until disambiguating information following the

Lexical Access of Polysemes 10

Page 11: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

word is encountered, such that the processor can choose one sense and suppress the others. Contextual information is used later in interpretation.

At present, there is little direct evidence to support one model of lexical access over the other.

However, findings from related fields provide indirect, theoretical support for each. Support for

the Underspecified-Early Selection Model would be compatible with findings that contextual

information is processed at the point of lexical access in other domains of language (see van

Berkum, 2008 and Federmeier, 2007 for reviews), such as counterfactuals (Nieuwland & Martin,

2012; Nieuwland & van Berkum, 2006), inferencing about the properties of the speaker (van

Berkum et al., 2008), and and semantic expectancy (Federmeier & Kutas, 1999). On the other

hand, support for the Underspecified-Late Selection Model would be compatible with models of

comprehension that posit a delayed contribution of context (e.g. Sanford & Garrod 2005, see

Nieuwland & Martin, 2012 for review). Testing the predictions made by these contrasting models

will be the primary goal of the experiment described below.

Experimental Studies

Our experiment was divided into two separate studies. We first conducted a norming

study, in which participants rated sentences, half of which were metonymic sentences and half of

which were literal counterparts to the metonymic sentences, for naturalness. Then, we conducted

a self-paced reading study in which participants read the sentence items from the norming study,

and their reading times were analyzed to assess encountered difficulty over varying biasing

contexts and sense interpretations of the polyseme.

Lexical Access of Polysemes 11

Page 12: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Experiment 1A: Norming Study

In the norming study, we elicited judgements of naturalness on a 1-7 scale for 26 items.

We found a main effect for Context on naturalness, with Metonymic Context items being rated as

significantly more natural than Neutral Context items. We removed the two sentence items

whose mean differences on the naturalness rating between Context conditions were largest, and

found no main effect of Context. These remaining 24 items were used in the self-paced reading

study, as described below.

Procedure

Sentence items were presented on a laptop computer screen. Participants read each

sentence pair at a natural pace, and then gave a score from 1 to 7 on the following question: ‘how

naturally did the second sentence flow from the first?’ The experiment lasted approximately 15

minutes.

Subjects

16 college-aged students from the Claremont Colleges were recruited for the norming

study. All participants were fluent English speakers. Participants were not compensated for their

time.

Items

All sentence pair items were created by the investigators. Each item contained four

sentence pairs, as illustrated in Table 1, which varied over biasing Context of the first sentence

(Metonymic or Neutral) and Verb of the second sentence’s proper-name polyseme (Literal-

Selecting or Figurative-Selecting). We used sentence pairs such as those in (4):

(4) Example Item From Norming Study

Lexical Access of Polysemes 12

Page 13: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

a. Metonymic Context, Literal-Selecting VerbIt was the height of English theater and many new plays were coming out.Shakespeare was bothered by the aspiring actors, who were eager for attention.

b. Metonymic Context, Figurative-Selecting VerbIt was the height of English theater and many new plays were coming out.Shakespeare was rehearsed by the aspiring actors, who were eager for attention.

c. Neutral Context - Literal-Selecting VerbIt was the height of English theater and the Globe was bustling with new talent.Shakespeare was bothered by the aspiring actors, who were eager for attention.

d. Neutral Context - Figurative-Selecting VerbIt was the height of English theater and the Globe was bustling with new talent.Shakespeare was rehearsed by the aspiring actors, who were eager for attention.

Within each item, the second sentence was identical in all four sentence pairs except for the main

verb, which determined what sense the comprehender would need to select for the sentence-

initial polyseme. The length and frequency of the second sentences’ disambiguating verbs were

controlled across the Verb condition, with frequency statistics calculated using the English

Lexicon Project’s Hyperspace Analogue to Language frequency data (Balota et al., 2007). For

the 24 items which were used in the self-paced reading study, there was no significant difference

between the literal-selecting (e.g. bothered) and figurative-selecting (e.g. rehearsed) verbs in

frequency (paired t-test, t(23) = 1.26, p = NS) or length (paired t-test, t(23) = 0.3027, p = NS).

Items were presented in random order, with filler items interspersed. All sentences used in

Experiment 1B appear in the Appendix.

Results

The data was subjected to by-subject and by-item ANOVAs, and further contrasts were

tested with paired t-tests. When controlling for differences between subjects, the main effect for

Lexical Access of Polysemes 13

Page 14: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Context on the naturalness ratings was significant, F1(1, 25) = 6.29, p < 0.05. This same main

effect was not significant when controlling for differences between items. A paired t-test revealed

a significant difference, t(15) = -2.16, p < 0.05 on naturalness ratings between sentence pairs that

were in the Neutral Context and Figurative-Selecting Verb conditions and sentence pairs that

were in the Metonymic Context and Figurative-Selecting Verb conditions, when controlling for

differences between subjects. All other t-test comparisons between conditions were not

significant. A comparison of means revealed that sentence pairs with Metonymic Context (M =

5.15, SD = 1.53) were rated as more natural than sentence pairs in which the first sentence had

Neutral Context (M = 4.99, SD = 1.60).

Since we aimed to have items which were balanced, we removed the two sentence items

whose differences in naturalness between the Metonymic and Neutral Context conditions were

greatest. We then subjected the new item set to the same ANOVAs as described above. We found

that when controlling for differences between subjects, the main effect for first-sentence context

on the naturalness ratings was not significant, F1(1, 15) = 2.49, p > 0.05. Likewise, when

controlling for differences between items, the main effect for first-sentence context on the

naturalness ratings was not significant, F2(1, 23) = 0.977, p > 0.05.

Discussion

In the norming study, participants reported the naturalness of 26 sentence pairs which

varied by the contents of the entire first sentence and the main verb of the second sentence. We

found a significant effect for naturalness across the items, with items biasing for a metonymic

interpretation in the first sentence being rated as more natural than items not biasing for either a

metonymic or literal interpretation in the first sentence. After removing the two items whose

Lexical Access of Polysemes 14

Page 15: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

differences in mean naturalness ratings were greatest across the Context condition, we found no

significant effect on naturalness ratings. We used the resulting 24 items for the self-paced reading

study described in Experiment 1B.

Experiment 1B: Self-Paced Reading Study

In the self-paced reading study, we examined the effects of context at the point of lexical

access in producer-for-product metonymic sentences. Our items were sentence pairs which were

drawn from the normed items in Experiment 1A, and which varied by Context of the first

sentence (Neutral or Metonymic) and Verb of the second (target) sentence’s proper-name

polyseme (Literal-Selecting or Figurative-Selecting). Within these sentence pairs, the target

sentence was divided into regions, as illustrated in (5):

(5) Example Item From Self-Paced Reading Study, with marked regions in target sentence.

a. Metonymic Context, Literal-Selecting VerbIt was the height of English theater and many new plays were coming out.Shakespeare / was bothered / by the aspiring actors, / who were eager / for attention.

b. Metonymic Context, Figurative-Selecting VerbIt was the height of English theater and many new plays were coming out.Shakespeare / was rehearsed / by the aspiring actors, / who were eager / for attention.

c. Neutral Context - Literal-Selecting VerbIt was the height of English theater and the Globe was bustling with new talent.Shakespeare / was bothered / by the aspiring actors, / who were eager / for attention.

d. Neutral Context - Figurative-Selecting VerbIt was the height of English theater and the Globe was bustling with new talent.Shakespeare / was rehearsed / by the aspiring actors, / who were eager / for attention.

The first sentences in (5a) and (5b) have Metonymic Context, meaning that they contain

information which we expected would bias the language processor towards interpreting the

Lexical Access of Polysemes 15

Page 16: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

proper-name polyseme in the target sentence using its figurative sense. Context-setting is

achieved in (5a) and (5b) with the mention of plays. In contrast, the first sentences in (5c) and

(5d) have Neutral Context, meaning that they do not contain information which we expected

would bias a language processor to interpret the proper-name polyseme in the second sentence

using its figurative sense rather than its literal sense, or vice versa.

The target sentence in (5a) and (5c) is from the Literal-Selecting Verb condition, meaning

that the main verb in those sentences, bothered, has lexical constraints which force the language

processor to interpret the polyseme, Shakespeare, using its literal sense. Specifically, bothered

takes an animate theme, which corresponds to the literal , ‘person-reading’ of Shakespeare. In

contrast, the target sentences in (5b) and (5d) are in the Figurative-Selecting Verb condition,

meaning that the main verb in those sentences, rehearsed, has lexical constraints which force the

language processor to interpret the polyseme using its figurative sense. Rehearsed must take an

inanimate, ‘rehearsable’ theme, which only corresponds to the figurative sense of Shakespeare

Crucially, in all of our items, the main verb in the second sentence determines what sense the

language processor must use in interpretation; selection of the wrong sense will lead to a

semantic mismatch upon reaching the verb, which will cause difficulty in reading (Traxler,

2012) .

The observed differences in processing of sentences such as (5a) - (5d) above helped to

determine whether the Underspecified-Early Selection Model or the Underspecified-Late

Selection Model best predicts the behavior of language processors. According to the

Underspecified-Early Selection Model, the language processor integrates salient contextual

information at the point of lexical access, and will immediately choose one sense of a polyseme

Lexical Access of Polysemes 16

Page 17: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

over the others. The Metonymic Context sentences in the sentence pairs above, (5a) and (5b),

strongly bias towards a figurative sense reading of Shakespeare on the second line. Therefore,

according to the Underspecified-Early Selection Model, the language processor will select the

figurative sense of Shakespeare at the point of lexical access of that word, and will move forward

in processing the next words with an interpretation that Shakespeare refers to the works of

Shakespeare, and not the playwright himself. If this is the case, then processors will encounter

difficulty in processing the sentence if information presented after Shakespeare necessitates that

they interpret Shakespeare using the word’s literal sense, as is the case in sentences such as (5a).

The Underspecified-Early Selection Model predicts that in sentences with Neutral Context, such

as (5c) and (5d), the lack of a strong biasing context will lead to processors maintaining all

senses of the polyseme Shakespeare until later disambiguating information becomes available.

Thus, the model predicts that there will be no difference in processing for when the polyseme

must be interpreted using its literal versus figurative sense when Context is Neutral.

According to the Underspecified Late-Selection Model, language processors will not

integrate information from a strongly biasing context to select one sense of a polyseme at the

point of lexical access. Therefore, in sentence pairs such as (5a) and (5b), the language processor

will not commit to the figurative sense of Shakespeare immediately upon encountering the

polyseme, despite that they were given a context which strongly biases towards the figurative

sense of Shakespeare. Processors will only commit to one sense of Shakespeare based on

information presented after the polyseme has been processed, so there will be no greater

difficulty in processing sentence pairs such as (5a), in which the processor must select the sense

of the polyseme that is incongruous to the prior biasing context, than the sentence pairs such as

Lexical Access of Polysemes 17

Page 18: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

(5b), in which the processor must select the sense of the polyseme that is congruous to the prior

biasing context. The Underspecified Late-Selection Model also predicts that in sentences with

neutral context, the lack of a strong biasing context will lead to processors maintaining all senses

of the polyseme Shakespeare until later disambiguating information becomes available.

Therefore, according to this model, there will be no difference in processing when the polyseme

must be interpreted using its literal versus figurative sense.

The critical test between these two contrasting models was in processing of sentences

such as (5a), wherein processors are given a strong prior biasing information towards the

figurative sense of the polyseme in the second sentence, but must nonetheless interpret the

polyseme using its literal sense due to the thematic role constraints of the main verb in the

second sentence. The Underspecified-Early Selection model predicts that processors will

encounter difficulty in processing these sentences, since processors will need to reinterpret the

polyseme and reselect its sense upon reaching the verb, whereas the Underspecified-Late

Selection Model predicts that processors will have no such difficulty, since all senses of the

polyseme will still be accessible to them upon reaching the main verb.

In the present experiment, we tested the predictions made by these two models using a

self-paced reading design (Aaronson & Scarborough, 1976; Mitchell & Green, 1978).

Participants read sentence pairs such as the ones in (5). The first sentence of each pair was

presented all at once, such that all the words in the sentence were on screen at one time. The

second sentence was presented by region, as in illustrated in Figure 2, with the polyseme, the

main verb, and the three post-verb regions all in separate regions from one another. Participants’

reading times for each region were recorded. Since encountering difficulty in reading has been

Lexical Access of Polysemes 18

Page 19: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

shown to correlate with longer reading times on the regions causing difficulty in self-paced

reading (Ferreira & Henderson, 1990; Traxler et al., 2002), we were able to evaluate the relative

difficulty that participants had in interpreting sentence pairs by comparing the relative reading

times for the verb and post-verb regions of the second sentence in each pair across context and

sense conditions.

Participants

38 native English-speaking participants were recruited to participate in the self-paced

reading study. All were students at the Claremont Colleges. None took part in the pretesting of

the sentence materials. Participants were compensated $10 for their time. The experiment lasted

approximately 30 minutes.

Items

We used the 24 items from the norming study whose differences in naturalness ratings

between the metonymic and neutral Context levels were least. Items varied across Context and

Verb conditions, as is illustrated in Table 1. These items comprise the first 24 items in the

Appendix.

Procedure

Each participant was run individually. The experimenter brought the participant into the

testing room, where the participant sat at a computer station with a keyboard in front of it. The

participant read through the instructions on-screen and completed a number of practice items

while the experimenter remained in the testing room. The instructions specified that participants

should read at a normal pace, and keep both hands on the keyboard during experimental trials.

Participants had the opportunity to ask the experimenter about the procedure at any time during

Lexical Access of Polysemes 19

Page 20: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

the instructions and practice trials. When the practice trials were over, the experimenter left the

room and the participant began the experimental trials.

During the experiment, participants read sentence pairs such as the one above using

Linger, a self-paced reading program. For each experimental item, the participant was initially

presented with two sentences on-screen, both of which had every character masked by a dash.

Each time the participant pressed the space bar, a new region was unmasked, and the previous

section was re-masked. After each experimental item, participants answered a forced-choice

comprehension question about the previous sentence pair. Each participant completed 24

experimental trials, which were interspersed randomly with 86 other filler trials and trials from

unrelated self-paced reading experiments.

Results

Prior to analysis, we excluded two participants’ data (5.2%) because the participants self

identified as potentially non-native English speakers. We also excluded all data from item 24

(4.7%) because an error in the formatting of the sentence items caused the item to be displayed

with too many regions. Finally, we removed all reaction times that were three standard deviations

or more away from the mean for the region that the reaction time corresponded to (1.8%). This

was done to eliminate responses in which the participant accidentally moved from one region to

the next without reading the first region or trials in which the participant may have been

distracted.

We first examined participants’ general comprehension of the sentence items by

analyzing question response data. Before removing all questions in which the response time was

greater than three standard deviations from the mean, we found that participants answered

Lexical Access of Polysemes 20

Page 21: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

incorrectly on 2.0% of trials. After removing all trials in which participants answered the

question incorrectly and removing responses that were more than three standard deviations away

from the mean, we conducted two-way ANOVAs to determine if participants had more difficulty

answering questions based on the corresponding sentences’ Context (Metonymic and Neutral)

and Verb (Figurative-Selecting and Literal-Selecting) conditions. We found a trend for the main

effect of the Verb when controlling for differences in subjects, F1(1, 35) = 3.38, p = 0.07, and

items, F2(1,22) = 3.22, p = 0.09. We then conducted t-tests to determine the direction of the

effect. We found a significant difference in Literal-Selecting and Figurative-Selecting Verb

conditions when the Context condition was Neutral, when controlling for differences in subjects,

t(35) = -2.19, p < 0.05, and that same effect trended when controlling for differences items, t(22)

= -1.93, p = 0.06. However, this effect was not significant when the Context was Metonymic.

Across all the items, we analyzed response times to all the regions of the target sentence

(Region 1 through Region 5) by subjecting the data to separate by-subject and by-item ANOVAs,

crossing Context and Verb conditions. Trials were only considered in these analyses if the

participant correctly answered the corresponding question correctly. All significant effects from

all ANOVAs that were run are reported here.

Region 3

When controlling for differences in subjects, there was a significant effect of both

Context, F1(1,35) = 22.71, p < 0.001, and Verb, F1(1, 35) = 27.73, p < 0.001, on response times

(RTs) for Region 3. When controlling for differences in items, both Context, F2(1,22) = 8.60, p <

0.01, and Verb, F2(1,22) = 13.32, p < 0.001, were significant for Region 3. The computed means

Lexical Access of Polysemes 21

Page 22: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

and standard deviations for Region 3 across Verb and Context can be seen in Table 2 below, and

the main effects of Context and Verb are illustrated in Figure 2.

We conducted t-tests across Context and Verb pairings to determine the direction of the

effects in Region 3. First, we examined the data to determine the effect direction of Context.

There was a significant difference in reaction times between Neutral Context - Figurative-

Selecting Verb and Metonymic Context - Figurative-Selecting Verb items, both when controlling

for differences in subjects, t(35) = 3.77, p < 0.001, and items, t(22) = 2.64, p < 0.05. Differences

in RTs between Neutral Context - Literal-Selecting Verb and Metonymic Context - Literal-

Selecting Verb items were significant when controlling for differences in subjects, t(35) = 2.39, p

< 0.05, and trended towards significance when controlling for differences items, t(22) = 1.98, p =

0.06. These results demonstrate that participants’ RTs were faster on items in which the Context

was Metonymic than Neutral, with greater differences observed when the Verb was Figurative-

selecting than Literal-selecting. Means and standard deviations for response times to Region 3

across the Context condition can be found in Table 3.

We then ran tests to determine the direction of the main effect of the Verb condition.

There was a significant difference in reaction times between Neutral Context - Figurative-

Selecting Verb and Neutral Context - Literal-Selecting Verb items, both when controlling for

subjects, t(35) = 4.05, p < 0.001, and items, t(22) = 4.23 p < 0.001. Differences between

Metonymic Context - Figurative-Selecting Verb and Metonymic Context - Literal-Selecting Verb

items were significant when controlling for differences in subjects, t(35) = 3.93, p < 0.001, and

trended when controlling for differences in items, t(22) = 2.03, p = 0.05. These results suggest

that, when the Context was Neutral, participants’ RTs were faster when the Verb condition was

Lexical Access of Polysemes 22

Page 23: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Literal-Selecting than Figurative-Selecting, but when the Context was Metonymic, there was no

difference in RTs between the Verb conditions. Means and standard deviations for response times

to Region 3 across the Context condition can be found in Table 4.

Discussion

The results from Experiment 1B suggest that participants had more difficulty processing

the target sentence when the polyseme needed to be interpreted using its figurative sense, as

compared to its literal sense. This effect was realized in Region 3, and was a spillover effect from

the difficulty encountered in the previous region, which contained the disambiguating verb.

Additionally, the results suggest that biasing context did have a facilitatory effect on

reinterpretation, since participants were faster to process the target sentence using the Figurative

Verb of the polyseme after reading the Metonymic Context sentence, as compared to the Neutral

Context sentence. These results suggest that participants selected the literal sense immediately

upon processing the polyseme based on non-contextual information, and used contextual

information to facilitate revision of the initial reanalysis.

General Discussion

In the present experiment, we tested whether salient contextual information biasing for

one sense of a polyseme over the others is used by the language processor to immediately make a

sense selection during lexical access. We found that context does not influence the language

processor’s first interpretation of a polyseme within a sentence; the language processor always

chooses a literal interpretation of the polyseme first, even when the preceding context biased for

a figurative interpretation of the polyseme. This suggests that the language processor’s lexical

Lexical Access of Polysemes 23

Page 24: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

access was first guided by some bias for a literal interpretation of the polyseme, and contextual

information was used in subsequent re-interpretations of that polyseme. We believe that this

preference for a literal interpretation may be the result of a default thematic-role analysis of the

syntactic structure of the target sentence. This interpretation of the results, along with other

interpretations, are discussed below.

Underspecified Model

The results obtained in the present experiment are not explicitly predicted by the

Underspecified Model of metonymy processing, as was proposed by Frisson and Pickering

(1999; 2001). According to their model, when a metonym is encountered, the language processor

activates an underspecified meaning of it, and all possible senses of the metonym are accessible

by the language processor once that underspecified meaning has been activated. Furthermore, the

model predicts that the language processor will use linguistic information, such as lexical

constraints, to home in on a specific sense of the metonym when such linguistic information is

available. Finally, according to Frisson and Pickering’s Underspecified Model, “a wrong sense

will never be assigned” (2001) to a polyseme, since all senses will remain available to the

language processor.

Our results demonstrate that the language processor can make a mistake in assigning a

sense to a polyseme. This was evident because when the language processor encountered a

figurative-selecting verb, it took longer to process Region 3, suggesting that the language

processor had initially assigned the wrong sense to the polyseme and had to reinterpret it. While

it remains possible that this result obtained because the metonymic context biased more strongly

Lexical Access of Polysemes 24

Page 25: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

for the literal sense of the polyseme than the figurative sense of the polyseme, such a conclusion

could only be drawn by further testing of our experimental items.

There are a number of models of polsemy processing which predict the above result, but

which remain unlikely on grounds independent to the results obtained here. A Literal-First

Model, in which the literal sense of the polyseme is always accessed first, correctly predicts our

experimental results; however, since a Literal-First Model is incompatible with the evidence on

polysemey processing presented by Frisson and Pickering (1999), we will not consider it here. A

Frequency-Based Model, in which senses are accessed in order of their frequency in natural

language, is unlikely for these same reasons.

We now propose two alternative explanations for the experimental results obtained here.

The first explanation is the Subject as Agent Principle:

(6) Subject as Agent PrincipleWhen the language processor encounters a noun phrase (NP) in sentence-initial position, it assigns that NP an agent theta-role, unless there is evidence to the contrary.

In reading the target sentences in our experiment, such as, Shakespeare was rehearsed by the

aspiring actors, who were eager for attention, the language processor may have assigned a

thematic role to the polyseme based upon the syntactic structure of the sentence. Specifically, the

processor may have read the proper noun in sentence-initial position and, based on the

information about the features of the noun and its position in the sentence, made a default

assignment of an agent role on the noun. Evidence from Clifton (1993) suggests that the

language processor assigns the agent theta-role early in processing, even before determining

whether the lexical item being assigned to the agent role is animate or inanimate. Then, based on

Lexical Access of Polysemes 25

Page 26: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

the assumption that the proper noun was an agent in the sentence, the language processor may

have determined that the proper noun was animate, since animates in sentence-initial position are

typically assigned agentive roles (Traxler, 2012). Since the literal sense of the proper noun would

yield a representation of the word such that it was animate, but the figurative sense of the proper

noun would not have, the language processor may have thus committed strongly to the literal

sense interpretation of the polyseme, and would have inhibited the figurative sense interpretation.

Evidence from Broca’s aphasics (Zurif, 1995) suggests that the language processor does perform

these kinds of simplistic syntactic analyses, and models of ‘good enough’ language processing

support the notion that even non-aphasic language processors readily make such assumptions

(Ferreira et al., 2002). Likewise, these findings fits within the broader framework of language

processing models which posit that the language processor will use any piece of linguistic

information that is available in order to perform interpretation as fast as possible (see van

Berkum, 2008).

The Underspecified Model can be made compatible with the Subject as Agent Principle

simply by expanding the scope of linguistic information which the language processor may use

to select one sense of a polyseme over the others. Under this revised version of the

Underspecified Model, the language processor may use information about a polyseme’s syntactic

position to assign a specific sense interpretation to the polyseme. Such a model would predict

that we could replicate the results found here using other syntactic structures, provided that those

structures also provide sufficient reason for the language processor to assign thematic roles even

when there are no lexical constraints on which thematic roles could be assigned. Furthermore,

this model could be tested using different polysemic relations, such as location-for-group. We

Lexical Access of Polysemes 26

Page 27: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

would expect that in such cases, the language processor would assign thematic roles different

than the agent-role assignment observed here, based on the features of those polysemes. For

example, if the metonymic relationship were location-for-group, it seems likely that the location

reading would be preferred based on syntactic information alone, since it is the literal reading.

Consider the following sentence pair in (7):

(7) Many delegations came to the Olympic Committee’s meeting to make proposals for the Summer Games.

Chicago was quizzed for an entire morning about hotel capacity and transportation.

In (7), the context sentence biases for the figurative interpretation of Chicago, which corresponds

to the group reading of the polyseme. However, based on our findings in Experiment 1B, we

believe it may be the case that the language processor will use structural cues to reach the literal

sense of polyseme first. Such a finding would confirm that the language processor uses structural

information to reinforce the literal sense reading of the polyseme, whether that literal sense is

agentive or not, rather than using the structural information only to support an agentive reading

regardless of what the polyseme is.

Another possible explanation, which we will call the Lexical Constraint Principle

explanation, claims that the the language processor uses the absence of lexical information as

reasoning for selection of the literal sense:

(8) Lexical Constraint PrincipleUnless there are immediate, lexical-level constraints on the polyseme which necessitate that the language processor do otherwise, the language processor will always assign a literal meaning to a polyseme.

Under the Lexical Constraint Principle explanation of our results, when the language processor

encounters a polyseme, such as Shakespeare, but that polyseme has no lexical constraints placed

Lexical Access of Polysemes 27

Page 28: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

upon it, then the language processor will make a default selection of the literal sense of that

polyseme.

The Underspecified Model is not compatible with the Lexical Constraint Principle.

Crucially, the Underspecified Model predicts that the representation of a polyseme will remain

underspecified until the language processor has sufficient linguistic information to select one of

its senses. However, the Lexical Constraint Principle predicts that when there is not sufficient

linguistic information available, the language processor will simply make a default sense

selection. Since Frisson and Pickering (1999) did not test the processing of polysemeies in the

absence of lexical constraints, it is impossible to rule out this possibility based upon their results.

However, the results presented by Frazier and Rayner (1990) are not predicted by the Lexical

Constraint Principle. In their experiment, participants read sentences such as the following:

(9) Sentence Items from Frazier and Rayner (1990) a. Apparently, the dinner wasn’t very enjoyable, ending early. b. Apparently, the dinner wasn’t very enjoyable, tasting burned.

In sentences such as the ones in (9), the Lexical Constraint Principle predicts that when the

language processor reaches the polyseme, dinner, and finds no lexical constraints upon it, the

language processor will assign a default sense to it. Since the polysemeic relationship between

the senses of dinner in (9) is not the same as the producer-product metonymic relationship we

considered in the present experiment, it is not clear which sense will be assigned by default here

under the Lexical Constraint Principle. However, the important prediction of the Lexical

Constraint Principle here is that it predicts that the language processor will assign some default

sense to dinner since there are no lexical constraints to guide sense selection of the polyseme. In

contrast to these predictions, Frazier and Rayner (1990) found that in instances of polysemy such

Lexical Access of Polysemes 28

Page 29: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

as the one in (9), the language processor activates an underspecified representation of dinner, and

does not commit to a specific sense of the polyseme until the disambiguating region (after the

second comma) has been reached. Thus, no default sense selection occurs upon first

encountering dinner, as was predicted by the Lexical Constraint Principle.

While both the Subject as Agent Principle and the Lexical Constraint Principle correctly

predict the results obtained in the present experiment, we reject the Lexical Constraint Principle

since it fails to predict the findings from Frazier and Rayner (1990) on polysemy processing.

However, further investigation into this phenomenon should be conducted in order to ensure that

this is indeed the correct conclusion. For example, it may be the case that the reason the Lexical

Constraint Principle does not correctly predict the results found in Frazier and Rayner (1990) is

that different metonymic relationships are processed differently, and that if Frazier and Rayner

(1990) were replicated using producer-for-product metonymies (as were used in the present

experiment), the predictions of the Lexical Constraint Principle would have been confirmed.

Indeed, such findings would be important because they would be the first evidence that different

metonymic relations are processed differently. Finally, more research should be done into

determining how structural information, such as syntactic positions of polysemes, is used by the

language processor in processing polysemy.

Integration of Context into Lexical Access of Polysemes

The results obtained in the present experiment support the notion that contextual

information has a relatively late effect on lexical access, as compared to lexical constraints and

structural information from word order. Even when the context preceding the polyseme biased

for the figurative sense of the polyseme, participants processed Region 3 significantly faster

Lexical Access of Polysemes 29

Page 30: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

when the polyseme needed to be interpreted using its literal sense as compared to its figurative

sense, suggesting that they always interpreted the polyseme using its literal sense first. This

literal sense reading must have been reached using information other than the preceding biasing

contextual information, since the contextual information biased for a figurative sense reading of

the polyseme. Thus, whatever information the language processor used to reach the literal sense

reading was integrated into lexical access before the contextual information, meaning that the

contextual information was used relatively late in lexical access. We have posited that the

language processor used structural information to arrive at the literal sense reading of the

polyseme. It would be incorrect to conclude that the contextual information was not used at all,

however, since the results show that when the polyseme needed to be interpreted using its

figurative sense, participants were faster to process Region 3 following a context biasing for the

figurative sense, as compared to a context biasing for neither sense. This result demonstrates that

the contextual information was used to facilitate reinterpretation of the polyseme, and that our

experiment was sensitive to its effect.

A remaining question, then, is whether there is competition between contextual and other

types of linguistic information in polyseme processing. It may be the case that when the language

processor encounters contextual information which strongly biases for the figurative

interpretation of a polyseme, and later encounters structural or lexical information biasing for a

literal interpretation of a polyseme, it encounters added difficulty associated with the cost of

inhibiting the figurative-biasing information source while selecting the literal sense of the

polyseme. Our results suggest that there is no such competition, since when the polyseme needed

to be interpreted using its figurative sense, they were faster to process Region 3 following a

Lexical Access of Polysemes 30

Page 31: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Metonymic Context as opposed to a Neutral Context. This result implies that competition does

not occur because the metonymic-biasing context primes the language processor to interpret the

polyseme with a sense that is incongruent with the structural information they eventually use to

arrive at the literal sense reading, whereas the neutral context does not prime for the incongruent

sense. However, this question should be explored further in order to understand the contributions

of context into the lexical access of polysemes. It may be the case that such competition effects

occur in regions other than Region 3, and that our self-paced reading measure was not sensitive

enough to detect the effect. Replicating the present study using eye-tracking would help to

elucidate whether such competition effects occur.

Our results are straightforwardly compatible with models of language processing which

posit that the language processor has some order in which it considers sources of linguistic

information, at least when first attempting to interpret a sentence (Rayner et al., 1983; Pickering

& Traxler, 1998). The results of Experiment 1B suggest that structural information is privileged

over contextual information in first-pass interpretation.

Future Directions

Future research should aim to determine whether the language processor truly employs a

static order in what linguistic information sources it uses in polyseme disambiguation, and if

there is a static order, what that order is. Additionally, more work must be done to understand the

nature of structural information in processing. This line of research would benefit greatly from

cross-linguistic research, since speakers of different languages may use structural information in

different ways, depending upon the word order of their languages. Finally, in coming to

understand polysemy, it will be important to determine if all metonymic relationships are

Lexical Access of Polysemes 31

Page 32: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

processed the same way. Some research (e.g. Frisson & Pickering, 1999) has suggested that all

metonymic relations are processed alike, though as was described above, the thematic-role

differences that stem from different metonymic relations may cause differences in processing.

Many of these research questions may be addressed through careful replication of the experiment

described here, using different metonymic relationships, contextual information, and more

sensitive measures, such as eye-tracking.

Conclusions

The present experiment provides evidence that in polysemy processing, the language

processor uses structural information to arrive at a literal sense interpretation of a polyseme even

before encountering lexical constraints on that polyseme which would otherwise guide sense

selection. The Underspecified Model of metonymy processing can be made compatible with this

new finding by expanding the scope of sources of linguistic information which can be used in

immediate sense selection to include information about the placement of the metonym in the

sentence’s word order. The present study also finds that context has a relatively late effect in

comparison to lexical constraints and structural information. In sum, these findings support the

notion that the language processor is ‘opportunistic,’ using all possible sources of information

during interpretation, and that the language processor consistently uses some sources of

linguistic information before others.

Lexical Access of Polysemes 32

Page 33: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

References

Aaronson, D., & Scarborough, H. S. (1977). Performance theories for sentence coding: Some

quantitative models. Journal of Verbal Learning and Verbal Behavior, 16

Balota, D.A., Yap, M.J., Cortese, M.J., Hutchison, K.A., Kessler, B., Loftis, B., Neely, J.H.,

Nelson, D.L., Simpson, G.B., & Treiman, R. (2007). The English Lexicon Project. Behavior

Research Methods, 39, 445-459.

Carroll, D. W. (2007). Psychology of language (5th ed.). USA: Wadsworth Publishing.

Clifton, C. (1993). Thematic roles in sentence parsing. Canadian Journal of Experimental

Psychology, 47(2)

Federmeier, K., D. (2007). Thinking ahead: The role and roots of prediction in language

comprehension. Psychophysiology, 44(4)

Federmeier, K., D., & Kutas, M. (1999). A rose by any other name: Long-term memory structure

and sentence processing. Journal of Memory and Language, 41

Ferreira, F., & Henderson, J. M. (1990). Use of verb information in syntactic parsing: Evidence

from eye movements and word-by-word self-paced reading. Journal of Experimental

Psychology, 16(4), 555-568.

Foraker, S. and Murphy, G.L. (2009, March). Polysemy in sentence comprehension: effects of

sense dominance. Poster presented at the 22nd Annual Meeting of the CUNY on Human

Sentence Processing, Davis, CA.

Lexical Access of Polysemes 33

Page 34: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Frazier, L., & Rayner, K. (1990). Taking on semantic commitments: Processing multiple

meanings vs. multiple senses. Journal of Memory and Language, 29, 181-200.

Frisson, S. (2009). Semantic underspecification in language processing. Language and

Linguistics Compass, 3(1), 111-127.

Frisson, S., & Pickering, M. J. (1999). The processing of metonymy: Evidence from eye

movements. Journal of Experimental Psychology, 25(6), 1366-1383.

Frisson, S., & Pickering, M. J. (2001). Obtaining a figurative interpretation of a word: Support

for underspecification. Metaphor and Symbol, 16(3&4), 149-171.

Frisson, S., & Pickering, M. J. (2007). The processing of familiar and novel senses of a word;

why reading dickens is easy but reading needham can be hard. Language and Cognitive

Processes, 22(4), 595-613.

Hogaboam, T. W., & Perfetti, C. A. (1975). Lexical ambiguity and sentence comprehension.

Journal of Verbal Learning and Verbal Behavior, 14, 265-274.

Klein, D. E., & Murphy, G. L. (2002). Paper has been my ruin: Conceptual relations between

polysemous senses. Journal of Memory and Language, 47

McElree, B., Frisson, S., & Pickering, M. J. (2006). Deferred interpretations: Why starting

dickens is taxing but reading dickens isn’t. Cognitive Science, 30, 181-192.

Mitchell, D. C., & Green, D. W. (1978). The effects of context and content on immediate

processing in reading. Quarterly Journal of Experimental Psychology, 30

Lexical Access of Polysemes 34

Page 35: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Nieuwland, M. S., & Martin, A. E. (2012). If the real world were irrelevant, so to speak: The role

of propositional truth-value in counterfactual sentence comprehension. Cognition, 122,

102-109.

Nieuwland, M. S., & Van Berkum, J. J. A. (2006). When peanuts fall in love: N400 evidence for

the power of discourse. Journal of Cognitive Neuroscience, 18(7)

Nunberg, G. (1979). The non-uniqeness of semantic solutions: Polysemy. Linguistics and

Philosophy, 3(2)

Nunberg, G. (1995). Transfers of meaning. Journals of Semantics, 12

Nunberg, G. (2004). The pragmatics of deferred interpretation: In L. Horn, & G. Ward (Eds.),

The handbook of pragmatics. Oxford: Blackwell.

Pickering, M. J., & Frisson, S. (2001). Processing ambiguous verbs: Evidence from eye

movements. Journal of Experimental Psychology: Learning, Memory and Cognition, 27(2)

Pickering, M. J., & Garrod, S. (2007). Do people use language production to make predictions

during comprehension? Trends in Cognitive Sciences, 11(3), 105-110.

Pickering, M. J., & Traxler, M. J. (1998). Plausibility and recovery from garden paths: An eye-

tracking study. Journal of Experimental Psychology, 24(4), 940-961.

Pylkkänen, L., Llinas, R., & Murphy, G. L. (2006). The representation of polysemy: MEG

evidence. Journal of Cognitive Neuroscience, 18(1), 97-109.

Pylkkänen, L., & Marantz, A. (2003). Tracking the time course of word recognition with MEG.

Trends in Cognitive Sciences, 7

Lexical Access of Polysemes 35

Page 36: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Pylkkänen, L., & McElree, B. (2006). The syntax-semantic interface: On-line composition of

sentence meaning. In M. J. Traxler, & M. A. Gernsbacher (Eds.), Handbook of

psycholinguistics (2nd ed., pp. 537-577). New York: Elsevier.

Rayner, K., Carlson, M., & Frazier, L. (1983). The interaction of syntax and semantics during

sentence processing: Eye movements in the analysis of semantically biased sentences.

Journal of Verbal Learning and Verbal Behavior, 22

Sanford, A. J., & Garrod, S. C. (2005). Memory-based approaches and beyond. Discourse

Processes, 39

Simpson, G. B. (1981). Meaning dominance and semantic context in the processing of lexical

ambiguity. Journal of Learning and Verbal Behavior, 20, 120-136.

Swets, B., Desmet, T., Clifton, C., & Ferreira, F. (2008). Underspecification of syntactic

ambiguities: Evidence from self-paced reading. Memory and Cognition, 36(1)

Swinney, D. A. (1979). Lexical access during sentence comprehension: (re)consideration of

context effects. Journal of Verbal Learning and Verbal Behavior, 18, 645-659.

Traxler, M. J. (2012). Semantic effects. Introduction to psycholinguistics (1st ed., pp. 159-161).

London: Wiley-Blackwell.

Traxler, M. J., Pickering, M. J., & McElree, B. (2002). Coercion in sentence processing:

Evidence from eye-movements and self-paced reading. Journal of Memory and Language,

47

van Berkum, Jos J. A. (2008). Understanding sentences in context: What brain waves can tell us.

Current Directions in Psychological Science, 17(6)

Lexical Access of Polysemes 36

Page 37: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

van Berkum, Jos J. A., van den Brink, D., Tesink, Cathelijne M. J. Y., Kos, M., & Hagoort, P.

(2008). The neural integration of speaker and message. Journal of Cognitive Neuroscience,

4(4)

Wheeldon, L., & Waksler, R. (2004). Phonological underspecification and mapping mechanisms

in the speech recognition lexicon. Brain and Language, 90

Zipf, G. K. (1945). The meaning-frequency relationship of words. Psychology, 33

Zurif, E.B. (1995). Brain regions of relevance to syntactic processing. In L. Gleitman & M.

Liberman, Eds., Invitation to Cognitive Science, Vol. I, 2nd edition. Cambridge, MA: MIT

Press

Lexical Access of Polysemes 37

Page 38: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Figures

Figure 1Regions of Target Sentence.

Figure 1. In the self-paced reading experiment, participants read the target sentence, which contained the metonym, region by region. In the experiment, the metonym was always in Region 1, the disambiguating verb was always in Region 2, and Regions 3-5 were postverbal spillover regions.

Figure 2Main Effects of Verb and Context Conditions in Region 3

!""#

!!"#

$""#

$!"#

%""#

%!"#

&""#

&!"#

'()*+,-./01/2/3-4)# 5(6/+,201/2/3-4)#

!"#$%&

#"'()"'*)

#+'

,"-.'

7/6849:(3#

;/*6+,2#

Lexical Access of Polysemes 38

Page 39: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Figure 2. Mean and standard error response times are displayed for Region 3 across Verb and Context conditions. Participants responded faster to Region 3 of the Literal-Selecting Verb sentences faster than Figurative-Selecting Verb sentences across Context conditions. This suggests that participants selected the literal sense of the proper-noun polyseme before reaching the disambiguating verb region (Region 2). Additionally, participants were faster to respond to Region 3 of the Metonymic Context - Figurative-Selecting Verb sentences than they were to respond to Region 3 of the Neutral Context - Figurative-Selecting Verb sentences. This suggests that participants used contextual information to facilitate reinterpretation of the polyseme, and thus that our apparatus was sensitive to the effects of the Context condition.

Tables

Table 1Sentence Items - Context and Verb Conditions

Table 1. Items varied over Context, with Metonymic Context sentences priming for the figurative sense interpretation of the proper-noun polyseme in the target sentence, and Neutral Context sentences priming for no sense of the polyseme over the others. Items also varied over Verb, with Figurative-Selecting Verb sentences containing a verb that necessitates the polyseme be interpreted using its figurative sense, and Literal-Selecting Verb sentences containing a verb that necessitates the polyseme be interpreted using its literal sense.

Lexical Access of Polysemes 39

Page 40: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Table 2Mean and Standard Deviation Response Times of Region 3 Across Verb and Context

VerbVerb

Context Figurative-Selecting Literal-Selecting

Metonymic 669 (407) 578 (299)

Neutral 788 (472) 632 (358)

Table 2. Means and standard deviation (SD) response times across Verb and Context condition levels for Region 3 are listed (SD values in parentheses). Importantly, participants were faster to respond to Region 3 in Metonymic Context - Literal-Selecting Verb items as compared to Metonymic Context - Figurative-Selecting Verb items.

Table 3Mean and Standard Deviation Response Times of Region 3 Across Context

ContextContext

Metonymic Neutral

623 (359) 710 (425)

Table 3. Means and standard deviation (SD) response times across the Context condition levels for Region 3 are listed (SD values in parentheses). Participants were faster to respond to Region 3 in Metonymic Context items as compared to Neutral Context items.

Table 4Mean and Standard Deviation Response Times of Region 3 Across Verb

VerbVerb

Figurative-Selecting Literal-Selecting

729 (444) 605 (331)Table 4. Means and standard deviation (SD) response times across the Verb condition levels for Region 3 are listed (SD values in parentheses). Participants were faster to respond to Region 3 in Figurative-Selecting Verb items as compared to Literal-Selecting Verb items.

Lexical Access of Polysemes 40

Page 41: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Lexical Access of Polysemes 41

Page 42: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Appendix

The following sentence pair items were used in Experiments 1A and 1B. All 26 items were used in Experiment 1A, and the first 24 items were used in Experiment 1B. The “ / ” marks denote where the regions began and ended when the sentence items were used in Experiment 1B. Items 25 and 26 are not marked with regions because they were not used in Experiment 1B.

1 Metonymic Context, Literal-Selecting Verb It was the height of English theater and many new plays were coming out. Shakespeare / was bothered / by the aspiring actors, / who were eager / for attention.

Metonymic Context, Figurative-Selecting VerbIt was the height of English theater and many new plays were coming out. Shakespeare / was rehearsed / by the aspiring actors, / who were eager / for attention.

Neutral Context, Literal-Selecting VerbIt was the height of English theater and the Globe was bustling with new talent. Shakespeare / was bothered / by the aspiring actors, / who were eager / for attention.

Neutral Context, Figurative-Selecting VerbIt was the height of English theater and the Globe was bustling with new talent. Shakespeare / was rehearsed / by the aspiring actors, / who were eager / for attention.

2 Metonymic Context, Literal-Selecting Verb In 1781, Vienna hosted a music festival where many popular orchestral pieces were played. Mozart / was invited / by Vienna's orchestra conductor, / which was / a great honor.

Metonymic Context, Figurative-Selecting VerbIn 1781, Vienna hosted a music festival where many popular orchestral pieces were played. Mozart / was conducted / by Vienna's orchestra conductor, / which was / a great honor.

Neutral Context, Literal-Selecting VerbIn the spring of 1781, a wonderful gala was held at the palace in Vienna. Mozart / was invited / by Vienna's orchestra conductor, / which was / a great honor.

Neutral Context, Figurative-Selecting VerbIn the spring of 1781, a wonderful gala was held at the palace in Vienna. Mozart / was conducted / by Vienna's orchestra conductor, / which was / a great honor.

3 Metonymic Context, Literal-Selecting Verb

Lexical Access of Polysemes 42

Page 43: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

A prominent publishing house recently decided to reprint mystery novels from the past decade. Stephen King / was contacted / by the publisher / shortly after / this decision was made.

Metonymic Context, Figurative-Selecting VerbA prominent publishing house recently decided to reprint mystery novels from the past decade. Stephen King / was printed / by the publisher / shortly after / this decision was made.

Neutral Context, Literal-Selecting VerbRecently, a prominent publishing house elected to move into a larger office. Stephen King / was contacted / by the publisher / shortly after / this decision was made.

Neutral Context, Figurative-Selecting VerbRecently, a prominent publishing house elected to move into a larger office. Stephen King / was printed / by the publisher / shortly after / this decision was made.

4 Metonymic Context, Literal-Selecting Verb Before Cubism became popular, no one cared when Cubist artwork was put on display. Picasso / was approached / by an obscure curator, / but nothing came of it / in the end.

Metonymic Context, Figurative-Selecting VerbBefore Cubism became popular, no one cared when Cubist artwork was put on display. Picasso / was exhibited / by an obscure curator, / but nothing came of it / in the end.

Neutral Context, Literal-Selecting VerbThere was a culture of close-mindedness around the turn of the 20th century in Paris. Picasso / was approached / by an obscure curator, / but nothing came of it / in the end.

Neutral Context, Figurative-Selecting VerbThere was a culture of close-mindedness around the turn of the 20th century in Paris. Picasso / was exhibited / by an obscure curator, / but nothing came of it / in the end.

5 Metonymic Context, Literal-Selecting Verb Recently, the New York Public Library held a public event at which famous poems were read. Maya Angelou / was sighted / at the event, / to no one's / great surprise.

Metonymic Context, Figurative-Selecting VerbRecently, the New York Public Library held a public event at which famous poems were read. Maya Angelou / was recited / at the event, / to no one's / great surprise.

Neutral Context, Literal-Selecting VerbA prominent poet's funeral was held just a few days ago.

Lexical Access of Polysemes 43

Page 44: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Maya Angelou / was sighted / at the event, / to no one's / great surprise.

Neutral Context, Figurative-Selecting VerbA prominent poet's funeral was held just a few days ago. Maya Angelou / was recited / at the event, / to no one's / great surprise.

6 Metonymic Context, Literal-Selecting Verb In 1879, the invention of a more reliable printing press allowed books to be produced faster. Oscar Wilde / was enlightened / by a London publisher / when this / exciting event occurred.

Metonymic Context, Figurative-Selecting VerbIn 1879, the invention of a more reliable printing press allowed books to be produced faster. Oscar Wilde / was circulated / by a London publisher / when this / exciting event occurred.

Neutral Context, Literal-Selecting VerbIn 1879, the British Prime Minister declared a year-long celebration of the arts in London. Oscar Wilde / was enlightened / by a London publisher / when this / exciting event occurred.

Neutral Context, Figurative-Selecting VerbIn 1879, the British Prime Minister declared a year-long celebration of the arts in London. Oscar Wilde / was circulated / by a London publisher / when this / exciting event occurred.

7 Metonymic Context, Literal-Selecting Verb Around 1900, Paris were flooded with counterfeit copies of important works of art. Monet / was saddened / by the forgers, / but the police / eventually caught them.

Metonymic Context, Figurative-Selecting VerbAround 1900, Paris were flooded with counterfeit copies of important works of art. Monet / was reproduced / by the forgers, / but the police / eventually caught them.

Neutral Context, Literal-Selecting VerbAround 1900, Paris was teeming with counterfeiters. Monet / was saddened / by the forgers, / but the police / eventually caught them.

Neutral Context, Figurative-Selecting VerbAround 1900, Paris was teeming with counterfeiters. Monet / was reproduced / by the forgers, / but the police / eventually caught them.

8 Metonymic Context, Literal-Selecting Verb In the 18th century, the Berlin Orchestra rehearsed well-known pieces before concerts.

Lexical Access of Polysemes 44

Page 45: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Beethoven / was glimpsed / at these sessions / by the violinists / and the cello players.

Metonymic Context, Figurative-Selecting VerbIn the 18th century, the Berlin Orchestra rehearsed well-known pieces before concerts. Beethoven / was practiced / at these sessions / by the violinists / and the cello players.

Neutral Context, Literal-Selecting VerbIn the 18th century, the Berlin Orchestra often held open rehearsals. Beethoven / was glimpsed / at these sessions / by the violinists / and the cello players.

Neutral Context, Figurative-Selecting VerbIn the 18th century, the Berlin Orchestra often held open rehearsals. Beethoven / was practiced / at these sessions / by the violinists / and the cello players.

9 Metonymic Context, Literal-Selecting Verb It was around 1950 that art collectors in New York became interested in truly avant garde works. Jackson Pollack / was solicited / by a gallery, / and his works / became famous.

Metonymic Context, Figurative-Selecting VerbIt was around 1950 that art collectors in New York became interested in truly avant garde works. Jackson Pollack / was displayed / by a gallery, / and his works / became famous.

Neutral Context, Literal-Selecting VerbCulture trends in New York during the 1950's were subject to overnight changes in direction. Jackson Pollack / was solicited / by a gallery, / and his works / became famous.

Neutral Context, Figurative-Selecting VerbCulture trends in New York during the 1950's were subject to overnight changes in direction. Jackson Pollack / was displayed / by a gallery, / and his works / became famous.

10 Metonymic Context, Literal-Selecting Verb A controversy at a Hollywood movie studio caused many movies to be released later than planned. Spielberg / was telephoned / by the studio / soon after / the scandal was over.

Metonymic Context, Figurative-Selecting VerbA controversy at a Hollywood movie studio caused many movies to be released later than planned. Spielberg / was distributed / by the studio / soon after / the scandal was over.

Neutral Context, Literal-Selecting Verb

Lexical Access of Polysemes 45

Page 46: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

A recent controversy at a movie production studio caused many problems in Hollywood. Spielberg / was telephoned / by the studio / soon after / the scandal was over.

Neutral Context, Figurative-Selecting VerbA recent controversy at a movie production studio caused many problems in Hollywood. Spielberg / was distributed / by the studio / soon after / the scandal was over.

11 Metonymic Context, Literal-Selecting Verb A few years ago, a bookstore decided to release updated editions of classic American novels. Vonnegut / was notified / after the announcement, / which did not / surprise anyone.

Metonymic Context, Figurative-Selecting VerbA few years ago, a bookstore decided to release updated editions of classic American novels. Vonnegut / was annotated / after the announcement, / which did not / surprise anyone.

Neutral Context, Literal-Selecting VerbBill Clinton proposed a new initiative to make American culture more accessible to foreigners. Vonnegut / was notified / after the announcement, / which did not / surprise anyone.

Neutral Context, Figurative-Selecting VerbBill Clinton proposed a new initiative to make American culture more accessible to foreigners. Vonnegut / was annotated / after the announcement, / which did not / surprise anyone.

12 Metonymic Context, Literal-Selecting Verb In Renaissance Era France, handwritten editions of books were illegally copied and sold. Voltaire / was duped / by a petty criminal / who was never / brought to justice.

Metonymic Context, Figurative-Selecting VerbIn Renaissance Era France, handwritten editions of books were illegally copied and sold. Voltaire / was forged / by a petty criminal / who was never / brought to justice.

Neutral Context, Literal-Selecting VerbIn Renaissance Era France, illegal activities often went completely unpunished. Voltaire / was duped / by a petty criminal / who was never / brought to justice.

Neutral Context, Figurative-Selecting VerbIn Renaissance Era France, illegal activities often went completely unpunished. Voltaire / was forged / by a petty criminal / who was never / brought to justice.

13 Metonymic Context, Literal-Selecting Verb

Lexical Access of Polysemes 46

Page 47: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

In the early 20th century, people in academia took great interest in contemporary literature. Fitzgerald / was phoned / by a Harvard professor, / whose research / was on novels.

Metonymic Context, Figurative-Selecting VerbIn the early 20th century, people in academia took great interest in contemporary literature. Fitzgerald / was edited / by a Harvard professor, / whose research / was on novels.

Neutral Context, Literal-Selecting VerbPeople in academia during the first part of the 20th century played a part in creating pop culture. Fitzgerald / was phoned / by a Harvard professor, / whose research / was on novels.

Neutral Context, Figurative-Selecting VerbPeople in academia during the first part of the 20th century played a part in creating pop culture. Fitzgerald / was edited / by a Harvard professor, / whose research / was on novels.

14 Metonymic Context, Literal-Selecting Verb Many Russian literary critics in the 1900's called for literature that was shorter in length. Tolstoy / was infuriated / after these sentiments / first appeared / in a Moscow newspaper.

Metonymic Context, Figurative-Selecting VerbMany Russian literary critics in the 1900's called for literature that was shorter in length. Tolstoy / was abridged / after these sentiments / were made public / in a Moscow newspaper.

Neutral Context, Literal-Selecting VerbAround 1900, Russian politicians called for a sweeping reform of the education system. Tolstoy / was infuriated / after these sentiments / first appeared / in a Moscow newspaper.

Neutral Context, Figurative-Selecting VerbAround 1900, Russian politicians called for a sweeping reform of the education system. Tolstoy / was abridged / after these sentiments / were made public / in a Moscow newspaper.

15 Metonymic Context, Literal-Selecting Verb At the American Literature convention, many attendees were busy reading some book or another. Philip Roth / was greeted / by dozens of people / as they waited / for the event to start.

Metonymic Context, Figurative-Selecting VerbAt the American Literature convention, many attendees were busy reading some book or another. Philip Roth / was skimmed / by dozens of people / as they waited / for the event to start.

Neutral Context, Literal-Selecting VerbRecently, the New York Public Library hosted its annual fundraiser.

Lexical Access of Polysemes 47

Page 48: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Philip Roth / was greeted / by dozens of people / as they waited / for the event to start.

Neutral Context, Figurative-Selecting VerbRecently, the New York Public Library hosted its annual fundraiser. Philip Roth / was skimmed / by dozens of people / as they waited / for the event to start.

16 Metonymic Context, Literal-Selecting Verb In the middle of the 20th century, there was a thriving industry based on books. Hemingway / was thanked / by a gracious editor, / who appreciated / the business.

Metonymic Context, Figurative-Selecting VerbIn the middle of the 20th century, there was a thriving industry based on books. Hemingway / was proofread / by a gracious editor, / who appreciated / the business.

Neutral Context, Literal-Selecting VerbIn the 1940's, the publishing industry was known for being very humble. Hemingway / was thanked / by a gracious editor, / who appreciated / the business.

Neutral Context, Figurative-Selecting VerbIn the 1940's, the publishing industry was known for being very humble. Hemingway / was proofread / by a gracious editor, / who appreciated / the business.

17 Metonymic Context, Literal-Selecting Verb Recently, American soldiers overseas were unhappy with the musical entertainment available to them. Kanye West / was interviewed / on the radio, / and troop morale / briefly improved.

Metonymic Context, Figurative-Selecting VerbRecently, American soldiers overseas were unhappy with the musical entertainment available to them. Kanye West / was broadcast / on the radio, / and troop morale / briefly improved.

Neutral Context, Literal-Selecting VerbRecently, American soldiers overseas were unhappy because they were often bored. Kanye West / was interviewed / on the radio, / and troop morale / briefly improved.

Neutral Context, Figurative-Selecting VerbRecently, American soldiers overseas were unhappy because they were often bored. Kanye West / was broadcast / on the radio, / and troop morale / briefly improved.

18

Lexical Access of Polysemes 48

Page 49: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Metonymic Context, Literal-Selecting Verb In 1903, many books were banned from schools because they discussed controversial topics. Twain / was angered / when these decisions / were announced / to the public.

Metonymic Context, Figurative-Selecting VerbIn 1903, many books were banned from schools because they discussed controversial topics. Twain / was rewritten / when these decisions / were announced / to the public.

Neutral Context, Literal-Selecting VerbAround the turn of the century, harsh censorship laws were put into effect. Twain / was angered / when these decisions / were announced / to the public.

Neutral Context, Figurative-Selecting VerbAround the turn of the century, harsh censorship laws were put into effect. Twain / was rewritten / when these decisions / were announced / to the public.

19 Metonymic Context, Literal-Selecting Verb During the 1830's, the Warsaw National Theater was constantly showcasing contemporary music. Chopin / was annoyed / by the music scholar / who determined / the schedule.

Metonymic Context, Figurative-Selecting VerbDuring the 1830's, the Warsaw National Theater was constantly showcasing contemporary music. Chopin / was arranged / by the music scholar / who determined / the schedule.

Neutral Context, Literal-Selecting VerbDuring the 1830's, the National Theater was Warsaw's cultural center. Chopin / was annoyed / by the music scholar / who determined / the schedule.

Neutral Context, Figurative-Selecting VerbDuring the 1830's, the National Theater was Warsaw's cultural center. Chopin / was arranged / by the music scholar / who determined / the schedule.

20 Metonymic Context, Literal-Selecting Verb In the 1820's, the people of Vienna greatly enjoyed vocal and choral music. Schubert / was cheered / by happy people / as they walked / around the city.

Metonymic Context, Figurative-Selecting VerbIn the 1820's, the people of Vienna greatly enjoyed vocal and choral music. Schubert / was hummed / by happy people / as they walked / around the city.

Lexical Access of Polysemes 49

Page 50: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Neutral Context, Literal-Selecting VerbDuring the 1820's, the people of Vienna were very proud of their city's culture. Schubert / was cheered / by happy people / as they walked / around the city.

Neutral Context, Figurative-Selecting VerbDuring the 1820's, the people of Vienna were very proud of their city's culture. Schubert / was hummed / by happy people / as they walked / around the city.

21 Metonymic Context, Literal-Selecting Verb The Budapest orchestral festival in 1845 was where some strange things happened in the music world. Liszt / was intrigued / by the orchestra conductor, / who was known / for being eccentric.

Metonymic Context, Figurative-Selecting VerbThe Budapest orchestral festival in 1845 was where some strange things happened in the music world. Liszt / was whistled / by the orchestra conductor, / who was known / for being eccentric.

Neutral Context, Literal-Selecting VerbStrange things happened at the Winter Ball in Budapest in 1845. Liszt / was intrigued / by the orchestra conductor, / who was known / for being eccentric.

Neutral Context, Figurative-Selecting VerbStrange things happened at the Winter Ball in Budapest in 1845. Liszt / was whistled / by the orchestra conductor, / who was known / for being eccentric.

22 Metonymic Context, Literal-Selecting Verb In 2010, production studios worked to acquire rights to popular books and make movies out of them. John Grisham / was persuaded / by a movie producer / who wanted / to make a film.

Metonymic Context, Figurative-Selecting VerbIn 2010, production studios worked to acquire rights to popular books and make movies out of them. John Grisham / was adapted / by a movie producer / who wanted / to make a film.

Neutral Context, Literal-Selecting VerbIn 2010, production studios bought the rights to media franchises and made movies out of them. John Grisham / was persuaded / by a movie producer / who wanted / to make a film.

Lexical Access of Polysemes 50

Page 51: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

Neutral Context, Figurative-Selecting VerbIn 2010, production studios bought the rights to media franchises and made movies out of them. John Grisham / was adapted / by a movie producer / who wanted / to make a film.

23 Metonymic Context, Literal-Selecting Verb In 1976, the New York Philharmonic held a concert where famous American musical pieces were played. Copland / was impressed / by the ensemble, / who were excited / about the opportunity.

Metonymic Context, Figurative-Selecting VerbIn 1976, the New York Philharmonic held a concert where famous American musical pieces were played. Copland / was performed / by the ensemble, / who were excited / about the opportunity.

Neutral Context, Literal-Selecting VerbOur college's orchestra was invited last year to play at a large concert hall in Los Angeles. Copland / was impressed / by the ensemble, / who were excited / about the opportunity.

Neutral Context, Figurative-Selecting VerbOur college's orchestra was invited last year to play at a large concert hall in Los Angeles. Copland / was performed / by the ensemble, / who were excited / about the opportunity.

24 Metonymic Context, Literal-Selecting Verb Over the past 15 years, illegal sharing of music has been a huge problem. Billy Joel / was enraged / by the internet hackers / who seemed / to have / no morals.

Metonymic Context, Figurative-Selecting VerbOver the past 15 years, illegal sharing of music has been a huge problem. Billy Joel / was pirated / by the internet hackers / who seemed / to have / no morals.

Neutral Context, Literal-Selecting VerbOver the past 15 years, illegal file sharing on the internet has been a huge problem. Billy Joel / was enraged / by the internet hackers / who seemed / to have / no morals.

Neutral Context, Figurative-Selecting VerbOver the past 15 years, illegal file sharing on the internet has been a huge problem. Billy Joel / was pirated / by the internet hackers / who seemed / to have / no morals.

25Metonymic Context, Literal-Selecting Verb

Lexical Access of Polysemes 51

Page 52: The Integration of Contextual Information in the Lexical ...linguistics.pomona.edu/files/Fishbein_thesis12.pdfcontext of talking about stock markets, a processor will know to interpret

In 1907, a huge mistake at an Irish printing press caused thousands of copies of famous books to be incorrectly formatted.Joyce was cautioned before too long, but the damage had already been done.

Metonymic Context, Figurative-Selecting VerbIn 1907, a huge mistake at an Irish printing press caused thousands of copies of famous books to be incorrectly formatted.Joyce was reprinted before too long, but the damage had already been done.

Neutral Context, Literal-Selecting VerbIn 1907, a fire in Dublin destroyed many homes and factories.Joyce was cautioned before too long, but the damage had already been done.

Neutral Context, Figurative-Selecting VerbIn 1907, a fire in Dublin destroyed many homes and factories.Joyce was reprinted before too long, but the damage had already been done.

26Metonymic Context, Literal-Selecting VerbIn the 1950’s, book sales in Great Britain plummeted because people felt that novels were too long to read.Orwell was reassured by a gifted editor who knew the industry well.

Metonymic Context, Figurative-Selecting VerbIn the 1950’s, book sales in Great Britain plummeted because people felt that novels were too long to read.Orwell was condensed by a gifted editor who knew the industry well.

Neutral Context, Literal-Selecting VerbIn the 1950’s, there was a great deal of criticism expressed towards intellectualism in Great Britain.Orwell was reassured by a gifted editor who knew the industry well.

Neutral Context, Figurative-Selecting VerbIn the 1950’s, there was a great deal of criticism expressed towards intellectualism in Great Britain.Orwell was condensed by a gifted editor who knew the industry well.

Lexical Access of Polysemes 52