Idioms

INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may be
from any type of computer printer.
The quality of this reproduction is dependent upon the quality o f the

copy submitted. Broken or indistinct print, colored or poor quality
illustrations and photographs, print bleedthrough, substandard margins,
and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted. Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by

sectioning the original, beginning at the upper left-hand com er and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in reduced
form at the back of the book.
Photographs included in the original manuscript have been reproduced

xerographically in this copy. Higher quality 6” x 9” black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UM I directly to
order.
UMI
A Bell & Howell Information Company
300 North Zeeb Road, Ann Arbor MI 48106-1346 USA
313/761-4700 800/521-0600
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A Cross-Linguistic Analysis of Idiom Comprehension
by Native and Non-Native Speakers
A Dissertation Presented
by
Heather Bortfeld
to
The Graduate School
in Partial Fulfillment of the
Requirements
for the Degree of
Doctor in Philosophy
in
Experimental Psychology
State University of New York
at Stony Brook
December 1997
UMI Number: 9822946
UMI Microform 9822946

Copyright 1998, by UMI Company. All rights reserved.
This microform edition is protected against unauthorized

copying under Title 17, United States Code.
UMI
300 North Zeeb Road
Ann Arbor, MI 48103
at Stony Brook
The Graduate School
Heather Bortfeld
We, the dissertation committee
for the above candidate for the
Doctor of Philosophy degree,
hereby recommended acceptance of this dissertation.
Susan BrennjiRfAssociate Professor, Department of Psychology
Arthur Sarptiel,
Samuel, Professor, Department of Psychology
Grover Whitehurst, ProfesiorJDepartment of Psychology
Richard Gerrig, Associate Professor, Department of Psychology
Eva Fedeplattay, Professor, Department of Philosophy

Stale University of New York at Stony Brook
/
This dissertation is accepted by the Graduate^cjiopl */ J,
ii
Abstract of the Dissertation
A Cross-Linguistic Analysis of Idiom Comprehension by Native and Non-Native
Speakers
by
Heather Bortfeld
in
Experimental Psychology
at Stony Brook
1997
This dissertation is a cross-linguistic analysis of idiom comprehension that
addresses two major points: 1) the manner in which people comprehend figurative
speech in their own language (in this case, English), and 2) the degree to which
people are able to apply conceptual knowledge from their first language (in this
case, English, Latvian, or Mandarin) to make sense of figurative speech used in

another (second or foreign) language.
Two series of studies illustrate how conceptual metaphors guide our
understanding (and creation) of idiomatic language: The first examines native and
non-native English speakers’ mental images for idioms. Even when explicitly
instructed to base their images on the phrases’ literal meanings, both native and
non-native English speakers (in this case, native Latvian speakers) formed
significantly more images that reflected both figurative and literal aspects of a
phrase than simply the literal meaning of the phrases’ words. That is, these images
iii
reflected the conceptual metaphors that guide people’s understanding of idioms’
figurative meanings. Furthermore, the same non-native speakers used idioms more
naturally following participation in the imaging task, indicating that they had
effectively mapped the phrases’ metaphorical underpinnings to their surface
structures.
Data from another imaging study show that non-native English speakers (in
this case, nadve Mandarin speakers) produced similar images based on idioms’
literal meanings both before and after being made aware of the phrases’ idiomatic
meanings. Just as native speakers do, the non-native speakers produced hybrid
images reflecting both the literal and figurative meanings of those phrase. These
results suggest that non-native speakers are able to infer the figurative meanings of
idiomatic phrases based on the phrases’ surface (or literal) structures.
Together, findings from the mental imaging studies support the claim (e.g.,
Kovecses, 1986; Lakoff, 1987; Gibbs, 1994) that the literal contents of most
idioms reflect universally-held conceptual metaphors that help us understand the
phrases’ figurative meanings (e.g., for the idiomatic phrase hit the ceiling, meaning
‘extreme and sudden anger,’ a speaker might respond to the request for a literal
image with “a man with flames shooung out of his feet rockets up to the ceiling and
hits his head,” apparently accessing the conceptual metaphor ANGER IS HEATED
FLUID IN A CONTAINER).
These results motivated a second series of experiments further examining

the degree to which native English speakers can analyze different types of idioms
from diverse languages (English, Latvian, and Mandarin) based on the phrases’
literal meanings. Idioms used in these categorization studies were selected based on
iv
initial rating studies conducted with native speakers of the three different languages.
Native English speakers then categorized idioms from the three languages according
to their figurative meanings (anger, control, secrecy, insanity, and revelation).
Response times and error rates indicate that speakers are able to identify the
figurative meanings of another language’s idiomatic phrases depending on how
analyzable (transparent), abnormally analyzable (metaphorically-based), or
unanalyzable (culturally- or historically based) they are.

Specifically, findings from the categorization studies show that native
English speakers categorize normally and abnormally analyzable idioms from their
native language equally quickly, while they take longer to categorize other
languages’ abnormally analyzable idioms than those that are normally analyzable.
This suggests that, although speakers are able to make the metaphorical mappings
necessary to categorize another languages’ metaphorical idioms correctly, they must

compute meaning on-line by accessing the idioms’ underlying conceptual
metaphors and that this takes measurable time.
On the other hand, speakers are able to correctly categorize fewer
unanalyzable idioms from another than from their own language, suggesting that
these idioms require particular cultural and/or historical knowledge to be
understood. Furthermore, unanalyzable idioms take the longest time to categorize

of the three idiom types, regardless of whether they are from one’s native or
another language. This highlights the more tenuous connections the surface
structures of these phrases have to any readily accessible conceptual metaphors.
Taken together, these results support the view that, when first exposed to a
language’s analyzable idiomatic phrases, adult non-native speakers are able to

understand those idioms’ figurative meanings by accessing relevant conceptual
metaphors that have been pre-established in their native language and map them to
the new phrases’ surface forms. Comprehension of abnormally-analyzable idioms
from one’s own language appears to have become automatic, such that these
lexicalizations of conceptual metaphors no longer require analysis. At the very
least, native speakers do not need to make the mappings between content of familiar
idioms and their metaphorical underpinnings. These mappings already exist.
This is not to say that adult native speakers no longer use conceptual
metaphors to understand language. To the contrary, each time we encounter new
applications of these old concepts (e.g., in recently coined colloquialisms, in slang,
in film and in literature), we must go through the process of accessing the
conceptual metaphors upon which the references are based in order to make sense
of them. The findings reported here highlight the degree to which we as language
comprehenders are able to flexibly accommodate old and new, familiar and
unfamiliar language through relying more or less on the interconnected bodies of
metaphorical meaning we have been developing our entire lives.
Table of Contents
List o f Tables.......................................................................x
A cknow ledgm ents..............................................................xiii

I. In tro d u ctio n ............................................................................ 1
Why study idioms, in particular?..................................... 3
Idiom processing................................................................ 5
C urrent hypothesis............................................................. 8
II. E xperim ent 1...................................................................... 13
M ethod........................................................................17
C o d in g ...................................................................... 22
R e su lts......................................................................27
D iscussion............................................................... 30
III. E xperim ent 2......................................................................32
M ethod......................................................................32
R e su lts...................................................................... 34
D iscussion................................................................35
IV. E xperim ent 3......................................................................36
M ethod......................................................................38
C o d in g ...................................................................... 47
R esu lts...................................................................... 48
D iscussion................................................................51
V. E xperim ent 4......................................................................54
Experiment 4A ...................................................... 57
M ethod...........................................................60
Results and Discussion................................62
Experim ent......4B .................................................... 63
M eth o d ...........................................................66
Results and Discussion................................67

Experim ent 4C .................................................... 69
M eth o d ...........................................................69
Results and Discussion.............................. 73
VI. Experim ent 5.........................................................................76

Norming Study 5A................................................76
M ethod...........................................................77
R e su lts...........................................................79
Experim ent 5B .................................................... 79
M eth o d ...........................................................79
R e su lts...........................................................81
D iscu ssio n .....................................................83
VII. General D iscussion..............................................................85

VIII. R eferences...............................................................................93
IX. T ab les...................................................................................... 98
X. A ppendices..............................................................................110
Appendix A........................................................... 110
Appendix B ........................................................... 112
Appendix C ........................................................... 122

Appendix D ........................................................... 133
v iii
A ppendix E ............................................................. 136
Appendix F ..............................................................137
Appendix G ............................................................. 140

Appendix H ............................................................. 141
Appendix 1...............................................................144
Appendix J ...............................................................145
Appendix K ..............................................................150
ix
List of Tables
Table 1. Idioms originally selected by Gibbs and O’Brien (1990) from Boatner,
Gates, and Makkai (1975) and Long and Summers (1979) and relevant to
Experiments 1,2 and 3. All idioms are either analyzable or abnormally analyzable.
Table 2. Three different native speakers’ responses to questions regarding their

image for the phrase go o ff your rocker (Experiment 1).
Table 3. Schemas derived from native English speakers’ most frequent responses
for English idioms (based on format followed by Gibbs and O’Brien, 1990) and
proportions of responses that matched them (Experiment 1).
Table 4. Type of images speakers reported during first and second mental imaging
sessions of Experiment 3, including images for phrases that were defined correctly
and incorrectly on pre-test.

Table 5. Proportions of responses given by native English speakers and native
Mandarin speakers for English idioms (during Mental Imaging Sessions I and II in
Experiment 3) that matched the most frequently reported image schemas described
by native English speakers (established in Experiment 1). Proportions are

presented by concept and are based on hybrid images only.
Table 6. Responses speakers gave for idioms’ figurative meanings during pre-and
post-tests (prior to and following Mental Imaging Session I of Experiment 3).

Table 7. Native English speakers’ response times for correct conceptual
categorization of three idiom types for the three languages tested in Experiments
4A, 4B, and 4C. Overall percentage correct is listed below the corresponding
response time.
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
Table 8. Native English speakers’ response times for correct conceptual
categorization of the three idiom types from English and Latvian (in Experiment 5).
Overall percentage correct is listed below the corresponding response time.
“Those involved in the making of things tend to have a favorite
metaphor with which to “image forth” the mysteries of
creativity. For me it is a memory from the school chemistry
laboratory of crystals growing in solution, the delicate blue

flakes suspended by threads and growing new facets until the
original is so transformed as to be no longer recognizable.”

—Andrew Miller (New York Times, October 12, 1997)
Acknowledgments
This study would not have been possible without the support of many
people. In particular, I am indebted to Susan Brennan for her unflagging
enthusiasm in pursuing answers to language production and processing questions.
Susan’s infectious excitement about research has motivated me from the very
beginning of my graduate career. I hope that I can maintain even half of her energy
level and excitement about the learning process.
Without Arthur Samuel, I would have limited my questions to higher level
processing issues. Arty’s attention to detail and profound ability to understand and
explain perceptual subtleties contributed fundamentally to my appreciation for the
use and understanding of language at all levels of the food chain.

I also thank Richard Gerrig for his insights and enthusiasm about my work.
Early on, Richard generously invited me to join him on a project that was consistent
with my own interests. This gesture has helped shape my current research
program. Richard’s presence on my dissertation committee provided an expert
voice from the field with the added wit of a comedian.

Thanks are also due to Russ Whitehurst and Eva Kittay, both of whom
served on my dissertation committee. These thoughtful scholars contributed
substantially to the methodological and theoretical bases of the present studies. I
could not have asked for a committee that worked better as a unit; I only hope that
the five members enjoyed their interaction as much as I learned from it.
Of course, I am indebted to all the people who participated in my studies,
both native and non-native speakers alike. Without their active imaginations (and
saint-like patience), the data that all of my arguments are based on would not be
available.
I am also indebted to several research assistants who eased my work load
by translating and/or transcribing hours of mental image interviews and
questionnaires. In particular, I would like to thank Rachael Hoina for her
impressive consistency and patience. Rachael helped me through several years’

worth of methodological tweaks and personal quirks. I wish her luck in her own
graduate studies, where I am certain she will find success.
Finally, I would like to acknowledge the degree to which Michael Ahrens

provided mental and moral guidance during the final year of this project His
Ubermensch qualities never did (and never will) cease to amaze me. Michael will
always serve as a model when I need to get my own house in order.
1
Introduction
Researchers interested in language have long tried to explain how it is that

we understand not only literal language (that is, a phrase meaning what is explicitly
stated), but also figurative language (that is, a phrase meaning something different
from what is explicitly stated). In doing so, many of the same researchers have
become involved in a debate about the nature of idiom comprehension. Metaphor is
considered the basis for various forms of figurative language, including idioms.
Numerous studies (e.g., Bobrow & Bell, 1973; Gibbs, 1980. 1986; Glucksberg,
Gildea, & Bookin, 1982; Swinney & Cutler, 1979) have examined whether idiom
processing, which is a specific form of metaphor interpretation, relies on special
mechanisms that are distinct from the processing used to comprehend literal
language.
In the course of trying to determine how we understand idioms in particular

and metaphor in general, the debate has included claims that this understanding is
based on universally-held conceptual metaphors (e.g., Gibbs, 1994) that guide our
use of all forms of language. At issue is whether creative language is a product of
certain cognitive processes used in everyday language (Gibbs, 1994; Glucksberg &
Keysar, 1990), or whether there is at least some special processing involved (e.g.,
Kittay, 1996). I will focus here on how this debate ignores the flexibility of our
language production and comprehension. That is, the various positions each reflect
one aspect of what we do every minute during which we produce or process
language.
Three possibilities are: 1) that pre-existing universal conceptual metaphors
guide people's comprehension of figurative language (e.g., Gibbs, 1994), 2) that
conceptual schemas are, instead, constructed on-line and in this way made available
during metaphor comprehension (Glucksberg, Brown, & McGlone, 1993), and 3)
that more individualistic, culture-based strategies underlie the comprehension and
production of figurative language (e.g., Kittay, 1996). I will argue that all of these
processes support our daily use of language, and to varying degrees.
The arguments used to justify any one of these positions have been based
largely on data collected from native speakers of English. However, any attempt at
resolution of this debate will require, at the very least, a preliminary comparative
analysis of idiom comprehension and processing across languages. The following
studies find that each of these viewpoints holds under certain circumstances (in this
case, under conditions created using a number of methodologies with several
languages). More important, these studies address the overriding question
regarding the possibility that the conceptual metaphors upon which figurative
language is based are universal.
First, I will briefly review some of the idiom processing research
conducted over the past several years. I will then present a proposal for testing
whether universal conceptual metaphors guide the comprehension of figurative
language. Based on my review of the idiom processing literature, I will argue that
no resolution of the previously outlined debate can be reached without cross-

linguistic experimentation that addresses two major topics: 1) the manner in which
people comprehend idiomatic expressions in their own language, and 2) the degree
3
to which people are able to apply conceptual knowledge from their first language to
make sense of figurative expressions used in another (second) language.
Why study idioms, in particular?
Psycholinguisdc analysis of idiomatic language can provide information

about how such language reflects underlying metaphorical thought (e.g., Gibbs,
Bogdanovich, Sykes, & Barr, 1997). One long-standing question is whether
idioms' literal meanings can account systematically for their associated non-literal
meanings. Early models of sentence comprehension had difficulty accounting for
idiomaticity precisely because those models were based on only the most literal
language. Since then, several researchers (e.g., Gibbs, 1994; Glucksberg, Brown,
& McGlone, 1993; Lakoff, 1987) have examined idiomatic expressions relatively
extensively, with a divergence of opinion on what these structures tell us about
figurative language use and comprehension.
Idioms are also remarkably frequent in everyday language use. Entire
dictionaries have been written about them. For example, Boatner, Gates, and
Makkai’s original (1975) Dictionary o f American Idioms presented over 4000

expressions occurring in English alone. The latest edition of this same dictionary
(Makkai, Boatner, & Gates, 1995) contains over 8000 English language idioms.
Increasingly, language scholars are acknowledging the wide range of idiomatic

phrases that people use daily. Clearly, idioms are abundant in spoken language
(any spoken language) and must be accounted for by any theory of language
comprehension. Since a variety of idioms are used frequently within any one
4
language and because idioms differ across languages, they pose a particularly
difficult problem for people learning a second language.
5
Idiom processing
Where early psycholinguistic models of sentence comprehension had

difficulty accounting for idiomaticity, several more recent models focused
specifically on whether idioms’ literal meanings can systematically map onto their
associated non-literal meanings. The debate regarding the nature of idiom
representation, storage, and access has led to specific disagreements about the
nature of metaphorical thought and its representation.

Bobrow and Bell’s (1973) early study presented literal and figurative
meaning as categorically distinct, such that people went into one of two “modes” of
processing (figurative or literal) depending on whether they were processing
figurative or literal language. Swinney and Cutler’s (1979) subsequent study
characterized the storage and retrieval of idioms as similar to that of any other
“word.” This view held that no unique processing mode was entered in the
comprehension of idioms; rather a parallel computation of both meanings (literal
and figurative) was initiated upon presentation of the first word in the idiom string,
resulting in a “race” by the system for meaning access. According to this view,
since an idiom works as a unit, where a literal phrase does not, idioms tended to
win the race (demonstrated by faster response times in a phrase classification task).
Indeed, Swinney and Cutler’s (1979) is one of several studies that have shown the
people process idioms more quickly than they do corresponding literal phrases
(Gibbs, 1980, 1986; Gibbs & Gonzales, 1985; Gibbs, Nayak, & Cutting, 1989;
Ortony, Schallert, Reynold, & Antos, 1978).
Gibbs (e.g., 1985, 1990, 1992, 1993, 1994) has focused more closely on
the analysis of (that is, the breaking down or decomposition of) idiomatic phrases
6
into smaller meaningful units. Gibbs and Nayak (1989) characterized idiomatic
phrases as belonging to one of three groups: normally decomposable (e.g., wear
out your welcome), abnormally decomposable (e.g., sitting on pins and needles),
or nondecomposable phrases (e.g., beat around the bush). There is a relatively
transparent relationship between the surface structure of normally decomposable
idioms and their figurative meanings, while that relationship is more metaphorical
for abnormally decomposable idioms. Nondecomposable idioms are those whose
surface structures have little to no relation to the figurative meanings they represent
\ I will use the term “analyzability” (rather than decomposability) in the studies
described here, since this term implies a more active role on the part of the language
processor.
Gibbs’s view is consistent with Nunberg, Wasow, and Sag’s (1994)

analysis of idiomaticity, criticizing the traditional view (e.g., Weinreich, 1969; Katz
& Postal, 1964) that idioms are noncompositional (that is, correspond to long
words). Analysis of the structure of idioms themselves motivates examination of
the relationship between metaphoric thought and idiomatic phrases’ actual content
Gibbs (1992) proposes that the meanings of idioms (except for unanalyzable
phrases—that is, phrases that are historically-based) are motivated by speakers'
implicit knowledge of the metaphors that underlie these phrases' figurative
interpretations.
Although the view that conceptual metaphors influence how people

understand figurative language in the broader sense is a difficult one to argue with,
others (e.g., Glucksberg, Brown, & McGlone, 1993) have taken issue with
Gibbs's proposal that these pre-existing concepts guide idiom comprehension as it
7
occurs on-line. Their alternative view (e.g., Glucksberg & Keysar, 1990;
Glucksberg, Keysar, & McGlone, 1992; Glucksberg, Brown, & McGlone, 1993;
Kreuz & Graesser, 1991; McGlone, 1996) is that conceptual structures are
constructed anew and made accessible in working memory during idiom

comprehension.
These views center on how idiomatic meaning is accessed; another view
focuses on its representation. For example, it has been suggested that idiom
comprehension is based on people’s ability to connect domains of meaning, or
semantic fields (e.g., Lakoff, 1987). According to this view, it is these “analogical
leaps across domains” (Kittay, 1996) that allow us to both produce and interpret
idioms. Cultural knowledge and experience play an important role in the
lexicalization of concepts in a given language. Therefore, this view allows for the
cultural differences found in different languages’ lexicalization of various
metaphorical concepts. These cultural differences are then what makes certain
idiomatic phrases appear arbitrary to cultural outsiders.
However, even members of a particular language group may find culturally-

based (or unanalyzable) phrases in their language arbitrary or opaque upon initial
reflection (e.g., as children first hearing a phrase). Yet as members of the group,
these individuals come to access the intended meanings of seemingly arbitrary
phrases automatically, such that the meanings are quickly and easily understood
upon hearing the phrases.
An examination of the process by which adult language learners use
conceptual knowledge from their native language to make sense of idioms in their
new language can help reveal the connections that we as native speakers make
automatically in our first language. In order to demonstrate the varying degree to
which we rely on conceptual metaphors for understanding figurative language, I
have conducted a cross-linguistic comparison of idiom comprehension that employs
several methodologies.
Current hypothesis
To determine whether the same conceptual metaphors underlie idiom
comprehension in diverse languages, I have conducted a cross-linguistic
examination of English, Latvian, and Mandarin idioms. An example of an
underlying conceptual metaphor would be: [anger = 'a heated substance under
pressure' (from Lakoff, 1987)], and may be expressed in English by phrases such
as blow your stack, flip your lid, and lose your cool. In Latvian, this anger
metaphor may be reflected in idioms such as splaut uguni (‘to spit fire’), aizsvilties
(‘to blaze up’), and varities no dusmam (‘to boil with anger’).
I have focused my analysis on language groups (Latvian and Mandarin)

relatively removed from the target language (English) (see Appendix A for
summaries of these languages' backgrounds) because this allows for a more
methodologically sound examination of the degree to which idioms are motivated

by universal conceptual metaphors, or alternatively, are culturally based and
therefore may seem arbitrary to those not privy to the cultural references in the
phrases.
I propose that the kernels of meaning upon which many idioms are based
are universal, and that the "stretching and warping" (Kittay, 1996) that ensues
when those concepts are lexicalized into any given language may be culturally
9
motivated. That is to say, this is not a black and white debate where idioms are
either universally based on conceptual metaphors or are unique to each cultural (and
linguistic) group such that the phrases are completely opaque to outsiders. Rather,
I will argue that there are universally-held conceptual metaphors motivating
idiomatic language in general, with further differentiation made based on whatever
cultural and linguistic subtleties are unique to the language groups into which the
concepts are lexicalized.

Specifically, I will argue that analyzable idioms are relatively transparently
based on universal metaphorical concepts (e.g., boiling mad for ANGER IS
HEATED FLUID IN A CONTAINER), that abnormally analyzable (or abnormally
decomposable) idioms reflect lexicalization of the same universal concepts, flavored
a bit more heavily by the culture in which they are used (e.g., the same anger
metaphor as flip your lid), and that unanalyzable idioms are relatively opaque to
anyone not privy to the cultural references embodied in the language (e.g., to get
one’s goat). The following experiments demonstrate that, just as idioms can be
viewed as located along a continuum of analyzability, our reliance on conceptual
metaphors for idiom comprehension runs along a continuum that reflects the depth
of understanding required by the situation. That is, where simply recognizing a
familiar idiom’s figurative meaning won’t necessarily require that we access the
phrase’s underlying metaphor, forming a mental image for that phrase is more
likely to motivate us to access the phrase’s metaphorical underpinnings. Most
important, based on data collected across two series of studies, I will argue that
these underlying conceptual metaphors are universal in nature.
10
Experiment 1 compares mental images based on 25 American English

idioms described by native English and native Latvian speakers learning English as
a foreign language. I examined the degree to which both sets of images reflect the
phrases’ underlying conceptual metaphors. Experiment 2 tests whether the same
non-native speakers use the 25 test idioms from Experiment 1 more idiomatically
(that is, more naturally) following completion of the mental imaging task than they
did prior to the task. Experiment 3 is an extension of the mental imaging paradigm,
comparing the mental images based on the same 25 test idioms described by native
English and native Mandarin speakers learning English as a second language both
before and after their being explicitly made aware of the phrases’ figurative
meanings.
If native and non-native speakers’ mental images for the same idioms are
significantly different in terms of the content they reflect, this will work against the
view that these phrases are based on universally-held conceptual metaphors.
Conversely, if native and non-native speakers report images that match image
schemas (pre-established based on other native English speakers’ images) a
significant amount of the time (e.g., greater than chance), this will indicate that the
speakers are analyzing the phrases by mapping them onto the same or similar
conceptual metaphors. Since the speakers are from diverse language backgrounds
(e.g., English, Latvian, Mandarin) yet are creating mental images for the same set
of 25 (English) idioms, such a finding would support claims regarding the

universality of these conceptual metaphors. Finally, if the imaging process does
guide speakers in their mapping of a phrase’s literal content to its figurative
11
referent, non-native speakers should be able to use the phrases more idiomatically
following the imaging task than they do prior to the task.
Specifically, I hypothesize that both native and non-native speakers’ images

(both from Experiment 1 and Experiment 3) will not differ significantly and that
these images will reflect the conceptual metaphors upon which the phrases are
based. Such a finding will support the view that these conceptual metaphors are
universally-held. I hypothesize that both native and non-native speakers’ images
from the first and second imaging sessions in Experiment 3 will reflect similar
changes from the first to the second imaging session, a product of both groups of
speakers’ familiarity with the imaging task rather than of their knowledge of the
phrases’ canonical figurative meanings. Finally, I hypothesize that non-native

speakers in Experiment 2 will use the 25 test idioms from the mental imaging task
completed in Experiment 1 more naturally following completion of the task than
they do prior to the task.

The second series of experiments (Experiments 4A, 4B, 4C, and 5) is a
closer examination of whether or not and to what degree native and non-native
speakers of a language can analyze idioms from that language in order to deduce
their figurative meanings. Native English speakers categorized idioms from both
their own and two other languages (whose phrases were literally translated into
English) according to the phrases’ figurative meanings in a forced choice task. If
non-native speakers can correctly categorize another language’s idiomatic phrases
based on what they perceive to be those phrases’ idiomatic interpretations, this will
provide the strongest evidence to date that such phrases are based on universally-
held conceptual metaphors.
12
I hypothesize that non-native speakers (in this case, the native English
speakers) will be able to correctly categorize normally analyzable idioms from the
two other languages (Latvian and Mandarin) significantly more often and faster than
they will abnormally analyzable idioms from those languages. In turn, they will
correctly categorize the foreign languages’ abnormally analyzable idioms more often
and faster than they will the unanalyzable idioms. Conversely, native English
speakers should correctly categorize both normally and abnormally analyzable
idioms from their own language equally quickly because they already have
established the metaphorical mappings of the abnormally analyzable phrases. This
is consistent with comparisons of native speakers’ processing of normally and
abnormally analyzable idioms that find the same equivalence in response times
(e.g., Gibbs, 1980, 1986; Gibbs & Gonzales, 1985; Gibbs, Nayak, & Cutting,
1989; Ortony, Schallert, Reynold. & Antos, 1978; Swinney & Cutler, 1979).
Finally, native English speakers will have the most difficulty correctly categorizing
unanalyzable English idioms and will take significantly longer to do so, also
consistent with findings from previous research. The same will hold for
unanalyzable idioms directly translated from other languages.
Together, the studies reported here critically test the hypothesis that
universally-held metaphorical structures constrain the surface forms with which
underlying concepts are represented in any language.
13
Experiment 1
Gibbs and O’Brien (1990) claim that “people’s knowledge about their
mental images for idioms is constrained by different conceptual metaphors” (p. 62).
In the study upon which this claim is based, the researchers tested native speakers
using familiar idioms, demonstrating that different conceptual metaphors appeared
to motivate or constrain the speakers’ mental images for those idioms. Gibbs and
O ’Brien argued that this demonstrates how idioms’ meanings are partially motivated
by conceptual metaphors that help us map information from one (source) domain
(e.g., for the phrase spill the beans, this takes the form of knowledge about what
happens when beans are spilled) to another (target) domain (e.g., for the same
phrase: knowledge about what happens when secrets are revealed). Such a
mapping would allow for the depth of meaning that a simple phrase such as spill the
beans carries for native speakers.

Traditionally, idioms have been referred to as “dead” metaphors whose
figurative meanings are directly stipulated in the mental lexicon (e.g., Cooper,
1986; Cruse, 1986; Fraser, 1970). Certainly historical linguists and other scholars
have documented the historical basis of many idioms, helping reveal the
metaphorical references from which the phrases stemmed. But, as is often pointed
out, native speakers tend not to know this historical information, yet demonstrate a
tacit knowledge of the metaphorical basis for the idioms nonetheless.
How does this knowledge come about? As Gibbs and O’Brien (1990)
report, when asked to form a mental image for a phrase like spill the beans, people
can do so without difficulty and can answer a variety of questions about that image.
14
Although there is some variation, generally native speakers report that the container
for the beans is something like a large jar or bowl about the size of a person’s head,
that the beans were supposed to remain in the container but were accidentally
spilled, that the beans go all over the place once spilled, and that they are not easy to
get back into the container.
There is no linguistic explanation for this specific, yet tacit, knowledge

about idioms. As Lakoff (1987) initially observed, people’s understanding of their
mental images for idioms is strongly constrained (to use his term) by conceptual
mappings between source and target domains. With regard to a phrase like spill the
beans, it seems clear that the beans represent ideas (e.g., IDEAS ARE PHYSICAL
ENTITIES) that can be passed from person to person (because the MIND IS A
CONTAINER). These two metaphors are consistent with the CONDUIT

metaphor, first discussed by Reddy (1979), regarding how exchange of
information is expressed in language (e.g., taking ideas out of the mind, putting
them into words, and handing them over to others). From this we can understand
that the mind (target domain) is like a container (source domain) and that the ideas
(target domain) in it are like physical entities (source domain), such as beans.
Because of this mapping, when we hear the phrase spill the beans, we don’t
typically just think about something being told to someone else. It is more likely
that we think about a quantity of information leaving someone’s possession and

going into someone else’s (or quite simply going everywhere) in a way that will
eliminate most (if any) chances for recovery.
Of course, to go from native English speakers demonstrating consistent
images for American English idioms to claims about the universality of those
15
phrases’ underlying metaphors is a stretch, to say the least. The following

experiment attempts to develop a better understanding of whether this mapping is to
universally-held conceptual metaphors or, alternatively, culturally-based
metaphorical notions. Specifically, this work extends Gibbs and O'Brien’s (1990)
original study regarding people's mental images for American English idiomatic
phrases. However, where Gibbs and O’Brien looked at native English speakers’
images for English idioms. Experiment 1 tests these researchers’ claims on two
different populations of speakers. One group consisted of native English speakers
and the other consisted of non-native English speakers whose native language
(Latvian) is only a distant relative to English.
Experiment 1: Native and non-native English speakers’ mental images

for American English idioms
In their study relevant to the questions raised here, Gibbs and O'Brien
found that native English speakers' images for idiomatic phrases with similar
figurative meanings were highly consistent, despite large differences in their surface
forms (e.g., keep it under your hat and hold your tongue). These native speakers'
responses to follow-up questions about the causes and effects of different events
within their images were also highly consistent. The researchers concluded that
these image schemas indicate that the conventional images and knowledge
associated with idioms are products of the conceptual metaphors (e.g., THE MIND
IS A CONTAINER and IDEAS ARE ENTITIES) that motivate the phrases'
figurative meanings.
In another study using the same task on native Italian speakers, Cacciari and
Glucksberg (1995) challenged Gibbs and O’Brien’s (1990) findings by showing
16
that the majority of images participants generated based on Italian idiomatic phrases
reflected the literal actions and events denoted in an idiom’s lexical makeup rather
than its underlying conceptual metaphor. Cacciari and Glucksberg argued that these
literal interpretations of idioms demonstrate that people do not generate images that
reflect idiomatic meanings, “let alone the conceptual metaphors that presumably
underlie those meanings,” (p. 290). They go on to claim that these literal images do
not evoke or reflect underlying conceptual metaphors and, in fact, interfere with
idiom comprehension (as demonstrated by idiom verification times).
Given these contradictory interpretations of results from data collected using
virtually identical techniques1, there may be a fundamental confusion about what
can be considered literal and what can be considered figurative. Indeed, in other
writings (e.g., Gibbs, 1994) Gibbs has argued that the distinction between literal
and figurative is not at all categorical and should instead be conceptualized as falling
along a continuum. In neither of these imaging studies were the instructions
published in full, so it is impossible to tell whether participants construed the
instructions as asking for a literal or figurative image. For example, Cacciari and
Glucksberg (1995) report only that participants were asked “to form a detailed
mental image for the string and to describe it as accurately as possible” (Cacciari &
Glucksberg, 1995, p. 288). One would assume that participants would (more often
than not) ask for clarification regarding which meaning the experimenters intended
this image to be based on. If no further instruction was given, then perhaps the
1 As far as we can tell, Cacciari and Glucksberg (1996) used a procedure identical to
that used by Gibbs and O’Brien (1990). However, they selected Italian idioms for
stimuli-based on normings for familiarity and semantic transparency—since they
conducted their study in an Italian university with Italian-speaking participants.
17
difference between the two studies’ results is due to their participants construing the
instructions differently.
Figurative interpretations of familiar idioms are made just as quickly as
literal ones by native speakers (Gibbs, 1980,1986), which suggests that the
conceptual metaphors can be direcdy or automatically accessed. If native speakers
were told to produce a literal image for an idiom, they should be likely to experience
interference from the figurative meaning. The findings reported by Gibbs and
O’Brien (1990) may reflect this inability on the part of native speakers to dissociate
literal from figurative meaning for familiar idioms. Non-native speakers, on the
other hand, are unfamiliar with the idioms so they don’t automatically associate
them with underlying conceptual metaphors and might be able to produce a strictly
literal image more easily, if instructed to. But if they produce the same kind of
hybrid images that native speakers produce, then it seems likely that there is on-line
mapping occurring to an underlying conceptual metaphor. For this reason, in

Experiment 1 I instructed native and non-native speakers (recently made aware of
the figurative meanings of the phrases) to form images based on the literal meanings
of the idiomatic phrases in question.
Method
The current study aims to determine whether native and non-native English
speakers’ images for the same (English) idioms are motivated or constrained in the
same way by conceptual metaphors. Specifically, it tries to resolve Gibbs and
O'Brien’s (1990) and Cacciari and Glucksberg’s (1995) contradictory findings by
testing two different groups of speakers: Native English speakers and native
Latvian speakers learning English as a foreign language. I compared images
18
formed by these two groups of speakers based on the literal meanings of a group of
American English idiomatic phrases.
Participants.
Native speakers: Twenty-five undergraduate students of the State
University of New York at Stony Brook received research credit for their
participation in this experiment. All were native speakers of English, and when
asked none reported the ability to speak another language to a degree that would be
considered functionally bilingual.
Non-native speakers: Twenty-five undergraduate students at the University

of Latvia volunteered to participate in this experiment. They were all in either their
first or second year at the University as English Literature majors. All were native
speakers of Latvian who also spoke Russian^ to varying degrees. All reported that
they had been studying English since high school (that is, for less than five years).
Materials. Stimuli were the 25 idiomatic phrases used by Gibbs and
O'Brien (1990) in their study examining the mental images that English speakers
associate with idiomatic phrases from that language. The phrases encompassed 5
concepts (e.g., anger, revelation, secretiveness, insanity, and control), with 5
different idioms expressing each concept (e.g., anger = blow your stack, flip your
lid, lose your cool, hit the ceiling, and foam at the mouth). All idioms were
abnormally analyzable (that is, metaphorically based), as defined by Gibbs and
Nayak (1989) in their taxonomy of idiom types. These five groups of idioms are
shown in Table 1.
2 Russian is a Slavic language, relatively closely related to Latvian in that both stem
from the Balto-Slavic branch of Indo-European languages.
19
I also used the six probe questions from Gibbs and O'Brien’s (1990)
original study. As noted by Gibbs and O’Brien, these questions are meant to reveal
participants’ beliefs about various aspects of their mental images. Probe question
data were in two forms: 1) two probe questions requiring yes/no responses (e.g.,
Intentionality—"Was the action done intentionally?"; Reversibility--"Is it difficult to
reverse the action?"), and 2) four probe questions eliciting open-ended responses
(e.g., Causation—"What caused the action to happen?"; Manner—"How was the
action performed?"; Consequence—"What happens as a result of the action?";
Negative Consequence—"What would happen if the action didn't happen?"). Each
probe question was altered to match the specific mental image that participants
described for each idiom. For example, if a subject reported that her image for
blow your stack was someone with fire and smoke shooting out of the top of her
head, they would be asked "What causes the fire and smoke to shoot out of the top
of this person's head?" for the Causation question.
Procedure-
Native speakers: Briefly, the methodology Gibbs and O'Brien (1990) used
in Experiment 1 of their study is as follows: Participants were asked to form and

describe their mental images based on the meanings of 25 different idiomatic
phrases (see Table 1). The 25 idioms were presented in a different random order
for each participant. Once participants described their mental image for an idiomatic
phrase’s literal meaning, they were asked the series of detailed probe questions
(outlined above) in an invariant order (e.g.. Causation, Intentionality, Manner,

Consequence, Negative Consequence, and Reversibility). I followed this format as
well. However, I also gave participants a list of the 25 idioms prior to and
20
following the mental imaging task, asking them to come up with sentences using
each phrase idiomatically. These data are relevant to Experiment 2.
Also note that for the mental imaging task, I departed from Gibbs and
O’Brien’s (1990) original protocol by explicitly instructing participants to base their
images on the phrases’ literal meanings. In their section on procedure, Gibbs and
O’Brien do not specify whether they told participants to form images based on the
phrases’ literal or their figurative meanings^. The same lack of a literal-figurative

distinction exists in Cacciari and Glucksberg’s (1995) procedure section. Since one
of the principle goals of Experiment 1 was to test non-native speakers’ ability to
analyze the structure of normally and abnormally analyzable English idioms, and to
test whether both native and non-native speakers’ literal interpretations of the
phrases can be distinguished from the canonical, figurative interpretations of the

same phrases, it was important for the purposes of this study that participants base
their mental images on the phrases’ literal meanings. In this way, similarities
among images for idioms describing a particular concept—each with quite different
structural and semantic content-wili be due to participants accessing the idioms'

underlying conceptual metaphors rather than to their simply accessing and basing
their images on the phrases’ figurative meanings.

Given that this is what Gibbs and O’Brien wanted to measure as well, and
despite the fact that they (apparently) did not specify literal images in their
experimental procedure, it seemed logical for the current study to specify that
3 It should be noted that personal communication with Gibbs failed to reveal whether
or not more explicit instructions were given. It is, thus, inferred that participants were
expected to produce images on whichever meaning they first understood and
imagined.
21
participants form their images based on the literal meanings of the idiomatic
phrases.
Finally, where Gibbs and O'Brien wrote down participants' images and
answers to probe questions, I audiotaped the entire interview with each participant.
The audiotaped interviews were later transcribed verbatim for coding.
Once participants described their mental image for an idiom, they were
asked the six probe questions regarding that image. The 25 idioms were presented
in random order for each participant. The probe questions were asked in an
invariant order 1) Causation, 2) Intentionality, 3) Manner, 4) Consequence, 5)
Negative Consequence, and 6) Reversibility. The entire process was repeated for
each of the 25 idioms and each interview took approximately one hour to complete.
With the participants’ consent, all interviews were audiotaped to be transcribed
later.
Idioms for which participants were unable to provide a mental image were
skipped and returned to at the end of the interview. If a participant was still unable
to provide a mental image for that phrase, the phrase was skipped entirely. The
interview was conducted according to a script, and very little discussion of
participants’ images took place, aside from simple clarification questions (e.g.,
"What do you see coming out of the person’s head?" when students described
"stuff” exploding out of the top of a person's head for blow your stack). The
experiment took less than one hour to complete.
Non-native speakers: The experimental protocol for the non-native speakers
was almost identical to that followed for the native speakers. However, since the
non-native speakers needed to first be made aware of the figurative meanings of the
22
test idioms, the initial procedure included a few additional steps that were as
follows: Upon scheduling an interview with the experimenter, participants were
given a list of idioms accompanied by the phrases’ figurative paraphrases. At that
time, participants were told to become familiar with the figurative meanings of the
idiomatic phrases, as they would be asked several questions about those phrases in
two weeks’ time (that is, during the scheduled interview).
Each participant was interviewed individually early in the academic
semester, prior to any discussion o f the test idioms in class. The experimental task
was first explained. Participants were told that they would be presented with the
idiomatic phrases (that they have been given two weeks earlier) one by one, and
that they were to form a mental image based on the literal meaning of that phrase
and to describe their image in detail. Participants were asked to report the first
image that came to mind, and to answer the probe questions in as detailed and
accurate a manner as possible. Participants were told that their performance was
not in any way being used as an evaluation of their language ability; rather, they
were told that the exercise was just to get them thinking about idioms more
analytically.
Coding. The interviews were transcribed and imported into the Sequence
program, a software package designed specifically for coding sequential linguistic
or behavioral data. The image transcripts were first coded for whether each image
reflected the literal or figurative meaning of the phrase, or some hybrid of the two
meanings. Responses to probe questions for images within each of the hybrid
image types were then coded for their general characteristics.
23
A naive research assistant and I used a mental image coding system that we
developed based on the native English speakers' descriptions of their mental images
and their answers to the six corresponding probe questions. Each of us coded all
native speakers’ images and probe question responses.
First, we tallied how many of the native speakers’ descriptions of their

images together with their answers to the probe questions fell into one of three
groups: those that were based entirely on the literal interpretations of the phrase,
those based entirely on the figurative interpretations of the phrase, and those that
represented a hybrid of both the figurative and literal meanings of the phrase. In
order to do this, we first established what constituted the three types of images for
each idiom. In order to be considered entirely literal, an image had to describe only
the literal contents of the phrase, with no reference to the underlying emotion or
desire. Conversely, those images that were considered entirely figurative had to
describe only the emotion or desire embodied in the figurative meaning of the
phrase. Hybrid images had to include references to both the concrete actions
described by a literal reading of the phrase, in addition to references to an
underlying (figuratively accurate) emotion or desire that motivated those actions.

For example, when a native speaker reported that his or her image for flip
your lid was of the top of a person's head being forcefully blown up, we scored
this as a hybrid image in which "some force causes the top of a container to release
pressure violently" (Gibbs & O'Brien, 1990). Table 2 lists examples of the three
types of images for the idiom go off your rocker from the insanity concept group.
Reliability was coded with satisfactory agreement between coders (96.5%
24
agreement, Cohen's Kappa = .95). We discussed all coding discrepancies until

agreement was reached.
Collapsing across concept groups, a total of 70% of the native English
speakers’ images qualified as hybrids of phrases’ figurative and literal meanings,

28% were based entirely on a literal reading of the phrases, and 2% clearly referred
to only the phrases’ figurative meanings.

Since the majority of images for phrases represented a hybrid of the two
meanings, we coded the hybrid images for their general characteristics. In order to
do this, each coder read through all of the participants’ hybrid images and
responses to the corresponding probe questions and independently developed five
schemas intended to capture the most common answers provided for each
conceptual group. Once we had established our schemas for the images’ general
characteristics, we met and discussed them until we had agreed on a single

representative schema for each concept (Table 3 includes the most frequent general
images and probe question responses).
Once these schemas were established, the two coders independently scored
the native speakers’ image descriptions and probe question responses for all images
(including those that had been classified as entirely literal or figurative) according to
whether or not they matched the schemas we had developed based on the hybrid
images. The Sequence program allowed us to score participants’ images and
responses grouped according to concept, and with the variable labels hidden, so
that we would not be biased by such information (e.g., whether the image had been
considered literal, figurative, or a hybrid; which participant provided that image;
which idiom was being described). However, each coder did view the image
25
schema key that had been developed earlier that corresponded to the particular
concept she was coding at that time.
We scored participants’ responses to the open-ended probe questions
(Causation, Manner, Consequence, and Negative Consequence) and images as
either matching or not matching the corresponding schemas. In order to be
considered a match, a general image had to include both the literal and figurative
parts of the image schema that we had identified for that image’s concept group.
For example, if a participant reported that she saw “a lion tamer cracking a whip to
discipline a lion” for the idiom crack the whip, this was coded as a match with the
general schema for control. That is, this general image includes a concrete action
(cracking the whip) for a purpose consistent with the figurative meaning of the
phrase (that order is imposed in some way by someone else). A general image was
considered a non-match if only a literal action was included (e.g., ‘‘a man jumping
too high and hitting his head on the ceiling” for hit the ceiling) with no reference to
the underlying figurative meaning of the phrase. Conversely, if the general image
included only a reference to the figurative meaning (e.g., “to sneak around with a
secret” for behind one’s back) with no reference to the literal contents of the phrase,
this was also considered a no-match.

In order for responses to the open-ended probe questions to be coded as
matches, they had to include at least one of the several characteristics listed in the
corresponding image schema. For example, if a person answered the Causation
question for the phrase blow your stack by saying “something upset them,” this
was counted as match since the image schema for that concept’s Causation probe
includes “stress, anger, and/or frustration.” Responses were coded as a non-match
26
if they did not include at least one aspect of that probe’s listing for the
corresponding concept. For example, when a participant reported that “someone
wanted to roll a ball to someone else” for the phrase keep the bcdl rolling, this was
coded as a non-match because the Causation probe for control had to include
“exertion of authority or control over others.”
General images and probe question responses were coded independently of
one another. That is, each general image and probe question response was coded
without consideration for what responses had preceded or followed it. This
prevented coders from “reading into” or projecting a response onto a probe question
based on supplementary information available in responses to other questions for
the same idiom.
Reliability was coded with satisfactory agreement between coders (92.4%
agreement, Cohen's Kappa = .91). Again, we discussed all coding discrepancies
until agreement was reached.

After we coded the general characteristics for native speakers’ images across
participants and across idioms within each of the five conceptual groups, we had a
summary of the most frequent image characteristics for each conceptual group,
according to our native speakers.
The measures outlined above established a baseline for the native speakers’
images and probe question responses against which to compare the non-native
speakers’ images and probe question responses. We coded each of the non-native
speakers’ images based on whether or not they matched the schema derived from
the native speakers’ responses for the same conceptual group. Non-native
speakers’ responses to the probe questions were coded in a similar manner.
27
Reliability was coded with satisfactory agreement between coders (94% agreement,
Cohen's Kappa = .92). Again, we discussed all coding discrepancies until we
came to an agreement
Results
Data from native Latvian and English speakers indicate that, for the most
part, both groups had very similar schemas underlying their mental images for
idioms with similar figurative meanings. An average of 74%4 of native English
speakers’ mental images for the different English idioms collapsed across the
different idiom groups matched the image schemas developed based on hybrids of
the phrases’ literal and figurative meanings (ranging from 61% to 89% for the five
concepts), a proportion that is significantly different from chance, rl(24) = 1.93, p

< .05; r2(24) = 1.72, p < .05 (with a very conservative null hypothesis of .50 that
assumes participants could only give two possible responses even though their
mental images could potentially be unlimited in number). An average of 72% of
non-native English speakers’ mental images for the different English idioms
collapsed across the different idiom groups matched the native speakers’ most
common schemas for hybrid images (ranging from 58% to 87%), also significandy
different from chance, fl(24) = 1.73, p < .05.
Native speakers’ images most frequendy represented a hybrid of the idioms’
figurative and literal meanings, describing both the figurative sense of the phrase
and more specific literal aspects of the events within the phrase. For example,
4 Note that this includes the 70% for native English speakers’ images that qualified
as hybrids, plus an additional 4% from those images that had been considered either
completely figurative or completely literal during the first round of coding. Since
information regarding the status of an image was hidden from coders, these 4% most
likely represent borderline cases.
28
idioms such as go off your rocker and lose your marbles both figuratively mean “to
go crazy”; presented with these idioms, both native and non-native speakers
reported a complete loss of stability (most often physical) on the part of the
individuals imagined for both phrases. That English and Latvian speakers imagined
things for the English Insanity idioms that conformed to this underlying schema
roughly the same percentage of the time is compelling. There is nothing in the
lexical makeup of these phrases that should constrain participants’ images to such a
degree. Despite this, participants’ protocols varied very little in the general events
that they imagined taking place for idioms with similar figurative meanings.
These findings support and extend Gibbs and O’Brien’s (1990) findings
regarding native English speakers’ images for English idioms. That is, the native
English speakers in the present study most often produced hybrid image schemas
that were similar within conceptual groups, as did the non-native English speakers
who had learned the figurative meanings of the English idioms just weeks prior to
being interviewed about their images for those idioms. It is notable that in a few
instances the non-native speakers’ images conformed more often to the native
speakers’ most frequent hybrid images than the native speakers’ responses
themselves. This may be attributed to the fact that several of the native English
speakers understood some of the idioms (which were presented to them out of
context) as currently-in-use, slang meanings of the terms (e.g., blow your stack as
meaning ‘spend all your money’ rather than ‘to get extremely angry’ or foam at the
mouth as meaning ‘to be impressed by and attracted to another person’ rather than
‘to go into a rage’). Regardless, the conformity of both the native and non-native
speakers’ images and responses to probe questions is impressive.
29
For comparison between idiom groups I computed the proportions of
participants’ images and responses that matched the image schemas developed for
the five groups of idioms, while collapsing across participants and idioms. As
noted by Gibbs and O’Brien (1990), there is no theoretical reason to predict
differences in the percent of participants’ reported images and answers to
corresponding probe questions that match. Therefore, I averaged the proportions
of each of these, producing a single proportion for each idiom group. An analysis
indicated no significant variability in matching percentages across the different

idiom groups (see Table 3).
In addition to their general image schemas, participants’ answers to the
probe questions illustrate participants’ understanding of their images. As I
mentioned earlier, the probe question data provide detailed information about
participants’ mental images. These data come in two forms: Some probe questions
required yes/no responses and others elicited open-ended responses that were coded
in the same manner as participants’ images. Again, I was interested in participants’

most frequent reposes to each probe question, collapsing across the five idioms in
each group. These means (for both native and non-native English speakers) are
based on the data presented in Table 3.
Overall, both groups of participants' responses to the different probe
questions about their images were highly consistent. Across the five idiom groups,
native English-speaking participants gave responses that matched the schemas 80%
of the time to the Causation questions, 86% to the Intentionality questions, 79% to
the Manner questions, 82% to the Consequence questions, 76% to the Negative
Consequence questions, and 75% of the time to the Reversibility questions. The
30
proportions for all the probes were greater than chance (with r1(24) = 2.03, 2.78,
1.98, 2.23, 1.81, and 1.79 respectively and p < .05 for each comparison). Non
native English speakers gave responses that matched the schemas 73% of the time
to the Causation questions, 82% to the Intentionality questions, 78% to the Manner
questions, 81% to the Consequence questions, 75% to the Negative Consequence
questions, and 72% of the time to the Reversibility questions. The proportions for
all but the reversibility probes were greater than chance (with fl(24) = 1.74, 2.19,
1.88, 2.25, and 1.78 respectively and p < .05 for each comparison).
Discussion
Gibbs and O'Brien (1990) found a significant number of similarities among
the images that native English speakers produced for analyzable idioms; finding the
same images among non-native English speakers for the same sample of idioms
further supports the claim that images for analyzable idioms are motivated by
universally-held conceptual metaphors.
The non-native English speakers’ data are consistent with Gibbs and
O’Brien’s findings for native English speakers. That is, native Latvian speakers’
mental images for English language idioms’ literal meanings were the same or
similar to those produced by native English speakers, as were their answers to the
standard set of probe questions asked regarding those images. In fact, the results
presented here for both the native and non-native speakers showed, on average,
slightly more conformity across images than Gibbs and O’Brien’s findings for
native English speakers.

This may be due in part to the interview format that was used to find out
about these native and non-native speakers’ images. That is, where Gibbs and
31
O’Brien wrote down participants’ abbreviated responses during the course of the
imaging task, I audio-recorded the entire discussion with each participant, allowing
me the freedom to ask much more detailed questions and to capture the breadth of
their insights regarding the idioms’ characteristics. Of course, the down-side of
such detailed interviews is that they allow for expectancy bias (on the part of the
experimenter) to influence the study’s outcome (though as I specified earlier, my
questions were carefully tailored to try to avoid this. I designed Experiment 3 to
eliminate any biasing confounds that Experiment 1’s interview format may have
introduced.
I also wanted to see whether the mental imaging task helped non-native
speakers use idioms more naturally than they were able to prior to completing the
task. That is, if the images native speakers form for idioms reflect mappings
between source and target domains, then non-native speakers should benefit from
attempting to do this themselves for phrases in the language they are learning. Of
course, the linguistic knowledge native speakers have within each of these domains
is what distinguishes them as native speakers. Nevertheless, adults who are able to
successfully learn another language do so in large part because they can adapt to the
way speakers of their new language lexicalize concepts. In order to do so, adult
learners need to analyze phrases from their new language in a way similar to that
promoted by the mental imaging task. Therefore, the Latvian speakers who
participated in Experiment 1 provided an ideal opportunity for testing this
hypothesis. These data are reported in Experiment 2.
32
Experiment 2
Experiment 2: Naturalness ratings for native and non-native English speakers’
use of idioms before and after mental imaging task
Experiment 2 tests the prediction that non-native speakers will use idiomatic
phrases from another language more naturally following participation in a task that
requires them to map the contents of idiomatic phrases to their underlying
conceptual metaphors.
To determine whether non-native speakers produced more natural sounding

sentences using the test idioms following the mental imaging task than they
produced prior to the mental imaging task, I had another group of native English
speakers compare the two sets of sentences for how “natural” they sounded.
Sentences produced by native speakers (also in Experiment 1) served as a
comparisons.
Method
Participants. Thirty undergraduate psychology students at the State
University of New York at Stony Brook judged sentences in order to fulfill a

research requirement for a psychology course. There were 16 women and 14 men,
and all identified themselves as native speakers of English.
Materials. A questionnaire was assembled from the sentences generated by

the native English speakers and the native Latvian speakers in Experiment 1 for five
of the 25 idiomatic phrases used in that experiment, both before and after they
33
participated in the mental imaging task. Each of the five idioms was randomly
selected from the set of five idioms for each concept group. That is, each idiom
selected to be used in Experiment 2 represented one of the five concept groups
(anger, revelation, insanity, control, and secrecy). For each of the five idioms,
pairs of example sentences were created from sentences produced by the 25 native
English speakers and the 25 Latvian speakers (that is, both before and after they
participated in the mental imaging task). This resulted in a total of 250 pairs of
sentences (125 from the native speakers; 125 from the non-native English
speakers). Members of each pair of sentences came from the same speaker. The
questionnaire asked, for each pair of sentences, (1) Which o f these sentences
sounds more natural? and (2) How natural does the sentence you chose sound?
The second question had a 7-point answer scale with 1 as "not very natural" and 7
as "very natural." The sentences were blocked by concept, such that each block
consisted of sentences based on the same idiomatic phrase representing a single

concept This resulted in five blocks with 50 pairs of sentences per block. Within
each block, pairs of sentences were presented in a random order (including
sentences from both the native and non-native English speakers), and within each
pair, sentences produced prior to and following the mental imaging task were
presented in each pair in a random order. No item appeared twice, as both native
and non-native English speakers produced unique sentences both before and after
the mental imaging task. We verified that each sentence in a pair differed,
guaranteeing that judgments of naturalness were not influenced by repetitions of
identical sentences within a pair.
34
Procedure. Each native-speaking judge completed the questionnaire

individually. Judges were told to compare each pair of sentences, choose the more
natural one, and then rate it for naturalness.
Results
Before and after choices. The sentences the Latvian speakers produced
using the idiomatic phrases after participating in the mental imaging task were
chosen by the native-speaking judges as more natural than those produced before,
62.8% of the time (greater than chance at t l (29) = 6.44, p <.001; t2( 123) = 3.21, p
< .003). Although this increase in naturalness may be simply a product of the
increased time the Latvian speakers knew the canonical definitions of the phrases, it
may also be that the Latvian speakers came to better understand how to use the
English idiomatic phrases by mapping the phrases to their underlying conceptual
metaphors during the mental imaging task. As was expected, the naturalness of
native speakers' expressions was not affected by their participation in the mental
imaging task; their After sentences were chosen over their Before expressions
52.9% of the time, no different from chance (tl(29) = 1.57, p > .10; r2( 123) = .52,
p > .10).
Naturalness ratings. Although the Latvian speakers learned to use the
English idiomatic phrases more naturally during the mental imaging task, the
sentences they produced were still rated as less natural than those produced by
native speakers, 4.01 to 5.47 (tl(29) = 6.50, p <.001; t2(248) = 7.76, p < .001).
35
Discussion
I take the findings from Experiment 2 as indicative of the degree to which
analysis of an idiom’s surface form required by the imaging task can help people
deduce an idiom’s metaphorical underpinnings. Despite the fact that the Latvian
speakers had studied the figurative meanings of the test idioms prior to participating
in the mental imagery task, the sentences they produced were judged to be less
natural sounding than those they produced following the task. Of course, this
increase in naturalness may have simply been due to the non-native speakers’
increased thinking about the phrases themselves and their figurative meanings.
Clearly a control study needs to be conducted that involves non-native speakers
dealing with idioms in some way that does not promote structural analysis (such as
studying them by rehearsal). Despite obvious limitations on this experiment’s
findings, the data highlight an area of possible application that is worth pursuing in
future research. Further studies are necessary to test the relevance of idiom analysis
as promoted by the mental imaging task to the subsequent naturalness with which
non-native speakers are able to use those phrases.
36
Experiment 3
Like English, Latvian is an Indo-European language. It is possible that the
similarities in interpretations for English idioms made by English and Latvian

speakers in Experiment 1 may be due to the two languages' common origins.
Although English and Latvian are only distantly related, this common-origin
confound raises some concerns that should be addressed using a language from a
different family. At this point, it is also relevant to test Gibbs and O’Brien’s
statement that their findings “[do] not imply that the meanings of idioms should be
predictable [even by native speakers of the language]” (p. 65). If it is the case that
the conceptual metaphors that constrain idioms are universal, as Gibbs and O’Brien
claim, then it seems that at least some (analyzable and abnormally analyzable)
idioms’ meanings (from whatever language) should be predictable by both native

and non-native speakers alike. In order to address these questions, I designed a
two part study similar to Experiment 1.
Experiment 3: Native and non-native English speakers’ before and after mental
images for American English idioms
Experiment 3 was designed to address three key issues: 1) whether or not
one's prior knowledge of the figurative meaning of an idiomatic phrase influences

the mental image one forms based on that phrase's literal meaning, 2) whether the
same similarities seen between the English and Latvian speakers’ images in
Experiment 1 would occur for native speakers of a language completely unrelated to
English, and 3) whether a detailed literal interpretation of non-literal phrases can
help either native or non-native speakers determine the non-literal meanings of those
37
phrases. In order to address these issues. Experiment 3 tests native Mandarin (a
Sino-Tibetan language) speakers’ mental images for English idioms both before and
after they’ve been informed of the figurative meanings of the phrases (which are the
same 25 as those used in Experiment 1). In addition to these changes, both native
English and native Mandarin speakers were tested before and after the first mental
imaging session. This was intended to measure the degree to which the literal
analysis required by the imaging task could help the native Mandarin speakers
determine the figurative meaning of the idiomatic phrases; data from the native
English speakers served as a control. Finally, a questionnaire format was
substituted for the interview format used in Experiment 1 to control for possible
interpersonal or interaction effects by the experimenter on participants' responses.
Where I asked the native Latvian speakers in Experiment 1 about their

mental images for idiomatic phrases only after I made them aware of those phrases’
figurative meanings, in Experiment 3 I asked native Mandarin speakers about their

images for English idioms both before and after they'd been informed of the
idioms' canonical figurative meanings. This allowed me to compare the degree to
which Mandarin speakers’ knowledge of the correct idiomatic meanings for phrases
(e.g., "to get angry" for blow your stack) enables them to access metaphors that
constrain the images they produce and report for those phrases. In addition to
including two sessions of mental imaging, I tested the Mandarin speakers for their
intuitions regarding the idioms' figurative meanings both prior to and following the
first mental imaging session. These pre- and post-tests served as a measure of the
degree to which a literal analysis of idiomatic phrases can help speakers determine
38
the figurative meaning of idiomatic phrases when there is little to no context to help
them.
Method
Participants.
Native speakers: Twenty-five undergraduate students of the State
University of New York at Stony Brook were paid five dollars for their
participation in the experiment There were 12 women andd 13 men, and all
identified themselves as native speakers of English. None reported the ability to
speak another language to a degree that would allow them to be considered
bilingual.
Non-native speakers: Participants were twenty-five non-native English
speakers from mainland China. Each was paid five dollars for their participation in
the experiment There were 14 women and 11 men, and all identified themselves as
native speakers of Mandarin Chinese.
Since Mandarin is a Sino-Tibetan language (quite distinct from any of the
Indo-European languages), testing a Mandarin-speaking population addresses the

common-origin confound that existed with the testing of native Latvian speakers.
Although it is common for people from mainland China to speak any number of
Sino-Tibetan dialects^ depending on what region of China they are from, formal
education takes place in Mandarin and all forms of media in the country are
transmitted in standard Mandarin. Because all of the native Mandarin-speaking
participants in Experiment 3 were new graduate students with several years of
5 As noted earlier, these so-called dialects tend to be entirely different languages,

although for political and cultural reasons they are typically not referred to as such.
39
higher education in China, they were all fluent in Mandarin. Given that any of the
dialects that the students might speak in addition to Mandarin are also unrelated to
Indo-European languages, their multi-lingualism did not present itself as a
confound for the present study. Participants were limited to native Mandarin
speakers from Mainland China in order to control for differences between

languages within the language family (e.g., Cantonese, any of several Min
dialects), as well as for differences in foreign language instruction techniques that
might exist between China and Taiwan (even though most Taiwanese people speak
Mandarin fluently).
In order to collect data from native Mandarin-speaking graduate students
who had only recently arrived in the United States, participants for Experiment 3
were recruited from English as a Second Language courses offered to new non
native English speaking graduate students by Stony Brook’s Department of Applied
Linguistics. Only those graduate students who had arrived in New York from
Mainland China within six months of taking the course were included in the study.
This served as a further control for the amount of idiomatic English participants had
been exposed to prior to participating in the study.

Further selection of participants was based on their pre-course English
competence. Upon acceptance to Stony Brook, all foreign graduate students must
take a Test of English as a Foreign Language (TOEFL), the score of which
determines whether each will attend ESL courses during the first year at Stony
Brook. Thus, students’ TOEFL scores provide a standard measure of their English
knowledge prior to entering the university. All of the non-native English speakers
who participated had TOEFL scores between 500 and 600.
40
Materials. Stimuli were the same 25 idiomatic phrases and six probe
questions used in Experiment 1. However, where Experiment 1 employed an
interview format. Experiment 3 was presented as a questionnaire. This
questionnaire protocol is a departure from the original protocol followed by Gibbs
and O’Brien (1990) and Experiment 1. Where the interview format is open to
experimenter bias (for example, positive or negative prompts for answers may be
given), the questionnaire format allows each participant to receive identical

instructions and presentation of test idioms. Five different randomized versions of
the questionnaire were prepared; each participant was randomly given one of these
five questionnaires. Native and non-native speakers completed identical mental

image questionnaires, except that the non-native speakers’ questionnaires consisted
of both the English idiomatic phrases and their translations in Mandarin. These
translations were direct, such that the literal meaning of the phrase when understood
in English was clearly conveyed in the Mandarin version as well. Instructions and
probe questions were presented in Mandarin.
Translations: Although the difficulties inherent in translating experimental
stimuli from one language to another are well-documented (e.g., see Au, 1983,
1984 for a translation-based argument against Bloom’s (1981) study supporting a
weak version of the Sapir-Whorf hypothesis), translation was a fundamental part of

the present study. Given the necessity of translation, every effort was made to
ensure that translations between English and Mandarin were accurate and reflected
the appropriate (figurative or literal) meanings of the instructions, experimental
stimuli, and participant responses.
41
Translations of the 25 American English idioms were completed by two
fully bilingual Mandarin/English speakers, and were later judged for accuracy by 5
native Mandarin speakers who had resided in the United States for 10 years or
more. All ambiguities (which amounted to two out of the 25 idioms) were
corrected, based on these 5 native speakers' recommendations (where the
ambiguities were actually noted by only 2 of the 5, indicating their subtlety). The
six probe questions that followed each idiom were also translated into Mandarin by
the same bilingual Mandarin/English speakers and compared for accuracy. No

significant semantic differences were noted between the two sets of translations
(given the literal nature of the probe questions, this is not surprising), ensuring that
each question's intended meaning was accurately represented. Again, the same 5
native Mandarin speakers judged the probe questions for clarity. Three sets of
instructions (for both the pre- and post-tests and for the mental imaging
questionnaire) were also translated into Mandarin and checked for accuracy
following the same procedure (see Appendix B for a complete set of questionnaires
and instructions, in both English and Mandarin).
Procedure. The basic procedure included a pre-test, the mental image
questionnaire, and a post-test similar to the pre-test. Participants were tested
individually. Non-native speakers were encouraged to write in Mandarin if they
found this more comfortable than English (almost all did so), so that they were able
to fully and fluently describe their mental images. At the first session, both native
English and native Mandarin speakers were given an English-only pre-test. Data
from the native English speakers established a baseline for how familiar native
speakers were with the test idioms. Data from the native Mandarin speakers was
42
used to determine how much idiomatic English they knew prior to participating in
the study. They were told to identify only those phrases they had heard or seen
used before, and to provide a figurative definition to each phrase they identified.
They then completed the mental image questionnaire. Questionnaires for native and
non-native speakers were identical, except that the non-native speakers’
questionnaire was written in Mandarin, with each idiom presented in English and
accompanied by a literal Mandarin translation. This was meant to help the non
natives in their literal analysis of each phrase.
Following completion of the questionnaire, participants were given a post
test (non-native speakers’ post-tests consisted of the 25 idioms in English with

accompanying Mandarin translations). Native speakers were told to simply provide
figurative definitions for as many of the phrases as they could. Non-native
speakers were asked to figure out as many of the phrases’ figurative meanings as
they could. Participants (both native and non-natives) were then given a study
sheet with the idiomatic definitions (in English) for the 25 idiomatic phrases. At
this time they scheduled a second session that took place approximately two weeks
from the date of the first session. The second session followed the same format as
the first, except that rather than completing a post-test, the non-native speakers were
asked to write down as many Mandarin language idioms as they could (these are
relevant to a subsequent experiment). Native speakers did not complete a post-test

following the second session.
First session.
Pre-test: Participants were given a list of the 25 American English idioms in
English. They were also given a set of instructions (written in Mandarin for the
43
non-native speakers) that instructed them to read through the list of idiomatic
phrases and to provide the figurative definition for those phrases that they had heard
or seen used before and thought that they knew. Mandarin speakers were told to
identify only those English phrases whose idiomatic meanings they thought that
they knew. That is, they were told not to guess what the phrases mean, only to
report what they already knew about English idioms. Mandarin speakers were
assured that they were not expected to know any of the idioms' figurative meanings
and that they would not be evaluated based on how many idioms they were able to
define at any point in the study.
The pre-test for the first session provided a measure of how many of the
idioms both native and non-native speakers knew prior to participating in the study.
It also served as a means for those who qualified to remain in the study to review
the phrases before receiving the mental image questionnaire. Pre-tests were scored
for accuracy. Those native English speakers who did not accurately define at least
15 of the 25 idiomatic phrases did not continue with the study; those non-native
English speakers who accurately defined more than ten of the 25 idiomatic phrases
did not continue with the study.

Session I quesdonnaire; Following completion of the pre-test and prior to
completing the mental image questionnaire, participants were provided with a
second set of instructions explaining the mental image procedure. Again, the native
Mandarin speakers received instructions in Mandarin. The instructions first
explained that the participants would be asked several questions about a set of
American English idioms (for the Mandarin-speakers, it was explained that the
idioms would be presented in English, followed by the literal Mandarin translation
44
of each phrase in parentheses). For each idiom, participants were instructed to

carefully read both the idiomatic phrase (and the accompanying literal translation, if
relevant) and to form a mental image based on the literal meaning of the idiomatic
phrase. They were instructed to briefly describe that image in the space provided,
describing the first mental image that came to mind upon reading each English
phrase.
They were then instructed to answer each of the six probe questions that
followed, referring specifically to the mental image they had just described. The
probe questions were meant to provide more specific information regarding what
participants understand about their mental images than the image schema themselves
can provide. Specifically, the probe question data provided information about
participants' knowledge of their mental images. Probe question data were in two
forms: 1) two probe questions requiring yes/no responses (these were the
Intentionality and Reversibility probes), or 2) four probe questions eliciting open-
ended responses (Causation, Manner, Consequence, Negative Consequence).
The instructions also included an example image for an idiomatic phrase that
was not one of the 25 test idioms (e.g., kick the bucket = a man in cowboy boots
out in a desert kicking a rusty tin bucket). This example was meant to help illustrate
the difference between "literal" and "figurative" (though the difference was also
clearly explained). Participants were instructed to answer all questions in as
detailed and accurate a manner as possible. The non-native speakers were also
assured that the questionnaire was in no way an evaluation of their language

knowledge and was, instead, an exercise meant to familiarize them with English
idioms.
45
During this first session, participants received one of two questionnaires

(either Form A or Form B), each of which had the test idioms a different random
order. They received the other questionnaire during their subsequent session. For
example, if they had received Form A for the first session, they were given Form B
for the second session (and vice-versa if they started with Form B). The order in
which the two forms were presented to participants was counterbalanced to control
for order effects.
Following completion of the questionnaire, all participants were given a
post-test (native Mandarin speakers' post-tests consisted of the 25 idioms in

English with accompanying Mandarin translations). Native English speakers were
asked to simply provide figurative definitions for as many of the phrases as they
could. Native Mandarin speakers were asked to figure out as many of the phrases'
figurative meanings as they could. All of the Mandarin speakers provided
definitions for at least 20 of the test idioms following the first mental imaging
questionnaire.
Finally, the experimenter provided a “study sheet” listing the 25 idioms (in
English only) along with the phrases' idiomatic meanings (also in English—as
defined in Boatner, Gates, & Makkai, 1995). At this point, the experimenter
scheduled a second session with the participant to take place approximately two
weeks following the first session. Participants were asked to become familiar with
the 25 idioms’ (as defined on the study sheet) for this second experimental session.
.Second session. Two weeks after they had completed the first mental image
questionnaire, participants returned to the lab to complete the pre-test and the mental
46
image questionnaire a second time. Again, participants were tested individually.

The second session followed the same basic protocol as the first.
Pre-test: At this point in the study, the pre-test was meant simply to get
participants thinking about the idioms again. As was the case prior to the first
imaging session, participants were given a list of the 25 American English idioms in
English. They were also given the same set of instructions (again written in
Mandarin for the non-nauve speakers) that told them to read through the list of
idiomatic phrases and to provide figurative definitions for those phrases that they
knew. Again, Mandarin speakers were assured that they were not expected to
know any of the idioms' figurative meanings and that they would not be evaluated
based on how many idioms they were able to define. These data were not used in
further analyses.
Session H questionnaire: Following the pre-test, participants received the

appropriate mental image questionnaire and were again instructed to form and
describe mental images based on the literal meanings of each phrase and to answer
the six probe questions for each idiom based on the image they had described.
They were told to describe the first image that came to mind and to try their best to
provide images for each of the 25 idioms.
Following completion of the second mental image questionnaire, the native
English speakers were paid for participating and debriefed. After the native
Mandarin speakers had finished the second mental image questionnaire, they were
asked to generate as many Mandarin idioms as they could that represented the five
concepts. Mandarin speakers were typically able to produce 10 to 15 idioms within
47
the 15 minute period o f time allotted for this exercise. They were then paid for
participating and debriefed.
Coding. Native English speakers’ responses to the first and second mental
image questionnaires were transcribed and imported into the Sequence program
(described in Experiment 1). Native Mandarin speakers’ responses (to both mental
image questionnaires and to the pre- and post-tests) were translated by two bilingual
Mandarin/English speakers, whose translations were compared for accuracy. No
significant semantic differences were noted between the two sets of translations,
ensuring that participants’ responses were accurately represented in the English
translations. The Mandarin speakers' responses were then transcribed and
imported into the Sequence program.
Using the proportions of native English speakers' most common responses

established in Experiment 1 as a comparison measure, two coders (again, the
experimenter and a research assistant) followed the same procedure that they
followed while coding the native Latvian speakers' images (in the second half of
Experiment 1). This involved comparing the native English and the native
Mandarin speakers’ responses (from Experiment 3) to the most common responses
generated by the native English speakers in Experiment 1 and coding them as either
matching or not matching those responses.
As in Experiment 1, the images were also coded for whether they were
based on the idiom's figurative or literal meaning, or instead represented a hybrid of
the two meanings. This was intended to illustrate the degree to which, even when
clearly instructed to base their images on idioms' literal meanings, people include
aspects of both the figurative and literal meanings in their images. We then coded
48
each of the three types of images for whether or not they matched the general
characteristics and probe question responses that were established for the five
concept groups in Experiment 1 (see Table 3 for these image schemas).
Along with participants' first and second session images and their responses
to the probe questions, the definitions that each participant (both native English and
native Mandarin speakers) provided during the pre- and post-tests preceding and
following the first mental imaging session were coded for accuracy. This allowed
us to measure the degree to which structural analysis of an idiom's literal meaning
guided speakers (of both English and Mandarin) to determine the phrase's
figurative meaning.
Each coder coded all participants' responses. Reliability was coded with
satisfactory agreement between coders (91.6% agreement, Cohen's Kappa = .90).
Results
The percentage of images reported by Mandarin and English speakers that
were coded as literal, figurative, or hybrids are shown in Table 4. Again, it is
notable that even when explicitly instructed to base their images on the literal
meanings of the phrases, both groups of speakers form hybrid images the majority
of the time. For the first session (collapsed across the different idiom groups), an
average of 69% of native English speakers gave hybrid images for the different
English idioms (ranging from 56.1% to 76.1%). For the second session this
average was 67% (ranging from 58.7% to 74.8%). An average of 47% of the
native Mandarin speakers’ images during the first session (again collapsed across
the different idiom groups) were hybrids (ranging from 27.7% to 57.4%). For the
second session this average was 54% (ranging from 27.1% to 73.4%). These data
49
indicate that it is much more difficult for native English speakers to separate the
literal from the figurative meanings of idiomatic phrases from their own language,
while native Mandarin speakers did not experience this difficulty as much (probably
in large part because they are not aware of the idioms’ figurative meanings at all).
Focusing on results from the hybrid images only, the data from Sessions I
and II indicate that both groups' responses consistently matched the concept
schemas. For the first session (collapsed across the different idiom groups), an
average of 76% of native English speakers gave responses that matched the
schemas for the different English idioms (ranging from 69.5% to 80.1%), a
proportion that is significantly different from chance, r(30) = 1.81, p < .05 (again
using a null hypothesis of .50). An average of 73% of the native Mandarin
speakers’ responses during the first session (again collapsed across the different
idiom groups) matched the schemas for the different English idioms (ranging from
66.5% to 78.5%), a result that is marginally different from chance, r(31) = 1.62, p
<.10. This indicates that native Mandarin speakers’ analyses of the phrases must
have given them an indication of the phrases’ figurative meanings, at least to some
degree.
For the second session, an average of 77% of native English speakers’
hybrid images for the different English idioms (again collapsed across the different
idiom groups) matched the schemas (ranging from 69.5% to 80.8%), a proportion
that is significantly different from chance, r(31) = 1.95, p < .05. An average of
76% of non-native English speakers’ hybrid images for the different English idioms
collapsed across the different idiom groups described similar general images
(ranging from 73% to 78.5%), significantly different from chance, f(31) = 1.93, p
50
< .05. These data indicate that the Mandarin speakers were beginning to associate
the literal and figurative meanings of the phrases following two weeks of knowing
the canonical, figurative definitions of the phrases. That is, during the second
session they were more likely to give responses that matched the image schemas
than they did during the first session.
Percentages of both English and Mandarin speakers' hybrid mental image
responses that matched the image schemas are presented in Table 5. Overall, both
groups of participants reported highly consistent responses to the different probe
questions about their images for the English idioms. This was particularly so for
the images produced during the second session of the mental image interview. For
the first session, across the five idiom groups native English speakers gave
responses that matched the schemas 70% of the time to the Causation questions,
74% to the Intentionality questions, 72% to the Manner questions, 86% to the
Consequence questions, 83% to the Negative Consequence questions, and 72% of

the time to the Reversibility questions. For the second session, across the five
idiom groups native English speakers gave responses that matched the schemas
70% of the time to the Causation questions, 76% to the Intentionality questions,
71% to the Manner questions, 87% to the Consequence questions, 84% to the
Negative Consequence questions, and 73% of the time to the Reversibility
questions.
For the first session. Mandarin speakers gave responses that matched the
schemas 67% of the time to the Causation questions, 74% to the Intentionality
questions, 68% to the Manner questions, 88% to the Consequence questions, 79%
to the Negative Consequence questions, and 59% of the time to the Reversibility
51
questions. For the second session. Mandarin speakers gave responses that matched
the schemas 76% of the time to the Causation questions, 76% to the Intentionality
questions, 66% to the Manner questions, 89% to the Consequence questions, 79%
to the Negative Consequence questions, and 68% of the time to the Reversibility
questions. Data from the probe questions indicate that when speakers’ images were
hybrids of phrases’ figurative and literal meanings, their intuitions about the
characteristics of their images tended to match the schemas based on native
speakers' most common responses.
Finally, data from the pre- and post-tests (for Session I; see Table 6)
support the hypothesis that the analysis of idioms’ surface structures promoted by
the imaging task would help Mandarin speakers deduce the idioms’ figurative
meanings. Where Mandarin speakers gave correct responses only 15% of the time
on the pre-tests, they correctly defined the test idioms 49.7% following completion
of the imaging questionnaire. This result makes a compelling argument for the
analyzability of some idiomatic phrases. It is notable that at no time prior to the
post-test were these non-native English speakers given information about what the
phrases mean figuratively. Their increase in correct figurative definitions for the
phrases must have come entirely from their analysis of the phrases’ surface
structure, a process that was guided by the imaging task.
Discussion
The results from Experiment 3 are consistent with findings in Experiment 1
regarding speakers’ mental images. That is, native Mandarin speakers' hybrid
images for the English language idioms matched schemas based on images
produced by native English speakers, as did their answers to the standard set of
52
probe questions asked regarding those images. Furthermore, following a two week
period during which they were made aware of the idioms’ figurative meanings, the
proportion of Mandarin speakers’ responses that matched the image schemas
increased.
Pre- and post-tests from the first imaging session indicate that speakers can
infer figurative meanings based entirely on an analysis of the surface structure of
the phrase (that is, without supporting context). Given that we hear and read
idioms most often in context, the degree to which these phrases are understandable
without any instruction regarding their canonical meanings is compelling.
Furthermore, although the images Mandarin speakers produced prior to

being informed of the phrases’ figurative meanings were more often based on literal
interpretations of the phrases than native English speakers’ images were, given the
increase in correct definitions of the idioms’ figurative meanings shown in the
Session I post-test results, it seems that the Mandarin speakers were engaged in a
process of analysis during which they mapped the phrases’ contents to their
corresponding conceptual metaphors. The similarities among native English
speakers’ images for idioms, as well as among native Latvian and Mandarin
speakers’ images for English idioms, may then be characterized as motivated or
constrained by that knowledge.

Again, there is no explanation in traditional linguistic analyses of
idiomaticity for such regularity in people’s knowledge about idiomatic phrases.
Data from the probe questions indicate that both native English and native Mandarin
speakers, upon reflection, are able to access very specific information about the
causes and effects of actions in their images for English idioms. As Gibbs and
53
O’Brien (1990) argued (based on data from native English speakers), these data
“demonstrate how our understanding o f idioms is motivated by different conceptual
metaphors” (p. 46). Where Gibbs and O’Briens’ data address the general issue of
whether idioms are motivated by underlying conceptual metaphors, the (similar)
results presented here from both nadve and non-native speakers o f English work
towards answering the question of whether or not these metaphors are universal.
Given the striking similarities between native English speakers’ and native Latvian
and Mandarin speakers’ images for English idioms, and their answers to probe
questions about those images, these studies support arguments for universal
conceptual metaphors.
One problem with the mental imaging procedure is that it does not reflect
on-line processing. Although the previous experiments indicate that speakers can
analyze idioms’ surface forms in order to deduce the phrases’ figurative meanings,
the imaging task does not give any indication of the time course of this process.
Experiment 4 was designed to address this issue.
54
Experiment 4
In the previous series of experiments, I outlined a general measure of
whether or not idioms are analyzable, based on whether or not the native Mandarin
speaking participants were able to predict an idiom’s figurative meaning based
simply on the phrase’s surface structure. Gibbs, Nayak, and Cutting (1989)
demonstrated that analyzability as Gibbs and Nayak (1989) had defined it earlier-
based on work by Nunberg (1978)—does influence processing. However, what
constitutes an analyzable idiom is still not entirely clear.
Using a simple (e.g., yes/no) meaningfulness confirmation task (e.g.. Is

this a meaningful phrase in English?), Gibbs, Nayak, and Cutting (1989) found
significantly different response times for native English speakers processing
normally analyzable, abnormally analyzable, and unanalyzable English language
idioms in addition to meaningful and nonmeaningful control strings. The
abnormally analyzable idioms produced the shortest response times, the
unanalyzable idioms produced the longest response times, and the normally
analyzable idioms produced response times that fell between the two other groups’
times. All these differences were significant The idioms used as stimuli had been
selected during another study where participants presented with a series of idioms
were asked “to decide whether the individual words in each expression made some
unique contribution to the phrase’s nonliteral interpretations” (Gibbs & Nayak,
1989, p. 108).
The notion of idiom analyzability has still been studied only with regard to
American English idioms. Although the previous series of studies points to a
55
relationship between analyzability and conceptual universalities, it is clear that

additional work needs to be done with other languages’ idioms. That is, before I
(or anyone else) can make claims about universality, it is necessary to test whether
idiom analyzability is a construct that holds cross-linguistically.
Nunberg (1978) was the first to suggest that the frozenness or flexibility of
idioms’ syntactic behavior was determined in large part by their semantic
decomposition. This characterization was the basis of Gibbs and Nayak’s (1989)
subsequent study in which they outlined their “idiom decomposition hypothesis”
(p. 104). This hypothesis, in broad strokes, holds that idioms are partially
analyzable (e.g., Fillmore, 1986; Gibbs, 1985; Gibbs & Gonzales, 1985; Nunberg,
1978; Wasow, Nunberg, & Sag, 1982) rather than simply assigned single semantic
representations unrelated to the meanings of their individual components, a claim
made in earlier idiom research (e.g., Bobrow & Bell, 1973; Swinney & Cutler,
1979). Gibbs and Nayak (1989) argued that the syntactic behavior of idioms is
determined based on speakers’ assumptions about how the meaning of idioms’

parts contribute to the figurative meanings of the whole.
In order to pursue this notion of analyzability, I designed the following

series of experiments to test whether idioms from other languages can be analyzed
based on their parts. I first separated idioms from three different languages
(English, Latvian, and Mandarin) into types (consistent with Gibbs and Nayak’s
(1989) characterization of idioms as being of three principle types; normally,
abnormally, and un analyzable idioms) by having native speakers of those
languages rate the idioms along the continuum that ostensibly represents the full
spectrum of idiom types or possibilities. I then examined native English speakers’
56
processing of these three idiom types from the three languages to determine how the
processing of idioms from one’s native language differs from the processing of
idioms from other languages. The cross-linguistic nature of these experiments will
highlight processing differences having to do with the automaticity of certain
idiomatic phrases for native speakers (e.g., one’s automatic access of an idiom’s
figurative meaning given that the phrase is from—and often repeated in—one’s native
language).
Ultimately, this methodology will highlight the differences in processing of
those idioms whose individual words have literal relations to their figurative
referents, those whose individual components have some metaphorical relation to
their idiomatic referents, and those whose individual components stem almost
entirely from historical and cultural influences on the conventionalization of phrases
in any particular language. The comparison of native English speakers’ processing

of idioms from their own and other languages will enable me to infer the degree to
which these conceptual metaphors are accessed when a phrase is understood
figuratively. Processing of abnormally analyzable idioms from one’s native

language should differ from processing of the same idiom types from another
language. That is, where phrases from one’s native language may be processed
automatically without the mapping of meaning from the phrases themselves to their
underlying metaphors, the lexical components of foreign languages’ phrases will
need to be mapped to the relevant metaphors in order for those phrases to be
understood figuratively.
The present experiments measure native English speakers’ ability to
correctly categorize idioms from their own and other languages according to the
57
concepts underlying their figurative meanings (anger, secrecy, revelation, insanity,
and control). Specifically, I compared the speed and accuracy with which
participants categorize the three types of idioms. The first experiment establishes a
baseline of English speakers’ categorization times for normally, abnormally, and un
analyzable idioms from their native language. The second and third experiments
measure whether and how quickly native English speakers can conceptually
categorize similar types of idioms that are literally translated into English from
genetically diverse language groups (Latvian and Mandarin).
Experiment 4A: Native English speakers’ conceptual categorization of

English idioms
Materials
In choosing stimuli for the conceptual categorization task, I concretized the

distinctions made by Gibbs and Nayak (1989) based on Nunberg’s (1978) work on
the semantic decomposition of idioms. I trained raters to locate numerous idioms

along a analyzability continuum so that I could group the idioms according to three
categories: normally analyzable, abnormally analyzable, and unanalyzable.
First, using idiom dictionaries and other sources, I compiled a large group
of idiomatic English phrases representing the five concept groups that would guide
the subsequent categorization task. I paired each of these idioms with their
figurative paraphrases, as defined in two standard English idiom dictionaries (e.g..

Long & Summers, 1979; Makkai, Boatner, & Gates, 1995). This resulted in a list
of 150 idioms, containing approximately 30 idioms for each of five concepts. All
of the idiomatic phrases followed a basic [Verb + Noun Phrase] structure.
58
Two native English speakers were trained to rate how transparent or opaque
they perceived the relationship to be between the 150 idioms' individual words and
their figurative referents. They were instructed to base this rating on a 7-point scale
anchored such that 1 = "transparent" and 7 = "opaque." The raters were doctoral
candidates in English literature who were naive to the experimental hypothesis
being tested. These raters were presented with a booklet containing a written set of
instructions (see Appendix C) and the 150 idiom-paraphrase pairs. The idioms
were randomly ordered. Every idiom-paraphrase pair was accompanied by a 7-
point Likert scale. Raters were instructed to use all 7 points of the 7-point scales.
Every idiom-paraphrase pair was also followed by a question regarding the

familiarity of that idiom (e.g., Is this idiom familiar or unfamiliar to you?) that
raters were instructed to answer based on whether they had previously used or
witnessed that idiom used in conversation or writing.
Transparent-opaque distinction. The instructions included in the booklet
paraphrased and expanded on Gibbs and Nayak’s (1989) original instructions to
their subjects (see Appendix C for our instructions). An idiom’s rating was based
on how many words in the phrase have transparent (or seemingly literal) relations
to their figurative referents (" 1" on the scale) or conversely, have completely
opaque (or seemingly arbitrary) relations to their figurative referents ("7" on the
scale). Those idioms whose individual components have metaphorical relations to
their idiomatic referents were to be placed in between the extreme ends of the
scales, such that the more metaphorical the relationship, the further up the scale
(towards 7) the idiom's rating would fall. That is, if only one word in a phrase is
metaphorically related to its figurative referent, while all other words are literally
59
related to their figurative referents, then raters were instructed to locate this idiom
lower on the scale (towards 1) than they would a phrase with two words that are
metaphorically related to their figurative referents (which would be placed nearer to
the middle of the scale). If an idiom contains both a metaphorically related word
and a seemingly arbitrary word, this phrase's rating would fall higher up the scale
(towards 7). Finally, if raters perceived both the verb and the noun phrase in an
idiom to have relatively arbitrary (or opaque) relations to their figurative referents,
they were to give the phrase a rating on or near 7.
Raters independently rated all of the idioms and indicated whether or not
they were familiar with each phrase. A reliability measure indicated satisfactory
agreement between them (88.6% agreement, Cohen’s Kappa = .87). Their ratings
were averaged and distributions of the means were compiled for each concept in
order to determine which fifteen idioms from each concept category were
consistently rated such that five fell on one end of the scale (on or near 1), five on
the other (on or near 7), and five in the middle (on or near 4). These five sets of
fifteen idioms were included in the categorization study. That is, the five idioms
that were consistently rated as being transparent for the five different concepts were
considered analyzable idioms to be used in subsequent categorization studies.

Those five that were consistently rated in the middle of the scale were considered
abnormally analyzable and those five that were consistently rated as being opaque
were considered unanalyzable. Only those idioms that were indicated as familiar by
both raters were included in the final set. In this way, the means distributions and
familiarity ratings for the 5 idiom concepts determined which 75 (out of the original
150) idioms were used in the subsequent categorization study.
60
Although Gibbs and Nayak forced their test idioms into one of their three
categories, there are subtle nuances that make an idiomatic phrase more or less
transparent or opaque. By using the system outlined above, I eliminated those
idioms that fall somewhere between the three types. This technique underscores
that idioms are differentiated along a continuum rather than as belonging to three
distinct groups.
Method
Experiment 4A established a baseline measure of the speed and accuracy
with which native English speakers are able to conceptually categorize the three
idiom types (as differentiated in the first half of the experiment) from their own
language in a forced choice task. Toward this end, native speakers categorized the
75 English idioms that had been selected via the norming technique just described
according to their representative concept (anger, secrecy, revelation, insanity, or
control).
Participants. A group of 34 undergraduate students (15 males and 19
females) of the State University of New York at Stony Brook received research
credit for their participation in the experiment All were native speakers of
American English and upon being asked, none reported that they were functionally
fluent in another language.

Materials. Materials for this study were the 75 idiomatic phrases that had
been selected based on the procedure oudined above (see Appendix D for a list of
these idioms).
Procedure. The idiomauc phrases were presented on a Macintosh
computer, using the SuperLab™ program. Each person completed the experiment
61
individually. Idiomatic phrases were presented on the screen one at a time. The
idioms were presented in a different random order to each participant The

keyboard was covered such that only 5 keys were exposed. These were labeled
with capital letters (R, I, C, A, or S) representing one of each of the five categories
(revelation, insanity, control, anger, or secretiveness), and the full name of each
category was located just above the appropriate key. Participants were asked to put
each idiom into the appropriate concept group based on its figurative meaning by
pressing the key that corresponded with that category. Prior to beginning the
experiment, participants read instructions regarding the procedure (see Appendix
E). Prior to beginning the actual experiment, participants completed a brief practice
session that included five idioms, one from each of the five concept categories. If
participants had no questions, they then began the experiment. Participants took
between 10 and 15 minutes to complete the entire experiment Response latencies

were measured from the onset of the word string to when participants pressed one
of the five buttons.
Analyses. For the series of studies that make up Experiment 4 , 1 report two
analyses for each result; FI or rl is the analysis by-subjects (with participants as the
random factor), and F2 or t2 is the analysis by-items. Where appropriate, I
conducted two planned comparisons of the effects of the three conditions,
comparing the three idiom types (that is, the normally analyzable versus the
abnormally analyzable idiom groups, and both of these versus the unanalyzable
idiom group). I did not look at concept as a variable as there was no theoretical
motivation to do so.
Results and Discussion
62
The response times to make the conceptual categorization decisions and the
proportion of idiom types (e.g., normally, abnormally, or unanalyzable) that were
correctly categorized are presented in Table 7. Response times from incorrectly
categorized phrases were eliminated from the analysis, as were response times
greater than 10 s (approximately 2 SDs from the mean).
Error rates for conceptual categorization show a pattern such that normally
analyzable (or transparent) idioms were most often correctly categorized (89.1%),
followed by abnormally analyzable idioms (76.4%), and unanalyzable (or opaque)
idioms (68.4%).
As predicted, differences in the analyzability of the idiom types were

reflected in differences in participants’ reaction times. An analysis of variance was
performed with Idiom Type (normally analyzable vs. abnormally analyzable vs.
unanalyzable) as a within-subject factor, revealing a significant effect of Idiom
Type, F I (2, 66) = 17.34, p < .001; F2(2, 73) = 3.07, p = .05. Participants took a
longer time to correctly categorize the unanalyzable idioms based on their
conceptual meanings than to do so for either the normally or abnormally analyzable

idioms. It appears that accessing the figurative meanings of both normally and
abnormally analyzable idioms is easier than it is for unanalyzable idioms. This may
be due to the fact that people can perform compositional analyses on the figurative
meanings of the phrases, much like they do for more literal language (Gibbs,
Nayak, & Cutting, 1989). Analysis of the individual means indicated that
participants were able to correcdy categorize normally and abnormally analyzable

idioms by concept faster (3720 and 3388 ms, respectively) than they could the
63
unanalyzable idioms (4326 ms, planned comparison, F l ( l , 66) = 30.25, p < .001;
F 2(l, 72) = 8.88, p < .01).
These results indicate that there are differences in people’s processing of

normally analyzable, abnormally analyzable, and unanalyzable idioms, adding
support to earlier evidence (e.g., Nunberg, 1978; Gibbs, Nayak, & Cutting, 1989)
that idioms range along a continuum of analyzability. Data from these English
idioms indicate that the speed with which people process idioms does, indeed,
depend on the degree to which the phrases are semantically analyzable.
To what degree does native English speakers’ preexisting knowledge of
idiomatic phrases in their native language influence their response times for
categorization of another language’s idioms? Experiment 4B was designed to

answer this question by having native English speakers perform the same
categorization task on idioms that were directly (that is, literally) translated from
another language.
Experiment 4B: Native English speakers’ conceptual categorization of
Latvian idioms
The previous experiment established a baseline for the speed and accuracy
with which native English speakers are able to categorize idioms from their own
language; Experiment 4B measures the speed and accuracy with which native
English speakers' are able to categorize another language’s idioms (in this case,
Latvian's) that have been directly (that is, literally) translated into English.
Materials
As in Experiment 4A, the first half of Experiment 4B was meant to establish

which Latvian idioms would be included as stimuli in the subsequent categorization
64
study. Again, given that Gibbs and Nayak’s (1989) definitions are somewhat
ambiguous regarding what makes an idiom normally, abnormally, or un analyzable,
a large body of Latvian idioms were submitted to the same rating system as the
English idioms were in Experiment 4A. The same guidelines were followed that
had been used to train the native English speaking raters in Experiment 4A.
Briefly, two native Latvian speakers were trained to rate how transparent or opaque
they perceived the relationship to be between the 150 idioms' individual words and
their figurative referents. This rating was based on a 7-point scale anchored such
that 1 = "transparent" and 7 = "opaque." The raters were 2 linguists educated at the
University of Latvia who were naive to the experimental hypothesis being tested.
Although both were able to speak English, neither had been raised in an English
speaking home and both had begun studying English relatively late (during high
school) in their academic careers. That is, neither of the raters spoke English
natively, nor had lived in an English speaking environment.
A list of 150 Latvian idioms was compiled, again containing 30 idioms for
each of five concepts. These concepts were the same as those included in
Experiments 1, 2, and 3 A (anger, insanity, authority, secretiveness, and
revelation). Each of the 150 idiomatic phrases followed the same basic [Verb +
Noun Phrase] syntactic structure. Coders were presented with a booklet containing
a written set of instructions (see Appendix C) and the 150 Latvian idioms paired
with paraphrases of their figurative meanings. The idioms were randomly ordered
for each coder. Each idiom-paraphrase pair was accompanied by a 7-point Likert
scale. Each idiom-paraphrase pair was also followed by a question regarding the
familiarity of that idiom. Coders were instructed to use all 7 points in the scale.
65
They were instructed to answer the familiarity question based on whether they had
previously used or witnessed that idiom used in conversation or writing.
As with the English idioms in Experiment 4A, a Latvian idiom’s rating was
described to the raters as based on how many words in the phrase have transparent
(or seemingly literal) relations to their figurative referents ("1" on the scale) or
conversely, have completely opaque (or seemingly arbitrary) relations to their
figurative referents ("7" on the scale). Those idioms whose individual components
have metaphorical relations to their idiomatic referents were to be placed along the
scale such that the more metaphorical the relationship, the further up the scale
(towards 7) the idiom's rating would fall. The same instructions that were included
in the English raters' booklets were translated into Latvian for the Latvian raters by
a bilingual English/Latvian speaking acquaintance of the author’s. Again, these
instructions paraphrased and expanded on Gibbs and Nayak's (1989) original
instructions, taking into account the change in terminology and distinctions between
"transparent" and "opaque”.

The two raters independently rated all of the idioms and indicated whether
or not they were familiar with each phrase. A reliability measure indicated
satisfactory agreement between them (86.7% agreement, Cohen's Kappa = .85).
Their ratings were averaged and a distribution of those means was compiled. A
means distribution for each concept was used to determine, for each of the five
concepts, which fifteen idioms were consistently rated such that five fell on one end
of the scale (on or near 1), five on the other (on or near 7), and five in the middle
(on or near 4). These fifteen idioms were included in the categorization study. In
this way, the means distributions for the 5 different idiom concepts determined
66
which 75 (out of the original 150) idioms were then used in the subsequent
categorization study. Only those idioms that were familiar to both raters were
included in the final set of 75 (see Appendix F for a list of these idioms).
Once it was determined which 75 idioms were to be used in the

categorization study, a bilingual Latvian/English speaker who was naive to the
experimental hypothesis translated each idiom directly (literally) into English.
These translations were then checked for accuracy by a second bilingual
Latvian/English speaker. Any ambiguous or problematic translations were noted
and the two Latvian/English bilingual speakers subsequently discussed the
ambiguities until they agreed upon an acceptable translation.
Method
Experiment 4B measured the speed and accuracy with which native English
speakers were able to conceptually categorize groups of the three types of idioms
that had been directly translated from another Indo-European language (Latvian) to
English. Toward this end, native English speakers participated in the timed
conceptual categorization of the 75 Latvian idioms.

Participants. A different group of 34 undergraduate students (16 female and
18 males) of the State University of New York at Stony Brook received research
American English and none reported that they were functionally fluent in another
language.
Procedure. The procedure was identical to that followed in Experiment 4A.
Prior to beginning the experiment, participants read instructions regarding the
67
conceptual categorization task that included relevant examples from Latvian (see
Appendix G).
As before, participants completed a brief practice session that included five
literally translated Latvian idioms, with one from each of the five concept
categories. Participants took between approximately 15 to 20 minutes to complete
the entire experiment

Native English speakers’ response times for making conceptual
categorization decisions and the proportion of idiom types (normally, abnormally,
or unanalyzable) that they correctly categorized are presented in Table 7. Response
times from incorrectly categorized phrases were eliminated from the analysis, as
were response times greater than 10 s (approximately 2 SDs from the mean).
Error rates for native English speakers’ conceptual categorization of Latvian
idioms directly translated into English show a pattern similar to that seen for
English idioms in Experiment 4A, although there was an overall reduction in the
number of idioms categorized correctly. Normally analyzable idioms again were
most often correctly categorized (79.4%), followed by abnormally analyzable
idioms (69.2%), and unanalyzable idioms (43.9%). The most noticeable drop in
English speakers’ correct conceptual categorization of Latvian idioms is in the
unanalyzable group. This is consistent with the characterization (outlined earlier) of
unanalyzable idioms as having extreme metaphorical bases or historical/cultural

bases that seems opaque to individuals not privy to the language in which these
idioms are used. In fact, it is notable that native English speakers were able to
68
correctly categorize over 40% of the so-called unanalyzable idioms, particularly

since there were five possibilities for them to choose from.
However, where native English speakers’ error rates for conceptual
categorization of Latvian idioms followed a pattern consistent with that seen in
Experiment 4A, their response times for those categorization decisions diverged
from the earlier results with English idioms. An analysis of variance was
performed with Idiom Type (normally analyzable vs. abnormally analyzable vs.
unanalyzable) as a within-subject factor. These analyses revealed a significant
effect of Idiom Type, F I (2, 66) = 39.93, p < .001; F2(2, 72) = 12.46, p < .001.
Further analysis of the individual means indicated that participants were able
to correcdy categorize normally analyzable idioms significantly faster (3791 ms)
than abnormally analyzable idioms (4421 ms), (planned comparison, FI (2, 66) =
11.92, p < .001; F2(2, 74) = 6.80, p < .001). This is opposite of the pattern seen
for English speakers’ categorization of English idioms in Experiment 4A, and is a
more intuitive one given that abnormally analyzable idioms are further down the
transparent-opaque continuum than are normally analyzable idioms. This indicates
that when certain idioms (e.g., abnormally analyzable ones) are highly familiar
(e.g., overleamed), they are processed differently than are those seen/heard for the
first time. When participants were able to correctly categorize unanalyzable idioms,
it took them significandy longer to do so than for either of the other idiom types
(5880 ms), (planned comparison, Fl(2, 44) = 11.92,p < .001; F2(2, 74) = 6.80,
p < .001), a finding that is consistent with results for English idioms.
Experiment 4C: Native English speakers’ conceptual categorization of
Mandarin idioms
69
The pattern of findings in Experiment 4B may have been in large part due to
the historical similarities between English and Latvian, since both are Indo-
European languages. Experiment 4C uses the same paradigm as Experiments 4A

and 4B, but replaces English and Latvian idiomatic phrases with Mandarin Chinese
idioms that have been literally translated into English. This change in language
family (Latvian and English are both Indo-European languages, while Mandarin is a
Sino-Tibetan language) was intended to replicate Experiment 4B while removing
any confound due to geneuc similarities between Latvian and English that might
influence the speed and accuracy with which speakers of one language are able to
determine the figurative meanings of the three types of idioms directly translated
from the other.
Method
As described in Materials for both Experiment 4A and 4B, Mandarin idioms
were rated by native Mandarin speakers in order to make the three-way analyzability
distinction. A set of Mandarin idioms was compiled with accompanying literal and
figurative English translations. The compilation was done by the native Mandarin
speaking participants who participated in Experiment 3.
Following the second session of Experiment 3, the native Mandarin
speakers were asked to compile as many idioms as possible from their own
language expressing a variety of meanings. They were asked to focus particularly
on generating idioms that expressed one of five concepts (e.g., anger,

secretiveness, revelation, insanity, and control). The Mandarin speakers were each
given a form with the five concepts listed as column headers. Each column
provided ample space for them to write down idiomatic phrases reflecting the
70
meaning of the corresponding header. Participants were asked to write the idioms
in Mandarin.
These idioms were compiled into one list by a native Mandarin-speaking
research assistant, and all redundancies were eliminated. This produced a collection
of idioms similar to that compiled for the English and Latvian rating task in
Experiments 4A and 4B. The research assistant then selected 150 of the Mandarin
idioms based on her familiarity with them such that there were 30 from each
concept group. These 150 idioms, along with their accompanying figurative
paraphrases, were included in a rating questionnaire identical to those used for the
English and Latvian idioms. The rating questionnaire was written using a Mandarin
language software package.
Two other native Mandarin speakers were then trained to analyze the 150
idiomatic phrases for how transparent, metaphorical, or opaque they seem in
relation to their figurative paraphrases. The raters were both graduate students (one
in Linguistics and the other in Experimental Psychology) who had received their
primary and secondary education in Mainland China and who had arrived in the
United States to begin their graduate studies within a year of this study. Both were
naive to the experimental hypothesis being tested. Although both were able to
speak English to some degree, neither had been raised in an English speaking home
and both had begun studying English relatively late (during college) in their
academic careers. That is, neither were native English speakers nor had lived in an
English speaking country prior to their graduate studies.

The Mandarin idioms were rated using the system that was used on the
English and Latvian idioms in Experiments 4A and 4B, and the same guidelines
71
were followed as those used to train native English and Latvian speaking raters in
the previous experiments. That is, the native Mandarin speaking raters were trained
to rate how transparent or opaque they perceived the relationships between the 150
idioms' individual words and their figurative referents to be. Again, ratings were
based on a 7-point scale anchored such that 1 = "transparent" and 7 = "opaque."

The idioms were randomly ordered for each rater. Each idiom-paraphrase pair was
also followed by a question regarding the familiarity of that idiom (e.g.,
familiar/unfamiliar). Raters were instructed to use all 7 points of the scale. They
were instructed to answer the familiarity question based on whether they had
previously used or witnessed that idiom used in conversation or writing.
Raters independently rated all of the idioms and indicated whether or not
they were familiar with each phrase. A reliability measure indicated satisfactory
agreement between them (82.1% agreement, Cohen's Kappa = .80). Their ratings
were averaged and a distribution of those means was compiled. As was done with
the English and Latvian idioms, means distribution for each concept was used to
determine which 15 Mandarin idioms were consistently rated at the two extremes
and the middle of the scale. These 15 idioms were selected for inclusion in the
categorization study, resulting in equal numbers of normally, abnormally, and
unanalyzable idioms for each concept. Only idioms that were familiar to both raters
were included in the final set of 75 (see Appendix H for a list of these idioms).
Once it was determined which 75 idioms were to be used in the
categorization study, a bilingual Mandarin/English speaker who was naive to the
experimental hypothesis translated each idiom directly (literally) into English.

These translations were then checked for accuracy by a second bilingual
72
Mandarin/English speaker and for fluency by the experimenter (a native English

speaker). Any ambiguous or problematic translations were noted and subsequendy
discussed until an acceptable translation had been determined.
Method
Just as in the second half of Experiment 4A and 4B, Experiment 4C
measured the speed and accuracy with which native English speakers conceptually
categorized those 75 Mandarin idioms selected via the norming technique, based on
their intuitions about each idiom’s figurative meaning.
Participants. A group of 34 undergraduate students of the State University
of New York at Stony Brook received research credit for their participation in the
experiment. All were native speakers of American English and none reported that
they were functionally fluent in another language. In addition, none reported
having been exposed to Mandarin Chinese to any degree as children or young

adults.
Procedure. The procedure was identical to that followed in Experiments 4A
and 4B. Prior to beginning the experiment, participants read the instructions for the
conceptual categorization task with relevant (Mandarin) examples (see Appendix I).
Prior to beginning the experiment, participants completed a brief practice
session that included Mandarin idioms that were not included in the experimental
trial, with one from each of the five concept categories. Participants took
approximately 20 minutes to complete the entire experiment.

categorization decisions and the proportion of idiom types (e.g., normally.
73
abnormally, or unanalyzable) that they correctly categorized are presented in Table
7. Response times from incorrectly categorized phrases were eliminated from the
analysis, as were response times greater than 10 s (approximately 2 SDs from the
mean).
Error rates for native English speakers’ conceptual categorization of
Mandarin idioms direcdy translated into English show a pattern similar to that seen
for native English speakers conceptually categorizing English idioms, although
there was an overall reduction in the number of idioms categorized correcdy.

Normally analyzable idioms were most often correcdy categorized (82.4%),
followed by abnormally analyzable idioms (59.3%), and unanalyzable idioms
(50.1%). The most nouceable drop in English speakers’ correct conceptual
categorization of Mandarin idioms is between the normally and abnormally
analyzable group. However, the difference between the number of abnormally
analyzable and unanalyzable idioms that were correcdy categorized is consistent

with the characterization (outlined earlier) of unanalyzable idioms having extreme
metaphorical bases or historical/cultural bases that seems opaque to individuals
unfamiliar with the language in which these idioms are used.
Again, it is notable that native English speakers were able to correcdy
categorize half of the so-called unanalyzable idioms. The fact that more of the
unanalyzable idioms were correcdy categorized from Mandarin than from Latvian is
surprising, given that Latvian is genetically more closely related to English than is
Mandarin. This may be due to the more descriptive or metaphorical nature of the
Mandarin idioms in general (the Mandarin stimuli are listed in Appendix H).
Regardless, there is a general decrease in the number of idioms accurately
74
categorized according to concept across the three idiom types (with increased error
rates from normally to unanalyzable idioms) that is consistent with the pattern of
error rates seen for both English and Latvian idioms in the previous two
experiments.
Native English speakers’ response times for those categorization decisions
mirrored results from Experiment 4B (for Latvian idioms), again diverging from
response times for English idiom categorization decisions. An analysis of variance
revealed a significant effect of Idiom Type, FI (2, 66) = 78.81, p < .001; F2(2, 74)
= 25.07, p < .001. Further analysis of the individual means indicated that
participants were able to correctly categorize normally analyzable idioms
significantly faster (4018 ms) than abnormally analyzable idioms (5123 ms,
planned comparison, F l( l, 66) = 74.21,/? < .001; F2(l, 72) = 7.62, p < .01).
Furthermore, when participants did correctly categorize unanalyzable idioms, it

took them significantly longer to do so than for either of the other idiom types
(5788 ms), (planned comparison, F l(l, 66) = 83.10,/? < .001; F2(2, 74) = 6.54,
p < .025).
Just as they do when processing idioms in their own language, native
English speakers demonstrated different processing times for categorizing translated
Mandarin idioms according to their figurative meanings, depending on whether the

phrases are analyzable, abnormally analyzable, or unanalyzable. Normally
analyzable Mandarin idioms were the easiest for participants to comprehend,
followed by abnormally analyzable idioms, while un analyzable idioms were the

most difficult to comprehend as meaning what they mean figuratively. The
response time differences for native English speakers’ conceptual categorization of
75
Mandarin idioms supports the notion that idioms can be characterized as normally,
abnormally, or un analyzable (Gibbs, Nayak, & Cutting, 1989). However, where
these researchers considered only native English speakers processing English
idioms, the present series of experiments shows the degree to which metaphorical
processing of familiar idioms becomes automatic for native speakers of a language.

Although native English speakers process both normally and abnormally analyzable
idioms equally quickly (with a trend towards faster processing of the abnormally
analyzable ones), their processing times for other languages’ abnormally analyzable
idioms demonstrate the time course for accessing new phrases’ underlying
metaphors.
Each experiment in this series was run independently of the others. In order
to accurately demonstrate the change in metaphorical processing that comes about as

a native speaker of a language, it was necessary to run a conceptual categorization
experiment that included idioms from both participants’ native language and from
some other, foreign language.
76
Experiment 5
Findings from Experiment 4 demonstrate different patterns of results for
native English speakers conceptually categorizing idioms from their own language
than when they categorize idioms that have been directly translated to English from
another language. I designed Experiment 5 in order to verify whether these
differential response time patterns hold using a within-subjects design. Given that
response times from both the Latvian and Mandarin categorization studies
(Experiments 4B and 4C) followed the same pattern, I included idioms from only
one of the two foreign languages in Experiment 5. Since the Latvian idioms were,
on average, the same length as the English idioms while the Mandarin idioms
tended to be longer (when directly translated into English), I coupled the Latvian
idioms with the English idioms for a two-language, within-subject design.

Norming Study 5A: Identification of English and Latvian idioms
The first half on Experiment 5 was designed with the goal of determining
which, if any, stimuli from Experiments 4A and 4B were problematic (e.g.,
English idioms not recognized as English; non-English idioms misrecognized as

English). Some of the English idioms used in the previous idiom categorization
experiments may not have been sufficiently familiar to justify their being used as
one in a set of "familiar" idioms. Alternatively, certain of the Latvian idioms, when
translated, may have seemed like English idioms to native English speakers.
Although both sets of stimuli chosen for use in Experiments 4A and 4B were
legitimate idioms from English and Latvian respectively, Experiment 5 was

designed with an eye towards controlling for source-language ambiguities between
the two sets of idioms. I tested the stimuli on a pool of native English speakers to
77
counter any biases that the language-sensitive coders (for the English idioms) or the
bilingual translators (for the Latvian idioms) may have introduced. That is, those
individuals may have been more familiar with idioms in general than the
undergraduates actually completing the conceptual categorization task.
Method
Participants. A group of 35 undergraduate students (20 females and 15
males) of the State University of New York at Stony Brook received research credit
for their participation in the norming study. All were native speakers of American
English and none reported that they were functionally fluent in another language.
Another 30 undergraduates (18 females and 12 males) received research credit for
their participation in a follow-up norming study. Again, all were native speakers of
language.
Materials. A questionnaire was designed that included both sets of the 75

English and 75 Latvian idioms that had been used in Experiments 4A and 4B. The
idioms were arranged such that each page of the questionnaire contained idioms
from both languages that represented one of the five concepts (e.g., anger, secrecy,
revelation, insanity, and control). The idioms within each concept group were
further divided into three groups based on whether they had been determined to be
normally analyzable, abnormally analyzable, or unanalyzable. Although the three

groups of idioms for each of the concepts were arranged in separate columns, they
were not labeled (e.g., as normally, abnormally, or unanalyzable) so as to avoid
biasing participants about the semantic structure of each idiom. Within each type
and concept group, the English and Latvian idioms were randomly ordered. That
78
is, each page consisted of idioms from both English and Latvian representing one
concept, such that there were 30 idioms per page. Within each concept group the
two languages' idioms were further divided into subgroups representing idiom
type, such that each subgroup contained 10 idioms (5 English and 5 Latvian). Next
to each idiomatic phrase were two blank spaces with headers containing the
questions: "Is this an English idiom?” and "How confident are you that it is or is
not an English idiom?" The pages of the questionnaire were randomly arranged for
each different participant.
Procedure. Each participant completed the questionnaire individually.
Participants were instructed to indicate whether or not a phrase was an English

idiom by writing either "Y" (for ‘yes’) or "N" (for ‘no’) in the first column of blank
spaces for each separate phrase. Then, once they had done that, they were
instructed to rate how confident they were in their decision, indicating "very
positive" with a 1 and "not very positive" with a 5. They were encouraged to use
the entire scale (between 1 and 5) when making these confidence ratings.
The instructions also explained that within the three groups of idioms on
each separate page there were roughly equal numbers of English idioms and idioms
that had been translated from another language. This was to motivate participants to
really think about whether or not they had previously heard each phrase used in
English, rather than to simply judge all of the transparent idioms as being English
(since these idioms make the most intuitive sense since they are completely
transparent) and all of the opaque idioms as being from another language (since
these idioms don’t necessarily make sense on the surface and so may be considered
foreign, even though familiar).
79
Results
Distributions were made of the mean percentage of native English speakers
who judged the English and the Latvian idioms as being English and non-English
phrases respectively. All those English idioms that were judged to be familiar
English phrases by fewer than 50% of the native English speakers were removed
from the pool of 75 English idiom stimuli and replaced with the next available
idioms from the original rating study that corresponded for both conceptual group
and idiom type (as determined by that phrase's rating as either normally,
abnormally, or unanalyzable). More than half of the participants judged five of the
original 75 English idioms to be non-English phrases. More than half judged six of
the original 75 Latvian idioms to be English phrases. These 11 problematic phrases

were replaced with idioms from the original rating procedure (the first half of both
Experiments 4A and 4B). Replacement phrases were judged (again in the manner
described above) by a different group of 30 native English speakers who positively

identified the five replacement English phrases as being English and the six
replacement Latvian phrase as being from some other language than English.
Experiment 5B: Native English speakers’ conceptual categorization

of English and Latvian idioms
Method
Just as in Experiments 4A, 4B, and 4C, Experiment 5B measured the speed
and accuracy with which native English speakers were able to categorize groups of
the three types of idioms from both English and Latvian. Native English speakers
participated in the timed categorization of the 75 English and the 75 Latvian idioms.
80
Participants. A different group of 34 undergraduate students (18 males and

16 females) of the State University of New York at Stony Brook received research
language.
Materials. Materials for this study were the 75 English and 75 Latvian
idiomatic phrases that had been selected and translated in Norming Study 5A (see
Appendix J for a full listing of these idioms).
Procedure. The 150 English and Latvian (directly translated into English)
idiomatic phrases were presented on a Macintosh computer, using the SuperLab™
program. Each person completed the experiment individually. The idiomatic
phrases were presented on the screen one at a time, and in a different random order
for each participant The computer keyboard was covered such that only 5 keys
(the five keys located directly above the space bar) were exposed. These were
labeled with capital letters (R, I, C, A, or S) representing one of each of the five
categories (revelation, insanity, control, anger, or secretiveness) and the full name
of each category was located just above the appropriate key. Participants were
asked to put each idiom into the appropriate concept group based on its figurative
meaning by pressing the key that corresponded with that concept Prior to
beginning the experiment participants read instructions explaining the procedure
(see Appendix J).
Prior to beginning the experiment participants completed a brief practice

session that included ten idioms (five English and five Latvian-all different from
those used in the experimental trial), with one of each of the five from each
81
language representing each of the five concept categories. If participants had no
questions, they were told to begin the experiment by pressing the space bar. Each
subsequent idiom appeared immediately following the previous idiom being
categorized via a key press for one of the five concepts. Participants took between
20 and 25 minutes to complete the entire experiment
Results
categorization decisions and the proportion of English and Latvian idiom types
(e.g., normally, abnormally, or unanalyzable) that they correctly categorized are
presented in Table 8. Response times from incorrectly categorized phrases were
eliminated from the analysis, as were response times greater than 10 s
(approximately 2 SDs from the mean).
Error rates for conceptual categorization of English and Latvian idioms are
similar to those seen for conceptual categorization of English or Latvian idioms

independently. For the English idioms, native English speakers again correctly
categorized normally analyzable idioms the most often (88.9%), followed by
abnormally analyzable idioms (75.7%), and unanalyzable idioms (68.6%). For the
Latvian idioms, native English speakers also most often correctly categorized
normally analyzable idioms, but significantly less often than for English idioms
(70.3%), followed by abnormally analyzable idioms (64.3%), and unanalyzable

idioms (41.1%). It should be noted that this group of native English-speaking
participants showed a slight reduction in their conceptual categorization
performance from the participants who categorized only Latvian idioms in
Experiment 4B. This may be a product of their having the more familiar English
82
idioms available for comparison, allowing them to feel less concerned about their
performance on the foreign language idioms. Regardless, the native English
speakers in Experiment 5 followed the same pattern of correct categorization as
those seen in both Experiment 4A and 4B, replicating both of those studies’ results
for categorization accuracy. Again, the most noticeable drop in participants’ correct
conceptual categorization of both English and Latvian idioms is in the unanalyzable
group.
As for the reaction times, the same differences in for the three idiom types
shown in Experiments 4A and 4B also were shown in this study. An analysis of

variance was performed with Idiom Type (normally analyzable vs. abnormally
analyzable vs. unanalyzable) and Language (English vs. Latvian) as within-subject
factors. These analyses revealed significant main effects for both Language (F l(l,
68) = 130.65, p < .001; F 2(l, 146)6 = 26.90, p < .001) and Idiom Type, Fl(2,
68) = 53.05, p < .001; F2(2, 145) = 9.45, p < .01), consistent with findings in
Experiments 4A and 4B. That is, native English speakers categorize idioms from
their own language much faster overall than they do idioms directly translated from
another language. These categorization times are further influenced by whether or
not the idioms are normally, abnormally, or un analyzable.
In addition, a significant interaction between Language and Idiom Type was
found (F I(2, 68) = 23.76, p < .001; F2(2, 142) = 3.71, p < .03), also consistent
with findings from Experiments 4A and 4B. That is, where English speakers
conceptually categorized abnormally analyzable idioms from their native language
6 The degrees of freedom here reflect the fact that certain items (two. to be exact)
were not categorized correctly by any of the participants.
83
the fastest (2665 ms, as compared to 3168 ms for normally analyzable idioms), it
takes them longer (3834 ms) to categorize abnormally analyzable idioms directly
translated from Latvian than normally analyzable idioms directly translated from that
language (3430 ms).
Discussion
This examination of native English speakers’ categorization of their own
and another language’s idioms according to the concepts they represent highlights
processing differences between pre-existing metaphorical mappings and those that
are created on-line. That is, where native English speakers are able to rapidly
access the meanings for familiar abnormally analyzable idioms from their native
language (ostensibly because those mappings already exist), they must create the
mappings anew for the same types of idioms that have been directly translated from
another language. The use of other languages’ idioms highlights the difference in
accessing time for these two very different processes.

The current debate regarding idiom processing focuses on whether we have
pre-stored mappings of conceptual metaphors or instead create those mappings as

we hear and comprehend a phrase (e.g., Gibbs, Bogdanovich, Sykes, & Barr,
1997; Glucksberg, Keysar, & McGlone, 1992; Glucksberg, Brown, & McGlone,
1993). The studies presented here address this question by setting up a situation
where it is clear whether the idioms are known or not (that is, with idioms from
one’s own and another language). Such a manipulation establishes a clear baseline
for the accessing times of familiar and unfamiliar idioms, the differences of which
should demonstrate the time-course of making new mappings from source to target
domains. Since it is nearly impossible to control for the breadth of speakers’
84
idiomatic knowledge in their native language, examining their processing of another

language’s idioms provides an ideal way to differentiate old and new metaphorical
mappings.
85
General Discussion
The studies reported here provide the most comprehensive evidence to date
that the use and understanding of idioms is guided by conceptual metaphors that
may be universal. By examining idiomatic phrases from multiple languages
together with what both native and non-native speakers do when confronted with
these phrases, I have shown that many idioms are, in fact, analyzable. From this, it
follows that the conceptual metaphors motivating the phrases are universal.
Experiment 1 compared native English and native Latvian speakers’ images
for American English idioms, examining the degree to which both sets of images
reflected underlying conceptual metaphors for the phrases. Given the same set of
idioms, native Latvian speakers produced mental images that were, on average, no
different from those produced by native English speakers. This was despite the fact
that the Latvian speakers had learned the figurative meanings of the phrases only
weeks prior to the mental imaging session, while the English speakers had used the
phrases across the course of their lifetimes as native English speakers.
I claim that the similarities between the Latvian and English speakers’
images were not due to prior knowledge of the phrases’ figurative meanings.
Rather, these similarities are more likely due to both groups of speakers’ analyses
of the relationship between the phrases’ surface structures and their figurative
meanings. Gibbs and O’Brien’s (1990) study found that when native English
speakers produced images for just the figurative paraphrases of idioms (e.g., ‘to get
angry’ for blow your stack), their mental images did not match those produced for
the idiomatic phrases themselves. That is, when asked to form an image based on
86
the figurative paraphrase of an anger idiom, they did not include any references to
heated fluid in a container in their descriptions. This counterintuitive finding helped

guide interpretation of the present data on native Latvian speakers’ images for
idioms.
Experiment 3 provided even stronger support for this claim by comparing
native English and native Mandarin speakers’ images for idioms both before and
after they were explicitly made aware of the phrases’ figurative meanings. Gibbs
and O’Brien’s (1990) findings suggest that knowledge of phrases’ figurative
meanings does not influence peoples’ mental images for those phrases. Experiment
3 found that both native and non-native speakers produced roughly the same mental
images during the two sequential imaging tasks for the same idiomatic phrases.
Although the native speakers in Experiment 3 produced more images and responses
to probe questions that matched native speakers’ image schemas (which was likely
due to practice with the task in Experiment 1), there was no difference between
native and non-native speakers’ images in Experiment 3.

The second series of experiments (Experiments 4A, 4B, 4C, and 5)
provides reaction time evidence about whether or not and how easily speakers can
analyze idioms from a foreign language in order to deduce their figurative
meanings. In a forced choice task, native English speakers categorized idioms from
both their own and two other languages (the phrases were literally translated from
Latvian and Mandarin) according to the phrases’ figurative meanings. Non-native

speakers (in this case, native English speakers) were able to correctly categorize
normally analyzable idioms from the two other languages significantly more often
and faster than they could correctly categorize abnormally analyzable idioms from
87
those languages. In turn, they correctly categorized the foreign languages’

abnormally analyzable idioms more often and faster than they did the unanalyzable
idioms.
The difference in response times for categorizing one’s own and another
language’s idioms suggests a difference in the processing required. Furthermore,

differences in the speed and accuracy with which speakers were able to categorize
the different idioms from either their own or another language reflect the degree to
which people can make sense of figurative language by analyzing it for underlying
metaphorical structures. It is particularly noteworthy that while normally analyzable
idioms from a foreign language were faster to process than abnormally analyzable
idioms from the same language, these two types of idioms were equally fast to
process in one’s own language. This response time difference supports arguments
for on-line mapping in addition to arguments for pre-stored structures. Overall, it
was most difficult for native English speakers to correctly categorize unanalyzable
English idioms and this took significantly longer for them to do. Data from these
studies provide compelling support for the view that many of these metaphorical
structures are universal. This would explain how native English speakers could
correctly categorize idioms directly translated from both Latvian (another Indo-
European language) and Mandarin (a non Indo-European language).
Analyzability itself can be conceptualized as ranging along a continuum,

with complete transparency anchoring one end of the continuum and complete
opacity anchoring the other. The present examination of native English speakers'
categorization of their own and other languages’ idioms according to their
underlying concepts clearly supports this view. That is, idioms range along a
88
continuum of analyzability (and thus, universality), whereby native and non-native

speakers process their own and other languages’ idioms from different parts of this
continuum differently.
Analyzable idioms from another language describing some concept (e.g.,
anger) were identified as such by native English speakers. However, depending on
whether that idiom was considered normally or abnormally analyzable, native

English speakers processed it more or less quickly. This points to the
nonindependence of analyzability and universality. When phrases were not
correctly categorized according to their figurative meanings by non-native speakers,
this suggested that the cultural flavoring of the language from which the phrases
were translated (or the fact that the idioms were somehow historically-based) made
the meanings too opaque for non-native speakers to identify.

Results from Experiment 2 point toward applications of the cross-linguistic
imaging task. This follow-up study was intended to measure the degree to which a
structural analysis of idiomatic phrases can help guide speakers (particularly non
native speakers) into using idioms more naturally. The Latvian speakers from
Experiment 1 were found to use the 25 test idioms more naturally following
completion of the mental imaging task than they did prior to the task. Of course
these results may be attributable to the Latvian speakers having more practice in
using the idioms. However, given that the Latvian speakers were well aware of the
English idioms’ figurative meanings at the time they produced the first set of
sentences, it seems clear that simple knowledge of idiomatic phrases’ figurative
meanings is not enough to allow one to use those phrases to their full figurative
potential. That is, if it is true that idioms are based on underlying conceptual
89
metaphors (and results from Experiment 1 suggest this to be the case), then an
analysis of the mappings between literal and figurative meanings should help one
use the phrases more naturally (e.g., in a manner that exemplifies the metaphorical
relationship between a literal phrase and its figurative interpretation). Findings
from Experiment 2 lay the groundwork for what could be a useful application in
second language instruction.
A more difficult question to answer is whether or not we access conceptual

metaphors (as Gibbs and his colleagues suggest) or compute the concept on-line (as
Glucksberg and his colleagues suggest) when processing figurative language.
Based on the results from the two series of studies reported here, I feel that both of
these positions are 1) too extreme and 2) focus on a tangential aspect of the more
interesting question. I will address these two points in turn.
As for the Glucksberg/Gibbs debate, I conclude that these extreme positions

are too extreme. This is not to say that neither position is correct In fact, I think
these positions are strongly held precisely because they are both correct, depending
on the situation. In support of Glucksberg’s view, the data I have reported here
indicate that people can and do create metaphorical mappings on-line. That non
native speakers are able to read (or hear) an idiomatic phrase out of context for the
first time (or in this case, an idiom translated literally from another language) and
correctly determine what it means when used figuratively indicates that we can and
do create metaphorical mappings on-line. On the other hand, in support of Gibbs’s
position, the fact that we consistently process normally analyzable idioms from our
own language no more quickly than abnormally analyzable idioms supports the
(contrary) view that these mappings are pre-existing and are applied automatically.
90
As for my second point, I regard both views as tangential to the bigger

picture. Both sides of the debate have unfortunately used methodology that
prevents them from answering questions about universality. Figurative language is
fascinating precisely because it is so prevalent and in such diverse forms across
languages. The processing debate has resulted in disregard for the qualities that
make figurative language so fascinating. That is, the important question is one of
universality: how is it that people who speak languages that are (quite literally) a
world apart can relatively easily map the same metaphor onto the underlying
concept?
The fact that well-translated writings from one language can be understood
and appreciated by speakers of another language attests to the importance of this
question. The structural metaphors tapped by a phrase must be universal in order
for the concept it represents to be understood by speakers of other languages. Our
ability to appreciate literature that is translated from around the world should help
guide us in our pursuit of the answer to this question, as indicated by data from the
cross-linguistic studies reported here. Of course, the art of translation depends on
the translator's ability to lexically structure metaphorical mappings from one
language in an acceptable (and understandable) form in another language. But truly
talented translators work to preserve the source language's original structural
representations (e.g., turns of phrase) so that readers of the same work in the target
language get to experience this other culture's manner of lexicalizing a familiar
concept It is this opportunity (that is, the opportunity to experience new

lexicalizations of old concepts) that makes foreign literature so delicious.
91
Regarding the interrelatedness of analyzability and universality, there is

much room for fruitful research. For example, an examination of cross-linguistic
confusions of figurative speech will prove informative regarding the degree to
which semantic fields are an appropriate representation of structural bodies of

meaning. If native speakers of a variety of languages are consistently confused by
the same concept groups across their own and other languages, this will add
support to the view that figurative speech taps into fields of meaning that overlap in
particular linguistic (or meaning) domains.
The very mention of "fields of meaning" begs the question of the nature of
metaphorical structures. Although I have consistently referred to LakofFs (1987)

characterization of conceptual metaphors as bold, linguistically represented thoughts
(e.g., ANGER IS HEATED FLUID IN A CONTAINER), I do not agree with
Lakoff that we necessarily carry such concepts around in our heads represented
linguistically. Rather, it seems more effective to think of conceptual metaphors as
structural bodies of meaning that are sensory in nature (e.g., visual, verbal, tactile).
That is to say, meaning takes on many forms which need not be linguistic.
Future work that will further characterize the structure of figurative speech
in general, and conceptual metaphors in particular, includes: additional
categorization studies where concept groups that are not represented by test idioms
(e.g., happiness, disgust) are added as possible responses, eliminating the degree
to which pariticipants’ responses are constrained and “cross-linguistic”

examinations of the same language as used by different communities of speakers
(e.g., American vs. British vs. Australian English), eliminating some of the risks
inherent in literally translating another language’s idioms. Such pursuits will help
92
flesh out the picture that is emerging from cross-linguistic research (e.g., the
studies reported here) regarding the nature of figurative language and the
universality of the metaphorical structures that such language represents.
93
References
Arlotto, A. (1972). Introduction to historical linguistics. Lanham, MD:

Univeristy Press of America.
Au,T. (1983). Chinese and English counterfactuals: The Sapir-Whorf hypothesis

revisited. Cognition, 15, 155-187.
Au, T. (1984). Counterfactuals: In reply to Alfred Bloom. Cognition, 17, 289-
302.
Boatner, M., Gates, J., & Makkai, A. (1975). A dictionary o f American idioms.
Woodbury, NY: Barron's Educational Series, Inc.
Bobrow, S., & Bell, S. (1973). On catching on to idiomatic expressions.

Memory and Cognition, I, 343-346.
Bynon, T. (1977). Historical Linguistics. Cambridge: Cambridge University
Press.
Cacciari, C., & Glucksberg, S. (1995). Understanding idioms: Do visual images
reflect figurative meanings? European Journal o f Cognitive Psychology, 7,
283-305.
Cacciari, C., & Tabossi, P. (1988). The comprehension of idioms. Journal o f
Memory and Language, 27, 668-683.
Cooper, D. (1986). Metaphor. London, England: Blackwell Publisher.
Cruse, D. (1986). Lexical semantics. Cambridge: Cambridge University Press.

Fraser, B. (1970). Idioms within a transformational grammar. Foundations o f
Language, 6, 22-42.
94
Gibbs, R. (1980). Spilling the beans on understanding and memory for idioms in
conversation. Memory and Cognition, 8, 449-456.

_______ . (1985). On the process of understanding idioms. Journal o f
Psycholinguistic Research, 14, 465-472.
_______ . (1986). Skating on thin ice: Literal meaning and understanding idioms
in conversation. Discourse Processes, 9, 17-30.
_______ . (1990). Psycholinguistic studies on the conceptual basis of idiomaticity.

Cognitive Linguistics, 1, 417-451.
_______ . (1992). What do idioms really mean? Journal of Memory and
Language, 31, 485-506.
_______ . (1993). Why idioms are not dead metaphors. In C. Cacciari & P.
Tabossi (Eds.), Idioms: Processing, structure and interpretation (pp. 57-
78). Hillsdale, NJ: Erlbaum and Associates.

_______ . (1994). The poetics o f mind. New York, NY: Cambridge University
Press.
Gibbs, R., & Gonzales, G. (1985). Syntactic frozenness in processing and
remembering idioms. Cognition, 20, 243-259.
Gibbs, R., & Nayak, N. (1989). Psycholinguistic studies on the syntactic
behavior of idioms. Cognitive Psychology, 21, 100-138.
_______ . (1991). Why idioms mean what they do. Journal o f Experimental
Psychology: General, 120, 93-95.
Gibbs, R., Nayak, N., Bolton, J., & Keppel, M. (1989). Speakers’ assumptions
about the lexical flexibility of idioms. Memory & Cognition, 17, 58-68.
95
Gibbs, R., Nayak, N„ & Cutting, C. (1989) How to kick the bucket and not
decompose: Analyzability and idiom processing. Journal o f Memory and
Language, 28, 576-593.

Gibbs, R., & O’Brien, J. (1990). Idioms and mental imagery: The metaphorical
motivation for idiomatic meaning. Cognition, 36, 35-38.
Gibbs, R., Bogdanovich, J., Sykes, J., & Barr, D. (1997). Metaphor in idiom
comprehension. Journal o f Memory and Language, 37, 141-154.
Glucksberg, S., Brown, M., & McGlone, M. (1993). Conceptual metaphors are
not accessed during idiom comprehension. Memory & Cognition, 21, 711-
719.
Glucksberg, S., Gildea, P., & Bookin, H. (1982). On understanding non-literal
speech: Can people ignore metaphors? Journal of Verbal Learning and

Verbal Behavior, 21, 85-98.
Glucksberg, S., & Keysar, B. (1990). Understanding metaphorical comparisons:
Beyond similarity. Psychological Review, 97, 3-18.

Glucksberg, S., McGlone, M., & Keysar, B. (1992). Metaphor understanding
and accessing conceptual schema: Reply to Gibbs. Psychological Review,
99, 578-581.
Katz, J., & Postal, P. (1964). An integrated theory o f linguistic descriptions.

Cambridge, MA: MIT Press.
Keysar, B., & Bly, B. (1995). Intuitions of the transparency of idioms: Can one
keep a secret by spilling the beans? Journal o f Memory and Language, 34,
89-109.
96
Kittay, E. (1987). Metaphor: Its cognitive force and linguistic structure. Oxford,
England: Clarendon Press.
Kittay, E. (1996). O f "men" and metaphors: Shakespeare, embodiment, and
filing cabinets. In T. Ward & S. Smith (Eds.), Creativity and concept.

American Philosophical Association.
Kittay, E., & Lehrer, A. (1981). Semantic fields and the structure of metaphor.
Studies in Language, 5, 31-63.
Kovecses, Z. (1986). Metaphors o f anger, pride, and love. Amsterdam: John
Benjamins Publishing.
Lakoff, G. (1987). Women, fire, and dangerous things: What categories reveal
about the mind. Chicago: University of Chicago Press.
Long, T., & Summers, D. 91979). Longman dictionary o f English idioms.
London: Longman.
Makkai, A., Boatner, M„ & Gates, J. (1995). A dictionary o f American idioms.
Hauppauge, NY: Barron's Educational Series, Inc.
Meillet, A. (1937). Introduction a I’etude comparative des langues indo-

europeennes. Paris: Hachette (reprinted by University of Alabama Press in
1964 edition).
Miller, G., & Johnson-Laird, P. (1976). Language and perception. Cambridge,

MA: Harvard University Press.
Nayak, N., & Gibbs, R. (1990). Conceptual knowledge in the interpretation of

idioms. Journal o f Experimental Psychology: General, 119, 115-130.
Nunberg, G. (1978). The pragmatics o f reference. Indiana University Linguistics
Club.
97
Nunberg, G., Sag, I., & Wasow, T. (1994). Idioms. Language, 70, 491-538.
Reddy, M. (1979). The conduit metaphor: A case of frame conflict in our
language about language. In A. Ortony (Ed.), Metaphor and thought (pp.
284-324). New York: Cambridge University Press.
Swinney, D., & Cutler, A. (1979). The access and processing of idiomatic
expressions. Journal o f Verbal Learning and Verbal Behavior, 18, 523-
534.
Weinreich, U. (1969). Problems in the analysis of idioms. In J. Puhvel (Ed.),
Substance and structure o f language. Berkeley, CA: University of
California.
98
Table 1 (adapted from Gibbs & O’Brien, 1990)
Idioms Originally Selected by Gibbs and O’Brien (1990) from Boatner. Gates, and
Makkai (1975) and Long and Summers ( 1979t and Relevant to Experiments 1. 2
and 3. All Idioms are either Analyzable or Abnormally Analyzable.
Anger
blow your stack
hit the ceiling
lose your cool
foam at the mouth
flip your lid
Control
crack the whip
lay down the law
call the shots
wear the pants
keep the ball rolling
Secretiveness
keep it under your hat
button your lips
hold your tongue
behind one's back
keep in the dark
Insanity
go off your rocker
lose your marbles
go to pieces
lose your grip
bounce off the walls
Revelation
spill the beans
let the cat out of the bag
blow the whistle
blow the lid off
loose lips
99
Table 2
Three Different Native Speakers’ Responses to Questions Regarding their Image
for the Phrase Go Offvour Rocker fin Experiment 1).
Literal
General Image: Getting off of my rocking chair to get food
Causation: hunger
Intentionality: yes, it was intentional
Manner casually standing up
Consequence: I get food
Negative Consq. more hunger
Reversibility: it’s not difficult to reverse, just sit back down
Figurative
General Image: The old man loses his mind
Causation: he’s getting too old
Intentionality: no
Manner: dotty (effect of old age)
Consequence: he gets put away
Negative Consq. he’d be normal
Reversibility: difficult too, because it happens sooner or later
Hybrid
General Image: An old woman is upset about something and loses her grip
on the chair she’s sitting in and falls off
Causation: loss of balance due to her agitation

Intentionality: no
Manner: surprisingly
Consequence: breaks her hip
Negative Consq. healthy hip
Reversibility: yes, it's hard to stop from happening once it’s happening
Question____________M ost frequent response____________________Native Non-native
Anger
General Image: Pressure forces some container to release its
contents violently upwards and outwards .76 .83
Causation: Stress, anger, and/or frustration .84 .86
Intentionality: Action is not intentional .83 .89
Manner: Forcefully, violently, uncontrollably .83 .85
Consequences: Pressure is released or relieved .69 .73
Negative Consq.: Container remains intact .53 .58
Reversibility: Action is difficult if not impossible to reverse .56 .70
Table 3
Schemas Derived from Native English Speakers’ most Frequent Eesponses for English Idioms
(Based.on Format Followed by.Gibbs.and O ’Brien, M Q) and.Proportions of Responses that Matched
them (Experiment 1).
100
Question Most frequent response J M y£ Non-native
Control
General Image: Action is taken by someone to impose order or
change someone else’s behavior .72 .64
Causation: Exertion of authority (control) over others .79 .61
Intentionality: Action is intentional .91 .89
Manner: Determinedly, forcefully .80 .78
Consequences: Some goal is accomplished, either positive or
negative depending on the agent .82 .77
Negative Conscq.: That goal is not accomplished .78 .73
Reversibility: Action is difficult to reverse .54 .51
Secrecy
General Image: Something is contained within something .89 .87
Causation Desire to have that thing not seen .93 .90
Manner: The thing remains contained .90 .89
Consequences: Object is kept hidden and unseen .95 .94
Negative Consq.: Object is seen and is general knowledge .91 .89
Reversibility: Action is quite simple to reverse .85 .79
Table 3 (contd.)
Ouestion Most frequent response Native Non-native

Insanity
General Image: A situation where a person or thing loses
physical/mental stability .74 .72
Causation Some internal or external stress .71 .69
Intentionality: Action is not intentional .92 .88
Manner: Control is lost at once and forcefully .69 .66
Consequences: Stability (usually physical) is lost .75 .73
Negative Consq.: Stability (usually mental) is maintained .66 .69
Reversibility: Action is extremely difficult to reverse .84 .85
Revelation
General Image: Something is tipped over or opened up such
that w hat’s in it is apparent to all .61 .58
Causation Extreme curiosity or stress/pressure .69 .65
Manner: Hastily, forcefully .72 .75
Consequences: Object or information is out/revealed .88 .90
Negative Consq: Object or information remains concealed .85 .86
Reversibility: Action is impossible to reverse .87 .75
Tabic 3 (contd.)
102
Concept Group Natives Non-natives

Session: / 11 / III
Image Tvpc
Anger Hybrid: 56.1 65.2 54.2 65.8
Literal: 40.0 29.0 32.9 12.3
Figurative: 3.9 5.8 12.9 21.9
Control Hybrid: 76.1 74.8 57.4 73.4
Literal: 22.6 21.9 37.4 18.8
Figurative: 1.3 3.2 5.2 7,8
Secrecy Hybrid: 75.9 74.2 56.8 60.0
Literal: 23.2 21.3 26.5 14.2
Figurative: 0.6 4.5 16.8 25.8
Insanity Hybrid: 72.3 64.5 37.0 41.3
Literal: 25.8 30.3 51.9 36.1
Figurative: 1.9 5.2 11.0 22,6
Revelation Hybrid: 65.8 58.7 27.7 27.1
Literal: 32.3 36.1 66.5 44.5
Figurative: 1.9 5.2 5.8 28.4
Table 4
Types o f Images that Speakers Reported during First and Second Mental Imaging Sessions of
Experiment 3. Including Images for Phrases that were Defined Correctly and Incorrectly on Pre-Test.
o
u>
Question____________M ost frequent image/response_________________ Native___________ Non-native
Session: I II___________/ II
Anger
General Image: Pressure forces some container to release its
contents violently upwards and outwards 73.9 80.2 66.5 74.6
Causation: Stress, anger, and/or frustration 71.0 81.3 68.4 85.8
Intentionality: Action is not intentional 61.3 63.9 53.5 56.8
Manner: Forcefully, violently, uncontrollably 78.7 82.6 71.0 76.8
Consequences: Pressure is released or relieved 85.8 93.5 84.5 90.3
Negative Consq.: Container remains intact 74.2 80.0 60.0 69.0
Reversibility: Action is difficult if not impossible to reverse 72.3 80.0 61.3 69.0
Table 5
Proportions ofResponses-GiveiLby Nativg-English speakers_and Maiive..Mandarin Speakers
for English Idioms (during M ental Imaging Sessions I and II in Experiment 3) that M atched the Most
Frequently Reported Image Schemas Described by Native English Speakers (Established in Experiment 1),
Proportions are Presented by Concept and arc Based on Hybrid Images Only.
10 4
Oucstion Most frequent image/response Native Non-native

Session: / 11 / II
Control
General Image: Order is imposed in some way by someone 77.4 76.2 73.6 78.5
Causation: Exertion of authority (control) over others 73.5 71.0 60.0 78.1
Intentionality: Action is intentional 91.6 95.5 91.5 91.6
Manner: Determinedly, forcefully 63.9 66.5 63.9 64.5
Consequences: Some goal is accomplished, either positive or
negative depending on the agent 89.0 86.5 89.7 92.3
Negative Conseq.: That goal is not accomplished 88.4 87.1 89.7 89.0
Reversibility: Action is difficult to reverse 58.1 50.3 46.5 55.5
Secrecy
General Image: Something is contained w/in something 80.1 80.8 74.8 75.5
Causation Desire to have that thing not seen 74.8 76.1 66.5 70.3
Manner: The thing remains contained 82.6 79.4 73.5 67.7
Consequences: Object is kept hidden and unseen 82.6 83.9 81.9 81.3
Negative Consq.: Object is seen and is general knowledge 82.3 82.6 70.3 71.0
Reversibility: Action is quite simple to reverse 67.7 71.0 70.3 72.9
Table 5 (contd.)
105
Ouestion Most frequent image/response Native Non-native

Session: .../ II I II
Insanity
General Image: A situation where a person or thing
loses physical/mental stability 76.5 77.3 77.1 76.9
Causation Some internal or external stress 69.0 63.2 70.8 69.9
Intentionality: Action is not intentional 67.7 71.0 69.5 72.4
Manner: Control is lost at once and forcefully 70.1 70.3 65.6 60.9
Consequences: Stability (usually physical) is lost 85.8 87.7 90.3 93.6
Negative Consq.: Stability (usually mental) is maintained 83.9 86.5 87.7 87.8
Reversibility: Action is extremely difficult to reverse 82.6 85.2 78.6 76.9
Revelation
General Image: Something is tipped over or opened up such
that what’s in it is apparent to all 72.7 69.5 70.8 73.0
Causation Extreme curiosity or stress/pressure 63.9 60.6 69.7 74.8
Manner: Hastily, forcefully 62.6 55.8 68.4 58.7
Consequences: Object or information is out/revealed 86.5 84.6 92.9 86.5
Negative Consq: Object or information remains concealed 87.1 84.5 85.2 80.6
Reversibility: Action is impossible to reverse 78.1 77.4 40.6 67.7
Table 5 (contd.)
106
Concept Group_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Natives_________ Non-natives

Test: Pre- Post-_______ Pro- Pnst-
Response
Anger Correct: 74.8 91.0 9.7 31.0

Incorrect: 7.1 6.5 10.3 58.1
Slang: 11.0 2.6 — --
Control Correct: 75.5 87.7 5.2 49.0

Incorrect: 5.2 1.9 22.6 36.8
Slang: 16.1 10.3 — —
Secrecy Correct: 91.6 97.4 45.2 82.6

Incorrect: 2.6 0.6 12.9 13.5
Slang: 3.2 1.9 — --
Insanity Correct: 76.8 90.3 5.8 41.0

Incorrect: 5.2 1.9 11.7 51.3
Slang: 15.5 7.7 — —
Revelation Correct: 81.3 96.1 12.3 45.2
Incorrect: 6.5 1.9 16.1 45.2
Slang: 8.4 1.9 — —
Table 6
Responses Speakers Gave for Idioms’ Figurative Meanings During Pre- and Post-Tests
(Prior to and following Mental Imaging Session I in Experiment 3).
107
108
Table 7
Native English Speakers' Response Times for Correct Conceptual Categorization of
Three Idiom Types for the Three Languages Tested in Experiments 4A. 4B. and
4C. Overall Percentage Correct is Listed Below the Corresponding Response
Time.
Source language for idiomatic stimuli
Idiom tvpe English (4A) Latvian f4B') Mandarin
Normally Analyzable 3720 3791 4018

(89%) (79%) (82%)
Abnormally Analyzable 3388 4421 5123

(76%) (69%) (59%)
Unanalyzable 4326 5880 5788

(68%) (44%) (50%)
109
Table 8
Native English Speakers' Response Times for Correct Conceptual Categorization of
the Three Idiom Types from English and Latvian (in Experiment 51. Overall
Percentage Correct is Listed Below the Corresponding Response Time.
Source language for idiomatic stimuli

Idiom type____________________________ English______Latvian_____
Normally Analyzable 3168 3430
(89 %) (70%)
Abnormally Analyzable 2665 3834

(76%) (64%)
Unanalyzable 3211 4984

(69%) (41%)
110
Appendix A: Background summaries of the three languages
included in subsequent studies
Latvian
Latvian is a member of the Balto-Slavic branch of the Indo-European
language family. The Balto-Slavic branch is split into two subgroups: Baltic and
Slavic. Speakers of Slavic languages are found throughout Eastern Europe. Slavic
languages are further grouped into three main branches: Eastern, Western, and
Southern. The most prominent Slavic language is Russian. Russian, Ukrainian,
and White Russian make up the Eastern group. The Western group is represented
by Polish, Czech, and Slovak. Bulgarian and Serbo-Croat are members of the
Southern group (Bynon, 1977).
Three languages make up the Baltic group of the Balto-Slavic branch: Old
Prussian, Lithuanian, and Latvian. While Old Prussian is now extinct, Lithuanian
and Latvian are still spoken by along the southern coast of the Baltic Sea in the
(now independent) countries of Lithuania and Latvia, both former Soviet Republics
(Bynon, 1977).
Mandarin
In East Asia, the most prominent language family is Sino-Tibetan. This
group has two main branches, the Tibeto-Burman and the Sinitic. In the Tibeto-
Burman branch, the most prominent languages are Tibetan, dating back to the
seventh century, and Burmese. There are also several so-called minor languages
(e.g., Lolo and Kachin). The Sinitic group includes most of the Chinese languages
that, for political and cultural reasons, are often referred to as dialects (Arlotto,
1981).
Sinitic languages that are currently in use are typically grouped into six main
divisions. Cantonese is spoken in Canton and Hong Kong. The Min dialects are
spoken mostly in the Fukien province and include the Amoy dialect, as well as
Taiwanese. The Wu dialects have Shanghai's language as their most prominent
member. Mandarin dialects resemble one another relatively closely (that is, with
effort a speaker of one Mandarin dialect can make him- or herself understood to a
Ill
speaker of another dialect). The Mandarin languages are subdivided into three
principle groups: the Northern is typified by the language spoken in Beijing and is
the standard of all China; the Southwestern group includes the Szechudn dialect;
Southeastern Mandarin is represented by the language of Nanking (Arlotto, 1981).
The Beijing dialect is relevant to the studies presented here.
English
English is a member of the Germanic branch of the Indo-European language
family. The Germanic branch breaks down into three groups: East, North, and
West. English belongs to the Western division. East Germanic stems from Gothic,
a language that can be tracked to early Biblical translations. Modem North
Germanic languages are Danish, Norwegian, Swedish, and Icelandic (Bynon,
1977).
West Germanic languages are often separated into three major subgroups
(based on a scheme developed by Meillet, 1937). These are: High Germanic, Low
Germanic, and Anglo-Frisian. Many of the High German dialects have been
around since the seventh century A.D. For example, both Yiddish and Modem
Standard German are High German dialects. Dutch, Flemish, and Afrikaans (an
offshoot of Dutch that was spoken by descendants of Dutch farmers in South
Africa) are examples of Low German. English is part of the Anglo-Frisian branch.
Old English (also referred to as Anglo-Saxon) dates back to the seventh and eighth
centuries A.D.; the Middle English period runs from the twelfth to fifteenth
centuries; New English has continued from then to now as more or less the same
language. The other member of Anglo-Frisian (known simply as Frisian), is still
spoken in some of the northern regions of Holland (Bynon, 1977).
A p p e n d i x B: Mental Image Qu est i o n n a i r e ^
g - g - B s w t t S i t m s n f M g g ■s g i g x M n M H - g
g - • M M S M
H -g g g ?
g g g -& g g g g tiB 9 te iJ » M g i • a M g a g g g g g W B B
m g s ® ’
S f fr M 'r ’Z X T g g ‘
113
This questionnaire is meant only to measure how many of the 25

idiomatic phrases you already know (prior to participating in this
study). You will be given both the English version of the phrase and
its literal translation in Mandarin.
You are NOT to guess what the phrases' idiom atic m eanings
a re. Rather, you are to check off only those phrases' whose
idiom atic meanings you know, and to provide the idiomatic
definitions for only those idioms.
This is not in any way a test of your intellectual or

linguistic abilities. It is simply m eant to measure your
prior exposure to idiom atic English
Please give your definitions in M andarin.
1) go off your rocker
2) crack the whip
3) blow your stack
4) spill the beans
5) keep it under your hat
6) loose lips
7) flip your lid
8) bounce off the wall
9) keep in the dark
10) keep the ball rolling
11) lay down the law
12) let the cat out of the bag
13) behind one’s back
14) hit the ceiling
15) lose your marbles
16) wear the pants
17) foam at the mouth
18) lose your cool
19) blow your whistle
20) go to pieces
21) lose your grip
22) hold your tongue
23) call the shots
24) button your lips
25) blow the lid off
116
m w m T B & £ 6 sm &«
* fc r n w r n a -E M ^ m m m • ^ t « s f k - « n - m ^m sst • f w * g
jt$ D K ’ 5xf§”KICK THE BUCKET’ ’ - J ttl " 3
W M 'SEffiRB JlSJH 3
m m : £ P A * H -M * i^ ’ S § T £ - T 3
s® : •
: fl&affiJtfeHTiPffiWfff- • SfPrTOJffiS : 3
m m : ffiro ® S ^ ± S T -< H G a ® 3
t w .v m m i •
f ^ 'C ^ « r a ! ^ J K t S j S S ® ? ® a E J i l B a t ± ’ ( fBISn : HSAW «*#$M*JtgSia&)

3a s a fP s ® a tR « E » « ^ w H 5 a iij
°
3 5rT C i#l«fcW «g
117
You are being asked to fill out a questionnaire about your mental
images based on 25 idiomatic phrases from the English language.
You will be given both the English version of each phrase and its
literal (or direct) translation in Mandarin. Translations are provided
only so that you better understand precisely what the idiom means.
You are to form a mental image of that idiom based on its literal
meaning. That is, form a picture in your mind based on the phrase's
actual words (their direct, literal meanings).
For example, fo r the phrase "kick the bucket" (which, in English,

means "to die" when used figuratively), you might imagine a man in
jeans and cowboy boots walking in a desert and kicking a rusty tin
bucket that he encounters along the way).
Once you've formed your image, write a brief description of it in the

space next to the words "mental image", and then answer the 6
questions that follow, specifically regarding that literal image.
In keeping with our example, you might answer each question like:
causation: The man came upon a bucket and kicked it.
i n t e n t i o n a l i t y : He kicked it on purpose.
manner: He kicked the bucket casually, without real thought.
consequence: The bucket gets another dent in it.
negative
consequence: The bucket would remain in its original position.
reversibility: Once he's kicked it, he cannot reverse that action.
Your images should project what the idiom would look like if the
phrase could literally happen (e.g., if one’s lip could literally flip).
Given this, many of your images may have a cartoonish quality; they
may be of events that can not really happen in our world, but that
can be imagined just as they are imagined by cartoonists.
Provide an image for each phrase and write all your answers in
M an d arin . Please provide sufficient written detail that your images
are understandable. This is in no way a test of your language skills;
118
rather, it is an exercise meant to familiarize you with idiomatic

English.
11 9
1) go off your rocker (®£«±&T3fc )

a. f f f f s a s s ff ln f ra s ?
b.
c. affluw ft& sssftttt ?
d. afBa&f&Messffis?
t ®nawafln&f^ - # a ± < + e ?
f. asfE W ftisiaflib^?
2) crack the whip )

a.
b. a flftff ttirs ftig ?
c. ?
d a flK fm ts n a ffs ?
e. fijn a w a aitiff • # a ± f t * ?
f. ss f& n ^ 0 a ® « 3 ^ ?
3) blow your stack (HfcMfWflff
a. ftJ S « 3 3 Ifll& fc S £ ?
b afflibf&tiraffifls ?
c.
d. H M M m tsm & ts ?
e. •# a s ifte ?
f. SSO K tfclB B Sfll& fls?
4) spill the beans ( *£B7a-?- )

a. ft* « 3 S 2 ffftft3 t£ ?
b. aisi& fB tw sw B ?
c.
d. a « » fi5 W * g * * fte ?
e issnisw aai&fE • # a * f + e ?
f. S S & K i& S id flftf'F ?
5) keep it under your hat ( SfcSEffttf-JTS-r-T)
a. {+ m m m i® m m 3L ?
b a « » ffJiw * a w s?
c a fl» f^ * » e s B i» i?
d. a iM & fm e « £ fi« ?
e. •a a ^ ft* ?
f. a s m K iiia fltt^ ?
6) loose lips ( I8SW# )
a.
b afliw iFaw w w *?
121
c.
d.
e. • # a * { te ?
f. & £ fiU £ 4 k la K « l» f ^ ?
7) flip your lid ( )
a. fH W K E » B » < 1 « £ ?
b. stffliftfmrsKNS?
C. a ffl» f £ £ 2 * 3 H 8 W ?
d. a fflttfw rts * a fte ?
e. ?
f &£fiU£lfc[BH2ffM&fl£?
8) bounce ofT the wail ( t£Jii_hi5[I3JR )
a. f ttE w s B a f f lis ^ a ± ?
b. afltt^wx^Bis ?
c.
d. a flttfttttS M ftff?
e. flB o ts m fln ^ * ft* ?
f. g-gtm & isnzffim ft ?
122
9) keep in the dark (

a. ft*S*SS2<S»J^a*L?
b. a n t t f B H f s n |s ?
d. ?
e. mnisw&mmiv • » a * f r e ?
r ssmnnfciHKHs®^ ?
10) keep the ball roiling ( M&fg3)j)

a .
b. i2 < !# )f£ £ W S & ^ S ?
c. i3 <3 M t m Z J 2 ?E B m ?
d. ia is n j^ ^ is ^ s fte ?
c •sa * f+ e ?
f. ftsaiii&iaaffli&m?
11) lay down the law (flFFj£&)

a. f te s is d H B f t^ a s .?
b. a « » f l w s t t |s ?
c. i» ll& fW l» e * « W ?
d.
123
e. fiso&ffisffttre • # a ± f t e ?
f. a S tK B ttls ia fln fE ?
12) let the cat out of the bag (SSSSSftfS^f

a. ft-eswaamftfEa^?
b a flu ftfm rm a ?
c.
d. a ffl» f« « i» » a fte ?
e. !g*a&wa*Bi&fis • # a £ f + e ?
f. S S « li4 5 tI s ia ® « J f ^ ?
13) behind one’s back (SAfftlf

a.
b. aflMfPSirsw"*?
c. a ffl» ^ a » B ® « tt?
d. afflttf»9®R*ftff?
e. fian»«rafli&^ • *a£ff-e ?
f. f t 5 f R l i ‘&0i2{B*!jfF?
14) hit the ceiling (HLhATEfi)
a.
b. a « » ft« w » w * ?
C. 3 flttfH £ S £ £ £ & F ?
d.
e. m a im & ffiM t' ?
f. ssmftlsISSflKrf^?
15) lose your marbles (^zrT
a.
b. i2 ® l & ^ £ W : S c f f i DJ I ?
c. ii® m {?m .zJS5tsx& j?
d. s M iw w e * * ff« ?
e. fls s n fa w a a ftff • a a ^ f t * ?
f. jKauaWsiaffllM*?
16) wear the pants
a. i+m mmimvjitgtSL ?
b. a flw i^ ftifw w ?
c. a « IW H M H E 3a * W ?
d. a * iw w fls * * ft* ?
e . < s$ n « w a ® » f^ - * a ± f t * ?
r. a s m iS E g ia « i& ^ ?
17) foam at the mouth )
a. itmzm&mmjttstZE.?
b. a M b ftm x K n c ?
c. IS fflM H & S S & g ft ?
d.
e. fttm m iz m th ft • fra * ff-« ?
f s s m K tiiis flib f^ ?
18) lose your cool )
a. fHSSKBM M M ^a*?
b ?
c a ffl» m j6 » ^ s 2 w ?
d ?
e. s s n jf tw a ( H » iiF - s a ± { + e ?
f. £«§?&l&ft[I]i2fS*a^?
19) blow your whistle (SKP^
a. f + e ^ s d n s K f ^ a ^ .?
b. iM ftfB M fS W W ?
c. a iH » fB S » S * * W ?
d.
126
e. w gm fiB W E • ?
f. S S « ttliB I9 ftfi£ ?
20) go to pieces (#fi£5?#i)
a. f t * W 0S M t t f l 5« £ ?
b. airofftfewsfityg?
c. ?
a.
e fgto^w iifSK ^ • ?
f. Sfr{Sli4UsUSfflKi^?
21) lose your grip(^S W 5S t^

a.
b. iH M b fftS friE W B ?
c sm m ft& zsfE B m ?
± s»M&flstt«*ftfte?
e. ■#a*.{+e ?
f.
22) hold your tongue (tfJ^fftflUii'SS)
a.
127
b.
c.
d. a « iw m ie « * ft* ?
e . ffittoa'fisiflnbf^ • ?
f. «S«UE4ScSIjlfllM^?
23) call the shots )
a. ft« S I§ G H @ « J^ a ^ ?
b ft ?
d.
e • # a ± f+ e ?
24) button your lips (ffiflif'FBTfllW)
a. fi« l* S K J ffl» ^ a * ?
b. a n « rf^ * * iE tt« s ?
c. S ffllM W fc S * S a 8 » l?
d. a ffl* jft$ W S * S ft-e ?
e tfS in& irilfiftfF • * a £ f t f f ?
f ftS W & IIE lffllJ fls ?
25) blow the lid off (SSfBSX^r)
a. f H e s is s n if t^ a ^ ?
b.
c. affl!W ftfe« * 5 B S tt?
d.
e . ?
f S5fS*|J&0i2{S*!j^?
129
You again will be given both the English versions of the phrases and
their literal translations in Mandarin.
This time, you are to try to figure out what the phrases' idiomatic
meanings are, based on the words involved and their literal
meanings. That is, what do you imagine native English speakers
mean when they use the phrases idiomatically?
T h is is n o t in any way a test of your intellectual o r

lin g u istic a b ilitie s. It is sim ply m ean t to d e te rm in e how
t r a n s p a r e n t th e E nglish p h ra s e s ' id io m a tic m e a n in g s are.
Please give your definitions in Mandarin.
130
1) go off your rocker (®£W_h££T3?5 )
2) crack the whip (3Slfc?:}£ S 4 ,»W§?$fif'E# )
3) blow your stack (Pfr8Bgro g g y
4) spill the beans (jfc tJ T it? -)
5) keep it under your hat (
6) loose lips ( )
7) flip your lid ( Hfttf-lS-T-)
8) bounce off the wall ( Q£AU:3I[II3£ )
9) keep in the dark ( {g&HBgir)
10) keep the ball rolling ( )
11) lay down the law (HIT&S)
12) let the cat out of the bag ( W S S f 'J g ^
13) behind one’s back (?£AWU’tfc
14) hit the ceiling (SLh^TES)
15) lose your marbles
16) wear the pants (9 9 3 )
131
17) foam at the mouth (W^WtS )
18) lose your cool (&£&&]&&)
19) blow your whistle (Ufcnqt
20) go to pieces (»& FSW)
21) lose your grip
22) hold your tongue (^3£fttf5i§sg)
23) call the shots («<£#**)
24) button your lips (lO&flfcWS)
25) blow the lid off («8123)
132
Appendix C: Experiment 1A: Coder instructions
Some analyzable idioms have words whose meanings directly relate to

their figurative interpretations. For example, the phrase pop the question has
two words that are transparently related to their individual figurative meanings.
Thus, the word pop is transparently related to the idea of 'suddenly asking' or
'suddenly proposing' while the word question refers to a particular kind of
question, namely a 'marriage proposal.’ Idiomatic phrases like this would be
rated as on or near 1 along a continuum that runs between 1 (= transparent) and
7 (= opaque).
On the other hand, there are idioms that are not analyzable. These
idioms have words whose meanings do not relate (directly or metaphorically)
to the idioms' figurative interpretations. For example, the phrase break a leg
has two words that are only opaquely (or seemingly arbitrarily) related to their
individual figurative meanings. That is, the words break and leg are only
opaquely related to the idea of 'performing extremely well in some event.’
Idiomatic phrases like this would be rated as on or near 7 along the transparent-
opaque continuum mentioned above.
Finally, there are idioms that are analyzable to varying degrees, but
abnormally so, such that certain or all of their individual words have metaphorical
relations to their figurative meanings. For some of these idioms, the relations
between the phrases' words and their figurative referents are not entirely
transparent, nor are they entirely opaque. Rather, the words can be understood as
being metaphorically related to their figurative referents. For example, the phrase
line one’s pocket means something like 'to embezzle funds.' Both the words line
and pocket have metaphorical relations to their figurative referents 'embezzle' and
'funds.' Idioms like this, whose words are metaphorically related to their figurative
referents, would be rated as being a 4 along the transparent-opaque continuum.
However, some idioms have one word that is relatively transparently related to its
figurative referent, and one that is metaphorically related. For example, the phrase
spill the beans means something like 'reveal a secret.’ Although there is a fairly
close relationship between spill and 'reveal', the word beans refers to 'secrets' in a
less direct, metaphorical way. Idioms like this would be rated between 1 and 4.
Conversely, those idioms with one word that is metaphorically related to its
figurative referent and one word that is opaquely (or arbitrarily) related would be
rated between 4 and 7.
133
Appendix D
English idiom types
S ec re tiv e n ess
Transparent.
to close one’s eyes to something
to keep something hush-hush
to keep to oneself about something
to cover one’s tracks
to do behind one’s back
Metaphorical:
to do behind the scenes
to have up one’s sleeve
to keep it under one’s hat
to sweep under the rug
to keep it under wraps
Opaque:
to beat around the bush
"mum’s" the word
to speak with a forked tongue
to have a skeleton in one’s closet
to sail under false colors
Revelation
Transparent.
to come out in the open
to get wise to something
to own up to something
to speak one’s mind
to tell it like it is
Metaphorical:
to air one’s dirty laundry
to blow the lid off
to blow the whisde on
to dish the dirt
to lay one’s cards on the table
Opaque:
to make no bones about
to see the writing on the wall
to open a Pandora’s box
to rat on someone
to shoot from the hip
134
A nger
Transparent.
to be sick and tired of
to bear a grudge against someone
to lose one’s cool about
to lose one’s temper
to get under one’s skin
Metaphorical:
to be at each other’s throats
to blow one’s stack
to flip one’s lid
to make one’s blood boil
to look daggers at someone
Opaque:
to give someone a dressing down
to have a bone to pick
to fly off the handle
to be in a lather
to get one’s goat
C ontrol
Transparent.
to have at one’s beck and call
to have a hand in
to get the upper hand
to have it over someone
to be in charge
Metaphorical:
to bring to terms
to set the ball rolling
to crack the whip
to put someone through their paces
to take the bull by the horns
Opaque:
to carry off the palm
to bring someone to bay
to lead somebody by the nose
to be the top banana
to have over a barrel
135
Insanity
Transparent.
to boggle one's mind
to be not all there
to be out of one's mind
to be all mixed up
to take leave of one’s senses
Metaphorical:
to go off the deep end
to have bats in one’s belfry
to come apart at the seams
to go to pieces
to lose one's marbles
Opaque:
to go ape over something
to be off the wall
to be a basket case
to drive one bananas with
to be mad as a hatter
136
Appendix E: Instructions for categorization of English idioms (Experiment 4A)
Idioms are phrases that mean something other than what they appear to
mean on the surface. That is, their figurative meanings are typically quite different
than their apparent, or literal, meanings. In this experiment you are being asked to
categorize American English idiomatic phrases according to their figurative
meanings. Your job is to select which of five categories each of the phrases fits
into. That is, you are to determine which of the five categories MOST represents
each phrase. Please try to categorize each phrase as quickly as possible without
sacrificing accuracy.
For example, the phrase to double-cross someone means to promise one
thing and deliver another, or to deceive. This phrase would be categorized most
accurately as having to do with the concept of secrecy, because it describes an act
that is done in secret. An example of another concept is to lower the boom on
someone. This phrase is said to describe when someone punishes someone else
strictly, checks or fully stops that person from doing what they were doing. This
phrase would most accurately be categorized as being one of control. The
remaining three concepts are revelation, insanity, and anger. Revelation involves
revealing information or letting out secrets (e.g., to spill the beans). Insanity is
becoming or being mentally unstable (e.g., to be out to lunch). Anger is, quite
simply, becoming or being angry (e.g., to raise one's hackles).
Some of the phrases you will see may not be familiar to you. All are
American English idioms. Please try your best to categorize them correctly based
on your native-speaker intuition about what each phrase might mean when used in
English.
Appendix F
Latvian idiom types
S ecretiveness
Transparent.
to do behind one’s ears
to command one’s tongue
to keep in a state of unknowing
to hold one’s tongue behind one’s teeth
to lock up one’s mouth
Metaphorical.
to keep as in a grave
to talk behind the stage’s curtain
to keep it in one’s chest
to scatter sand in someone’s eyes
for one's tongue to itch
Opaque'.
to cast something in white
to keep in one’s own skin
to sew on a beard
to keep something behind seven locks
to arch like an eel does
Revelation
Transparent.
to lift into the light
to blow all over the region
to be a pot of gossip
the mouth takes the lead
to go from mouth to ear
Metaphorical:
for someone to unfasten their tongue
to try to hide water in a sack
to spill water
to open the sluices
to release something in the wind
Opaque:
to let gin out of the bottle
to tumble out like an old fence
to let stones roll to somebody
like scales falling from eyes
to blow ducklings
Anger
Transparent.
for someone to boil over
for someone to blaze up
for someone to begin to boil inside
to split with anger
for someone to heat up over something
Metaphorical:
to burst out of one's skin
to let one’s mind loose
for someone to spit fire
for someone to speak with bared teeth
to let out a storm
Opaque:
for one’s wisdom to fall to pieces
for a lung to come out
to go out of the frames
to jump on harrows
to fall in someone’s fur coat
Control
Transparent.
for someone to lead the meeting
to be the head of something
to be an overseer for something
to throw out the boss
to play boss at something
Metaphorical:
to pull the reins
to lead the parade
for one to play first violin
to be on the ship’s bridge
to act as if in a man’s shoes
Opaque:
for one to be at the syringe
to cleave with an ax
to put everything onto shelves
to keep within the frames
to be like “amen” in church
139
Insanity
Transparent:.
for one to go out of order
to leave one’s reason
to lose one’s brain
when all five senses are not home
for something to darken one’s mind
Metaphorical.
for someone to become unyoked
to come off one's hinges
for someone to be short of film
to let go of the reins
to walk on blowing sand
Opaque'.
to have a squeezed head
to go to butter
for one to turn into a Swede
to go in all heaven’s directions
to be caught like a fish
140
Appendix G: Instructions for categorization of Latvian idioms (Experiment 4B)
categorize idiomatic phrases according to their figurative meanings. These phrases
have been literally translated from another language. Your job is to select which of
five categories each of the phrases fits into. That is, you are to determine which of
the five categories MOST represents each phrase. Please try to categorize each
phrase as quickly as possible without sacrificing accuracy.
For example, in Latvian the phrase to blow mist into someone's eyes means
to hide information from someone by way of leading them astray. This phrase
would be categorized most accurately as having to do with the concept of secrecy,
because it describes an act that keeps information secret or unknown to another
person. An example of another concept is to live under someone's heel. This
phrase describes someone whose actions are controlled by another. This phrase
would most accurately be categorized as being one of control. The remaining three
concepts are revelation, insanity, and anger. Revelation involves revealing
information or letting out secrets (e.g., to pull the end o f the bag open). Insanity is
becoming or being mentally unstable (e.g., to dissolve into pieces). Anger is, quite
simply, becoming or being angry (e.g., fo r blood to climb into one's cheeks).
Some of the phrases you will see may not make much sense to you. All are
literally-translated Latvian idioms. Please try your best to categorize them based on
your intuition about what each phrase might mean when used figuratively in another
language.
141
Appendix H
Mandarin idiom types
Secretiveness
Transparent.
to shut up and not to say
to chat stealthily
not know by men, not perceived by ghosts
to be in deep silence and rarely talk
unknown to anybody
Metaphorical:
to have a knife in the smile
for there to be words in words
to bury one’s first name
to whitewash a mistake and gloss over a fault
not a drop of water could trickle through
Opaque:
embedded deeply as if it is empty
the body lives behind the curtain
to be covered in a drum
like entering five miles into the cloud
good things sink to the bottom
Revelation
Transparent.
for there to suddenly be great understanding
to divulge the plot
the true situation comes clear
to disclose one’s feelings of the heart
one word speaks the truth
Metaphorical:
even the streets know what has happened
for one’s speaking to wake another from dreams
participants are confused but the audience is clear
the family disgrace is waved outside
to cover the head but expose the tail
Opaque:
when something happens at eastern windows
to turn back the old debts
the ground doesn’t have three hundred dollars
there is another village behind the trees and flowers
to open the liver and empty the gall bladder
142
A nger
Transparent.
both humans and gods are angry
one’s rage cannot be stopped
to scold with dirty words at length
to speak out grievances furiously
one big quarrel
Metaphorical:
to harden one’s contenance like a board
with swords drawn and bows stretched
to help chum up waves and billows
with thirty feet of flame sparkling above one’s head
a furious glow that soars up to the sky
Opaque:
even Buddha could shoot fire
to bite hard on one’s teeth
for one’s hair to stand up and tip off one’s hat
too much grudging will break the intestine
as if dog’s blood were poured over one’s head
Control
Transparent
to stop something once the order is commanded
to hold great power
to own the entire situation
to have the whole game under one’s manipulation
to operate something at one’s will
Metaphorical:
to use one hand to cover the sky
to close the window tightly before it rains
to perform a task according to one’s heart and hands
the weak are the prey of the strong
to turn round the heavens and earth
Opaque:
a monk comes to see the Buddha’s face
the big fish eats the small fish
to kill the chicken for a monkey to see
to lay down a golden hook
to first cut the head and then report to the emperor
143
Insanity
Transparent.
with no cause and no reason
the state of the mind is not at ease
to be inconsistent in joy and anger
to have unclear spirit and intelligence
to be frightened and out of control
Metaphorical:
the spirit is out of the house
lost soul and fallen spirit
for one’s nerves to misfire
to be beyond rescue
to run without legs
Opaque:
the man of Qi worries about the sky falling
seven up and eight down
to fish the moon out of the water
chickens fly and dogs jump
Szechwan dogs bark at the sun
144
Appendix I: Instructions for categorization of Mandarin idioms (Experiment 4C)
categorize idiomatic phrases according to their figurative meanings. These phrases
have been literally translated from another language. Your job is to select which of
five categories each of the phrases fits into. That is, you are to determine which of
the five categories MOST represents each phrase. Please try to categorize each
phrase as quickly as possible without sacrificing accuracy.
For example, in Mandarin Chinese the phrase to act in the dark means to
hide something from others by doing it so as not to be seen. This phrase would be
categorized most accurately as having to do with the concept of secrecy, because it
describes an act that keeps information secret or unknown to another person. An
example of another concept is to hold the bull's ears. This phrase describes
someone who controls another's actions. This phrase would most accurately be
categorized as being one of control. The remaining three concepts are revelation,
insanity, and anger. Revelation involves revealing information or letting out
secrets (e.g., to tear o ff the surface coating o f a picture). Insanity is becoming or
being mentally unstable (e.g., to be completely lawless and godless). Anger is,
quite simply, becoming or being angry (e.g., to roar like thunder).
Some of the phrases you will see may not make much sense to you. All are
literally-translated Mandarin idioms. Please try your best to categorize them based
on your intuition about what each phrase might mean when used figuratively in
another language.
Appendix J: English/Latvian idiom types
S e c r e tiv e n e s s
E n g lis h L a tv ia n
Transparent:
to close one’s eyes to something to do behind one’s ears
to keep something hush-hush to command one’s tongue
to keep to oneself about something to keep in a stale of unknowing
to cover one’s tracks **1 to hold the tongue behind one’s teeth
to do behind one’s back to lock up one’s mouth
Metaphorical:
to do behind the scenes to keep as in a grave
to have up one’s sleeve to talk behind the stage’s curtain
to keep it under one’s hat to keep it in one’s chest
to sweep under the rug to scatter sand in som eone’s eyes
to keep it under wraps for one's tongue to itch
Opaque:
to beat around the bush to cast something in white
"mum's" the w ord *2 to lock one’s heart with nine locks
to speak with a forked tongue to sew on a beard
to have a skeleton in one’s closet to keep something behind seven locks
*to sell somone down the river to arch like an eel does
* in d ic a te s slig h t m o d ific a tio n s o f o r ig in a l p h rase (fr o m E xp erim en t 4 ) b a sed o n r esu lts fro m N o r m in g S tu d y 5 A
2 in d ic a tes n e w p h ra ses (rep la cem en ts o f ph rase u s e d in E xp erim en t 4 ) b a s e d o n r esu lts fro m N o rm in g S tu d y 5 A
145
Revelation
English Latvian
Transparent:
to come out in the open to lift into the light
to get wise to something to blow all over the region
to own up to something to be a pot of gossip
to speak one’s mind the mouth takes the lead
to tell it like it is to go from mouth to ear
Metaphorical:
to air one’s dirty laundry for som eone to unfasten their tongue
to blow the lid off to try to hide water in a sack
to blow the whistle on to spill w ater
to dish the dirt to open the sluices
to lay one’s cards on the table to release something in the wind
Opaque:
to make no bones about to let gin out of the bottle
to see the writing on the wall to tum ble out like an old fence
to open a Pandora’s box to let stones roll to somebody
to rat on someone like scales falling from eyes
to shoot from the hip to blow ducklings
English Latvian
Transparent:
to be sick and tired of for someone to boil over
to bear a grudge against someone for someone to blaze up
to lose one’s cool about for someone to begin to boil inside
to lose one’s temper to split with anger
to get under one’s skin *to be in anger’s power
Metaphorical'.
to be at each other’s throats to burst out o f one's skin
to blow one’s stack to let one’s mind loose
to Hip one’s lid for someone to spit fire
to make one’s blood boil for someone to speak with bared teeth
*to foam at the mouth to let out a storm
Opaque:
*to jum p down som eone’s throat for one’s wisdom to fall to pieces
to have a bone to pick for a lung to com e out
to fly off the handle to go out o f the frames
to be in a lather to jum p on harrows
to get one’s goat to fall in som eone’s fur coat
English Latvian
Transparent:
to have at one’s beck and call for someone to lead the meeting
to have a hand in *lo say one tough word
to get the upper hand to be an overseer for something
to have it over someone to throw out the boss
to be in charge to play boss at something
Metaphorical:
to bring to terms *to live under one’s slipper
to set the ball rolling to lead the parade
to crack the whip for one to play first violin
*to blaze a trail to be on the ship’s bridge
to take the bull by the horns to act as if in a m an’s shoes
Opaque:
*to be a brass hat for one to be at the syringe
to bring someone to bay to cleave with an ax
**to lead by the nose to put everything onto shelves
to be the top banana *to climb on som eone’s head
**to have someone over a barrel to be like “amen” in church
English Latvian
Transparent:
to boggle one's mind for one to go out of order
to be not all there to leave one’s reason
to be out o f one's mind to lose one’s brain
to be all mixed up when all five senses are not home
to take leave o f one’s senses for something to darken one’s mind
Metaphorical:
to go off the deep end for someone to become unyoked
to have bats in one’s belfry *to pull oneself to pieces
to come apart at the seams for someone to be short of film
to go to pieces to let go of the reins
to lose one's marbles to walk on blowing sand
Opaque:
to go ape over something to have a squeezed head
to be off the wall to go to butter
to be a basket case for one to tum into a Swede
to drive one bananas with to go in all heaven’s directions
to be mad as a hatter to be caught like a fish
150
Appendix K: Instructions for categorization of English/Latvian idioms

(Experiment 5B)
categorize idiomatic phrases according to their figurative meanings. The phrases
you will be seeing are a mix of English idioms and idioms that have been literally
(or directly) translated from another language. Your job is to select which of five
categories each of the phrases fits into. That is, you are to determine which of the
five categories MOST represents each phrase. Please try to categorize each phrase
as quickly as possible without sacrificing accuracy.
For example, the phrase to double-cross someone means to promise one
thing and deliver another, or to deceive when used in English. This phrase would
be categorized most accurately as having to do with the concept of secrecy,
because it describes an act that is done in secret. However, other languages express
such ideas differently. For example, in Latvian the phrase to blow mist into
someone's eyes means to hide information from someone by way of leading them
astray. This phrase would also be categorized most accurately as having to do with
the concept of secrecy, because it describes an act that keeps information secret or
unknown to another person.
An example of another concept is to lower the boom on someone. This
phrase, when said in English, describes when someone punishes someone else
strictly, checks or fully stops that person from doing what they were doing. This
phrase would most accurately be categorized as being one of control. Another
example of this concept is to live under someone's heel. This Latvian phrase
describes someone whose actions are controlled by another. This phrase would
also most accurately be categorized as being one of control.
The remaining three concepts are revelation, insanity, and anger.
Revelation involves revealing information or letting out secrets (e.g., to spill the
beans, in English; to pull the end o f the bag open, in Latvian). Insanity is
becoming or being mentally unstable (e.g., to be out to lunch, in English; to
151
dissolve into pieces, in Latvian). Anger is, quite simply, becoming or being angry
(e.g., to raise one's hackles, in English; for blood to climb into one's cheeks, in
Latvian).
Half of the phrases you will be asked to categorize during this experiment
are English idioms and half are literally-translated Latvian idioms. Many of the
phrases will be familiar, while some may not make much sense to you. Please try
your best to categorize them all based on your intuition about what each phrase
means, either when used as fixed English phrases or as figurative phrases in
another language.
IMAGE EVALUATION
TEST TARGET ( Q A - 3 )
✓
■£123 |2 ^
1.0
S “ HUi
12.0
l.l
1.8
1.25 1.4 1.6
150mm
IIV H G E .Inc
1653 East Main Street
Rochester, NY 14609 USA
Phone: 716/482-0300
Fax: 716/288-5989

Idioms

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Idioms

Uploaded by

Copyright:

Available Formats

INFORMATION TO USERS

The quality of this reproduction is dependent upon the quality o f the

Oversize materials (e.g., maps, drawings, charts) are reproduced by

Photographs included in the original manuscript have been reproduced

by Native and Non-Native Speakers

The Graduate School

in Partial Fulfillment of the

State University of New York

UMI Microform 9822946

This microform edition is protected against unauthorized

for the above candidate for the

Doctor of Philosophy degree,

hereby recommended acceptance of this dissertation.

Susan BrennjiRfAssociate Professor, Department of Psychology

Grover Whitehurst, ProfesiorJDepartment of Psychology

Richard Gerrig, Associate Professor, Department of Psychology

Eva Fedeplattay, Professor, Department of Philosophy

case, English, Latvian, or Mandarin) to make sense of figurative speech used in

These results motivated a second series of experiments further examining

analyzable (transparent), abnormally analyzable (metaphorically-based), or

unanalyzable (culturally- or historically based) they are.

necessary to categorize another languages’ metaphorical idioms correctly, they must

On the other hand, speakers are able to correctly categorize fewer

these idioms require particular cultural and/or historical knowledge to be

understood. Furthermore, unanalyzable idioms take the longest time to categorize

structures of these phrases have to any readily accessible conceptual metaphors.

language’s analyzable idiomatic phrases, adult non-native speakers are able to

applications of these old concepts (e.g., in recently coined colloquialisms, in slang,

in film and in literature), we must go through the process of accessing the

metaphorical meaning we have been developing our entire lives.

A cknow ledgm ents..............................................................xiii

II. E xperim ent 1...................................................................... 13

III. E xperim ent 2......................................................................32

IV. E xperim ent 3......................................................................36

Results and Discussion................................67

VI. Experim ent 5.........................................................................76

VII. General D iscussion..............................................................85

Appendix B ........................................................... 112

Appendix C ........................................................... 122

Appendix G ............................................................. 140

Table 2. Three different native speakers’ responses to questions regarding their

and incorrectly on pre-test.

by native English speakers (established in Experiment 1). Proportions are

post-tests (prior to and following Mental Imaging Session I of Experiment 3).

Overall percentage correct is listed below the corresponding response time.

laboratory of crystals growing in solution, the delicate blue

original is so transformed as to be no longer recognizable.”

enthusiasm in pursuing answers to language production and processing questions.

explain perceptual subtleties contributed fundamentally to my appreciation for the

use and understanding of language at all levels of the food chain.

program. Richard’s presence on my dissertation committee provided an expert

voice from the field with the added wit of a comedian.

Of course, I am indebted to all the people who participated in my studies,

questionnaires. In particular, I would like to thank Rachael Hoina for her

impressive consistency and patience. Rachael helped me through several years’

Finally, I would like to acknowledge the degree to which Michael Ahrens

always serve as a model when I need to get my own house in order.

Researchers interested in language have long tried to explain how it is that

In the course of trying to determine how we understand idioms in particular

languages). More important, these studies address the overriding question

no resolution of the previously outlined debate can be reached without cross-

Why study idioms, in particular?

Psycholinguisdc analysis of idiomatic language can provide information

Makkai’s original (1975) Dictionary o f American Idioms presented over 4000