You are on page 1of 17

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/223824454

The comprehension of idioms

Article  in  Journal of Memory and Language · December 1988


DOI: 10.1016/0749-596X(88)90014-9

CITATIONS READS

325 3,763

2 authors, including:

Cristina Cacciari
Università degli Studi di Modena e Reggio Emilia
89 PUBLICATIONS   2,230 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

LA FORMACION DE LA CONCIENCIA FIGURATIVA EN LA ETAPA DE EDUCACIÓN PRIMARIA: EL HUMOR Y LA FRASEOLOGÍA (FIGURKID) (FFI2016-76047-P) View project

Gender Stereotyping in language View project

All content following this page was uploaded by Cristina Cacciari on 25 July 2019.

The user has requested enhancement of the downloaded file.


JOURNAL OF MEMORY AND LANGUAGE 27,66-83 (1988)

The Comprehension df Idioms

CRISTINA CACCIARI AND PATRIZIA TABOSSI


Dipartimento di Psicologia, Universitci di Bologna
The access of idiomatic expressions was investigated in three cross-modal priming ex-
periments. When the idiomatic string was predictable, subjects were faster at performing a
lexical decision to idiomatically related targets than to literally related targets (Experiment
1). When the string was not recognizable as idiomatic until its completion, then subjects
were faster on the target literally related to the last word (Experiment 2). Lexical decision
for the same materials used in Experiment 2, when target words were presented 300 ms after
the end of idiom was heard, was faster both for targets related to the last word and for the
idiomatic meaning of the string, relative to control targets (Experiment 3). Neither the direct
access model nor the lexical representation model seems adequate to explain the present
findings. A new hypothesis concerning the representation and the processing of idioms is
presented. 63 1988 Academic Press, Inc.

Typically, an idiom is characterized as a However, insofar as this view does not al-
string of words whose semantic interpreta- low for the comprehension of idioms, these
tion cannot be derived compositionally expressions are problematic for any theory
from the interpretation of its parts. “Shoot of language processing. They are even
the breeze,” for instance, cannot be under- more challenging if one considers how fre-
stood by putting together shoot, the, and quently they appear in the language. Boat-
breeze in the same way in which “eat the ner, Gates, and Makkai’s (1975) Dictionary
soup” can be understood. Thus, idiomatic of American Idioms, for example, includes
expressions defy the standard view of lan- 4000 expressions, a list which is far from
guage comprehension according to which exhaustive. It is clear, therefore, that the
understanding a sentence entails at least comprehension system cannot deal with
recognizing the individual words in the sen- them simply as exceptions that must be
tence, retrieving their meanings from the learned but do not affect the normal func-
mental lexicon, and combining them ac- tioning of the system.
cording to their grammatical relations. Although the issue is important, and its
theoretical relevance has long been ac-
The order of authors’ names is alphabetical. The knowledged by linguists (Fraser, 1970),
present research was supported by Grant CT. 1131 psycholinguists have devoted little atten-
from the Italian National Council for Research (CNR)
and by Grant 8600499 from Minister0 Pubblica Istruz- tion to the way in which idioms are identi-
ione. Correspondence and requests for reprints con- fied and comprehended. To date, there are
cerning this article should be sent to Cristina Cacciari, two main views. According to the lexical
Dipartimento di Psicologia, Viale Berti-Pichat, 5, representation hypothesis, idioms are
40127 Bologna Italia. We thank Corrado Cavallero, stored in the mental lexicon as individual
Eugenio Pasini, and the Istituto di Elettronica e Elet-
trotecnica, Universita di Padova without whose help items and are retrieved in the same way as
the experimental materials would have never been any other word. This retrieval process is
made. We are also grateful to Oliviero Stock, Ino Flo- initiated as soon as the first word of an id-
res d’Arcais, Sam Glucksberg, Gary Dell, Mike iomatic expression is encountered and runs
Tanenhaus, Curt Burgess, Bob Peterson, and Lucia in parallel with the computation of the lit-
Colombo for helpful discussions. We also thank Isa-
bella Cupelli for her assistance in collecting and ana- eral meaning of the expression (Swinney &
lyzing data and Linda Ginzel for improving the English Cutler, 1979).
form. An alternative view has been proposed
668
0749-5%x/88 $3.00
Copyri&t 0 1988 by Academic Press, Inc.
Au rights of reproduction in any form reserved.
COMPREHENSION OF IDIOMS 669

by Gibbs (1980, 1986), who claims that the presented after a short story containing an
identification of an idiom does not occur in idiomatic expression (e.g., “he kept it un-
parallel with the computation of its literal der his hat”). The story biased either the
meaning. Rather, it precedes the derivation idiomatic interpretation of the expression
of the literal meaning, which is accom- or its literal interpretation, as illustrated in
plished only if the idiomatic interpretation the following example:
of the string fails to integrate with the con-
text. Idiom story.
Evidence in support of either of these hy- Alice was telling her friend about her
potheses is ambiguous. One reason is that brother Bob.
the hypotheses lead to similar predictions, “Bob saw someone steal a diamond brace-
although Gibbs’ direct access hypothesis let in a jewelry shop.”
does make stronger claims than the lexical “Did he tell anyone?” the friend asked.
representation hypothesis. Results showing “No,” Alice replied,
that accessingthe meaning of an idiom from “he kept it under his hat.”
the mental lexicon is faster than computing Literal story.
its literal interpretation are compatible with Grace and Mary were watching a magician
both hypotheses. On the contrary, both hy- perform at their school.
potheses seem to reject the possibility of For one trick he held out a rabbit and made
the literal interpretation being faster than it disappear.
the idiomatic interpretation. However, in Grace couldn’t figure it out.
the lexical representation hypothesis the “Where is it?” she asked.
two processes are thought to start in paral- “It’s obvious,” Mary replied,
lel. Findings showing no difference in com- “he kept it under his hat.”
prehension times of idiomatic and literal in-
terpretation are easier to accommodate The target phrase was either a para-
within this framework than in Gibbs’. Here, phrase of the nonliteral interpretation of the
the idiomatic interpretation must be faster idiom (e.g., “he didn’t tell anyone”), a
than its literal counterpart, because it is the paraphrase of its literal interpretation (e.g.,
only one initiated automatically upon pre- “it’s beneath its cap”), or an unrelated
sentation of the string. Indeed, results from phrase (e.g., “it happened to Sally”).
several studies are compatible with both Gibbs found that subjects made faster deci-
views, showing faster reaction times for the sions about the nonliteral paraphrase than
figurative interpretation of an idiomatic ex- about either the literal paraphrase or the
pression than for its literal interpretation, unrelated phrase, regardless of the story
both in and out of context (Gibbs, 1980; which they had been presented.
1986; Ortony, Schallert, Reynolds, & An- One problem with these studies is that
tos, 1978). they do not necessarily reflect the way peo-
Swinney and Cutler (1979) had subjects ple process idiomatic expressions during
judge whether a string of words was a comprehension.
meaningful phrase. Decision times were Probably the only on-line study pub-
faster for those strings which had both an lished so far is that by Estill and Kemper
idiomatic and a literal interpretation (e.g., (1982). Their subjects monitored for a tar-
“break the ice”) than for those that had get word in a sentence (e.g., WALLS),
only a literal interpretation (e.g., “break which was the same as a word specified in
the cup”). advance (word condition: e.g., WALLS),
A similar pattern of results was obtained rhymed with a prespecified word (rhyme
by Gibbs (1986), whose subjects had to de- condition: e.g., FALLS), or belonged to a
cide the meaningfulness of a target phrase prespecified category (category condition:
670 CACCIARI AND TABOSSI

e.g., PART OF A BUILDING) (Marslen- Also, as Mueller and Gibbs (1987) have
Wilson & Tyler, 1980). The target word, recently pointed out, intuition suggeststhat
which was the last word of an idiom (e.g., idiomatic expressions differ not only in how
“climbing the walls”), occurred in one of frozen they are, i.e., the kind of operations
four sentential contexts: a sentence con- that they can undergo (Fraser, 1970), but in
taining the idiom and biasing its literal in- other ways that may be relevant to how
terpretation (e.g., “Orville was interested they are processed and understood. For ex-
in spiders and could sit for hours and watch ample, according to definition, “feel out”
them climbing the walls of the garden.“); a is an idiom as much as “spill the beans,”
sentence containing the idiom and biasing although ordinary speakers’ intuition about
its nonliteral interpretation (e.g., “By the the two expressions is likely to be very dif-
fourth day in the hospital, Orville was ferent. Namely, the nonliteralness of the id-
climbing the walls to go home.“); a sen- iom will be easily acknowledged for the lat-
tence containing the idiom and ambiguous ter expression but not for the former. More-
about its interpretation (e.g., “Orville over, even “very idiomatic” idioms differ
hated prison and was climbing the walls to from one another with regard to how early
get out.“); and finally, a sentence contain- in the string they can be identified, even
ing a nonidiomatic expression (e.g., “Or- without taking context into account.
ville began the renovation of his old house Although these factors have not been
by knocking out the walls.“). Estill and considered in current research on idiom
Kemper (1982) found that word monitoring comprehension, they are likely to influence
was significantly slower in nonidiomatic the way in which people understand these
contexts than any of the idiomatic contexts, expressions. The present study is an initial
which did not differ reliably from one an- exploration of how a subclass of idiomatic
other. The authors interpreted the lack of expressions is processed during sentence
difference among the three sentences con- comprehension. These expressions have
taining the idiomatic expression as evi- the following characteristics:
dence in support of the lexical representa-
tion hypothesis. 1. Syntactically they have the structure:
These studies, whether on-line or off- V (NP) (PP)(PP), where at least one of the
line, have treated idioms as a type of ambi- constituents in parentheses is present.
guity: “kick the bucket” is an ambiguous 2. They are not ambiguous in that they
expression meaning either “kick the pail” have no literal counterpart. According to
or “die” in a fashion similar to the way in Ruwet (1983) they are opaque.
which “bank” is ambiguous as “money”
or “river bank.” Consistent with this view, 3. Their interpretation cannot be deter-
the authors have focused their attention on mined by trying to make sense of their lit-
how fast people can understand the two eral meaning and must be part of the lin-
meanings. Likewise, the two theories differ guistic knowledge of the listener.
mainly in that they suggest different time 4. The idiomatic string can be completed
courses in the comprehension of each literally until its last word.
meaning.
While many idioms can also be inter- This procedure specified idioms that are
preted as literal sentences, this is not an probably processed in the same way. Also,
essential characteristic. None of the follow- this choice has practical advantages at this
ing idioms, for instance, can be interpreted stage of research: It allows different hy-
literally: “by and large,” “look after,” potheses to make distinct predictions that
“lay down the law, ” “shoot the breeze,” can be tested with an ordinary on-line tech-
“build castles in the air.” nique such as the cross-modal priming par-
COMPREHENSION OF IDIOMS 671

adigm (Swinney , Onifer, Prather, & Hir- EXPERIMENT 1


shkowitz, 1979). Method
Let us suppose that after listening to the
sentence Subjects. A total of 35 undergraduates
(20 women and 15 men) volunteered to par-
After the excellent performance, the tennis ticipate in an experimental session lasting
player was in seventh heaven. about 20 min. None of them had previously
a subject makes a lexical decision to either participated in an experiment of this sort.
SAINT, semantically related to heaven, or Material and design. Twenty familiar
to HAPPY, semantically related to the idioms,’ selected according to the criteria
meaning of the idiom, or to the unrelated outlined above, were presented in a random
control UMBRELLA. In the sentential order to 15 subjects, who were requested to
context preceding the idiomatic string, or in give paraphrases of the expressions. Nine
the string itself before heaven, no cue is idioms, whose meanings were correctly
given to suggest that an idiom is going to paraphrased by at least 95% of the subjects,
occur. Intuitively, one would imagine that were obtained for use in the experiment.
in this case comprehension goes on compo- For each of the nine idiomatic expres-
sitionally until one recognizes heaven. sions, a sentence was constructed so that
Since this word does not tit literally in the the information preceding the idioms biased
context, the sentence will be reconsidered neither the idiomatic nor the literal comple-
and eventually reinterpreted idiomatically. tion of the string (e.g., “After the excellent
If the hypothesis that has been ascribed to performance, the tennis player was in sev-
Bobrow and Bell (1973) (Gibbs, 1980; Swin- enth heaven.“-“ After the excellent per-
ney & Cutler, 1979) can ever be shown to formance, the tennis player was in seventh
have psychological validity, it will be in a position.“). Three visually presented target
case as the one detailed above, where no words were paired with each sentence: a
cues to the idiomatic interpretation are word semantically associated with the
given until the end of the string. Thus, ac- meaning of the idiom (idiom target); a word
cording to this hypothesis, one should ex- associated with the meaning of the last
pect responses to SAINT to be faster than word in the string (literal target); and an
those to HAPPY or UMBRELLA. unrelated word to be used as control (con-
The opposite prediction is made by Gibbs’ trol target).
hypothesis, according to which the idiom- The idiom targets were selected accord-
related target HAPPY should be faster than ing to the answers produced by the 15 sub-
both SAINT and UMBRELLA. jects mentioned above. Ten different judges
Finally, the lexical representation hypothe- were used to select the literal targets. They
sis predicts that both SAINT and HAPPY were presented with the final words of the
should be faster than the unrelated control. nine idioms in a random order and re-
In fact, if the idiom is accessed as a lexical quested to produce the first associates that
item, its meaning must be available by the came to mind. A control experiment was
end of the string, thus facilitating HAPPY. run to determine that the lexical decision
Moreover, if the literal processing of the i Experiments 1,2, and 3 were conducted with Ital-
expression is carried out in parallel, at least ian subjects and Italian materials which are listed in
until some cue favors the idiomatic inter- the Appendix along with their corresponding English
pretation, in the present case heaven must translations. Since most of the Italian idioms have no
be recognized and accessed, thereby facili- exact English equivalent, we have given both a word-
by-word translation and the English meaning or equiv-
tating SAINT. In order to test these hy- alent for each idiom. The English translations were
potheses, Experiment 1 was devised. used in the text.
672 CACCIARI AND TABOSSI

times for the nine idiom targets, the nine the filler sentences, the pulse occurred at
literal targets, and nine arbitrarily chosen, the offset of one word chosen in such a way
unrelated controls were comparable in iso- as to cover the entire range of positions
lation. These words were included in a list within sentences (i.e., beginning, middle,
of 304 items (152 words and 152nonwords). and end). The pulse caused a word to be
The list was presented in four randomized displayed on the screen of an Apple micro-
blocks to 9 subjects, who performed a lex- computer for 1500 ms. It also started a dig-
ical decision task on the items. The mean ital timer which either stopped when the
reaction times (RTs) and standard devia- subjects pressed the space bar or reset au-
tions (SDS) for the idiomatic target words, tomatically after 5 s.
the literal target words, and the control In order for the same idiom to be pre-
words were the following: 519 ms (SD 34), sented only once to each subject, three sets
520 ms (SD 31), and 520 ms (SD 19). They of visual targets were constructed. In one
did not differ significantly. set the visual target paired with an experi-
In addition to the test materials, filler ma- mental sentence was related to the idiom-
terials were also prepared. They consisted atic meaning; in another list it was associ-
of 30 “formal” and 21 “colloquial” sen- ated with the meaning of the last word in
tences. The latter, which were obtained by the string and in the third list it was the
using colloquial lexical items, but did not control target. The filler visual targets were
contain idiomatic expressions, were de- the same in the three sets. In each set, three
vised in order to avoid the experimental targets were idiomatic targets, three were
sentences being noticed by the subjects as literal targets, and three were controls,
strange or awkward in the list. Each of yielding a total of nine experimental tar-
these 51 filler sentences, which were ap- gets.
proximately of the same length as the ex- One set of 60 visual words was paired
perimental sentences, was paired with a vi- with the list of sentences and randomly as-
sual target. Fifteen formal and 15 colloquial signed to an equal number of subjects who
sentences were paired with target visual acted as their own controls.
nonwords, whereas the remainder were Procedure. Subjects were tested individ-
paired with target visual words. Three vi- ually. They sat in a sound-attenuated room
sual filler words had a semantic association in front of an Apple microcomputer screen
with one word of the paired formal sen- connected to a stereo tape recorder. They
tence and three with one word of the paired were instructed to perform a lexical deci-
informal sentence. sion task to strings that would appear on the
The 60 sentences were recorded in a pre- screen while listening to a sentence. They
specified random order by a male speaker were told to press the space bar of the Ap-
on one channel of a tape recorder. There ple keyboard with their dominant hand (3
was a 5-s interval between sentences. A subjects were left-handed) if the string was
lOOO-Hzpulse,* inaudible to the subjects, an actual word and to do nothing otherwise.
was placed automatically on the other After 10practice trials, one of the three sets
channel of the tape, precisely at the offset paired with the list of sentences was pre-
of one word in each sentence. In the exper- sented to each subject. Reaction times to
imental sentences, the pulse was at the end lexical decisions were recorded.
of the last word of the idiomatic string. In The subjects were informed that they
should pay attention to the sentences, be-
*Pulses are placed automatically by using a pro-
gram developed at the Istituto di Elettronica e Elettro-
cause at the end of the experimental session
tecnica, LJniversit6 di Padova and run on a Nova com- they would be asked questions about them.
puter. In fact, immediately after the end of the
COMPREHENSION OF IDIOMS 673

sequence, the subjects were given 20 sen- occurs. These results are compatible with
tences, each printed on one card. Half of the results of Estill and Kemper (1982),
the sentences had been presented in the ex- who found an advantage of the idiomatic
periment and the remainder had not. The over the literal interpretation of an idiom-
latter were derived from the sentences the atic string. Hence, the results seem to sup-
subjects actually heard. Subjects who failed port Gibbs’ hypothesis that people under-
to reach a score of 60% correct recognitions stand idioms without even trying to com-
were excluded from subsequent analyses. pute their literal interpretation.
Among the 30 subjects who reached the Although Gibbs’ hypothesis can account
threshold, the level of correct recognitions for the present data, it does not explain the
was 75%. processes during comprehension that give
rise to the idiomatic interpretation. In fact,
Results the materials used in Experiment 1 give no
The mean percentage of errors was 1.5. apparent indication either in the sentential
In order to reduce variability, data points + context preceding the idiomatic string or in
2 SDS from the mean RTs of each subject the string itself prior to its last word that an
(2.97% of responses) were excluded from idiom has been encountered. Even assum-
the analyses. ing that the retrieval of the idiomatic mean-
Table 1 shows the mean RTs and SDSfor ing is initiated at the beginning of the string
the correct responses under the three ex- in parallel with its literal interpretation, it is
perimental conditions. The results showed difficult to imagine that people would sus-
a reliable main effect both in the analysis by pend the computation of the literal meaning
subjects (F(2,58) = 7.03, MSe 3071; p < of the sentence before accessing its last
.005) and in the analysis by materials word. Thus, before this conclusion is ac-
(F(2,16) = 4.20, MSe 1585; p < .05). Min cepted, a number of alternatives should be
F’(2,37) = 2.63 was not significant, considered.
Planned nonorthogonal comparisions First, the results of Experiment 1 are
showed that the idiom targets were re- puzzling only if they genuinely reflect the
sponded to faster than the literal targets, on-line process of idiom comprehension.
which did not differ significantly from the Otherwise, they merely replicate a consis-
control targets (Idiom vs. Literal: F(1,58) tent body of evidence obtained with post-
= 7.32,~ < .Ol; Literalvs. Control: F(1,58) perceptual tasks (Gibbs, 1980, 1986). In-
= 0.8, n.s.). deed, it has been claimed that the two tasks
most commonly used in cross-modal stud-
Discussion ies-naming and lexical decision-are both
sensitive to semantic factors, but differ in
The findings of Experiment I suggest that that the former reflects only immediate,
the interpretation of an idiomatic expres- perceptual processes, whereas the latter
sion is immediately available upon its pre- can reflect postperceptual as well as per-
sentation. There is no evidence that access ceptual processes (Seidenberg, Waters,
to the meaning of the last word in the string Sanders, & Langer, 1984). Hence, it would
be easy to conclude that the results of Ex-
TABLE 1 periment 1, obtained using the lexical deci-
MEAN REACTION TIMES AND STANDARD sion task, reflect postperceptual phenom-
DEVIATIONS (ms) FOR LEXICAL DECISION TO ena. However, there are reasons to doubt
TARGET WORDS IN EXPERIMENT 1
this explanation. Data collected in several
Idiomatic Literal Control studies using this technique indicate that it
599 (82) 638 (97)
can reflect on-line processes (Swinney, On-
651 (69)
ifer, Prather, & Hirshkowitz, 1979) without
674 CACCIABI AND TABOSSI

being susceptible to backward priming ef- captain, even if it is going to be completed


fects, at least in sentential contexts. More- as captive. It is possible that, despite pos-
over, there is evidence that in shallow lan- sible alternatives, subjects had a bias to-
guages, such as Italian and unlike English, ward the idiomatic interpretation of some
lexical decision is sensitive to semantic fac- of the expressions after the presentation of
tors, whereas naming is not. Thus this evi- an initial fragment of them. If this were the
dence challenges, at least for these lan- case, these could be identified as idioms be-
guages, the adequacy of this task to tap lex- fore enough information for the retrieval of
ical processing and suggests the the last word was encountered or identi-
appropriateness of the lexical decision in- fied.
stead (Katz & Feldman, 1983). While the If this intuition is correct, one could ex-
advantages and disadvantages of these pect that in a completion task where people
tasks are still under investigation, it is rea- are given initial fragments of these nine id-
sonable at least to consider that the findings ioms out of context, at least some of them
in Experiment 1 may not be an artifact of could be completed idiomatically. Indeed, a
the particular experimental paradigm. test of this sort was given to two groups of
A second possibility is that the materials 15 subjects each. They were asked to com-
(lists of sentences and sets of visual words) plete each string: one group at the third
allowed the subjects to enter an “idiom word and the other at the second word from
mode” of language processing (Bobrow & the end (e.g., “To be in . . .” and “to be in
Bell, 1973). They might have realized the seventh”). The string fragments were pre-
purpose of the experiment and developed a sented in a different random order for each
strategy to detect idioms. Or, more simply, subject. The results showed that the strings
they may have become attuned to the oc- were completed idiomatically in 34% of the
currence of idioms, therefore developing a cases at three words from the end, and in
bias toward anticipating the completion of a 45% of the casesat two words from the end.
sentence as idiomatic. However, the pro- When the same percentages were com-
portion of sentences containing idioms in puted without considering one idiom,
the list was only 15%, and such a propor- which was recognized much less than any
tion is usually not considered sufficiently of the others, the scores rose to 37.5 and
high to draw the subject’s attention to one 50%, respectively.
type of material in an experimental list Experiment 2 was devised to examine
(Tweedy & Lapinski, 1981). what meanings are available with idiomatic
A third hypothesis, which will be inves- strings that are not predictable in advance
tigated in the next experiment, concerns of their completion. That is, the last word
the nature of the idioms employed. They must be accessed in order for the idiom to
were selected so that they could be com- be recognized as such. If this were the case,
pleted equally well either literally or idiom- facilitation for the target related to that
atically; e.g., “to be in seventh . . .” could word should be observed.
be completed with position as well as with
heaven, among other words. Moreover, as EXPERIMENT 2
we have already stated, contexts created no
bias either way. However, it can be argued As in Experiment 1, in the test trials sub-
that a string such as “to be in seventh . . .” jects listened to a sentence containing an
can call to mind its idiomatic completion, idiom. At the end of the idiomatic string
regardless of whether other completions they were visually presented with a target
are possible. A similar phenomenon can be word to which they had to perform a lexical
found in word recognition, where a frag- decision task, as illustrated in the following
ment such as “capt” is likely to call to mind example:
COMPREHENSION OF IDIOMS 675

The girl decided to tell her boyfriend to go Three words were paired with each of 12
to the devil, once and for all. idioms for use as visual targets. One word,
AWAY obtained by selecting the most frequent ex-
HORNS pression used by 15 judges to paraphrase
TROUT each string, was related to its meaning (id-
iomatic target); one word, obtained by ask-
The sentences biased neither the idiom- ing a new panel of subjects simple word
atic nor the literal completion of the string. associations, was related to the meaning of
The idioms had the same characteristics as the last word in each string (literal target),
those used in Experiment 1, except that and the third word was an unrelated control
they were pretested and selected so that (control target).
people were not likely to complete the frag- In order to ensure that lexical decisions
ments preceding the last word of these ex- to the selected targets did not differ signif-
pressions idiomatically. icantly in isolation, a pretest similar to the
one described in Experiment 1 was per-
Method formed with 20 subjects. The mean RTs and
Subjects. A total of 33 undergraduates SDS for the idiomatic target words, the lit-
(23 women and 10 men) volunteered for the eral target words, and the control target
experiment, which lasted about 25 min. words were 537 ms (SD 20), 535 ms (SD
They had not previously participated in an 21), and 538 ms (SD 18), respectively.
experiment of this sort. In addition to the experimental materials,
Materials and design. A group of 40 fa- filler materials were constructed. They con-
miliar idioms, different from those used in sisted of 12 informal sentences and 48 for-
Experiment 1, was selected. To ensure that mal sentences, approximately the same
the initial parts of these expressions did not length as the experimental sentences. Half
call to mind the idioms, the selected strings of these (6 informal and 24 formal) were
were placed in sentences as informatively paired with visual target words, whereas
poor as possible (e.g., “He told him to go to the remainder were paired with legal non-
the devil.“). These sentences were mixed words. Six of the visual target words paired
with 175 filler sentences (e.g., “The man with formal sentences were associated with
was leaning on the door.“). Five lists were one of the words in the preceding context,
then created, each of which contained sen- whereas the remaining 24 filler targets were
tence fragments in a random order. The unrelated.
length of the fragments varied in the five The 72 sentences (12 experimental and 60
lists, so that in List 1 the fragments were filler) were recorded on one channel of a
the shortest (e.g., “He told . . .“) and in tape recorder by a male speaker in a pre-
List 5 they were the longest (e.g., “He told specified random order. There was a 5-s in-
him to go to the . . .“). In general, the terval between sentences. On the second
shortest fragments included only the verb channel of the tape recorder an inaudible
of the idiomatic expressions, whereas the pulse of 1000 Hz was placed at the offset of
longest fragment included all but their last one word in each sentence. In the case of
word. Each list was given to 15 subjects, the experimental sentences, the pulse oc-
who were instructed to make a sentence out curred at the end of the last word in the
of the fragments. This procedure allowed idiomatic strings, whereas in the case of the
us to select 12 expressions that were not filler sentences it was placed in such a way
completed idiomatically by more than 5% as to cover the entire range of positions in
of the subjects and were therefore suitable the sentences. The pulse caused a visual
for the purpose of Experiment 2. word to be displayed on the screen of the
676 CACCIARI AND TABOSSI

same microcomputer used in the previous (F(2,22) = 3.56, MSe 2882; p < .05). Min
experiment for 1500 ms and started the F’(2,47) = 2.31 was not significant.
timer as in Experiment 1. Planned nonorthogonal comparisons
Three sets of 72 visual targets to be showed that the responses to literal target
paired with the list of sentences were also words were faster than those to idiomatic
prepared. The fdler words and nonwords target words, which responses were not re-
were the same in the three sets. While in liably faster than those to the control tar-
one set the visual target paired with an ex- gets (Literal vs. Idiomatic: F(1,52) = 5.17,
perimental sentence was the idiomatic one, p < .05; Idiomatic vs. Control: F(1,52) =
in another set it was the literal one, and in 0.9, n.s.).
the third set it was the unrelated control. In The present results are clearly different
each set, there was a total of 12 experimen- from those of Experiment 1. When there is
tal visual targets: 4 were idiomatic, 4 were no bias either in the sentence or in the idi-
literal, and 4 were controls. omatic string to suggest to a listener that
Each set, paired with the list of sen- (s)he is being presented with an idiom, the
tences, was randomly assigned to an equal figurative interpretation is not automati-
number of subjects, who acted as their own cally activated, even at the end of the
controls. string. Thus, these data are counter to both
Procedure. The procedure was the same the direct accesstheory and the lexical rep-
as that in Experiment 1. The only differ- resentation theory. The data support the
ence was that halfway through the experi- view that, at least with idioms of this type
mental session there was a brief rest inter- and without contextual cues, idiomatic
val. Six subjects failed to reach the thresh- strings are initially processed only literally.
old in the recognition test and were not Obviously, if this conclusion is correct, one
considered in the subsequent analyses. could expect that sometime after the last
Among the 27 subjects who reached the word in the string-i.e., after the word has
threshold, the mean level of correct recog- been retrieved and has failed to be inte-
nitions was 74.9%. grated in the preceding context-its idiom-
atic interpretation, which has not been ob-
Results and Discussion served at the exact end of the idiom, should
The mean percentage of errors was 2.5. be detected. In order to test this prediction,
In order to reduce variability, data points -+ Experiment 3 was designed.
2 SDS from the mean RTs of each subject
(3.09% of responses) were excluded from EXPERIMENT 3
analyses. Table 2 shows the mean RTs and
SDS of the correct responses under the Experiment 3 was identical to Experi-
three experimental conditions. The main ment 2, except that visual target words
factor was significant both in the analysis were presented to subjects 300 ms after the
by subjects (F(232) = 6.60, MSe 2540; p < last word of the idiom. This delay has been
.Ol) and in the analysis by materials shown to be long enough to allow people to
perform postperceptual, integrative tasks
TABLE 2 such as resolving lexical ambiguity (Seiden-
MEAN REACTION TIMES AND STANDARD berg, Tanenhaus, Leiman, dz Bienkowsky,
DEVIATIONS (ms) FOR LEXICAL DECISION TO 1982) or interpreting unambiguous words
TARGET WORDS IN EXPERIMENT 2 (Whitney, McKay, Kellas, & Emerson,
Idiomatic Literal Control 1985). It is therefore reasonable to assume
that after a delay the responses to idiomatic
676 (97) 644 (72) 694 (95)
targets, which in Experiment 2 failed to be
COMPREHENSION OF IDIOMS 677

facilitated, should be faster than those to The present findings suggest that when
their controls. no cues are provided by the context and
Subjects. A total of 34 undergraduates there is no bias toward the idiomatic com-
(26 women and 8 men) volunteered for the pletion of a string, some time is required
experiment, which lasted about 25 min. before the meaning of the idiomatic expres-
Four subjects failed to meet the require- sion can be activated. Indeed, this activa-
ments of the recognition test (see above), tion was observed 300 ms after the end of
and their data were not analyzed. the idiom. It is noteworthy, however, that
Materials and design. The materials and the visual word related to the meaning of
design were the same as those used in Ex- the last word in the string is still activated.
periment 2, except that the lOOO-Hzpulse Evidently, contrary to Gibbs, not only is
placed on the second channel of the tape there a time at which the comprehension of
recorder that was delayed by 300 ms. an idiomatic interpretation is carried out,
Procedure. The procedure was the same but interpreting the expression idiomati-
as that in Experiments 1 and 2. Among the cally is a rather slow process. In fact, the
subjects who reached the threshold in the literal meaning of one of its elements is still
recognition test, the mean percentage of active at 300 ms, i.e., at least 100 ms longer
correct responses was 73.09. than the time usually required to resolve
lexical ambiguity (Seidenberg et al., 1982).
Results and Discussion
GENERAL DISCUSSION
The mean percentage of errors was 2.4.
In order to reduce variability, data points + With predictable idioms, only the idiom-
2 SD from the mean RTs of each subject atic meaning was activated by the time the
(3.05% of responses) were excluded from last word of an idiomatic string was en-
analyses. The mean RTs and SDSof correct countered (Experiment 1). With nonpre-
responses under the three experimental dictable idioms, only the literal meaning of
conditions are reported in Table 3. that last word was available (Experiment 2)
There was a significant main effect both until some time later (Experiment 3) when
in the analysis by subjects (F(2,58) = 6.09, both idiomatic and literal meanings were
MSe 2333; p < .Ol) and in the analysis by activated. These findings are inconsistent
materials (F(2,22) = 3.73, MSe 2212; p < with either the lexical representation or the
.05). Min F’(2,50) = 2.31 was not signiti- direct access models of idiom comprehen-
cant. Planned nonorthogonal comparisons sion.
showed that responses to both idiomatic These findings rule out the possibility
and literal targets, which did not differ from that the results could be due to the experi-
each other, were faster than those to the mental task-lexical decision- and do not
controls (Idiomatic vs. Literal: F(1,58) = reflect how idioms are processed by people
0.58, ns.; Idiomatic vs. Control: F(1,58) = during language comprehension. In fact,
6.18, p < .02; Literal vs. Control: F(l,58) = the data observed in Experiment 2 are very
11.89; p < .005). difficult to account for on the basis of post-
perceptual strategies to which lexical deci-
TABLE 3 sion is often thought to be sensitive. More-
MEAN REACTION TIMES AND STANDARD over, the different pattern of results ob-
DEVIATIONS (ms) FOR LEXICAL DECISION TO tained using the same paradigm and either
TARGET WORDS IN EXPERIMENT 3
identical or comparable materials suggests
Idiomatic Literal Control that the task is indeed sensitive to the var-
ious experimental conditions.
630 (86) 618 (105) 661 (119)
If these findings reflect the way in which
678 CACCIARI AND TABOSSI

people understand idioms-at least the tion recognizable. The individual words
class of idioms considered here-they do that participate in the configuration are the
not support current models. In particular, same lexical items that are accessed ordi-
Experiment 2 failed to show initial activa- narily during comprehension. But, in addi-
tion of the idiomatic meaning of the string, tion, they take part in all the configurations
and Experiment 3 showed late activation of representing idiomatic meanings. Take, for
both idiomatic meaning and of the literal instance, is a lexical entry which is acti-
meaning of its last word. According to the vated in order to understand a sentence like
direct access hypothesis, the idiomatic in- “The boy took the book”; but it is also part
terpretation should always be faster than of configurations such as “take the bull by
the literal interpretation. According to the the horns, ” “take to heart.”
lexical representation hypothesis, the idi- One can hypothesize, analogous to what
omatic interpretation of the string that runs happens in word recognition (Marslen-
parallel to the literal interpretation could be Wilson, 1987), that a configuration, i.e., an
available at the end of the string since no idiom, cannot be recognized before a cer-
structural analysis but only its retrieval tain amount of information has been re-
from the mental lexicon is required. Nei- ceived. But how can this amount be speci-
ther is there any plausible explanation for tied?
the presence of the literal interpretation of At least those idioms that are verbal
the last word of the string 300 ms after the phrases seem to be semantically structured
end of the idiom. Thus, our results support such that some parts in the idiomatic string
neither the direct access theory nor the lex- are more relevant than others for detecting
ical representation theory. the idiom (Cacciari & Stock, 1987). Let us
Most of the available empirical evidence suppose as a working assumption that ev-
did not discriminate between these two hy- ery VP idiom has such a part, which we will
potheses. However, earlier research has refer to as the key of the string, and that the
ruled out the view that people perform a string cannot be recognized until its key has
literal analysis of an idiomatic string before been accessed from the mental lexicon.
they initiate a nonliteral interpretation. One Whether the recognition of the idiom oc-
point established by the present study is curs immediately after the key has been de-
that such evidence does not reflect the on- tected can depend upon a number of fac-
line process of comprehension; rather it re- tors: whether the key occurs early or late in
flects the outcome of that process. Hence, the string and whether the string occurs in a
it can hardly clarify how people come to context that biases the figurative interpre-
understand idioms while understanding a tation of the string. We do not necessarily
sentence. On balance, the somewhat com- imply that the key is fixed through all pos-
monsense hypothesis that the literal inter- sible contexts. We simply assume that in
pretation of a string is attempted first and the absence of a context which could antic-
the idiomatic meaning activated when this ipate the recognition of an idiom (much like
fails seems too simplistic given the results a context can anticipate the identification of
of Experiment 1. a word) the recognition of the idiomatic
An alternative hypothesis is that idioms configuration cannot take place until after
are not encoded as separate entries in the the key has been accessed. According to
mental lexicon. Rather, their meaning is as- this hypothesis, a configuration is made up
sociated with particular configurations of of the same lexical items that need to be
words and becomes available-in lexical activated during the comprehension of lit-
processing terms, is accessed-whenever eral discourse. So one claim of the present
sufBcient input has rendered the conflgura- view is that there is only one processing of
COMPREHENSION OF IDIOMS 679

an idiomatic string. This is literal until, dicts that the interpretation that will be
sometime after the activation of its key, the available first depends upon how early the
configuration emerges. At this point, any idiomatic configuration becomes available.
remaining lexical items in the string may If the identification point of an idiom occurs
not be literally processed. early in the string, by the end of the string
A crucial issue is the specification of the only the figurative interpretation should be
recognition point of an idiom, i.e., the iden- available. If the identification does not oc-
tification of its key. Unfortunately, no log- cur until the end of the string, then the lit-
ical criterion such as the uniqueness point eral interpretation should be observed be-
recently proposed for word identification fore the nonliteral one emerges.
(Marslen-Wilson, 1987) seems to apply to Third, one problem related to the view
idioms. However, one way of establishing that idioms are mentally represented as
this point is to ask independent judges to morphologically complex lexical items is
complete ever increasing idiomatic frag- the fact that they can undergo syntactic
ments in a sort of gating task (Grosjean, processing (Peterson, Burgess, Dell, &
1980). Eberhard, 1987). In our approach, idioms
Though admittedly sketchy, this model are processed like any other string and a
could account for the present results. When configuration can be activated regardless of
the key is the last word in a string, as in which part occurs first. Hence the present
Experiment 2, this word must be accessed view also accounts for the syntactic parsing
before the idiomatic configuration can be- of idioms.
come detectable. In fact, we found a facil- Within the present framework, the fact
itation effect only for the literal target re- that some or all of the lexical items in an
lated to the last word. Only after this point idiom are activated and its key is always
can the idiom be identified, as suggestedby activated, easily allows for aspects of the
Experiment 3. However, when the key oc- meaning of those items being available
curs earlier in the string (Experiment l), the while understanding the idiom.
last word may not even be accessed, al- Needless to say, this view has a number
though some information from it such as its of problems. One of the greatest is related
initial phonemes is probably extracted to to the notion of a key of an idiom. So far,
allow the identification of the idiom. we have used this notion as if (a) it were a
In addition to accounting for our data, single content word and (b) there were only
this model has a number of interesting fea- one in each idiomatic string. While both as-
tures. First, each word is represented in the sumptions were made for the sake of sim-
lexicon only in one form and need not be plicity (and indeed they held in Experi-
marked as literal or idiomatic. Particularly ments 2 and 3) they need not be true in
with words such as be, take, set, in, and general On many occasions an idiom be-
out, this seems to be a characteristic distin- comes identifiable after a preposition, an
guishing this approach from previous article, or the lack of or the unusual occur-
views. rence of one such lexical item. Moreover,
Second, even though we have treated there are idioms for which it is difficult to
only verbal idioms, the present approach establish which word plays the role of the
seems to be potentially more general than key, the intuition being that what is crucial
prior models. It provides a general treat- is the cooccurrence of those words (e.g.,
ment for idioms both with and without a “kick the bucket”). In other cases there
literal interpretation. Both kinds are taken seemsto be more than one such word, as in
to be represented and processed in the “take the bull. . . .” Obviously, if the no-
same way. For those expressions which do tion of a key is going to be of any use in the
have both interpretations, this view pre- comprehension of idioms, a much fuller
680 CACCIARI AND TABOSSI

specification of it than the crude character- the way in which idioms can be recognized
ization provided here is necessary. depending upon how early in the string they
A further issue that requires more de- become identifiable. That this factor should
tailed attention is the notion of configura- be relevant is hardly surprising if one con-
tion. Clearly, in our framework, there is a siders what is being done in the related field
distributed representation of the idioms in of word recognition. However, no attention
the mental lexicon (Hinton, McClelland, & has been given to this important aspect by
Rumelhart, 1986). We assume, as was men- current psychological research on idioms.
tioned before, that activation of the config- The importance of the experimental para-
uration can be initiated from any of its ele- digm used is clearly shown by the present
ments. This has the clear advantage of eas- findings.
ily accounting for the syntactic flexibility of
many idioms. However, what is not clear is APPENDIX
how our model can account for the fact that Listed below are the test sentences and targets (id-
listeners know the syntactic transforma- iomatic, literal, and control) for Experiments 1 and 2 in
tions of familiar idioms. Whereas some Italian, translated word-by-word into English and with
nonpermissible transformations do not pre- the meaning or the corresponding English form of the
Italian idiom. In most cases, there is no English equiv-
vent the listeners from detecting the idiom, alent, so in order to respect the form of the idiomatic
some do, as in “the bucket was kicked by expression in Italian-in the original version-the
John,” where the passive form appears translation into English is necessarily stilted. The En-
more suggestive of a literal interpretation. glish version was used in the text.
Finally, this study has not taken into ac-
count how context affects the recognition Experiment 1
of an idiom. The well-known role of context 1. I1 ragazzo pensava the sue fratello fosse nato con
in literal language comprehension is proba- la camicia.
bly more crucial with idioms. In fact, what -FORTUNATO
is peculiar about these expressions is that if -CRAVATTA
taken literally they sometimes make no SORGENTE
sense but, nevertheless, they do fit in an The young man thought that his brother was born
appropriate context. It is very likely there- with the shirt. (was born with a silver spoon in his
fore that the detection of this incongruency mouth)
can play a role in the identification of an -LUCKY
idiomatic expression (R. R. Peterson, per- -TIE
sonal communication). -SPRING
In conclusion, although the hypothesis il- 2. Il commesso inform6 il collega the il principale
lustrated above is incomplete and tentative, era andato all’altro mondo.
it accounts for data that would be very dif- -MORTO
ficult to explain within the standard mod- -UNIVERSO
els. Furthermore, it possessesa number of -0ROLOGIO.
interesting features that make future re- The clerk informed the colleague that the boss had
search worthwhile. The present study sug- gone to another world. (had died)
gests that experimental results which seem -DEAD
to show that the idiomatic interpretation of -UNIVERSE
a string is easier or as easy to accessas its -WATCH
literal interpretation are largely due to the 3. 11 malato di mente era convinto the i medici gli
inadequacy of the experimental paradigms avrebbero fatto la pelle.
used. When a technique more likely to tap -UCCISO
on-line processing is used, a pattern of re- SUDORE
sults emerges that suggests differences in -NEGOZIO
COMPREHENSION OF IDIOMS 681

The mentally ill person was convinced that the doc- The actor asked the producer to put in a good word
tors would have done him the skin. (would have killed for him with the director.
him) -FAVOR
-KILLED -LANGUAGE
-SWEAT -GLOVE
-SHOP
9. 11bambino protestava perch6 non voleva andare
4. La moglie rimprover il mar&o di aver fatto un a letto con le galline.
buco nell’acqua.
-PRESTO
-FALLITO -0CHE
-FONTE -SAPONE
-PARRUCCA
The child protested because he didn’t want to go to
The wife reproached her husband for having made a bed with the hens. (early)
hole in the water. (for having been unsuccessful)
-EARLY
-FAILED -GOOSE
-FOUNTAIN -SOAP.
-WIG
5. Dopo l’ottima prestazione, il tennista era al set-
timo cielo. Experiment 2
-FELICE
STELLE 1. Appena saputi i risultati il candidato liberale ac-
-RISPETTO cus6 il colpo, inaspettato, della perdita delle elezioni.
After the excellent performance, the tennis player
SCONFITTO
was at the seventh sky. (in seventh heaven) -FUCILE
-HAPPY -BUCO
-STARS
As soon as the candidate knew the results, he ac-
-RESPECT
cused the stroke (he resigned himself to the) unex-
6. I1 giovane disse alla tidanzata the non vedeva piu pected of the loss of the election.
in la del suo naso.
-DEFEATED
-0TTus0 -RIFLE
-0RECCHIE -HOLE
-MARE
2. 11regista telefon6 al produttore per chiedergli di
The young man told his girlfriend that she couldn’t venire a cape, repidamente, de1 problema.
see beyond the end of her nose.
-ESITO
-STUPID -cOLLO
-EARS -PENNE
-SEA
The director called up the producer asking him to
7. 11 direttore si lagn6 con la segretaria the aveva come to the head (to decide about) quickly of the
fatto un lavoro coi piedi. problem.
-MALE -RESOLUTION
SCARPE -NECK
-0RTO -FEATHERS
The manager complained to the secretary who did a 3. A causa di diverse diBicolta, era andato a monte
job with the feet. (badly) definitivamente l’intero progetto.
-BADLY -FALLITO
-SHOES -VETTA
-GARDEN -CHIOMA
8. L’attore chiese al produttore di aiutarlo co1 Because of several troubles, the whole project def-
regista mettendo una buona parola. initely has gone to the mountain. (has failed)
-FAVORE -UNSUCCESSFUL
-LINGUAGGIO -PEAK
-GUANTO -MANE
682 CACCIARI AND TABOSSI

4. Quando fu letto il testamento, gh eredi si accor- -HORNS


sero the era stat0 dato fondo, ormai, a tutto il patri- -TROUT
monio. 9. Per l’ennesima volta, il ragazzo cadde in piedi.
-FINITO
-SALVO
-BO’ITIGLIA
SCARPE
-BRODO -CASTAGNA
When the testament was read, the heirs realized that
Once again, the boy landed on his feet.
all the assets had been given bottom. (depleted)
SAVE
-FINISHED -SHOES
-BOTTLE
-CHESTNUT
-BROTH
10. La maestra si accorse the hhnno stava s&e
5. Durante uno dei sohti htigi, la donna accus6 il
sue.
marito di fare lo struzzo.
-TIMID0
-FUGA -PROPRIETA
-uovo -LINEA
-0ROLOGIO
The teacher realized that the pupil stayed on his
During one of the usual quarrels, the woman ac-
own. (the pupil kept to himself)
cused her husband of making the ostrich. (to play os-
trich) -SHY
-PROPERTY
-ESCAPE -LINE
-EGG
-WATCH 11. Tutti sapevano the l’uomo si era fatto in quat-
tro, pur di fare studiare il figho.
6. La Signora non mar@ la torta perch6 aveva la
luna. -AIUTO
-NUMERO
-NERVOSO -0RECCHIE
--CIELO
-VESTITO Everybody knows that the man made himself in four
(the man did everything he could) to have his son
The woman didn’t eat the cake because she had the study.
moon. (she was in a bad mood)
-HELP
-NERVOUS -NUMBER
-SKY -EARS
-DRESS
12. I parenti gli dissero the la moglie era restata di
7. L’impiegato non and6 via soltanto perch6 era al sale dopo la brutta notizia.
Verde.
-MALE
-DENARO -MARE
-ERBA -FONTE
-BOlTA
His relatives told him that his wife was left of salt
The clerk didn’t go on holiday only because he was (his wife was struck dumb) after the horrible news.
at the green. (was broke)
-BAD
-MONEY -SEA
-GRAss SPRING.
-BLOW
8. La ragazza si decise a mandate al diavolo, una REFERENCES
voha per tutte, il tidanzato.
BOATNER, M., GATES, J., & MAKKAI, A. (1975). Dic-
-LONTANO tionary of American Idioms. New York: Baron’s
-CORNA EducationalSeries.
-TROT A BOBROW, S., & BELL, S. (1973).On catchingon to
The girl decided to tell her boyfriend to go to the idiomaticexpressions.Memory & Cognition, 1,
343-346.
devil, once and for all.
CACCIARI,C., & STOCK, 0. (1987).Processingidioms
-AWAY without buildingcastlesin the air. Manuscript
COMPREHENSION OF IDIOMS 683

submitted for publication. Estill, R., & Kemper, PETERSON, R. R., BURGESS, C., DELL, G. S., & EBER-
S. (1982). Interpreting idioms. Journal ofPsycho- HARD, K. (1987). Syntactic analysis during idiom
linguistic Research, 6, W-568. processing. Paper presented at the 28th Annual
FRASER, B. (1970). Idioms within a transformational Meeting of the Psychonomic Society, Seattle,
grammar. Foundations of Language, 6, 1,2242. 1987.
GIBBS, W. R. (1980). Spilling the beans on understand- RUWET, N. (1983). Du bon usage des expressions id-
ing and memory for idioms in context. Memory & iomatiques. Recherches Linguistiques, 11, 5-&1.
Cognition, 8, 149-156. SEIDENBERG, M. S., TANENHAUS, M. K., LEIMAN,
GIBBS, R. (1986). Skating on thin ice: Literal meaning J. M., & BIENKOWSKI, M. (1982). Automatic ac-
and understanding idioms in conversation. Dis- cess of the meaning of ambiguous word in context:
course Processes, 9, 17-30. Some limitations to knowledge-based processing.
GROSJEAN, F. (1980). Spoken word recognition pro- Cognitive Psychology, 14, 489-537.
cesses and the Bating paradigm. Perception & Psy- SEIDENBERG, M. S., WATERS, G. S., SANDERS, M.. &
chophysics, 28, 267-283. LANGER, P. (1984). Pre- and postlexical loci of
HINTON, G. E., MCCLELLAND, J. L., & RUMELHART, contextual effects on word recognition. Memory
D. E. (1986). Distributed representations. In J. M. & Cognition, 12, 315-328.
McClelland, D. E. Rumelhart, & the PDP Re- SWINNEY, D., & CUTLER, A. (1979). The access and
search Group (Eds.), Parallel Distributed Pro- processing of idiomatic expressions. Journal of
cessing: Vol. 1. Cambridge, MA: MIT Press. Verbal Learning and Verbal Behavior, 18, 523-
KATZ, L., & FELDMAN, L. B. (1983). Relations be- 534.
tween pronunciation and recognition of printed SWINNEY, D. A., ONIFER, M., P~ATHER, P., & HIR-
words in deep and shallow orthographies. Journal
SHKOWITZ, M. (1979). Semantic facilitation across
of Experimental Psychology: Learning, Memory
sensory modalities in the processing of individual
and Cognition, 9, 157-166.
words and sentences. Memory di Cognition, 7,
MARSLEN-WILSON, W. D. (1987). Functional parallel- 159-165.
ism in spoken word recognition. Cognition, 25,
TWEEDY, J. R., & LAPINSKY, R. H. (1981). Facilitat-
309/310, 71-102.
ing word recognition: Evidence for strategic and
MARSLEN-WILSON, W. D., & TYLER, L. K. (1980). automatic factors. The Quarterly Journal of Ex-
The temporal structure of spoken language com-
perimental Psychology: Human Experimental
prehension. Cognition, 8, l-71.
Psychology, 33(l), 51-60.
MUELLER, R. A. G., & GIBBS, R. (1987). Processing
idioms with multiple meanings. Journal of Psy- WHITNEY, P., MCKAY, T., KELLAS, G., & EMERSON,
cholinguistic Research, 16(l), 63-81. W. A., JR. (1985). Semantic activation of noun
ORTONY, A., SCHALLERT, D., REYNOLDS, R., & AN- concepts in context. Journal of Experimental Psy-
chology: Learning, Memory, & Cognition. 11,
TOS, S. (1978). Interpreting metaphors and idioms:
Some effects of context on comprehension. Jour- 126-135.
nal of Verbal Learning and Verbal Behavior, 17, (Received December 21, 1987)
465-478. (Revision received May 13, 1988)

View publication stats

You might also like