Language Learning 51:2, June 2001, pp. 257–280
Resource Consumption as a Function
of Topic Knowledge in Nonnative
and Native Comprehension
Michael D. Tyler
University of Western Sydney
Previous research suggests that low-level receptive language processes are less developed for nonnative than for
native listeners, yet experienced nonnatives seem to comprehend effortlessly in everyday situations. One possible
explanation is that experienced nonnatives use topic
knowledge to reduce working memory (WM) requirements.
Native and experienced nonnatives attended to Bransford
and Johnson’s (1972) Washing Text while performing a
concurrent task, with half of each group given the topic of
the passage. Scores on the concurrent task were compared
with baseline to index WM consumption. The results
showed a relatively greater WM consumption for nonnatives than natives when the topic was unavailable, suggesting that nonnatives rely more than natives on topic
knowledge in comprehension. Implications for foreign language learning are discussed.
Michael D. Tyler, Macarthur Auditory Research Centre, Sydney.
I would like to thank Dr. Kate Stevens and Prof. Denis Burnham, from the
Macarthur Auditory Research Centre, Sydney, for their helpful comments on
earlier drafts of this paper, and Bruno Di Biase from the University of Western
Sydney’s Language Acquisition Research Centre for his assistance in recruiting volunteers for the study.
Correspondence concerning this article may be sent to Michael D. Tyler,
Macarthur Auditory Research Centre, Sydney, University of Western Sydney,
Locked Bag 1797, South Penrith Distribution Centre, South Penrith, New
South Wales, 2560, Australia. Internet: [email protected]
257
258
Language Learning
Vol. 51, No. 2
There are many people who live in an environment where
their native language is not spoken and who conduct their daily
business using a nonnative language. There has been a great deal
of research examining the nonnative speech of these individuals,
but far less on their receptive language capabilities. Most researchers studying second language (L2) acquisition agree that if
people start learning a language as an adult, their accent will be
clearly distinguishable from that of a native speaker (e.g., Flege,
Yeni-Komshian, & Liu, 1999; Scovel, 1988). However, research in
the area of L2 speech perception has also led to the idea that
nonnatives may not only speak with an accent, but also hear with
an accent (Jenkins, Strange, & Polka, 1995). For example, there is
a large body of literature demonstrating that native Japanese
listeners have poorer discrimination of English /r/ and /l/ than
native English listeners (e.g., MacKain, Best, & Strange, 1981;
Sheldon & Strange, 1982), even after 12 years of naturalistic
exposure to English (Takagi & Mann, 1995). One possible explanation for this finding is that the low-level processes involved in
the perception of these phonemic contrasts are less developed for
nonnative than for native listeners.
The results of several other studies support the notion that
nonnative low-level processes are less developed than native processes. For example, using grammaticality judgement tasks,
Johnson, Shenkman, Newport, & Medin (1996) showed that nonnative speakers have an indeterminate grammatical system. Emmorey and Corina (1990) administered a gating task to late
learners of American Sign Language and to native signers, and
they found that native signers were able to guess the identity of
the sign significantly earlier than nonnative signers. They argued
that this demonstrated differences in processing at the early
stages of lexical recognition, perhaps reflecting differences in
processing capacity or in the structure of lexical representations.
Mayo, Florentine, and Buus (1997) administered the Speech Perception in Noise test to monolingual English speakers and to early
and late bilingual Mexican-Spanish speakers. They found that
English speech was intelligible at higher levels of noise for
Tyler
259
monolinguals and early bilinguals than for late bilinguals, which
is also consistent with the idea that nonnative listeners have less
developed low-level language processes than native listeners.
If it is the case that nonnative low-level processes are less
developed than native low-level processes, then one might expect
some difficulty in comprehending speech in the nonnative language. The concept of difficulty is accounted for by theories of
attention. For example, Navon and Gopher (1979) defined difficulty as “the subjective feeling of strain accompanying involvement in demanding tasks” (p. 215). They stated that when a task
is demanding, larger amounts of mental resources are required to
perform the task. With practice and experience, performance
becomes automatic rather than controlled, the amount of required
resources decreases, and the task becomes less demanding
(Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977). A more
recent theory that deals specifically with resource consumption in
language comprehension is Just and Carpenter’s (1992) theory of
working memory (WM). According to this theory, human comprehension ability is constrained by a pool of WM resources that fuels
both the computation and storage of information. Difficulty in
comprehension occurs when the amount of available WM is too
small to cope with the demands of storage and computation. In the
case of foreign language listening, at the beginning stages of
learning, attention is focused on low-level processes such as phoneme discrimination and word recognition. Large amounts of WM
resources are required for these unfamiliar tasks, so performance
is difficult. The negative effect of this difficulty on motivation and
the associated increase in anxiety is well documented (MacIntyre
& Gardner, 1991; MacIntyre & Gardner, 1994; Schumann, 1998).
According to theories of attention, with practice and experience,
performance in all of these domains should become automatic, and
fewer WM resources should be required for L2 listening. Indeed,
in everyday conversation, experienced nonnatives living and
working in nonnative language environments appear to experience no more difficulty in comprehension than native listeners.
However, if low-level receptive language processes are less
260
Language Learning
Vol. 51, No. 2
developed for nonnative listeners than for native listeners, then a
larger resource allocation would be expected for nonnatives than
natives during language comprehension. Therefore, there seems
to be a gap between observed performance in the real world and
the predicted performance from experimental results.
One possible factor contributing to the seemingly effortless
comprehension abilities of experienced nonnatives is topic knowledge. That is, the topic of conversation can provide cues to meaning. For example, Bransford and Johnson (1972) showed, with
native listeners and written stimuli, that the recall of information
from vague passages could be improved by the provision of a
relevant picture or topic. For example, in one experiment they
presented the following passage:
The procedure is actually quite simple. First you arrange
things into different groups. Of course, one pile may be
sufficient depending on how much there is to do. If you have
to go somewhere else due to lack of facilities that is the
next step, otherwise you are pretty well set. It is important
not to overdo things. That is, it is better to do too few things
at once than too many. In the short run this may not seem
important but complications can easily arise. A mistake
can be expensive as well. At first the whole procedure will
seem complicated. Soon, however, it will become just another facet of life. It is difficult to forsee any end to the
necessity for this task in the immediate future, but then
one never can tell. After the procedure is completed one
arranges the materials into different groups again. Then
they can be put into their appropriate places. Eventually
they will be used once more and the whole cycle will then
have to be repeated. However, that is part of life. (p. 722)
Participants who were not given the topic of this passage were able
to recall less information than those who were given the topic,
which was “Washing Clothes.”
Similar findings have been reported for students learning a
foreign language (see Schmidt-Rinehart, 1994, for a review). Students are able to recall more information from foreign language
passages when the topic is familiar than when it is unfamiliar. For
Tyler
261
example, Markham and Latham (1987) presented Muslim and
Christian students with novel foreign language texts describing
religious practices from these faiths. They found that the Muslim
students were able to recall more information from the Muslim
texts, and the Christian students could recall more information
from the Christian texts.
Rather than recruiting foreign language students from different backgrounds, Wolff (1987) used Bransford and Johnson’s
(1972) Balloon Story to manipulate access to the topic directly.
Like the Washing Text, the Balloon Story is vague, but when a
contextual picture was provided, the passage was easy to follow.
Wolff presented the Balloon Story to foreign language students,
with half also given the contextual picture to aid comprehension.
The results replicated Bransford and Johnson’s findings, because
superior recall was obtained when the picture was provided compared to when there was no picture. Wolff interpreted these
results, and the results of similar tasks using other passages, in
terms of information processing theory. He argued that the
bottom-up processes that nonnative listeners use to decode the
speech signal are underdeveloped, so they must resort to top-down
processing to achieve comprehension. In other words, when the
topic is not known, nonnative listeners need to rely on their
underdeveloped low-level processes for comprehension, but if the
topic is available, then knowledge about the world can be used to
aid comprehension, and less information is required from the
speech signal. In terms of working memory, if the topic is known,
then nonnatives need to rely less on low-level processes for comprehension, leaving more resources for higher-level comprehension processes.
Therefore, it is possible that the seemingly effortless comprehension performance of nonnative listeners in everyday situations
is due to knowledge of topic. That is, if nonnatives “hear” with an
accent, then topic knowledge may be more beneficial for nonnative
comprehension than for native comprehension. The aim of the
present study is to ascertain the effect of topic knowledge on the
allocation of WM resources in experienced nonnative comprehension
262
Language Learning
Vol. 51, No. 2
compared with native comprehension. It is predicted that the
difference in the availability of WM resources between when the
topic is given and when no topic is given will be greater for
nonnatives than for natives.
Before describing the task used here for measuring WM
availability during comprehension, it is necessary to address some
theoretical issues. The first section will justify the selection of
nonnative participants for the study and the second section will
describe a recent theory of comprehension that is compatible with
Just and Carpenter’s (1992) theory of WM.
Experience Versus Proficiency in the Selection of Nonnative Listeners
In my description of the target nonnative group for the
present study, I have used the term “experienced,” rather than
“proficient,” the latter being more common in second language
acquisition research. In a review of the assessment of nonnative
language proficiency, Thomas (1994) identified four main methods:
Impressionistic judgement, institutional status, in-house assessment instruments, and standardized test scores. The use of standardized test scores would be the most appropriate of these four
for the present study if proficiency were the criterion of selection.
However, these tests often suffer from the “comparative fallacy”
(Bley-Vroman, 1983), owing to their inherent comparisons with
native-speaker norms. That is, it is possible that the outcome of
nonnative language acquisition is qualitatively different from
native language acquisition, so a bias may be introduced if certain
nonnatives are excluded because they fail to perform at nearnative levels on language tests (Bley-Vroman, 1989, 1990). The
distinction between native and nonnative is also controversial,
because the issue of maturational constraints on language acquisition is still a matter of debate (e.g., Birdsong, 1999; Flege et al.,
1999). Within this literature, there appears to be a consensus that
the outcome of nonnative language acquisition after the age of 15
is different from early nonnative language acquisition and from
Tyler
263
native language acquisition, but there is disagreement about the
source of the difference.
Both controversies can be avoided by using the criteria of
Johnson et al. (1996) to select proficient nonnatives. Their nonnative participants were not exposed to English before the age of 10,
arrived in the United States after the age of 15, and had been
residing there for at least 5 years. Furthermore, all participants
were graduate students who were exposed to English on a daily
basis. According to Johnson et al., individuals who conform to
these criteria have reached an end-state in their language acquisition, such that their proficiency, relative to native speakers, will
no longer improve or deteriorate. The present study recruited
nonnative listeners who conformed to these criteria, because both
of the controversies would thus be avoided. The nonnative participants of Johnson et al. were all older than 15 when immersed in
an English speaking environment and so were outside the influence of maturational constraints, and comparison with native
speaker performance was not required, because a criterion of
experience rather than proficiency was used.
Comprehension as the Formation of a Mental Model
Comprehension is not a unitary concept. As Glenberg, Kruley,
and Langston (1994) have pointed out, we can talk of comprehending words, phrases, sentences, or discourse. Furthermore,
comprehension processes are not necessarily specific to language, because people can also comprehend the physical and
social worlds through other sensory observations. The present
study is concerned with the comprehension of everyday spoken
language, that is, language that occurs at the discourse level.
Comprehension gained from other sources therefore needs to be
carefully controlled.
Recently, researchers have begun to examine comprehension
processes in more detail. Comprehension is now seen as tantamount
to the formation of a mental structure of a situation described in
language, or observed in reality. These mental structures are
264
Language Learning
Vol. 51, No. 2
known as situation models, or mental models (see Zwaan & Radvansky, 1998, for a review). Mental models are not the same as
Bartlett’s (1932) well known “schema” construct, which has been
used previously in L2 acquisition research (e.g., Anderson & Pearson, 1984; Carrell, 1984, 1987). Schemata are representations of
canonical or stereotypical situations, whereas mental models are
representations of specific situations in space and time. For example, knowledge of the general procedure involved in washing
clothes is represented by a schema, but if a person tells a story of
their experience doing the washing, then that specific situation is
represented by a mental model. So, although schemata are not
mental models, they can be used to aid in the construction of a
mental model by providing a framework of expectations for a
specific situation. If the listener possesses, for example, a schema
for washing clothes, then the storyteller needs to provide less
information for the listener to construct a coherent mental model.
Another property of mental models is that they are constrained by working memory (Glenberg & Langston, 1992; Zwaan
& Brown, 1996). If the processes required to form a coherent
mental model require more WM resources than are available, then
the reader or listener will experience difficulty. Furthermore, Just
and Carpenter (1992) state that a reduction in WM demands can
be achieved by the preactivation of relevant concepts, relations,
and schemas. In other words, the use of topic knowledge to activate
an existing mental model reduces WM demands. On the other
hand, if the reader or listener is not aware of the topic, then the
mental model must be constructed from the beginning, which
requires a far greater amount of WM.
An On-Line Task to Measure Resource Allocation in Comprehension
Bransford and Johnson (1972), using the Washing Text,
showed that access to topic can improve recall, so this passage will
also be used in the present study to manipulate access to topic.
However, the dependent measure in this, and in other studies that
have investigated the effect of topic knowledge on comprehension,
Tyler
265
has been recall of information. This measure does not necessarily
provide information about WM resource allocation during comprehension, so it is inappropriate for the present study. Furthermore,
the use of recall as a dependent variable is problematic, because
it necessarily follows comprehension and may involve processes of
construction or reconstruction (Trabasso & Suh, 1993). Any differences in recall performance could be attributable to problems at
the stages of encoding, storage, or retrieval of information. A more
appropriate task to measure WM resource allocation during comprehension is one which takes its measurements while comprehension is occurring. In other words, an “on-line” task is required.
Takano and Noda (1993) devised an on-line task to measure
the decline in the ability to perform a concurrent task during
foreign language use. Their participants completed double-digit
calculations on a sheet of paper, while answering questions at
regular intervals. They found that participants made less correct
calculations when answering questions in their nonnative language than in their native language. Takano and Noda (1995)
called this the “Foreign Language Effect.”
The question-and-answer paradigm used in Takano and
Noda’s (1993) study could be altered to accommodate the presentation of the Washing Text. However, one problem is that there is
no record of the timecourse of performance on the task, because
the responses to the calculations are recorded on a sheet of paper.
A more appropriate task would be a computer task requiring only
single keystrokes. Another problem is that Takano and Noda
(1993) clearly stated that the task does not measure nonnative
language processing difficulty per se, but only the decline in
thinking ability which accompanies it. For the present study, a
task is required that measures comprehension difficulty as directly as possible. This can be achieved by selecting a competing
task that requires WM resources, but only minimally involves
linguistic processing. One such task is the verification of singledigit calculations (De Rammelaere, Stuyven, & Vandierendonck,
1999; Lemaire, Abdi, & Fayol, 1996), in which participants are
presented with a sum (e.g., 2 + 4 = 6) and must indicate whether
266
Language Learning
Vol. 51, No. 2
the total is correct or incorrect. This task is also amenable to
computer presentation and requires only a single keystroke for
responding “correct” or “incorrect.”
Therefore, in the present experiment, the participants were
instructed to divide their attention equally between verifying the
totals of single-digit calculations and trying to remember facts
from the Washing Text, which was presented simultaneously
through headphones. Half of the participants in the native and
nonnative groups were presented with the topic of the Washing
Text. If the participants divided their attention equally between
the two tasks, then the number of correct calculations achieved in
the verification task will provide an indication of WM resource
consumption while participants were attending to the Washing
Text and remembering information for later recall. Even though
the results of the recall task were not used, it was important that
the participants be asked to recall the information to ensure that
they were attending to the passage. To overcome individual differences on the calculation verification task, the number of correct
verifications in the dual-task condition was subtracted from the
number of correct verifications in a baseline condition, in which
the participants performed a verification task on its own, to obtain
a “working memory index score.”
In summary, the experiment was a 2 × 2 (Language Background × Topic) between-subjects design with “WM index score”
as the dependent variable. It was hypothesized that some nonnative language processes might require effortful processing, so
nonnative listeners would use topic knowledge to assist their
comprehension more than would native listeners. If this was the
case, then an interaction between Language Background (Native/
Nonnative) and Topic (Topic/No Topic) was to be expected, with a
greater difference in WM consumption between nonnative and
native listeners when no topic is provided than when the topic is
provided.
Tyler
267
Method
Participants
There were 30 native Australian English listeners and 30
nonnative listeners in the study. The participants in each group
were randomly allocated to the Topic condition (Native: Mean
age = 32.1, SD = 14.5; Nonnative: Mean age = 44.2, SD = 11.4) or
the No Topic condition (Native: Mean age = 28.1, SD = 12.3;
Nonnative: Mean age = 38.9, SD = 10.8), with 15 participants in
each group. As the mean ages appear different, a 2-way ANOVA
was performed on the age data, which revealed a significant main
effect for Language Background (F(1,56) = 12.9, p < .05, ηp2 = .187),
but not for Topic (F(1,56) = 2.1), and no interaction (F(1,56) = 0.05).
This means that there was no significant age difference between
participants in the Topic and No Topic conditions, but there was a
significant overall age difference between the native and nonnative listeners. This significant main effect for age should not affect
the results of the experiment, because an interaction was predicted for the difficulty index scores between Language Background and Topic, and there is no significant interaction for age.
The volunteers were obtained from classes at the University
of Western Sydney, Australia, and from the general public by
word-of-mouth. All participants reported having normal hearing.
The nonnative listeners were all university students or professional people who used English on a daily basis, and they also
conformed to the nonnative speaker criteria of Johnson et al.
(1996). There were no significant differences between the nonnative Topic and No Topic groups in mean age of arrival in Australia
(t(28) = –.47, p > .05; Topic: M = 24.6, SD = 8.0; No Topic: M = 25.9,
SD = 6.8), or for the mean age at which the participants began
learning English (t(28) = .38, p > .05; Topic: M = 17.8, SD = 6.3; No
Topic: M = 16.9, SD = 6.1). However, there was a significant
difference between the groups on mean length of stay in Australia
(t(28) = 2.1, p < .05, d = .77; Topic: M = 19.5, SD = 9.9; No Topic:
M = 12.8, SD = 7.1). The nonnative participants came from a
268
Language Learning
Vol. 51, No. 2
variety of language backgrounds, as shown in Table 1. The random
allocation of participants into the Topic or No Topic condition
resulted in a slightly larger number of participants from Asian
language backgrounds in the No Topic condition (n = 7 of 15) than
the Topic condition (n = 3 of 15).
Materials and Apparatus
Bransford and Johnson’s (1972) Washing Text was recorded
by a native Australian English male speaker at a rate of 153.5
words per minute, for a duration of 68 seconds, and then digitized
at a sampling rate of 11.025 kHz. A practice passage was also
recorded, but digitized at a sampling rate of 7.35 kHz to lower
computer memory requirements. The practice passage consisted
Table 1
Distribution of nonnative language backgrounds in the Topic and
No Topic conditions
Condition
Language
Cantonese
Czech
Danish
French
German
Hungarian
Indonesian
Japanese
Khmer
Lebanese
Romanian
Serbian
Spanish
Vietnamese
Topic
No Topic
1
1
1
2
1
1
1
3
1
1
1
1
1
4
2
1
4
3
Tyler
269
of a series of sentences describing a country, but the name of the
country was not given, making it similar to the No Topic condition.
The calculation stimuli for the verification task were adapted
from Geary, Cormier, Goggin, Estrada, and Lunn’s (1993) study.
They imposed the following constraints on the stimuli to ensure
that the items were of equal difficulty: (a) The augends and
addends must be single digits, ranging in value from 2 to 9 (for
example, 10 + 1 = 11 is unacceptable, because both the augend [10]
and the addend [1] are outside this range); (b) No calculation may
contain an augend and an addend of the same value (for example,
3 + 3 = 6 is not an acceptable calculation); (c) Incorrect totals must
not deviate from the correct total by more than 2; (d) The pool of
calculations must not contain more than four consecutive calculations requiring the same response. The application of these constraints resulted in a pool of 56 correct calculations. Another list
of 56 calculations was constructed, where one quarter of the
calculations were each randomly allocated an incorrect total with
a deviation of –2, –1, +1, or +2. Three further lists were constructed, such that the distribution of incorrect deviations was
orthogonal. This resulted in 4 lists of 112 items, which is a greater
number of items than participants could complete in 68 seconds.
Two of these lists were used in the experiment, and another was
reduced to 20 correct calculations and 20 incorrect calculations for
the practice session.
The experiment was run with SuperLab software (Haxby,
Parasuraman, Lalonde, & Abboud, 1993) on a range of Macintosh
computers, ranging from LC to Power Macintosh 7300. To overcome any nonrandom variation due to screen refresh rates between computers, the presentation of each calculation was
preceded by a 200ms delay minus the presentation time. The
passages were presented through headphones, and the participants responded to the verifications by using the index fingers of
both hands, one for correct and the other for incorrect. A green tick
(check mark) was present on one side of the screen throughout the
experiment to indicate which hand to use for a correct response
and a red cross was present on the opposite side of the screen. The
270
Language Learning
Vol. 51, No. 2
location of the tick (left or right) was counterbalanced across
participants.
Procedure
There were four sections in the experiment: a practice session
for the calculation verification task, the single-task condition, the
practice passage, and the dual-task condition. First, the participants were given a practice session of 36 calculation verifications
to familiarize them with the task. Second, in the single-task
condition, participants had 68 seconds to complete the calculation
verifications as quickly and as accurately as possible. Third, the
same practice passage was presented to all participants, who were
told that the passage described a country. They were instructed to
try to remember as many ideas as possible from the passage to
write down at its conclusion. As recall was not the dependent
variable, the modality of responding was unimportant. The participants were aware that the passage was for practice only, so the
results were not analyzed. In the final section, the participants
were instructed to divide their attention equally between verifying
the calculations as quickly and as accurately as possible, and
trying to remember information from the Washing Text for later
recall. For participants in the Topic condition, the dual-task was
preceded by a sentence announcing that the passage would be
about washing clothes. To avoid overwhelming the participants,
the 68-second passage began after 4 calculations had been completed. At the conclusion of the passage, the participants were
asked to write down whatever they could remember from the
passage, but these responses were not analyzed, because of the
problems involved with using recall as a dependent variable.
Results
For each participant, the number of correct calculations completed in the dual-task condition was subtracted from the number
of correct calculations completed in the single-task condition to
Tyler
271
obtain the “WM index score,” where a larger score corresponds to
greater WM resource consumption in the dual-task condition. The
results are shown in Figure 1. A 2 × 2 (Topic × Language Background) factorial ANOVA revealed significant main effects for
Topic (F(1,56) = 6.0, p < .05, ηp2 = .10) and Language Background
(F(1,56) = 8.5, p < .05, ηp2 = .13) and a significant interaction
between Language Background and Topic (F(1,56) = 4.2, p < .05,
ηp2 = .07), showing that the difference in difficulty level between
Topic and No Topic was greater for nonnative listeners than for
native listeners.
It is also interesting to look at the difference between native
and nonnative listeners in the Topic condition. A t-test of this
comparison, with the alpha rate adjusted to .01 because the test
was post hoc, revealed that the difference was not significant
(t(28) = .68, p > .5), showing that the difficulty index score was the
same for natives and nonnatives in the Topic condition. Thus,
processing difficulty seems to be equivalent in the Topic condition
for native and nonnative listeners. There was also no significant
difference between the scores for native listeners in the Topic and
No Topic conditions (t(28) = .29, p > .7).
Figure 1. Working memory index scores for native and nonnative listeners
in the Topic and No Topic conditions.
Error bars represent 95% confidence intervals.
272
Language Learning
Vol. 51, No. 2
Extraneous factors that may have affected the results need
to be investigated. There was a difference between nonnatives in
the Topic and No Topic conditions on their length of stay in
Australia. However, there was no significant correlation between
difficulty index values and length of stay (r = –.07, p = .73),1 so the
difference between Topic and No Topic cannot be attributable to
this imbalance. There was also an imbalance of language background between the Topic and No Topic conditions. Takano and
Noda (1995) showed that the Foreign Language Effect was
enhanced by interlanguage dissimilarity, so the imbalance of
participants from Asian and European language backgrounds
may have affected the results. In the Topic condition, there were
11 participants from European language backgrounds, but only
3 participants from Asian language backgrounds (see Table 1), and
a t-test with these numbers would grossly violate the assumptions
of the t-distribution. On the other hand, in the No Topic condition,
where the difference between Topic and No Topic was most pronounced, the distribution of languages was balanced, with 7 people
from Asian language backgrounds and 8 from European language
backgrounds. A t-test was performed and no significant difference
was found (t(13) = –.426, p = .68). Therefore, the results of the
experiment can not be attributable to the imbalance of language
backgrounds between the nonnative participants in the Topic and
No Topic conditions.
Discussion
The results support the hypothesis. There was a significant
interaction between Language Background and Topic, demonstrating that the difference in WM consumption between the Topic
and No Topic conditions was greater for experienced nonnative
than native listeners. In other words, access to topic seems to have
had a small effect on the scores of native listeners, but a large effect
on the scores of nonnative listeners. From Figure 1, it can be seen
that there was only a small difference between natives and nonnatives in the Topic condition, and a post-hoc t-test showed that
Tyler
273
this difference was not significant. Thus, when the topic of the
Washing Text was available, the WM consumption of nonnatives,
as measured by the WM index score, was no different from that of
natives. On the other hand, when the topic was unavailable,
nonnative WM consumption was greater than that of natives.
These results are consistent with the idea that experienced
nonnatives not only tend to speak with an accent, but they also
hear with an accent (Jenkins et al., 1995). That is, the low-level
processes used by nonnative listeners to decode the speech signal,
which provide input for the formation of a mental model, are
either inappropriate for the target language, or not as efficient
as those used by native listeners. For example, the processes may
be underdeveloped, as suggested by Wolff (1987); they may consume more WM resources; or they may be qualitatively different
from native processes (Bley-Vroman, 1989, 1990). When the topic
was unknown, the listeners needed to rely solely on information
from the speech signal for comprehension, but when the topic was
available, less information was needed from the speech signal to
adapt an existing mental model to fit the described situation. That
is, even if nonnative language processes are not as efficient as
native processes, when the topic was available, nonnatives could
still attend to the passage using a similar amount of effort as
natives, because less information was needed from the speech
signal for the formation of a mental model. By extension, in
everyday situations, nonnatives would not be expected to experience comprehension difficulty through WM resource limitations,
because most daily situations are predictable, and topic knowledge
can no doubt be used to aid comprehension. Difficulty in comprehension would only be expected in unfamiliar situations, where
topic knowledge is unavailable. As a consequence of this, it is
predicted that experienced nonnative comprehenders have welldeveloped strategies for extracting topic information from situations to assist in comprehension.
For the native listeners, it is interesting to note that there
was no significant difference between WM index scores in the
Topic and No Topic conditions. This means that, on average, the
274
Language Learning
Vol. 51, No. 2
reduction in available WM resources between baseline and dualtask was the same for natives in the Topic and No Topic conditions.
There are several possible explanations for this finding. First, it
could reflect the sensitivity of the dual task for indexing WM
consumption. Perhaps the verification of single-digit calculations
is not sufficiently demanding to reveal a difference between the
native Topic and No Topic groups, even if one exists. Even if this
is the case, it is not problematic for the present study, because an
interaction between Language Background and Topic was predicted and obtained. It is not the differences between pairs of
groups which are important, but the overall relationship among
the four groups. Nevertheless, the incorporation of a larger
number of passages, ranging in difficulty, could perhaps be included in future studies, because the difference between native
Topic and No Topic should vary as a function of passage difficulty.
Second, it may reflect different listening strategies for the natives
and nonnatives. It is possible that the presence of the dual task
changed the way in which the participants comprehended the
passage, and this may have exerted a differential influence on
nonnatives than on natives. For example, in the No Topic condition,
natives may have been content to concentrate on the sentence
level, whereas nonnatives may have striven to find coherence in
the discourse. Even if this is the case, the results may still reflect
differences in low-level processing. Nonnatives may need to focus
on the discourse level to compensate for their less developed
low-level processes. Third, there could be an effect of anxiety.
Nonnatives in the No Topic condition may have been more susceptible to language anxiety than their native counterparts, and this
may have been exacerbated by the demands of the dual task. The
inclusion of a larger number of passages would also help to address
this possibility. If language anxiety is responsible for the effect
observed in the present study, then the effect should diminish with
later passages, as the participants become accustomed to the task.
Language anxiety might also be reduced if a method were used
that does not rely on a dual task. On-line methods that also index
resource consumption, and that could be adapted to suit the
Tyler
275
present experiment include pupillometry (Hyönä, Tommola, &
Alaja, 1995; Just & Carpenter, 1995), event-related potentials
(ERPs) (Raney, 1993), and the auditory moving window technique
(Ferreira, Henderson, Anes, Weeks, & McFarlane, 1996).
Future research could also examine low-level processing in
more detail. The results of the present study suggest that, on
average, low-level processes are less efficient in nonnative than
native listeners, but the processes that cause the greater WM
consumption are unable to be identified using the current experimental design. The results of studies showing poor perception of
English /r/ and /l/ by Japanese speakers (e.g., MacKain et al., 1981;
Sheldon & Strange, 1982) suggest the possibility that speech
perception processes may be a contributing factor. Further studies
could attempt to identify those processes that are problematic for
nonnative listeners and ascertain whether the same processes are
responsible across language groups and across individuals.
The main implications of the present study are for foreign
language teaching. On the one hand, a student’s poor underlying
language processing may be masked by the effects of topic knowledge. That is, if topic is used to aid comprehension, then it may
inhibit the development of low-level foreign language processes
from controlled to automatic processes. One solution would be to
remove the influence of topic from foreign language comprehension, but this would be difficult to achieve, if not impossible. Rather,
there are techniques that could be used to develop those low-level
processes in isolation, rather than in a discourse context. For
example, in the field of foreign language speech perception, Logan,
Lively, and Pisoni (1991; Lively, Logan, & Pisoni, 1993) developed
a training task that led to an improvement in Japanese listeners’
discrimination of English /r/ and /l/ and also retention of this
improvement. The advantage of this task is that it uses real-world
exemplars, such as the minimal pairs “grass” and “glass”, rather
than phonemes in isolation. Their technique could be modified to
accommodate any speech contrast, provided that minimal pairs of
that contrast exist in the target language. Furthermore, Hardison
(1998) demonstrated, with both Japanese and Korean learners of
276
Language Learning
Vol. 51, No. 2
English, that this type of training on words spoken in isolation
generalizes to the earlier recognition of words in connected speech.
Thus, the incorporation of such training into regular language
classes may allow phoneme perception to improve independently
of comprehension. In time, these processes may become automatic,
thus reducing the working memory requirements of speech perception processes.
On the other hand, experienced nonnative listeners seem to
comprehend effortlessly in everyday situations, so foreign language graduates may nevertheless perform well using their foreign language even without the development of low-level
processes. If a student’s goal is the rapid acquisition of a foreign
language, then perhaps it would be more beneficial to teach
strategies for the identification and extraction of topic from situations, rather than focusing on the improvement of low-level
processes. For example, students could be taught to analyze a
situation for nonlinguistic information, which may provide assistance in narrowing down the number of possible discourse topics.
This, in turn, would increase the probability that topic could be
used to aid the formation of a mental model.
There are many people living in foreign language environments who conduct their daily business using a nonnative language. The results of the present study may be extended to suggest
that in everyday situations, where topic knowledge is available,
comprehension is no more difficult for experienced nonnative listeners than for native listeners. However, if topic knowledge is unavailable, comprehension will be more difficult for the nonnative
listeners. This is interpreted to suggest that the low-level processes used in decoding the speech signal are less efficient in
experienced nonnative listeners than in native listeners. The
application of this finding to foreign language teaching depends
upon the goals of the learner. Students aiming for long-term
mastery of a language may benefit from such measures as phoneme discrimination training, whereas students who aim for rapid
acquisition, such as travellers, may receive a greater benefit from
Tyler
277
training in extracting nonlinguistic information from a situation
to aid comprehension.
Revised version accepted 13 October 2000
Note
1
A single correlation was performed on the nonnative scores, collapsed across
the Topic and No Topic conditions, because each separate correlation would
be based on only 15 observations. For reference, the individual correlations
are provided here. In the Topic condition, r = .10 and p = .72, and in the No
Topic condition, r = .21 and p = .45.
References
Anderson, R. C., & Pearson, P. D. (1984). A schema-theoretic view of basic
processes in reading comprehension. In P. D. Pearson (Ed.), The handbook
of reading research (pp. 255–292). New York: Longman.
Bartlett, F. C. (1932). Remembering: A study in experimental and social
psychology. Cambridge, UK: Cambridge University Press.
Birdsong, D. (Ed.). (1999). Second language acquisition and the Critical Period
Hypothesis. Mahwah, NJ: Lawrence Erlbaum.
Bley-Vroman, R. (1983). The comparative fallacy in interlanguage studies:
The case of systematicity. Language Learning, 33, 1–17.
Bley-Vroman, R. (1989). What is the logical problem of foreign language
learning? In S. M. Gass & J. Schachter (Eds.), Linguistic perspectives on
second language acquisition. Cambridge, UK: Cambridge University
Press.
Bley-Vroman, R. (1990). The logical problem of foreign language learning.
Linguistic Analysis, 20, 3–49.
Bransford, J. D., & Johnson, M. K. (1972). Contextual prerequisites for
understanding: Some investigations of comprehension and recall. Journal
of Verbal Learning & Verbal Behavior, 11(6), 717–726.
Carrell, P. L. (1984). Schema theory and ESL reading: Classroom implications
and applications. Modern Language Journal, 68(4), 332–343.
Carrell, P. L. (1987). Content and formal schemata in ESL reading. TESOL
Quarterly, 21(3), 461–481.
De Rammelaere, S., Stuyven, E., & Vandierendonck, A. (1999). The contribution of working memory resources in the verification of simple mental
arithmetic sums. Psychological Research, 62, 72–77.
278
Language Learning
Vol. 51, No. 2
Emmorey, K., & Corina, D. (1990). Lexical recognition in sign language:
Effects of phonetic structure and morphology. Perceptual & Motor Skills,
71(3, Pt. 2), 1227–1252.
Ferreira, F., Henderson, J. M., Anes, M. D., Weeks, P. A., & McFarlane, D. K.
(1996). Effects of lexical frequency and syntactic complexity in spokenlanguage comprehension: Evidence from the auditory moving-window
technique. Journal of Experimental Psychology: Learning, Memory, &
Cognition, 22(2), 324–335.
Flege, J. E., Yeni-Komshian, G. H., & Liu, S. (1999). Age constraints on
second-language acquisition. Journal of Memory & Language, 41(1),
78–104.
Geary, D. C., Cormier, P., Goggin, J. P., Estrada, P., & Lunn, M. C. E. (1993).
Mental arithmetic: A componential analysis of speed-of-processing across
monolingual, weak bilingual, and strong bilingual adults. International
Journal of Psychology, 28, 185–201.
Glenberg, A. M., Kruley, P., & Langston, W. E. (1994). Analogical processes in
comprehension: Simulation of a mental model. In M. A. Gernsbacher (Ed.),
Handbook of psycholinguistics (pp. 609–640). San Diego: Academic Press.
Glenberg, A. M., & Langston, W. E. (1992). Comprehension of illustrated text:
Pictures help to build mental models. Journal of Memory & Language, 31,
129–151.
Hardison, D. M. (1998). Spoken word identification by native and nonnative
speakers of English: Effects of training, modality, context and phonetic
environment. In R.H. Mannell & J. Robert-Ribes (Eds.), Proceedings of the
Fifth International Conference on Spoken Language Processing (Vol. 5,
1875–1878). Sydney, Australia: Australian Speech Science and Technology
Association, Inc. (ASSTA).
Haxby, J. V., Parasuraman, R., Lalonde, F., & Abboud, H. (1993). SuperLab:
General-purpose Macintosh software for human experimental psychology
and psychological testing. Behavior Research Methods, Instruments, &
Computers, 25(3), 400–405.
Hyönä, J., Tommola, J., & Alaja, A.-M. (1995). Pupil dilation as a measure of
processing load in simultaneous interpretation and other language tasks.
Quarterly Journal of Experimental Psychology. A. Human Experimental
Psychology, 48A(3), 598–612.
Jenkins, J. J., Strange, W., & Polka, L. (1995). Not everyone can tell a “rock”
from a “lock”: Assessing individual differences in speech perception. In
D. J. Lubinski & R. V. Dawis (Eds.), Assessing individual differences in
human behavior: New concepts, methods, and findings (pp. 297–325). Palo
Alto, CA: Davies-Black Publishing.
Tyler
279
Johnson, J. S., Shenkman, K. D., Newport, E. L., & Medin, D. L. (1996).
Indeterminacy in the grammar of adult language learners. Journal of
Memory & Language, 35(3), 335–352.
Just, M. A., & Carpenter, P. A. (1992). Paradigms and processing in reading
comprehension. Journal of Experimental Psychology, 111, 228–238.
Just, M. A., & Carpenter, P. A. (1995). The intensity dimension of thought:
Pupillometric indices of sentence processing. In J. M. Henderson, M.
Singer, & F. Ferreira (Eds.), Reading and language processing
(pp. 182–211). Mahwah, NJ: Lawrence Erlbaum.
Lemaire, P., Abdi, H., & Fayol, M. (1996). The role of working memory
resources in simple cognitive arithmetic. European Journal of Cognitive
Psychology, 8(1), 73–103.
Lively, S. E., Logan, J. S., & Pisoni, D. B. (1993). Training Japanese listeners
to identify English /r/ and /l/. II. The role of phonetic environment and
talker variability in learning new perceptual categories. Journal of the
Acoustical Society of America, 94(3, Pt. 1), 1242–1255.
Logan, J. S., Lively, S. E., & Pisoni, D. B. (1991). Training Japanese listeners
to identify English /r/ and /l/: A first report. Journal of the Acoustical
Society of America, 89, 874–886.
MacIntyre, P. D., & Gardner, R. C. (1991). Methods and results in the study
of anxiety and language learning: A review of the literature. Language
Learning, 41, 85–117.
MacIntyre, P. D., & Gardner, R. C. (1994). The subtle effects of language
anxiety on cognitive processing in the second language. Language Learning, 44, 283–305.
MacKain, K. S., Best, C. T., & Strange, W. (1981). Categorical perception of
English /r/ and /l/ by Japanese bilinguals. Applied Psycholinguistics, 2(4),
369–390.
Markham, P., & Latham, M. L. (1987). The influence of religion-specific
background knowledge on the listening comprehension of adult secondlanguage students. Language Learning, 37, 157–170.
Mayo, L. H., Florentine, M., & Buus, S. (1997). Age of second-language
acquisition and perception of speech in noise. Journal of Speech & Hearing
Research, 40, 686–693.
Navon, D., & Gopher, D. (1979). On the economy of the human-processing
system. Psychological Review, 86, 214–255.
Raney, G. E. (1993). Monitoring changes in cognitive load during reading: An
event-related brain potential and reaction time analysis. Journal of Experimental Psychology: Learning, Memory, & Cognition, 19, 51–69.
Schmidt-Rinehart, B. C. (1994). The effects of topic familiarity on second
language listening comprehension. Modern Language Journal, 78,
179–189.
280
Language Learning
Vol. 51, No. 2
Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human
information processing: I. Detection, search, and attention. Psychological
Review, 84(1), 1–66.
Schumann, J. H. (1998). The neurobiology of affect in language: A supplement
to Language Learning. Language Learning, 48(suppl.), 1–341.
Scovel, T. (1988). A time to speak: A psycholinguistic inquiry into the critical
period for human speech. New York: Newbury House.
Sheldon, A., & Strange, W. (1982). The acquisition of /r/ and /l/ by Japanese
learners of English: Evidence that speech production can precede speech
perception. Applied Psycholinguistics, 3(3), 243–261.
Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic human
information processing. II. Perceptual learning, automatic attending and
a general theory. Psychological Review, 84(2), 127–190.
Takagi, N., & Mann, V. A. (1995). The limits of extended naturalistic exposure
on the perceptual mastery of English /r/ and /l/ by adult Japanese learners
of English. Applied Psycholinguistics, 16(4), 379–405.
Takano, Y., & Noda, A. (1993). A temporary decline in thinking ability during
foreign language processing. Journal of Cross-Cultural Psychology, 24,
445–464.
Takano, Y., & Noda, A. (1995). Interlanguage dissimilarity enhances the
decline of thinking ability during foreign language processing. Language
Learning, 45, 657–681.
Thomas, M. (1994). Assessment of L2 proficiency in second language acquisition research. Language Learning, 44, 307–336.
Trabasso, T., & Suh, S. (1993). Understanding text: Achieving explanatory
coherence through on-line inferences and mental operations in working
memory. Discourse Processes, 16, 3–34.
Wolff, D. (1987). Some assumptions about second language comprehension.
Studies in Second Language Acquisition, 9, 307–326.
Zwaan, R. A., & Brown, C. M. (1996). The influence of language proficiency
and comprehension skill on situation-model construction. Discourse Processes, 21, 289–327.
Zwaan, R. A., & Radvansky, G. A. (1998). Situation models in language
comprehension and memory. Psychological Bulletin, 123(2), 162–185.