Content uploaded by Stephen Matthews
Author content
All content in this area was uploaded by Stephen Matthews on Dec 20, 2013
Content may be subject to copyright.
Aspectual Asymmetries in the Mental Representation of Events:
Significance of Lexical Aspect
Foong Ha Yap (yap_foong_ha@cuhk.edu.hk)
Department of Linguistics and Modern Languages
Chinese University of Hong Kong
Shatin, Hong Kong, China
Stella Wing Man Kwan (skwan17@gmail.com)
Department of Linguistics and Modern Languages
Chinese University of Hong Kong
Shatin, Hong Kong, China
Emily Sze Man Yiu (emilyyiu2000@yahoo.com)
Department of Linguistics and Modern Languages
Chinese University of Hong Kong
Shatin, Hong Kong, China
Patrick Chun Kau Chu (patrickhk@cuhk.edu.hk)
Department of Linguistics and Modern Languages
Chinese University of Hong Kong
Shatin, Hong Kong, China
Stella Fat Wong (w_f_stella@hotmail.com)
Department of Linguistics and Modern Languages
Chinese University of Hong Kong
Shatin, Hong Kong, China
Stephen Matthews (matthews@hkucc.hku.hk)
Department of Linguistics
University of Hong Kong
Pokfulam, Hong Kong, China
Yasuhiro Shirai (lingpitt@pitt.edu)
Department of Linguistics
University of Pittsburgh
2816 Cathedral of Learning
Pittsburgh, PA 15260, USA
Abstract
Aspect contributes important temporal information for the
construction of situation models in the human mind. Previous
studies examining the effect of grammatical aspect on
accomplishment verbs (e.g. bake a cake) show that perfective
sentences/utterances are processed faster than imperfective
ones (Madden & Zwaan, 2003; Chan et al., 2004; Yap et al.,
2004, in press). The present study, however, shows strong
interaction between lexical aspect and grammatical aspect.
More specifically, the results show that perfective facilitation
is found on accomplishment verbs, while imperfective
facilitation is found on activity verbs. We suggest that this is
because the inherent atelic nature of activity verbs matches
the unbounded features associated with imperfectives.
Keywords: aspect; temporal processing; situation models
Introduction
We construct situation models in our mind as we listen to
narratives, using various cues such as temporal and spatial
information, agent intentionality, and causality, among
many others (Zwaan & Radvansky, 1998). With respect to
temporal information, aspectual cues are known to play a
very important role in cognitive representations.
Aspect refers to different ways of viewing the temporal
properties of a situation (Comrie, 1976). There are two
major types of aspect—namely, lexical aspect and
2410
grammatical aspect. Lexical aspect refers to situation types
that are distinguished on the basis of temporal properties
such as dynamism, durativity, and telicity. Vendler (1967)
distinguishes four basic situation types, namely states (e.g.
know), activities (e.g. play), accomplishments (e.g. bake a
cake), and achievements (e.g. break). Others have often
included a fifth category, often referred to as semelfactives
(e.g. cough or skip, iteratively). Table 1 distinguishes each
verb type based on their temporal characteristics.
Table 1: Situation types denoted by verbs and their
temporal properties (Smith 1991)
Dynamic Durative Telic
State - + -
Activity + + -
Accomplishment + + +
Achievement + - +
Semelfactive + + +
Grammatical aspect, on the other hand, allows us to view
a situation as bounded or unbounded, telic or atelic, i.e.
whether or not our temporal focus includes an initial or
final endpoint (Comrie 1976; Smith 1991). We often make
a two-way distinction for grammatical aspect: perfective
and imperfective. As defined in Comrie (1976), perfective
aspect allows us to view an event as a whole (hence
‘bounded’ or ‘external view’), while imperfective aspect
constrains us to focus on the internal stages of an event
(hence ‘unbounded’ or ‘internal view’).
Different aspectual cues have been shown to have
differential effects on processing. Magliano and Schleich
(2000) showed that imperfectives yield slower decay rates
in memory recall tasks. More recently, Madden and Zwaan
(2003) have shown that there is an advantage in processing
speeds for perfective constructions in English. That is,
perfective aspect (e.g. He wrote a letter) is processed faster
than imperfective aspect (e.g. He was writing a letter).
Perfective facilitation was also found in tenseless
languages such as Cantonese and Mandarin. In Cantonese,
aspectual asymmetry in processing was manifested as
follows: sentences with perfective zo2 were processed
significantly faster than those with imperfective progressive
gan2 (Chan, Yap, Shirai & Matthews, 2004). Similar results
were also found in Mandarin, where sentences with
perfective le were processed significantly faster than those
with imperfective progressive zai (Yap, Chan, Shirai, Tan,
Matthews & Li, 2004). A similar study on Japanese likewise
showed that sentences with perfective –ta were processed
faster than those with imperfective –teiru (Yap, Inoue,
Shirai, Matthews, Wong & Chan, in press).
Note that in all these studies, verbs used to compare the
effect of grammatical aspect were accomplishments, and the
role of lexical aspect itself has yet to be examined. In the
present study, we investigate the effect of lexical aspect (as
well as grammatical aspect) in the construction of situation
models by examining whether perfective facilitation holds
across other verb types. In particular, we extend our study to
activity verbs.
Current Study
Temporal reference system of Cantonese
Cantonese is a Chinese dialect spoken in the southern part
of China. It lacks grammaticalized tense but is rich in
aspectual distinction. Temporal reference is generally
inferred from information provided through temporal
adverbs (e.g. ji5cin4 ‘before’, ji1gaa1 ‘now’, ting1jat6
‘tomorrow’), lexical aspect, and grammatical aspect
markers (Matthews & Yip, 1994).
Specific objectives of the present study
In this study, we examine whether different verb classes—
accomplishments and activities—have different effects on
cognitive processing. Accomplishment verbs are dynamic,
durative and telic (i.e. bound by an inherent endpoint),
while activity verbs are dynamic, durative but atelic (i.e.
without an inherent endpoint). Examples (in Cantonese) are
shown in (1) and (2) below:
(1) Accomplishment (ACC): [+durative] [+telic]
sik6 go3 ping4gwo2
eat CL apple
‘eat an apple’
(2) Activity (ACT): [+durative] [-telic]
jau4seoi2
swim
‘swim’
We also examine the effects of 4 Cantonese grammatical
aspect markers on activity and achievement verbs. These
comprise of two imperfective markers, gan2 and hai2dou6,
and two perfective markers, zo2 and jyun4. Gan2 and
hai2dou6 denote a progressive sense—the former is more
dynamic, being derived from a verb meaning ‘to hold
tightly’; the latter is more stative, consistent with its
derivation from a locative verb meaning ‘be here, at’. Zo2
and jyun4 induce a completive sense; the former highlights
the relevance of a completed event to a subsequent or later
situation (much like the English perfect has + V-en), while
the latter focuses more on the sense of completion itself.
Examples of an activity verb (jau4seoi2 ‘swim’) with
perfective aspect markers zo2 and jyun4 are shown in (3)
and (4), and the same activity verb with imperfective gan2
and hai2dou6 are shown in (5) and (6) respectively.
(3) Perfective zo2
Go3 laam4zai2 jau4-zo2 seoi2
CL boy swim-ASP water
‘The boy has swum.’
(4) Perfective jyun4
Go3 laam4zai2 jau4-jyun4 seoi2
CL boy swim-ASP water
‘The boy has swum.’
(5) Imperfective gan2
Go3 laam4zai2 jau4-gan2 seoi2
CL boy swim-ASP water
‘The boy is swimming.’
2411
(6) Imperfective hai2dou6
Go3 laam4zai2 hai2dou6 jau4 seoi2
CL boy ASP swim water
‘The boy is swimming.’/
‘The boy swims here.’
The aims of our study are thus twofold. First, we
examine whether perfective facilitation in processing speed
is observed across different perfective and imperfective
markers. Second, we investigate if lexical aspect also
influences processing speeds. More specifically, we
examine if accomplishment and activity verbs yield
different aspectual asymmetries, contra Madden and Zwaan
(2003).
Overall Methodology
Forced-choice utterance-picture matching tasks were used
in our experiments. Auditory stimuli (i.e. spoken utterances)
instead of visual stimuli were used. We compared the
reaction time and accuracy rate of perfectives and
imperfectives. All the sentences constructed for testing are
natural and frequent in the language. The combination of
verbs and aspect markers is also natural. The participants in
this study were native Cantonese speakers at grades 12 and
13 in local secondary schools in Hong Kong (mean age
group approximately 18).
During the experiments, for each test item, the
participants were first presented with an utterance involving
either a perfective or imperfective marker. The participants
were then immediately shown a pair of pictures with one
picture depicting an ongoing action and another picture
depicting a completed action. The participants had to match
which of the pictures best describes the utterance they had
just heard by pressing the corresponding key on the
keyboard. If the correct picture appeared on the left, the
subject had to press the A key. If the correct picture
appeared on the right, the subject had to press the 5 key on
the number pad to the right of the keyboard.
The ISI between the onset of stimulus and target is
2200ms. The pictures remained for a maximum of 3
seconds. Failure to respond within 3 seconds was regarded
as an incorrect response. No outlying items were removed
in the analyses. Reaction times were recorded using
millisecond INQUISIT software. The mean reaction time
was compared across matched perfectives and matched
imperfectives, and the significance was tested by a series of
ANOVAs. An extra 8 pairs of utterance-picture items were
used as trials for the participants to practice at the beginning
of the experiment. All the auditory and picture stimuli in the
experiments were counterbalanced. In the analysis, only
correct responses were analyzed.
Altogether, we conducted a series of 3 experiments using
this methodology, with some configurational variations
related to number of verb classes (to test for the effect of
lexical aspect), type of grammatical aspect markers (to test
for the robustness of different types of perfective and
imperfective markers), number of participants, and number
of sentences used (see Table 2 for a summary).
Figure 1: Utterance-picture matching task
A. Presentation of either a perfective or
imperfective utterance:
Go3 laam4zai2 taan4-gan2 kam4
CL boy play-ASP piano
‘The boy is playing the piano.’
B. Presentation of a pair of pictures depicting
completed versus ongoing events:
Table 2: Experimental setups
Experiment Verb
Class
Aspect
markers
No. of
participants
No. of
sentences*
1a ACC zo2, gan2 18 20
1b ACC jyun4,
hai2dou6
18 20
2a ACT zo2, gan2 18 20
2b ACT jyun4,
hai2dou6
18 20
3 ACT,
ACC
zo2, gan2 32 24
ACC: Accomplishment; ACT: Activity
* (excluding trial sentences)
Experiment 1
Experiment 1 is conducted to establish a baseline of contrast
between perfective and imperfective aspect for subsequent
experiments in this study. Previous literature has indicated
that a perfective advantage is observed with
accomplishment verbs (Madden & Zwaan, 2003; Chan et al.,
2004; Yap et al., 2004, in press). Previous work on
Cantonese examined a perfective-imperfective contrast
between zo2 and gan2 (Chan et al., 2004); the present study
adds another perfective-imperfective set, namely jyun4 and
hai2dou6 respectively). This allows us to examine whether
the perfective advantage is robust across different types of
perfective and imperfective aspect markers.
Methodology
A forced-choice utterance-picture matching task was used,
as described above, with Experiment 1a testing for the
effects of perfective zo2 and imperfective gan2 on
accomplishment verbs, and Experiment 1b replicating the
preceding study using perfective jyun4 and imperfective
hai2dou6.
2412
Results
For experiment 1a, a one-way repeated measure ANOVA
was conducted to compare the reaction time of perfective
marker zo2 and imperfective marker gan2 in the context of
accomplishment verbs. The dependent variable was
subjects’ reaction time in the picture-matching task. The
within-subjects factor was grammatical aspect with 2 levels
(perfective marker zo2 and imperfective marker gan2). The
effect of grammatical aspect was significant, F(1,141) =
10.82, p = .001. The result shows that the mean reaction
time of the constructions with perfective zo2 is significantly
faster than those with imperfective gan2. The mean reaction
time of the constructions with perfective zo2 is 941ms
(SD=242), while that of the constructions with imperfective
gan2 is 1032ms (SD=289). See Table 3.
Table 3: Results of experiments 1a and 1b
Experiment 1a Experiment 1b
Verb Class Accomplishment Accomplishment
Aspect
markers
zo2 vs. gan2 jyun4 vs. hai2dou6
Mean length
of utterance
1.94 secs
SD= .12
1.82 secs
SD= .19
Perfective Mean=941ms
SD=242
Mean=964ms
SD=295
Imperfective Mean=1032ms
SD=289
Mean=1113ms
SD=396
Significance p = .001 p < .001
Perfective
advantage
Perfective
advantage
For experiment 1b, a similar one-way repeated measure
ANOVA was conducted to compare the reaction time of
perfective marker jyun4 and imperfective marker hai2dou6
in the context of accomplishment verbs. The dependent
variable again was participants’ reaction time in the picture-
matching task. The within-subjects factors were
grammatical aspect with 2 levels (perfective jyun4 and
imperfective hai2dou6). The effect of grammatical aspect
was significant, F(1,164) = 15.54, p < .001. Results show
that there is a perfective advantage. The mean reaction time
of the constructions with perfective jyun4 is significantly
faster than those with imperfective hai2dou6. The mean
reaction time of perfective utterances with jyun4 is 964ms
(SD=295), while that of utterances with imperfective
hai2dou6 is 1113ms (SD=396).
The results from both experiments 1a and 1b show that
there is a perfective advantage with accomplishment verbs.
This advantage is robust across two different types of
perfective aspect markers in Cantonese, zo2 and jyun4.
Experiment 2
We further examine whether the perfective advantage found
with accomplishment verbs is persistent across other verbs,
in this particular case, with activity verbs. Positive results
would suggest that aspectual asymmetry in processing
speed is strongly biased in favor of perfective aspect
markers. However, a negative result (i.e. one that shows
imperfective facilitation), would indicate that our mental
representations are influenced by subtle and complex
interactions between both lexical and grammatical aspect.
Methodology
The same forced-choice utterance-picture matching task
used in Experiment 1 above was adopted, with one
important change to the configuration: activity verbs were
used instead of accomplishment verbs. Again two pair-wise
experiments were used to test for the robustness of
aspectual asymmetry (if any) across different types of
grammatical markers. Experiment 2a tested for the effects
of perfective zo2 and imperfective gan2 on activity verbs,
while experiment 2b tested for the effects of perfective
jyun4 and imperfective hai2dou6 on the same set of verbs.
Results
Perfective advantage was not found for activity verbs in
experiments 2a and 2b. In experiment 2a, the effect of
grammatical aspect was significant, F(1,165) = 5.09, p
= .025. For activity verbs, imperfective utterances were
processed significantly faster than perfective utterances.
The mean reaction time of imperfective gan2 utterances
(mean=1125ms, SD=367) is significantly faster than the
mean reaction time of perfective zo2 utterances
(mean=1211ms, SD=379). In experiment 2b, the effect of
grammatical aspect was also significant, F (1,176) = 5.91, p
= .016. The mean reaction time of imperfective
constructions with hai2dou6 is 1064ms (SD=425), which is
faster than the mean reaction time of perfective
constructions with jyun4 (mean=1171ms, SD=431). See
Table 4.
Table 4: Results of experiments 2a and 2b
Experiment 2a Experiment 2b
Verb Class Activity Activity
Aspect
markers
zo2 vs. gan2 jyun4 vs. hai2dou6
Mean length
of utterance
1.7 secs
SD= .15
1.7 secs
SD = .15
Perfective Mean=1211ms
SD=379
Mean=1171ms
SD=431
Imperfective Mean=1125ms
SD=367
Mean=1064ms
SD=425
Significance p = .025 p = .016
Imperfective
facilitation
Imperfective
facilitation
Discussion
Results from experiments 2a and 2b demonstrate that there
is no perfective advantage for activity verbs. The findings
indicate that inherent lexical aspect (accomplishment or
activity) plays an important role in the formation of mental
representation of events. Accomplishment verbs impose an
inherent telic endpoint to the event, while activity verbs
draw reference to an event without focusing on the
endpoint(s). The intrinsic atelic nature of activity verbs
2413
matches the features associated with imperfective aspect
markers gan2 and hai2dou6. Imperfective markers impose
reference to the internal stages of the event, which matches
the atelic nature of activity verbs. Processing time could be
delayed when perfective markers impose an ‘induced’
endpoint on activity verbs.
On the other hand, perfective facilitation is observed for
accomplishment verbs, as seen in experiments 1a and 1b. A
perfective processing advantage is seen in accomplishment
verbs but not in activity verbs since the telic nature of
accomplishment verbs matches the characteristics of
perfective aspect markers zo2 and jyun4. Perfective aspect
markers are compatible with (and apparently reinforce) a
‘boundedness’ representation. We thus see that inherent
lexical aspect plays a significant role in the construction of
mental representations of events.
Experiment 3
This follow-up experiment examines whether aspectual
asymmetry is found in environments involving two verb
classes: activity and accomplishment. The asymmetrical
relationship between perfective and imperfective
constructions can be more clearly seen in pair-wise
situations in experiments 1 and 2. As mentioned previously,
perfective facilitation was found in experiments 1a and 1b
with accomplishment verbs, while imperfective facilitation
was found in experiments 2a and 2b with activity verbs.
Experiment 3 is needed to determine if the above aspectual
asymmetries persist in more complex environments where
more than one verb type is found. More specifically, the
research question is whether the same perfective advantage
(found in experiments 1a and 1b) and imperfective
facilitation (found in experiments 2a and 2b) can be found
in a setting where both accomplishment and activity verbs
are included in the stimuli set.
Methodology
The forced-choice utterance-picture matching task used
above is expanded to include two (instead of one) verb types,
namely activity and accomplishment verbs. In terms of
grammatical aspect markers, perfective zo2 and
imperfective gan2 were used.
Results
A two-way repeated measures ANOVA was conducted to
compare the effect of grammatical aspect and lexical aspect
on processing time. The dependent variable was
participants’ reaction time in picture-matching task. The
within-subjects factors were lexical aspect with 2 levels
(accomplishments and activities) and grammatical aspect
with 2 levels (aspect markers gan2 and zo2). The main
effect of grammatical aspect was significant, F(1,167) =
6.63, p = .011. The interaction of lexical x grammatical
aspect effect was also significant, F(1,167) = 5.79, p = .017.
The main effect of lexical aspect was not significant,
F(1,167) = .73, p = .394. See Table 5.
Table 5: Results of experiment 3
Grammatical aspect Lexical aspect
Perfective zo2 Imperfective gan2
Accomplishment
(N=168)
Mean=1199 ms
SD=376
Mean=1193 ms
SD=430
Activity
(N=168)
Mean=1239 ms
SD=445
Mean=1096 ms
SD=326
Mean length of utterance = 1.8 secs; SD= .19
Main effect of grammatical aspect effect: p = .011
Main effect of lexical aspect effect: p = .394
Lexical x grammatical interaction effect: p = .017
A paired-samples t-test was conducted to investigate the
significant interaction. The reaction time of imperfective
marker gan2 (1091ms) (SD=321) is significantly faster than
perfective marker zo2 (1241ms) (SD=435) within activity
verbs, t(178) = -4.063, p < .001. The difference of reaction
times of gan2 and zo2 within accomplishment verbs is not
significant, t(167) = - .146, p = .884. See Table 6.
Imperfective facilitation, but not perfective facilitation, was
found in the presence of two verb classes. This shows that
imperfective facilitation with activity verbs is more robust
when compared to perfective facilitation with
accomplishment verbs.
Table 6: Follow-up t-test on interaction effect of
experiment 3
Perfective zo2 Imperfective
gan2
Signifi-
cance
ACC
(N=168)
Mean=1199ms
SD=376
Mean=1193ms
SD=430
p = .884
ACT
(N=179)
Mean=1241ms
SD=435
Mean=1091ms
SD=321
p < .001
Note: The mean and SD of perfective zo2 and imperfective gan2 with
activity verbs are different in Tables 5 and 6 due to differences in N. In
Table 5, N =168; in Table 6, N=179.
Discussion
Neighborhood density appears to be a contributing factor,
yielding a ‘squishing effect’. Since accomplishment verbs
comprise of an ongoing phase and a telic (endpoint) phase,
perfective facilitation in accomplishment verbs is more
fragile. In this experiment, in which subjects are exposed to
both activity and accomplishment verbs, the presence of
activity verbs could be inducing subjects to become more
responsive to the ongoing phase of accomplishment verbs.
This could then sometimes ‘wipe out’ the perfective
advantage often associated with accomplishment verbs.
An experiment of pair-wise design (like the ones in
experiments 1 and 2) and a complex experiment with 2x2
design (like experiment 3) are conducted for different
purposes. A pair-wise experiment can show the simple main
effect of grammatical aspect in the context of one verb class.
A complex experiment can show the behavior of
grammatical aspect in the context of two (or more) verb
classes. The results of our experiments show that perfective
facilitation with accomplishment verbs can be teased out in
the clinical environment of a pair-wise design. However,
2414
experiment 3 of a complex design shows that perfective
facilitation is more fragile in the context of two verb classes,
while imperfective facilitation with activity verbs remains
robust.
We also conducted an alternative way of analyzing the
results of experiments 1 and 2, i.e. we combined
experiments 1a and 2a for analysis. This is a split plot
design in which the lexical aspect (with 2 levels:
accomplishment and activity) is a between-subjects factor
while grammatical aspect (with 2 levels: gan2 and zo2) is a
within-subjects factor. This way of analysis of combining
experiments 1a and 2a is similar to the complex design of
experiment 3, except that lexical aspect is a between-
subjects factor. The results of this analysis (repeated-
measures ANOVA) reveal the following: the main effect of
grammatical aspect is significant (p < .001), the main effect
of lexical aspect is not significant, and the interaction effect
is not significant (p=1.000). (Crucially different, the
interaction effect of experiment 3 is significant, p = .017).
The significant grammatical aspect effect and non-
significant lexical aspect effect in this analysis is the same
as those for complex experiment 3. What is different is the
interaction effect between this analysis (which combined
experiments 1 and 2) and the complex design of experiment
3. This could be explained by the fact that subjects in
experiments 1 and 2 were exposed to one verb class only.
The interaction effect found in experiment 3 could not be
revealed in this combined analysis. A complex design which
includes 2 (or more) verb classes mimics real-time
processing in language comprehension. Thus, conducting a
pair-wise experiment and a complex experiment provides
different insights.
General Discussion
The accuracy rates of matched perfective and matched
imperfective in the present study is higher than that in the
English study by Madden and Zwaan (2003). In their study,
participants were sensitive only to perfective cues. The
accuracy rate of perfective constructions was 76% and that
of imperfectives was 56%. However, different results were
found for Chinese. Participants were sensitive to both
perfective and imperfective cues in Cantonese and
Mandarin (Chan et al., 2004; Yap et al., 2004; this study).
In the Cantonese study (Chan et al., 2004), the accuracy rate
for perfectives is 95% and that of imperfectives is 91%. In
the Mandarin study (Yap et al., 2004), the accuracy rate of
perfectives is 87% and that of imperfectives is 82%. In the
present study on Cantonese, the average accuracy rate for
perfectives and imperfectives is over 85% for each
experiment. The results, as summarized in Table 7, show
that there is high sensitivity to both perfective and
imperfective cues in Mandarin and Cantonese, both of
which are tenseless but rich in aspectual markings.
The Cantonese data also reveal that both perfective and
imperfective facilitations were found in utterance-picture
matching tasks. Imperfective facilitation was robust with
activity verbs. Perfective facilitation was found with
accomplishment verbs; however this perfective advantage
was fragile in complex environments involving more than
one verb type. There is therefore an interaction between
lexical aspect and grammatical aspect on cognitive
processing and the construction of mental representation of
events. Future work could further investigate the effect of
other verb classes (e.g. achievements, states and
semelfactives) in the processing of aspectual information.
Table 7: Accuracy rates in the present study
Experiment 1a 1b 2a 2b
Verb Class ACC ACC ACT ACT
Perfective 90% 95% 92% 93%
Imperfective 89% 92% 97% 96%
Overall
Accuracy
89.5% 93.5% 94.5% 94.5%
Acknowledgements
We gratefully acknowledge support from Direct Grant 2004-06
(#2010255) from the Chinese University of Hong Kong and
Competitive Earmarked Research Grant 2005-07 (#2110122) from
the Research Grants Council of Hong Kong. We also thank Lai
Chim Chow, Irene Lam, and participating schools for technical
assistance; Calvin Chan and Kimmee Lo for picture drawing; and
Edson Miyamoto and Him Cheung for valuable comments.
References
Chan, Y.H., Yap, F.H., Shirai Y., & Matthews, S. (2004). A
perfective-imperfective asymmetry in language processing:
Evidence from Cantonese. Proceedings of the 9th International
Symposium on Chinese Languages and Linguistics, (pp. 383-
391). Academia Sinica and the Graduate Institute of Linguistics,
National Taiwan University, Taipei.
Comrie, B. (1976). Aspect. Cambridge, UK: Cambridge University
Press.
Madden, C.J. & Zwaan, R.A. (2003). How does verb aspect
constrain event representation? Memory & Cognition, 31, 663-
672.
Magliano, J.P. & Schleich, M.C. (2000). Verb aspect and situation
models. Discourse processes, 29, 83-112.
Matthews, S. & Yip, V. (1994). Cantonese: A comprehensive
grammar. London: Routledge.
Smith, C. (1991). The parameter of aspect. Dordrecht: Kluwer
Academic Press.
Vendler, Z. (1967). Linguistics in Philosophy. Ithaca, NY: Cornell
University Press.
Yap, F. H., Chan, Y.H., Shirai, Y., Tan, L.H., Matthews, S. & Li, P.
(2004). A perfective-imperfective asymmetry in the human
mind: Evidence from Cantonese and Mandarin. Paper presented
at the 28th International Congress of Psychology, Beijing,
August 8-13.
Yap, F. H., Inoue, Y., Shirai Y., Matthews, S., Wong, Y.W., &
Chan, Y.H. (in press). Aspectual asymmetries in Japanese:
Evidence from a reaction time study. Japanese/Korean
Linguistics, vol. 14. Stanford, CSLI.
Zwaan, R.A. & Radvansky, G.A. (1998). Situation models in
language comprehension and memory. Psychological Bulletin,
123, 162-185.
2415
