interpreting elliptical verb phrases at different

LANGUAGE AND SPEECH, Vol. 30, Part 2,1987
145
INTERPRETING ELLIPTICAL VERB PHRASES AT DIFFERENT
TIKES OF DAY: EFFECTS OF PLAUSIBILITY AND
ANTECEDENT DISTANCE*
JANE OAKHILL
and
ALAN GARNIIAhl
Utiiversiry of Srtssex
This paper reports an experiment that used the self-paced reading technique to explore
subjects’ on-line processing of short texts at two times of day (early morning and late
afternoon). A previous experiment (Oakhill, 1986b) has shown that, when reading simple
texts, subjects adopt a “less effortful” strategy in the morning; they delay processing until
it is needed to answer a question. In the present experiment, we used a more difficult
anaphor resolution task that resulted in more errors. Under these circumstances, the morning
subjects increased their on-line processing effort. They needed to spend longer reading the
more difficult texts in order to achieve the same level of accuracy as subjects tested in the
afternoon.
Keywords: time of day, text comprehension, amphora
INTRODUCTION
A number of studies have shown that memory for text varies with time of day
(Laird, 1925; Folkard, hlonk, Bradbury and Rosenthall, 1977; Folkard and hlonk,
1980; Oakhill, 1986a). Immediate memory for text tends t o decline over the day,
while delayed memory shows the opposite trend; better recall is associated with original
presentation later in the day. However, the reasons for these changes have remained
obscure.
One finding that may be relevant is that of Folkard (1979), who showed that, when
learning lists, subjects tended to engage in maintenance processing in the morning and in
elaborative processing based on the items’ meanings in the afternoon. If subjects show
the same changes in bias when processing text, they will be more likely to integrate
and elaborate on the ideas in it, producing a better integrated and more coherent
memory trace in the afternoon. Thus the explanation of time-of-day effects on memory
for text may lie in differences in on-line text processing.
A previous on-line experiment (Oakhill, 1986b) showed that subjects tested in the
morning used qualitatively different text comprehension strategies from those tested
in the afternoon. The strategy employed by the morning subjects can be characterized
as less “effortful.” These subjects tended to delay the integration of information in
*
We would like to thank Peter Colquhoun f o r comments on an earlier draft of this
paper.
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
146
Ellip tical VPs
short texts until it was needed t o answer a question. Subjects tested in the afternoon,
by contrast, dealt with referential difficulties in the text immediately. They took
longer to read the texts, and less time to answer subsequent questions.
However, the texts used by Oakhill (1986b) were short and easy to understand.
The integration task required subjects to resolve pronouns in order to answer questions.
Even though the morning subjects did not seem to resolve the pronouns as they read
them, their ability t o answer the questions was slightly better than that of the afternoon
group (the groups made 13.5%and 17.7% errors respectively).
The present experiment explores the effects of time of day on text processing when
the integration of information from different parts of the text is more difficult. The
passages used in this experiment contained a different type of anaphor: verb-phrase
ellipsis (WE). Verb-phrase ellipsis occurs when a sentence contains a “dangling”
auxiliary verb or infinitive marker with a verb phrase “missing” after it. Sentences such
as I wanted to and Peter ivas contain examples of WE. Elliptical verb phrases are
common and not usually difficult to understand. However, they can be made much
more difficult than pronouns, because, unlike pronouns, these verb phrases necessarily
take their meaning from the exact wording of the preceding text. Verbatim memory,
unlike memory for content, lasts for only a comparatively short period (Sachs, 1967;
Johnson-Laird and Stevenson, 1970).
The way in which the interpretation of elliptical verb phrases depends on surface
form is illustrated in the following examples. The second sentences in the two passages
differ in meaning, and their meanings can only be derived if the precise wording (not
simply the gist meaning) of the first sentence is accessible.
The Gauls had been captured by the Romans.
The Britons had too.
The Romans had captured the Gauls.
The Britons had too.
In the earlier experiment (Oakhill, 1986b), morning subjects delayed their interpretation of pronouns with no effect on their ability to answer simple questions. The first
question that the present .experiment was designed to answer was: What d o morning
subjects do when the integration task is harder? To delay the interpretation of an
elliptical verb phrase may cause comprehension problems, because the exact wording on
which its interpretation depends may have been forgotten. There are two possible
Strategies that subjects might adopt. The first is to continue t o delay and become more.
error prone. The second strategy, which subjects may adopt because they realize that the
task is difficult, is to adjust their reading times to take account of the difficulty.
A further question investigated in this experiment was: Do morning and afternoon
subjects use different strategies t o process WE? This question arises because there are
different ways in which elliptical verb phrases can be interpreted (see hlurphy, 1985).
One way is effortful but accurate. The other is easier, but prone to error. hloming
subjects may use this latter strategy.
The first strategy “copies” a previous verb phrase into the ellipsis and, hence, requires
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
J. Oakhill arid A . Gariihani
147
precise memory for surface form. The second interprets the ellipsis with reference to a
“more abstract representation.” Murphy does not make any more specific suggestion
about what the more abstract representation might be. He suggests that copying occurs
when the antecedent of an elliptical verb phrase is in the preceding sentence. Otherwise
the alternative strategy is used. Murphy’s reason for this suggestion is that copying is
difficult when the antecedent is more distant because of loss of memory for surface form.
However, copying always produces the right interpretation, provided that the correct
verb phrase is copied. Any other strategy may lead t o errors. Murphy’s ideas, together
with the finding of Oakhill (1968b), lead to the following prediction. With distant antecedents, morning subjects should be less likely than afternoon subjects to use the difficult
but accurate copying strategy. They are more likely t o fall back on a less effortful, but
riskier, strategy.
One such strategy, one that does indeed rely on a more abstract representation, is
to compute a probable reading, based on general knowledge about the situation being
described. For example, in the context:
The elderly patient had been examined by the doctor, during the ward round.
the linguistically correct interpretation of the elliptical verb phrase the child had, too is
highly plausible. Computing a plausible reading produces the correct interpretation.
However, in the same context, the iairse had, too will be incorrectly interpreted by
this strategy. It means that the nurse had been examined by the doctor during the ward
round. This event is unlikely - much less likely than the one described by the
“suggested” interpretation: the nurse had examined the elderly patient.
- There are two ways in which the plausible interpretation strategy can take over from
the copying strategy. The more extreme hypothesis is that, at some level of difficulty,
copying is simply abandoned and the plausible interpretation is the only one available.
Readers will then assign contextually plausible, rather than linguistically correct, interpretations to elliptical clauses. The alternative is that, as copying becomes more difficult,
and its results less reliable - because what is to be copied cannot be remembered properly
- more weight is given to the plausible interpretation. This hypothesis assumes that
people always try to use memory for surface form to interpret elliptical verb phrases.
But as the distance between the anaphor and antecedent increases, they have difficulty
in recovering the surface form of the antecedent. Given their uncertainty about surface
form, and hence about the meaning of the ellipsis, subjects look for other clues t o its
interpretation.
Both these hypotheses predict that, when the correct interpretation and the plausible
interpretation coincide, any effect of distance will be reduced or eliminated, since
memory for content is more readily accessible than memory for surface form. However,
when there is a conflict between the two interpretations, the hypotheses make different
predictions. The first hypothesis predicts that, under such circumstances, elliptical VPs
that result in implausible readings will be systematically misinterpreted. Furthermore,
it predicts that, since surface form is ignored, then the incorrect interpretations should
be produced relatively quickly. The second hypothesis also predicts that people will
make errors that favour plausible interpretations of elliptical clauses and that these
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
148
Elliptical VPs
errors will increase with distance. However, contrary to the previous hypothesis, it
predicts that such errors will have long latencies. This prediction derives from the
assumption that people always attempt to recover surface form to understand elliptical
verb phrases. When there is some doubt about the interpretation suggested by the copying
strategy, and the hint of a conflict between that interpretation and the plausible one,
there will be interference that produces a delay in assigning an interpretation. The weaker
the memory for surface form, the more slowly the correct interpretation will be assigned,
and the more prone it will be to interference.
These predictions can be applied straightforwardly to time of day. hlorning subjects
prefer less effortful processing, so they should tend t o rely on niemory for content rather
than memory for surface form, especially when their task is made more difficult by
separating the anaphor and its antecedent.
The present experiment compared the interpretation of elliptical verb phrases at two
times of day, under two conditions. In the first, the linguistically correct interpretation
and one suggested by context were both plausible. In the second, the correct reading was
implausible, but the suggested reading was plausible. The distance between the ellipsis and
its’ antecedent was varied by including or excluding an intervening phrase or clause. It
was predicted that plausibility should have a greater effect with greater distance, since
surface form becomes less accessible, and thus that the use of distinct strategies would
become more marked in the distant antecedent condition.
Distance was varied behveetz subjects. The reason was that we expected the passages
in which the elliptical clause directly followed the antecedent t o be interpreted correctly
most of the time. I f distance had varied within subjects, it might have become obvious
that the passages often described highly implausible situations. The subjects might, therefore, have been more circumspect in the “difficult” condition, and not relied on
plausibility as much as in normal reading.
hlETHOD
Subjects
The subjects were 48 volunteers from the staff and student population of Sussex
University. They were paid for participating.
Materials
Twenty-four sets of materials were written for the experiment. Four related passages
were constructed from each set. Each passage had three sentences. The first provided
some setting information. The last contained an elliptical VP, which took its meaning
from the second sentence. There were two versions of the third sentence, so that in half
of the passages it had a plausible reading, while in the other half it had an implausible
reading. Each passage was followed by a question. There were two questions for each
version (plausible/implausible), one of which had a “yes” answer, and the other “no.’’
The correct answer to the question always depended on the elliptical clause having been
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
J. Oakliill arid A . Ganikani
149
interpreted correctly. There was both a long and a short form of the second sentence
in each set of materials, t o vary the distance between the ellipsis and its antecedent.
In the long form an extra phrase or clause was appended after the VP that served as the
antecedent to the elliptical VP in the third sentence. This additional material appeared
as a separate frame on the visual display; it was thematically related to the previous
passage, but did not contain a VP that could act as the antecedent of the following
anaphor. An example passage in its eight different versions is shown in Table 1.
Apparatus
The experiment was controlled on-line by an SWTPC 6809 micro-computer system.
The scntences and questions were presented one at a time in the centre of a TeleVideo
TVI-912 VDU. Between the subject and the screen were two response button boxes one labelled “yes” and one “no.”
Desigii
The subjects were randomly assigned t o two equal-sized groups - one was tested
at 9 A.M. and the other at 5 P.hl. Within each time-of-day group, distance was varied
between subjects, for the reasons given in the Introduction, although this factor was
within materials. Plausibility varied within subjects and within materials. For the
questions, there was an additional factor of whether the correct answer was “yes” or
“no.” This factor also varied within subjects and within materials.
Only one of the eight different versions of each material was to be presented to each
subject. Eight counterbalanced lists of materials were, therefore, constructed and assigned
randomly to subjects. The order of presentation of passages within each list was
randomized independently. The materials from the present experiment were randomly
intermixed with materials from another experiment on anaphora, which acted as fillers
to distract subjects from the structure of the materials.
Procedure
The subjects were instructed t o press the ‘yes’ button t o display the successive
sentences of the passage. This button was always located on the side of their dominant
hand. Each frame of the passage and the question remained on the screen until the
subject pressed a button. The instructions emphasised that they should read the passages
at normal reading speed, but that they should answer the questions as quickly as possible,
by pressing the appropriate response button. Before each passage, subjects saw the
display: ‘$$ NEXT TRIAL $$.’ When they were ready to start reading the passage,
they pressed the ’yes’ button to display the first frame of the first passage. Further
button presses displayed the remaining parts of the passage and the question. Between
the passages, after the question had been answered, there was a one-second interval,
during which the screen remained blank. The instructions emphasized that subjects could
pause as long as they wished when they saw the ‘next trial’ prompt. There were 12
practice trials, four of which were related to the present experiment.
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
150
Elliptical VPs
TABLE1
The eight passages derived from one of the sets of materials in the experiment. The
intervening clauses, included only in the “distant” condition are shown in parentheses.
(Note: Each line corresponds to one display in the experiment.)
PLAUSIBLE “YES”
It had been a busy morning in the hospital.
The elderly patient had been examined by the doctor.
(during the ward round.)
The child had too.
Did the doctor examine the child?
PLAUSIBLE “NO”
It had been a busy morning in the hospital.
The elderly patient had been examined by the doctor.
(during the ward round.)
The child had too.
Did the child examine the elderly patient?
IhlPLAUSIBLE “YES”
It had been a busy morning at the hospital.
The elderly patient had been examined by the doctor.
(during the ward round.)
The nurse had too.
Did the doctor examine the nurse?
IhlPLAUSIBLE “NO”
It had been a busy morning in the hospital.
The elderly patient had been examined by the doctor.
(during the ward round.)
The nurse had too.
Did the nurse examine the elderly patient?
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
151
J. Oakliill and A , Gartihain
TABLE2
Mean percentage errors in question answering, as a function of time of day,
plausibility and antecedent distance
Plausible
Implausible
13.3
44.3
No intervening clause
9.O
26.3
5 P.hl.
Intervening clause
8.3
32.7
No intervening clause
7.O
24.3
9 A.M.
Intervening clause
RESULTS
Question-annveriig errors
The percentage of errors in the main conditions of the experiment are shown in Table
2. There was n o main effect of time of testing on question-answering accuracy, and no
interactions with this factor. There was a main effect of plausibility, F1(1,32) = 66.54,
p < 0.0001, n ( l , 2 3 ) = 81.05, p < 0.001. Subjects made only 9% errors when the
correct interpretation of the ellipsis was plausible, but 32% when it was implausible.
There was also a main effect of whether the answer to the question was “yes” or “no”;
more errors (27%vs. 14%)were made on questions to which the correct answer was
“yes”: F1(1,32) = 14.96, p < 0.001, F2(1,23) = 24.77, p < 0.001. This result was
not predicted, but it might be explained by the fact that the subject of “no” questions
had the same (surface) role in the elliptical sentence. There was also a main effect of
distance: There were 25%errors with an intervening clause, but only 17%without one.
This effect was marginal by subjects: F1(1,32) = 3.71, p = 0.06, but significant by
materials: F2(1,23) = 11.28, p < 0.003. This factor also produced a marginally
significant interaction with plausibility, F1(1,32) = 3.54, p = 0.07, F2(1,23) = 3.74,
p = 0.07. As Table 2 shows, the interaction arises because the effect of plausibility on
errors tended t o be larger with an intervening clause than without. The fact that the
interaction failed to reach significance may be partly explained by a “floor effect” for
the plausible passages. Although these passages produced 9%errors, a large proportion of
these can be explained by incorrect response execution. Typical error rates in self-paced
reading experiments are 5-1076.
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
Elliptical VPs
152
TABLE 3
hiean clause-reading times as a function of time of day,
plausibility and distance to antecedent (times in msec)
Plausible
Implausible
9 A.M.
Intervening clause
2048
3638
No intervening clause
1936
2527
Intervening clause
1934
2960
No intervening clause
1663
2340
5 P.hl.
Clause-reading times
The mean reading times are shown in Table 3. Reading times greater than 20 seconds
or less than 500 nisec (only 1 data point in 1152) were excluded from the analysis. The
missing data point was replaced by the subject-by-condition mean in the by-subjects
analysis and the item-by-condition mean in the by-materials analysis. Analyses of variance
showed a main effect of plausibility, F1 (1,32) = 6 8 . 4 8 , ~
< 0.0001, n ( l , 2 3 ) = 69.14,
p < 0.0001, and a main effect of distance F1(1,32) = 6 . 7 7 , <
~ 0.02, F2(1,23) = 24.35,
p < 0.0001. The elliptical clauses took longer to read when an implausible correct interpretation clashed with a plausible one suggested by context (2866 vs. 1895 msec), and
longer to read when there was an intervening clause (2645 vs. 21 17 msec). There was
also an interaction between these two factors: F1(1,32) = 8.28, p <0.008,F2(1,23) =
5.36, p < 0.03. Table 3 shows that the interaction arose because the effect of plausibility
was greater when there was an intervening clause (1308 vs. 633 msec). There was a main
effect of time of day on clause reading tinics that was highly significant by materials:
F(1,23) = 23.14, p < 0.001 but not by subjects: F(1,32) = 2.37, p = 0.14. Subjects in
the morning took longer (2537 msec) than subjects tested in the afternoon (2224 msec)
to read the clauses.
The differences between the various conditions were tested using the Newman-Keuls
procedure. There were no significant differences between the morning and afternoon
subjects for the plausible passages, but the difference between the morning subjects’
reading time for implausible passages with an intervening clause (678 msec) exceeded.
the critical difference (646 nisec) at the 1% level of significance.
Questioii-answering times
The mean question-answering times are shown in Table 4. As in the analysis of the
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
153
J. Oakliill arid A . Ganihani
TABLE4
hlean question-answering times, as a function of time of day,
plausibility and distance t o antecedent (times in nisec)
Plausible
Implausible
9 A.M.
Intervening clause
2690
3963
No intervening clause
2405
309 1
5 P.M.
Intervening clause
2532
3154
No intervening clause
21 15
2496
elliptical-clause reading times, data points greater than 20 seconds or less than 500 msec
were replaced (3 data points fell outside this range). In the question-answering times,
there was a main effect of time of day that was significant by subjects, F(1,32) = 4.92,
p < 0.04, and by materials, F(1,23) = 19.38, p < 0.001. Subjects in the morning took
longer to answer the questions (3037 msec) than subjects tested in the afternoon (2574
nisec). There was also an interaction between time of day and plausibility: F1(1,32) =
8.73, p < 0.006, F2(1,23) = 7.66, p < 0.02. This interaction arises because the subjects
tested in the morning were particularly slow to answer the questions about the
implausible passages. Otherwise, the pattern of results was very similar t o that for reading
times: There was a main effcct of plausibility F1(1,32) = 82.42, p < 0.0001, F2(1,23)
= 36.72, p < 0.0001; a niain effect of whether or not there was an intervening clause
between the antecedent and target sentence, F1(1,32) = 7.09, p < 0.02, F2(1,23) =
22.05, p < 0.0001; and an interaction between these factors, F1(1,32) = 6.44, p <
0.02, n(l,23) = 5 . 4 3 , ~< 0.03.
As for the clause reading times, Newman-Keuls tests revealed n o significant differences
in question-answering for the plausible passages, but the morning subjects were
significantly slower t o answer questions about the implausible passages both with and
without an intervening clause. Both differences exceeded the critical difference (450
msec) a t the 1% level of significance.
DISCUSSION
Previous research (Oakhill, 1986b) has shown that, when reading simple texts, subjects
adopt a “less effortful” strategy in the morning; they delay processing until it is needed
t o answer a question about the text. The first question we addressed in the present
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
154
Elliptical VPs
study was: How would subjects cope with a more difficult integration task at different
times of day? The task we chose, the interpretation of elliptical verb phrases with
iniplausible meanings, was one in which a delay in processing was likely t o result in an
increase in question-answering errors. We were, therefore, particularly interested in
whether morning subjects would read the texts more slowly than afternoon subjects
or produce more errors.
The error rates confirmed that the task used in the present experiment was more
difficult than the pronoun-resolution task used previously. The results showed that the
morning subjects achieved the same level of accuracy as the afternoon subjects, even in
the more difficult, distant antecedent, condition. However, in order to achieve this level
of performance they had to increase their processing effort; the harder task forced them
t o engage in some integrative processing as they were reading the text, rather than postponing the processing until the question was presented. They took longer to read the
anaphor-containing clauses than did the afternoon subjecjs, especially in the more
difficult (implausible) condition.
The sinular error patterns of the morning and afternoon subjects suggest that both
groups were aware that some of the passages were implausible, and were attempting to
interpret them correctly. However, the morning subjects found the task difficult when
content and surface form conflicted. Both their reading times and their questionanswering times were longer for the implausible passages, particularly when a clause
intervened between the anaphor and its antecedent. In fact, it was only in this latter
condition that the reading times of the morning subjects were significantly longer than
those of the afternoon group.
The finding that morning subjects took longer to answer the questions as. well. as
longer to read the text could be explained in one of two ways. It may be that, even in
this more difficult task, they still did not complete processing of the anaphor until they
were confronted with the questions. Alternatively, they may have found it more difficult
to retrieve or t o check the relevant information necessary t o answer the question. Our
data d o not enable us to distinguish between these possibilities, but the retrieval
explanation is certainly a possible one. Other studies have shown that morning subjects
are slower on tasks that require meanings to be retrieved from memory (hlillar, Styles
and Wastell, 1980; Tilley and Warren, 1983).
A further question that the experiment addressed was whether the morning and
afternoon subjects used different strategies for processing verb-phrase ellipsis. We
predicted that, particularly in the distant anaphor condition, in which surface form is less
accessible, subjects would tend to use a plausible interpretation strategy to interpret
elliptical VPs. This alternative to the syntactically correct “copying” procedure leads to
errors in passages that describe unlikely situations. Furthermore, we predicted that
subjects tested in the morning might be especially likely to adopt this less-effortful
plausible interpretation strategy in the distant anaphor condition.
In fact, our results are incompatible with the hypothesis that either morning or
afternoon subjects tried to interpret the elliptical VPs solely by reference to memory
for content. Where the plausible and correct interpretations were in conflict, subjects
did not simply choose the plausible interpretation quickly and make a high proportion
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
J. Oakliill arid A . Gartihatti
155
of errors. Although a substantial number of the questions were answered wrongly, the
most frequent response was to interpret the passage correctly, but t o do so slowly. These
results suggest that both morning and afternoon subjects were trying to interpret the
ellipses using memory for surface form. However, because their memory for the exact
wording of the passages was poor, and difficult to access, they could not help being
influenced by other, plausible, interpretations for what they had read, interpretations
suggested by their general knowledge of how things usually happen. The further from
its antecedent an elliptical VP is, the greater the interference effects produced by those
alternative interpretations, presumably because people are more uncertain about surface
form with increasing distance. These results that do not relate directly to the time-of-day
hypotheses are discussed in niore detail elsewhere (Garnharn and Oakhill, in press).
In summary, the results.showed that when the correct interpretation of an elliptical
VP is different from a more plausible one suggested by context, people’s ability to give
a correct interpretation of that VP diminishes. Not only do they make more errors, but
they also take longer to produce an interpretation and to answer a question that depends
on their having done so. A plausible meaning suggested by context produces interference
even when the ellipsis immediately follows its antecedent. However, the effect is
considerably enhanced when there is an intervening phrase or clause. The effects on
reading and question-answering times were particularly marked for subjects tested in the
morning. Their level of accuracy and the pattenr of their times were similar to those of
subjects tested in the afternoon. However, the very long latencies that the morning
subjects showed for the implausible texts indicate that they had more difficulty than the
afternoon subjects in understanding the anaphors when the syntactically correct interpretation and the plausible one were in conflict.
The results of this experiment, taken together with those of Oakhill (1986b), show
that the effects of time of day on text comprehension cannot be characterized in any
simple manner. The effects depend, in part, on the difficulty of the processing required
to integrate information in different parts of the text. hlorning subjects apparently
prefer to delay processing if they can. If, however, it becomes apparent t o them that
such delaying tactics will lead to failures in comprehension, they do the processing
on-line; however, they find on-line processing more time-consuming than afternoon
subjects.
The complexity of time of day effects is not surprising, given the variety of different
skills required to understand even the simplest of texts. However, the results of the
present experiment help t o explain some previous findings on memory for text and time
of day. Experiments on text nieniory in which subjects have been allowed to read at their
own pace (Folkard and hlonk, 1980; Laird, 1925), showed that immediate memory is
superior in the morning and decreases over the day. These experiments also produced
some evidence that morning subjects read more slowly than afternoon subjects. Another
(unpublished) experiment by hlonk, Adam, Conrad and Folkard (cited by hlarks and
Folkard, 1984) also showed that subjects tested at 8 A.hl. read a text more slowly than
subjects tested later in the day. As Oakhill (1986b) points out, it is not possible to
delay much integrative processing when reading a long complex text (cf. Dooling and
Lacluiian, 1971; Bransford and Johnson, 1972). If the integration is not achieved as the
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
156
Elliptical VPs
text is read, it probably never will be. If the morning subjects in hlonk etal.’s experiment
were attempting to understand the texts, then they would have had to carry out a certain
amount of integrative processing on-line. Furthermore, some of this processing would
probably be difficult enough to slow morning subjects down relative to those tested in
the afternoon. These considerations could explain why morning subjects spent more time
reading the texts. The fact that the morning subjects remembered the stories better is
explained by a general processing principle expounded, for example, by Johnson-Laird,
Gibbs and de hlowbray (1978). The more processing at input, the better the memory
trace.
In some ways, it would be more satisfying if there were an explanation of the underlying causes of the time-of-day effects investigated here. Until recently, the results of
many studies showing changes in performance with time of day were interpreted in
terms of subjects’ putative level of arousal at the time of testing. Physiological measures
taken as indicants of arousal generally show higher values later in the day than early
in the morning, and performance changes were related t o these changes in arousal.
However, a theory of cognitive performance rhythins in which arousal is assumed t o be
unidimensional and rhythmic, does not work (see Eysenck, 1982, for a detailed critique).
For example, it has been shown that different psychological rhythms adjust at diffcrent
rates to changes in the sleeplwake cycle (Folkard and hlonk, 1980), a result that is
impossible to reconcile with the idea that arousal is a general mediator of performance
rhythms. hlore recent experiments (hlonk, Weitzman, Fookson, hloline, Kronauer and
Gander, 1983; Folkard, Wever and Wildgruber, 1983), in which subjects have to adjust
to light/dark cycles that are greater or less than 24 hours, suggest’ that there are at least
two “oscillators” controlling performance rhythms, one of which is linked to the
temperature rhythm. Both of these studies showed that performance on simple tasks
that do not involve memory followed the temperature oscillator. Folkard et al. (1983)
showed that performance on more complex tasks followed a different pattern of
adaptation. However, this research is still in its infancy, and it remains to be determined
what other psychological and physiological functions are governed by this second
oscillator. It is likely t o be some time before these questions are answered, and, even
when they are, it may not be immediately apparent why certain rhythms are linked. In
the meantime, it is useful to gain some further insight into the processing changes that
underlie memory differences at different times of day, rather than to speculate on the
physiological causes of these changes.
REFERENCES
BRANSFORD. J.D. and JOHNSON, h1.K. (1972). Contextual prerequisites for understanding: Some
investigations of comprehension and recall. Jozrnial of Verbal Learribig and Verbal Behavior,
11,717-726.
DOOLING, D.T. and LACHhlAN, R. (1971). Effects o f comprehension on retention o f prase. Jounial
of Experiniental Psychology, 88, 216-222.
EYSENCK, h1.W. (1982). Artention and Arousal: Cogliitioiz and Performarice. Berlin: Springer Verlag.
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016
J. Oakhill aiid A . Gariihmii
157
FOLKARD, S. (1979). Time of day and level of processing. hletnory and Cognition, 7,247-252.
FOLKARD, S. and LIONK, T.H. (1980). Circadian rhythms in human memory. British Journal of
PSyClIOlOgy, 71,295-307.
s., hIONK. T.H., BRADBURY, R. and ROSENTHALL, J. (1977). Time of day effects
in school children’s immediate and delayed recall of meaningful material. British Journal of
PSyCliOlo~,68,4550.
FOLKARD, s., WEVER, R.A. and WILDGRUBER, C. (1983). Multi-oscillatory control of circadian
rhythms in human performance. Nature, 305,223-226.
GARNHAAI, A. and OAKHILL, J.V. (in press). Interpreting elliptical verb phrases. Quarterly Journal
o f Experiniental PsyclioIogy.
JOHNSON-LAIRD, P.N. and STEVENSON, R. (1970). Memory for syntax. Nature, 227,412413.
JOHNSON-LAIRD,P.N., GIBBS, G. and DE hIOWBRAY. J. (1978). Meaning, amount of processing
and memory for words. hlemory and Cognition, 6,372-375.
LAIRD, D.A. (1 925). Relative performance of college students as conditioned by time of day and day
of week. Journal o f Experiiiierital Psycliology, 8 , 5 0 4 3 .
hIARKS, h1.N. and FOLKARD, S. (1984). Diurnal rhythms in cognitive performanec. In J. Nicholson
and H. Beloff (eds.), Psychology Survey, 5 (pp. 63-94). British Psychological Society.
LIILLAR, K., STYLES, B.C. and WASTELL, D.G. (1980). Time of day and rctrkval from long-term
memory. British Journalof Psycliology. 71,407414.
MONK, T.H., WEITZLIAN,E.D., FOOKSON, J.E.. BIOLINE, h1.L.. KRONAUER, R.E. and GANDER,
P.H. (1983). Task variables determine which biological clock controls circadian rhythms
in human performance. A’atzire, 304, 543-545.
hlURPHY, G.L. (1985). Processes of understanding amphora. hlcrnory and Language, 24,290-303.
OAKHILL, J.V. (1986a). Effects of time of day and information importance on adults’ memory for
a short story. Quarterly Journal o f Experiniental Psychology, %A, 4 19430.
OAKHILL, J.V. (1986b). Effects of time of day on the integration of information in text. British
Jounzal o f Psychology, 77.481488.
SACHS, J.S. (1967). Recognition memory for syntactic and semantic aspects of connected discourse.
Percep tiori and Psy chopliysics, 2 , 4 37 4 4 2.
TILLEY, A. and WARREN, P. (1983). Retrieval from semantic memory at different times of day.
Journal o f Experimental Psychology: Learning, hlemory and Cognition, 9, 718-724.
FOLKARD,
Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on April 8, 2016