Você está na página 1de 4

Journal of Experimental Psychology

1969, Vol. 79, No. 1, 18-21

MONITORING EYE MOVEMENTS DURING THE LEARNING


OF LOW-HIGH AND HIGH-LOW MEANINGFULNESS
PAIRED-ASSOCIATE LISTSx
P. D. McCORMACK AND T. E. MOORE
Carleton University, Ottawa, Ontario, Canada

Eye movements of 5s were monitored as they attempted to learn an H-L or


an L-H m' PA list. An earlier study revealed that more sustained attention
is directed to low-m' than to high-m' materials. Predictions arising from
this finding were compared with those stemming from the 2-stage con-
ceptualization. The data of this study gave more support for the latter
position.

A series of verbal learning investigations Hannah, 1967), one group of Ss (Group


(e.g., McCormack & Haltrecht, 1966) in H) learned a PA list comprised of high-m'
which eye movements were photographed PA syllables while a second group (Group
has provided support for the position that L) was required to learn a list made up of
the learning of paired-associate (PA) lists low-m' items. Eye movements were photo-
is a two-stage process (Underwood, Run- graphed during the first eight trials with 5"s
quist, & Schulz, 1959) involving response being interrupted for approximately 1 min.
consolidation and S-R hookup. These between Trials 4 and 5 to enable film
studies have used PA materials of middle changeover. The two-stage notion would
meaningfulness (m') (Noble, 1961) and the lead to the prediction that the fixation func-
study-test method. As learning progressed, tions of Group L would cross over later in
response viewing time decreased while time time than those of Group H, the low-m'
spent viewing the stimulus increased. These responses in the former case requiring a
stimulus and response functions have crossed longer period of consolidation. This was not
over during early learning trials resulting the case since both groups exhibited the
in a disproportionate amount of stimulus usual functions, with response viewing pre-
viewing on later trials. Presumably, once dominating only on very early trials. Nor
response consolidation has taken place, did the two groups differ with respect to
stimulus viewing predominates despite the their eye movement patterns, the usual S-R
fact that the study-test method demands that to S-R-S viewing being obtained. There
the stimulus receive double exposure. The was some indication, however, that Group L
Ss also have consistently evidenced a pattern ,9s experienced difficulty with consolidation
of looking first to the stimulus and then to of the low-m' responses. This appeared in
the response during early learning trials, but the form of a regression to predominant re-
later in learning have shown a pattern of sponse viewing on Trial 5 following the film
S-R-S viewing. Such phenomena have been change. This renewed attention to the re-
demonstrated to be stable from day to day sponses appeared only in the data of Group
(McCormack, Haltrecht, & Hannah, 1967) L and was consistent with the overall in-
and to be absent in nonlearning situations feriority in the learning of those Ss as com-
(McCormack, Haltrecht, & Hannah, 1966). pared to that of Group H.
In another investigation in which eye A somewhat unexpected outcome of this
movements were monitored (McCormack & study was that the total fixation time of
J-This research was supported by a grant-in-aid Group L 5"s exceeded that of Group H,
from the Associate Committee on Experimental which could not be accounted for by less
Psychology of the National Research Council of attention over trials on the part of the latter
Canada (Grant APA-78). The authors are in-
debted to Wayne Fingus, Ted Hannah, and Kathy 5"s, a phenomenon occasionally seen in other
Haycock for data collection. experiments. This finding of more sus-
18
EYE MOVEMENTS AND THE TWO-STAGE MODEL 19

tained attention to low-m' materials leads to end of every 4 trials to enable E to change film.
the prediction that response viewing will The lists were presented in each of four different
random orders in an attempt to minimize serial
predominate during the learning of a list learning and the session was terminated at the
comprised of high-m' stimuli and low-m' completion of 16 trials. The fixation and eye move-
responses (List H-L) and that stimulus ment data were analyzed for the first 8 trials only.
viewing will prevail on a list with low-m'
stimuli and high-m' responses (List L-H). RESULTS AND DISCUSSION
On the other hand, the two-stage conceptual-
ization would lead to the expectation that Learning data.It may be predicted from
response viewing would exceed stimulus the two-stage conceptualization that L-H
viewing during the early learning trials of items will be learned with greater facility
List L-H and that it would be further ex- than H-L pairs, response consolidation in
tended in time during the learning of an the former case requiring relatively less time.
H-L list. The present experiment was per- Two recent reviews (Goss & Nodine, 1965;
formed with the purpose of evaluating these Underwood & Schulz, 1960) have indicated
alternatives. that this is generally the case. The findings,
however, have been by no means unequivo-
METHOD cal. An examination of the correct response
Subjects and materials.The 5s were 21 male data of the present investigation reveals
and 15 female students from introductory psychol- some support for this prediction, the L-H
ogy courses who were assigned at random to the group being superior to the H-L group over
two groups, Group H-L and Group L-H, with the the last 8 of the 16 trials. However, an
restriction that there be 10 males and 8 females in analysis of variance (ANOVA) of these
the former group and 11 males and 7 females in the
latter. Each list was made up of seven pairs of data failed to reveal any statistically depend-
CVCs, the L items having m' values ranging from able Trials X Groups interaction, the only
1.23 to 1.55 while the m' values of the H items reliable effect being that for trials, F (15,
ranged from 2.33 to 2.48. The H-L list was as 510) = 65.19, p < .001. One possible rea-
follows: YAN keb, MOK tuh, FIR sew, FUT daq, son that the analysis did not demonstrate
SOG fiy, HEV cug, CAW rwv. These pairs were rer
versed to form the L-H list. superior performance on the part of the L-H
Apparatus and procedure.The lists were pro- group, apart from the equivocality of earlier
jected by means of a timer-controlled Carousel findings, stems from the relative brevity of
unit to a white-surfaced screen 1.4 m. from S the two lists, a restriction imposed by the
and1 2.4 m. from the projector. Eye movements nature of the apparatus. In order to have
were recorded by means of a head-mounted eye-
marker camera (Mackworth & Thomas, 1962) four different serial orders of each list, and
consisting of a 2 X 8 mm. motion-picture unit film- at the same time provide continuous filming
ing at a speed of IS frames per second and a over each four-trial segment, no more than
periscope that reflects a spot of light from the seven CVC pairs could be used.
cornea to the film. The film, when developed,
shows the PA slides and, superimposed on each, a Fixation and eye movement data.The
bright circular spot indicating the approximate major fixation findings are summarized in
position of actual fixation. The data were analyzed Fig. 1 where mean number of frames per
frame by frame by examining the S-R slides syllable pair is shown as a function of the
through a film viewer and determining whether S first eight trials for each group: the solid-
was fixating the stimulus or the response. This
method of analysis was highly dependable, the line curves depicting fixations of the stim-
between- and within-E reliability coefficients both ulus, while response fixations are represented
being .98. by broken-line functions. For any given
The seven CVC pairs were presented consecu- trial, total frames of stimulus and response
tively on any given trial. During this study phase, may be calculated by adding the ordinate
S was instructed to examine the pairs in an
attempt to learn them. Following a 2-sec. inter- values for the open and closed circles. The
val, the seven stimuli were exposed with S at- difference between this value and 21 (1.4 X
tempting to recall the appropriate responses dur- 15) represents frames involving fixations of
ing this test phase. Slides changed every 2 sec. with
a 1.4-sec. exposure of each. The intertrial interval the center portion of the display, blinking,
was 2 sec. and a 1-min. rest was introduced at the looking away, etc. The data of Group H-L,
20 P. D. McCORMACK AND T. E. MOORE

H-L two-stage notion provided that it is assumed


that Stage 1 precedes Stage 2 in time.
ANOVA of the fixation data shows these
effects to be statistically dependable.
ANOVA for the H-L group revealed a pre-
ponderance of response viewing in the form
of a reliable conditions (S vs. R) effect,
F (1, 255) = 46.51, p < .001, as well as a
convergence of the stimulus and response
functions over trials which was revealed
by a dependable Trials X Conditions inter-
action, F (7, 255) = 2.15, p < .05. ANOVA
of the L-H data indicated only a Trials X
Conditions interaction, F (7, 255) =2.24,
p < .05, which reveals the statistical depend-
ability of the transition from response to
stimulus viewing on the part of these ^"s.
V.. These analyses were supplemented by data
of that trial for each 5" where the stimulus
was first viewed out of proportion to the
response. The mean of 4.78 for Group H-L
was found to be reliably larger than that of
2.72 for Group L-H, t (34) = 3.61, p <
.001; thus, the functions converged for both
groups, the crossover point occurring earlier
for Group L-H.
In agreement with earlier studies where
FIG. 1. Mean number ot frames per syllable response consolidation was expected to be
pair as a function of the first eight trials for the extended (e.g., McCormack, Hannah, Brad-
two groups of 5s.
ley, & Moore, 1967) was the increased view-
ing of the response by 14 of the 18 5"s of
which indicate a preponderance of response Group H-L from Trials 4 to 5 where a
viewing, are consistent with the earlier rind- 1-min. interval was introduced to enable E
ing that low-m' materials evoke sustained to change film. The binomial probability of
attention. This tendency to view the re- 14 or more ,9s exhibiting this phenomenon
sponse out of proportion to the stimulus is .0154. Thus, response consolidation ap-
throughout most of the first half of the pears to be incomplete at the time of the
learning task is also in line with what would first film change for Ss of the H-L group,
be expected on the basis of the two-stage a finding which is also in line with the two-
notion, since the consolidation of these stage notion. The eye movement data also
low-m' responses should be relatively diffi- were subjected to ANOVA with one move-
cult. The two-stage conceptualization also ment being recorded for S-R viewing, two
is supported by the finding that the stimulus movements for an S-R-S pattern, etc. Ear-
and response functions of Group H-L con- lier studies have shown a transition from
verge with trials, indicating a transition from S-R to S-R-S viewing as learning pro-
Stage 1 to Stage 2. The data of Group gresses, a phenomenon interpreted as sup-
L-H, on the other hand, are not totally con- porting the two-stage concept. This also
sistent with the prediction that low-m' ma- was the case in the present investigation,
terials will receive greatest attention since with a reliable trials effect revealing that
the response is viewed out of proportion to the slope of the function relating eye move-
the stimulus during early learning trials, a ments to trials was dependable greater than
finding which would be predicted from the zero, F (7, 238) = 4.11, p < .001. Although
EYE MOVEMENTS AND THE TWO-STAGE MODEL 21

the function for Group H-L lagged behind MACKWORTH, N. H., & THOMAS, E. L. Head-
that for Group L-H for the four trials, this mounted eye-marker camera. Journal of the
Optical Society of America, 1962, 52, 713-716.
was not demonstrated to be statistically de- McCoRMACK, P. D., & HALTRECHT, E. J. Moni-
pendable since the F ratio for Trials X toring eye movements under two conditions of
Groups did not reliably exceed its expected presentation of paired-associate materials.
value of unity under the null hypothesis. Canadian Journal of Psychology, 1966, 20, 154-
Conclusion.The data of the present in- 159.
MCCORMACK, P. D., HALTRECHT, E. J., & HANNAH,
vestigation are consistent with the two-stage T. E. Monitoring eye movements in nonlearn-
notion but are not in agreement, particularly ing situations. Psychonomic Science, 1966, 6,
those of Group L-H, with what would be 371-372.
expected from the findings of the earlier McCoRMACK, P. D., HALTRECHT, E. J., & HANNAH,
study where 5s were observed to attend T. E. Monitoring eye movements during the
learning of successive paired-associate lists.
more to low-m' materials. The early cross- Journal of Verbal Learning and Verbal Behavior,
over of the fixation functions for Group L 1967, 6, 950-953.
of the earlier investigation, however, does MCCORMACK, P. D., & HANNAH, T. E. Monitor-
not support the two-stage conceptualization ing eye movements during the learning of high
since these responses would be expected to and low meaningfulness paired-associate lists.
Psychonomic Science, 1967, 8, 443-444.
be relatively difficult to consolidate and so MCCORMACK, P. D., HANNAH, T. E., BRADLEY,
should lead to more sustained attention dur- W. J., & MOORE, T. E, Monitoring eye move-
ing the consolidation phase. On the other ments under conditions of high and low intralist
hand, the increased viewing of the response response (meaningful) similarity. Psychonomic
following the film change on the part of Science, 1967, 8, 517-518.
McGuiRE, W. J. A multiprocess model for paired-
Group L 5s, a phenomenon observed for associate learning. Journal of Experimental Psy-
Group H-L in the present study, is in line chology, 1961, 62, 335-347.
with the two-stage notion. In order to ac- NOBLE, C. E. Measurement of associative value
count for all of the data gathered so far, (a), rated associations (a') and scaled meaning-
fulness (m') for the 2100 CVC combinations of
it may become necessary to invoke a multi- the English alphabet. Psychological Reports,
stage model such as that proposed by Mc- 1961, 8, 487-521.
Guire (1961) or by Underwood, Ekstrand, UNDERWOOD, B. J., EKSTRAND, B. R., & KEPPEL, G.
and Keppel (1965). Such a model might An analysis of intralist similarity with experi-
emphasize certain aspects over others where ments on conceptual similarity. Journal of
Verbal Learning and Verbal Behavior, 1965, 4,
pairs are homogeneous with respect to m', 447-462.
as in the earlier study, and alter its emphasis UNDERWOOD, B. J., RUNQUIST, W. N., & SCHULZ,
for situations where pairs are heterogeneous R. W. Response learning in paired-associate
with respect to m', as in the present investi- lists as a function of intralist similarity. Journal
gation. of Experimental Psychology, 1959, 58, 70-78.
UNDERWOOD, B. J., & SCHULZ, R. W. Meaningful-
REFERENCES ness and verbal learning. Chicago: Lippincott,
1960.
Goss, A. E., & NODINE, C. F. Paired-associates
learning. New York: Academic Press, 196S. (Received January 2, 1968)

Você também pode gostar