Sie sind auf Seite 1von 6

Journal of Experimental Psychology

1965, Vol. 70, No. 1, 51-56

PAIRED-ASSOCIATES LEARNING WITH VARYING


RELATIVE PERCENTAGES OF OCCURRENCE
OF ALTERNATIVE RESPONSE MEMBERS:
INFLUENCE OF INSTRUCTIONS 1
ALBERT E. GOSS
University of Massachusetts

Acquisition and performance with divergent PA units were determined


for 4 relative percentages of occurrence of alternative response members
(RP) combined with 4 list conditions under matching, regular, and
maximizing instructions. Under all 3 instructions and irrespective of
RP, percentages of anticipations and actual occurrences of more frequent
response members were essentially the same for a single unit presented at
constant, short, and at variable, long interunit intervals. With RP of
70-30, 80-20, and 90-10, for a single unit presented with randomly
paired intervening stimuli and for 4 units, matching instructions reduced
differences between anticipations and actual occurrences. Maximizing
instructions increased those differences but did not occasion 100%
anticipation.

Considered together, the three ex- was empirical: to determine direction


periments reported here provide in- and extent of effects of different sets
formation about effects of different of instructions on acquisition and
sets of instructions on acquisition and performance both across RP and list
performance with divergent paired- conditions and for different RP and
associates (PA) units. These units list conditions alone and in combina-
were each constituted of a stimulus tion. Another reason for varying
member and two response members, instructions was theoretical: to deter-
one of which accompanied the stim- mine whether or not theories to ex-
ulus member on P% of its occurrences, plain acquisition and performance
the other of which accompanied the with divergent PA units need be
stimulus member on (1 — P)% of its elaborated to allow for effects of
occurrences. Relative percentages of instructional variables. The w-choice
occurrence of alternative response or probability-learning task (Estes,
members (RP) were varied as were 1964) can be considered a special case
list conditions which represented dif- of divergent PA units. Therefore,
ferences in number of units and findings of effects of instructional
manner of presentation of single units. variables on acquisition and perform-
One reason for varying instructions ance with divergent PA units might
1
Supported by Contract 2691 (00) be-
have interesting and important rami-
tween the Office of Naval Research, Training fications for theories of probability
and Personnel Branch, and the University of learning.
Massachusetts. Marilyn E. Sugerman col- METHOD
laborated on one experiment. Nancy C.
Farrick was responsible for running half of Design.—Three different sets of instruc-
the 5s of that experiment and all of those of tions were combined orthogonally with RP
the other two. She and Helen L. Rowell of 60-40, 70-30, 80-20, and 90-10 and with
contributed to analyses of the data of the list conditions of 4 units or 1 unit and three
latter experiments and of all three experiments different manners of presentation of the
combined as well as to preparation of the single unit. The three different manners
manuscript. were with a constant 2-sec. interval between
51
52 ALBERT E. GOSS

each presentation of the single unit; with the two later experiments were in the times
intervals from 2 to 38 sec. between pres- the experiments were run and in Es. The
entations about a mean of 20 sec.; and with former experiment was run during the spring
intervals from 2 to 38 sec. between pres- of one year, the experiment with matching
entations about a mean of 20 sec., but with instructions was run in the fall of that year,
the intervals filled by presentations of ran- and the experiment with maximizing instruc-
domly paired stimulus and response members. tions was run in the spring of the next year.
These different manners of presentation are There is no reason to believe that 5s of the
described in greater detail by Goss and three experiments, who were drawn from
Sugerman (1961). Coded by them as 1 unit, among undergraduates enrolled in the course
1 unit spaced, and 1 unit random, they are, in introductory psychology during those
henceforth, coded here as 1, IS, and 1R, semesters, differed in any ways related to
respectively. The 4-unit list is coded here acquisition with divergent PA units.
as 4. The 5s in the Goss and Sugerman experi-
The three different sets of instructions are ment were run by two female Es. No differ-
referred to as regular, matching, and maximiz- ences attributable to Es were observed. In
ing. The regular instructions were those the two later experiments, one of these Es ran
employed for acquisition with conventional all of the 5s.
PA units of each stimulus member accom- Lists.—The divergent PA units were con-
panied by but one response member. The stituted of CVCs of high meaningfulness and
5s were not informed that each stimulus low formal similarity to each other. The
member would be accompanied by two units were WOM accompanied by TEX or
alternating response members. Nor were LIK, RUF by CEN or DOL, FES by vie or BAL,
they told the manner in which they should and NAV by HOB or SUR. When presented as
anticipate those members except to spell out four units, TEX, DOL, Vic, and HOB all occurred
the stimulus they expected to occur before on 60%, 70%, 80%, or 90% and LIK, CEN,
the response member was exposed. BAL, and SUR all occurred on 40%, 30%, 20%,
The matching instructions involved in- or 10% of the presentations of their stimulus
forming 5s that each stimulus member would members, or the converse. When presented
be accompanied by two response members as 1 unit, which response member was the
which would alternate. The 5s were told more frequent was also counterbalanced.
further to anticipate the response members Goss and Sugerman (1961) describe other
with frequencies which were the same as the details of the lists and of the semirandom
actual frequencies of occurrence of those orders of presentation of the units.
members. The maximizing instructions also Apparatus and procedure.-—A Hunter card-
involved informing 5s that each stimulus master was employed to present the units;
member would be accompanied by response the rate was 2:2 sec. There were 120 trials
members which would alternate. They were with the 4-unit lists with all units presented
told further, however, to always anticipate on each trial; there were 120 presentations
the more frequent response member of a unit. of the single units under the 1, IS, or 1R list
The design was accomplished in three conditions.
successive experiments. The first experiment, Subjects.—Within each experiment there
with regular instructions, has been described were 16 combinations of RP and list condi-
previously (Goss & Sugerman, 1961). Once tions. These combinations were administered
acquisition and performance with divergent in counterbalanced cycles until the requisite
PA units under regular instructions had been number of 8 5s within each had been reached.
determined for RP and list conditions alone Assignment to each combination was in order
and in combination, comparisons with ac- of 5s' appearance within a constraint of equal
quisition and performance under other sets numbers of males and females within each
of instructions were warranted and desirable. combination. There were 128 5s within each
The second experiment was with matching experiment, for a total of 384. All 5s were
instructions and the third experiment was naive with respect to exposure to lists of
with maximizing instructions. divergent PA units. Some 5s within each
Goss and Sugerman varied RP from 50-50 experiment, however, had served previously
to 100-0 in successive changes of 10%. Since in other experiments with conventional PA
instructions to match or to maximize are of units.
little or no consequence for RP of 50-50 or
100-0, these levels were omitted in the two RESULTS AND DISCUSSION
later experiments. Except for the omission
of RP of 50-50 and 90-10, the only known As previously, the response measure
differences between the first experiment and was percentage of anticipations of the
INSTRUCTIONS AND RESPONSE FREQUENCY IN PAs 53

95.6*
92

85.2
80.7

1 IS 1R 4 1 IS 1H 1 IS 1R
MATCHING REGULAR MAXIMIZING
FIG. 1. Means of percentages of anticipations of the more frequent response members for
each of the combinations of instructions and lists in relation to actual percentages of occurrence
of the more frequent response members. (Differences significant at p < .05 are indicated by
an asterisk.)
54 ALBERT E. GOSS

TABLE 1
ANALYSES OF VARIANCE ON PERCENTAGES OF ANTICIPATIONS OF THE MORE
FREQUENT RESPONSE MEMBERS DURING ALL Six AND
THE LAST THREE BLOCKS OF TRIALS

All Six Blocks Last Three Blocks


Source
df MS F df MS F
Instructions (I) 2 2,458.08 3.97* 2 2,232.75 7.11**
RP 3 89,733.75 145.09** 3 41,646.13 132.61**
List conditions (L) 3 2,313.34 3.74* 3 4,690.09 14.93**
I X RP 6 1,929.71 3.12** 6 1,302.86 4.15**
I XL 6 1,049.72 2.15 6 581.73 1.85
RPXL 9 767:16 1.57 9 553.78 1.76
I X RP X L 18 488.34 18 334.82 1.07
Error (b) 336 618.47 336 314.04
Trials (T) 5 37,680.46 327.20** 2 360.30 6.17**
TX I 10 197.79 1.72 4 45.41
T X RP 15 161.09 1.40 6 90.05 1.54
T XL 15 3,039.13 26.39** 6 119.23 2.04
T X I X RP 30 86.36 12 42.86
T XIXL 30 158.70 1.38 12 56.32
T X RP X L 45 172.26 1.50* 18 30.04
T X I X RP X L 90 117.05 1.02 36 85.91 1.47*
Error (w) 1672" 115.16 664" 58.38

» 8 estimated values subtracted from df.


*p< .05.
**-p < .01.

more frequent response members dur- instructions were slight. Fewer an-
ing successive blocks of 20 trials. For ticipations occurred under both of
brevity, means of percentages for these sets of instructions than under
successive separate blocks of trials are maximizing instructions. Both of the
not shown. Means of percentages highly significant Fs for RP reflected
during the last three blocks of trials the direct relationship between per-
combined are presented at the top of centages of anticipations and of actual
each rectangle of Fig. 1. Means of occurrences of the more frequent
percentages of anticipations are first response members found in other
examined in terms of differences experiments (Brown, Osterhout, &
among their absolute values through Voss, 1962; Voss, Thompson, &
blocks of trials as functions of in- Keegan, 1959). The significant Fs
structions, RP, and list conditions. for list conditions are considered
Direction and extent of differences below in connection with effects of
between anticipations of response trials.
members and their actual occurrence The Fs for I X RP were also
are then examined. significant. For all six and for the
Absolute values.—Table 1 summa- last three blocks of trials, percentages
rizes analyses of variance on percent- of anticipations were in the order
ages of anticipations during each of regular < matching < maximizing
all six and the last three blocks of with RP of 60-40 and 70-30, in the
trials. In each analysis, the F for order matching < maximizing < reg-
instructions was significant. Both ular with RP of 80-20, and in the order
across all six and the last three blocks maximizing < matching < regular
of trials, differences between anticipa- with RP of 90-10. None of these
tions under matching and regular orders is the same as the overall order
INSTRUCTIONS AND RESPONSE FREQUENCY IN PAs 55

for instructions of matching = regular during Trials 1-20, and the converse
< maximizing. during the last four blocks of trials.
Neither of the Fs for 1 X L was The pattern of this interaction sub-
significant. Across all six blocks of stantially accounts for the difference
trials, maximizing instructions pro- between the orders of the list condi-
duced more anticipations than did tions across all six and across the last
regular or matching instructions for three blocks of trials. Across all six
all but the 1 list condition. Across blocks, the initial advantage of the
the last three blocks of trials, maxi- IS list condition was sufficient to
mizing instructions produced more produce the order 1 < 1R < 4 < IS.
anticipations than did matching and During the last three blocks, however,
regular instructions for all four of the the order was 1 < IS < 1R = 4.
list conditions. Thus, while the Both Fs for list conditions were
higher percentages of anticipations significant.
under maximizing than under match- The significant T X RP X L inter-
ing and regular instructions suggested action reflected a greater reversal in
by the overall Fs for instructions did the advantage of the 1 and IS to the
not hold at each RP, with the one 1R and 4 list conditions between
exception, the higher percentages did Trials 1-20 and the last four blocks of
hold for each list condition. trials with RP of 60-40 and 70-30
Examination of the rank orders of than with RP of 80-20 and 90-10.
means for each of the 16 combinations For the last three blocks of trials,
of RP and list conditions under the only one interaction involving trials
different instructions also indicated was significant. However, the pattern
that maximizing instructions did, in of changes through trials precluded
general, produce more anticipations meaningful interpretation of this
than did regular or matching in- interaction.
structions. However, neither of the Anticipations and actual occurrences.
Fs for I X RP X L was significant. —Figure 1 indicates the direction and
The highly significant F for trials extent of differences between percent-
through all six blocks reflected a ages of anticipations of response
typical negatively accelerated increase members and their actual occurrences.
in percentage of anticipations of the The heterogeneity of the variances
more frequent response members. among the combinations of instruc-
The F for trials through the last three tions, RP, and list conditions was not
blocks was also significant. More considered an important limitation on
anticipations occurred during Trials interpretation of the outcomes of the
101-120 than during Trials 61-80 for analyses of variance. But such heter-
which the mean was slightly higher ogeneity seemed an important limita-
than the mean for Trials 81-100. tion on assessment of differences
However, none of these differences between percentages of anticipations
was sufficient to be of practical im- and of actual occurrences. Accord-
portance or to vitiate the subsequent ingly, a two-tailed t (7) was computed
treatment of the last 60 trials as for each of the 48 combinations with
essentially homogeneous. an error term based on the SD for
The significant T X L interaction that particular combination. Differ-
for all six blocks was attributable to ences significant at p < .05 are noted
higher percentages of anticipations in Fig. 1.
with the 1 and IS list conditions than Under all three sets of instructions,
with the 1R and 4 list conditions only 2 of the 24 fa for RP combined
56 ALBERT E. GOSS

with 1 and IS list conditions were task approximated the 1 list condition
significant. Under matching instruc- of the present experiments; the RP was
tions, 2 of the 8 t's for RP combined 70-30. Under Group 2 instructions,
with 1R and 4 list conditions were percentages of anticipations of the more
frequent stimulus were close to 70%.
significant. In contrast, under regular
Under Group 6 instructions, percentages
and maximizing instructions, respec- of anticipations rose to about 90%. In
tively, 5 and 6 of the 8 t'a for RP com- contrast, with 1 and IS list conditions
bined with 1R and 4 were significant. under maximizing instructions, anticipa-
The pattern of nonsignificant and tions matched actual occurrences at each
significant t's, in conjunction with the level of RP. One or more of many
outcome of the analysis of variance for differences between the McCracken et al.
the last three blocks of trials, supports and the present experiments in lists and
four conclusions. First, regardless of procedures might underlie the discrepant
instructions, during the last three blocks outcomes. Thus, for example, main-
of trials percentages of anticipations for tenance of instructions-induced sets may
combinations of RP with 1 and IS list have been less difficult with their simple
conditions matched actual percentages of nonmechanical arrangement for present-
occurrence. Two significant t's among ing the stimuli than with the complicated,
24 are best interpreted as due to chance. noisy cardmaster of the present ex-
Second, regardless of instructions, per- periments.
centages of anticipations for RP of 60-40 Fourth, regardless of any differences
combined with 1R and 4 list conditions in effects of regular and maximizing
also matched actual percentages of instructions, under these instructions
occurrence. Because of the extent of the anticipations tend to exceed actual occur-
increase in the percentage of anticipa- rences for RP for 70-30, 80-20, and 90-10
tions over 60%, the combination of RP combined with the 1R and 4 list
of 60-40 and the 4 list condition under conditions.
maximizing instructions may be an
REFERENCES
exception.
Third, instructions to match may have BROWN, A., OSTERHOUT, C., & Voss, J. F.
reduced differences between anticipa- The effect of one probabilistic verbal
tions and actual occurrences for RP of association upon the concomitant acquisi-
70-30, 80-20, and 90-10 combined with tion of a second association. J. verbal
the 1R and 4 list conditions. Instruc- Learn, verbal Behav., 1962, 1, 48-53.
tions to maximize may have increased ESTES, W. K. Probability learning. In A. W.
Melton (Ed.), Categories of human learning.
those differences for the same combina- New York: Academic Press, 1964.
tions. The largest absolute differences Goss, A. E., & SUGERMAN, M. Paired-
were under instructions to maximize for associates learning with varying relative
RP of 60-40 combined with the 4 and of percentages of occurrence of alternative
70-30 combined with the 1R and 4 list response members. /. exp. Psychol., 1961,
conditions. Thus, theories of learning 62, 24-34.
and performance with divergent PA MCCRACKEN, J., OSTERHOUT, C., & Voss,
um°ts, including theories of probability J. F. Effects of instructions in probability
learning, may have to be elaborated to learning. /. exp. Psychol., 1962, 64, 267-
allow for effects of different sets of 271.
instructions. Voss, J. F., THOMPSON, C. P., & KEEGAN,
The Group 2 and Group 6 instructions J. H. Acquisition of probabilistic paired
associates as a function of S-Ri, S-R2
employed by McCracken, Osterhout, probability. /. exp. Psychol., 1959, 58,
and Voss (1962) resembled the matching 390-399.
and maximizing instructions of the
present experiments, respectively. Their (Received April 6, 1964)

Das könnte Ihnen auch gefallen