Sunteți pe pagina 1din 5

Journal of Experimental Psychology

1971, Vol. 88, No. 1, 26-30

EFFECTS OF BLANK VEMSUS NONINFORMATIVE FEEDBACK AND


"RIGHT" AND "WRCNG" ON RESPONSE REPETITION IN
PAIRED-A ^SOCIATE LEARNING
DAVID RIMM, 1 RONALD ROESCH, RONALD PERRY, AND CHRIS PEEBLES
A rizona State University

The experiment investigated the effect of Informationless feedback and


"Right" and "Wrong" on response repetition in paired-associate learning.
Informationless feedback consisted of either the standard Blank or a redundant
reminder of the two response alternatives, with 5s informed in both cases that
such feedback conveyed no information. Both types of feedback resulted in
an equally strong tendency for 5 to repeat responses which they followed, with
the probability of repetition only slightly less than that obtained for "Right."
"Wrong" resulted in marked response suppression. The findings relating to
Informationless feedback were interpreted as supporting Buchwald's view of
the effect of Blank feedback.

The effect on 5 of not receiving feedback on occasion followed later by positive re-
(i.e., of receiving a Blank) following a re- inforcement; whereas incorrect or inappro-
sponse has been a source of considerable priate responding might have been followed
controversy. Both Buchwald (1962) and regularly by some variety of punishment.
Levine, Leitenberg, and Richter (1964) Such an individual would be expected to
have suggested that in the absence of in- have a very strong tendency to interpret
formation specifying the meaning of Blank, Blank as "Right." If, for such an S,
5 tends to respond as though Blank means instead of providing him with Blank feed-
"Right." Spence (1964, 1966a) presented back following certain responses, E simply
data which partially supported this hy- reminded him what the response alterna-
pothesis, with the important modification tives were (i.e., noninformative feedback),
that the information value of Blank also such feedback would be devoid of any real
depends upon whether it occurs within a information. At the same time, 5 pre-
context of "Right" or "Wrong" feedback. sumably would be much less likely to inter-
More recently, Buchwald (1969) contended pret it as correct, since in the past such
that the effect (on probability of repetition) feedback had not been correlated with ap-
of a Blank is identical to that obtained propriate behavior or positive reinforce-
when the feedback is "Right" or "Wrong" ment. Following this line of reasoning, one
but 5 fails to recall the feedback. Thus, would predict that the probability of re-
according to Buchwald (1969), 5 treats the sponse repetition, given the noninformative
Blank as if it had no information value feedback (NIF), would be considerably less
rather than viewing it as confirmation of than that obtaining when the usual Blank
the hypothesis which generated his response was employed. On the Buchwald (1969)
(Levine et al., 1964). hypothesis, however, it would be expected
If, as Levine et al. (1964) propose, S does that the probability of repetition following
interpret Blank as "Right," this has inter- NIF would not differ from the probability
esting implications with respect to the of repetition following Blank feedback.
typical 5s in vivo learning history. For In the present investigation, 5 was pre-
such an individual, correct or appropriate sented with a paired-associate learning task
behavior may have frequently resulted in in which his response was followed by
no immediate feedback of any sort, perhaps "Right," "Wrong," or Informationless feed-
1 back. For half of the 5s, Informationless
Requests for reprints should be sent to David C.
Rimm, Department of Psychology, Arizona State feedback consisted of the standard Blank
University, Tempe, Arizona 8S281. (BL); for the remaining half, such feedback
26
INFORMATION FEEDBACK IN PA LEARNING 27

consisted of E reminding 5 what the two list, the cards were shuffled and the list was presented
response alternatives were (NIF). For again, with this procedure continuing until 5 had
completed a total of 14 trials.
both groups of Ss, feedback was response The procedure for 5s in the NIF condition was
contingent with the sequence of types of identical to the above except that Informationless
pairs (i.e., types of feedback) random so as feedback consisted of E reminding 5 that "one or
to eliminate possible context or contrast two was correct" with 5 informed at the outset that
effects (Spence, 1964). Finally, to "neu- such feedback indicated that he had been correct
half of the time and incorrect half of the time.
tralize" further Informationless feedback,
5s in both conditions were informed in ad- RESULTS
vance that such feedback was in fact in-
tended to convey no information. Since the principle purpose of the study
was to examine response repetition as a
METHOD function of feedback type, the data are
presented in the form of sequential response
Subjects.—The 5s were 16 males and 16 females
enrolled in an introductory psychology course at repetition probabilities for Informationless
Arizona State University whose participation ful- feedback, "Right," and "Wrong."
filled a course requirement. Eight males and 8 The data presented in Table 1 show a
females were randomly assigned to the BL group, marked trial by trial increase in the proba-
with the remaining 16 5s assigned to the NIF bility of response repetition (i.e., the proba-
condition.
Materials.-—The 14 stimulus words, which were bility that a response occurring on Trial N
printed on individual 3 X 5 in. cards, were taken is repeated on Trial N + 1) for both types
from a list prepared by Paivio, Yuille, and Madigan of Informationless feedback for both sexes.
(1968). They were concrete nouns homogeneous A three-way mixed analysis of variance
with respect to concreteness (C ranging from 6.35
to 6.96), imagery (I ranging from 5.73 to 6.60) and (Edwards, 1968, p. 287) revealed a highly
meaningfulness (m ranging from 5.81 to 6.96). significant trials effect, F (12, 336) = 20.66,
Procedure.-—The 5s were informed that they were p < .005, a significant sex effect, F (1, 28)
to learn to associate the numbers one or two to each = 9.75, p < .005, and a significant Trials
of the 14 stimulus words, with "one" being the X Sex interaction, F (12, 336) = 3.15,
correct response for 7 of the words and "two" correct
for the remaining 7. Each 5 in the BL group was p < .005. The latter two significant effects
told that immediately following his response, E corroborate a trend which is apparent upon
would provide oral feedback consisting of "Right" visual inspection of these data: Females
which meant he had been correct, or "Wrong" tended to repeat response with a higher
which meant he had been incorrect, or E would say probability than males over early trials,
nothing. 5 was told that half of the time "nothing"
meant he had been incorrect. but by Trials 10 or 11 the sex differences
For seven of the pairs, following a correct response, disappeared as both sexes began to ap-
E said "Right" and, following an error, E remained proach an asymptote.
silent. For the remaining seven, following a correct Of primary concern is the effect of em-
response, E remained silent and, following an error,
E said "Wrong." The task was self-paced and, when ploying the two different types of Informa-
E responded, it was immediately after 5 had given tionless feedback, which did not approach
his response. Following presentation of the entire significance. The remaining interactions
TABLE 1
MEAN PROBABILITY OF A REPETITION FOLLOWING INFORMATIONLESS FEEDBACK

Trial
Snh'prf rl rl
2 3 4 s 6 7 8 9 10 11 12 13 14

Male
BL .52 .56 .52 .65 .62 .78 .84 .75 .81 .91 .94 .99 .91
NIF .60 ,57 .52 .69 .72 .65 .73 .80 .91 .87 .88 1.0 .93
Female
BL .63 .82 .76 .79 .90 .89 .90 1.0 .99 .93 .94 .98 .97
NIF .66 .82 .83 .76 .84 .85 .86 .88 .72 .88 .86 .89 .94
28 D. RIMM, R. ROESCH, R. PERRY, AND C. PEEBLES

TABLE 2
MEAN PROBABILITY OF A REPETITION FOLLOWING "RIGHT" FEEDBACK
Trial
q , . J J

2 3 4 5 6 7 8 9 10 ll 12 13 14

Male
BL .56 .63 .66 .74 .79 .91 .77 .92 .85 .95 .98 1.0 1.0
NIF .52 .61 .67 .66 .56 .88 .78 .81 .93 .95 .97 .96 .94
Female
BL .75 .78 .89 .91 .88 1.0 1.0 1.0 .98 .98 1.0 .98 1.0
NIF .66 .81 .89 .77 .89 .90 .98 .93 .91 .90 .94 .98 1.0

were also nonsignificant. Thus while both nomial probability, p < .025). For female
sexes showed very strong tendencies to Ss, the comparable mean difference was
repeat responses which had been followed .057, with 14/16 5s showing a higher proba-
by Informationless feedback, whether such bility following "Right" (binomial proba-
feedback was in the form of a Blank or in bility, p < .003). Males and females did
the form of a reminder of the response not differ significantly with respect to the
alternatives appears to have made abso- relative effects of the two classes of feed-
lutely no difference. back, / (30) = 1.33. Thus, "Right" was
Table 2 presents the probability of repe- more effective in facilitating response repe-
tition following "Right." These results tition than Informationless feedback. It
generally parallel those obtained for Infor- should be pointed out, however, that the
mationless feedback. The analysis of magnitude of the differences are quite small
variance again revealed a significant trials relative to the overall effects of either
effect, F (12, 336) = 14.12, p < .005, a "Right" or Informationless feedback on
significant sex effect, F (1, 28) = 11.33, response repetition.
p < .005, and a significant Sex X Trials For 5s in all conditions, response repeti-
interaction, F (12, 336) = 1.94, p < .025. tion following "Wrong" feedback was not
As was the case for Informationless feed- possible for 160 out of 416 trials (32 5s X 13
back, females showed a greater tendency trials) because, on the prior trial, 5 failed
than males to repeat response following to emit any responses which received
"Right" over early trials, with the differ- "Wrong." This suggests that "Wrong"
ence disappearing over later trials. was quite effective in suppressing respond-
As might have been expected, 5s run ing. To examine the effect of "Wrong," a
under the two different types of Informa- somewhat different method of data analysis
tionless feedback did not differ in their re- is required since no probability weighting
sponse to "Right." The Sex X Feedback can be assigned a trial for which repetition,
interaction was also nonsignificant, as was a priori, is impossible. For those trials for
the Trials X Feedback interaction and the which repetition was possible, for each 5,
Sex X Trials X Feedback interaction. the probabilities of repetition following
Of interest is the relative effectiveness of "Wrong" were vincentized into two blocks,
"Right" and Informationless feedback in each containing an equal number of trials
facilitating response repetition. For all 16 (with the median trial excluded when the
males (collapsing across types of Informa- total number of trials was odd). These
tionless feedback), the mean probability of data appear in Table 3.
a repetition (averaged over all trials) fol- An analysis of variance revealed a sig-
lowing "Right" exceeded the mean proba- nificant trial block effect, F (1, 28)
bility of a repetition following Information- = 21.82, p < .005. The effect of feedback
less feedback by .055, with 13 of 16 5s was significant, F (1, 28) = 4.96, p < .05,
showing a higher average repetition of re- and the effect of sex approached signifi-
sponse probability following "Right" (bi- cance, F (1, 28) = 2.91, p < .10. None
INFORMATION FEEDBACK IN PA LEARNING 29

of the remaining effects approached sig- TABLE 3


nificance. The significant and near sig- MEAN PROBABILITY OF A REPETITION FOLLOWING
nificant effects of feedback and sex can be "WRONG" FEEDBACK (EXCLUDING
TRIALS WHERE REPETITION
attributed to a single extraordinary small WAS IMPOSSIBLE)
cell mean (females, BL, Block 2). To
reach significance (Scheffe, 1953, a. = .10) Trial block
a mean difference must exceed .23, and it
1 2
is clear from Table 3 that the only pair-wise
comparisons for sex and feedback which Males
approach this value involve this cell mean. BL .41 .26
NIF .52 .27
While the occurrence of one highly dis- Females
crepant cell mean may be of some interest BL .44 .04
in its own right, this additionl analysis NIF .46 .25
tends to weaken any conclusion affirming
differences attributable to either sex or in a random sequence which included pairs re-
feedback. ceiving "Right" and "Wrong" as well, so that
Another way to assess the effect of initially at least, any contrast effects (Spence,
"Wrong" is to examine the number of trials 1966a) should have balanced out. It might be
prior to total or almost total suppression of argued that the typical 5's learning history
responses scheduled to receive "Wrong." favored his equating Blank with "Right" to
For 74% of the 5s, when two consecutive such a degree that he was compelled to ignore
trials did occur for which repetition follow- the information he initially received about its
ing "Wrong" was impossible because no information value or that he simply forgot
"Wrong" responses occurred on the pre- what he had been told about Blank. While it
vious trial, no additional "Wrong" re- is not at all difficult to reconstruct a learning
history which would account for S's interpret-
sponses were emitted. The mean number ing Blank as "Right," an equally plausible
of trials prior to the first of these two trials account for why 5 would likewise interpret
was as follows: Males, BL, 9.13; males, "one or two was correct" as confirming his re-
NIF, 11.00; females, BL, 7.25; females, sponse is not readily forthcoming, nor is it in
NIF, 9.13. Neither the main effects nor the interest of parsimony to assume that while
the interaction approached significance. Blank and NIF appeared to have had quali-
Taken together, these two analyses of tatively identical effects on repetition proba-
the effects of "Wrong" feedback suggest a bility, they operated through different
definite suppressing effect of such feedback mechanisms.
while providing little evidence for a dif- The present data, then, do not tend to favor
the view put forth by Levine et al. (1964), that
ferential effect due to sex or type of Infor- Blank confirms 5's response. On the other
mationless feedback. hand, these results are in accord with Buchwald
(1969), who assumes that Blank is functionally
DISCUSSION equivalent to "Right" or "Wrong" feedback
which is forgotten (ergo, Blank is equivalent
The principal finding of the present investi- to NIF or to no feedback at all). If the Buch-
gation was that Informationless feedback fa- wald (1969) hypothesis is correct, the high
cilitated the same high degree of response asymptotic probabilities observed for both In-
repetition whether the feedback was in the formationless feedback conditions indicate the
form of the standard Blank or in the form of a cumulative effect on recall of repeated expo-
redundant statement describing the response sure to a given item in the absence of any cor-
alternatives. The overall effects were nearly rective feedback.
as potent as that obtained when S received One rather compelling reason for assuming
"Right" feedback. This was so in spite of the that 5s do equate'Blank with "Right" is that
fact that S was informed at the outset that they sometimes say that they do during post-
Informationless feedback indicated that he experimental interviews (Spence, 1966a,
was as likely incorrect as correct and in spite 1966b).*In the Spence studies, 5s were not in-
of the fact that Informationless items appeared formed about the information value of Blank
30 D. RIMM, R. ROESCH, R. PERRY, AND C. PEEBLES

Such a procedure might favor equating Blank back to e|ffect response repetition. Whether as
with "Right," although there exists an in- a result of poor recall, or the aversive effects of
triguing alternative possibility that 5 repeats "Wrong," 5 happened to change his response
his response following Blank solely because he following "Wrong," he would then receive
is able to recall the response (Buchwald, 1969), Informationless feedback which would tend to
but later concludes (rationalizes) that Blank favor his repeating the second response
must have meant "Right" since it correlated throughout the remainder of the experiment.
with response repetition.
For both "Right" and Informationless feed- REFERENCES
back, female 5s tended to approach asymptote BUCHWALD, A. M. Variations in the apparent
more rapidly than males. One might interpret effects of "Right" and "Wrong" on subsequent
this as supporting the functional equivalence behavior. Journal of Verbal Learning and Verbal
of these two forms of feedback. However, Behavior, 1962, 1, 71-78.
according to Buchwald (1969), covert recall is BUCHWALD, A. M. Effects of "Right" and "Wrong"
independent of feedback; applying this to the on subsequent behavior: A new interpretation.
present results, if under Informationless feed- Psychological Review, 1969, 76, 132-143.
EDWARDS, A. L. Experimental design in psychologi-
back the recall for females differed from that cal research. New York: Holt, Rinehart, &
obtained for males, then under "Right" feed- Winston, 1968.
back, performance for females should similarly LEVINE, M., LEITENBERG, H., & RICHTER, M. The
differ from performance for males since per- Blank Trials Law: The equivalence of positive re-
formance under "Right" also includes the inforcement and non-reinforcement. Psychologi-
recall factor. cal Review, 1964, 71, 94-104.
Had contrast (the tendency for Blank, or its MOSBERG, L. Response elimination in noncorrec-
functional equivalent, to assume reinforcing tion paired-associates learning. Journal of Ex-
perimental Psychology, 1970, 83, 94-100.
value opposite to feedback to which it is PAIVIO, A., YUILLE, J. C., & MADIGAN, S. A. Con-
juxtaposed) been an important source of creteness, imagery, and raeaningfulness values of
variance in the present investigation, one would 92S nouns. Journal of Experimental Psychology,
have expected repetition probability following 1968, 76 (1, Pt. 2).
Informationless feedback to decline with trials RICHARDS, W. S. The effect of variations in pattern-
rather than increase. This is because the ing of feedback and type of prefeedback trial ex-
number of "Wrong"s delivered to S declined posures on response repetition following verbal
to near zero over the 14 trials, whereas the outcomes. Unpublished doctoral dissertation,
number of "Right"s tended to increase. By Arizona State University, 1970.
RIMM, D. C., & BIGGS, B. Supplementary report:
contrast, Informationless feedback might have Effects of variations in interpair interval and type
been expected to take on a negative valence, of feedback on paired-associate learning. Journal
thereby suppressing responses which it fol- of Verbal Learning and Behavior, 1969, 8, 159-160.
lowed. Such was clearly not the case, although SCHEFFE, H. A. A method for judging all contrasts
the present investigation can hardly be viewed in the analysis of variance. Biometrika, 1953, 40,
as a fair test of contrast since an effort was 87-104.
made to minimize such effects. Additionally, SPENCE, J. T. Verbal discrimination performance
there is evidence that contrast may influence under different verbal reinforcement combina-
the valence of Blank feedback (Richards, 1970 ; tions. Journal of Experimental Psychology, 1964,
67, 195-197.
Spence, 1964). SPENCE, J. T. Effects of verbal reinforcement com-
Consistent with some previous findings bination and instructional condition on the per-
(e.g., Mosberg, 1970; Rimm & Biggs, 1969), formance of a problem-solving task. Journal of
the present experiment provides support for Personality and Social Psychology, 1966, 3,
the view that "Wrong" does lead to suppres- 163-170. (a)
sion of responses which it follows, given a small SPENCE, J. T. The effects of verbal reinforcement
number of response alternatives. However, combination on the performance of a four-alterna-
the tendency to emit very few "Wrong" re- tive discrimination task. Journal of Verbal Learn-
ing and Verbal Behavior, 1966, 5, 421-428. (b)
sponses during latter trials in the present study
also reflects the power of Informationless feed- (Received August 15, 1970)

S-ar putea să vă placă și