JOURNAL
OF EXPERIMENTAL
Implicit
CHILD
PSTCHOLOGY
Associative
Recognition JAMES
W.
6, 52-66 (1968)
Responses by Young
HALL Northwestern
AND
WILLIAM
and
False
Children1
B.
WARE
University
Two experiments concerned with implicit verbal behavior in young rhildren are reported. In Exp. I, 86 elementary school children were presented a word list containing several critical stimulus words (e.g., SCISSORS) that possess strong natural-language associates (e.g., CUT). Next, a recognition list was presented that included (a) the associates of the critical stimulus words, and (b) control words (i.e., words not associated with any words on the learning list). It was assumed that if the experimental words had been elicited as implicit responses by the critical stimulus words, then these experimental words frequently would be falsely recognized as having appeared on the learning list. False recognitions did occur more frequently to experiment,al than to control words, confirming results obtained with adults by other investigators. Experiment II employed 24 kindergarten and 24 third-grade children. Prior to presentation of the learning list certain of the critical stimulus words from that list were used as stimulus words in PA learning, with nonassociated words used as response terms (e.g., SCISSORS-BATH). Then the learning and recognition lists were presented. No false recognition effect was obtained either for the associates developed during PA learning or for the natural-language associates of the critical stimulus words that had been used in PA learning. However, a reliable false recognition effect was obtained for natural language associates of critical stimulus words that had not been used in PA learning. These results suggest that the PA learning functioned to temporarily extinguish the natural-language associates but did not raise the strength of the “new” responses to the point that they nere produced as IARs. Experiment II also examined false recognition as a function of CA, and found this effect reliably greater for the younger than for the older Ss. This finding is discussed with respect to the development of verbally mediated behavior. ‘This research was supported by a grant from Northwestern University. The investigators are indebted to the following Evanston school officials for their cooperation: Mr. Frank Christensen, Director of Pupil Services, Mr. Francis X. Vogel, Principal of Central School, and Mrs. Berniece Hoffman, Principal of Lincoln School. Thanks also are due Albert Erlebacher for his helpful advice regarding statistical analyses. 52
IMPLICIT
ASSOCIATIVE
RESPONSES
53
Implicit responses have been given increasing attention in theoretical developments relating to complex behavior. One kind of implicit response made to a verbal unit has been called by Bousfield, Whitmarsh, and Danick (1958) the representational response (RR). The RR is the act of perceiving a particular verbal unit. A second implicit response has been called the implicit associative response (IAR) by IJnderwood (1965). The IAR is conceived as being elicited by the stimulus properties of the RR due to prior associations of the two. These associations may occur because of the normal language experiences of an individual in which case the particular IAR to a given word often is assumed to be predictable from word-association norms. They may also be established experimentally as in mediational transfer studies employing paired-associate (PA) learning. Recently Underwood (1965) reported an experiment designed to examine conditions under which IARs occur to verbal units and the subsequent effect of these IARs on accuracy of recognition. A list of 200 words was presented to college students who were asked to report whenever they recognized a word that had appeared earlier in the list. Late in the list certain words (experimental or E words) appeared which were strong natural-language associates of particular words (critical stimulus or CS words) that had appeared earlier in the list. It was assumed that if the experimental words had previously been elicited by critical stimulus words as IARs, these E words frequently would be reported as having appeared before. This prediction was confirmed by Underwood, in that for three of the five classes of words used, false recognition of E words was reliably more frequent than for control (C) words. Davis (1967) recently has reported results essentially in agreement with those of Underwood. In an extension of Underwood’s reasoning, Wallace (1967133, assuming greater IAR activity for normal compared to retarded Ss, predicted a lower rate of TAR-produced false recognitions by retardat,es. Using a procedure similar to that described above, Wallace confirmed this precliction. Wallace (1967a) has described a second experiment in which associates developed in the laboratory through paired-associate (PA) training apparently replaced natural-language associates as IARs to the CS words. In the present paper two experiments are reported in which a modification of the false recognition procedure used by Underwood (1965) was employed to study IAR production in young children. EXPERIMENT
I
Experiment I was designed simply to discover whether or not the basic false-recognition phenomenon reported for adults by Underwood is re-
54
HALL
producible with young children. sentation of a free-learning (FL) of a recognition list.
AND
WARE
This was accomplished by aural preword list, followed by aural presentation
Method Subjects. Ss were 86 children (38 boys and 48 girls) ranging from 5 to 7 years of age, and enrolled in a public elementary school in Evanston, Illinois. Design and procedure. The word lists and the function of each word are shown in Table 1. The procedure for each S (run individually) consisted of aural presentation of a lbword list (List la or lb) at a 4-second rate with Ss instructed to repeat each word aloud and to try to remember the words. After approximately 15 minutes, during which Ss were engaged in a nonverbal problem solving t.ask, a 24-word list (List 2) was presented aurally and Ss were instructed to identify those words that had appeared in the first list. List. 2 contained three conditions of words. Experimental words (E words) were strong natural-language associates of the CS words on the FL list, but had not themselves appeared in FL. The E words are those expected to occur frequently during FL as IARs to the CS words, resulting in frequent false recognition when they later are presented in List 2. For example, it was expected that, for many Ss when SCISSORS was presented during FL, CUT would occur as an IAR, resulting in a tendency to falsely recognize CUT when it was presented in List 2. Control words (C words) had not’ appeared in FL, nor were they associates of any FL words. Repeated words (R words) appeared in both FL and List 2. Forty-four of the Ss received List la and 42 received List lb during FL. As shown in Table 1, the non-R words in List 2 served as either E or C words depending on whether a particular S had received List la or lb. For example, for Ss who received List la, CUT served as an E word because of its association with SCISSORS, but for Ss receiving List lb, CUT served as a C word. The mean frequency with which the CS words elicit the corresponding E words for first- and second-grade children, in norms recent’ly provided by Palermo and Jenkins (1966)) is 34.8%. Results
and Discussion
Major int,erest was in false recognition rates for E and C words. Of the total of 646 false recognitions possible for each of these conditions, 101 occurred for E words and 45 for C words. That is, the false recognition rate for E words was more than twice that for C words (15.6% vs. 7.0%). The mean number of false recognitions per S was 1.17 (SD = 1.24) for E
IMPLICIT
ASSOCIATIVE
55
RESPONSES
TABLE 1 LISTS AND WORDS USED IN EXPERIMENT
I
Free learning List lb
List la SCISSORS FINGER ROUGH SPIDER BLUE MOUNTAIN NEEDLE CARS GREEN BIBLE HOUSE MAN THIRSTY EAGLE RUNNING SWEET
CS cs R cs R R cs R cs cs R R cs R R cs
BLOSSOM BED ROUGH LAMP BLUE MOUNTAIN SHEEP CARS HOUSE BATH IT MAN EATING EAGLE RUNNING FRUIT
List 2 (recognition)
cs cs R cs R R cs R R cs cs R cs R R cs
v Omitted for analysis because of formal similarity
MOUNTAIN IS WATER LIGHT GRASS ROUGH RUNNING HAND CARS CUT VEGETABLE LAMB, HOUSE MAN SLEEP FOOD WEB PIN BLUE CANDY TUB BOOK FLOWER EAGLE to LAMP
R Eb, Ea, Eb, Ea, R R Ea, R Ea, Eb, Eb, R R Eb, Eb, Ea, Ea, R Ea, Eb, Ea, Eb, R
Ca Cb Ca Cb Cb Cb Ca Ca Ca Ca Cb Cb Cb Ca Cb Ca
of List lb.
words and .52 (SD = 1.11) for C words. The mean difference of .65 was reliably greater than zero, t(85) = 5.01, p < .Ol. That this difference was not due simply to a few extreme cases is apparent from the fact that more E than C words were falsely recognized by 44 Ss while the reverse was true for only 10 Ss, a highly reliable difference (p < .Ol) as shown by the sign test. Neither age nor sex was reliably related to false recognition rate. The mean number of correct recognitions of R words per S was 6.13. Since there were eight R words per S, the recognition rate for these words was 76.6%. The higher false recognition rate for E than for C words is consistent with the hypothesis that (a) IARs based on previous associations are likely to occur when a familiar word is presented, and (b) such occurrence is likely to lead to confusion as to whether the associate had appeared
56
HALL
AND
WARE
explicitly (on the presented list) or implicitly. The absence of age differences was not interpreted to mean that IAR-produced false recognition is independent of age, since sampling from the three age levels was not random within the school, and there was some reason to suspect the presence of a bias in this respect. EXPERIMENT
II
Experiment II was designed to examine (a) the modifiability of the nature of IARs by preliminary PA training, and (b) age differences in IAR-produced false recognition.
Method Subjects. Ss were 24 5- and 6-year olds who had just completed kindergarten, and 24 8- and g-year olds who had just. completed third grade; each S was drawn randomly from the summer session of a public elementary school in Evanston. Each age group contained 14 girls and 10 boys. llesign. As in Exp. I, each S was presented a FL list followed by a recognition list. However, prior to FL all Ss were given PA training with six pairs of words, with low within- and between-pair associations. Table 2 shows t)he lists and function of each word. The design results in seven
LISTS
AND WORDS
TABLE USED
2 IN EXPERIMENT
II
Recognitionl PA training bath-joy scissors-red eating-car bible-closer sheep-hard fingers-dog
FLa bath scissors eating
spider bed it girl eagle baby
PA response
CS,
CS,
joy red car closer hard dog
BE
BC
Natural association
C
tub cut A’Er food book lamb A’C hand web sleep A’E2 is
R
light chair guns girl eagle baby
R
a Each of the CS and R words appeared three times each appeared once. b Five filler words appeared in the recognition list.
during
FL,
and
13 filler
words
IME’LICIT
ASSOCIATIVE
RESPONSES
57
conditions of words (within-&) in the recognition list, three words per condition. BE words had occurred as responses during PA training, and their corresponding PA stimuli were presented in FL. BC words also had occurred in PA training, but their corresponding PA stimuli did not appear in the FL list. A’E, words are natural-language associates of the PA stimuli that appeared in FL. A’C words are natural-language associates of the PA stimuli that did not appear in FL. A’E, words are natural-language associates of words that appeared in FL but not in the PA list. These words correspond to the E words of Exp. I. C words neither occurred earlier nor were associates of any of the words that appeared earlier. These words correspond to the C words of Experiment I. R words, as in Esp. I, actually had appeared in FL. A counterbalancing procedure ensured that each group of three naturallanguage associates served in a different recognition condition for each third of the Ss, so that Table 2 shows only one of three arrangements of words in lists. Thus, for example, for another group of Ss, BIBLE, SHEEP, and FINGERS appeared in the FL list (as CS, words), but not. in the PA list, and BOOK, LAMB, and HAND served in the A’E, condition. The mean frequency with which the nine natural associates were elicited by their respective stimulus words in the Palermo and Jenkins (1966) norms was 36.8% for first grade and 36.0% for fourth grade. Each of the six CS and three R words appeared three times in the FL list, while each of 13 filler words appeared once. Several predictions were made with respect to the various conditions described above. First, if PA training results in the establishment of the PA responses as IARs when t.heir corresponding PA stimuli are presented in FL, false recognition of these words (BE words) would be more frequent than for BC words. The BC words are necessary in order to distinguish false recognitions due to IARs in FL from recognitions due simply to interlist (PA-FL) confusions. Second, if during PA training the natural-language associates of PA stimuli undergo temporary extinction, then presentations of the PA stimuli during FL will not elicit these (the A’E, words) as 1,4Rs, reducing false recognit,ion for this condition. In that case the A’E, words would not be recognized any more frequently than would the A’C words. Third, if IAR production is occurring during the course of the expcriment, the A’E, words would be falsely recognized more frequently than would the C words. This essentially is a replication of Exp. I. Finally, it was anticipated that IAR-produced false recognition would be greater for the older than for the younger children. This prediction was based in part on Wallace’s (1967b) finding of a lower rate of IAR-
58
HALL
AND
WARE
produced false recognitions for retardates than for Ss of normal intelligence. It was based also on the assumption that. IARs play an important role in verbal mediation processes which apparently become increasingly active across the age range sampled here (Kendler, 1963). Procedure. The PA list was presented aurally with a 4- to a-second rate, using the anticipation method. Order of presentation was varied randomly from trial to trial within two blocks of eight trials each. Ss were stopped after two consecutive correct t,rials within a limit of 16 trials. After completion of the PA phase, S was given (via a tape recorder) the FL instructions: “Now we are going to play a different word game. I am going to say some words to you, then later I will see how well you remember these words. I will say each word two times, then you say the word and try to remember it.” The instructions were presented twice. Words in the FL list were presented orally at a 6-second rate. The recognition list words were presented orally at a 4-second rate with Ss instructed to indicate by saying “yes” or (‘no” whether or not the word “was one that I just read to you.” A tape recorder was used for all presentations, and Ss were run individually. Results
and Discussion
The mean number of trials during PA training was 9.33 (SD = 1.58) for third grade and 11.42 (SD = 1.42) for kindergarten children. Three of the third-grade and six of the kindergarten children failed to meet the criterion of two consecutive correct trials within the 16-trial limit. In other words, a slightly higher level of learning may have been achieved by the older children, despite the nonsignificance of the difference between mean number of PA trails, t (23) = 1.53, p > .lO. In view of the results described next, it is difficult to see any problem in interpretation created by this difference, if in fact it existed. TABLE PERCENT
RECOGNITIONS AND PER SUBJECT
3
MEAN NUMBER IN EXPERIMENT
OF RECOGNITIONS
II Condition
-
Grade
Kindergarten ?JO Recognition Mean no. Recognition/S Third grade y0 Recognition Mean no. Recognition/S
BE
BC
A’E,
.4’C
A’Ez
20.8 0.62
20.8 0.62
9.7 0.29
5.6 0.17
23.6 0.71
12.5 0.38
12.5 0.38
8.3 0.25
4.2 0.12
6.9 0.21
C
R
0 0
87.5 2.62
4.2 0.12
94.4 2.83
IMPLICIT
ASSOCIATIVE
59
RESPONSES
Table 3 shows the mean number of recognitions per S for each condition at each age level. (Except in the case of R words, these are false recognitions.) Using conditions BE, A’E,, BC, and A’C, an analysis of variance was performed to examine the main effects of and interactions among (a) appearance of the PA stimulus during FL, (b) type of associate, and (c) grade level. The only reliable F was that for type of associate, F(1,46) = 9.45, p < .Ol, with more false recognition of experimental (BE and BC) than of natural (A’E, and A’C) associates. Apparently a number of Ss were unable to distinguish between a word having occurred during FL and its occurrence earlier during PA training. It is clear that false recognitions due to the occurrence of IARs (either natural or experimental) during FL did not occur at either age level. These results suggest the possibility that PA training served to extinguish temporarily the usual IAR, but was not sufficient to establish the PA responses as IARs during FL. Analysis of variance was employed to compare the standard experimental and control conditions (A’E, vs. C) at the two age levels. The main effecb of condition was reliable, F(1,46) = 15.16, p < .OOl, confirming the results of Exp. I. The interaction between condition and age also was reliable, F (1,46) = 9.45, p < .Ol, indicating that IAR-produced false recognit.ions were more frequent for younger than for older children. In fact, it appears that IAR-produced false recognitions did not occur among the older children. This age difference is of particular interest since precisely the opposite results were anticipated. One interpretation of these unexpected results is that IAR production decreases rather than increases over this age range. However, an alternative hypothesis that seems more plausible is that the older children, though producing as many or more IARs, are better able to discriminate between the previous appearance of a word on the presented list vs. its earlier occurrence as an IAR only. One implication of this hypothesis is that interlist (PA-FL) confusion would be greater for the younger Ss. An examination of false recognitions of BC compared with C words seems to confirm this prediction, in that the interaction between word condition and age was significant, F(1,46) = 18.85, p < .OOl. However, this difference was produced mainly by a few Xs who falsely recognized either two or all three of the BC words. Thus, it seems fairly probable that the age difference was in misinterpretations of the recognition instructions, rat,her than in interlist confusions. REFERENCES BOUSFIELD,
in verbal
W.
A., WHITMARSH, generalization.
G. 9., Psychological
AND
DANICK,
Report,
J. J. Partial 1958,
4, 703-713.
response
identities
60
HALL
DAVIS, G. 9. Recognition
AXD
WARE
memory for visually presented homophones. Psychological 1967, 20, 227-233. KENDLER, TR.~CY S. Development of mediating responses in children. Monogrnplis of the Society of Research in Child Developmcuf. 1963. 28, (2), 33-52. PALERMO, D. S., AND JENKINS, J. J. Oral word association norms for citildren iir grades one through four. Pennsylvania State Gniversity Research Bulletin So. 60, 1966. UNDERWOOD, B. J. False recognition produced by implicit verbal responses. Jowlal of Experimental Psychology, 1965, ‘70, 122-129. WALLACE, W. P. False recognition produced by laboratory-established associative responses. Psychonomic Science, 1967a, 7, 139-140. W.~LL.~CE, W. P. Implicit associative response occurrence in learning u-ith retarded subjects: A supplementary report. Journal of Educational Psychology, 1967b, 58, 110-114. Report.