Category typicality effects in children's memory performance: Qualitative and quantitative differences in the processing of category information

Category typicality effects in children's memory performance: Qualitative and quantitative differences in the processing of category information

JOURNAL OF EXPERIMENTAL CHILD 35, 329-344 (1983) PSYCHOLOGY Category Typicality Effects in Children’s Memory Performance: Qualitative and Quantit...

1MB Sizes 1 Downloads 61 Views

JOURNAL

OF EXPERIMENTAL

CHILD

35, 329-344 (1983)

PSYCHOLOGY

Category Typicality Effects in Children’s Memory Performance: Qualitative and Quantitative Differences in the Processing of Category Information DAVID F. BJORKLUNDAND BARBARAE. THOMPSON Florida

Atlantic

University

Kindergarten, third, and sixth graders (6, 9, and 12 years of age, respectively) received a cue-at-input/cue-at-output recall task, using category typical and atypical items that were based on either (1) children’s conceptions of item typicality, or (2) adults’ conceptions of item typicality. At each grade level, recall was greater with the child-defined lists than with the adult-defined lists, and typical items were recalled to a greater extent than atypical items. Further analyses revealed that the recall of typical items varied as a function of children’s typicality ratings of items, and that the “typicality effect” in the adult-norm condition was due primarily to the children not realizing that many of the atypical items were appropriate category exemplars. In contrast, typicality effects in the child-norm condition were attributed to qualitative differences in the judged “goodness of example” of the typical and atypical items. The results were discussed in terms of the appropriateness of typicality as a dimension of children’s natural language concepts, the role of age differences in knowledge base in affecting performance on a cognitive task, and of the importance of using child-generated norms in studies of children’s processing of category information.

Previous research has indicated that adults judge some members of natural language concepts to be more “typical” or representative of their categories than others (e.g., Rosch, 1973, 1975; Uyeda & Mandler, 1980), and that differences in word typicality influence performance on a variety of cognitive tasks, including those involving category verification (e.g., Rosch, 1973; Rips, Shoben, & Smith, 1973), same-different reaction The research is based on a thesis by the second author in partial fulfillment of the Master of Arts degree at Florida Atlantic University. This research was supported by a grant to the first author from The Spencer Foundation. We thank the staff and students of Henderson University School, Bocd Raton Middle School. Coral Springs Elementary School. James S. Hunt Elementary School, and Pompano Beach Middle School for their cooperation in the execution of this study. Portions of this research were presented at the meeting of the Society for Research in Child Development. Boston. 1981. Requests for reprints should be sent to David F. Bjorklund. Department of Psychology, Florida Atlantic University, Boca Raton. FL 3343 I. 329 0022-0965183 $3.00 Copyright Q 1983 hy Academic Press. Inc. All right\ of reproduction in any form reserved.

330

BJORKLUND

AND

THOMPSON

times (e.g., Rosch, 1975), and free recall (e.g., Greenberg & Bjorklund, 1981; Keller & Kellas, 1978). A number of investigators have reported age differences in the number of items children identify as category members, with the majority of age differences being attributed to young children’s tendencies to deny that familiar but atypical (or peripheral) items are appropriate category instances (e.g., Anglin, 1977; Saltz, Soller, & Sigel, 1972; Bjorklund, Thompson, & Ornstein, Note 1). Furthermore, in a recent investigation where children were asked to select from lists of words examples of familiar categories and rate them in terms of typicality (Bjorklund et al., Note l), it was found that although the typicality judgments of children were significantly correlated with those of adults, there were also many discrepancies. That is, not all typical (or atypical) category members as judged by adults were similarly judged by children. Other researchers have evaluated children’s processing of typical and atypical category exemplars, and report that, at least for children 9 years of age and older, processing differences between typical and atypical items are similar to those found for adults (Duncan & Kellas, 1978; Rosch, 1973). However, in their developmental priming study, Duncan and Kellas (1978) suggest that variations in semantic prototypes between young children and older subjects seem likely. It is important to note that in the experiments cited above, adult judgments of what constitutes typical and atypical category exemplars have been used. While such a procedure provides valuable information concerning how children process adult-defined category exemplars, a different set of interpretations may arise if subjects were given cognitive tasks with the typicality of the items being defined in terms of age-mates’ judgments rather than in terms of those of adults. In addition to the above, using child-defined rather than adult-defined category exemplars would increase the likelihood that children would be familiar with the categorical nature of the materials. A number of researchers have recently stressed the importance that a child’s knowledge base may play in the execution of a cognitive task. More precisely, what a child knows will affect how that information is processed (see Chi, 1978; Ornstein & Naus, in press). With respect to category typicality, the use of adult-defined norms may not reveal how children process typical versus atypical category exemplars, per se, because many of the adult-defined items may not be so perceived by the children. In such cases, age differences in information processing are confounded with age differences in knowledge base. To the extent that children’s concepts differ from those of adults, the use of child-defined typicality norms may provide a different picture of children’s concept utilization and one that is more accurate with respect to processing differences as a function of differences in judged typicality.

TYPICALITY

EFFECTS

IN RECALL

331

The present experiment used a cued recall task with kindergarten, third, and sixth grade children to evaluate processing differences when child-generated versus adult-generated typicality norms were used as stimuli. Children were given category exemplars paired at input with their category labels, with these labels subsequently being used as retrieval cues. Cued recall was deemed appropriate for assessing category utilization in children, because previous research has found that the perceived relationship between the target and cue words affect the probability of recall. For example, words that are strong associates of target items are more effective retrieval cues for adults than are weak associates (e.g., Thompson & Tulving, 1970), and category labels serve as better mediators of recall than do function words for third and sixth grade children (Bjorklund, Greenberg, Hurlbert, & Thompson, Note 2). In general, the effectiveness of a cue is a function of a subject’s perceived relationship between the target and cue words, with “the specific semantic relationships between the input cue and the target . . . [being] (Salzberg, 1976), . . . an important aspect of the encoding representation” and thus of the efficacy of the retrieval cue. In the present experiment, it was expected that levels of cued recall would be greater for the typical than for the atypical category exemplars when both child-generated and adult-generated typicality norms were used. However, differences in the pattern of recall of typical and atypical items were expected to vary as a function of which set of norms were used, and possibly as a function of grade level. METHOD

Subjects One hundred ninety-two children, 64 each from the kindergarten (40 boys and 24 girls), third (29 boys and 35 girls), and sixth (32 boys and 32 girls) grades, served as subjects. The children were selected from four public schools in Palm Beach and Broward Counties, Florida, and were predominantly from middle class homes. Mean ages were 5 years 11 months (range, 5 years 3 months to 6 years 5 months), 8 years 10 months (range, 8 years 0 months to 9 years 9 months), and 11 years 9 months (range, 10 years 11 months to 13 years 4 months) for the kindergarten, third, and sixth grade children, respectively. At each grade level, 32 children were randomly assigned to each of two experimental conditions, balancing as closely as possible for sex and school. Normative

Study

The typicality of all items at each grade level was based on norms collected by Bjorklund et al. (Note 1). In that study, children in the kindergarten, third, and sixth grades, and college students were read lists of words from 12 natural language categories and asked to select from

332

BJORKLUND

AND THOMPSON

each list all members of the designated category (several noncategory exemplars were included as foil items in each list). Following item selection for each category, subjects were asked to rate each item in terms of how typical it was of its category. Children were told to pretend that there was a person from outer space who knew nothing about Earth, and that they could help the spaceman understand our world by answering some questions for the experimenter. They were to rate items in terms of how good an example each one was of its category on a 3point scale so that the spaceman could better understand the meaning of these categories. To simplify the rating task, the children’s judgments were obtained by having them point to one of three black and white schematic faces placed on the table in front of them. Each face displayed a different expression. The “smiling” face corresponded to “very good” category exemplars, ones that would be very helpful to the spaceman in understanding the category; the “straight” face corresponded to “okay” category exemplars, ones that would help the spaceman somewhat to understand the category, but not as well as the “very good” items; and the “frowning” face corresponded to “poor” category exemplars, items that were category members, but ones that would not help the spaceman understand the category very well. Children were told not to rate items in terms of how much they liked a particular item, but only in terms of how good an example it was of its category. The adults were tested as a group and participated in a rating procedure similar to that used by Rosch (1975). Each item received a score from each subject ranging from 0 (not selected as a category exemplar) to 3 (a highly typical or “very good” category instance).’ Task and Materials

All children were given one cued recall trial with typical and atypical category items that were paired with their category labels at input, with those labels serving as retrieval cues at output. For children in the childnorm condition, the typicality of the to-be-remembered items was based ’ The use of a 3-point rating scale for items included as category members was necessary so that the children would be able to perform the task adequately. It is possible that the use of this scale may have resulted in a loss of discriminability for the adults, relative to when a 7-point scale is used, as in Rosch (1975) and Uyeda and Mandler (1980). However, Spearman rank-order correlations computed between the typicality ratings of the adult subjects in the Bjorklund et al. study and those of Rosch (1975) were generally significant (for seven of eight categories, p < .Ol), with a median correlation of .94. Rank-order correlations were slightly lower when the Bjorklund et al. adult ratings were compared with those of Uyeda and Mandler (1980), but were significant (p < .05) for 8 of IO categories, with a median value of .82. Thus, although the 3-point scale may have resulted in some loss of discriminabihty for the adults, the correlational analyses indicate considerable agreement between the Bjorklund et al. and the Rosch (1975) and Uyeda and Mandler (1980) studies, despite differences in the rating scales.

TYPICALITY

EFFECTS

IN RECALL

333

upon judgments of same-age peers who had participated in the Bjorklund et al. (Note 1) investigation. For children in the adult-norm condition, typicality of the items in the lists for recall was based upon the adult norms collected by Bjorklund et al. Two word lists (A and B) were generated separately at each grade level for the child-norm condition. Further, two lists of words (C and D) were constructed for use at all grade levels for the adult-norm condition. Examples of these word lists are presented in Table 1. Some of the items in the child-norm lists at each age were also present in the adult-norm lists (i.e., some items overlapped in the two conditions). Although this dimension was not intentionally manipulated in the present experiment, the percent of items found in both conditions, collapsing across the two lists at each age, was 41% typical and 27% atypical for the kindergarten children, 36% typical and 55% atypical for the third graders, and 59% typical and 59% atypical for the sixth graders. The two lists (A and B, or C and D) were used with equal frequency for subjects in their respective conditions. Each list of words comprised 22 items. One-half of the items in each list were typical exemplars cued by their superordinate category labels, and the remaining items were atypical exemplars cued by their appropriate category labels. Within each list, two typical or atypical items were paired with the category labels Birds, Body parts, Clothing, Fruit, Furniture, Musical instruments, Vehicles (i.e., Things that take you places), Tools, Vegetables, and Weapons. Due to limitations imposed by an inclusion criterion (discussed below), only one item per list was paired with the category labels Buildings and Dogs. Those category labels paired with typical items in Lists A and C were paired with atypical items in Lists B and D, and vice versa. For example, in the child-norm condition at the kindergarten level, List A contained the category label Clothing paired with the typical exemplars dress and pants, whereas the atypical items socks and tie were paired with this category label in List B. Thus, both typical and atypical exemplars from a category were tested, although each subject received only typical (or only atypical), items from a single category. In order for an exemplar to be included as a list item at a particular grade level, at least 75% of same-age subjects in the Bjorklund et al. study had to include that item as an appropriate category exemplar. It should be noted, however, that with this stipulation, items used in the adult-norm condition need not be included by at least 75% of the children in the Bjorklund et al. study. Although this dimension was not intentionally manipulated, the percent of adult-defined atypical items in the adult-norm lists not included by at least 75% of same-age peers as category members was 55% (12 of 22), 27% (6 of 22), and 23% (5 of 22) for the kindergarten, third, and sixth graders, respectively.

334

BJORKLUND

AND

TABLE SAMPLE

Pair

Child-norm Typical

LISTS

Category condition-kindergarten items

items

German Spear

Weapon Tool Tool

Sword Screwdriver

Vehicle Vehicle

Train

Wrench Airplane Fingers Elbow

part part

Clothing Clothing

Pants Dress

Building Vegetable

School Potato

Vegetable Furniture Furniture

Squash Desk Bed

Musical Musical

instrument instrument

Bird Bird

Typical

condition-used items

at all

grade

Lemon levels

Dog Bird Bird

Collie Bluejay Parrot

Tool Tool Musical

Screwdriver Wrench Flute

Musical Body Body

Atypical

items

Accordion Tamborine Eagle Owl Cherries

Fruit Fruit Adult-norm

Exemplar

Dog Weapon

Body Body

Atypical

I

WORD

type

THOMPSON

instrument instrument part part

Guitar Fingers Foot

Weapon Weapon

Rifle Sword

Building

Igloo

Furniture Furniture Vehicle Vehicle Clothing Clothing

Lamp Television Bicycle Tricycle Hat Scarf

Fruit Fruit Vegetable

Lemon Lime Mushroom Onion

Vegetable

shepherd

TYPICALITY

EFFECTS

IN RECALL

335

Four random presentation orders were generated for each list, with the stipulation that not more than two items of either classification type (typical or atypical), or two items from the same category be presented consecutively. Four random cuing orders were also generated, with the additional stipulation that the last list category presented at input not be the first list category cued at output. In order to minimize confounding of results due to other relevant semantic dimensions, the typical and atypical items selected for use within each list and across lists between the child-norm and adult-norm conditions at each grade level were matched, as closely as possible. in terms of word frequency (Thorndike & Lorge, 1944: Kucera & Francis, 1967) and category frequency (Battig & Montague, 1969: Posnansky. 1978). In addition, the mean typicality ratings for typical items in both child-norm lists at each grade level were matched as closely as possible, and likewise for the atypical items. Mean typicality ratings, based on judgments made by subjects in the Bjorklund et al. (Note 1) normative study, are presented for the typical and atypical items for each list by condition and grade in Table 2. Note that although the lists were matched in typicality ratings within each condition, children’s typicality ratings were generally higher for the child-norm than for the adult-norm lists. However, the adults’ ratings were actually somewhat higher for both the typical and atypical items than those of the children’s’ Procedure

Each child was tested individually by a female experimenter. The children were told that they were going to be presented with several short lists of words, one list at a time, which they would be required to remember. Following this explanation, the children were presented with an g-item list of words for free recall, and were subsequently tested in traditional free recall fashion. Upon completion of this practice list, the children were told that for the remaining word lists, each to-beremembered item would be paired with a “hint” word that would help them to remember the target. In addition, the children were told that ’ Due to an effort to match lists on a variety of factors, the items chosen for inclusion in this study were not necessarily the highest rated (or for the atypical items, the lowest rated) items at a grade level from the Bjorkhmd et al. study. Rather, for any given category and at each grade level, typical items were chosen so that they had higher ratings than atypical items for that same category, although, whenever possible, we attempted to maximize rating differences between the typical and atypical items. Thus, it would be more appropriate (although more awkward) to refer to the items used in this experiment as “relatively typical” and “relatively atypical” rather than as typical and atypical, per se. Accordingly, contrast between typical and atypical items, and between typical (or atypical) child-norm and adult-norm items, should be viewed as reflecting differences between items that vary on a continuum of “typicality,” rather than as reflecting differences between items that represent two ends of a dichotomy.

336

BJORKLUND

MEAN

TYPICALITY

AND THOMPSON TABLE 2 3.00) FOR ITEMS IN THE CHILD-NORM

RATINGS (MAX. = ADULT-NORM

AND

LISTS BY GRADE

Child norm -___ List A List B

Adult norm List C

List D-

Kindergarten TYP ATYP

2.37 (.24) 1.78 (.20)

2.35 C.16) 1.73 (.24)

2.09 (.31) 1.53 (.40)

2.25 (.22) 1.53 (.38)

Third TYP ATYP

2.63 (.12) 1.87 (.25)

2.67 (.13) 1.81 (.27)

2.47 t.19) 1.63 (.29)

2.45 C.19) 1.76 (Sl)

Sixth TYP ATYP

2.89 (.08) 2.08 (.38)

2.83 (.07) 2.02 (.45)

2.84 (.09) 1.69 (.33)

2.73 t.16) 2.09 (.45)

2.96 (.03) 2.41 (.24)

2.94 (.05) 2.33 (.46)

Adults TYP ATYP

Note. Norms from Bjorklund et al. (Note 1). Each mean based on N = 11. Standard deviations in parentheses; TYP, typical; ATYP, atypical.

these “hint” words would later be presented alone, and that their job would be to tell the experimenter what words had been previously presented with the hint words (i.e., cued recall). Following these instructions, children received a second practice list, and later, one test list appropriate to their assigned condition. Word pairs were read at a rate of one pair every 5 set, and the children were asked to repeat both items after the experimenter. At output, the cues were also presented at a rate of one cue approximately every 5 sec. Upon completion of the test list, each child was thanked for his/her participation. RESULTS

A preliminary analysis, assessing possible sex differences in the number of words recalled, indicated that performance did vary with sex, F( 1, 180) = 8.91, p < .005, with boys recalling significantly more items (M = 6.97) than girls (M = 6.00). Because sex did not interact significantly with any other factor, however, all subsequent analyses were collapsed over sex. The mean number of words recalled by condition and typicality at each grade level are presented in Table 3. These data were evaluated in a grade (3) x condition (2) x typicality (2) analysis of variance, with repeated measures on the last factor. The analysis of these data indicated that recall was greater in the child-norm (M = 6.69) than in the adultnorm (M = 6.15) condition, F(1, 186) = 4.16, p < .OOl, and that typical

TYPICALITY

EFFECTS TABLE

337

IN RECALL

3

MEAN NUMBER OF WORDS RECALLED BY CONDITION, TYPICALITY AND GRADE (MAXIMUM RECALL = 11)

Typical

Atypical

Means

Kindergarten Child norm Adult norm Means

5.63 4.59 5.11

4.75 4.06 4.41

5.19 4.33 4.76

Third Child norm Adult norm Means

7.38 6.59 6.98

5.88 5.66 5.77

6.63 6.13 6.38

Sixth Child norm Adult norm Means

8.81 8.09 8.45

7.72 7.88 7.80

8.27 7.98 8.13

Grade

items were recalled to a greater extent (M = 6.85) than atypical items (it4 = 5.99), F(1, 186) = 31.65, p < .OOl. This typicality effect implies that typical items from a category are recalled to a greater extent than atypical items from that same category. However, it is possible that such an effect could reflect large differences between typical and atypical items from only a very few categories, rather than typicality effects within all categories. This, however, was not the case for the child-norm condition, in which recall was greater with the typical than with the atypical exemplars for 10 of the 12 categories (the exceptions, FRUIT and WEAPONS), p = .02 (binomial test). This effect was not significant for the adult-norm lists (typical > atypical for 8 of 12 categories, p = .19), indicating that the typicality effect found in the adult-norm condition was not uniform across categories. In addition to the significant main effects, the condition x typicality interaction approached significance, F(1, 186) = 3.78, p = .053. Further analysis of this interaction indicated that recall of typical items was significantly greater in the child-norm (M = 7.27) than in the adult-norm (M = 6.43) condition, F(1, 372) = 7.48, p < .Ol. However, the recall of atypical items was comparable between the two conditions (means = 6.11 and 5.86 for the child-norm and adult-norm conditions, respectively), F(1, 372) < 1. Although the difference in recall between the child-norm and the adult-norm conditions varied somewhat with age (mean child-norm minus adult-norm difference = .86, .50, and .29 for the first, third, and sixth grades, respectively), the grade x condition interaction was not significant, F(2, 186) < 1.

338

BJORKLUND

AND

THOMPSON

Differences in recall between the child-norm and adult-norm conditions were further investigated in a series of analyses that involved examinations of the nature of the typical and atypical items in the child-norm and adult-norm lists. In all of these analyses, “percentage recall scores” were derived for each subject in the relevant conditions. This measure will be explained in the context of the first of these analyses. Recall of Category

Typical Exemplars

The first analysis involved an examination, at each grade level, of recall for typical items that appeared in both the child-norm and adultnorm lists (i.e., “overlap” items) versus recall for typical items which did not overlap between the lists (i.e., “nonoverlap” items, or items found exclusively in child-norm or adult-norm lists). As was previously noted, 41% (9 of 22) of the typical items in the child-norm lists at the kindergarten level were also included as typical items in the adult-norm lists. The corresponding “overlap” figures were 36% (8 of 22) and 59% (13 of 22) for the third and sixth graders, respectively. It was hypothesized that since items included in the adult-norm lists only were judged to be typical category exemplars by adult standards and did not necessarily represent typical category members for the children, levels of recall for these items would be lower than levels of recall for items included as typical exemplars in both the child-norm and adult-norm lists (i.e., the “overlap” items). In contrast, typical items found in the childnorm lists only were defined as typical exemplars by age mates of the subjects tested here. Accordingly, levels of recall for these items should be at least as great as recall for the “overlap” items, and possibly greater. In order to assess these predictions, the following analysis was performed. For each subject in both the child-norm and adult-norm conditions, two percentage recall scores were calculated-one for typical “overlap” items (i.e., typical items found in both conditions at the appropriate grade level), and one for typical “nonoverlap” items (i.e., typical items found only in the condition in which the child participated). As was noted above, the percentage of typical items that “overlapped” with the adult-norm condition varied with grade level. Thus, although there were differences in the proportion of “overlap” versus “nonoverlap” items among the three grades, equal weight was given to the “overitems for subjects in the analysis. lap” and “nonoverlap” The mean percent recall of overlap and nonoverlap typical items by condition is presented in Table 4. These data were analyzed in a grade (3) x condition (2) x overlap (2) analysis of variance, with repeated measures on the last factor. This analysis produced a significant condition x overlap interaction, F(1. 186) = 10.23. p < .Ol. Subsequent tests revealed that for subjects in the child-norm condition, typical nonoverlap items (i.e., items unique to the child norm lists) were recalled significantly

TYPICALITY

EFFECTS TABLE

IN RECALL

339

4

MEAN PERCENT RECALL OF OVERLAP AND NONOVERLAP ITEMS BY CONDITION NOll-

Condition Child-norm Adult-norm

Overlap”

overlap”

Mean

61

71

66

62

56

59

n Typical items found in both the child-norm and adult-norm lists. ’ Typical items found exclusively in either childnorm or adult-norm lists.

more than typical overlap items (i.e., items found in boll) zht &-lilct-- and adult-norm lists), t(186) = 2.85, p < .Ol. This effect was reversed in the adult-norm condition, with the overlap items being recalled significantly more than the nonoverlap items, r(186) = 1.71, p < .05 (one-tailed test). Furthermore, recall of the nonoverlap items in the child-norm lists was significantly greater than recall of the nonoverlap items in the ;Idult-norm lists, t(372) = 3.94, p < .OOl. This result indicates that recall of typical items in the child-norm condition exceeded recall of typical items in the adult-norm condition because of the items that differed between the two conditions (i.e., the nonoverlap items). One factor which may have contributed to the above effect is differences in judged typicality between the nonoverlap child-norm and adultnorm typical items. As can be seen in Table 2. children’s ratings of typical items in the child-norm condition exceeded their ratings ot’ typical items in the adult-norm condition. A closer inspection of itew ratings revealed that this difference was due primarily to differences in ratings of the nonoverlap items. Overall, the child-norm. nunoverlap typical items were rated higher (i.e., more typical) by children 10 11~ qj:,rE.Iund et al. (Note I) normative study (mean typicalii:i, rating = 3.65 1, than were the adult-norm, nonoverlap items (M = 2.37). with items found in both lists (i.e., the overlap items) having an intermediate mean value (A4 = 2.58).

Recall of Category Atypical Exemplars

The next series of analyses was based upon an examinatkon of the nature of the atypical items found in the adult-norm lists. and involved deriving percentage recall scores only for subjects in the adult-norm condition. As was previously stated, recall ul’ tit~pi~al ;~~:oi> MW corn. parable in the child-norm and adult-norm conditions. in order to evaluate further this finding, it was hypothesized that children would show greater recall for adult-defined atypical category exemplars which they consid-

340

BJORKLUNDANDTHOMPSON

ered to be category members (defined here as those items included by at least 75% of the children in the Bjorklund et al. study) than for atypical items which they did not consider to be members of their designated categories (defined here as those items which were included by less than 75% of the children in the aforementioned study). To reiterate, of the 22 atypical items in the adult-norm lists, 12 (55%), 6 (27%), and 5 (23%) atypical items were not included by at least 75% of the kindergarten, third, and sixth grade children, respectively, in the Bjorklund et al. study. Percent recall scores were calculated for each subject in the adult-norm condition for those atypical items that were included by 75% or more of the children in the normative study (i.e., “included” items) and for those included by less than 75% of the children (i.e., “not included” items). Thus, although there were differences in the proportions of atypical items classified as “included” versus “not included” among the three grade levels, equal weight was given to the “included” and “not included” items in the following analysis. These data were analyzed in a grade (3) x inclusion (2) analysis of variance, with repeated measures on the last factor. This analysis produced a significant effect of inclusion, F(1, 93) = 14.33, p < .OOl, with “included” atypical items being recalled significantly more (58%) than “not included” atypical items (45%). The grade x inclusion interaction was not significant, F(2, 93) < 1. These results suggest that the “typicality effects” found for children in the adult-norm condition might be due not so much to differences in perceived typicality between the typical and atypical exemplars, but to the fact that many of the atypical items were not judged by same-age peers to be category members. In fact, large typicality effects were found for subjects in the adult-norm condition when contrasting the recall of typical exemplars (59%) with the recall of the “not included” atypical items (45%), F(1, 93) = 29.39, p < .OOl, similar to that found in the child-norm condition. However, when recall of the typical items was contrasted with recall of only the “included” atypical items (58%), a significant typicality effect for children in the adult-norm condition was effects with respect to the eliminated, F(1, 93) < 1. Thus, “typicality” adult-norm lists seem to be due to the presence of what, for many children, were noncategory exemplars. DISCUSSION

In the present experiment, differences in recall as a function of judged typicality were found, at all ages tested, for both child- and adult-defined sets of category exemplars. In other words, in both the child-norm and adult-norm conditions, typical category exemplars were recalled better than atypical category instances. This result is consistent with the theorizing of Rosch (1973, 1975) and others suggesting that items which

TYPICALITY

EFFECTS

IN RECALL

341

are more similar to their category prototype (here represented by the category labels or “cues”) fit better with a subject’s semantic memory organization, and therefore, result in higher levels of performance than less prototypical or “atypical” items. The presence of typicality effects in the child-norm condition is a finding of central importance in the current research, as this result is the first indication of processing differences between items judged by children to be typical and atypical of their categories (i.e., with respect to the Bjorklund et al. norms). In Rosch’s (1973, 1975) terms, the childgenerated norms collected by Bjorklund et al. (Note 1) seem to have “psychological reality.” In other words, since performance differences between child-defined typical and atypical category exemplars have been found, this suggests that children’s conceptions of how representative items are of their categories are cognitively meaningful. In addition to the observed typicality effects, differences in overall recall performance between the child-norm and adult-norm conditions were found. Recall in the child-norm condition, in which child-defined typical and atypical items were used, exceeded recall in the adult-norm condition, in which adult-defined exemplars served as stimuli. It should be noted, however, that because of selection restrictions concerning which items were chosen for inclusion in the experiment (see Footnote 2), differences between the child- and adult-norm conditions for the typical items should not necessarily be viewed as representing differences between child and adult prototypes. The typical items used in each list represent “relatively typical” items for the age group which generated them, and not necessarily the category prototypes, per se. Nevertheless, the finding of differences between the child-norm and adult-norm conditions is consistent with the results of others (e.g., Bjorklund & Zeman, 1982; Chi, 1978; Lindberg, 1980) which suggests that the degree of knowledge children possess with respect to the to-be-remembered items can have important consequences for performance on memory tasks. To elaborate, children often demonstrate enhanced levels of recall performance when memory is assessed in tasks using materials that are meaningful and well-known to them, in comparison to when more traditional materials (i.e., items which are more familiar to adults) are used. The present finding suggests that this perspective is also important for developmental research in concept development and concept utilization. In other words, it can be argued that the use of child-generated materials in investigations of concept utilization may be more relevant, and may better assess the nature of children’s concepts, than when adult-defined materials are used. As will be further discussed below, the use of childdefined materials in cognitive tasks may not only result in overall enhanced levels of performance, but may also reveal differences in the qualitative nature of performance on these tasks.

342

ICl~IKKl~U\;L) AND THOMPSON

With respect to the adult-norm lists, the overall low levels of recall for adult-defined t~picczl items appears to be due to the presence of several “typical” items in these lists which were judged to be highly representative category instances by adults but not necessarily by children. That is, the “nonoverlap” typical items in the adult-norm lists were recalled to a lesser extent than were typical items found in both the child-norm and adult-norm lists (i.e., the “overlap” items). However, the reverse ~‘ar: true I;N the child-norm lists. In other words, the typical item< tiw~t,i ili ::-IC i ./r!‘/ti ,11+)7717 li5r.t on/y were recalled to a greater extent than typical items found in both the child-norm and adult-norm lists. This implies that the “nonoverlap” typical items found in the child-norm lists tended to enhance levels of recall for the typical items in this condition, whereas, the “nonoverlap” typical items found in the adult-norm lists tended lo lower levels of recall for the typical items. This latter result suggests that when the knowledge possessed by children is different from that possessed by the adults who generated the sets of items, children will show decrements in performance. Stmilar r‘i>ult!, ww found for the atypical items in the adult-norm lists: [hat i\% the extent of knowledge possessed by children with respect to these items d/so influenced overall levels of recall in the adult-norm condition. This was supported by the finding that “included” atypical items ii.c.. items included as category members by at least 75% of the children in the Bjorklund et al. normative study) were recalled to a greater extent than the remaining “not included” atypical instances found in the adult-norm lists (i.e., these items may be assumed not to be part of the category knowledge possessed by many of the children in this the fact that “typicality effects” were obexperiment]. Furthermore. served at all in the adult-norm condition is due largely to the presence of the “not included” atypical items, as demonstrated by the result that when these items were removed from the analysis, the recall differences between the typical and atypical category exemplars in the adult-norm lists were eliminated. Furthermore, whereas typical items were recalled to a greater extent than atypical items for all but two categories with the child-norm lists (p = .02), significant within category typicality effects were not found with the adult-norm lists (p = .19). This finding suggests that typic:ilit!. ~l‘t’cct\ in the adult-norm condition were attributable to large differcnc:es M ithin a few categories, and that children’s processing differences between adult-judged typical and atypical items were not generalizable across categories. In conclusion, differences in the nature of the “typicality effects” were found between the child-norm and adult-norm conditions. When childdefined tnaterials were used in the present experiment, processing differences between the typical and atypical exemplars could be attributed to qualitative differences in children’s conceptions of how representative

TYPICALITY

EFFECTS

IN RECALL

343

items are of their categories. However, when adult-defined materials were used, the converse was true. Processing differences between the typical and atypical items in the adult-norm lists were apparently due more to a quantitative lack of category knowledge (i.e., not realizing that many atypical items were appropriate category members), rather than to a difference in the semantic nature of the items, per se. This suggests the need for a more cautious evaluation of the results obtained by previous investigators (e.g., Duncan & Kellas, 1978; Rosch, 1973) in developmental studies of typicality effects when adult-defined typical and atypical category exemplars were used as stimuli. Perhaps their “typicality effects” were not so much the consequence of item “goodper se, but were more a reflection of quantitative differences in ness,” the knowledge bases of adults and children. REFERENCES Anglin, J. M. Word, object, and conceptual development. New York: Norton, 1977. Battig, W.. & Montague, W. Category norms for verbal items in 56 categories. Journal of Experimental Psychology Monographs, 1969, 80, l-45. B.jorklund. D. F., & Zeman, B. R. Children’s organization and metamemory awareness in their recall of familiar information. Child De\~e/opment. 1982. 53, 799-810. Chi, M. T. H. Knowledge structure and memory development. In R. Siegler (Ed.). Children’s thinking: What deve/ops:‘Thirteenth Annual Carnegie Symposium on Cognition. Hillsdale. N.J.: Erlbaum. 1978. 73-96. Duncan, E. M., & Kellas, G. Developmental changes in the internal structure of categories. Journal of Experimental Child Psychology, 1978, 26, 328-340. Greenberg, M. S., & Bjorklund, D. F. Category typicality in free recall: Effects of feature overlap or differential category encoding? Journal of Experimental Psychology: Human Learning and Memory, 1981, I, 145-147. Keller, D., & Kellas, G. Typicality as a dimension of encoding. Journal of Experimental Psychology:

Human

Learning

and Memory,

1978, 4, 78-85.

Kueera. H., & Francis. W. N. Computational trntr/.vsis c$present-dtry American English. Providence, R.1.: Brown Univ. Press, 1967. Lindberg, M. The role of knowledge structures in the ontogeny of learning. Journal of Experimental Child Psychology, 1980, 30, 401-410. Omstein, P. A., & Naus, M. Effects of the knowledge base on children’s memory processes. In J. B. Sidowski (Ed.), Contemporary issues in experimental psychology. Hillsdale, N.J.: Erlbaum, in press. Posnansky. C. J. Category norms for verbal items in 25 categories for children in grades 2-6. Behavior Research Methods & It~strrrmerttcrtion. 1978. IO, 8 19-832. Rips, L. J., Shoben, E. J., & Smith, E. E. Semantic distance and the verification of semantic relations. Journal of Verbal Learning and Verbal Behavior, 1973, 12, l-20. Rosch, E. On the internal structure of perceptual and semantic categories. In T. E. Moore (Ed.), Cognitive development and the acquisition of language. New York: Academic Press, 1973. Rosch, E. Cognitive representations of semantic categories. Journal of Experimental Psychology: General, 1975, 7, 192-233. Saltz, E., Soiler, E., & Sigel. I. E. The development of natural language concepts. Child Development, 1972, 43, 1191-1202. Salzberg, P. M. On the generality of encoding specificity. Journal of Experimental Psychology: Human Learning and Memory. 1976, 2. 586-596.

344

BJORKLLJND

Thompson, cues.

AND THOMPSON

D. M., & Tulving, E. Associative encoding and retrieval: Weak and strong Journal

of Experimental

Psychology,

1970,

86,

255-262.

Thorndike, E. L.. & Lorge, 1. The teacher’s word book of 30,000 words. New York: Teachers College, Columbia University, Bureau of Publications, 1944. Uyeda, K. M., & Mandler, G. Prototypicality norms for 28 semantic categories. Behavior Research

Methods

& Instrumentation,

1980.

REFERENCE

12,

587-595.

NOTES

I. Bjorklund. D. F.. Thompson. B. E.. & Ornstein. P. A. Developmental trends in children‘s typicality .iudgments. Behavior Research Methods & Instrumentation. in press. 2. Bjorklund, D. F., Greenberg, M. S., Hurlbert. B. E., & Thompson, B. E. Children’s cued retrieval and the structure of semantic memory. Paper presented at meeting of Southeastern Conference on Human Development, Atlanta, Ga., April 1978. RECEIVED:

December IO. IYXI: REVISED:March 3. 1982: May II. 1982.