Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
104
changes in perceptual aspects of items and, as such, is considered to be a data-driven test (Jacoby, 1983b; Jacoby & Dallas,
1981; Roedigcr & Blaxton, 1987b; Roedigcr, Wcldon, &
Challis, 1989). In addition, this test is not affected by manipulations that vary conceptually driven processing such as
levels of processing manipulations (Jacoby & Dallas, 1981).
Thus, we expected that word identification would not be
affected by organization, and that performance on this test
would be dissociated from performance on the categoryproduction test.
Five experiments were conducted to examine the effects of
organization on implicit and explicit tests. In Experiment 1,
we studied the effects of organization on category production
and free and cued recall by varying the method of presentation
of the items. Study words from a category were presented
either in a blocked order to encourage organization or in a
random order. In Experiment 2. we examined the generality
of the effect of organization on category production and free
and cued recall by using both high- and low-frequency instances of categories. In Experiment 3, we manipulated organization by giving organizational instructions rather than
by varying list structure. In Experiment 4, the time course of
forgetting in category production was compared with that of
free and cued recall. Finally, in Experiment 5, the effect of
method of presentation (blocked vs. random) on category
production was compared with its effect on word identification.
Experiment 1
The purpose of this experiment was to compare the effects
of organization on category production and free and cued
recall. One criterion proposed for classifying tests as data- or
conceptually driven is that data-driven tests should not be
affected by manipulations that vary elaborative or conceptually driven processing such as organization, whereas conceptually driven tests should benefit from these manipulations
(Hashtroudi, Ferguson, Rappold, & Chrosniak, 1988; Roediger & Blaxton, 1987a, 1987b; Roediger, Weldon, & Challis,
1989). To examine whether category production benefits from
conceptually driven processing, we used two study conditions
in which organization was varied by changing list structure
(Bower et al., 1969; Cofer et al., 1966). In the first condition,
subjects were presented with categorically related words (e.g.,
eagle, falcon, sparrow) from several different categories (e.g.,
animals, fruits) in a blocked order for study. In the second
condition, the same categorically related words were presented
in a random order. It was expected that the list structure in
the blocked condition would encourage subjects to interrelate
the items actively and to form a unitized memory representation.
The blocked/random organizational manipulation has
been used widely in free and cued recall, and has been shown
to be effective in improving performance on these explicit
memory tests (e.g., Bower el al., 1969; Cofer et al., 1966;
Dallett, 1964; Horton & Cofer, 1975). As with the previous
studies, we expected that blocked presentation would result
in higher recall than random presentation. The comparison
of the effect of method of presentation (blocked vs. random)
105
Method
Subjects. The subjects were 96 undergraduates from George
Washington University who received course credit for their participation. Subjects were all tested individually in this experiment and
in all the other experiments reported in this article.
Design and materials. The design was a 2 x 3 factorial with
method of presentation (blocked and random) and type of test
(category production, free recall, and cued recall) as between-subjects
variables. Sixteen subjects were tested in each condition.
The materials were selected from the Battig and Montague (1969)
norms in accordance with the procedures described in Graf et al.
(1985). From each of the 12 categories (a sport, a fruit, a piece of
furniture, a kitchen utensil, a bird, a color, a four-footed animal, an
article of clothing, a tree, a musical instrument, a part of the human
body, a vegetable), six common instances (e.g., softball, skiing, boxing,
volleyball, polo, hunting) were selected. These items were not ranked
among the 10 most frequent instances, but each item was listed by at
least 10 subjects in the sample of 400 subjects used in the Battig and
Montague normative study. The average rank of the selected instances
was 23.30 (range: 11-48).
Six of the 12 categories were assigned to one list and six to another,
resulting in two lists of 36 items (6 categories of 6 items each).
Subjects heard one of the lists (study list) during the learning phase;
the other list served as a new list at test. Across subjects, each list
served equally often as a study list and as a new list. The two lists
were matched in "category potency" (Battig & Montague, 1969), that
is, in the number of instances produced by subjects in 30 s for a given
category (List 1, M = 8.60; List 2,M = 7.80) and in the average rank
of the items {List 1, M= 23.08; List 2, M = 23.42).
Procedure. At acquisition, a list of 36 items was presented auditorily at a rate of 3 s per item. Half of the subjects heard the instances
of the categories in a random order with the restriction that no 2
items from the same category occurred adjacently. The other half of
the subjects heard the instances in a blocked order with all the items
from one category occurring adjacently. Subjects were instructed to
learn the list for a later memory test, but the type of test was not
specified.
Immediately after the study list, subjects were given one of the
three different types of tests. In the category-production condition,
subjects were given the 12 category titles one at a time in a random
order. They were asked to produce eight instances for each category
title. Eight instances were required to increase the likelihood that
subjects produce instances other than the most common ones (Graf
et al., 1985). Six of the 12 categories were always from the study list,
and the other 6 were from the list that had not been presented to that
subject (new list). Subjects were instructed to "say eight things that
belong to that category as fast as possible.1' However, no time limit
was specified. Subjects rarely had difficulty generating eight instances.
To familiarize the subjects with the procedure, they were given a
sample category title (a relative) and three examples of that title
(father, brother, and sister) with instructions to produce five more
instances of that category as fast as they could (Graf et al., 1985).
In the free-recall condition, subjects were instructed to write down
as many words as they could remember from the study list. They
were given 3 min for recall but most subjects finished faster.
The cued-recall condition differed from the category-production
condition only with respect to instructions. Subjects were given the
12 category titles one at a time, and were told that some of the
categories would refer to words from the study list and others would
not. They were instructed to "use the category names as cues to help
recall the words from the study list." Again no time limit was specified.
Blocked
Random
.32
.09
.23
.20
.10
,10
.41
.29
.39
.27
106
Method
Subjects. The subjects were 144 undergraduates from the same
subject pool used in Experiment 1.
Design and materials. The design was a 2 x 2 x 3 factorial with
presentation method (blocked and random), frequency of the category
instances (high-frequency and low-frequency instances), and test type
(category production, free recall, and cued recall) as between-subjects
variables. Twelve subjects were tested in each condition.
107
Table 2
Experiment 2: Proportions of Responses in Category Production, Free Recall, and Cued
Recall as a Function of Method of Presentation and Category Instance Frequency
Method of presentation
Measure
Category production
High-frequency instances
Old
New
Priming
Low-frequency instances
Old
New
Priming
Free recall
High-frequency instances
Old
Low-frequency instances
Old
Cued recall
High-frequency instances
Old
Low-frequency instances
Old
Blocked
Random
.57
.36
.21
.50
.34
.16
.28
.06
.22
.20
.08
.12
.57
.47
.42
.34
.58
.45
.42
.28
108
Method
Subjects. The subjects were 48 undergraduates from the same
subject pool as in the other experiments.
Design and materials. The design was a 2 x 2 factorial with type
of instructions (organizational instructions and no instructions) and
test type (category production and free recall) as between-subjects
variables. Twelve subjects were tested in each condition. The stimulus
materials were the same as in Experiment 1, except that the items in
the study list were presented randomly.
Procedure. At study, subjects in the no-instructions condition
were informed that they would be presented with a list of words
belonging to six categories, and that learning the category names
might improve their performance on a later memory test. Category
names were read to subjects before list presentation. Subjects in the
organizational instructions condition were given the same instructions, but they were additionally told that, at study, they should try
to group the members of the different categories in their minds. All
other details of the study presentation were the same as in the previous
experiments.
Following study, subjects received either a category-production test
or a free-recall test. All details of these tests were identical to those
described in the previous experiments.
Organizational
instructions
No
instructions
.34
.09
.25
.28
.12
.16
,41
.30
at study than for subjects who did not receive these instructions, F(\, 22) = 5.75, MSt = 0.009. Additional analyses of
the category-production data showed that priming occurred
in both the organizational instructions and no-instructions
conditions (both ts > 1.80). Analyses of the free-recall data
also showed that subjects who received instructions to organize at study had higher recall than those who did not, F(l,
22) - 7.78, MSe = 0.010. It appears that having subjects
actively organize the categorized words at study improved
performance in both category production and free recall. An
overall ANOVA with test type as a variable was consistent
with the separate ANOVAs. The interaction of instructions
and test type was not significant, F < 1.
As in the previous experiments, examination of the first
three items produced at test showed that subjects use different
strategies in category' production and recall. In category production, only 33% of the first three words produced at test
were target words from the study list, whereas in free recall
94% of the first three words recalled were target words.
Furthermore, there was no increase in the production of target
words from the first two (34%), to the middle two (32%), and
the last two (27%) categories, F{2, 44) = 1.19, MS, = 0.026,
for either type of instructional condition, (F < 1, for the
interaction of category position and type of instructions).
In short, it appears that instructions to organize items at
study improved both category-production and free-recall performance. These results also indicate that contiguous presentation of category instances at study was not necessary to
demonstrate an effect of organization on category production;
instructions to organize enhanced priming performance even
when the items were presented randomly. Similarly, the effect
of organization on category production could not be attributed to the detection of the category names and automatic
strengthening of category-in stance associations because these
processes were likely to occur in both instructional conditions.
Finally, the results of this experiment combined with those of
the first two experiments clearly show that organizational
activities, induced either by list structure or by instructions,
improved performance on category production and had similar effects on free recall.
Experiment 4
Several studies have reported experimental dissociations
between priming tests and explicit memory tests as a function
of delay. Priming is maintained over long retention intervals,
109
110
Table 4
Experiment 4: Proportions of Responses in Category
Production, Free Recall, and Cued Recall as a Function of
Method of Presentation and Delay
Delay
Measure
Category production
Blocked
Old
New
Priming
Random
Old
New
Priming
Free recall
Blocked
Old
Omin
45 min
24 hr
.33
.10
.23
.25
.11
.14
.23
.10
.13
.22
.11
.11
.22
.11
.11
.20
.09
.11
.42
.33
.22
.34
.24
.18
.48
.33
.30
.30
.23
.20
Random
Old
Cued recall
Blocked
Old
Random
Old
the overall analysis was that in contrast to the separate analyses, the three-way interaction of method of presentation,
delay, and test type was not significant, F< 1. As mentioned
earlier, because of the scaling differences between category
production and recall tests, the separate analyses are preferable.
One explanation for the short-lived effect of organization
on category production is that subjects use explicit retrieval
in the immediate test, but do not use this strategy with longer
retention intervals when the study items are less accessible.
However, if this were the case, the probability of producing
items from the study list would be greater in the immediate
than in the delayed tests. In fact, the percentage of the first
three items at test that were from the study list did not differ
across the retention interval (29%, 21%, and 29%, for immediate, 45-min delay, and 24-hr delay, respectively), F < 1,
suggesting that subjects do not use different strategies in the
immediate and delayed category-product ion tests.
As in the first three experiments, there was evidence that
subjects in category production did not use explicit retrieval.
Of the first three items produced in the category-production
test, collapsed across delay, only 269c were from the study
list. In contrast, 94% and 83% of the first three items in free
recall and cued recall, respectively, were from the study list.
Similarly, as in the other experiments, there was no evidence
that production of targets from the study list increased for
either the random or blocked presentation when subjects
"caught on" to the goal of the test. The percentage of target
words produced from the study list was 27%, 24%, and 22%
for the first two, the middle two, and the last two categories
at test, respectively. This difference was significant, F(2. 188)
= 3.92, MSt 0.019, but contrary' to expectations, NewmanKeuls tests showed that production of targets was greater in
the first two than in the last two categories at test. The
interaction of category position and method of presentation
was not significant, F{2, 188) - 1.95, MS, = 0.019.
In addition to the item analyses, the dissociation between
category production and cued recall as a function of delay
provided evidence that category-production performance was
not mediated by explicit retrieval. Following the logic of the
retrieval intcntionality criterion (Schacter et al, 1989), the
different time course of forgetting in category production and
cued recall and the short-lived effect of organization on category production compared with its long-lasting effect on cued
recall suggest that subjects do not treat category production
like cued recall.
Overall, the results of Experiment 4 showed that, as in the
previous three experiments, priming in category production
was increased by organization. However, the separate analyses
suggested that the effects of organization did not last over the
delay interval. Apparently, the enhancement of category production by conceptually driven processing is short-lived. Freeand cued-recall performance was also improved by organization but, as in other experiments (Horton & Cofer, 1975;
Mandler et al., 1969), organization improved performance
even after a delay interval.
The results of the present experiment also showed that
priming in category production declined over time, although
significant priming occurred even after a 24-hr delay. Thus,
111
Experiment 5
In this experiment, we compared the effect of organization
on category production to its effect on the word-identification
priming test. The first four experiments clearly established
that category production is a conceptually driven test, as it
benefited from organization. It was expected, therefore, that
performance on this test would be dissociated from performance on a data-driven priming test such as word identification
as a function of organization. Word identification is considered to be data driven because performance on this test does
not benefit from manipulations that vary conceptual or elaborative processing such as levels of processing or generation
(Jacoby, 1983b; Jacoby & Dallas, 1981). In addition, word
identification is sensitive to changes in perceptual aspects of
information such as changes in modality of presentation
between study and test and, as such, can be classified as data
driven (Clarke & Morton, 1983; Jacoby & Dallas, 1981;
Kirsner, Milech, & Standen, 1983). It appears that, in contrast
to category production, word identification should not be
affected by organization.
Method
Subjects. Sixty-four subjects from the same subject pool as in the
earlier experiments participated in this study.
Design and materials. The design was a 2 x 2 factorial with
method of presentation (blocked and random) and test type (category
production and word identification) as between-subjects variables.
Sixteen subjects were tested in each condition.
The stimulus materials were selected from the Battig and Montague
(1969) norms according to the procedures described in Experiment
1, except that because the present experiment included word identification, we chose only words that were four to six letters long. This
experiment also differed from Experiment 1 in that 16 categories (a
bird, a piece of furniture, a sport, a kitchen utensil, a tree, a fourfooted animal, a part of the human body, a musical instrument, an
article of clothing, a weapon, a dance, a color, a substance for flavoring
food, an insect, a fish, a part of a building) were used instead of 12.
In addition, seven common instances instead of six (e.g., turkey,
raven, pigeon, dove, wren, parrot, lark) were selected. The average
rank of the selected category members was 23.29 (range: 11-48).
Eight ofthe 16 categories were assigned to one list and 8 to another,
resulting in two lists of 56 items each (8 categories of 7 items each).
Across subjects, both lists served equally often as the study list and
the new list. The two lists were matched in category potency (List 1,
M = 7.97; List 2, M = 7.84) and category rank (List 1, M = 23.10;
List 2, M = 23.48). The test list for the word-identification test was
randomly ordered with the restriction that no two words from the
same category could occur adjacently.
All study words were typed in uppercase letters in the center of 3
x 5 in. index cards. The word-identification test was presented with
Table 5
Experiment 5: Proportions of Responses in Category
Production and Word Identification as a Function of Method
oj Presentation
Method of presentation
Measure
Category production
Old
New
Priming
Word identification
Old
New
Priming
Blocked
Random
.28
.09
.19
.23
.11
.12
.47
.28
.19
.55
.34
.21
112
General Discussion
The results of the present series of experiments clearly
indicate that organization improved performance on category-production and recall tests. The results of Experiment 1
showed that, as with free and cued recall, blocked presentation
relative to random presentation increased priming in category
production. Experiment 2 demonstrated that this finding was
reliable and was obtained with both high-frequency and lowfrequency instances of categories. In Experiment 3, it was
found that varying organization by instructions, rather than
by method of presentation, also had similar effects on priming
and free recall; performance improved in both cases. Finally,
the results of Experiment 5 showed that the effect of organization on priming was limited to the conceptually driven
category-production test and was not observed with the datadriven word-identification test.
Although these results indicate that the category-production
test and recall tests are similar, other aspects of the data
suggest certain differences between the two types of tests. For
example, in Experiment 2, free and cued recall were higher
for high-frequency than for low-frequency instances of categories, whereas there wras no significant difference between
the two types of instances in category-production priming.
Thus, in terms of category-instance frequency, performance
on category production was dissociated from performance on
free and cued recall.
In addition, in Experiment 4, the effect of method of
presentation on priming in category production was evident
only when subjects were tested immediately. In contrast, in
free and cued recall, the effect of organization was long lasting.
Furthermore, the results of this experiment showed that the
time course of forgetting in category production was different
from free and cued recall. The rate of forgetting seemed to be
slower in the category-production test than in the recall tests.
Taken together, these differences suggest that somewhat different processes may be involved in the two types of tests.
13
pair in the list) facilitated performance on both the stemcompletion priming test and the explicit cued-recall test.
Grouping (e.g., generating related sentences for word pairs to
form a story), on the other hand, enhanced performance on
the cued-recall test but not on the stem-completion test. The
latter finding seems to be inconsistent with our results that
organization improved priming in category production.
There are, however, a number of differences between the
two studies. One difference is that Graf and Schacter (1989)
compared the effect of grouping on memory for newly acquired associations rather than on memory for items that
have a preexisting representation, such as the words used in
the present experiments. In addition, grouping processes in
Graf and Schacter's experiment required forming new interitem relationships fa story) among word pairs that were unrelated to each other, whereas grouping in our study involved
using already existing interitcm relationships among categorized words. Finally, compared with the category-production
test used in our study, the stem-completion test used by Graf
and Schacter emphasizes processing of individual word pairs
rather than processing of relational information among word
pairs. Therefore, the relational information acquired by
grouping the items at study may not affect performance on
the stem-completion test. These differences may have contributed to the discrepant findings of the effects of grouping on
priming tests in the two studies.
In general, the results of the present experiments arc consistent with the transfer-appropriate processing account of the
differences between implicit and explicit tests (Roediger &
Blaxton, 1987b; Roediger, Weldon, & Chain's, 1989). This
framework focuses less on the type of memory test (implicit
vs. explicit) than on the match of mental operations performed at study and test. In this view, performance will benefit
to the extent that operations involved in the test match those
used during study. Conceptually driven processing at study
benefits performance on both implicit and explicit tests that
rely on conceptually driven processing. Thus, in the present
experiments, organization had parallel effects on the implicit
test of category production and the explicit tests of free and
cued recall because all of these tests rely on conceptually
driven processing. In addition, our results demonstrate that
conceptually driven processing at study had different effects
on two implicit tests when one test was data driven and the
other was conceptually driven. Organization increased priming in the conceptually driven category-production test but
not in the data-driven word-identification test.
Nevertheless, there are certain findings in the present experiments that cannot be readily accommodated within the
transfer-appropriate processing framework (Roediger, Weldon, & Challis, 1989). For example, the framework does not
predict or explain why variables other than those involving
the manipulation of data-driven and conceptually driven
processing have different effects on implicit and explicit tests.
Hence, it is not clear how the framework can account for the
differential effect of category-instance frequency on category
production and recall or the short-lived effect of organization
on category production. Likewise, the framework has not
addressed the issue of the differences in the time course of
forgetting for implicit and explicit tests and, therefore, does
References
Battig, W. F., & Montague, W. E. (1969). Category norms for verbal
items in 56 categories: A replication and extension of the Connecticut category norms. Journal of Experimental Psychology Monograph, #0(3,"PL 2).
14