Memory & Cognition
1981, Vol. 9 (2),164-168
The effects of search criteria and retrieval cue
availability on memory for words
A. G. R. McCLELLAND, R. E. RAWLES, and F. E. SINCLAIR
University College London, London WCIE 6BT, England
The effects of search criteria upon recall performance following a word classification task
were investigated. Rival hypotheses were tested, one based on the number of cues available
at retrieval and the other on the amount of processing carried out during classification.
After being presented with two-component category names, subjects classified words on the
basis of either category membership or the number of semantic components the words shared
with a target category, and then they received an unexpected recall test. Experiment 1
favored an amount-of-processing explanation, but it was felt that redundancy in the stimulus
materials might invalidate this conclusion. Experiment 2 confirmed this suspicion and clearly
supported an explanation in terms of retrieval cue availability. The notion of partial semantic
processing of words is discussed, and it is argued that the number of congruent encodings
formed during classification can be used as an index of elaboration and subsequent memory
performance.
Johnson-Laird, Gibbs, and de Mowbray (1978) differential recall was predicted and obtained, based on
recently reported the results of two experiments which the different number of decisions required during
suggested that some elements of the meaning of a word classification.
can be processed without having to process them all. In their paper, Johnson-Laird et al. (1978) specifically
They manipulated the amount of processing required dismissed an explanation of their results in terms of
to classify a list of words as members or nonmembers retrieval cues because, unlike some previous studies
of a given target category and obtained improved recall (e.g., Frase & Kammann, 1974; Klein & Saltz, 1976),
with increased processing. they held the number and type of word categories
The results were of great interest, not only because constant across encoding conditions, thus ruling out
they appeared to resolve the controversy surrounding an account in terms of differences in the number or
the retrieval of semantic components of words (i.e., the nature of the retrieval cues available. However, a
are all or only some of the components of a word retrieval explanation is possible if one considers the cate-
retrieved when it is encountered in a sentence), but gory names' components rather than the category names
also because the experimental manipulation promised an per se as being the most likely retrieval cues. Thus with
independent measure of elaboration (Craik & Tulving, a two-component category name, there are potentially
1975) in terms of the number of classificatory decisions two cues available for target items, one cue for words
made about a word that yield pertinent information. As with one component, and no obvious cue for words
Johnson-Laird et al. (1978) point out, this has the with neither of the target components.
advantage of leading to more explicit predictions con- In order to test this alternative hypothesis, it is
cerning the fate of nontarget items in a classification necessary to vary the number of decisions made about
experiment. For example, in their second experiment, a word during classification independently of the
they employed two-component categories, such as number of components it has in common with the target
"consumable solid." On the basis of their argument, category. In the situation in which subjects have to
it should take a subject 2.0 decisions to classify a target search for target items (target search), as in Johnson-
item (e.g., bread), 1.5 decisions, on average (assuming Laird et al.'s (1978) experiments, these two factors
random access of the components), to classify a word are always confounded. To achieve the required indepen-
with one component in common with the target cate- dence, a new task was employed in the present study in
gory (e.g., sherry, iron), and 1.0 decisions to classify an addition to target search. This new search criterion
item with neither of the components (e.g., shampoo). (component search) consisted of asking subjects to
The last two types of words are both nontargets, but indicate for each word in the list the number of semantic
components the word had in common with the target
The authors would like to thank Nigel Harvey for his helpful category. With a two-component category name, the
comments, A. R. Jonckheere for statistical advice, and the subjects must make two decisions about each word in
Social Sciences Research Council (Great Britain) for supporting
part of this research. Requests for reprints should be sent to
the list, irrespective of whether it has two, one, or none
A. G. R. McClelland, Department of Psychology, University of the components in common with the target category.
College London, Gower Street, London WClE 6BT, England. The first experiment was a partial replication of
Copyright 1981 Psychonomic Society, Inc. 164 0090-502X/81 /020164-05 $00.75/0
, SEARCH, RETRIEVAL, AND MEMORY 165
Johnson-Laird et al.'s (1978) Experiment 2, employing of the words from the list as they could. Three minutes were
component-search instructions in addition to the target- given for recall.
search instructions they had used. Clearly, for the
target-search condition, the same results obtained by Results
Johnson-Laird et al. would be predicted on the basis of The proportions of words recalled at the three
either an "amount-of-processing" or a "number of component levels and the overall recall under the two
retrieval cues" hypothesis. However, for the component- search conditions are shown in Table l , The particular
search conditions, there are two distinct alternatives. target category searched for did not have a reliable
Amount of processing would predict no difference in effect on recall performance [F(3,24) = 1.06]. Recall
recall across the three component levels and a conse- following component search was significantly better
quent interaction between search strategy (target or than that following target search [F(1 ,24) = 6.31,
component) and component number. An explanation p < .025]. The number of semantic components in
in terms of the number of retrieval cues available would common with the target category reliably affected recall
predict a pattern of results similar to those for target [F(2,48) = 18.14, p< .001], and this factor interacted
search and, consequently, no interaction. with both search task [F(2,48)= 10.37, p< .001] and
category type [F(6,48) = 3.86, P < .05]. Planned
EXPERIMENT I comparisons showed that under target-search conditions,
component number was a reliable effect [F(2,24) =
Method
The subjects were run in two groups, one group receiving 33.72, P < .001], but under component-search condi-
target-search and the other receiving component-search classi- tions, no such effect was found [F(2,24) = .46] .
fication instructions. Within these groups, subjects were assigned Further analysis revealed a significant trend across
atrandom to one of the four following target categories: consum- components in the target-search condition (page's
able solids, consumable liquids, nonconsumable solids, and non-
consumable liquids. Each subject searched through the same word
L=2l8, p<.OOl), but no trend in the component-
list, although the order of presentation was randomized across search condition (page's L = 192).
subjects. Following the classification task, an unexpected free
recall test was administered.
Materials. The stimulus list consisted of 10 words from each Discussion
of the four categories (bread, cake, garlic, peanut, pear, radish, The results indicate that amount of processing signif-
rice, salmon, steak, toffee; coffee, gravy, lager, lemonade, milk, icantly affects recall and that the number of decisions
sherry, soup, syrup, vinegar, vodka; bronze, concrete, gravel, made about a word can be used as an index of elabora-
jade, nylon, plastic, ruby, silk, wool, zinc; acid, dye, glue, ink,
lotion, paint, paraffin, perfume, shampoo, varnish). As far as
tion. With target-search instructions, the results replicate
possible, the words were balanced across categories for length those obtained by Johnson-Laird et al. (1978, Experi-
and frequency of usage (Kucera & Francis, 1967). The list ment 2), whereas with component-search instructions,
was presented in a booklet consisting of a front cover, a page no differences were found in recall performance across
with the target category printed on it, and four pages of words, the three component levels. Thus, it appears that it is
each word printed in uppercase and placed on a separate line.
The words were assigned at random to the pages, and the the number of decisions made during encoding rather
order of pages in the booklets was also randomized. than the number of congruent components available as
Subjects. Thirty-two volunteer subjects participated in the retrieval cues that is the crucial variable affecting recall
experiment. They were mostly junior and senior high school performance.
students, together with some 1st -year undergraduates from the
department of psychology, University College London. None of However, in comparison with results from other
the subjects had previously taken part in an experiment of this experiments, the present data are rather surprising. For
kind. example, Craik and Tulving (1975) noted that in the
Procedure. Subjects were assigned at random 10 one of the classification tasks they used, words that were responded
experimental conditions and were run under either target- or
to positively (i.e., category members) were generally
component-search instructions, respectively. Both groups were
informed that the experiment was concerned with words and better recalled and recognized than were words responded
how people classified them into categories. Subjects were told to negatively. Indeed, they introduced the notion of
that the booklets in front of them consisted of a page giving the elaboration in order to explain this result, arguing that
name of a target category followed by four pages of words. positive instances could be encoded in an integrated
Under target-search instructions, they were asked to read the
category name carefully and to proceed through the booklet fashion with the encoding question because they were
putting a checkrnark by category members and a cross by non- , "congruous" with it (Schulman, 1974), whereas negative
members. Under component-search instructions, an example
of each type of category/item pairing was given. The category Table I
used was "dangerous fish," and the two-, one-, and zero- Proportions of Words Correcdy Recalled in Experiment 1
component words presented as examples were "SHARK,"
Number of Target Category Components
"POISON," "TROUT," and "PAPER," respectively. The
Possessed by the Words
subjects were instructed to go through the booklet putting a 2,
1, or 0 by each word as an indication of how many components Two One Zero Overall
that word had in common with the target category. Both groups
were asked to work as quickly and accurately as possible and to Mean SO Mean SO Mean SO Mean SO
close their booklets as soon as they had finished. All subjects
completed the task within 2 min and were instructed to turn Target Search .48 .13 .27 .10 .16 .12 .29 .16
their booklets over and write down on the back cover as many Component Search .39 .15 .36 .12 .35 .15 .37 .20
1981, Vol. 9 (2),164-168
The effects of search criteria and retrieval cue
availability on memory for words
A. G. R. McCLELLAND, R. E. RAWLES, and F. E. SINCLAIR
University College London, London WCIE 6BT, England
The effects of search criteria upon recall performance following a word classification task
were investigated. Rival hypotheses were tested, one based on the number of cues available
at retrieval and the other on the amount of processing carried out during classification.
After being presented with two-component category names, subjects classified words on the
basis of either category membership or the number of semantic components the words shared
with a target category, and then they received an unexpected recall test. Experiment 1
favored an amount-of-processing explanation, but it was felt that redundancy in the stimulus
materials might invalidate this conclusion. Experiment 2 confirmed this suspicion and clearly
supported an explanation in terms of retrieval cue availability. The notion of partial semantic
processing of words is discussed, and it is argued that the number of congruent encodings
formed during classification can be used as an index of elaboration and subsequent memory
performance.
Johnson-Laird, Gibbs, and de Mowbray (1978) differential recall was predicted and obtained, based on
recently reported the results of two experiments which the different number of decisions required during
suggested that some elements of the meaning of a word classification.
can be processed without having to process them all. In their paper, Johnson-Laird et al. (1978) specifically
They manipulated the amount of processing required dismissed an explanation of their results in terms of
to classify a list of words as members or nonmembers retrieval cues because, unlike some previous studies
of a given target category and obtained improved recall (e.g., Frase & Kammann, 1974; Klein & Saltz, 1976),
with increased processing. they held the number and type of word categories
The results were of great interest, not only because constant across encoding conditions, thus ruling out
they appeared to resolve the controversy surrounding an account in terms of differences in the number or
the retrieval of semantic components of words (i.e., the nature of the retrieval cues available. However, a
are all or only some of the components of a word retrieval explanation is possible if one considers the cate-
retrieved when it is encountered in a sentence), but gory names' components rather than the category names
also because the experimental manipulation promised an per se as being the most likely retrieval cues. Thus with
independent measure of elaboration (Craik & Tulving, a two-component category name, there are potentially
1975) in terms of the number of classificatory decisions two cues available for target items, one cue for words
made about a word that yield pertinent information. As with one component, and no obvious cue for words
Johnson-Laird et al. (1978) point out, this has the with neither of the target components.
advantage of leading to more explicit predictions con- In order to test this alternative hypothesis, it is
cerning the fate of nontarget items in a classification necessary to vary the number of decisions made about
experiment. For example, in their second experiment, a word during classification independently of the
they employed two-component categories, such as number of components it has in common with the target
"consumable solid." On the basis of their argument, category. In the situation in which subjects have to
it should take a subject 2.0 decisions to classify a target search for target items (target search), as in Johnson-
item (e.g., bread), 1.5 decisions, on average (assuming Laird et al.'s (1978) experiments, these two factors
random access of the components), to classify a word are always confounded. To achieve the required indepen-
with one component in common with the target cate- dence, a new task was employed in the present study in
gory (e.g., sherry, iron), and 1.0 decisions to classify an addition to target search. This new search criterion
item with neither of the components (e.g., shampoo). (component search) consisted of asking subjects to
The last two types of words are both nontargets, but indicate for each word in the list the number of semantic
components the word had in common with the target
The authors would like to thank Nigel Harvey for his helpful category. With a two-component category name, the
comments, A. R. Jonckheere for statistical advice, and the subjects must make two decisions about each word in
Social Sciences Research Council (Great Britain) for supporting
part of this research. Requests for reprints should be sent to
the list, irrespective of whether it has two, one, or none
A. G. R. McClelland, Department of Psychology, University of the components in common with the target category.
College London, Gower Street, London WClE 6BT, England. The first experiment was a partial replication of
Copyright 1981 Psychonomic Society, Inc. 164 0090-502X/81 /020164-05 $00.75/0
, SEARCH, RETRIEVAL, AND MEMORY 165
Johnson-Laird et al.'s (1978) Experiment 2, employing of the words from the list as they could. Three minutes were
component-search instructions in addition to the target- given for recall.
search instructions they had used. Clearly, for the
target-search condition, the same results obtained by Results
Johnson-Laird et al. would be predicted on the basis of The proportions of words recalled at the three
either an "amount-of-processing" or a "number of component levels and the overall recall under the two
retrieval cues" hypothesis. However, for the component- search conditions are shown in Table l , The particular
search conditions, there are two distinct alternatives. target category searched for did not have a reliable
Amount of processing would predict no difference in effect on recall performance [F(3,24) = 1.06]. Recall
recall across the three component levels and a conse- following component search was significantly better
quent interaction between search strategy (target or than that following target search [F(1 ,24) = 6.31,
component) and component number. An explanation p < .025]. The number of semantic components in
in terms of the number of retrieval cues available would common with the target category reliably affected recall
predict a pattern of results similar to those for target [F(2,48) = 18.14, p< .001], and this factor interacted
search and, consequently, no interaction. with both search task [F(2,48)= 10.37, p< .001] and
category type [F(6,48) = 3.86, P < .05]. Planned
EXPERIMENT I comparisons showed that under target-search conditions,
component number was a reliable effect [F(2,24) =
Method
The subjects were run in two groups, one group receiving 33.72, P < .001], but under component-search condi-
target-search and the other receiving component-search classi- tions, no such effect was found [F(2,24) = .46] .
fication instructions. Within these groups, subjects were assigned Further analysis revealed a significant trend across
atrandom to one of the four following target categories: consum- components in the target-search condition (page's
able solids, consumable liquids, nonconsumable solids, and non-
consumable liquids. Each subject searched through the same word
L=2l8, p<.OOl), but no trend in the component-
list, although the order of presentation was randomized across search condition (page's L = 192).
subjects. Following the classification task, an unexpected free
recall test was administered.
Materials. The stimulus list consisted of 10 words from each Discussion
of the four categories (bread, cake, garlic, peanut, pear, radish, The results indicate that amount of processing signif-
rice, salmon, steak, toffee; coffee, gravy, lager, lemonade, milk, icantly affects recall and that the number of decisions
sherry, soup, syrup, vinegar, vodka; bronze, concrete, gravel, made about a word can be used as an index of elabora-
jade, nylon, plastic, ruby, silk, wool, zinc; acid, dye, glue, ink,
lotion, paint, paraffin, perfume, shampoo, varnish). As far as
tion. With target-search instructions, the results replicate
possible, the words were balanced across categories for length those obtained by Johnson-Laird et al. (1978, Experi-
and frequency of usage (Kucera & Francis, 1967). The list ment 2), whereas with component-search instructions,
was presented in a booklet consisting of a front cover, a page no differences were found in recall performance across
with the target category printed on it, and four pages of words, the three component levels. Thus, it appears that it is
each word printed in uppercase and placed on a separate line.
The words were assigned at random to the pages, and the the number of decisions made during encoding rather
order of pages in the booklets was also randomized. than the number of congruent components available as
Subjects. Thirty-two volunteer subjects participated in the retrieval cues that is the crucial variable affecting recall
experiment. They were mostly junior and senior high school performance.
students, together with some 1st -year undergraduates from the
department of psychology, University College London. None of However, in comparison with results from other
the subjects had previously taken part in an experiment of this experiments, the present data are rather surprising. For
kind. example, Craik and Tulving (1975) noted that in the
Procedure. Subjects were assigned at random 10 one of the classification tasks they used, words that were responded
experimental conditions and were run under either target- or
to positively (i.e., category members) were generally
component-search instructions, respectively. Both groups were
informed that the experiment was concerned with words and better recalled and recognized than were words responded
how people classified them into categories. Subjects were told to negatively. Indeed, they introduced the notion of
that the booklets in front of them consisted of a page giving the elaboration in order to explain this result, arguing that
name of a target category followed by four pages of words. positive instances could be encoded in an integrated
Under target-search instructions, they were asked to read the
category name carefully and to proceed through the booklet fashion with the encoding question because they were
putting a checkrnark by category members and a cross by non- , "congruous" with it (Schulman, 1974), whereas negative
members. Under component-search instructions, an example
of each type of category/item pairing was given. The category Table I
used was "dangerous fish," and the two-, one-, and zero- Proportions of Words Correcdy Recalled in Experiment 1
component words presented as examples were "SHARK,"
Number of Target Category Components
"POISON," "TROUT," and "PAPER," respectively. The
Possessed by the Words
subjects were instructed to go through the booklet putting a 2,
1, or 0 by each word as an indication of how many components Two One Zero Overall
that word had in common with the target category. Both groups
were asked to work as quickly and accurately as possible and to Mean SO Mean SO Mean SO Mean SO
close their booklets as soon as they had finished. All subjects
completed the task within 2 min and were instructed to turn Target Search .48 .13 .27 .10 .16 .12 .29 .16
their booklets over and write down on the back cover as many Component Search .39 .15 .36 .12 .35 .15 .37 .20
