Prior Knowledge and Memory - American Psychological Association

3 downloads 0 Views 974KB Size Report
The presentation of a familiar word activates related concepts, and, once encoded, related concepts interfere with memory far the word actually presented.
Journal of Experimental Psychology: Learning, Memory, and Cognition 1987, Vol. 13, No. 1,54-63

Copyright 1987 by the American Psychological Association, Inc. 0278-T393/87/SOO.75

Prior Knowledge and Memory: The Episodic Encoding of Implicitly Activated Associates and Rhymes Douglas L. Nelson, Maria Teresa Bajo, and Jose Canas

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

University of South Florida The presentation of a familiar word activates related concepts, and, once encoded, related concepts interfere with memory far the word actually presented. Presented words that activate larger numbers of related concepts are generally more difficult to recall than those that activate smaller numbers. The purpose of the present experiments was to explore the effects of study-trial orientation and test delay on the encoding of both rhyme- and meaning-related concepts. The results of Experiments 1 and 2 indicate that meaning-related concepts are encoded and interfere with memory for the presented target regardless of study-trial orientation. Interference is obtained even when the study-trial context emphasizes phonemic information and subjects are incidentally oriented to rating rhyme properties. However, this interference effect disappears when the test trial is delayed. In contrast, the results of Experiments 1-4 indicate that rhyme-related concepts are encoded and interfere with memory for the presented target only when subjects explicitly attend to the rhyme dimension. Once oriented, this interference effect is found after a relatively long delay. These differences are attributed to differences in attentional processing. The encoding of meaning-related concepts results from relatively automatic processes and the encoding of rhyme-related concepts requires subjects to attend to rhyme.

Experiencing a familiar concept implicitly activates related concepts. This associative principle predates psychology as a separate discipline and, though it has the status of a truism, relatively little is known about how the related concepts influence memory for the concept that is directly experienced. Rir example, an idea expressed in conversation by a speaker typically activates related knowledge in a listener. Some of this knowledge may be brought to conscious awareness and some may not. One interesting problem concerns the manner in which this related knowledge affects the listener's memory for the information actually presented by the speaker. This problem is interesting partly because it seems to touch on a number of problem areas, including the roles of automatic and controlled processing (e.g., Schneider & Shiflnn, 1977; Shiflrin & Schneider, 1977), traditional memory principles involving interference (e.g., Roediger, 1974; Underwood, 1945; Watkins, 1975), and encoding specificity (Tulving, 1983). Our approach to this problem has been to present subjects with familiar words known to be connected to few or to many related concepts and then to test memory for the item actually shown {e.g., Nelson, 1981; Nelson, Bajo, & Casamieva, 1985; Nelson & Friedrich, 1980). This research has indicated that words connected to a larger number of related concepts are less likely to be recalled in the presence of retrieval cues. That is, concepts that are not presented by the experimenter but are

implicitly activated and encoded in the task can interfere with memory for words that are presented. This interference effect is found regardless of whether related concepts consist of phonemically or meaningfully related items. In either case, targets having larger phonemic or meaning sets are less likely to be recalled. This result is independent of the domain of information provided by the test cue because each type of target-set size effect is found with cues that provide either phonemic or meaning information. This independence suggests that target-set size effects are at least partially linked to encoding processes. Furthermore, the importance of encoding processes is also manifested in findings showing that these effects are heavily dependent on study-trial context (Nelson, 1981). Phonemically related target-set size effects are contingent on emphasizing phonemic information during study. For example, rhyme-set size effects are not found when the target is studied in the absence of a rhyme context (e.g., studying SHIRT alone), but they are found when the target is studied in the presence of rhyming context (e.g., HURT-SHIRT). In contrast, meaning-set size effects are obtained when the target is studied in the absence of modirying context (e.g., SHIRT alone), and they are not found when it is studied in the presence of a meaning-related context (e.g., SLEEVE-SHIRT).

One explanation for these results assumes that the encoding of both types of related concepts in the general episodic context is determined by contextual factors (Nelson, 1981). Concepts that are phonemically related to the target are encoded when the study context emphasizes rhyme, and concepts that are meaningfully related to the target are encoded when the context fails to indicate a specific meaning. The encoding of these related concepts as determined by study context can affect the status of these concepts as extralist cues (Tulving, 1983). In addition, this encoding can also reduce the retrievability of the

This research was supported by Grant MH 16360 from the National Institute of Mental Health to Douglas L. Nelson. Special thanks go to Barbara tuhn and Mary LaLomia for their help in data collection. Correspondence concerning this article should be addressed to Douglas L. Nelson, Department of Psychology, University of South Florida, Tampa, Florida 33620. 54

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

PRIOR KNOWLEDGE representation corresponding to the target. Cued recall requires the production of the studied target and, to the extent that it must be recovered from a pool of encoded phonemicaUy or meaningfully related concepts, retrieval inhibition effects associated with the recovery of nontarget items should reduce the likelihood of recall (e.g., Raaijmalcers & Shiffirin, 1981; Roediger, 1974). The effect of retrieval inhibition, or cue overload, should be greater for larger than for smaller encoded sets (Nelson, McEvoy, & Friedrich, 1982;Watkins, 1975). This interpretation of target-set size effects assumes that they emanate from the encoding of related concepts during study that place greater or lesser burdens on the test cue during retrieval. For example, when HURT-SH1RT is studied, phonemically related concepts, such as FLIRT, DIRT, CURT, PERT, and so forth, presumably are encoded. When the extralist meaning cue SLEEVE is presented at test, a search through the meaning-related concepts defined by this cue in initiated. Given that the representation corresponding to SHIRT is accessed, it is used as an internal cue to instantiate episodically encoded information, including phonemic and meaning information. In this example, recovering the phonemic representation for SHIRT simultaneously recovers the prior phonemic encoding of HURT, FLIRT, DIRT, and so forth, which reduces the likelihood of recovering the specific phonemic representation corresponding to the presented target. The assumption is that the episodically encoded related concepts are recovered in parallel as a set. The larger the encoded set, the greater are the chances for retrieval inhibition, because the likelihood of recovering nontarget concepts increases. (See Nelson et al., 1982, for additional details.) The implicit encoding of related concepts assumes a critical role in this interpretation. In the meaning domain, this encoding presumably reflects the action of relatively automatic processes. College-age subjects habitually and unconsciously focus on the meaning of familiar words and need no special self-instruction or environmental support for encoding such information. Consistent with this assumption the magnitude of meaning-set size effects appears to be uninfluenced by changes in encoding orientation. These effects remained unchanged across variations in intentional learning instructions and incidental instructions requiring pleasantness ratings, concreteness ratings, or ratings of number of associates (Nelson, Bajo, & Casanueva, 1985). Hence, meaning-set size effects appear to at least meet the intentionality criterion for assuming automaticity(e.g., Hasher &Zachs, 1979; Schneider &Shiffrin, 1977). In the phonemic domain, the encoding of related concepts appears to reflect the combined action of both controlled and automatic processing. Rhyme-set size effects associated with the target are found only when rhyme is explicitly emphasized at study. Only when attention is specificaUy focused on rhyme are rhyme-related concepts then automatically activated and encoded so that they become connected to the episodic context, along with the representations for cue and target. The initial focusing of attention on rhyme is assumed to reflect controlled processing on the assumption that college-age subjects ordinarily do not attend to phonemic information in a familiar word. However, at present, little is known about the potential effect of variations in encoding orientation on processing within this domain. It could be that related-rhyme concepts are automatically encoded whenever rhyme is present, regardless of whether

55

subjects attend to rhyme or not. Hence, the results of manipulations of orientation will contribute to a better understanding of how and when prior knowledge will contribute to episodic memory performance. The primary aim of the present experiments was to explore the effects of study-trial orientation on the encoding of rhymerelated concepts. Pairs of rhymes were presented during the study trial with rhyme-set size varied in all experiments. Equal numbers of target words in each study list belonged to either large or small rhyme sets as determined by normative measures made prior to the experiments. In Experiment 1, rhyme pairs were studied under intentional learning or under incidental conditions that oriented them toward rhyme or toward meaning. In the intentional condition, subjects were told to pay attention to the rhyme relation and to remember as many targets as possible. In the incidental-rhyme condition, subjects rated each pair in terms of how well the words rhymed even though, technically speaking, all pairs rhymed equally well. In the incidental-meaning condition, subjects rated each pair on how meaningfully related the words were to each other. Subjects in both incidental conditions were told that their ratings would be used to make lists for other experiments and no mention was made of a possible test. Expectations for the results were predicated on the assumed nature of the controlled process. If rhyme-set size effects are contingent on consciously attending to rhyme, it should be possible to modify the magnitude of these effects by varying encoding orientation. These effects theoretically should be strongest in the incidental-rhyme condition and weakest in the incidental-meaning condition. The respective likelihood of attending and of not attending to rhyme should be maximized under these conditions. For example, if subjects can focus exclusively on meaning, rhyme-set size effects should not be found even though the pairs obviously rhyme. Alternatively, and contrary to theoretical expectations, rhyme-set size effects might be equally apparent under all conditions. Thisfindingcould mean that rhyme-related concepts are encoded whenever rhyme contexts are present or it could mean that the instructional manipulation did not differentially engage the control process. The secondary aim of Experiment 1 was to re-evaluate the assumption that the encoding of meaning-related concepts is automatic. In previous work, the intentional-incidental nature of the encoding orientation was varied, and this variation failed to influence the magnitude of the meaning-related target-set size effect (Nelson, Bajo, & Casanueva, 1985). An even more stringent test of the automaticity assumption would involve manipulations of encoding orientation when the targets are presented in a rhyme context. A controlled aspect to the encoding of implicitly activated meaning concepts might be revealed when the information presented allows subjects to split the focus of their attention between the different domains of meaning and rhyme. To evaluate this possibility, meaning-set size of the target was crossed with rhyme-set size of the target and with encoding orientation. Hence, the targets were divided evenly among four sets: small meaning/small rhyme, small meaning/ large rhyme, large meaning/small rhyme, and large meaning/ large rhyme. A controlled aspect to the encoding of meaningrelated concepts would be shown if meaning-set size effects are weakest when subjects are incidentally oriented to rhyme and

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

56

D. NELSON, M. BAJO, AND J. CANAS

strongest when they are incidentally oriented to meaning. Alternatively, if meaning-set size effects reflect automated encoding procedures, then this effect should be apparent across all conditions, even when the targets are studied in a rhyme context. Both meaning-related and rhyme-related responses may be encoded in parallel. In Experiment 2 the pairs were studied under intentional instructions or under incidental instructions that oriented subjects to rhyme. In addition, testing was immediate or it followed after 24 hr of distraction and delay. In the meaning domain, attention shifts from the memory task to a abstractor task requiring several minutes of multiplication was sufficient to eliminate meaning-set size effects. Words having larger meaning sets were more difficult to recall only when their related concepts were functionally connected to the episodic context (Nelson, Bajo, & Casanueva, 1985). On the assumption that rhyme-set size effects require conscious attention to rhyme such that this information becomes a part of what is remembered about the study experience, these effects should be apparent even after long periods of distraction and delay. In contrast, on the assumption that subjects need not attend consciously to rhyme to encode it, these effects should be as fragile as those associated with meaning. Finally, Experiments 3 and 4 evaluated the potential effects of encoding orientation on rhyme effects when the pairs were both rhyme and meaning related (e.g., SNEEZEFREEZE).

Experiment 1 Method Design. The design formed a 3 X 2 x 2 mixed-model factorial. Encoding orientation (intentional, incidental rhyme, or incidental meaning) was manipulated between subjects, and rhyme-set size (small or large) and meaning-set size (small or large) were manipulated within subjects. Subjects. There were 42 subjects who served in the experiment, 14 in each orienting condition with half of these assigned to each list All were selected from courses in introductory psychology under an incentive system. The experiment was run in replication blocks of 6 subjects so that every 6th subject completed a replication of the experiment, including lists. Within blocks, the order of conditions was randomly determined. Materials. Controlled association norms for rhyme and meaning were collected by presenting the to-be-normed cues to large groups of individuals (n& = 12S-242). Rhyme norms were obtained for 396 word endings. Ending sounds were tape-recorded in a female voice and pronounced twice prior to 5 s of silence. During the silent period, subjects wrote the first word they thought of that rhymed with the sound just beard. Meaning norms were obtained for 800 words and in this task subjects were asked to provide thefirstword that came to mind that was meaningfully associated to the presented item (see Nelson & McEvoy, 1979, for details). In both sets of norms, single rather than multiple responses were used to avoid problems with response chaining and retrieval inhibition and because the first response provides a better measure of set size (Joelson & Herrmann, 1978). Test-retest reliability in both sets of norms is high with Spearman rank correlations ranging from .75 to .90. These norms were used to estimate set size by counting the total number of different but appropriate words given to each cue. Relative strength was estimated by the portion of subjects giving any particular response. Because most of the items in the meaning norms were taken from responses given in the rhyme-norming task, items

could be cross-referenced in the two normative procedures. For example, the target word flag has relatively large rhyme and meaning sets, consisting of 18 and 19 items, respectively. These norms were used to construct two 24-item lists of target words that were presented in the study phase. Each list was subdivided into four sets of 6 words representing all combinations of small and large rhyme and meaning sets. Mean rhyme- and meaning-set sizes in the small/small combinations were 6.33 (SD = 1.25) and 7.67 (SD = 2.68), and in the small/large combination they were 5.92 (SD = 2.21) and 20.58 (SD = 3.29). In the large/small condition these means were 22.42 {SD = 6.64) and 6.84 (SD = 1.92), andfinally,in the large/large condition they were 21.00 (SD = 8.56) and 19.67 (SD = 3.76). Listed in this order, the rhyme study-context cues, the targets, and the meaningrelated test cues for Lists 1 and 2 are shown in Table 1. Note that meaning-related test cues were used to avoid confounding rhyme-set size of the target with rhyme-set size of the test cue. The normative set size of these cues was equated in all conditions, and it averaged 11.52 (SD = 3.65). Normative cue-to-target strength was also equated and averaged .31 (SD = .11). Hence, with the meaning-related words as cues in the normative task, the above targets were given by 31 % of the sample. Procedure. Subjects participated in individual sessions and received a single study-test trial. All cue-target pairs were typed in uppercase letters, with the cue presented to the left and underlined, and all pairs were presented by a Kodak carousel slide projector driven at a 3-s rate. Subjects were required to read each rhyme pair aloud when it appeared on the screen. In the intentional condition subjects were told to concentrate and remember as many target words as possible. No information on the nature of the test was provided at this point. In the incidental conditions, they were asked to rate each pair on a scale from 1 to 3 by saying a number. Rhyme subjects were asked to rate each pair on how well the words rhymed with each other; meaning subjects were told to rate each pair on how meaningfully related the words were to each other. Explicit examples were provided to facilitate understanding. Mean rhyme ratings averaged 2.65 and did not differ as a function of either rhyme- or meaning-set size; similarly, meaning ratings averaged 1.49 and did not differ as a function of rhyme- or meaning-set size. lest trial instructions were read immediately following the last rhyme pair shown at study. These instructions took about 30 s to read, and they described the nature of the semantic relation between the test cues and the targets. The average time between the study of a rhyme pair and its test was approximately 1.5-2.0 min. During the test trial each meaning-related cue was shown in uppercase letters, underlined, and subjects were required to read it aloud before attempting target recall. This test trial was self-paced and guessing was encouraged. The order of presentation of both study pairs and cues was randomized for each subject. Finally, two subjects in the incidental conditions reported that they suspected a memory test and they were replaced.

Results The results showing probability of correct recall are presented in Table 2. Rhyme-set size affected recall in every condition and did not appear to vary with orientation. In each comparison, targets belonging to larger rhyme categories were not as likely to be recalled as those belonging to smaller rhyme categories. This table also shows that meaning-set size of the target tended to affect the likelihood of recall. Words having larger meaning-related sets were generally not as likely to be recalled. However, as shown in the bottom row, this effect tended to be reduced when subjects were incidentally oriented to rhyme during study, and this reduction was most apparent when rhymeset size was large.

57

PRIOR KNOWLEDGE

Table 1 Materials Used in Experiment I List 2

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Listl Set

Study context

Small rhyme/small meaning

MONEY LOUD GOOF LOVE PUP LIFE

HONEY CLOUD ROOF SHOVE CUP WIFE

Small rhyme/large meaning

YEAST LOT HARP GOOSE HOOD MILK

targe rhyme/small meaning

Large rhyme/large meaning

Target

Test cue

Study context

Target

BEE SKY CEILING PUSH MUG SPOUSE

SOUND PUP CALF LOVE FETCH LIFE

ROUND CUP LAUGH SHOVE SKETCH WIFE

CIRCLE MUG JOKE PUSH DRAW SPOUSE

FEAST ROT SHARP MOOSE WOOD SILK

BANQUET DECAY POINTED ELK TIMBER SATIN

BUNCH SILK LIVER BROTH HOOD HOST

LUNCH MILK RIVER CLOTH WOOD GHOST

DINNER COW STREAM FABRIC LOG HAUNT

MIGHT FAIL LIP SOB TEA BED

NIGHT NAIL SKIP JOB KEY THREAD

EVENING TACK HOP TASK LOCK NEEDLE

TRASH SLED KEY FAIL SOB MAD

CASH THREAD BEE PAIL JOB SAD

MONEY NEEDLE HONEY BUCKET TASK GRIEF

HURT CANE NAG SAP PINK FEEL

SHIRT TRAIN FLAG MAP THINK STEAL

SLEEVE RAIL SALUTE ATLAS BRAIN THIEF

CANE MESS FLIRT SINK BAKE BOARD

RAIN DRESS DIRT THINK SNAKE SWORD

CLOUD SKIRT GROUND BRAIN MONGOOSE SABER

These trends, however, were only partially confirmed by the results of the analysis of variance (ANOVA). Rhyme and meaning-set size effects were reliable, with respective values of f\\,39) = 22.09, MSe - .026, and f{U 39) - 17.06, MSC = .039. The .05 significance level was used for all results reported. The interaction between encoding orientation and meaning set size was not significant, F{2,39) = 2.02, MSt = .039, and hence, the apparent trend for reduced meaning-set size effects with a rhyme orientation was not reliable. The analysis also indicated that encoding orientation was not a reliable source of variance, F\2, 39) = 1.98, MSe = .102, and that all other sources produced fs < 1. Discussion The results indicate that words that theoretically activate larger sets of phonemically and meaningfully related concepts

Test cue

are not as likely to be recalled. Thisfindingreplicates the results of past studies (e.g., Nelson & Friedrich, 1980; Nelson, Bajo, & Casanueva, 1985). The present results, however, also show that neither rhyme- nor meaning-set size effects depend on the nature of the encoding orientation. The incidental orientation to rhyme failed to increase the magnitude of the rhyme-set size effect and the incidental orientation to meaning failed to decrease the magnitude of this effect. This finding is contrary to the assumption that the encoding of rhyme-related concepts reflects the action of a controlled process. This information may be automatically activated and encoded whenever rhyme is emphasized during the study trial. However, these results could also mean that the presence of rhyme context for each target during study overwhelmed the encoding-orientation manipulation. In other words, the instructional manipulation may simply have failed to engage the control process differentially, and under other conditions, the encoding of implicitly activated rhyme

Table 2 Probability of Correct Recall as a Function ofEncoding Orientation (EO) and Rhyme- and Meaning-Set Size ofthe Target Meaning-set size of target Incidental EO Intentional EO

Rhyme

Meaning

Rhyme-set size

Small

Large

Small

Large

Small

Large

Small Large M

.73 .63 .68

.60 .52 .56

.69 .51 .60

.58 .51 .55

.68 .52 .60

.46 .33 .40

.62 .53

58

D. NELSON, M. BAJO, AND J. CANAS

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

may be more subject to strategic manipulation. This possibility was examined in the remaining experiments. The results of the meaning-set size manipulation were more in line with the expectation that these effects are mediated by automatic processes. They were apparent, even though each target was encoded in the presence of a rhyming word, and at least in the statistical sense, encoding orientation did not differentially influence their magnitude. However, there was a trend in the data showing a reduced effect when subjects had been incidentally oriented to rhyme during study, but possibly because of lack of sensitivity in the statistical analysis, this trend was not reliable. An attempt was made to replicate this trend in Experiment 2.

Table 3 Probability of Correct Recall as a Function of Encoding Orientation (EO), Meaning- and Rhyme-Set Size ofthe Target, and Test Delay Meaning-set size of target

Intentional EO Test delay

Rhyme-set size

Immediate

Small Large Small Large

Delayed

Incidental rhyme EO

Small

Large

Small

Large

.71 .57

.54 .51 .46 .45

.67 .57 .52 .45

.42 .38 ,47 .32

.56 .44

Experiment 2 In Experiment 2, subjects studied pairs of rhymes under intentional or under incidental rhyme instructions, with half of the subjects given an immediate test and the remainder given a test after a 24-hr delay. As in the previous experiment, rhymeand meaning-set size of the target were factorially combined. Thus, the purpose of the experiment was to evaluate the effect of test delay and to determine if the trend toward a reduced meaning-set size effect under the rhyme orientation could be replicated. On the delayed test, rhyme-set size effects were expected to be more robust than meaning-set size effects. The assumption underlying this expectation was that information encoded as a result of direct attention is more likely to be recovered at test compared to information encoded as a result of automatic processes (e.g., Balota, 1983). Hence, if the encoding of rhyme-related concepts is more likely to reflect the action of an attentional control process, then rhyme-set size effects should be more apparent after the delay.

Method Design. The experimental design formed a 2 X 2 X 2 X 2 mixedmodel factorial. Encoding orientation (intentional or incidental) and test delay (immediate or delayed) were manipulated between subjects, with rhyme- (small or large) and meaning- (small or large) set size of the target varied within subjects. Subjects. There were 64 subjects who served in the experiment, with 16 in each between-subjects condition and with 8 assigned to each list. Materials and procedure. The lists and general procedures were identical to those used for comparable conditions in Experiment 1. Subjects in the delayed testing conditions were dismissed, following the study trial, after an appointment for their return was made for the next day at the same time. No mention of the testing procedure was made at this time.

Results The results are shown in Table 3. Probability of correct recall tended to be lower on the delayed (.46) than on the immediate (.55) test. In addition, targets belonging to large rhyme sets (.46) were not as likely to be recalled as those from small rhyme sets (.55). The magnitude of this rhyme-set size effect did not appear to vary with either test delay or with encoding orientation. That is, rhyme-set size effects were as apparent on the delayed test as

on the immediate test and they were not magnified by incidentally orienting subjects toward rhyme during study. The results in Table 3 also suggest that targets having larger meaning sets were generally more difficult to recall, particularly when testing immediately followed study. The probabilities of recall on the immediate test for small and large meaning sets were, respectively, .63 and .46; on the delayed test these values were .49 and .43. Finally, whereas intentional instructions (.53) produced slightly higher recall than incidental rhyme instructions (.48), encoding orientation did not appear to alter the magnitude of the meaning-set size effect obtained on the immediate test. Hence, the reduced meaning-set size effect observed in the incidental-rhyme condition of Experiment 1 was not replicated. If anything, the meaning-set size effect tended to be slightly stronger in the incidental-rhyme condition than in the intentional condition. The results of the ANOVA showed that test delay, i*Xl, 60) = 5.34, MSe = .090, rhyme-set size, F{ 1,60) = 11.82, MS* = .035, and meaning-set size, F(l, 60) = 15.81, MSe = .058, were all reliable sources of variance. Encoding orientation was not reliable, f{lt 60) = 2.04, and neither was the interaction between meaning-set size and test delay, i^l, 60) = 2.70. However, a Fisher's least significance difference (LSD) of .08 indicated that the difference between small and large meaning sets was significant on the immediate but not on the delayed test. Finally, all other sources produced Fs of less than or near unity.

Discussion The results of Experiment 2 indicate that, when testing is immediate, targets with larger numbers of meaning-related associates are not as likely to be recalled. This effect is uninfluenced by whether subjects are oriented to intentional learning or to rating the pairs for their rhyming qualities under incidental instructions. Hence, the somewhat reduced effect of meaning-set size found in the incidental rhyme condition in Experiment 1 was not replicated in Experiment 2. Consistent with previous findings, meaning-set size effects are unaffected by variations in encoding orientation, and this conclusion holds even when targets are encoded in the presence of rhyme-context words. In contrast, as in previous experiments, these effects are vulnerable to the effects of distraction and delay. Meaning-set size

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

PRIOR KNOWLEDGE

effects are larger on immediate than on delayed tests. Taken altogether, this pattern of findings is consistent with the conclusion that meaning-set size effects reflect the influence of automatic processes. Conscious attention is not required to activate meaning-related concepts. Instead, these concepts are implicitly and automatically activated during the encoding of the target, and as long as they remain activated, they interfere with the retrievability of the target representation. The results of manipulations of rhyme-set size are quite different. Even though meaning-related cues were presented to aid retrieval, rhyme-set size effects were apparent on the immediate test and were equally robust after the 24-hr delay. The greater durability of these effects relative to those associated with meaning presumably reflects the influence of controlled processing. Actively paying attention to rhyme during study seems to ensure that rhyme-related concepts will be activated and encoded during study and retrieved during test even after a relatively long interval. Nevertheless, attempts to manipulate this attentionai control process by varying encoding orientation failed. The magnitude of rhyme-set size effects in both experiments did not change as attention was drawn toward or away from the phonemic dimension. However; although it questions the validity of the controlled processing assumption, this finding may mean only that the present manipulations of encoding orientation failed to differentially engage the control process. Merely presenting the target in a rhyme context may be insufficient for activating related rhyme concepts. Subjects may have to attend explicitly to rhyme for rhyme-set size effects to emerge. This possibility was explored in Experiments 3 and 4.

Method

Experiment 3

Results

The manipulations of encoding orientation in Experiments 1 and 2 may have Jailed because of the nature of the pairs presented during study: All were pairs of rhymes that were unrelated in terms of meaning. The unidimensional nature of the relation between the study context words and their targets may have made it unlikely that the manipulation of orientation instructions could be effective. For example, in Experiment 1, even though a portion of the subjects were asked to rate pairs on how meaningfully related the words were to each other, the fact that the words rhymed would be difficult to miss. The attentionai control process may be more easily manipulated when at least some of the pairs are both phonemicafly and meaningfully related. In Experiment 3, all the study pairs were rhyme related, with half of these pairs being low in meaning-relatedness (e.g., SNEEZE-PEAS) and half being high in meaning relatedness (e.g., SNEEZE-FREEZE). Meaning relatedness was crossed with rhyme-set size and with encoding orientation. Subjects were given either intentional instructions, incidental instructions that oriented them to rhyme, or incidental instructions that oriented them to meaning. Given the presence of meaning-related rhymes and an orientation toward meaning, subjects may not attend to rhyme and rhyme-set size effects may not emerge. Such a result would suggest that merely presenting rhyme pairs is insufficient for engaging the control process that results in the encoding or related rhyme concepts.

59

Design, The design formed a 3 x 2 x 2 mixed-model factorial. Encoding orientation was manipulated between subjects (intentional, incidental rhyme, or incidental meaning), with meaning relatedness (high or low) and rhyme-set size (small or large) varied as within-subjects factors. Subjects, There were 42 subjects who served in the experiment, with 14 serving in each encoding orientation and 7 receiving each list. They were drawn from the same source as before. Materials. The Ksts were constructed from the rhyme norms used in the earlier studies. Pairs varying in meaning relatedness were selected by searching through the words associated with small and large rhyme sets in order tofindappropriate pairs. Hence, meaning relatedness was determined by experimenter judgment. Pools of items were created and two 24-item lists were constructed by selecting the best exemplars within this pool. Each list was subdivided into four sets of six words representing all combinations of meaning relatedness and rhyme-set size. Mean rhyme-set size was equated in each relatedness condition in each list. It averaged 7.30 (SD = 1.67) and 24.58 (SD = 4.90) for small and large rhyme sets. Relative rhyme strength was also equated in all conditions for both targets and their context cues and it averaged .12 (SD =.12). Finally, as in the earlier experiments, meaning-related cues were shown at test The meaning-set size of the test cues and their relative strength in relation to their targets was equated in all conditions. Meaning-set size of the test cues averaged 11.52 (SD = 3.65), and cueto-target strength averaged .31 (SD = .11). The materials for both lists are shown in Table 4; note that targets appearing in a high-meaningrelated pairing in List 1 appeared in a low-meaning-related pairing in List 2. Procedure. The procedures and instructions were identical to those used for Experiment 1.

The results presented in Table 5 indicate that the probability of recall was lower in every condition when rhyme-set size was large. This effect did not appear to differ as a function of either encoding orientation or meaning relatedness. This trend was confirmed by the results of the statistical analysis, which indicated that only rhyme-set size was a reliable source of variance, Fil, 39) = 16.38, MSt = .029. The probabilities of recall for small and large rhyme sets were .62 and .51, respectively. Encoding orientation, F{\, 39) < 1, MSe = .071, and meaning relatedness, F(\, 39) = 1.22, MSC = .069, were not significant, aod all interactions produced Fs less than unity.

Discussion These results indicate that rhyme-set size effects were equally apparent for rhyme pairs that were either high or low in meaning relatedness. Furthermore, specifically orienting subjects toward meaning did not change this relation. These results may mean that presenting rhyme pairs is sufficient to focus attentionai control processes on phonemic information and to produce an encoding that includes rhyme-related concepts. Alternatively, it is possible that the mere presence of the rhymes that were low in meaning relatedness was sufficient to orient subjects toward phonemic information for both types of pairs. Experiment 4 was designed to evaluate this possibility.

60

D. NELSON, M. BAJO, AND J. CANAS

Table 4 Materials Used in Experiment 3 Listl

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Set

List 2

Study context

Target

Small rhyme/meaning related

BRUNCH LOUD HOUSE ROMANCE GOOD FROWN

LUNCH CROWD MOUSE DANCE WOOD CLOWN

DINNER MOB CAT WALTZ TIMBER CIRCUS

Large rhyme/meaning related

SNEEZE FAIR SPRAIN GREED HIGH BEE

FREEZE HAIR PAIN NEED SKY FLEA

Small rhyme/meaning unrelated

FIFE LIVER SLOTH PLEDGE NOOSE EGG

Large rhyme/meaning unrelated

BRAKE STAG LINE BORE MODE PEEL

Target

Test cue

LIFE MOTH LEDGE GOOSE PEG SHIVER

WIFE CLOTH EDGE MOOSE LEG RIVER

SPOUSE FABRIC BORDER ELK CRUTCH STREAM

THAW COMB HURT WANT STAR TICK

CAKE BRAG VINE DOOR LOAD SQUEAL

BAKE FLAG WINE FLOOR ROAD STEAL

BROIL POLE CORK CEILING MAP THIEF

WIFE RIVER CLOTH EDGE MOOSE LEG

SPOUSE STREAM FABRIC BORDER ELK CRUTCH

BUNCH CLOUD BLOUSE LANCE HOOD NOUN

LUNCH CROWD MOUSE DANCE WOOD CLOWN

DINNER MOB CAT WALTZ TIMBER CIRCUS

BAKE FLAG WINE FLOOR ROAD STEAL

BROIL POLE CORK CEILING MAP THIEF

TEASE PAIR LANE REED LIE TEA

FREEZE HAIR PAIN NEED SKY FLEA

THAW COMB HURT WANT STAR TICK

Experiment 4 In Experiment 4 all of the study pairs were meaning related and half of these pairs were also rhyme related. Subjects were given intentional remembering instructions or they were incidentally oriented toward meaning, and rhyme-set size was manipulated as in previous experiments. Under the intentional orientation, rhyme-set size effects were expected when rhyme relatedness was high, but not when it was low (e.g., SNEEZEFREEZE but not ICE-FREEZE). Thisfindingsimply would replicate previous research showing that rhyme-set size effects are contingent on emphasizing rhyme (e.g., Nelson & Friedrich, 1980). Under the incidental-meaning orientation, expectations were

Table 5 Probability ofCorrect Recall as a Function ofEncoding Orientation (EO), Rhyme-Set Size, and Meaning Relatedness Rhyme-set size of target Incidental EO Intentional EO Meaning relatedness High

Low M

Rhyme

Meaning

Small

Large

Small

Large

Small

Large

.52 .64 .58

.45 .54 .50

.63 .58 .61

.51 .45 .48

.69

.52 .57 .55

.64 .67

Test cue

Study context

based on whether subjects automatically encode rhyme-related concepts whenever rhyme is present. If merely presenting rhyme pairs is sufficient to initiate such encoding, then rhymeset size effects should be apparent whenever rhyme relatedness is high. In this case, the results of this experiment simply would replicate Experiments 1-3 in showing that rhyme-set size effects are apparent whenever the context cue consists of a rhyming word. Such effects should be apparent even when all pairs in the list are meaningfully related and the encoding orientation emphasizes meaning. Alternatively, if the encoding of rhyme-related concepts requires the action of an attentional control process, rhyme-set size effects should only be found when rhyme relatedness is high and subjects are encoding information under an intentional orientation. Only in this condition would subjects be likely to attend to and notice the rhyme relation.

Method Design. The design formed a 2 X 2 X 2 factorial with encoding orientation manipulated between subjects, and with rhyme relatedness and rhyme-set size varied within subjects. Subjects. There were 36 subjects who served in the experiment with 18 serving in each orienting condition and with 9 given each list. All were taken from the same source as in the previous experiments. Materials. As in the previous experiments, two 24-item lists were constructed using both rhyme and meaning norms. Pairs high in rhyme and meaning relatedness and their associated tests cues were taken from Experiment 3 (e.g., BRUNCH LUNCH, DINNER). Pairs low in rhyme but

high in meaning relatedness were selected so that the targets still varied

61

PRIOR KNOWLEDGE

Table 6 Probability ofCorrect Recoil as a Function ofEncoding Orientation (EO), Rhyme-Set Size, andRhyme Relatedness Rhyme-set size of target

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Intentional EO

Incidental meaning EO

Rhyme relatedness

Small

Large

Small

Large

High Low

.70 .55

.47 .55

.58 .52

.56 .49

in rhyme-set size. In List 1 the context cues, targets, and test cues for targets having small rhyme sets were OVEN COOK, BAKE; FLOW RIVER, BRIDGE: DRAW PAINT, ART: HAND TOUCH, CONTACT: EVENING DUSK, DAWN; MARRY wire, SPOUSE. Those for targets having larger rhyme sets were ROBBERY STEAL, THIEF; SHIP SEA, OCEAN: BOTTLE WINE, CORK: MOP FLOOR, CEILING: PATH ROAD, MAP: PIN SEW, THIMBLE.

In List 2 the items for small and large sets, respectively, that were low in rhyme relatedness were BAG LUNCH, DINNER: SMILE LAUGH, JOKE: RODENT HOUSE, CAT; BALLET DANCE, WALTZ: STICK WOOD, TIMBER: CLOUD SOFT, PILLOW and I£E FREEZE, THAW: BLONDE HAIR, COMB: FEEL PAIN, HURT: SQUINT EYE, SIGHT: BLUE SKY, STAR: SHIP SEA, GULL.

Each list was subdivided into four sets of words representing combinations of rhyme relatedness and rhyme-set size. As before, mean rhyme-set size was equated in each relatedness condition, and it averaged 7.02 (SD = 1.56) for small rhyme sets and 25.79 (SD = 5.73) for large rhyme sets. Relative rhyme strength of the targets was also equated in all conditions and averaged. 13 (SD = .13). In addition, the meaningset size denned by the test cue and the relative strength of the cue in relation to its target was also equated. These values averaged 11.15 (SD = 3.69) and .31 (SD = .11). Finally, even though the study-context cues were included at study only to bias encoding toward meaning, their strength in relation to their targets was equated in each condition of rhyme relatedness and rhyme-set size. It averaged. 12 (SD =.11). Procedure. The procedures and instructions were identical to those used for Experiment 1.

Results The results indicated in Table 6 show that rhyme-set size influences the probability of recall but only when rhyme relatedness was high and instructions required intentional remembering. When instructions incidentally oriented subjects toward meaning, rhyme-set size appeared to have little or no effect even when rhyme relatedness was high. These trends were supported by the results of an ANOVA and subsequent tests. Although the effect of rhyme relatedness approached criterion for significance, FIX, 34) = 3.25, MSC - .043, the only significant source was the three-way interaction of Encoding Orientation X Rhyme-Set Size X Rhyme Relatedness, F( 1,34) = 5.52, MSt = .034. A Fisher's LSD of .12 indicated that the rhyme-set size effect was reliable when rhyme was emphasized and subjects were intentionally oriented but not when they were semantically oriented or when rhyme was not emphasized.

Discussion These results suggest that rhyme-related concepts are not automatically activated and encoded whenever rhyme is present.

The control process responsible for orienting subjects to particular domains of information apparently must be engaged for rhyme-set size effects to occur. When all of the pairs of the list are meaningfully related and subjects are incidentally oriented toward processing meaning, subjects apparently focus attentional processes on meaning and do not automatically encode rhyme information.

General Discussion The results of this series of experiments suggest that a familiar concept can activate related concepts, and when these related concepts are encoded and connected to the general episodic context, they interfere with memory for the item actually presented. The results replicate target-set size effects for both meaning and phonemically related information: Words having larger numbers of related concepts are generally more difficult to recall (Nelson, Bajo, & Casanueva, 1985; Nelson & Friedrich, 1980). This difficulty could emanate from either differential encoding weakness or from effects associated with retrieval inhibition. According to the encoding weakness explanation, words having larger sets naturally activate greater numbers of related concepts. Either through retroactive interference (e.g., Underwood, 1945) or through spreading activation (Anderson, 1983), the encoded representation of the target would weaken in direct proportion to the number of currently activated items in the set (Nelson, Bajo, & Casanueva, 1985). The encoding weakness hypothesis explains why target-set size effects should be found because it suggests that these targets are simply more weakly represented. However, this hypothesis suggests that these effects should be as apparent in recognition and in free recall as they are when recall is prompted with extralist cues. At present, such effects have not been found in any of the 25 experiments designed to detect them in these two tasks (Nelson, Canas, Casanueva, & Castano, 1985; Nelson et al., 1982). Thus, encoding weakness can be dismissed as a viable explanation. The more likely explanation for target-set size effects assumes that the encoding of related concepts during the study trial places greater or lesser burdens on the retrieval cue presented during test. The cued-recall task requires the production of a single response, the presented target. Because the study trial experience extends beyond the encoding of the target representation to include related representations, the task for the subject is to recover the representation that was directly experienced. Test cues are specifically related to the target representation and are frequently outside the domain of information emphasized at study, for example, as when HURT SHIRT is studied, and SHIRT is cued by SLEEVE. Hence, test cues must provide access to information specified in the cue, and, in turn, this information reactivates episodically encoded information about the target. In the example, SLEEVE must ultimately reactivate the episodically encoded representation for SHIRT, and because this representation includes HURT, FLIRT, DIRT, and so on, interfer-

ence is the result. Test cues that access targets having larger numbers of encoded concepts generate a larger pool of similar alternatives and more retrieval inhibition that is due to the retrieval of nontarget items (Raaijmakers & Shiffrin, 1981). Note that this retrievat?ility explanation for target-set size effects is similar to that offered for cue-set size effects (Nelson et al.,

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

62

D. NELSON, M. BAJO, AND ) . CANAS

1982). The difference is an emphasis on the number of encoded concepts in the case of target-set size effects as compared to the number of concepts existing in prior knowledge networks in the case of cue-set size effects. The occurrence of cue-set size effects requires a search through prior-knowledge networks that is based on information provided in the cue. In contrast, the occurrence of target-set size effects is contingent on related concepts being encoded at study and reactivated at test. In either case, larger sets are associated with lower levels of recall. This finding is analogous to fan effects (e.g., Anderson, 1983) and to cue overload effects (Watkins, 1975, 1979), except that the information generating the interference arises from prior knowledge instead of from direct laboratory experience. Earlier results and the present results suggest that the encoding of meaning-related concepts reflects the action of relatively automatic processes. Meaning-set size effects are found when subjects are asked to remember the words and when they are given the incidental tasks of rating them for pleasantness, concreteness or for number of associates (Nelson, Bajo, & Casanueva, 1985). As Experiments 1 and 2 show, these effects are also present when the targets are encoded in the presence of cues that bias subjects toward rhyme. Most important, the magnitude of these effects remains the same when the targets are encoded in a rhyme-study context and subjects are also incidentally oriented toward judging these rhyme relations. Even when subjects are strongly biased toward encoding rhyme, meaningrelated concepts are encoded and connected to the episodic context. Finally, as shown previously and in Experiment 2, these effects are fragile. Once subjects are distracted from the memory task, words having larger meaning-related sets are not any more difficult to recall than those with smaller sets. All of thesefindingsare consistent with the assumption that meaningset size effects reSect the automatic encoding of implicitly activated concepts and with the idea that unattended information is easy to disrupt. Hence, retrieval inhibition associated with the recovery of targets having relatively large meaning sets will be apparent only when attention is consistently maintained on the memory task. In contrast to meaning, the encoding of rhyme-related concepts is more clearly determined by the action of attentional control processes. Unlike with meaning, subjects are not habitually oriented toward encoding word sound, and they must be explicitly directed to rhyme to encode rhyme-related concepts. Previousfindingsindicate that rhyme-set size effects are found only when the target is encoded in the presence of a rhyming word (Nelson & Friedrich, 1980). However, as Experiment 4 indicates, merely presenting rhyme pairs is insufficient. In this experiment, all pairs were meaningfully related, and when subjects were incidentally oriented to rating meaning, rhyme-set size effects were not found even when the study context consisted of another rhyming word. Hence, the present results suggest that encoding implicitly activated rhymes requires attention to rhyme. The study context must consist of a rhyming word and subjects must attend to rhyme information for rhyme-set size effects to emerge. In fact, once these two conditions are met, the results of Experiments 1-3 suggest that further attempts to emphasize rhyme by asking subjects to rate rhyme will not increase the magnitude of the effect, nor will attempts to channel processing toward meaning decrease the

effect. In all three studies, encoding orientation had no differential influence on the magnitude of rhyme-set size effects. These manipulations failed presumably because the two critical conditions were already met. The study-context cues consisted of rhyming words, and because all pairs in the list rhymed, noticing rhyme was likely no matter what encoding orientation was encouraged by the instructions. Once subjects attend to rhyme, and rhyme-related concepts are activated and encoded, this representation has longer lasting effects than a comparable encoding of meaning-related concepts. In contrast to meaning, rhyme-set size effects are apparent even after a 24-hr period of distraction and delay. The difference is more likely to be related to differences in attentional processes than to inherent differences in the nature of the information. The encoding of rhyme-related concepts appears to require subjects to attend consciously to rhyme, whereas the encoding of meaning-related concepts appears to be the result of relatively automatic processing. Information encoded as a result of direct attention is more likely to have memorial consequences than information that is automatically activated and encoded (e.g., Balota, 1983). Hence, even though meaning-related cues were presented at test, subjects remember that the memorial activities of the previous day involved attending to rhyme. This memory serves to reactivate the implicit encoding operations concerning rhyme with the consequence that rhymeset size effects are still apparent after the long delay. Finally, one last comment about target-set size effects and a potential relation to elaborative encoding needs to be mentioned. Words with larger sets would seem to be more elaborately encoded in memory, and hence, the negative effects of larger set size would seem to contradict findings showing that more elaborative encodings often produce superior memory performance (e.g., Craik & Tulving, 1975). However, this contradiction may be more apparent than real. Prior-knowledge sets theoretically consist of related concepts that have been connected to the target through years of language experience. The concepts within a particular set are associatively and semantically similar to the target, and the more of these that are encoded and then reactivated at test, the less the relative distinctiveness of the episodic representation of the target. In contrast, laboratory-induced elaborative encodings encourage subjects to use or develop concepts that modify the encoding of the target so that its episodic representation is relatively disinct (e.g., Craik & Tulving, 1975). Hence, although these two types of elaborations produce apparently opposing effects, both may be mediated by the distinctiveness of the encoding. In this view, relative distinctiveness is a characteristic of the episodic encoding, and the assumption is that its effects are manifested through the process of search and retrieval inhibition. With larger sets, the encoding of the target is less distinct and, as subjects search, the effects of retrieval inhibition are apt to be more apparent. Although this account is speculative, it resolves the apparent contradiction infindingsby noting that encoding distinctiveness may underlie both effects. In fact, when subjects are encouraged to elaborate the encoding of the target in relation to a specific meaning cue and thus develop a distinctive encoding, meaning-set size no longer influences recall (Nelson & Friedrich, 1980).

PRIOR KNOWLEDGE

Journal of Experimental Psychology: Human Learning and Memory, 5,292-314. Anderson, J. R. (1983). A spreading activation theory of memory. JourNelson, D. L., McEvoy, C. L., & Friedrich, M. A. (1982). Extralist cuing nal of Verbal Learning and Verbal Behavior, 22,261-295. and retrieval inhibition. Journal of Experimental Psychology: LearnBalota, D. A. (1983). Automatic semantic activation and episodic meming, Memory, and Cognition, 8, 89-105. ory encoding. Journal of Verbal Learning and Verbal Behavior, 22, Raaijmakers, J. G. W., & Shiffrin, R. M. (1981). Search of associative 88-104. memory. Psychological Review, 88, 93-134. Craik, F. I. M., &Tulving, E. (1975). Depth of processing and the retenRoediger, H. L. III. (1974). Inhibiting effects of recall. Memory and Cogtion of words in episodic memory. Journal ofExperimental Psycholnition, 2, 261-269. ogy; General, 104, 268-294. Schneider, N., & Shifirin, R. M. (1977). Controlled and automatic inHasher, L., & Zacks, R, T. (1979). Automatic and effortful processes in formation processing: I. Detection, search, and attention. Psychologimemory. Journal of Experimental Psychology: General, 108, 356cal Review, 84, 1-66. 388. Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic inJoel son, J. M., & Hernnan, D. J. (1978). Properties of categories in formation processing: II. Perceptual learning, automatic attending, semantic memory. American Journal ofPsychology 91,101-114. and a general theory. Psychological Review, 84, 12 7-190. Nelson, D. L. (1981). Many are called but few are chosen: The influence Tulving, E. (1983). Elements of episodic memory. Oxford University of context on the effects of category size. In G. H. Bower (Ed.), The Press. psychology oflearning and motivation: Advances in research and theUnderwood, B. J. (1945). The effect of successive interpolations on retory(Vol 15, pp. 129-162). New York Academic Press. roactive and proactive inhibition. Psychological Monographs, 59Q, Nelson, D. L., Bajo, M. T., & Casanueva, D. (1985). Prior knowledge Whole No. 273). and memory: The influence of natural category size as a function of Watkins, M. J. (1975). Inhibition in recall with extralist "cues." Journal intention and distraction. Journal of Experimental Psychologyof Verbal Learning and Verbal Behavior, 14, 294-303. Learning, Memory, and Cognition, 11, 94-105. Watkins, M. J. (1979). Engrains as cuegrams and forgetting as cue overNelson, D. L., Canas, J., Casanueva, D. M., ACastano, D. (1985). Prior load: A cueing approach to the structure of memory. In C. R. Puff knowledge and recognition. American Journal of Psychology, 98, (Ed.), Memory organization and structure (pp. 347-372). New York: 379-397. Academic Press. Nelson, D. L., & Friedrich, M. A. (1980). Encoding and cuing sounds and senses. Journal of Experimental Psychology: Human Learning Received February 11, 1986 andMemory, 6, 717-731. Nelson, D. L., & McEvoy, C. L. (1979). Encoding context and set size. Revision received May 12,1986 •

References

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

63