Journal of Experimental Psychology: Learning, Memory, and Cognition 2011, Vol. 37, No. 5, 1236 –1242

© 2011 American Psychological Association 0278-7393/11/$12.00 DOI: 10.1037/a0023548

RESEARCH REPORT

The Interplay Between Value and Relatedness as Bases for Metacognitive Monitoring and Control: Evidence for Agenda-Based Monitoring Nicholas C. Soderstrom and David P. McCabe Colorado State University Two experiments are reported examining how value and relatedness interact to influence metacognitive monitoring and control processes. Participants studied unrelated and related word pairs, each accompanied by point values denoting how important the items were to remember. These values were presented either before or after each pair in a between-subjects design, and participants made item-by-item judgments of learning (JOLs) predicting the likelihood that each item would be remembered later. Results from Experiment 1 showed that participants used value and relatedness as cues to inform their JOLs. Interestingly, JOLs increased as a function of value even in the after condition in which value had no impact on cued recall. Participants in Experiment 2 were permitted to control study time for each item. Results showed that value and relatedness were simultaneously considered when allocating study time. These results support a cue-weighting process in which JOLs and study time allocation are based on multiple cues, which may or may not be predictive of future memory performance, and complements the agenda-based regulation model of study time (Ariel, Dunlosky, & Bailey, 2009) by providing evidence for agenda-based monitoring. Keywords: metacognition, judgments of learning, metacognitive monitoring, study time allocation, value

has yet to be examined with respect to metacognitive monitoring (i.e., memory predictions). Predictions regarding later memory performance are often studied using judgments of learning (JOLs; see Nelson, 1996). The typical JOL procedure involves asking participants to assess the likelihood, most commonly on a 0%–100% scale, that they will remember a particular item on a later test. JOLs can then be compared with test performance to determine metacognitive accuracy. A number of prior studies have shown that JOLs are sensitive to word-pair relatedness. That is, when both related pairs (e.g., Injury–Hurt) and unrelated pairs (e.g., Knee–Boat) are studied, participants give relatively higher JOLs for related pairs (e.g., Castel, McCabe, & Roediger, 2007; Dunlosky & Matvey, 2001; Rhodes & Castel, 2008). Furthermore, relatedness is a potent cue that is not easily discounted. For example, Carroll, Nelson, and Kirwan (1997) elicited JOLs from participants after they had learned related pairs to a criterion of two correct recalls and unrelated pairs to a criterion of eight correct recalls. Although the unrelated pairs were more likely to be recalled later, related pairs were still given higher JOLs.1 To our knowledge, the only published study exploring the effect of value on JOLs was reported by Koriat, Ma’ayan, and Nussinson (2006). They presented low or high values (1 vs. 3) before word

The current study examined the extent to which the importance, or value, of information has an impact on awareness of learning (i.e., metacognitive monitoring) and whether value can supersede relatedness as the basis for metacognitive judgments. Furthermore, we investigated the influence of these variables on metacognitive control (i.e., the self-regulation of study; see Nelson & Narens, 1990, for a discussion regarding the relationship between metacognitive monitoring and control). Ariel, Dunlosky, and Bailey (2009) recently developed the agenda-based regulation model of study time allocation that suggests that control processes are affected by one’s agendas, or goals. Ariel et al. showed that reward (i.e., value) superseded item difficulty as the basis for selecting items for restudy, leading to the conclusion that “learners assess task constraints prior to study and then construct an agenda that aims to efficiently achieve the current task goals within those constraints” (p. 38; see also Dunlosky, Ariel, & Thiede, in press). To date, agenda-based regulation has been demonstrated for metacognitive control (e.g., restudy choices), but it

This article was published Online First May 16, 2011. Nicholas C. Soderstrom and David P. McCabe, Department of Psychology, Colorado State University. David P. McCabe passed away unexpectedly on January 11, 2011. His death is a great loss to everyone who knew him; he will be missed dearly. The first experiment was completed in partial fulfillment of the master’s degree for Nicholas C. Soderstrom at Colorado State University. Thanks go to Matthew Rhodes for helpful comments throughout this project and to Chelsea Crouch, Kayla Phillips, and Tim Pallaoro for their efforts in data collection. Correspondence concerning this article should be addressed to Nicholas C. Soderstrom, Department of Psychology, Colorado State University, Fort Collins, CO 80523-1876. E-mail: [email protected]

1

It should be noted that the mechanism responsible for relatedness effects on JOLs is debated, as these could arise from fluency (i.e., differences in how easily the items are processed) or knowledge (i.e., an explicit theory about how relatedness affects memory). The current study is simply concerned with the fact that the effect of relatedness on JOLs is robust (e.g., Dunlosky & Matvey, 2001). 1236

AGENDA-BASED MONITORING

pairs (unrelated and related) were studied and had participants make item-by-item JOLs. Both study time and JOLs increased with value, with related pairs receiving higher JOLs than unrelated pairs at both value levels. The experiments we report also investigated the influence of value on JOLs, but our study differed from this previous research in two important ways. First, the current study included conditions in which value was presented before and after each item to control for value-based encoding strategies. Koriat et al.’s effects of value could have been mediated through encoding processes because value always occurred before the presentation of an item for study. Our after condition, therefore, is important to establish that value has a direct influence on JOLs that is independent of the influence of value on encoding. Thus, discovering whether this value-based dissociation occurs between JOLs and recall provides a critical test to establish the influence of agenda-based monitoring. Second, whereas Koriat et al. used only two value levels, six levels were used in the current study. Previous research indicates that the saliency of value as a cue increases with multiple levels (see Dunlosky & Thiede, 1998), so six levels of value may produce robust value-based effects that may even partly overshadow the effects of relatedness. The general aim of the current study was to investigate how value and relatedness interact to influence JOLs and study time allocation. Value was manipulated by pairing numbers with studied items, with higher values indicating greater value (cf. Castel, Benjamin, Craik, & Watkins, 2002; Watkins & Bloom, 1999). It is important to note that these values were orthogonally crossed with relatedness. This allowed an examination of the effects of value and relatedness on JOLs and study time, and offered a way to investigate the idea of cue-weighting, in which multiple cues are integrated in forming JOLs (e.g., Koriat, 1997). Another, more specific, aim of this study was to examine the influence of value on JOLs when value was expected to have no impact on memory performance. We anticipated that participants would erroneously predict that more important information would be better remembered than unimportant information, even when this was not the case (see Kassam, Gilbert, Swencionis, & Wilson, 2009). Such data would provide support for the idea that monitoring and control processes in part reflect agenda-based monitoring. Thus, we predicted that value would have a differential impact on memory predictions and memory performance, depending on when value was provided as a cue. Finally, we expected that both value and relatedness would influence study time allocation.

Experiment 1 In Experiment 1, we sought to determine the effects of value and relatedness on metacognitive monitoring alone. Participants studied related and unrelated word pairs, each of which were paired with a value ranging from 1 to 6, denoting how important each item was to remember later. It is important to note that the value either preceded or followed the word pair, thus allowing a determination of the direct effects of value on JOLs, independent of value-mediated encoding processes.

Method Participants. Eighty undergraduates from Colorado State University took part in the experiment and received course credit

1237

for their participation. Forty participants were randomly assigned to each of the two between-subjects conditions (i.e., before and after). Materials. Studied items included 52 word pairs, four of which (the first two and the last two) were not scored. Thus, 48 pairs were included in the final analyses. Of these pairs, half were weakly related (e.g., Leak–Boat) and half were unrelated (e.g., Prize–Beach; taken from Castel, McCabe, Roediger, & Heitman, 2007). Two study lists were created that counterbalanced the cues used for the related and unrelated pairs. Thus, all participants studied the same 48 targets, but the cue type (unrelated, related) was counterbalanced across the two lists. For a given participant, related or unrelated cues were paired with one of two sets of 24 targets, and the type of cue words used for each set was counterbalanced across participants. Two study lists were then created. Within each of the two study lists each pair was accompanied by a value ranging from 1 to 6. These values were randomly assigned to unrelated and related items, with the constraint that both unrelated and related pairs were presented with equal numbers of each value. That is, of the 24 pairs in each relatedness category, four were paired with 1, four with 2, four with 3, and so on. The presentation of all studied stimuli and the recording of responses were programmed with E-Prime 2.0 experiment software (Psychology Software Tools Inc., Pittsburgh, PA). Immediately after the study list, participants engaged in a paper-and-pencil demographic questionnaire that acted as a brief distractor. Tested items included the 48 cue words followed by a dash and a blank line indicating that the target word needed to be recalled (e.g., Leak–______). In total, four different paper-and-pencil test sheets were created for purposes of counterbalancing, and cues were randomly intermixed. Each test sheet included half of the cue words on the left side of the sheet and the other half of the cue words on the other side. Procedure. Participants were tested in small groups of up to four, seated individually. Participants were informed that they would be studying and making memory predictions for word pairs with the expectation that on a later test their memory for the second word of each pair would be tested when presented with the first word of each pair. Depending on the condition, participants were told that the value of each word pair would be presented either before or after the pair. Instructions emphasized that the goal was to maximize the total value of remembered word pairs as designated by the number accompanying the pair. In addition, participants were told that this maximizing of scores could be achieved in several ways, such as remembering the high-value words or remembering many words.2 During the study phase, word pairs accompanied with values were presented one at a time in the center of the screen. In the 2

One might draw parallels between the current methodology and that employed in a directed-forgetting paradigm in which participants are told after studying each item to either remember or forget the item (see MacLeod, 1998). Indeed, in the current after conditions, the importance of remembering an item is associated with an item after study. Unlike directed-forgetting experiments, participants in the current study were not told to forget any items, and in fact, were told that the value of each item would be counted toward their overall score.

SODERSTROM AND MCCABE

1238

before condition, a value ranging from 1 to 6 was presented for 1 s, and then the value was removed from the screen and replaced with a word pair for 4 s. The after condition was identical, but the order of the value and word pair was switched. In both conditions, following each pair–value or value–pair sequence, the participant was prompted for his or her JOL with a screen reading, “Chances you will remember? (0%–100%).” JOLs were self-paced and were typed by the participant on the computer screen. After each prediction was made, participants pressed the ENTER key, which advanced the screen to the next studied item. After JOLs had been made for all of the word pairs, participants completed a brief demographic questionnaire as a filler task (approximately 2 min), prior to the test. During the cued recall test, participants were asked to write down the target word that was presented with each cue word earlier. Participants were told that either guessing or leaving the space blank were acceptable responses if the target word could not be remembered. Once all participants completed the test, they were given debriefing forms and dismissed from the experiment.

Results F values, mean square errors (MSEs), and effect sizes are reported for all statistical tests for which the F values were greater than 1. The alpha level for statistical tests was .05 unless otherwise noted. Only significant interaction effects are reported. To reduce noise in the data, we collapsed the analyses across the lowest values (1 and 2) and the highest values (5 and 6; cf. Castel, Farb, & Craik, 2007). Furthermore, this method of collapsing the data perhaps best constitutes a comparison between low- and highvalue items. A breakdown of all observations from each value level for Experiment 1 can be found in Table 1. As shown in Figure 1, JOLs increased with both value and relatedness, and the influence of these variables was the same for the before and after conditions. To assess this finding, we analyzed the impact of value and relatedness on JOLs by conducting a 2

Figure 1. Mean judgment of learning (JOL) and percentage correct cued recall as a function of value and relatedness for the before (Panel A) and after (Panel B) conditions in Experiment 1.

(timing: before vs. after) ⫻ 2 (value: low vs. high) ⫻ 2 (relatedness: unrelated vs. related) mixed-model analysis of variance (ANOVA). A main effect of value was evident, F(1, 78) ⫽ 87.86, MSE ⫽ 155.03, ␩2p ⫽ .53, indicating that JOLs increased with

Table 1 Mean Judgment of Learning (JOL) and Percentage Correct Recall as a Function of Value and Relatedness for the Before and After Conditions in Experiment 1 Value Timing

1

2

3

4

5

6

Before JOLs Unrelated Related Recall Unrelated Related

27.0 (15.9) 54.1 (21.5)

26.5 (13.0) 54.2 (17.0)

30.4 (15.4) 59.2 (16.1)

31.8 (15.4) 67.0 (16.0)

39.2 (19.1) 62.3 (19.1)

39.9 (21.1) 70.7 (16.5)

16.9 (24.9) 64.4 (28.2)

26.3 (27.7) 53.8 (24.4)

21.9 (23.5) 53.1 (23.5)

20.0 (22.1) 58.8 (26.9)

30.6 (28.0) 49.4 (31.3)

31.3 (27.6) 73.1 (21.5)

After JOLs Unrelated Related Recall Unrelated Related Note.

31.1 (18.4) 59.5 (18.8)

32.2 (14.6) 53.4 (17.8)

34.6 (15.1) 64.5 (17.7)

36.9 (18.9) 68.1 (17.0)

45.1 (20.1) 65.4 (19.3)

49.5 (21.8) 70.3 (18.4)

30.6 (28.0) 70.6 (21.8)

34.4 (28.7) 60.6 (30.4)

25.6 (26.2) 60.6 (27.1)

29.4 (26.5) 63.1 (26.6)

31.9 (28.3) 53.1 (26.1)

31.9 (31.5) 74.4 (20.0)

Standard deviations are in parentheses.

AGENDA-BASED MONITORING

1239

only to the before condition in which participants have value information before study. Relatedness should affect study time in both conditions such that unrelated pairs are studied longer than related pairs. Finally, in the before condition, value should positively affect study time within each level of relatedness. This pattern would be consistent with the recently proposed agendabased regulation model of study time allocation (Ariel et al., 2009) because it would demonstrate that one’s goals are taken into account with respect to study behavior.

value. Furthermore, a main effect of relatedness was found, F(1, 78) ⫽ 247.76, MSE ⫽ 200.86, ␩2p ⫽ .76, indicating that JOLs were higher for related pairs than for unrelated pairs. As shown in Figure 1, value had different effects on cued recall for the before and after conditions. In particular, a 2 (timing: before vs. after) ⫻ 2 (value: low vs. high) ⫻ 2 (relatedness: unrelated vs. related) mixed-model ANOVA on recall indicated that there was no main effect of value, F(1, 78) ⫽ 2.14, MSE ⫽ 191.90, ␩2p ⫽ .03. However, there was a reliable main effect of relatedness, F(1, 78) ⫽ 210.61, MSE ⫽ 418.76, ␩2p ⫽ .73, indicating that recall for related pairs was greater than for unrelated pairs; however, these effects were qualified by a reliable Timing ⫻ Value interaction, F(1, 78) ⫽ 5.15, MSE ⫽ 191.89, ␩2p ⫽ .06. We explored this interaction by conducting separate 2 (value: low vs. high) ⫻ 2 (relatedness: unrelated vs. related) repeatedmeasures ANOVAs for the before and after conditions. For the before condition, reliable main effects of value, F(1, 39) ⫽ 6.10, MSE ⫽ 219.13, ␩2p ⫽ .14, and relatedness, F(1, 39) ⫽ 135.12, MSE ⫽ 340.32, ␩2p ⫽ .77, were evident. In the after condition, only a main effect of relatedness was found, F(1, 39) ⫽ 84.98, MSE ⫽ 497.20, ␩2p ⫽ .69; there was no main effect of value, F ⬍ 1. Thus, in the before condition both value and relatedness positively affected recall (Figure 1A), whereas in the after condition only relatedness positively affected recall (Figure 1B).

Method Participants. Sixty undergraduates from Colorado State University took part in the experiment and received course credit for their participation. There were 30 participants in each of the two conditions (i.e., before and after). Materials and procedure. Experiment 2 was identical to Experiment 1 with the exception that instead of a fixed presentation rate for each word pair, the study duration of each item in Experiment 2 was self-paced, allowing participants to study items as long as necessary. When the study time for each item was deemed sufficient, participants pressed the ENTER key to advance.

Results Similar to Experiment 1, all analyses for Experiment 2 were conducted after collapsing across the first two value levels (1 and 2) and the last two value levels (5 and 6), and nonsignificant interactions are not reported. A breakdown of all observations from each value level for Experiment 2 regarding study time allocation, JOLs, and cued recall can be found in Tables 2 and 3, respectively. As shown in Figure 2, both unrelated and high-value items received more study time in the before condition, whereas in the after condition only relatedness positively affected study time. To assess this finding, we conducted a 2 (timing: before vs. after) ⫻ 2 (value: low vs. high) ⫻ 2 (relatedness: unrelated vs. related) mixed-model ANOVA. A main effect of value was evident, F(1, 58) ⫽ 6.46, MSE ⫽ 1.45E7, ␩2p ⫽ .10, indicating that study time increased with value. Furthermore, a main effect of relatedness was found, F(1, 58) ⫽ 23.32, MSE ⫽ 9.01E7, ␩2p ⫽ .29, indicating that study time was greater for unrelated pairs than for related pairs. Finally, the Value ⫻ Timing interaction was significant, F(1, 58) ⫽ 4.04, MSE ⫽ 9,043,259, ␩2p ⫽ .07.

Discussion To summarize the results from Experiment 1: In both the before and after conditions, value and relatedness both contributed to participants’ JOLs, supporting the notion of cue-weighting (e.g., Koriat, 1997). In particular, JOLs were greatest for related and high-value items, and regardless of the timing of value, JOL patterns were the same. However, for recall, the timing of value did matter. As expected, in the before condition, recall increased as a function of both value and relatedness. In contrast, in the after condition, recall differed only as a function of relatedness (in favor of related pairs), with no impact of value on memory performance.

Experiment 2 In Experiment 2, we examined how value and relatedness affect metacognitive control, as assessed by study time allocation. We predicted that learners would study high-value items longer than they would study low-value items. Such a pattern would apply

Table 2 Mean Study Time Allocation (in ms) as a Function of Value and Relatedness for the Before and After Conditions in Experiment 2 Value Timing Before Unrelated Related After Unrelated Related Note.

1

2

3

4

5

6

6,646 (3,435) 4,484 (2,135)

6,284 (2,899) 5,862 (2,162)

6,744 (3,267) 5,574 (2,167)

7,051 (3,779) 5,688 (2,700)

7,939 (3,999) 5,156 (2,522)

6,998 (4,204) 6,500 (3,640)

6,775 (3,792) 4,816 (2,127)

6,016 (3,177) 5,929 (3,111)

6,343 (3,561) 5,182 (2,576)

6,049 (3,231) 5,145 (2,413)

7,264 (5,496) 4,524 (2,673)

5,765 (3,471) 6,395 (3,544)

Standard deviations are in parentheses.

SODERSTROM AND MCCABE

1240

Table 3 Mean Judgment of Learning (JOL) and Percentage Correct Recall as a Function of Value and Relatedness for the Before and After Conditions in Experiment 2 Value Timing

1

2

3

4

5

6

Before JOLs Unrelated Related Recall Unrelated Related

32.3 (17.6) 59.8 (19.9)

40.5 (22.0) 61.1 (19.1)

38.1 (21.7) 64.4 (17.3)

34.7 (21.2) 71.2 (15.5)

47.9 (19.8) 66.9 (18.3)

47.8 (12.0) 74.2 (23.3)

37.5 (29.9) 63.3 (30.6)

43.3 (28.6) 75.8 (30.4)

27.5 (26.6) 65.0 (25.1)

40.0 (33.2) 70.0 (24.0)

57.5 (32.3) 60.0 (22.4)

44.2 (33.9) 82.5 (19.9)

After JOLs Unrelated Related Recall Unrelated Related Note.

29.3 (17.3) 53.9 (23.2)

37.2 (20.1) 52.5 (18.8)

41.5 (16.9) 56.4 (13.1)

40.3 (21.1) 66.8 (16.1)

49.0 (19.3) 66.2 (18.2)

53.4 (22.3) 72.7 (17.4)

32.5 (27.2) 68.3 (31.4)

41.7 (24.9) 69.2 (26.8)

33.3 (26.5) 60.0 (23.3)

31.7 (27.8) 61.7 (24.3)

45.0 (29.7) 56.7 (24.5)

36.7 (28.4) 76.7 (19.6)

Standard deviations are in parentheses.

Planned comparisons were conducted to investigate the Value ⫻ Timing interaction. In particular, separate 2 (value: low vs. high) ⫻ 2 (relatedness: unrelated vs. related) repeated-measures ANOVAs were conducted for the before and after conditions. In the before condition, main effects of value, F(1, 29) ⫽ 7.80, MSE ⫽ 2.32E7, ␩2p ⫽ .21, and relatedness, F(1, 29) ⫽ 19.26, MSE ⫽ 6.02E7, ␩2p ⫽ .40, were found. For the after condition, a main effect of relatedness was again found, F(1, 29) ⫽ 6.99, MSE ⫽ 3.24E7, ␩2p ⫽ .19, but unlike the before condition, the main effect of value was not reliable, F ⬍ 1. Thus, whereas in the before condition study time increased as a function of both value and relatedness, only relatedness had an effect in the after condition. As shown in Figure 3, and consistent with Experiment 1, JOLs increased with both value and relatedness, the influence of these variables being independent of the timing of value. This was assessed by conducting a 2 (timing: before vs. after) ⫻ 2 (value: low vs. high) ⫻ 2 (relatedness: unrelated vs. related) mixed-model ANOVA. A main effect of value was evident, F(1, 58) ⫽ 47.67,

Figure 2. Study time allocation (in ms) as a function of value and relatedness for the before and after conditions in Experiment 2. Error bars indicate standard error of the means.

MSE ⫽ 11,648, ␩2p ⫽ .45, indicating that JOLs increased with value. Furthermore, a main effect of relatedness was found, F(1, 58) ⫽ 198.60, MSE ⫽ 27,018, ␩2p ⫽ .77, indicating that JOLs were higher for related pairs than for unrelated pairs. Thus, these analyses showed that, regardless of condition, JOLs increased with value and relatedness, replicating the effects found in Experiment 1. To examine the impact of value and relatedness on recall, we conducted a 2 (timing: before vs. after) ⫻ 2 (value: low vs. high) ⫻ 2 (relatedness: unrelated vs. related) mixed-model ANOVA (see Figure 3). There was no main effect of value, F(1, 58) ⫽ 3.19, MSE ⫽ 708.98, ␩2p ⫽ .05, but there was a main effect of relatedness, F(1, 58) ⫽ 171.56, MSE ⫽ 43,000, ␩2p ⫽ .75, indicating that recall for related pairs was greater than for unrelated pairs.

Discussion Results from Experiment 2 indicate that value and relatedness affected study time allocation in the before condition, such that within each level of relatedness high-value items were studied longer than low-value items. However, in the after condition, only relatedness influenced study time. These data support the agendabased regulation model because participants were able to adjust their study time when they were given different goals for study of different items. When the goal was given after study, value had no effect. Despite these differences in study time allocation as a function of value in the before and after conditions, value still influenced JOLs. Thus, control did not affect monitoring in the after condition (i.e., participants did not use study time as a cue to inform their JOLs); instead, better memory was predicted for higher value items despite value’s having no influence on cued recall. Of course, unlike participants in the after condition, those in the before condition were able to use value as a cue to inform their study behavior.

AGENDA-BASED MONITORING

Figure 3. Mean judgment of learning (JOL) and percentage correct cued recall as a function of value and relatedness for the before (Panel A) and after (Panel B) conditions in Experiment 2.

General Discussion We investigated the ways in which both value and relatedness informed metacognitive monitoring and control. In two experiments, value (ranging from 1 to 6) was orthogonally crossed with word pair relatedness and either preceded or followed each item. Results from Experiment 1 showed that, whereas JOLs increased as a function of both value and relatedness regardless of the timing of value, only in the before condition did value positively affect recall. That is, participants believed that high-value information would be better remembered than low-value information, even in the after condition in which value had no impact on recall. Experiment 2 extended these findings to a metacognitive control component—specifically, study time allocation. In the before condition, both value and relatedness influenced study time allocation, such that unrelated items were studied longer than related items were, and high-value items were studied longer than low-value items were. By contrast, in the after condition, only relatedness influenced study time allocation. Given that value information was presented after an item was studied, it did not have any influence on study time allocation. The patterns of JOLs in both experiments indicate that multiple cues—in this case, value and relatedness— can simultaneously inform JOLs. This is consistent with the idea of cue-weighting (Koriat, 1997), in which multiple factors inform JOLs. The finding that these cues operated independently, reflected by the lack of an interaction, is consistent with past research (e.g., Koriat et al.,

1241

2006). However, the current study extended this finding to conditions in which one of the cues (value) was presented after each studied item, thus allowing an examination of how value affects JOLs independent of value-mediated encoding processes. One might expect that those who were presented with value after study would not use value as a cue, whereas those who were made aware of each item’s value beforehand would use value as a cue. However, we found no difference in the influence of value on JOLs based on timing. Notably, whereas value and relatedness informed JOLs in similar ways regardless of the timing of value, recall was sensitive to value only in the before condition. Thus, a value bias in JOLs was evident such that people believe that more valuable information will be better remembered relative to less valuable information even in cases when value does not confer a memory benefit. In general, this supports Koriat’s (1997) cue utilization framework, which proposes that JOLs reflect influences based on cues that may or may not be diagnostic of future performance, and is inconsistent with the direct-access view (e.g., Cohen, Sandler, & Keglevich, 1991; Hart, 1967), which posits that JOLs and memory performance are both based on memory strength (but see Jang & Nelson, 2005). Thus, cues such as value may drive JOLs regardless of how diagnostic they are of memory performance. Value and relatedness were also shown to simultaneously influence study time allocation (Experiment 2). However, as predicted, this was observed only in the before condition in which the value of each item was known before it was studied. If value came afterward, this cue was not used to allocate study time. This finding is clearly in line with the agenda-based regulation model of study time, which places importance on, among other things, the reward structure of the task (Ariel et al., 2009). Also complementing this model is our added emphasis on value-directed predicting, a finding that might be viewed in light of agenda-based monitoring, in which people assess their task goals and monitor their learning with these goals in mind. Such agenda-based monitoring has real-world implications. Take the example of a person who just witnessed a crime and is now en route to the police station to report it. The current findings suggest that the value, or importance, of the witnessed events will influence both what this person believes will be remembered later and what is actually remembered later. However, if the eyewitness does not realize that a crime had been witnessed until after the event has taken place, the witness might overpredict how much information will be remembered about the event. For example, a person may see someone running out of a grocery store only to find out minutes later that that person had robbed the store. In other words, the importance of the event had not been assigned until after it had been witnessed, thereby preventing an influence on encoding the event. The current study with its finding of valuebiased metacognitive judgments would suggest that, in this case, the eyewitness would show inflated confidence in the ability to accurately recall the events (see Kassam et al., 2009). With the exception of a few studies (Ariel et al., 2009; Gerlach, 2008; Koriat et al., 2006), value, or importance, has not been given much consideration in metacognitive research. This is surprising given that selectively attending to important information plays a critical role in efficient memory functioning (see Castel, 2007). We hope that this study will not only demonstrate how the impact of value on metacognition might be explored but will also initiate

SODERSTROM AND MCCABE

1242

and facilitate a dialogue among memory researchers as to why value is an important avenue for future study.

References Ariel, R., Dunlosky, J., & Bailey, H. (2009). Agenda-based regulation of study-time allocation: When agendas override item-based monitoring. Journal of Experimental Psychology: General, 138, 432– 447. doi: 10.1037/a0015928 Carroll, M., Nelson, T. O., & Kirwan, A. (1997). Tradeoff of semantic relatedness and degree of overlearning: Differential effects on metamemory and on long-term memory. Acta Psychologica, 95, 239 – 253. doi:10.1016/S0001-6918(96)00040-6 Castel, A. D. (2007). The adaptive and strategic use of memory by older adults: Evaluative processing and value-directed remembering. In A. S. Benjamin & B. H. Ross (Eds.), The psychology of learning and motivation (Vol. 48, pp. 225–270). London, England: Academic Press. Castel, A. D., Benjamin, A. S., Craik, F. I. M., & Watkins, M. J. (2002). The effects of aging on selectivity and control in short-term recall. Memory & Cognition, 30, 1078 –1085. doi:10.3758/BF03194325 Castel, A. D., Farb, N. A. S., & Craik, F. I. M. (2007). Memory for general and specific value information in younger and older adults: Measuring the limits of strategic control. Memory & Cognition, 35, 689 –700. doi:10.3758/BF03193307 Castel, A. D., McCabe, D. P., & Roediger, H. L., III. (2007). Illusions of competence and overestimation of associative memory for identical items: Evidence from judgments of learning. Psychonomic Bulletin & Review, 14, 107–111. doi:10.3758/BF03194036 Castel, A. D., McCabe, D. P., Roediger, H. L., III, & Heitman, J. L. (2007). The dark side of expertise: Domain-specific memory errors. Psychological Science, 18, 3–5. doi:10.1111/j.1467-9280.2007.01838.x Cohen, R. L., Sandler, S. P., & Keglevich, L. (1991). The failure of memory monitoring in a free recall task. Canadian Journal of Psychology/Revue canadienne de psychologie, 45, 523–538. doi:10.1037/ h0084303 Dunlosky, J., Ariel, R., & Thiede, K. W. (in press). Agenda-based regulation of study-time allocation. In P. A. Higham & J. Leboe (Eds.), Constructions of remembering and metacognition: Essays in honour of Bruce Whittlesea. New York, NY: Macmillan. Dunlosky, J., & Matvey, G. (2001). Empirical analysis of the intrinsicextrinsic distinction of judgments of learning (JOLs): Effects of relatedness and serial position on JOLs. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27, 1180 –1191. doi:10.1037/ 0278-7393.27.5.1180

Dunlosky, J., & Thiede, K. W. (1998). What makes people study more? An evaluation of factors that affect self-paced study. Acta Psychologica, 98, 37–56. doi:10.1016/S0001-6918(97)00051-6 Gerlach, K. D. (2008). Value-directed remembering. Unpublished manuscript. Hart, J. T. (1967). Memory and the memory-monitoring process. Journal of Verbal Learning and Verbal Behavior, 6, 685– 691. doi:10.1016/ S0022-5371(67)80072-0 Jang, Y., & Nelson, T. O. (2005). How many dimensions underlie judgments of learning and recall? Evidence from state-trace methodology. Journal of Experimental Psychology: General, 134, 308 –326. doi: 10.1037/0096-3445.134.3.308 Kassam, K. S., Gilbert, D. T., Swencionis, J. K., & Wilson, T. D. (2009). Misconceptions of memory: The Scooter Libby effect. Psychological Science, 20, 551–552. doi:10.1111/j.1467-9280.2009.02334.x Koriat, A. (1997). Monitoring one’s own knowledge during study: A cue-utilization approach to judgments of learning. Journal of Experimental Psychology: General, 126, 349 –370. doi:10.1037/00963445.126.4.349 Koriat, A., Ma’ayan, H., & Nussinson, R. (2006). The intricate relationship between monitoring and control in metacognition: Lessons for the causeand-effect relation between subjective experience and behavior. Journal of Experimental Psychology: General, 135, 36 – 69. doi:10.1037/00963445.135.1.36 MacLeod, C. M. (1998). Directed forgetting. In J. M. Golding & C. M. MacLeod (Eds.), Intentional forgetting: Interdisciplinary approaches (pp. 1–57). Mahwah, NJ: Erlbaum. Nelson, T. O. (1996). Consciousness and metacognition. American Psychologist, 51, 102–116. doi:10.1037/0003-066X.51.2.102 Nelson, T. O., & Narens, L. (1990). Metamemory: A theoretical framework and some new findings. In G. H. Bower (Ed.), The psychology of learning and motivation (pp. 125–173). New York, NY: Academic Press. Rhodes, M. G., & Castel, A. D. (2008). Memory predictions are influenced by perceptual information: Evidence for metacognitive illusions. Journal of Experimental Psychology: General, 137, 615– 625. doi:10.1037/ a0013684 Watkins, M. J., & Bloom, L. C. (1999). Selectivity in memory: An exploration of willful control over the remembering process. Unpublished manuscript.

Received October 20, 2010 Revision received February 21, 2011 Accepted February 21, 2011 䡲