Internalizing Versus Externalizing Control: Different ... - APA PsycNET

Journal of Experimental Psychology: Learning, Memory, and Cognition 2014, Vol. 40, No. 4, 1064 –1071

© 2014 American Psychological Association 0278-7393/14/$12.00 DOI: 10.1037/a0035786

Internalizing Versus Externalizing Control: Different Ways to Perform a Time-Based Prospective Memory Task Tracy Huang

Shayne Loft

The University of Queensland

The University of Western Australia

Michael S. Humphreys 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.

The University of Queensland Time-based prospective memory (PM) refers to performing intended actions at a future time. Participants with time-based PM tasks can be slower to perform ongoing tasks (costs) than participants without PM tasks because internal control is required to maintain the PM intention or to make prospective-timing estimates. However, external control can be gained, and internal control minimized, by checking clocks or by using PM reminders. We present 3 experiments that examined how individuals externalize and internalize control of time-based PM tasks. The control condition performed a lexical decision task only, whereas the PM conditions were additionally required to make a time-based PM response after 11 min. We manipulated whether participants received a reminder, and whether clock checking was discouraged. In Experiments 1 and 3, no cost was found under standard clock check conditions. In contrast, when participants were discouraged from clock checking (Experiments 2 and 3), significant costs were found, accompanied by a decrease in clock checking. PM reminders prompted participants to check the clock, and improved PM accuracy if those reminders were expected. However, there was no evidence that participants could localize the internal or external control of the PM task to after the presentation of an expected reminder (Experiment 3). We conclude that much of the need for internal control can be transferred to the external world by performing a well-practiced task such as clock checking, which reminds participants of the PM task and reduces the internal control required to maintain the intention to perform the PM task. Keywords: prospective memory, reminder, external control, internal control

an ongoing task activity and the PM task. For example, participants may be asked to respond to the PM task by pressing a certain response key after 10 min has elapsed while performing an ongoing task.

Prospective memory (PM) refers to the processes underlying the task of remembering to perform deferred actions in the future and is crucial for our daily functioning (Kliegel, McDaniel, & Einstein, 2008). In event-based PM tasks, individuals need to remember to perform the deferred action when a cue (target) is encountered in their environment. For example, an individual may intend to pick up a loaf of bread when passing the grocery store on the way home from work. Time-based PM tasks lack this external cue to trigger the remembering of the intended action and instead require the deferred action to be made after or during a specified time. An everyday example of a time-based PM task is remembering to pick up your child from school at a specified time (e.g., at 3 p.m.). In a typical time-based PM experiment, attention is divided between

Internal and External Control Although there are several theoretical frameworks of eventbased PM (e.g., Einstein & McDaniel, 2005; Smith & Bayen, 2004), there is a comparative lack of knowledge regarding time-based PM. A common method for investigating the cognitive processes underlying PM is to measure costs exacted upon ongoing tasks. Costs refer to when participants with PM task requirements are slower to perform ongoing tasks than participants without PM tasks, and are indicative that resources have been allocated to the PM task (for a review, see Smith, Hunt, McVay, & McConnell, 2007). Some studies have demonstrated costs associated with a time-based PM task (Cook, Marsh, Clark-Foos, & Meeks, 2007; Hick, Marsh, & Cook, 2005; Marsh, Hicks, & Cook, 2006), whereas others have not reported costs (Cicogna, Nigro, Occhionero, & Esposito, 2005; Jäger & Kliegel, 2008; Occhionero, Esposito, Cicogna, & Nigro, 2010). In these studies, participants had unlimited access to a clock, and participants on average checked the clock more than once per minute. There is some preliminary evidence that costs can increase with decreased clock use. Waldum and

This article was published Online First February 17, 2014. Tracy Huang, School of Psychology, The University of Queensland, St. Lucia, Queensland, Australia; Shayne Loft, School of Psychology, The University of Western Australia, Perth, Western Australia, Australia; Michael S. Humphreys, School of Psychology, The University of Queensland, St. Lucia, Queensland, Australia. This research was supported in part by Australian Research Council Grant DP12010311, awarded to Shayne Loft. Correspondence concerning this article should be addressed to Tracy Huang, School of Psychology, The University of Queensland, St. Lucia, QLD 4072, Australia. E-mail: [email protected] 1064

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.

TIME-BASED PM CONTROL PROCESSES

Sahakyan (2013) had participants perform a lexical decision task and instructed them to make a PM response after 10 min. Participants were assigned to either a control condition, a low-consequence PM condition (no penalty for clock checks), or a high-consequence PM condition (penalty for clock checks). Clock checking was reduced and costs were greater for the highconsequence condition compared with the low-consequence condition. Furthermore, there were no significant costs for the low-consequence condition. We argue that there are two ways that individuals can control time-based PM task requirements. Costs could arise because internal control processes are required to make prospective timing estimates. Prospective timing tasks require participants to make a target response after a target duration has elapsed and have been shown to negatively impact concurrent task performance, suggesting that prospective timing is resource-demanding (Brown & Merchant, 2007; Taatgen, van Rijn, & Anderson, 2007; Zakay & Block, 1996). Zakay and Block claimed that the timing mechanisms outlined in their attentional-gate model of prospective timing likely contribute to time-based PM. When individuals attend to the passage of time, a pacemaker produces temporal pulses that are collected in an accumulator. Once the number of collected pulses matches the target response time, the internal clock signals that it is time to make the PM response. However, note that, even if prospective timing is accurate, a PM failure can occur if the PM intention is not retrieved. Thus, internal control may also be required to maintain the intention to perform the PM task. Individuals can externally control the time-based PM task by checking a clock. Prospective timing should be less crucial when clock checking is unlimited. Furthermore, there may be less need to maintain the intent to perform the PM task because by checking the clock, participants effectively remind themselves about the PM task. In line with this, there is growing evidence in the retrospective memory literature that a memory access operation (retrieving a word from a list) recovers contextual information that is then used in the next memory access operation (Humphreys, Murray, & Koh, in press; Sederberg, Howard, & Kahana, 2008). However, does the participant with unlimited clock access not have to maintain the intent to periodically check the clock? Not necessarily, as it is possible that when the clock-checking response occurs frequently and recently, it will occur again with little need to maintain an intention to do so. Moreover, clock checking may be largely driven by a cue in the environment (e.g., the visible presence of the space bar). We conducted three experiments to examine different ways individuals externalize or internalize control to a time-based PM task. In Experiment 1, participants had unlimited clock access. We were interested in whether we would observe significant costs when clock checking was unlimited and internal control minimized. In Experiment 2, we discouraged clock checking by asking participants to only check the clock when necessary in order to examine costs when internal control processes were more heavily relied on. In Experiment 3, we manipulated clock-checking discouragement. In addition, as outlined below, we examined the impact of another factor that may influence the degree of external control in time-based PM tasks: the provision of explicit reminders.

1065

Further Externalizing Control: The Provision of Reminders Kvavilashvili and Fisher (2007) asked participants to telephone the experimenter after 7 days and to record when they recollected this time-based PM task. Most recollections (45%) were triggered by external cues such as a phone, a calendar, a clock, or hearing words (e.g., forgot) associated with the experiment, suggesting that environmental cues are vital in naturalistic time-based PM retrieval. A handful of studies have demonstrated that explicit reminders (Cook, Marsh, & Hicks, 2005; Henry, Rendell, Phillips, Dunlop, & Kliegel, 2012) or linking time-based PM tasks to specific ongoing task contexts (Cook et al., 2005, 2007; Marsh et al., 2006) can improve PM accuracy. Explicit reminders have also been shown to prompt clock checking and improve PM accuracy, irrespective of whether these reminders were expected (Henry et al., 2012) or unexpected (Cook et al., 2005). Furthermore, participants show no costs and reduced clock checks if they are told that the target time for the PM response will occur in a subsequent phrase of the experiment (Cook et al., 2005, 2007; Marsh et al., 2006), thereby localizing internal and external control of the time-based PM task to the relevant ongoing task context. We examined the effect of reminders on PM accuracy and compared clock checking and costs before and after unexpected (Experiments 1 and 2) and expected (Experiment 3) reminders were provided. We were interested in the extent to which clock checking and internal control increased after reminders. In Experiment 3, we examined the extent to which participants expecting a reminder were able to localize external and internal control of the time-based PM task to the time period after the reminder, and the effect that not receiving an expected reminder had on these control processes and resulting PM accuracy.

Experiment 1 Experiment 1 consisted of three conditions: control, PM no reminder, and PM (unexpected) reminder. The control condition performed two blocks of a lexical decision task. Those in the PM condition also performed two blocks of lexical decisions; in the second block, participants were also required to make a single time-based PM response 11 min into the ongoing task. The control condition allowed us to contrast the effects of embedding the PM task in the ongoing task with the effects of performing the ongoing task twice without the PM task (Loft, Kearney, & Remington, 2008; Smith et al., 2007). Despite the associated reduction in power for the PM accuracy analyses, we used a single-response PM task because it more closely emulates time-based PM in naturalistic settings (Kvavilashvili & Fisher, 2007) and because it provides a cleaner opportunity for contrast between internal and external control. For example, if we had included multiple PM response requirements, the control processes used for the second time-based PM response (or any subsequent response) are likely to have been affected by the control processes used for earlier responses or the perceived success of those earlier responses. Participants were allowed to make as many clock checks as they deemed necessary (standard conditions). Both our theorizing regarding the external and internal control of time-based PM tasks and results from a subset of past studies (e.g., Jäger & Kliegel,

1066

HUANG, LOFT, AND HUMPHREYS

2008; Waldum & Sahakyan, 2013) suggest that we may not necessarily find costs in Experiment 1 when participants have a single-response time-based PM task with lack of constraints on clock checking. Later, we then contrast this condition with the conditions in Experiments 2 and 3 in which more emphasis was placed on internal control by discouraging clock checking. We expected the reminder to prompt clock checking and to possibly improve PM in Experiment 1.


Method Participants. One hundred ninety-five first-year psychology students (154 women) at the University of Queensland, Australia, between the ages of 17 and 48 years (M ⫽ 19.4, SD ⫽ 2.9), participated for course credit. Sixty-five participants were randomly assigned to each group. Materials. Four hundred eighty medium-frequency five-letter words were randomly chosen from the Sydney Morning Herald database (Dennis, 1995). Nonwords were created by replacing the vowels in the word with a randomly selected vowel. For example, diary was presented as deory. One hundred twenty words and 120 nonwords were assigned to List A, and the other 120 words and 120 nonwords were assigned to List B. The assignment of List A and List B to the two lexical decision blocks was counterbalanced. The presentation of stimuli within blocks was randomized. Procedure. Participants removed watches and cell phones before entering the testing room. The experiment duration was 35 min, and participants were assigned separate testing cubicles. Participants in the three conditions (control, PM no reminder, and PM reminder) had to complete a lexical decision task in Block 1. Participants were informed that a letter string would appear on screen, and they had to decide as quickly and accurately as possible whether the string was a word. Responses were made using home keys (D ⫽ nonword, K ⫽ word). The trial sequence was as follows: The first 250-ms display was a focus point marked by a symbol “⫹” displayed in white on a black background, followed by a letter string that remained on the screen until a response was detected. The intertrial interval was calculated by subtracting lexical decision response time and fixation time from 3 s (duration of each trial) (Hicks et al., 2005). After completing the first block of lexical decision trials, participants were given the lexical decision instructions again. Participants in PM conditions were given an additional instruction to press the F1 key 11 min after Block 2 had commenced. They were told they could press the space bar to check elapsed time. In Block 2, participants in the reminder condition were presented with an unexpected reminder (“Please remember to press the F1 key at the appropriate time”) 6 min into the task.

Results and Discussion An alpha level of .05 was used for all statistical tests in Experiments 1–3. We had a power of .96 to detect the cost to the ongoing lexical decision task reported by Hicks et al., (2005; d ⫽ .57) and a power of .997 to detect the cost to the ongoing lexical decision task reported by Marsh et al. (2006; d ⫽ .73). PM accuracy. PM was scored correct when a response was recorded within ⫾20 s (e.g., Hicks et al., 2005) of the 11th-min target time. A chi-square test revealed no significant difference in

PM accuracy between the reminder (50/65 participants) and no reminder (44/65 participants) conditions, ␹2(1, N ⫽ 130) ⫽ 1.38, p ⫽ .24. A further five participants in the reminder condition made early PM responses, and two made late PM responses. A further two participants made early PM responses in the no reminder condition. These early/late PM responses were considered PM failures. Ongoing task performance. Several types of trials were excluded from the ongoing task analysis: PM false alarms, trials when participants checked the clock and the two trials after, and lexical decision trials after the 11th min of both blocks. Consistent with previous research (Hicks et al., 2005; Waldum & Sahakyan, 2013), accuracy on the lexical decision task was near ceiling (95%), and there was no evidence of costs to accuracy (ts ⬍ 1). The key dependent variable was the average response time to words in the lexical decision task (Hicks et al., 2005; Waldum & Sahakyan, 2013). Responses greater than three standard deviations above a participant’s grand mean for a given block were excluded from the analysis (3.75% of responses were trimmed). Table 1 shows lexical response times to correctly responded word trials as a function of condition. There was no difference in Block 1 response time between the PM conditions and control condition, or between the reminder and no reminder conditions (both ts ⬍ 1). Difference scores were calculated by subtracting the response times of Block 1 from the response times of Block 2 (Loft et al., 2008; Smith et al., 2007). We analyzed the difference scores before and after 6 min, but there was no main effect of this time interval variable, and it did not interact with condition (Fs ⬍ 1). For simplicity, we collapsed across the time interval variable when conducting planned contrasts. The planned contrasts revealed no significant difference between the PM conditions and the control condition, t(193) ⫽ 1.70, p ⫽ .09, d ⫽ .28, or between the reminder and no reminder conditions, t(128) ⫽ 1.27, p ⫽ .21, d ⫽ .23.

Table 1 Response Times (Correct Word Trials) to the Lexical Decision Task as a Function of Reminder (Reminder vs. No Reminder) in Experiment 1 (Standard Instruction) and in Experiment 2 (Discourage Instruction), and as a Function of Reminder (Reminder vs. No Reminder) and Instruction (Standard vs. Discourage Instructions) in Experiment 3

Experiment 1 Reminder (standard) No reminder (standard) Control Experiment 2 Reminder (discourage) No reminder (discourage) Control Experiment 3 Reminder (standard) No reminder (standard) Reminder (discourage) No reminder (discourage) Control Note.

Block 1

Block 2

Difference score

605 (93) 609 (92) 608 (94)

627 (122) 611 (128) 600 (103)

22 (72) 2 (97) ⫺8 (59)

591 (94) 615 (88) 620 (105)

616 (101) 627 (96) 611 (116)

25 (42) 12 (64) ⫺9 (53)

602 (90) 607 (107) 583 (88) 595 (99) 573 (83)

605 (104) 605 (123) 608 (106) 628 (109) 570 (86)

3 (66) ⫺3 (66) 24 (50) 33 (98) ⫺3 (51)

Standard deviations appear in parentheses.



We examined whether ongoing task response times were functionally related to PM accuracy. No significant correlation was found between ongoing task response time and PM accuracy for the reminder condition, rpb(65) ⫽ .03, p ⫽ .76, or for the no reminder condition, rpb(65) ⫽ .03, p ⫽ .80. Clock checking. Participants checked the clock 22 times on average (reminder, M ⫽ 28; no reminder, M ⫽ 16), approximately two clock checks per min. This clock checking frequency is high relative to previous studies in which approximately one clock check per min has been reported (Hicks et al., 2005; Marsh et al., 2006; Waldum & Sahakyan, 2013). A 2 Condition (reminder vs. no reminder) ⫻ 2 Interval (before 6 min vs. after 6 min) mixed analysis of variance (ANOVA) was performed on clock checking during Block 2. There was a main effect of interval, indicating that more clock checks were made after 6 min, F(1, 128) ⫽ 171.11, p ⬍ .001, ␩p2 ⫽ .57. A main effect of condition indicated that participants in the reminder condition checked the clock more frequently than participants in the no reminder condition, F(1, 128) ⫽ 22.57, p ⬍ .001, ␩p2 ⫽ .15. However, these main effects were qualified by a significant interaction between condition and interval, F(1, 128) ⫽ 18.95, p ⬍ .001, ␩p2 ⫽ .12. The difference in clock checking between the first 6 min and the last 6 min was greater for the reminder condition (M ⫽ 19.20, SD ⫽ 11.58) than for the no reminder condition (M ⫽ 10.27, SD ⫽ 8.61), t(128) ⫽ 4.983, p ⬍ .001, d ⫽ .88. To examine how immediate this increased clock checking was, we examined clock checking within 20 s after the reminder was presented. A chi-square analysis showed that more participants in the reminder condition (N ⫽ 56) checked the clock during this time period than participants in the no reminder condition (N ⫽ 17), ␹2(1, N ⫽ 130) ⫽ 47.52, p ⬍ .001. No significant correlations were found between lexical decision response time and clock checking for the reminder condition, r(65) ⫽ .11, p ⫽ .36, or for the no reminder condition, r(65) ⫽ ⫺.02, p ⫽ .82. However, clock checking and PM accuracy were positively correlated for the reminder condition, rpb(65) ⫽ .42, p ⬍ .001, and the no reminder condition, rpb(65) ⫽ .63, p ⬍ .001, indicating that clock checking increased the likelihood that participants performed the PM task.

Experiment 2 Experiment 1 did not demonstrate costs for the time-based PM conditions. The first aim of Experiment 2 was to attempt to increase costs by encouraging participants to minimize clock checks. When external clock use decreases, participants should rely more on internal control to maintain the intention to make the time-based PM response, or to estimate time. The second aim was to replicate the immediate increase in clock checking caused by presenting unexpected reminders. Third, by discouraging clock checking and thereby increasing reliance on internal control, we may be more likely to find increased costs for the reminder condition compared with the no reminder condition after the reminder is presented. Finally, given the positive correlations between clock checking and PM accuracy in Experiment 1, we may find that the provision of the alternative form of external control (reminders) significantly improves PM accuracy when clock checking is minimized in Experiment 2.

1067

Method Participants. One hundred ninety-five first-year psychology students (141 women) at the University of Queensland, Australia, between the ages of 17 and 42 (M ⫽ 20.2, SD ⫽ 2.9), participated for course credit. Sixty-five participants were randomly assigned to each group. Materials and procedure. Experiment 2 was identical to Experiment 1, with two exceptions. First, participants were instructed to respond to the PM task by saying “time’s up” instead of pressing the F1 key. We decided not to use the F1 key as it may have reminded participants of the PM task due to its visible presence. The microphone used to record the PM response was hidden, and the testing cubicles were soundproof. Second, participants were instructed to remember to perform the PM task using the least number of clock checks as possible (“The goal is to say time’s up at exactly 11 min using the least number of clock checks”).

Results and Discussion PM accuracy. The PM accuracy data replicated Experiment 1. The chi-square test revealed no significant difference in PM accuracy between the reminder (50/65 participants) and no reminder (44/65 participants) conditions, ␹2(1, N ⫽ 130) ⫽ 1.38, p ⫽ .24. A further three participants in the reminder condition made early PM responses, and eight made late PM responses. A further three participants made late PM responses in the no reminder condition. Ongoing task performance. The same data exclusion and trimming techniques as in Experiment 1 were used in Experiment 2. Accuracy on the lexical decision task was near ceiling (95%), and there were no costs to accuracy (ts ⬍ 1). Table 1 presents correct lexical decision response times as a function of condition. There was no difference in Block 1 response time between the PM conditions and control condition, t(193) ⫽ 1.17, p ⫽ .24, d ⫽ .17, or between the reminder condition and no reminder condition, t(128) ⫽ 1.51, p ⫽ .13, d ⫽ .26. As in Experiment 1, we analyzed difference scores before and after 6 min, but there was no main effect of this time interval variable, and it did not interact with condition (smallest p ⫽ .18). Planned contrasts revealed that difference scores were greater for the PM conditions compared with the control condition, t(193) ⫽ 3.29, p ⬍ .001, d ⫽ .50, but there was no significant difference between the reminder and no reminder condition, t(128) ⫽ 1.35, p ⫽ .18, d ⫽ .24. Thus, we found significant cost for the PM conditions when participants were discouraged from clock checking. We conclude that participants relied more on internal control processes in Experiment 2 when clock checking was discouraged, leading to increased costs. A significant correlation was found between lexical decision response time and PM accuracy for reminder condition, rpb(65) ⫽ .25, p ⫽ .04, but not for the no reminder condition, rpb(65) ⫽ .16, p ⫽ .21. For the reminder condition at least, this suggests that that the costs in Experiment 2 represented internal control processes that were functional to time-based PM performance. Clock checking. The clock checking manipulation was effective, as there was a large reduction in clock checks (M ⫽ 5.96, SD ⫽ 5.01) compared with Experiment 1 (M ⫽ 22.33, SD ⫽ 16.41). Despite fewer clock checks, PM accuracy did not decline



1068

in Experiment 2 compared with Experiment 1, and was in fact identical. A 2 Condition (reminder vs. no reminder) ⫻ 2 Intervals (before 6 min vs. after 6 min) mixed ANOVA was performed on clock checking during Block 2. A main effect of interval indicated that more clock checks were made after 6 min (M ⫽ 4.15, SD ⫽ 3.95), compared with before 6 min (M ⫽ 1.09, SD ⫽ 1.41), F(1, 128) ⫽ 99.30, p ⬍ .001, ␩p2 ⫽ .44. There was a marginal main effect of condition; participants in the reminder condition (M ⫽ 3.00, SD ⫽ 2.38) checked the clock more frequently than participants in the no reminder condition (M ⫽ 2.23, SD ⫽ 2.92), F(1, 128) ⫽ 3.41, p ⫽ .06. However, in contrast to Experiment 1, there was no interaction between condition and interval, F(1, 128) ⫽ 1.21, p ⫽ .27, which is likely due to the low base rate of clock checking. However, as in Experiment 1, the reminder prompted immediate clock checking. More participants in the reminder condition (N ⫽ 33) checked the clock than participants in the no reminder condition (N ⫽ 12) during the 20 s after the reminder was presented, ␹2(1, N ⫽ 130) ⫽ 14.90, p ⬍ .001. As in Experiment 1, no significant correlations were found between lexical decision response time and clock checking for the reminder, r(65) ⫽ .22, p ⫽ .07, or for the no reminder condition, r(65) ⫽ ⫺.04, p ⫽ .71. Also replicating Experiment 1, clock checking and PM accuracy were positively correlated for the reminder, rpb(65) ⫽ .42, p ⬍ .001, and for the no reminder condition, rpb(65) ⫽ .45, p ⬍ .001.

Experiment 3 In Experiment 1, we found no evidence of costs to ongoing tasks under standard time-based PM conditions. In Experiment 2, we discouraged clock checking, and the PM conditions demonstrated a cost. We conducted a third experiment that directly manipulated clock checking discouragement. We expected greater costs in the discourage condition compared with the standard condition. We also included a control condition to provide another test of whether there are significant costs for participants in standard conditions with unlimited clock access. In Experiment 3, we also examined how participant’s metacognitive beliefs regarding the presentation of a reminder would influence costs and clock checking. In Experiment 3, all participants were informed as part of the instruction set that a reminder would be presented during Block 2. In addition, we manipulated whether participants actually received this reminder during Block 2. This creation of expectation may increase the likelihood that PM accuracy will be statistically higher for those who receive the reminder compared with those who do not. In addition, we examined the extent to which participants expecting a reminder localized clock checking and costs to the time period after the presentation of the reminder. Presumably, the least burdensome strategy would be to limit external and internal control of the PM task until after presentation of the expected reminder.

Method Participants. Two hundred ten first-year psychology students (176 women) at the University of Queensland, Australia, between the ages of 17 and 50 (M ⫽ 18.9, SD ⫽ 5.2), participated for

course credit. Thirty-five participants were randomly assigned to one of four PM conditions, and 70 participants were assigned to the control condition. Materials and procedure. Experiment 3 was identical to Experiment 2, with the exception that half the participants in the reminder and no reminder conditions were allocated to the standard condition (Experiment 1 instructions), and the other half were allocated to the discourage condition (Experiment 2 instructions). Participants in the PM conditions were told, “You will be reminded of the PM task during the lexical decision task. The reminder will read ‘Please remember to say time’s up at the appropriate time.’” However, only half the participants actually received this reminder.

Results and Discussion PM accuracy. The number of participants who made the PM response within ⫾20 s of the target time were as follows: no reminder-standard condition (20/35 participants, with a further one early response), no reminder-discourage condition (21/35, with a further one early and two late responses), reminder-standard condition (28/35, with a further two early and one late response), and reminder-discourage condition (24/35, with a further two early and three late responses). We ran a logistic regression, with reminder (no reminder ⫽ 0, reminder ⫽ 1) and instruction (standard ⫽ 0, discourage ⫽ 1) entered as categorical independent variables to predict the binary PM accuracy criterion (0 ⫽ fail, 1 ⫽ success). Reminder was a significant predictor of PM accuracy (OR ⫽ 3, p ⫽ .04), but neither instruction (OR ⫽ 1.13, p ⫽ .80) nor the Reminder ⫻ Instruction interaction (OR ⫽ 49, p ⫽ .33) predicted PM accuracy. Ongoing task performance. The same data exclusion and trimming techniques as those in Experiments 1 and 2 were used in Experiment 3. Accuracy of the lexical decision task was near ceiling (95%), and there was no accuracy cost (Fs ⬍ 1). Table 1 shows lexical decision response times. A one-way ANOVA revealed there were no significant differences between conditions for Block 1 response time, F(4, 209) ⫽ 1.11, p ⫽ .35, ␩p2 ⫽ .02. We analyzed the difference scores using a 2 Instruction (discourage vs. standard) ⫻ 2 Reminder (reminder vs. no reminder) ⫻ 2 Interval (before 6 min vs. after 6 min) mixed ANOVA. There was a main effect of instruction, F(1, 136) ⫽ 5.56, p ⫽ .02, ␩p2 ⫽ .04, with greater difference scores for the discourage conditions (M ⫽ ⫹29 ms) compared with standard conditions (M ⫽ 0 ms). There were no other main effects of interactions (all Fs ⬍ 1). The lack of main effect of time interval, or interaction between interval and condition, indicates that in the reminder condition, costs were not confined to the trials after the expected reminder had been presented. We then conducted planned contrasts, and as there were no effects of reminder on costs, we collapsed across this variable. Planned contrasts revealed that difference scores were greater for the PM discourage instruction conditions compared with the control condition, t(138) ⫽ 2.87, p ⫽ .005, d ⫽ .48, but there was no significant difference between PM standard conditions and the control condition (t ⬍ 1, d ⫽ .05). Thus, consistent with the results of Experiments 1 and 2, we did not find significant costs to the ongoing task when clock checking was unrestricted, but found significant costs when clock checking was discouraged.



No correlation was found between lexical decision response time and PM accuracy for the reminder, rpb(35) ⫽ ⫺.31, p ⫽ .07, or the no reminder standard conditions, rpb(35) ⫽ .16, p ⫽ .33. In the discourage conditions, no correlation was found between lexical decision response time and PM accuracy for the reminder, rpb(35) ⫽ .07, p ⫽ .65, or for the no reminder conditions, rpb(35) ⫽ .21, p ⫽ .23. Clock checking. A 2 Instruction (discourage vs. standard) ⫻ 2 Reminder (reminder vs. no reminder) ⫻ 2 Interval (before 6 min vs. after 6 min) mixed ANOVA was performed on clock checking. A main effect of interval indicated that more clock checks were made after 6 min (M ⫽ 6.38, SD ⫽ 8.07), compared with before 6 min (M ⫽ 3.25, SD ⫽ 5.59), F(1, 136) ⫽ 70.12, p ⬍ .001, ␩p2 ⫽ .34. A main effect of reminder indicated that participants in the reminder conditions (M ⫽ 6.21, SD ⫽ 7.50) checked the clock more frequently than participants in the no reminder conditions (M ⫽ 3.43, SD ⫽ 5.59), F(1, 136) ⫽ 8.88, p ⬍ .001, ␩p2 ⫽ .06. There was a main effect of instruction, F(1, 136) ⫽ 43.03, p ⬍ .001, ␩p2 ⫽ .24, with participants in the standard condition (M ⫽ 7.87, SD ⫽ 8.44) performing more clock checks than those in the discourage condition (M ⫽ 1.76, SD ⫽ 1.82), indicating that the clock checking manipulation was effective. However, these main effects were qualified by interactions between interval and instruction, F(1, 136) ⫽ 24.97, p ⬍ .001, ␩p2 ⫽ .15, between reminder and interval, F(1, 136) ⫽ 17.45, p ⬍ .001, ␩p2 ⫽ .11, and by a significant three-way interaction between reminder, interval, and instruction, F(1, 136) ⫽ 11.27, p ⬍ .001, ␩p2 ⫽ .07. Follow-up analysis showed the increase in clock checking between the first 6 min and the last 6 min was greater for the reminder condition (M ⫽ 7.82, SD ⫽ 6.47) than for the no reminder condition (M ⫽ 2.18, SD ⫽ 5.36) when using standard instructions, F(1, 68) ⫽ 15.76, p ⬍ .001, ␩p2 ⫽ .19, but not when using discourage instructions, F(1, 68) ⫽ 1.68, p ⫽ .19, ␩p2 ⫽ .02. This finding is consistent with Experiments 1 and 2, in which we found that presentation of the reminder increased clock checking for the reminder condition compared with the no reminder condition when using standard (Experiment 1), but not discourage (Experiment 2), instructions. We then examined the extent to which the reminder immediately prompted clock checking. When receiving standard instructions, more participants in the reminder (N ⫽ 28) condition checked the clock in the 20 s after the reminder was presented than participants in the no reminder (N ⫽ 9) condition, ␹2(1, N ⫽ 70) ⫽ 20.69, p ⬍ .001. Similarly, when receiving discourage instructions, more participants in the reminder (N ⫽ 19) condition checked the clock in the 20 s after the reminder was presented than participants in the no reminder condition (N ⫽ 4), ␹2(1, N ⫽ 70) ⫽ 14.57, p ⬍ .001. In the first 6 min of the ongoing task (before the reminder was presented to the reminder conditions), clock checking frequency for the standard (M ⫽ 5) and discourage (M ⫽ 1) conditions in Experiment 3 was similar to Experiment 1 (standard clock; M ⫽ 6) and Experiment 2 (discourage clock; M ⫽ 1), respectively. Therefore, there was no evidence that the participants attempted to limit checking the clock until after the expected reminder had been presented. When using standard instructions, no significant correlation was found between lexical decision response time and clock checking for the reminder condition, r(35) ⫽ .06, p ⫽ .74, but unlike

1069

Experiments 1 and 2, there was a significant correlation for the no reminder condition, r(35) ⫽ .42, p ⫽ .01. A significant correlation between lexical decision response time and clock checking was also observed when using discourage instructions for the reminder, r(35) ⫽ .33, p ⫽ .04, and the no reminder condition, r(35) ⫽ .32, p ⫽ .05. Replicating Experiments 1 and 2, clock checking and PM accuracy was significantly positively correlated for all four conditions: reminder-standard, rpb(35) ⫽ .30, p ⫽ .05; no reminderstandard, rpb (35) ⫽ .49, p ⬍ .001; discourage-reminder, rpb(35) ⫽ .46, p ⬍ .001; and discourage-no reminder, rpb(35) ⫽ .59, p ⬍ .001.

General Discussion Together, the three experiments demonstrate different ways individuals externalize and internalize control of time-based PM tasks. In Experiment 1 (standard clock check) and Experiment 3 (standard clock check), we found no cost to an ongoing lexical decision task when participants had a time-based PM task requirement, compared with a control condition. The discourage clock check conditions in Experiments 2 and 3 did demonstrate a cost, accompanied by a decrease in clock checking. Thus, when external control increased and internal control decreased (standard conditions), no cost was observed. Conversely, when internal control increased and external control decreased (discourage conditions), a cost was observed; this implies that control in a time-based PM task operates on a continuum. With reduced external control, more internal control is required to make prospective timing estimates (Taatgen et al., 2007; Zakay & Block, 1996) or to maintain the intention to perform the PM task, consuming attentional capacity. Our results also indicate that internal control was either consistently applied throughout the ongoing task or was not applied at all. That is, in Experiment 3, there was no evidence that participants waited until after the expected reminder to internally control the PM task, which would have been a rational strategy. This finding aligns with Zakay and Block’s attentional-gate model of prospective timing, which would assume that our participants needed to continuously accumulate pulses in order to accurately match them to the time-based PM target time. However, the significant number of participants who checked the clock immediately (within 20 s) after the expected reminder in Experiment 3 could have started this prospective timing process after that clock check. It may be the case that this less burdensome internal control strategy was not adopted because fewer resources are required to maintain the intent to make prospective timing estimates if they are more regularly practiced. When clock checking occurs frequently, there is less need to internally control. The likely reason for this is that after a bit of practice, clock checking becomes a response that readily occurs in the context of the experiment. We cannot tell at this stage whether or not this occurs in response to attending to something in the environment such as the space bar. Another possibility is that the resource requirements are small for a response that occurs frequently and recently. Nevertheless, it seems likely that much of the need for internal control can be transferred to the external world by performing a well-practiced task such as clock checking. More generally, it also seems likely that when the chain of associations involved in producing the PM task are strong (there is pre-


1070


experimental strength or the task has been frequently and/or recently practiced in the experimental context), the need for internal control (e.g., maintaining an intent, time estimation) will be reduced. The pattern of clock checking in Experiment 3 provides some support for this conjecture. The expectation that a reminder was going to be presented in Experiment 3 had little effect on the frequency of clock checking in the first 6 min, when compared with the first 6 min of Experiment 1 and Experiment 2. A rational strategy would be to defer clock checking until after the expected reminder. The fact that this strategy was not widely adopted may suggest that the clock checking is most reliable when it is practiced frequently. Participants may have felt the need to check the clock before the expected reminder to ensure that they continued to check the clock after the reminder. Finally, it also seems possible that there is no need to maintain an intention to respond at 11 min. That is, each time the clock is checked, it is highly likely to remind the participant that he or she also needs to respond at the 11th min. The significant correlations in Experiment 3 between costs and clock checking could be argued as problematic for our proposition that frequent clock checks can be carried out with little need for internal control. However, it is important to note that finding a positive correlation within conditions between costs and clock checking (in one experiment) is not incompatible with finding a negative correlation between costs and clock checking between conditions (across three experiments). The positive correlation within conditions may have been produced by the expectation that there would be a reminder, and maybe this influenced some participants to think or worry about the reminder. More generally, the participants who checked the clock frequently when there was little or no apparent need to check (they were told that they would be reminded) may not be representative of those who check the clock frequently when clock checking is perceived to be necessary. This selection effect could have produced the positive correlations within groups in Experiment 3 even though for most participants, the adoption of a high clock checking regime was reducing costs. We found no evidence in Experiments 1 or 2 that unexpected reminders improved PM accuracy, although care must be taken in interpreting these results given the trend toward greater accuracy with unexpected reminders and our relative lack of power to detect a change in PM (one data point per participant). However, in Experiment 3, the expected reminders did improve PM accuracy. As planned, there was more power to detect changes with the dependent measures of clock checking and costs. We found consistent evidence that clock checking increased in the period after unexpected (Experiments 1 and 2) and expected (Experiment 3) reminders were presented, compared with the time period before reminder presentation, including in the (immediate) time period 20 s after the presentation of the reminder. We also consistently found that clock checking was correlated with PM accuracy. It is important to note that in naturalistic settings, clocks are not always present, and hence it is important to understand the underlying mechanisms of time-based PM when clock checking occurs more sporadically. This notwithstanding, despite the fact that we found costs when clock checking was discouraged, self-report data from naturalistic settings indicate that individuals rarely rehearse time-based PM tasks or make prospective timing estimates, but that time-based PM tasks are more likely triggered by external cues in the environment (Kvavilashvili & Fisher, 2007). Presumably, individuals expect that they will encounter these external

cues in their day-to-day activities. In Experiment 3, PM performance was improved by presenting participants with an expected reminder. The current data also suggest that internal and external control can be equally as effective. That is, there was no effect of clock checking discouragement in Experiment 3 on PM accuracy, and PM accuracy rates in Experiment 1 (standard) and Experiment 2 (discourage) were identical. However, the high number of clock checks by participants under standard conditions, coupled with the collection of self-reports reported by Kvavilashvili and Fisher in naturalistic settings, suggests that individuals prefer to externalize the control of time-based PM tasks. In summary, we identified two modes of control in time-based prospective memory tasks: internal and external control. The current results firmly establish that there are costs involved in either maintaining the intent to perform the PM task and/or to maintaining an internal record of the passage of time with internal control. However, with external control, the costs to maintain the intent to perform the PM task, and the need to make prospective timing estimates, are clearly reduced. By drawing on the retrospective memory literature, we proposed that performing a clock check can reinstate the intent to perform the PM task just as the retrieval of an item from a study list reinstates aspects of the study context (Humphreys et al., in press). It also seems likely that wellpracticed responses such as clock checking can be performed with a minimal amount of central resources. The current results also offer some insight into how reminders and the expectation of a reminder work.

References Brown, S. W., & Merchant, S. M. (2007). Processing resources in timing and sequencing tasks. Perception & Psychophysics, 69, 439 – 449. doi: 10.3758/BF03193764 Cicogna, P. C., Nigro, G., Occhionero, M., & Esposito, M. J. (2005). Time-based prospective remembering: Interference and facilitation in a dual task. European Journal of Cognitive Psychology, 17, 221–240. doi:10.1080/09541440340000556 Cook, G. I., Marsh, R. L., Clark-Foos, A., & Meeks, J. T. (2007). Learning is impaired by activated intentions. Psychonomic Bulletin & Review, 14, 101–106. doi:10.3758/BF03194035 Cook, G. I., Marsh, R. L., & Hicks, J. L. (2005). Associating a time-based prospective memory task with an expected context can improve or impair intention completion. Applied Cognitive Psychology, 19, 345– 360. doi:10.1002/acp.1082 Dennis, S. (1995). The Sydney Morning Herald Word Database. Noetica: Open Forum, 1(4). Einstein, G. O., & McDaniel, M. A. (2005). Prospective memory: Multiple retrieval processes. Current Directions in Psychological Science, 14, 286 –290. doi:10.1111/j.0963-7214.2005.00382.x Henry, J. D., Rendell, P. G., Phillips, L. H., Dunlop, L., & Kliegel, M. (2012). Prospective memory reminders: A laboratory investigation of initiation source and age effects. Quarterly Journal of Experimental Psychology, 65, 1274 –1287. doi:10.1080/17470218.2011.651091 Hicks, J. L., Marsh, R. L., & Cook, G. I. (2005). Task interference in time-based, event-based, and dual intention prospective memory conditions. Journal of Memory and Language, 53, 430 – 444. doi:10.1016/j .jml.2005.04.001 Humphreys, M. S., Murray, K. M., & Koh, J. Y. (in press). Using a pre-existing associative relationship and other class attribute information in accessing episodic memories. Journal of Experimental Psychology: Learning, Memory, and Cognition.


TIME-BASED PM CONTROL PROCESSES Jäger, T., & Kliegel, M. (2008). Time-based and event-based prospective memory across adulthood: Underlying mechanisms and differential costs on the ongoing task. Journal of General Psychology, 135, 4 –22. doi:10.3200/GENP.135.1.4-22 Kliegel, M., McDaniel, M. A., & Einstein, G. O. (2008). Prospective memory: Cognitive, neuroscience, developmental, and applied perspectives. New York, NY: Taylor & Francis. Kvavilashvili, L., & Fisher, L. (2007). Is time-based prospective remembering mediated by self-initiated rehearsals? Role of incidental cues, ongoing activity, age, and motivation. Journal of Experimental Psychology: General, 136, 112–132. doi:10.1037/0096-3445.136.1.112 Loft, S., Kearney, R., & Remington, R. (2008). Is task interference in event-based prospective memory dependent on cue presentation? Memory & Cognition, 36, 139 –148. doi:10.3758/MC.36.1.139 Marsh, R. L., Hicks, J. L., & Cook, G. I. (2006). Task interference from prospective memories covaries with contextual associations of fulfilling them. Memory & Cognition, 34, 1037–1045. doi:10.3758/BF03193250 Occhionero, M., Esposito, M. J., Cicogna, P. C., & Nigro, G. (2010). The effects of ongoing activity on time estimation in prospective remembering. Applied Cognitive Psychology, 24, 774 –791. doi:10.1002/acp.1585 Sederberg, P. B., Howard, M. W., & Kahana, M. J. (2008). A contextbased theory of recency and contiguity in free recall. Psychological Review, 115, 893–912. doi:10.1037/a0013396

1071

Smith, R. E., & Bayen, U. J. (2004). A multinomial model of event-based prospective memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30, 756 –777. doi:10.1037/0278-7393.30.4.756 Smith, R. E., Hunt, R. R., McVay, J. C., & McConnell, M. D. (2007). The cost of event-based prospective memory: Salient target events. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33, 734 –746. doi:10.1037/0278-7393.33.4.734 Taatgen, N. A., van Rijn, H., & Anderson, J. (2007). An integrated theory of prospective time interval estimation: The role of cognition, attention, and learning. Psychological Review, 114, 577–598. doi:10.1037/0033295X.114.3.577 Waldum, E. R., & Sahakyan, L. (2013). A role for memory in prospective timing informs timing in prospective memory. Journal of Experimental Psychology: General, 142, 809 – 826. doi:10.1037/a0030113 Zakay, D., & Block, R. A. (1996). The role of attention in time estimation processes. In M. A. Pastor & J. Artieda (Eds.), Time, internal clocks and movement (pp. 143–164). Amsterdam, the Netherlands: North-Holland/ Elsevier Science. doi:10.1016/S0166-4115(96)80057-4

Received July 1, 2013 Revision received December 5, 2013 Accepted December 18, 2013 䡲

E-Mail Notification of Your Latest Issue Online! Would you like to know when the next issue of your favorite APA journal will be available online? This service is now available to you. Sign up at http://notify.apa.org/ and you will be notified by e-mail when issues of interest to you become available!