Psychology and Aging 2013, Vol. 28, No. 4. 910-922
© 2013 American Psychological Association 0882-7974/ ! 3/$ 12.00 DOI: 10.1037/aO034347
Prospective Memory and Aging: Evidence for Preserved Spontaneous Retrieval With Exact but Not Related Cues Hillary G. Mullet
Michael K. ScuUin
Ftirman University and Duke University
Furman University and Emory University School of Medicine
Theodore J. Hess
Rachel B. Scullin
Furman University
Furman University and Emory University Nell Hodgson Woodruff School of Nursing
Kathleen M. Arnold
Gilles O. Einstein
Furman University and Duke University
Furman University
We examined whether normal aging spares or compromises cue-driven spontaneous retrieval processes that support prospective remembering. In Experiment 1, young and older adults performed prospectivememory tasks that required either strategic monitoring processes for retrieval (nonfocai) or for which participants relied on spontaneous retrieval processes (focal). We found age differences for nonfocai, but not focal, prospective-memory performance. Experiments 2 and 3 used an intention-interference paradigm in which participants were asked to perform a prospective-memory task (e.g., press "Q" when the word money appears) in the context of an image-rating task and were then told to suspend their prospective-memory intention until after completing an intervening lexical-decision task. During the lexical-decision task, we presented the exact prospective-memory cue (e.g., money; Experiments 2 and 3) or a semantically related lure (e.g., wallet; Experiment 3), and we inferred spontaneous retrieval from slowed lexical-decision responses to these items relative to matched control items. Young and older adults showed significant slowing when the exact prospective-memory cue was presented. Only young adults, however, showed significant slowing to the semantically related lure items. Collectively, these results partially support the multiprocess theory prediction that aging spares spontaneous retrieval processes. Spontaneous retrieval processes may become less sensitive with aging, such that older adults are less likely to respond to cues that do not exactly match their encoded targets. Keywords: involuntary memory, elderly, monitoring, intention interference, goal fulfillment
Prospective memory is memory to perform actions in the future. In a typical laboratory prospective-memory experiment (Einstein & McDaniel, 1990), participants perform an ongoing task (e.g., rating the pleasantness of words) and are also given instructions for a prospective-memory task (e.g., press the "Q" key when the word flower appears). Because the prospective-memory cue flower
occurs in the context of the ongoing task, this word does not by nature demand a special response (i.e., participants could simply give a pleasantness rating). Thus, participants must recall without extemal prompting to suspend ongoing task responding and instead press the "Q" key (McDaniel & Einstein, 2007). Given the pervasiveness of health-related prospective-memory tasks for
Hillary G. Mullet, Department of Psychology, Furman University and Department of Psychology & Neuroscience, Duke University; Michael K. Scullin, Department of Psychology, Furman University and Department of Neurology, Emory University School of Medicine; Theodore J. Hess, Department of Psychology, Furman University; Rachel B. Scullin, Department of Psychology, Furman University and Nell Hodgson Woodruff School of Nursing, Emory University; Kathleen M. Arnold, Department of Psychology, Furman University and Department of Psychology & Neuroscience, Duke University; Gilles O. Einstein, Department of Psychology, Furman University. Portions of this research were presented at the 51st Annual Meeting of the Psychonomic Society, St. Louis, Missouri, and the 2012 Cogni-
tive Aging Conference, Atlanta, Georgia. This research was supported in part by the Furman Advantage Program. Hillary G. Mullet is supported by National Science Foundation Graduate Research Fellowship; Michael K. Scullin is supported by the National Institutes of Health and an Emory University Cottrell Fellowship and Kathleen M. Arnold is supported by NIH Institute of General Medical Sciences Grant T32GM081739-05. We appreciate the assistance of Tyler Harrison and Jamie Roberts. Correspondence concerning this article should be addressed to Hillary G. Mullet, Psychology & Neuroscience, Duke University, Box 90086, Durham, NC 27708-0086. E-mail: hillary.mullet@duke .edu 910
AGING AND SPONTANEOUS RETRIEVAL older adults (e.g., remembering to take medication), an important question is how normal aging affects prospective-memory abilities. According to Craik (1986), the high self-initiated retrieval demands of prospective-memory tasks should make these tasks especially difficult for older adults. The preparatory attentional processes and memory processes (PAM) theory (Smith & Bayen, 2006), which assumes that successful prospective remembering requires preparatory attentional processes that decline with age, makes the same prediction. Consistent with these views, young adults have outperformed older adults in many experiments (e.g., Kvavilashvili, Kombrot, Mash, Cockbum, & Milne, 2009; Maylor, 1993, 1996; Park, Hertzog, Kidder, Morrell, & Mayhom, 1997; West, Hemdon, & Covell, 2003). By contrast, in other research, age differences have been small, nonexistent, or reversed (e.g.. Cherry & LeCompte, 1999; Einstein & McDaniel, 1990; Kvavilashvili et al., 2009; Reese & Cherry, 2002; Rendell & Thomson, 1999). Several factors may contribute to these inconsistent findings, including age and ability differences across older adult samples (Kvavilashvili et al., 2009; Reese & Cherry, 2002) and ceiling effects that could obscure age differences (Uttl, 2011). Another possibility is that normal aging impairs different prospective-memory processes at different rates. According to the multiprocess theory (McDaniel & Einstein, 2000), prospectivememory retrieval can be accomplished by a resource-demanding monitoring process or by relatively automatic spontaneous retrieval processes in which cues that are strongly associated with the intended action can trigger retrieval of that intention in the absence of monitoring. From this perspective, because older adults have fewer resources available to monitor (Braver & West, 2008), prospective memory tasks that require monitoring should show age-related declines in performance. From the view that aging does not affect relatively automatic retrieval processes (e.g., Cohn, Emrich, & Moscovitch, 2008; Craik, 1986; Dywan & Jacoby, 1990; Schmitter-Edgecomb, 1999), however, older adults should be able to use spontaneous retrieval processes as effectively as young adults. Meta-analyses of the prospective-memory literature have produced mixed results regarding the multiprocess theory's predictions (Henry, MacLeod, Phillips, & Crawford, 2004; Kliegel, Jäger, & Phillips, 2008; Uttl, 2011). All of the meta-analyses agree that prospective-memory tasks that require high levels of strategic processing (e.g., nonfocal tasks) result in large age differences. However, the meta-analyses disagree with regard to whether there are age differences (Kliegel et al., 2008; Uttl, 2011) or not (Henry et al., 2004) for prospective-memory tasks that bias participants toward using more automatic processes. For example, Uttl (2011) reported large age differences with both focal and nonfocal prospective-memory cues, and he argued that his findings "strongly contradict McDaniel and Einstein's claims that prospective memory with focal cues is spared by aging due to automatic, obligatory, or reflexive retrieval (p. 9)." An important limitation of the above-mentioned meta-analyses, however, is that the included studies were not appropriately designed to provide a clear test of the multiprocess theory's predictions. A critical requirement for evaluating whether spontaneous retrieval is spared with aging is that prospective-memory retrieval must be evaluated in the absence of monitoring; otherwise, it is unclear whether one is measuring spontaneous retrieval or moni-
911
toring processes (Einstein & McDaniel, 2010). None of the studies included in the Kliegel et al. (2008) and Uttl (2011) meta-analyses met this critical requirement because the studies did not measure ongoing task costs (e.g., compromised speed on the ongoing task) when participants had a prospective-memory demand. Recent research has shown that older adults sometimes monitor more often (Horn, Bayen, & Smith, 2013; Gao, Cheung, Chan, Chu, & Lee, 2012; McDaniel, Einstein, & Rendell, 2008; Mioni & Stablum, 2013) than young adults and sometimes less often (Albiriski, Sedek, & Kliegel, 2012; Smith & Bayen, 2006). Thus, when older adults showed equivalent prospective-memory performance to young adults in previous research, this may have been because they were engaging in more monitoring, which is consistent with their generally more conscientious approach to laboratory tasks (e.g., Jackson & Balota, 2012). Conversely, when young adults showed better prospective-memory performance than older adults in previous research, this may have been due to them supplementing spontaneous retrieval with preparatory monitoring. Because of this limitation as well as other limitations (e.g., using focal and nonfocal cues that have not been shown to be equivalent in monitoring difficulty; Scullin, McDaniel, Shelton, & Lee, 2010), no studies have tested the multiprocess theory's prediction that spontaneous retrieval is spared with age in a situation in which monitoring is clearly measured and eliminated. In the present research, we used two different approaches to eliminate monitoring so as to appropriately test for spontaneous retrieval in older adults. In the flrst experiment, we compared young and older adults' performance on focal and nonfocal prospective-memory tasks, and we measured costs to the ongoing task to confirm the absence of monitoring in the focal condition (e.g.. Smith, 2003). In Experiments 2 and 3, we used the intentioninterference paradigm (Cohen, Dixon, & Lindsay, 2005; Einstein et al., 2005). In this paradigm, participants encoded a prospectivememory task within the context of an image-rating ongoing task and then were told to suspend their intention while performing an intervening lexical-decision task. Critically, the prospectivememory target cues were still presented during the lexical-decision phase, even though participants were not expecting those cues (and thus did not monitor; Knight et al., 2011; Rummel, Einstein, & Rampey, 2012). Several studies with young adults have observed slower responding to target cues than to control items during this suspended phase (Brewer, Knight, Meeks, & Marsh, 2011; Cohen et al., 2005; Cohen, Kantner, Dixon, & Lindsay, 2011; Einstein et al., 2005; Knight et al., 2011; McDaniel & Scullin, 2010; Rummel et al., 2012; Scullin, Einstein, & McDaniel, 2009), and the typical interpretation is that slowed processing of those cues reflects spontaneous noticing, discrepancy detection, or intention retrieval. Our goal was to test whether older adults show similar slowing when encountering these cues in unexpected contexts. We also tested the possibility that age differences in prospective memory may be most pronounced among the oldest participants. Past research using prospective-memory tasks that require monitoring suggests that age differences may be mainly driven by the oldest adults (old-old; 71 years and older, e.g., Kvavilashvili et al., 2009; see Shelton et al., 2011, for related evidence). Considering the possibility that age differences may emerge primarily among the old-old, we tested samples of older adults that spanned a large age range in all three experiments.
MULLET ET AL.
912 Experiment 1
We first examined the effects of normal aging on focal and nonfocal prospective-memory performance. McDaniel and Einstein (2000; see Maylor, 1996, for a similar idea) distinguished between focal processing, in which characteristics of the prospective-memory target that were processed at encoding are processed as part of the ongoing task (i.e., at retrieval), and nonfocal processing, in which the encoded characteristics are not likely to be processed at retrieval. For example, in the context of an ongoing lexical-decision task, a word-target cue (e.g., tortoise) is focal because the ongoing task directs processing of the semantic information that was processed at encoding. However, a target of any word beginning with the letter "t" is nonfocal because the ongoing task does not direct processing of first letters (Einstein & McDaniel, 2005). Using young participants, recent research has shown that processes involved in spontaneous retrieval respond to focal cues, but that monitoring is required to search for nonfocal cues (e.g., Scullin, McDaniel, & Einstein, 2010). The goal of the present experiment was to examine the effects of normal aging on spontaneous retrieval and resource-demanding monitoring, which support focal and nonfocal prospective remembering, respectively. Importantly, we tested for these processes in a situation in which we (a) discouraged monitoring by strongly emphasizing the importance of the ongoing task, and (b) used focal and nonfocal prospective-memory cues that have been shown to be equivalent in monitoring difficulty (Scullin, McDaniel, Shelton, & Lee, 2010).
Method Participants and design. The 36 young adults were Furman University undergraduates who received course credit, and the 48 older adults {n = 30 young-old and n = 18 old-old) resided in the community and received monetary compensation. Participant characteristics for all three experiments are presented in Table 1. The design was a 3 X 3 mixed factorial with the betweensubjects variable of age group (young, young-old, old-old) and the within-subjects variable of block (control, focal prospectivememory task, nonfocal prospective-memory task).' Procedure. First, participants received instructions for the lexical-decision task. A fixation cross appeared for 500 ms, followed by a letter string. Participants were told to indicate as quickly and accurately as possible whether each string represented a word by pressing either the labeled "Yes" (5) or "No" (6) key on the number pad. Each string remained on the screen until participants responded, and a response triggered the presentation of the next trial. Participants completed 10 lexical-decision practice trials and received speed and accuracy feedback. They then completed three blocks of 270 lexical-decision trials (counterbalanced across participants): a control block consisting of the lexical-decision task alone, a block with a focal prospective-memory task, and a block with a nonfocal prospective-memory task. For the control block, participants read instructions stating that they did not have a prospective-memory task (except for participants who completed the control block first). For the focal and nonfocal prospectivememory blocks, participants read that there was a secondary interest in their ability to remember to complete an intended action in the future. Participants were instructed to press the "Q" key on
the keyboard if the prospective-memory target occurred in the current block of lexical-decision trials. During the focal block, the prospective-memory target was the word channel (or printer., in the counterbalanced condition). During the nonfocal block, the target was a word that began with the letter "p" (or "c" in the counterbalanced condition), which occurred once each in the words process, printer, and package (or channel, culture, and created). Participants were told to press the "Q" key as soon as they remembered to do so, even if several trials had passed since the occurrence of the prospective-memory target. For each block, participants were reminded to "Please remember that your main goal is to respond quickly and accurately to the word judgment task. Devote your full attention to the word-judgment task." After reading instructions for each block, participants repeated them to the experimenter. Next, participants completed two filler tasks: a 30-s pattern-comparison speed-of-processing measure (Salthouse & Babcock, 1991) and a 21-item vocabulary test. The first vocabulary test (between repeating instructions and beginning Block 1) was the Mill Hill Vocabulary Scale (Raven, Raven, & Court, 1988). The other two vocabulary tests (between repeating instructions and beginning Blocks 2 and 3) were not used to collect data and were simply filler tasks. For both the focal and nonfocal prospective-memory blocks, the targets occurred on Trials 105, 185, and 265. The prospective-memory targets were widely spaced, and the onset of the first target was delayed to discourage monitoring (Loft, Kearney, & Remington, 2008). The entire procedure lasted about 60-75 min. Materials. Lexical-decision items were selected from Balota et al.'s (2007) English Lexicon Project and were five to eight letters in length. We created three unique lists, each containing 135 words and 135 nonwords. Seventy-five words (and nonwords) were presented once and 20 words (and nonwords) were repeated three times within each list. Across lists, the words were closely matched on length, frequency, and average response time (RT), and the nonwords were closely matched on length and average RT. One list was randomly assigned to each of the control, focal, and nonfocal blocks.
Results An alpha level of .05 was used for all statistical tests. Cohen's d or Ti^ was used to estimate effect size. Prospective-memory performance. A prospective-memory response was scored as correct if participants pressed the "Q" key during the presentation of the prospective-memory target or within five trials afterward (scoring the responses according to a stricter criterion produced the same pattern of results). We conducted a 3 X 2 mixed ANOVA with the between-subjects variable of age group (young, young-old, old-old) and the within-subjects variable ' Because we wanted to discourage monitoring, we used especially strong speed-emphasis instructions for half of the older participants. The only significant effect involving this manipulation was a main effect on lexical-decision RTs, F(2, 81) = 33.75, MSE = 48,678.68, r|^ = .45. The young adults were fastest, followed by the older adults who received speed-emphasis instructions, and then the older adults who received typical instructions. No other main effects and no interactions involving the speed-emphasis manipulation approached significance, and thus we collapsed speed emphasis and typical instructions across the two groups of older adults for all statistical analyses.
AGING AND SPONTANEOUS RETRIEVAL
913
Table 1
Participant Characteristics for the Three Age Groups in Experiments 1, 2, and 3 Experiment Experiment 1 Age Age Range Education % Female Vocabulary Experiment 2 Age Age Range Education % Female Vocabulary Experiment 3 Age Age Range Education % Female Vocabulary
Young
Young-old
Old-old
F
P
18.83 (0.94) 17-21 12.80(0.94) 69.4 .69 (.08)
65.40(2.81) 60-69 17.30(2.31) 66.7 .74-.78(.15)
76.72 (4.23) 71-85 16.19(2.59) 50.0 3.92-.02(.13)
3877.61
.05 and -n^s < .03). In the prospective-memory condition, young participants tended to respond more slowly to target words than to control words, r(30) = 2.03, p = .05, d = 0.74, consistent with the intention-interference effect for young adults (Cohen et al., 2005; Einstein et al., 2005). Central to the concerns of this research, older adults demonstrated a significant intentioninterference effect, with significantly slower responding to targets than to controls, i(30) = 3.25, p = .003, d = 1.19. In the recognition-memory condition, the comparison of target and control RTs did not reach statistical significance for young, i(31) = ^ After the lexical-decision task, participants completed a block of 80 experimental image-rating trials with either an embedded recognitionmemory task or a prospective-memory task. Due to experimenter error, the target words did not appear during the image-rating trials for some of the young adults in the prospective-memory condition. Because of this error, and because a postexperimental questionnaire indicated that some older adults believed that the prospective-memory task was still suspended during these trials (i.e., that they should still be withholding the prospective-memory response), we decided not to discuss this phase of the experiment in the paper. Scullin et al. (2011) found that, when this contextual confusion was removed, older adults performed similarly to young adults in a focal prospective-memory task.
916
MULLET ET AL. 800 ••ftrget DConlrol
700 600 500 400 300 200 100 0
•
• 1I
1I • •• I1 I• ^
Young
Older Prospective Meraoiy
1
Young
Older
Retrospective MemtHv
Eigure 4. Mean target and control word RTs during the lexical-decision task in Experiment 2. Error bars represent SEM.
\.l\,p = .10, i/ = 0.61, or older, i(31) = 1.38,p = .18,^ = 0.50, participants.^ Finally, we also examined RTs on the lexical-decision task for the young-old ( .18 and ti^s < .03).
is the case in the present research as Einstein et al. (2005, Experiment 5) found significantly greater slowing in a prospectivememory condition (relative to a recognition-memory condition) in young adults. Moreover, the means for the young adults were nominally in the right direction, and any evidence for a three way interaction disappeared when using a different RT trimming procedure.** What seems critical in light of the issues of the present paper is that our results indicate significant slowing to prospectivememory cues (relative to control items) in the absence of monitoring in young adults and at least as much slowing by older adults. A notable limitation of the intention-interference approach, however, is that we cannot determine whether the intention per se is retrieved (as is the case with the overt prospective-memory responses in Experiment 1). At minimum, the prospective-memory cues caused the cognitive system to react in such a way that the cues were noticed and interfered with quick performance on the lexical-decision task. Because we confirmed no monitoring during the suspended phase, to observe slowing at all, the cognitive systems in the older adults had to spontaneously notice the significance of the cues, spontaneously detect a memory-related discrepancy, or spontaneously and reflexively retrieve the associated intention. We explore this issue further in the General Discussion.
Experiment 3 Discussion This experiment was the flrst to demonstrate the intentioninterference effect in older adults in the absence of evidence for monitoring during the test phase. Experiment 2's results converged well with those of Experiment 1, and our analysis of RTs was not hmited by ceiling effects (cf Uttl, 2011). Statistically, our results indicated significantly greater slowing in the prospective-memory condition than the recognition-memory condition. Although the three-way interaction did not meet conventional levels of signiflcance, there was a trend toward a significant effect and examination of Figure 4 suggests that this effect was mainly driven by the older adults. We are unsure of why this
As part of our goal to provide a comprehensive examination of spontaneous retrieval in older adults, we investigated possible boundary conditions in Experiment 3. We again used the intention*We also examined RTs that were trimmed two standard deviations from each individual's mean, and this produced the same statistical conclusions, except that the interaction between target/control item and prospective-/recognition-memory condition was only marginally significant, F(l, 122) = 3.49, p = .06, and, in the young adult prospectivememory condition, the intention-interference effect clearly surpassed the alpha threshold, i(30) = 3.21, p = .003, d = 1.17. Also, there was no evidence of a three-way interaction with age, F{\, 122) < .01, MSE = 2,053.14,/) = .99, Ti=^ < .01.
AGING AND SPONTANEOUS RETRIEVAL interference procedure, but this time tested whether cues that were semantically related to the prospective-memory targets could also trigger spontaneous retrieval processes. Some research suggests that older adults might be as sensitive to related cues as young adults. For example, in a naturalistic prospective-memory test, Kvavilashvili and Fisher (2007) observed that young and older adults were often reminded of a prospective-memory intention when seeing a variety of cues that were related to the intention of phoning the experimenter. Another possibility arises from the involuntary autobiographical memory literature (Mace, 2007), in which researchers have found that older adults may have fewer involuntary autobiographical memory retrievals than young adults (Bemtsen & Rubin, 2002; Schlagman, Kliegel, Schulz, & Kvavilashvili, 2009; Schlagman, Kvavilashvili, & Schulz, 2007; but see Rubin & Bemtsen, 2009). Because autobiographical memory retrievals may be driven by a wide range of cues (Baars, Ramamurthy, & Franklin, 2007), the implication of these studies is that spontaneous retrieval processes may be less responsive to a diversity of cues in older adults than in young adults.
Method Participants and design. The 48 young adults were Furman University undergraduates who received monetary compensation or course credit, and the 48 community-dwelling older adults (n = 21 young-old and n = 21 old-old) received monetary compensation. Participants were tested individually or in groups of two to four. No participants had mean control trial RTs that were greater than four standard deviations from their group mean. The design of this experiment was a 2 X 2 X 2 mixed factorial design, in which we varied age (young, older) and type of target in the suspended phase (exact cue, associate lure) between subjects and lexical-decision item (target, control) within subjects. Procedure. The experimental procedure was closely modeled after Experiment 2 (see Figure 3) with a few exceptions. Unlike Experiment 2, there was only a prospective-memory condition: All participants performed an image-rating task and a lexical-decision task in which prospective-memory cues were unexpectedly presented. The prospective-memory target words were money and teeth (or water and tree in the counterbalanced condition). In this experiment, participants encoded the prospective-memory task by writing the sentence, "If I see money or teeth (or water or tree) during the image-rating task, then I will press the 'Q' key." This implementation-intention instruction (GoUwitzer, 1999) was used to ensure strong encoding of the intention. The experimenter confirmed that each participant understood the prospectivememory instructions. In this experiment, there were 312 lexical-decision trials, consisting of 144 word trials, 144 nonword trials, 12 target-word trials, and 12 control-word trials. Participants were randomly assigned to either the exact cue or the associate lure condition. In the exact cue condition, each target and control word was presented six times during the lexical-decision task. In the associate lure condition, we selected three words that were forwardly associated (Nelson, McEvoy, & Schreiber, 1998) with each target (or control) word and presented these words two times each. In both conditions, word repetitions were always separated by more than 11 trials and the
917
target- and control-word sets and positions were counterbalanced across participants. The entire procedure lasted 30-45 min. Materials. Lexical-decision task filler words and nonwords (nontargets and noncontrols) were selected from the Balota et al. (2002) norms and their order was randomized for each participant, with the stipulation that no more than three words (or nonwords) could occur in succession. The words were medium frequency, four to nine letters, and ranged in mean RTs from 600-750 ms. The nonwords were matched on length and RTs. No filler words were forwardly associated with target or control words (Nelson et al., 1998). Seventy-two items (36 words and nonwords) appeared twice each and 24 items (12 words/nonwords) appeared six times each, thereby matching the frequency of occurrence of the associate lures, exact cues, and matched control items. Thus, any increase or decrease in the speed of responding to target and control items could not be attributed to their presentation frequency. The associated lures for money v/eie profit, wallet, and fund; for teeth: gums, braces, and dentist; for tree: sap, leaf, and forest; and for water: faucet, splash, and fountain. The mean forward associations ranged from .60 to .65 (Nelson et al., 1998). Money served as the control for tree, and teeth served as the control for water, and vice versa. Each target-control pair (and associates of these pairs) was matched on mean lexical-decision RT and forward associative strength (Nelson et al., 1998).
Results As in Experiment 2, we compared mean correct RTs to target and control items in order to test for intention interference (see Figure 5). We conducted a 2 X 2 X 2 mixed ANOVA that included the between-subjects variables of age (young, older) and cue condition (exact cue, associative lure) and the within-subjects variable of item (control, target). Older adults (M = 735 ms. So = 111.11) were slower than young adults (M = 545 ms, SD = 111.67), F(l, 92) = 40.31, MSE = 43,050.32,p < .001, TI^ = .30. Consistent with the results of Experiment 2, participants were significantly slower in their lexical-decision responses to target items {M = 657 ms, SD = 195.96) relative to control items (M =
• Target
900
• CODlIOl
800 700
s 600
400 300 200 100
Young
Older Exact
Young
Older Associate
Figure 5. Mean target and control word RTs during the lexical-decision task in Experiment 3. Error bars represent SEM.
918
MULLET ET AL.
622 ms, SD = 164.24), F(l, 92) = 17.77, MSE = 3,338.63, p < .001, T|^ = .15. This main effect was qualified by a significant two-way interaction with cue condition, F(l, 92) = 4.30, MSE = 3,338.63, p = .04, r\^ = .04, and a marginally significant (p = .054) three-way interaction with cue condition and age group, F( 1, 92) = 3.81, MSE = 3,338.63, ti^ = .03.' As in Experiment 2, to determine the source of these interaction effects, we conducted separate f tests comparing target and control RTs for each of the four between-subjects conditions. Replicating Experiment 2, for the exact cue condition, there was significant slowing to target items relative to control items for both young adults (AÍT^get = 521 ms, SD = 121.02, Mco„„o, = 488 ms, SD = 65.55), /(23) = 2.10, p = .047, d = 0.88, and older adults (^Target = 761 ms, SD = 206.96, Mcontroi = 689 ms, SD = 146.90), i(23) = 3.40, p < .01, d = 1.42. For the associative lure condition, there was significant slowing to target items relative to control words for young adults (MT^rget = 600 ms, SD = 136.35, Mcontroi = 570 ms, SD = 108.93), i(23) = 2.21,p = .04, d = 0.92, but not for older adults {MT-^^^, = 747 ms, SD = 198.46, ^Control = 742 ms, SD = 170.89), t
© 2013 American Psychological Association 0882-7974/ ! 3/$ 12.00 DOI: 10.1037/aO034347
Prospective Memory and Aging: Evidence for Preserved Spontaneous Retrieval With Exact but Not Related Cues Hillary G. Mullet
Michael K. ScuUin
Ftirman University and Duke University
Furman University and Emory University School of Medicine
Theodore J. Hess
Rachel B. Scullin
Furman University
Furman University and Emory University Nell Hodgson Woodruff School of Nursing
Kathleen M. Arnold
Gilles O. Einstein
Furman University and Duke University
Furman University
We examined whether normal aging spares or compromises cue-driven spontaneous retrieval processes that support prospective remembering. In Experiment 1, young and older adults performed prospectivememory tasks that required either strategic monitoring processes for retrieval (nonfocai) or for which participants relied on spontaneous retrieval processes (focal). We found age differences for nonfocai, but not focal, prospective-memory performance. Experiments 2 and 3 used an intention-interference paradigm in which participants were asked to perform a prospective-memory task (e.g., press "Q" when the word money appears) in the context of an image-rating task and were then told to suspend their prospective-memory intention until after completing an intervening lexical-decision task. During the lexical-decision task, we presented the exact prospective-memory cue (e.g., money; Experiments 2 and 3) or a semantically related lure (e.g., wallet; Experiment 3), and we inferred spontaneous retrieval from slowed lexical-decision responses to these items relative to matched control items. Young and older adults showed significant slowing when the exact prospective-memory cue was presented. Only young adults, however, showed significant slowing to the semantically related lure items. Collectively, these results partially support the multiprocess theory prediction that aging spares spontaneous retrieval processes. Spontaneous retrieval processes may become less sensitive with aging, such that older adults are less likely to respond to cues that do not exactly match their encoded targets. Keywords: involuntary memory, elderly, monitoring, intention interference, goal fulfillment
Prospective memory is memory to perform actions in the future. In a typical laboratory prospective-memory experiment (Einstein & McDaniel, 1990), participants perform an ongoing task (e.g., rating the pleasantness of words) and are also given instructions for a prospective-memory task (e.g., press the "Q" key when the word flower appears). Because the prospective-memory cue flower
occurs in the context of the ongoing task, this word does not by nature demand a special response (i.e., participants could simply give a pleasantness rating). Thus, participants must recall without extemal prompting to suspend ongoing task responding and instead press the "Q" key (McDaniel & Einstein, 2007). Given the pervasiveness of health-related prospective-memory tasks for
Hillary G. Mullet, Department of Psychology, Furman University and Department of Psychology & Neuroscience, Duke University; Michael K. Scullin, Department of Psychology, Furman University and Department of Neurology, Emory University School of Medicine; Theodore J. Hess, Department of Psychology, Furman University; Rachel B. Scullin, Department of Psychology, Furman University and Nell Hodgson Woodruff School of Nursing, Emory University; Kathleen M. Arnold, Department of Psychology, Furman University and Department of Psychology & Neuroscience, Duke University; Gilles O. Einstein, Department of Psychology, Furman University. Portions of this research were presented at the 51st Annual Meeting of the Psychonomic Society, St. Louis, Missouri, and the 2012 Cogni-
tive Aging Conference, Atlanta, Georgia. This research was supported in part by the Furman Advantage Program. Hillary G. Mullet is supported by National Science Foundation Graduate Research Fellowship; Michael K. Scullin is supported by the National Institutes of Health and an Emory University Cottrell Fellowship and Kathleen M. Arnold is supported by NIH Institute of General Medical Sciences Grant T32GM081739-05. We appreciate the assistance of Tyler Harrison and Jamie Roberts. Correspondence concerning this article should be addressed to Hillary G. Mullet, Psychology & Neuroscience, Duke University, Box 90086, Durham, NC 27708-0086. E-mail: hillary.mullet@duke .edu 910
AGING AND SPONTANEOUS RETRIEVAL older adults (e.g., remembering to take medication), an important question is how normal aging affects prospective-memory abilities. According to Craik (1986), the high self-initiated retrieval demands of prospective-memory tasks should make these tasks especially difficult for older adults. The preparatory attentional processes and memory processes (PAM) theory (Smith & Bayen, 2006), which assumes that successful prospective remembering requires preparatory attentional processes that decline with age, makes the same prediction. Consistent with these views, young adults have outperformed older adults in many experiments (e.g., Kvavilashvili, Kombrot, Mash, Cockbum, & Milne, 2009; Maylor, 1993, 1996; Park, Hertzog, Kidder, Morrell, & Mayhom, 1997; West, Hemdon, & Covell, 2003). By contrast, in other research, age differences have been small, nonexistent, or reversed (e.g.. Cherry & LeCompte, 1999; Einstein & McDaniel, 1990; Kvavilashvili et al., 2009; Reese & Cherry, 2002; Rendell & Thomson, 1999). Several factors may contribute to these inconsistent findings, including age and ability differences across older adult samples (Kvavilashvili et al., 2009; Reese & Cherry, 2002) and ceiling effects that could obscure age differences (Uttl, 2011). Another possibility is that normal aging impairs different prospective-memory processes at different rates. According to the multiprocess theory (McDaniel & Einstein, 2000), prospectivememory retrieval can be accomplished by a resource-demanding monitoring process or by relatively automatic spontaneous retrieval processes in which cues that are strongly associated with the intended action can trigger retrieval of that intention in the absence of monitoring. From this perspective, because older adults have fewer resources available to monitor (Braver & West, 2008), prospective memory tasks that require monitoring should show age-related declines in performance. From the view that aging does not affect relatively automatic retrieval processes (e.g., Cohn, Emrich, & Moscovitch, 2008; Craik, 1986; Dywan & Jacoby, 1990; Schmitter-Edgecomb, 1999), however, older adults should be able to use spontaneous retrieval processes as effectively as young adults. Meta-analyses of the prospective-memory literature have produced mixed results regarding the multiprocess theory's predictions (Henry, MacLeod, Phillips, & Crawford, 2004; Kliegel, Jäger, & Phillips, 2008; Uttl, 2011). All of the meta-analyses agree that prospective-memory tasks that require high levels of strategic processing (e.g., nonfocal tasks) result in large age differences. However, the meta-analyses disagree with regard to whether there are age differences (Kliegel et al., 2008; Uttl, 2011) or not (Henry et al., 2004) for prospective-memory tasks that bias participants toward using more automatic processes. For example, Uttl (2011) reported large age differences with both focal and nonfocal prospective-memory cues, and he argued that his findings "strongly contradict McDaniel and Einstein's claims that prospective memory with focal cues is spared by aging due to automatic, obligatory, or reflexive retrieval (p. 9)." An important limitation of the above-mentioned meta-analyses, however, is that the included studies were not appropriately designed to provide a clear test of the multiprocess theory's predictions. A critical requirement for evaluating whether spontaneous retrieval is spared with aging is that prospective-memory retrieval must be evaluated in the absence of monitoring; otherwise, it is unclear whether one is measuring spontaneous retrieval or moni-
911
toring processes (Einstein & McDaniel, 2010). None of the studies included in the Kliegel et al. (2008) and Uttl (2011) meta-analyses met this critical requirement because the studies did not measure ongoing task costs (e.g., compromised speed on the ongoing task) when participants had a prospective-memory demand. Recent research has shown that older adults sometimes monitor more often (Horn, Bayen, & Smith, 2013; Gao, Cheung, Chan, Chu, & Lee, 2012; McDaniel, Einstein, & Rendell, 2008; Mioni & Stablum, 2013) than young adults and sometimes less often (Albiriski, Sedek, & Kliegel, 2012; Smith & Bayen, 2006). Thus, when older adults showed equivalent prospective-memory performance to young adults in previous research, this may have been because they were engaging in more monitoring, which is consistent with their generally more conscientious approach to laboratory tasks (e.g., Jackson & Balota, 2012). Conversely, when young adults showed better prospective-memory performance than older adults in previous research, this may have been due to them supplementing spontaneous retrieval with preparatory monitoring. Because of this limitation as well as other limitations (e.g., using focal and nonfocal cues that have not been shown to be equivalent in monitoring difficulty; Scullin, McDaniel, Shelton, & Lee, 2010), no studies have tested the multiprocess theory's prediction that spontaneous retrieval is spared with age in a situation in which monitoring is clearly measured and eliminated. In the present research, we used two different approaches to eliminate monitoring so as to appropriately test for spontaneous retrieval in older adults. In the flrst experiment, we compared young and older adults' performance on focal and nonfocal prospective-memory tasks, and we measured costs to the ongoing task to confirm the absence of monitoring in the focal condition (e.g.. Smith, 2003). In Experiments 2 and 3, we used the intentioninterference paradigm (Cohen, Dixon, & Lindsay, 2005; Einstein et al., 2005). In this paradigm, participants encoded a prospectivememory task within the context of an image-rating ongoing task and then were told to suspend their intention while performing an intervening lexical-decision task. Critically, the prospectivememory target cues were still presented during the lexical-decision phase, even though participants were not expecting those cues (and thus did not monitor; Knight et al., 2011; Rummel, Einstein, & Rampey, 2012). Several studies with young adults have observed slower responding to target cues than to control items during this suspended phase (Brewer, Knight, Meeks, & Marsh, 2011; Cohen et al., 2005; Cohen, Kantner, Dixon, & Lindsay, 2011; Einstein et al., 2005; Knight et al., 2011; McDaniel & Scullin, 2010; Rummel et al., 2012; Scullin, Einstein, & McDaniel, 2009), and the typical interpretation is that slowed processing of those cues reflects spontaneous noticing, discrepancy detection, or intention retrieval. Our goal was to test whether older adults show similar slowing when encountering these cues in unexpected contexts. We also tested the possibility that age differences in prospective memory may be most pronounced among the oldest participants. Past research using prospective-memory tasks that require monitoring suggests that age differences may be mainly driven by the oldest adults (old-old; 71 years and older, e.g., Kvavilashvili et al., 2009; see Shelton et al., 2011, for related evidence). Considering the possibility that age differences may emerge primarily among the old-old, we tested samples of older adults that spanned a large age range in all three experiments.
MULLET ET AL.
912 Experiment 1
We first examined the effects of normal aging on focal and nonfocal prospective-memory performance. McDaniel and Einstein (2000; see Maylor, 1996, for a similar idea) distinguished between focal processing, in which characteristics of the prospective-memory target that were processed at encoding are processed as part of the ongoing task (i.e., at retrieval), and nonfocal processing, in which the encoded characteristics are not likely to be processed at retrieval. For example, in the context of an ongoing lexical-decision task, a word-target cue (e.g., tortoise) is focal because the ongoing task directs processing of the semantic information that was processed at encoding. However, a target of any word beginning with the letter "t" is nonfocal because the ongoing task does not direct processing of first letters (Einstein & McDaniel, 2005). Using young participants, recent research has shown that processes involved in spontaneous retrieval respond to focal cues, but that monitoring is required to search for nonfocal cues (e.g., Scullin, McDaniel, & Einstein, 2010). The goal of the present experiment was to examine the effects of normal aging on spontaneous retrieval and resource-demanding monitoring, which support focal and nonfocal prospective remembering, respectively. Importantly, we tested for these processes in a situation in which we (a) discouraged monitoring by strongly emphasizing the importance of the ongoing task, and (b) used focal and nonfocal prospective-memory cues that have been shown to be equivalent in monitoring difficulty (Scullin, McDaniel, Shelton, & Lee, 2010).
Method Participants and design. The 36 young adults were Furman University undergraduates who received course credit, and the 48 older adults {n = 30 young-old and n = 18 old-old) resided in the community and received monetary compensation. Participant characteristics for all three experiments are presented in Table 1. The design was a 3 X 3 mixed factorial with the betweensubjects variable of age group (young, young-old, old-old) and the within-subjects variable of block (control, focal prospectivememory task, nonfocal prospective-memory task).' Procedure. First, participants received instructions for the lexical-decision task. A fixation cross appeared for 500 ms, followed by a letter string. Participants were told to indicate as quickly and accurately as possible whether each string represented a word by pressing either the labeled "Yes" (5) or "No" (6) key on the number pad. Each string remained on the screen until participants responded, and a response triggered the presentation of the next trial. Participants completed 10 lexical-decision practice trials and received speed and accuracy feedback. They then completed three blocks of 270 lexical-decision trials (counterbalanced across participants): a control block consisting of the lexical-decision task alone, a block with a focal prospective-memory task, and a block with a nonfocal prospective-memory task. For the control block, participants read instructions stating that they did not have a prospective-memory task (except for participants who completed the control block first). For the focal and nonfocal prospectivememory blocks, participants read that there was a secondary interest in their ability to remember to complete an intended action in the future. Participants were instructed to press the "Q" key on
the keyboard if the prospective-memory target occurred in the current block of lexical-decision trials. During the focal block, the prospective-memory target was the word channel (or printer., in the counterbalanced condition). During the nonfocal block, the target was a word that began with the letter "p" (or "c" in the counterbalanced condition), which occurred once each in the words process, printer, and package (or channel, culture, and created). Participants were told to press the "Q" key as soon as they remembered to do so, even if several trials had passed since the occurrence of the prospective-memory target. For each block, participants were reminded to "Please remember that your main goal is to respond quickly and accurately to the word judgment task. Devote your full attention to the word-judgment task." After reading instructions for each block, participants repeated them to the experimenter. Next, participants completed two filler tasks: a 30-s pattern-comparison speed-of-processing measure (Salthouse & Babcock, 1991) and a 21-item vocabulary test. The first vocabulary test (between repeating instructions and beginning Block 1) was the Mill Hill Vocabulary Scale (Raven, Raven, & Court, 1988). The other two vocabulary tests (between repeating instructions and beginning Blocks 2 and 3) were not used to collect data and were simply filler tasks. For both the focal and nonfocal prospective-memory blocks, the targets occurred on Trials 105, 185, and 265. The prospective-memory targets were widely spaced, and the onset of the first target was delayed to discourage monitoring (Loft, Kearney, & Remington, 2008). The entire procedure lasted about 60-75 min. Materials. Lexical-decision items were selected from Balota et al.'s (2007) English Lexicon Project and were five to eight letters in length. We created three unique lists, each containing 135 words and 135 nonwords. Seventy-five words (and nonwords) were presented once and 20 words (and nonwords) were repeated three times within each list. Across lists, the words were closely matched on length, frequency, and average response time (RT), and the nonwords were closely matched on length and average RT. One list was randomly assigned to each of the control, focal, and nonfocal blocks.
Results An alpha level of .05 was used for all statistical tests. Cohen's d or Ti^ was used to estimate effect size. Prospective-memory performance. A prospective-memory response was scored as correct if participants pressed the "Q" key during the presentation of the prospective-memory target or within five trials afterward (scoring the responses according to a stricter criterion produced the same pattern of results). We conducted a 3 X 2 mixed ANOVA with the between-subjects variable of age group (young, young-old, old-old) and the within-subjects variable ' Because we wanted to discourage monitoring, we used especially strong speed-emphasis instructions for half of the older participants. The only significant effect involving this manipulation was a main effect on lexical-decision RTs, F(2, 81) = 33.75, MSE = 48,678.68, r|^ = .45. The young adults were fastest, followed by the older adults who received speed-emphasis instructions, and then the older adults who received typical instructions. No other main effects and no interactions involving the speed-emphasis manipulation approached significance, and thus we collapsed speed emphasis and typical instructions across the two groups of older adults for all statistical analyses.
AGING AND SPONTANEOUS RETRIEVAL
913
Table 1
Participant Characteristics for the Three Age Groups in Experiments 1, 2, and 3 Experiment Experiment 1 Age Age Range Education % Female Vocabulary Experiment 2 Age Age Range Education % Female Vocabulary Experiment 3 Age Age Range Education % Female Vocabulary
Young
Young-old
Old-old
F
P
18.83 (0.94) 17-21 12.80(0.94) 69.4 .69 (.08)
65.40(2.81) 60-69 17.30(2.31) 66.7 .74-.78(.15)
76.72 (4.23) 71-85 16.19(2.59) 50.0 3.92-.02(.13)
3877.61
.05 and -n^s < .03). In the prospective-memory condition, young participants tended to respond more slowly to target words than to control words, r(30) = 2.03, p = .05, d = 0.74, consistent with the intention-interference effect for young adults (Cohen et al., 2005; Einstein et al., 2005). Central to the concerns of this research, older adults demonstrated a significant intentioninterference effect, with significantly slower responding to targets than to controls, i(30) = 3.25, p = .003, d = 1.19. In the recognition-memory condition, the comparison of target and control RTs did not reach statistical significance for young, i(31) = ^ After the lexical-decision task, participants completed a block of 80 experimental image-rating trials with either an embedded recognitionmemory task or a prospective-memory task. Due to experimenter error, the target words did not appear during the image-rating trials for some of the young adults in the prospective-memory condition. Because of this error, and because a postexperimental questionnaire indicated that some older adults believed that the prospective-memory task was still suspended during these trials (i.e., that they should still be withholding the prospective-memory response), we decided not to discuss this phase of the experiment in the paper. Scullin et al. (2011) found that, when this contextual confusion was removed, older adults performed similarly to young adults in a focal prospective-memory task.
916
MULLET ET AL. 800 ••ftrget DConlrol
700 600 500 400 300 200 100 0
•
• 1I
1I • •• I1 I• ^
Young
Older Prospective Meraoiy
1
Young
Older
Retrospective MemtHv
Eigure 4. Mean target and control word RTs during the lexical-decision task in Experiment 2. Error bars represent SEM.
\.l\,p = .10, i/ = 0.61, or older, i(31) = 1.38,p = .18,^ = 0.50, participants.^ Finally, we also examined RTs on the lexical-decision task for the young-old ( .18 and ti^s < .03).
is the case in the present research as Einstein et al. (2005, Experiment 5) found significantly greater slowing in a prospectivememory condition (relative to a recognition-memory condition) in young adults. Moreover, the means for the young adults were nominally in the right direction, and any evidence for a three way interaction disappeared when using a different RT trimming procedure.** What seems critical in light of the issues of the present paper is that our results indicate significant slowing to prospectivememory cues (relative to control items) in the absence of monitoring in young adults and at least as much slowing by older adults. A notable limitation of the intention-interference approach, however, is that we cannot determine whether the intention per se is retrieved (as is the case with the overt prospective-memory responses in Experiment 1). At minimum, the prospective-memory cues caused the cognitive system to react in such a way that the cues were noticed and interfered with quick performance on the lexical-decision task. Because we confirmed no monitoring during the suspended phase, to observe slowing at all, the cognitive systems in the older adults had to spontaneously notice the significance of the cues, spontaneously detect a memory-related discrepancy, or spontaneously and reflexively retrieve the associated intention. We explore this issue further in the General Discussion.
Experiment 3 Discussion This experiment was the flrst to demonstrate the intentioninterference effect in older adults in the absence of evidence for monitoring during the test phase. Experiment 2's results converged well with those of Experiment 1, and our analysis of RTs was not hmited by ceiling effects (cf Uttl, 2011). Statistically, our results indicated significantly greater slowing in the prospective-memory condition than the recognition-memory condition. Although the three-way interaction did not meet conventional levels of signiflcance, there was a trend toward a significant effect and examination of Figure 4 suggests that this effect was mainly driven by the older adults. We are unsure of why this
As part of our goal to provide a comprehensive examination of spontaneous retrieval in older adults, we investigated possible boundary conditions in Experiment 3. We again used the intention*We also examined RTs that were trimmed two standard deviations from each individual's mean, and this produced the same statistical conclusions, except that the interaction between target/control item and prospective-/recognition-memory condition was only marginally significant, F(l, 122) = 3.49, p = .06, and, in the young adult prospectivememory condition, the intention-interference effect clearly surpassed the alpha threshold, i(30) = 3.21, p = .003, d = 1.17. Also, there was no evidence of a three-way interaction with age, F{\, 122) < .01, MSE = 2,053.14,/) = .99, Ti=^ < .01.
AGING AND SPONTANEOUS RETRIEVAL interference procedure, but this time tested whether cues that were semantically related to the prospective-memory targets could also trigger spontaneous retrieval processes. Some research suggests that older adults might be as sensitive to related cues as young adults. For example, in a naturalistic prospective-memory test, Kvavilashvili and Fisher (2007) observed that young and older adults were often reminded of a prospective-memory intention when seeing a variety of cues that were related to the intention of phoning the experimenter. Another possibility arises from the involuntary autobiographical memory literature (Mace, 2007), in which researchers have found that older adults may have fewer involuntary autobiographical memory retrievals than young adults (Bemtsen & Rubin, 2002; Schlagman, Kliegel, Schulz, & Kvavilashvili, 2009; Schlagman, Kvavilashvili, & Schulz, 2007; but see Rubin & Bemtsen, 2009). Because autobiographical memory retrievals may be driven by a wide range of cues (Baars, Ramamurthy, & Franklin, 2007), the implication of these studies is that spontaneous retrieval processes may be less responsive to a diversity of cues in older adults than in young adults.
Method Participants and design. The 48 young adults were Furman University undergraduates who received monetary compensation or course credit, and the 48 community-dwelling older adults (n = 21 young-old and n = 21 old-old) received monetary compensation. Participants were tested individually or in groups of two to four. No participants had mean control trial RTs that were greater than four standard deviations from their group mean. The design of this experiment was a 2 X 2 X 2 mixed factorial design, in which we varied age (young, older) and type of target in the suspended phase (exact cue, associate lure) between subjects and lexical-decision item (target, control) within subjects. Procedure. The experimental procedure was closely modeled after Experiment 2 (see Figure 3) with a few exceptions. Unlike Experiment 2, there was only a prospective-memory condition: All participants performed an image-rating task and a lexical-decision task in which prospective-memory cues were unexpectedly presented. The prospective-memory target words were money and teeth (or water and tree in the counterbalanced condition). In this experiment, participants encoded the prospective-memory task by writing the sentence, "If I see money or teeth (or water or tree) during the image-rating task, then I will press the 'Q' key." This implementation-intention instruction (GoUwitzer, 1999) was used to ensure strong encoding of the intention. The experimenter confirmed that each participant understood the prospectivememory instructions. In this experiment, there were 312 lexical-decision trials, consisting of 144 word trials, 144 nonword trials, 12 target-word trials, and 12 control-word trials. Participants were randomly assigned to either the exact cue or the associate lure condition. In the exact cue condition, each target and control word was presented six times during the lexical-decision task. In the associate lure condition, we selected three words that were forwardly associated (Nelson, McEvoy, & Schreiber, 1998) with each target (or control) word and presented these words two times each. In both conditions, word repetitions were always separated by more than 11 trials and the
917
target- and control-word sets and positions were counterbalanced across participants. The entire procedure lasted 30-45 min. Materials. Lexical-decision task filler words and nonwords (nontargets and noncontrols) were selected from the Balota et al. (2002) norms and their order was randomized for each participant, with the stipulation that no more than three words (or nonwords) could occur in succession. The words were medium frequency, four to nine letters, and ranged in mean RTs from 600-750 ms. The nonwords were matched on length and RTs. No filler words were forwardly associated with target or control words (Nelson et al., 1998). Seventy-two items (36 words and nonwords) appeared twice each and 24 items (12 words/nonwords) appeared six times each, thereby matching the frequency of occurrence of the associate lures, exact cues, and matched control items. Thus, any increase or decrease in the speed of responding to target and control items could not be attributed to their presentation frequency. The associated lures for money v/eie profit, wallet, and fund; for teeth: gums, braces, and dentist; for tree: sap, leaf, and forest; and for water: faucet, splash, and fountain. The mean forward associations ranged from .60 to .65 (Nelson et al., 1998). Money served as the control for tree, and teeth served as the control for water, and vice versa. Each target-control pair (and associates of these pairs) was matched on mean lexical-decision RT and forward associative strength (Nelson et al., 1998).
Results As in Experiment 2, we compared mean correct RTs to target and control items in order to test for intention interference (see Figure 5). We conducted a 2 X 2 X 2 mixed ANOVA that included the between-subjects variables of age (young, older) and cue condition (exact cue, associative lure) and the within-subjects variable of item (control, target). Older adults (M = 735 ms. So = 111.11) were slower than young adults (M = 545 ms, SD = 111.67), F(l, 92) = 40.31, MSE = 43,050.32,p < .001, TI^ = .30. Consistent with the results of Experiment 2, participants were significantly slower in their lexical-decision responses to target items {M = 657 ms, SD = 195.96) relative to control items (M =
• Target
900
• CODlIOl
800 700
s 600
400 300 200 100
Young
Older Exact
Young
Older Associate
Figure 5. Mean target and control word RTs during the lexical-decision task in Experiment 3. Error bars represent SEM.
918
MULLET ET AL.
622 ms, SD = 164.24), F(l, 92) = 17.77, MSE = 3,338.63, p < .001, T|^ = .15. This main effect was qualified by a significant two-way interaction with cue condition, F(l, 92) = 4.30, MSE = 3,338.63, p = .04, r\^ = .04, and a marginally significant (p = .054) three-way interaction with cue condition and age group, F( 1, 92) = 3.81, MSE = 3,338.63, ti^ = .03.' As in Experiment 2, to determine the source of these interaction effects, we conducted separate f tests comparing target and control RTs for each of the four between-subjects conditions. Replicating Experiment 2, for the exact cue condition, there was significant slowing to target items relative to control items for both young adults (AÍT^get = 521 ms, SD = 121.02, Mco„„o, = 488 ms, SD = 65.55), /(23) = 2.10, p = .047, d = 0.88, and older adults (^Target = 761 ms, SD = 206.96, Mcontroi = 689 ms, SD = 146.90), i(23) = 3.40, p < .01, d = 1.42. For the associative lure condition, there was significant slowing to target items relative to control words for young adults (MT^rget = 600 ms, SD = 136.35, Mcontroi = 570 ms, SD = 108.93), i(23) = 2.21,p = .04, d = 0.92, but not for older adults {MT-^^^, = 747 ms, SD = 198.46, ^Control = 742 ms, SD = 170.89), t
