Visual Word Recognition Without Central Attention ... - Semantic Scholar

Psychology and Aging 2006, Vol. 21, No. 3, 431– 447

Copyright 2006 by the American Psychological Association 0882-7974/06/$12.00 DOI: 10.1037/0882-7974.21.3.431

Visual Word Recognition Without Central Attention: Evidence for Greater Automaticity With Advancing Age Mei-Ching Lien

Philip A. Allen

Oregon State University

University of Akron

Eric Ruthruff

Jeremy Grabbe

University of New Mexico

University of Akron

Robert S. McCann

Roger W. Remington

National Aeronautics and Space Administration Ames Research Center

Johns Hopkins University

The present experiments examined the automaticity of word recognition. The authors examined whether people can recognize words while central attention is devoted to another task and how this ability changes across the life span. In Experiment 1, a lexical decision Task 2 was combined with either an auditory or a visual Task 1. Regardless of the Task 1 modality, Task 2 word recognition proceeded in parallel with Task 1 central operations for older adults but not for younger adults. This is a rare example of improved cognitive processing with advancing age. When Task 2 was nonlexical (Experiment 2), however, there was no evidence for greater parallel processing for older adults. Thus, the processing advantage appears to be restricted to lexical processes. The authors conclude that greater cumulative experience with lexical processing leads to greater automaticity, allowing older adults to more efficiently perform this stage in parallel with another task. Keywords: aging, dual-task interference, visual word recognition, divided attention

Because word reading plays a critical role in human cognition, considerable effort has been devoted in recent years to understanding the underlying mechanisms. One important issue is the degree to which word recognition (or lexical access) is automatic. Many studies have shown that it is automatic in the sense that it does not require the specific intention to read words. One example comes from the well-known Stroop paradigm, whereby participants are slow to name the ink color of a word that spells an incongruent color name (e.g., the word red in the color green; see MacLeod, 1991, for a review). People cannot avoid reading the word even when doing so strongly interferes with their assigned task. Although word recognition does not require an intention to read, there is evidence that it does require certain attentional resources. Lachter, Forster, and Ruthruff (2004), for example, found that words produced no repetition priming unless they received spatial

attention1 (see also Lachter, Ruthruff, Lien, & McCann, 2006; McCann, Folk, & Johnston, 1992; but see Brown, Roos-Gilbert, & Carr, 1995). There is also evidence that word recognition requires central attention; that is, word recognition on one task cannot take place while central attention is devoted to selecting a response to another task (McCann, Remington, & Van Selst, 2000). However, another dual-task study used somewhat different methods (different age groups and a different input modality for the primary task) and obtained a different outcome (Allen et al., 2002). Although no firm conclusions can yet be drawn, as we discuss later in detail, these studies suggest two interesting hypotheses. One is that older adults are better able than younger adults to perform word recognition without central attention. Another hypothesis is that visual word recognition can proceed in parallel with another visual task but not with an auditory task. The present study examines both of these hypotheses.

Mei-Ching Lien, Department of Psychology, Oregon State University; Philip A. Allen and Jeremy Grabbe, Department of Psychology, University of Akron; Eric Ruthruff, Department of Psychology, University of New Mexico; Robert S. McCann, National Aeronautics and Space Administration Ames Research Center, Moffett Field, California; Roger W. Remington, Applied Physics Laboratory, Johns Hopkins University. This research was supported by funding from the Oregon State University Research Incentive Programs, the College of Liberal Arts at Oregon State University, and National Aeronautics and Space Administration Grant NCC 2-1325. Correspondence concerning this article should be addressed to MeiChing Lien, Department of Psychology, Oregon State University, Corvallis, OR 97331. E-mail: [email protected]

Locus-of-Slack Logic One direct way to establish whether a particular mental process (e.g., word recognition) requires central attention is to determine whether this process can be carried out while central attention is devoted to another task. Before we describe the details of this 1

Johnston, McCann, and Remington (1995) provided evidence for two dissociable forms of attention. Input attention, which is identical to what we call spatial attention, limits parallel perceptual processing of multiple stimuli, whereas central attention limits parallel processing in higher mental functions, such as response selection. 431

432

LIEN ET AL.

approach, however, it is necessary to provide background on dual-task methodology and findings. Probably the most widely used dual-task paradigm is the psychological refractory period (PRP) paradigm, in which two tasks are separated by a variable time interval (called the stimulus onset asynchrony [SOA]). The task presented first is called Task 1, and the task presented second is called Task 2. The typical PRP finding is that the response time (RT) for Task 1 (RT1) is roughly constant across SOAs, but the RT for Task 2 (RT2) increases sharply as the SOA decreases (i.e., when the tasks are demanded at nearly the same time). This robust form of dual-task interference, known as the PRP effect (Telford, 1931), has been found even when tasks have no input conflicts or output conflicts. Many studies have found evidence that the PRP effect occurs largely because central stages (e.g., response selection, decision making) for Task 1 and Task 2 do not operate in parallel (for reviews, see Lien & Proctor, 2002; Pashler, 1994; Pashler & Johnston, 1989; see also Meyer & Kieras, 1997, for a different point of view2). This central bottleneck, shown in Figure 1, has been reported to occur even with exceptionally easy tasks (e.g., Lien, McCann, Ruthruff, & Proctor, 2005; Lien, Proctor, & Allen, 2002). Because central stages of Task 2 are postponed until central stages of Task 1 are completed, this central bottleneck creates a period of cognitive slack between the perceptual and central stages in Task 2 at short SOAs (Pashler, 1984; Schweickert, 1978; Welford, 1952). At long SOAs, however, the tasks are performed more or less independently, so there is no period of cognitive slack. According to this central bottleneck model, variables that primarily affect the duration of Task 2 central processing should produce additive effects with SOA. However, any variable affecting earlier Task 2 stages should produce effects that are underadditive with SOA. In brief, the reason is that any lengthening of prebottleneck stages of Task 2 (Stage 2A in Figure 1) can be absorbed into the cognitive slack at short SOAs but not at long SOAs (see Pashler, 1994; Schweickert, 1978). For ease of discussion, we refer to this as a slack effect. By observing whether manipulations of a certain stage interact additively or underaddi-

Figure 1. The temporal relations between central processing stages of Task 1 and Task 2 at short stimulus onset asynchrony (SOA) in the psychological refractory period paradigm, as suggested by the central bottleneck model. This model assumes that perceptual and response initiation/execution stages of Task 2 can operate in parallel with any stage of Task 1 but that central stages of Task 2 cannot start until central stages of Task 1 have been completed. 1A, 1B, and 1C are the perceptual, central, and response initiation/execution stages of Task 1, respectively. 2A, 2B, and 2C are the corresponding stages for Task 2. S1 ⫽ stimulus for Task 1; S2 ⫽ stimulus for Task 2; R1 ⫽ response for Task 1; R2 ⫽ response for Task 2.

tively with SOA, one can determine whether that stage is subject to the central bottleneck (i.e., whether it requires central attention). This logic, used in a large number of dual-task studies, has become known as locus-of-slack logic (see McCann & Johnston, 1992; Pashler, 1984; Schweickert, 1978).

Does Word Recognition Require Central Attention? Two recent PRP studies used locus-of-slack logic—within the framework of the central bottleneck model—to determine whether word recognition requires central attention, but they reached seemingly opposite conclusions (Allen et al., 2002; McCann et al., 2000). In both studies, Task 2 required participants to determine whether a letter string formed a word or nonword, also known as a lexical decision task. To manipulate the duration of Task 2 word recognition, the authors in both studies selected a set of lowfrequency words and a set of high-frequency words, with frequency defined in terms of occurrences in the English language (Kucˇera & Francis, 1967). If word recognition requires central attention, then the effect of word frequency on RT2 should be additive with the effects of SOA. However, if word recognition does not require central attention, then the word frequency effect on RT2 should be underadditive with the effects of SOA (i.e., producing a slack effect for word frequency). McCann et al. (2000) investigated the reliance of word processing on central attention in a series of experiments involving younger adults. In Experiments 1 and 2, participants made manual keypresses to a high- or low-pitched tone for Task 1 and a word/nonword judgment (lexical decision) for Task 2. The key finding was that word frequency effects were roughly additive with the effects of SOA. Experiments 3 to 6 replicated the roughly additive effects of word frequency and SOA when Task 2 was a speeded naming task rather than a lexical decision task. McCann et al. therefore concluded that Task 2 word recognition was postponed until central attention switched from Task 1 to Task 2. In other words, they concluded that word recognition requires central attention. Allen et al. (2002) used a similar paradigm to look for age differences in the ability to perform tasks in parallel. Given one widely held theory that older adults exhibit a decrement in processing resources (e.g., Craik & Salthouse, 2000), the authors hypothesized that older adults should consistently show less efficient parallel processing than younger adults across a wide variety of tasks and processing stages, including memory retrieval and response selection. That is, older adults should have more difficulty than younger adults in performing Task 2 (a lexical decision task) while central attention is devoted to Task 1. Whereas McCann et al. (2000) used an auditory discrimination (high- vs. low-pitched tones) for Task 1, Allen et al. (2002) used a visual 2 These authors pointed out that the central bottleneck might reflect a strategic choice rather than a structural limitation. On the one hand, attempts to induce participants to choose to bypass the central bottleneck have generally failed (e.g., Ruthruff, Pashler, & Klaassen, 2001). On the other hand, there is some evidence that bypassing is possible with relatively easy tasks after multiple practice sessions (Hazeltine, Ruthruff, & Remington, 2006; Hazeltine, Teague, & Ivry, 2002; Ruthruff, Hazeltine, & Remington, in press; Ruthruff, Van Selst, Johnston, & Remington, 2006; Schumacher et al., 2001). Note, however, that the present tasks were not especially easy or highly practiced.

AGING AND WORD RECOGNITION

shape discrimination (triangle vs. rectangle). The word or nonword for Task 2 was presented inside the shape, so that there was no competition for spatial attention between tasks. Contrary to McCann et al., Allen et al. found an underadditive interaction between word frequency effects and SOA in both of their experiments, for both younger and older adults. On the basis of locus-of-slack logic, this result suggests that word recognition can proceed without central attention. Allen et al. argued that when a process is highly overlearned (or automatized; e.g., Hasher & Zacks, 1979), as might be the case for lexical processing, it can operate in parallel with central stages on another task (see also Logan & Schulkind, 2000). Although Allen et al. (2002) found an underadditive interaction of word frequency and SOA (i.e., the slack effect) for both older and younger adults, the effect was more pronounced for older adults than for younger adults— especially in Experiment 2 (which included more trials than Experiment 1). This finding suggests, contrary to the general-resource decrement theory of cognitive aging (Craik & Salthouse, 2000), that older adults actually show more efficient parallel processing than younger adults, at least for certain highly overlearned mental processes. Given the potential importance of this finding for theories of cognitive aging, Allen et al. (2002) attempted to provide converging evidence. In particular, they derived a measure of the degree of parallel processing, which they called general savings. It is equal to the time it should take participants to complete the tasks if performed strictly sequentially (estimated as the sum of long-SOA RT1 and long-SOA RT2) minus the time that it actually took participants to complete both tasks at the shortest SOA (the time between Task 1 stimulus onset and Task 2 response, which can be expressed simply as SOA plus RT2). General savings ⫽ 共RT1 Longest SOA ⫹ RT 2Longest SOA兲 ⫺ 共SOAShortest⫹RT 2Shortest SOA兲.

(1)

If there is no parallel processing, then general savings should be equal to zero. To the degree that some parallel processing has occurred, general savings should be positive. Allen et al. (2002) argued that if older adults are better able than younger adults to process the tasks in parallel, then they should show a larger general savings score. Unlike the larger slack effects observed for older adults (relative to younger adults), Allen et al. found no age differences in general savings scores. They concluded that older adults showed at least as much parallel processing, overall, as did younger adults. Nevertheless, there are three reasons why Allen et al.’s (2002) findings regarding general savings are inconclusive. First, RT1 increased as SOA increased, and the increase was much stronger for younger adults than for older adults. These results suggest that participants often withheld the response to Task 1 so that it could be initiated together with the response to Task 2 (a phenomenon known as response grouping; Borger, 1963; Pashler & Johnston, 1989). Consequently, the observed long-SOA RT1 does not necessarily reflect the actual time required to complete Task 1 processing. Second, Allen et al. used a relatively long SOA (250 ms) as their shortest SOA, limiting the opportunity for parallel processing and the opportunity to observe savings. Finally, Allen et al. measured general savings averaged across high- and lowfrequency words, which might not have provided the most sensitive test for age differences. Given that lexical access (the stage

433

that older adults are hypothesized to perform in parallel with Task 1) is longer for low-frequency words than for high-frequency words, the best opportunity to observe the increase in general savings is for low-frequency words in isolation. As we discuss later, the present study addresses each of these three issues and provides a more sensitive test of the hypothesis that older adults are better able than younger adults to carry out word recognition in parallel with Task 1.

The Present Study McCann et al. (2000) and Allen et al. (2002) used a PRP paradigm to study the automaticity of word recognition but arrived at seemingly conflicting conclusions. There are two possible explanations for the discrepancy between these two studies. One hypothesis, which we call the input modality hypothesis, is that visual word recognition can proceed in parallel with a visual Task 1 (as in Allen et al.) but not with an auditory Task 1 (as in McCann et al.). When both Task 1 and Task 2 are visual (which we call the visual–visual [VV] condition) and the Task 1 stimulus is in the same location as the word/nonword of Task 2 (as in Allen et al.), spatial attention to the visual Task 1 stimulus would also cover the word/nonword of Task 2. In contrast, when an auditory Task 1 is combined with the visual Task 2 (which we call the auditory– visual [AV] condition), participants have more leeway to direct spatial attention away from the location of the word/nonword. The motivation to divert spatial attention is that it would temporarily prevent the Task 2 word from being identified (Lachter et al., 2004) and thus reduce the interfering effects of Task 2 processing on Task 1 processing. According to this hypothesis, the lack of spatial attention, rather than the lack of central attention, caused the failure of McCann et al.’s participants to recognize the Task 2 word/nonword in parallel with an auditory Task 1. To evaluate this input modality hypothesis, we had participants in the present Experiment 1 perform both the AV and the VV conditions in separate experimental sessions. In both the AV and the VV conditions, an unfilled visual shape was presented as an auditory stimulus was played. However, only the visual shape was relevant to Task 1 in the VV condition, and only the auditory stimulus was relevant to Task 1 in the AV condition. The advantage of this approach is that the stimulus conditions are identical in the AV and VV conditions. If lexical processing of Task 2 cannot proceed in parallel with central processing of the auditory Task 1, then additive effects of word frequency and SOA should be obtained. If lexical processing on Task 2 can proceed in parallel with central processing of the visual Task 1, then an underadditive effect should be observed. Another hypothesis, which we call the age hypothesis, is that the degree of parallelism of word recognition with the performance of another task increases with age. Because word reading is a skill that develops over time, the ability to perform word recognition in parallel with another task might improve as the cumulative experience with words increases (Hasher & Zacks, 1979; Logan & Schulkind, 2000). Therefore, older adults might show more efficient parallel processing of Task 2 word recognition with Task 1 than younger adults. That is, older adults might show an underadditive interaction between the effects of word frequency and SOA, as found by Allen et al. (2002), whereas younger adults might show a more additive interaction, as found by McCann et al. (2000).

LIEN ET AL.

434

To evaluate this age hypothesis, we examined slack effects for both younger adults and older adults (who performed both the AV and the VV conditions). To look for converging evidence that older adults are better able than younger adults to perform word recognition in parallel with Task 1, we also estimated the general savings (see Equation 1) for both age groups. As we show, Experiment 1 reveals parallel processing for older adults in both AV and VV conditions but not for younger adults. Thus, Experiment 2 was designed to test our hypothesis that the increase in parallel processing for older adults is restricted to lexical processes.

Experiment 1 The purpose of Experiment 1 was to evaluate (a) whether visual word recognition can proceed in parallel with a visual Task 1 (as in Allen et al., 2002) but not with an auditory Task 1 (as in McCann et al., 2000) and (b) whether older adults show more efficient parallel processing of Task 2 word recognition with Task 1 than younger adults.

Method Participants. A total of 48 participants (24 younger adults and 24 older adults) were tested. Younger adults were undergraduate students at the University of Akron who participated in partial fulfillment of a course requirement. Their mean age was 25 years, with a range of 19 to 44 years. Older adults were community-dwelling individuals and were paid $20 per hour for their participation. Their mean age was 71 years, with a range of 59 to 83 years. All participants had normal or corrected-to-normal vision. Each participant performed two sessions, one for the AV condition and one for the VV condition. Half of the participants performed the AV condition first, whereas the other half performed the VV condition first. Both sessions were conducted during a single visit to the laboratory, although participants were given a break between sessions. The entire experiment lasted less than 2 hr. All participants were tested on the Wechsler Adult Intelligence Scale— Revised (WAIS–R; Wechsler, 1981) Vocabulary and Digit Symbol Substitution Task subscales. For the Digit Symbol task, younger adults (mean score ⫽ 67.3) scored significantly higher than older adults (mean score ⫽ 50.8), t(46) ⫽ 3.73, p ⬍ .001. For the WAIS–R Vocabulary test, older adults showed higher scores (mean score ⫽ 53.3) than younger adults (mean score ⫽ 45.7), t(46) ⫽ 2.03, p ⬍ .05. Apparatus and stimuli. Stimulus presentation, timing, and data collection were controlled via IBM-compatible microcomputers driven by E-Prime software (Schneider, Eschman, & Zuccolotto, 2002). For Task 1, a visual shape appeared on the screen at the same time that an auditory stimulus was sounded. The shape was an unfilled circle or square. The circle was 6 cm in diameter, and the square was 6 cm long on each side. At a viewing distance of 55 cm, both shapes subtended horizontal and vertical visual angles of 6.23°. The auditory stimulus was either a pure tone or white noise (similar to a hissing sound). For the AV condition, Task 1 was to respond to the sound and ignore the shape. For the VV condition, Task 1 was to respond to the shape and ignore the sound. In both the AV and the VV conditions, the Task 2 stimulus was a word or a nonword presented inside the circle or square. Each letter was presented in lowercase in Courier New font, size 18, and was approximately 0.8 cm in height and 0.6 cm in width. At a viewing distance of 55 cm, each letter subtended a visual angle of 0.83° ⫻ 0.63°. The word stimuli were taken from the Kucˇera and Francis (1967) norms to form two different stimulus lists. Because each participant completed two sessions (one for the AV condition and one for the VV condition), we generated two different word lists (see Appendixes A and B). One word list (List 1), essentially the same as in Allen et al. (2002), included low-frequency words ranging from 10 to 30 occurrences and high-frequency words

ranging from 240 to 1,016 occurrences. The other word list (List 2) also included low-frequency words ranging from 10 to 30 occurrences but included high-frequency words ranging from 151 to 236 occurrences (taken from Allen, Wallace, & Weber’s, 1995; medium high-frequency category).3 We formed nonwords by changing one of the letters of a word stimulus. Each word or nonword appeared only once during the experimental trials for an individual participant. The presentation of the stimulus lists in the AV and VV conditions was counterbalanced. That is, half of the participants received List 1 in the AV condition and List 2 in the VV condition, whereas the other half of the participants received List 1 in the VV condition and List 2 in the AV condition. Thus, although the two lists were not identical, each was used equally often in each condition for both younger and older adults. Design and procedure. To begin the session, each participant performed two practice blocks. The first block contained 32 trials with a constant 1,500-ms SOA. The purpose of this block was to provide an opportunity for participants to learn the tasks without having to deal with both tasks at the same time and to encourage participants to perform the two tasks without grouping together the responses for Task 1 and Task 2. The second practice block contained 72 trials with the same set of SOAs used in the experimental blocks. Each participant was then given a total of 432 regular trials, divided into six blocks of 72 trials each. These trials consisted of the 144 possible combinations of the following five variables: Task 1 stimulus type (tone/noise for the AV condition or circle/square for the VV condition), word type (word or nonword), letter-string length (four, five, or six letters), word frequency (high or low), and SOA (50, 100, 300, 500, 700, or 900 ms). Therefore, there were three occurrences of each combination of conditions (although with a different word each time) for each modality condition (AV and VV). These conditions were selected randomly within a session, with the restriction that each combination of conditions must occur equally often during the regular trials. The participant pressed the space bar of the keyboard to initiate the first trial of each block. The fixation cross was then presented for 500 ms in the center of the screen. One hundred ms after offset of the fixation cross, the Task 1 stimulus (S1) was presented: The auditory stimulus was sounded (for 100 ms), and the shape appeared in the screen center (until response). The Task 2 stimulus (S2; a string of letters) followed S1 after one of six SOAs (50, 100, 300, 500, 700, or 900 ms, randomly selected within blocks) and remained on the screen until a response was recorded. In the AV condition, participants were asked to respond to the auditory S1 by pressing the z key with their left middle finger for the tone and the x key with their left index finger for the noise. In the VV condition, participants were asked to respond to the visual S1 by pressing the z key with their left middle finger for the circle and the x key with their left index finger for the square. For Task 2 (which was identical for the AV and VV conditions), participants were asked to press the ⬎ key with their right index finger if the letter string formed a word and to press the ? key with their right middle finger if the letter string formed a nonword. Participants were instructed to respond to S1 before S2 and to respond as quickly and accurately as possible to both tasks. They were also asked to maintain at least 90% accuracy and to respond to Task 1 within 3 s for younger adults and within 5 s for older adults. The use of a time-out for Task 1 discouraged response grouping. Feedback for incorrect responses was presented on the screen for 1,200 ms.

3

The reason List 2 included a different range of high-frequency words is that no other high-frequency words could be found in the same range as those in List 1. Allen et al. (1995) found no RT difference between the ranges of very high-frequency words and medium high-frequency words. Similarly, we found no difference in performance between these ranges in the present study. Therefore, the present data analyses were collapsed across Word List 1 and List 2.

AGING AND WORD RECOGNITION

Results Trials were excluded from the RT analyses if the response to either task was incorrect. In addition, for younger adults, trials were excluded if RT was either shorter than 100 ms or longer than 3,000 ms; approximately 0.8% of trials were eliminated because of these RT cutoffs. For older adults, trials were excluded if RT was either shorter than 100 ms or longer than 4,000 ms; approximately 0.9% of trials were eliminated because of these cutoffs. The proportions of errors (PEs) for Task 1 (PE1) and Task 2 (PE2) were determined without regard to whether the response for the other task was correct. The independent variables included in the data analyses were age group (younger vs. older), modality condition (AV vs. VV), S2 word frequency (high vs. low), S2 lexicality (word vs. nonword), SOA (100, 300, 500, 700, and 900 ms),4 and participants. S1 type (tone/noise in the AV condition and circle/square in the VV condition) and session (first vs. second) had little effect5 and therefore were not included as factors in the final data analyses. An alpha level of .05 was used to determine statistical significance. Because word frequency and lexicality are not orthogonal, they cannot be included in the same crossed analysis of variance (ANOVA). Thus, we conducted two different ANOVAs. The ANOVA of our primary interest, which included the S2 word frequency variable, was conducted for words only.6 The secondary ANOVA, which excluded the S2 word frequency variable, was conducted to examine the lexicality effect (word vs. nonword). In this particular data analysis, only significant main effects or interactions involving S2 lexicality and age group were reported. Note that our tests for the automaticity of word recognition were based on the interaction between word frequency and SOA, not the interaction between lexicality (word vs. nonword) and SOA. In brief, the reason is that the lexicality variable likely influenced not only word recognition but also response selection and response execution. Word frequency effects. Mean RT1s and RT2s are shown in Figure 2 for both the AV and the VV conditions. PE data for Task 1 are shown in Table 1, and PE data for Task 2 are shown in Table 2. For RT1, the main effect of age group was significant, F(1, 46) ⫽ 57.27, p ⬍ .0001, MSE ⫽ 505,731; mean RT1 was 634 ms for younger adults and 982 ms for older adults. There were also main effects of modality condition, F(1, 46) ⫽ 8.34, p ⬍ .01, MSE ⫽ 155,107, and SOA, F(4, 184) ⫽ 13.84, p ⬍ .001, MSE ⫽ 15,575. Mean RT1 was 74 ms longer in the AV condition (845 ms) than in the VV condition (771 ms). Mean RT1s were 863, 797, 781, 790, and 810 ms at the 100-, 300-, 500-, 700-, and 900-ms SOAs. The interaction of modality condition and SOA was significant, F(4, 184) ⫽ 3.90, p ⬍ .01, MSE ⫽ 8,371. For the AV condition, mean RT1 was roughly constant across SOAs (RT1s ⫽ 878, 842, 817, 829, and 858 ms at the 100-, 300-, 500-, 700-, and 900-ms SOAs). For the VV condition, mean RT1 was longer at the shortest SOA than at the other SOAs (RT1s ⫽ 848, 751, 745, 751, and 762 ms at the 100-, 300-, 500-, 700-, and 900-ms SOAs), which might partly be because S2 laterally masked the visual S1. Overall, the data suggest that response grouping rarely occurred at long SOAs for both younger and older adults. The PE1 data showed a significant main effect of SOA, F(4, 184) ⫽ 5.33, p ⬍ .001, MSE ⫽ 0.0012; PE1 was slightly higher at the shortest SOA than at the other SOAs (PE1s were .03, .02, .02,

435

.02, and .02, at 100-, 300-, 500-, 700-, and 900-ms SOAs, respectively). No other effects were significant. For RT2, the main effect of age group was significant, F(1, 46) ⫽ 50.57, p ⬍ .001, MSE ⫽ 532,424; mean RT2 was 742 ms for younger adults and 1,077 ms for older adults. The main effect of SOA was also significant, F(4, 184) ⫽ 363.44, p ⬍ .0001, MSE ⫽ 15,438; RT2 increased as SOA decreased (RT2s ⫽ 1,181, 968, 845, 788, and 765 ms at the 100-, 300-, 500-, 700-, and 900-ms SOAs). Thus, a sizeable PRP effect of 416 ms was obtained. The main effect of S2 word frequency was also significant, F(1, 46) ⫽ 125.39, p ⬍ .001, MSE ⫽ 6,775. RT2 was 60 ms shorter when S2 was a high-frequency word (879 ms) than when it was a low-frequency word (939 ms). The interaction of age group and SOA was also significant, F(4, 184) ⫽ 19.55, p ⬍ .0001, MSE ⫽ 15,438, reflecting that the PRP effect was smaller for younger adults (321 ms) than for older adults (511 ms). Most important, the Age Group ⫻ S2 Word Frequency ⫻ SOA interaction on RT2 was significant, F(4, 184) ⫽ 4.28, p ⬍ .01, MSE ⫽ 6,436, reflecting that older adults showed underadditivity between word frequency and SOA but younger adults did not (see Figure 2). For older adults, the word frequency effect was ⫺2, 59, 51, 64, and 98 ms at 100-, 300-, 500-, 700- and 900-ms SOAs. For younger adults, the word frequency effect was 92, 45, 48, 69, and 71 ms at the 100-, 300-, 500-, 700- and 900-ms SOAs. The four-way interaction of age group, modality condition, S2 word frequency, and SOA was not significant (F ⬍ 1.0). For PE2, there were main effects of age group, F(1, 46) ⫽ 30.33, p ⬍ .0001, MSE ⫽ 0.0119, and S2 word frequency, F(1, 46) ⫽ 57.28, p ⬍ .0001, MSE ⫽ 0.0080. PE2 was higher for younger adults (.07) than for older adults (.03). PE2 was also higher for low-frequency words (.07) than for high-frequency words (.03). The interaction of age group and S2 word frequency was also significant, F(1, 46) ⫽ 8.89, p ⬍ .001, MSE ⫽ 0.0080; the word frequency effect on PE2 was .06 for younger adults but was only .03 for older adults. No other effects were significant. Lexicality effects. Mean RTs to Task 1 and Task 2 are shown in Figure 3 for both the AV and the VV conditions. For RT1, the only significant main effect or interaction involving S2 lexicality was the main effect of S2 lexicality, F(1, 46) ⫽ 5.70, p ⬍ .05, MSE ⫽ 4,500, and its interaction with age group, F(1, 46) ⫽ 5.28, p ⬍ .05, MSE ⫽ 4,500. RT1 was 7 ms shorter when S2 was a word (RT1 ⫽ 808 ms) than when it was a nonword (RT1 ⫽ 815 ms). Younger adults showed no effect of lexicality on RT1 (RT1 ⫽ 634 ms for both words and nonwords), whereas older adults showed a slightly shorter mean RT1 when S2 was a word (RT1 ⫽ 982 ms) than when it was a nonword (RT1 ⫽ 996 ms). No effects involving S2 lexicality on PE1 were significant. 4

Initially, Experiment 1 included six different SOAs (50, 100, 300, 500, 700, and 900 ms). Because of a technical difficulty, however, the software did not present an SOA of less than 100 ms. Therefore, we did not analyze data from the 50-ms trials, which left only five different SOAs (100, 300, 500, 700, and 900 ms) in the final data analyses. 5 The same pattern of results was obtained even when the data analysis included the S1-type variable. The session variable also had little influence on the data and did not modulate any interactions involving SOA and frequency (Fs ⱕ 1.75). 6 The high- versus low-frequency distinction does not apply to nonword stimuli. Consequently, the word frequency effect was examined only for the word stimuli.

436

LIEN ET AL.

Figure 2. Mean response times (RTs) for Task 1 and Task 2 in Experiment 1 for younger adults and older adults in the auditory–visual (AV) condition and the visual–visual (VV) condition as a function of stimulus onset asynchrony (SOA; 100, 300, 500, 700, and 900 ms) and Stimulus 2 word frequency (high vs. low).

For RT2, there was a main effect of lexicality, F(1, 46) ⫽ 57.21, p ⬍ .0001, MSE ⫽ 22,573; mean RT2 was shorter when S2 was a word (RT2 ⫽ 909 ms) than when it was a nonword (RT2 ⫽ 961 ms). For PE2, there were significant main effects of group, F(1, 46) ⫽ 35.38, p ⬍ .0001, MSE ⫽ 0.0371, and lexicality, F(1, 46) ⫽ 7.54, p ⬍ .01, MSE ⫽ 0.0141. PE2 was higher for younger adults (PE2 ⫽ .087) than for older adults (PE2 ⫽ .034). Participants made more errors when S2 was a nonword (PE2 ⫽ .068) than when it was a word (PE2 ⫽ .053). The interaction between group and lexicality was also significant, F(1, 46) ⫽ 6.18, p ⬍ .05, MSE ⫽ 0.0141. The lexicality effect on PE2 was .029 for younger adults and .001 for older adults. The three-way interaction of S2 lexicality, modality condition, and SOA was significant on PE2, F(4, 184) ⫽ 2.68, p ⬍ .05, MSE ⫽ 0.0026. For the AV condition, there was a trend for the lexicality effect on PE2 to decrease as SOA increased (the effect on PE2 was .023, .023, .009, ⫺.008, and .004 at the 100-, 300-, 500-, 700-, and 900-ms SOAs). For the VV condition, there was no consistent trend across SOAs (the lexicality effect on PE2 was .031, .010, .018, .027, and .011 at the 100-, 300-, 500-, 700-, and 900-ms SOAs).

Discussion In Experiment 1, older adults produced underadditive effects of word frequency and SOA in both the AV and the VV conditions, whereas younger adults produced additive effects in both conditions. Thus, age influenced the pattern of results, but Task 1 modality (auditory vs. visual) did not. On the basis of locus-ofslack logic and the central bottleneck model, these results indicate that word recognition required central attention for younger adults but not for older adults. Thus, these findings support the hypothesis that the efficiency of lexical access improves with age. An alternative hypothesis that deserves consideration is that the lack of slack effects for younger adults occurred simply because these participants had insufficient slack time to absorb the effects of word frequency. This hypothesis seems unlikely, however, given that younger adults did show a substantial PRP effect of 321 ms, which should be more than long enough to absorb the 65-ms word frequency effect. Nevertheless, it is conceivable that younger adults would have shown more underadditivity had they produced even longer slack times. To address this possibility, we rank ordered younger adults on the basis of their mean RT1 (which

AGING AND WORD RECOGNITION

437

Table 1 Proportion of Errors on Task 1 for Younger Adults and Older Adults in Experiment 1 as a Function of Stimulus Onset Asynchrony (SOA), Modality, Stimulus 2 Lexicality, and Stimulus 2 Word Frequency SOA Condition and word type

100

300

500

700

900

Younger adults Auditory–visual condition Nonword High-frequency word Low-frequency word Visual–visual condition Nonword High-frequency word Low-frequency word

.027 .032 .025

.022 .021 .021

.024 .019 .019

.026 .014 .028

.019 .014 .026

.035 .028 .037

.026 .023 .032

.017 .028 .007

.017 .007 .009

.019 .023 .028

Older adults Auditory–visual condition Nonword High-frequency word Low-frequency word Visual–visual condition Nonword High-frequency word Low-frequency word

.038 .039 .041

.030 .030 .029

.027 .023 .030

.027 .032 .019

.026 .023 .028

.024 .021 .021

.019 .014 .012

.012 .009 .009

.010 .009 .005

.010 .007 .007

should be positively correlated with the amount of slack time; see Figure 1). We then examined the RT2 data for the 12 younger participants who exhibited the longest RT1 and hence the longest slack times. Even these younger adults with the longest slack times failed to show an interaction of S2 word frequency and SOA, F(4,

44) ⫽ 1.08, p ⬎ .05, MSE ⫽ 3,603; the word frequency effect was 98, 60, 50, 66, and 66 ms at the 100-, 300-, 500-, 700- and 900-ms SOAs. Another method of examining the effect of slack duration for younger adults is to analyze just the half of the trials for each participant in which RT1 was the longest. Even the longer half of

Table 2 Proportion of Errors on Task 2 for Younger Adults and Older Adults in Experiment 1 as a Function of Stimulus Onset Asynchrony (SOA), Modality, Stimulus 2 Lexicality, and Stimulus 2 Word Frequency SOA Condition and word type

100

300

500

700

900

Younger adults Auditory–visual condition Nonword High-frequency word Low-frequency word Visual–visual condition Nonword High-frequency word Low-frequency word

.086 .049 .079

.088 .035 .098

.092 .052 .119

.092 .060 .143

.072 .037 .096

.092 .046 .079

.086 .037 .113

.81 .031 .088

.93 .039 .100

.085 .035 .113

Older adults Auditory–visual condition Nonword High-frequency word Low-frequency word Visual–visual condition Nonword High-frequency word Low-frequency word

.036 .023 .048

.031 .019 .040

.038 .027 .045

.030 .014 .043

.033 .021 .047

.040 .025 .050

.036 .019 .054

.036 .021 .058

.032 .012 .045

.031 .023 .039

438

LIEN ET AL.

Figure 3. Mean response times (RTs) for Task 1 and Task 2 in Experiment 1 for younger adults and older adults in the auditory–visual (AV) condition and the visual–visual (VV) condition as a function of stimulus onset asynchrony (SOA; 100, 300, 500, 700, and 900 ms) and Stimulus 2 lexicality (word vs. nonword).

younger adults’ RT data failed to show an interaction of S2 word frequency and SOA, F(4, 92) ⫽ 2.44, p ⬎ .05, MSE ⫽ 9,574; the word frequency effect was 109, 34, 34, 63, and 69 ms at the 100-, 300-, 500-, 700-, and 900-ms SOAs. Consequently, there is no evidence that the failure of younger adults to produce an underadditive interaction between word frequency effects and SOA was due to insufficient slack time. For a converging measure of the amount of parallel processing (in addition to slack effects), we also computed the amount of general savings for each participant (see Equation 1). Again, general savings reflects the degree to which actual dual-task performance was faster than strictly serial processing of the two tasks. For this analysis we used data from the low-frequency words only, because they should show the largest benefit of performing lexical access in parallel with Task 1 central operations. There was a significant age difference in general savings, F(1, 46) ⫽ 12.67, p ⬍ .001, MSE ⫽ 110,697; younger adults showed general savings of 193 ms, whereas older adults showed general savings of 434 ms. Given that responses were slower overall for older adults than for younger adults, the larger general savings for older adults than for younger adults could be due to a generalized slowing of all

mental processes with age (e.g., Cerella, 1990; Salthouse, 1996). If we assume equal parallelism for both age groups, generalized slowing of all processing stages by the same multiplicative factor k would increase general savings by the same factor k (see Appendix C). Therefore, to evaluate whether the increase in general savings for older adults exceeded that predicted by generalized slowing, we also used a more conservative test. That is, we computed the proportional savings for both younger adults and older adults as follows: Proportional savings ⫽

共RT1Longest SOA ⫹ RT2Longest SOA) ⫺ (SOAShortest ⫹ RT2Shortest SOA) . (2) (RT1Longest SOA ⫹ RT2Longest SOA)

When we used this more conservative measure, there was still a trend for older adults to show larger savings than younger adults (.20 vs. .14), F(1, 46) ⫽ 3.43, p ⫽ .07, MSE ⫽ 0.0207. In summary, the general and proportional savings scores suggest that older adults showed more overall parallel processing. Thus, they provide converging evidence that older adults were better able than

AGING AND WORD RECOGNITION

younger adults to perform the word recognition Task 2 in parallel with Task 1.

Experiment 2 We have hypothesized that older adults show greater slack effects because they are better able to overlap Task 2 lexical processing with Task 1 central operations. An alternative hypothesis, however, is that older adults are less able to block out Task 2 processing while performing Task 1. A related hypothesis is that older adults simply chose a strategy of processing more of Task 2 in parallel with Task 1. To determine whether the increased parallel processing in older adults is due specifically to superior lexical processing (as opposed to a general inability to block out Task 2 or a processing strategy), we conducted a conceptual replication of Experiment 1 using a nonlexical Task 2. Because word frequency influences the stimulus categorization stage, our goal was to find a nonlexical task that would allow a similar manipulation of stimulus categorization. The task we chose was based on an experiment reported by Johnston and McCann (2006), in which participants decided whether a box was narrow or wide. We manipulated the difficulty of this categorization by using two different narrow boxes and two different wide boxes, one of which was close to the arbitrary boundary between narrow and wide (the difficult condition) and one of which was far from the boundary (the easy condition). Thus, we manipulated the difficulty of a stimulus categorization stage in both Experiment 1 (lexical task) and Experiment 2 (nonlexical task), and this manipulation was followed by a response selection stage. Johnston and McCann (2006) found additive effects between this box-width manipulation and SOA in a sample of younger adults, which suggests that this categorization (like lexical access) did not proceed in parallel with Task 1 central operations. The key question in the present experiment is whether the same findings will be obtained in a sample of older adults (i.e., additive effects of box-width difficulty and SOA). If older adults have an advantage specific to lexical processing, as we have proposed, then there is no reason to expect older adults to produce especially large slack effects with a nonlexical Task 2. However, if older adults produce larger slack effects because they are generally less able to block out Task 2 processing, then one might expect results similar to those of Experiment 1 (i.e., larger slack effects for older adults). Likewise, if older adults are simply more likely to choose not to block out Task 2 processing in general, then one might expect results similar to those of Experiment 1. For younger adults, we expect to replicate Johnston and McCann’s (2006) results.

Method Participants. A total of 48 participants (24 younger adults and 24 older adults) were tested. All younger adults were undergraduate psychology students at Oregon State University who participated in exchange for course credit. Their mean age was 20.58 years, with a range of 18 to 32 years. Thirteen older adults were recruited from the community surrounding Oregon State University, and 11 older adults were drawn from the same participant pool as in Experiment 1 (i.e., from the University of Akron community). Older adults were paid $20 per hour for their participation. Their mean age was 68 years, with a range of 61 to 78 years. None had participated in Experiment 1.

439

As in Experiment 1, all participants were tested on the WAIS–R Vocabulary and Digit Symbol Substitution Task subscales. The participants’ performance was similar to performance in Experiment 1. For the Digit Symbol task, younger adults (mean score ⫽ 65.3) scored significantly higher than older adults (mean score ⫽ 54.5), t(46) ⫽ 3.63, p ⬍ .001. For the WAIS–R Vocabulary test, older adults showed higher scores (mean score ⫽ 59.8) than younger adults (mean score ⫽ 50.6), t(46) ⫽ 5.59, p ⬍ .001. Apparatus, stimuli, and procedure. The apparatus, stimuli, and procedure were similar to those in Experiment 1, except as noted in the following. Each participant received only one session. Task 1 involved a tone/noise discrimination task, like the AV condition of Experiment 1, except that there was no irrelevant shape. Task 2 was to decide whether boxes (1.4 cm in height, subtending a visual angle 0.20°) were narrow or wide. Two different narrow and wide boxes were used to form an easy and a difficult discrimination condition. For the easy condition, the narrow box was approximately 1 cm in width (subtending a visual angle of 0.17°), and the wide box was about 3.8 cm in width (subtending a visual angle of 0.66°). For the difficult condition, the narrow box was approximately 1.8 cm in width (subtending a visual angle of 0.29°), and the wide box was about 2.6 cm in width (subtending a visual angle of 0.435°). Participants were asked to press the ⬎ key with their right index finger for a narrow box and to press the ? key with their right middle finger for a wide box.

Results The data analysis was similar to that of Experiment 1. Application of the RT cutoff values eliminated 1.33% of trials for younger adults and 1.18% of trials for older adults. Data were analyzed as a function of age group (younger vs. older), Task 2 difficulty (easy vs. difficult), SOA (50, 100, 300, 500, 700, and 900 ms), and participants. RT data for Task 1 and Task 2 are shown in Figure 4, and PE data are shown in Table 3. For RT1, the main effect of age group was significant, F(1, 46) ⫽ 24.85, p ⬍ .001, MSE ⫽ 1,215,571; mean RT1 was 584 ms for younger adults and 813 ms for older adults. There was also a main effect of SOA, F(5, 230) ⫽ 18.74, p ⬍ .0001, MSE ⫽ 21,535, reflecting a slight increase in RT1 at short SOAs. This effect was more pronounced for older adults (see Figure 4), leading to a significant interaction between age group and SOA, F(5, 230) ⫽ 4.45, p ⬍ .001, MSE ⫽ 21,535. Age group also interacted with Task 2 difficulty, F(1, 46) ⫽ 4.83, p ⬍ .05, MSE ⫽ 17,187; RT1 was 8 ms longer in the easy condition (588 ms) than in the difficult condition (580 ms) for younger adults but was 16 ms shorter in the easy condition (805 ms) than in the difficult condition (821 ms) for older adults. The PE1 data showed a significant main effect of SOA, F(5, 230) ⫽ 13.38, p ⬍ .0001, MSE ⫽ 0.0027, with smaller PE1s at long SOAs (PE1s were .04, .03, .02, .02, .02, and .01, at the 50-, 100-, 300-, 500-, 700-, and 900-ms SOAs, respectively). No other effects were significant. For RT2, the main effect of age group was significant, F(1, 46) ⫽ 44.52, p ⬍ .0001, MSE ⫽ 1,311,322; mean RT2 was 674 ms for younger adults and 992 ms for older adults. There was also a main effect of SOA, F(5, 230) ⫽ 475.78, p ⬍ .0001, MSE ⫽ 32,866; the overall PRP effect was 474 ms. The interaction of age group and SOA was significant, F(5, 230) ⫽ 20.07, p ⬍ .0001, MSE ⫽ 32,866; the PRP effect was 375 ms for younger adults and 573 ms for older adults. Age group also interacted significantly with Task 2 difficulty, F(1, 46) ⫽ 10.00, p ⬍ .01, MSE ⫽ 36,981; the Task 2 difficulty effect on RT2 was larger for older adults (138 ms) than for younger adults (87 ms).

LIEN ET AL.

440

Figure 4. Mean response times (RTs) for Task 1 and Task 2 in Experiment 2 for younger adults and older adults as a function of stimulus onset asynchrony (SOA; 50, 100, 300, 500, 700, and 900 ms) and Task 2 difficulty (easy vs. difficult).

Although the Task 2 difficulty effects on RT2 were roughly additive with SOA (see Figure 4), the interaction of SOA and Task 2 difficulty was statistically significant, F(5, 230) ⫽ 2.58, p ⬍ .05, MSE ⫽ 11,643. The Task 2 difficulty effects were 89, 114, 106, 104, 121, and 141 ms at the 50-, 100-, 300-, 500-, 700-, and 900-ms SOAs, respectively. The three-way interaction of these variables with age group was not significant, F(5, 230) ⬍ 1.0. Separate data analyses for each age group revealed that the interaction of Task 2 difficulty and SOA failed to reach significance both for younger adults, F(5, 115) ⫽ 1.38, p ⫽ .24, MSE ⫽ 10,213, and older adults, F(5, 115) ⫽ 1.81, p ⫽ .12, MSE ⫽ 13,074. For younger adults, the effects of Task 2 difficulty were

80, 84, 78, 69, 94, and 118 ms at the 50-, 100-, 300-, 500-, 700-, and 900-ms SOAs. For older adults, the effects of Task 2 difficulty were 98, 144, 134, 138, 148, and 164 ms at the 50-, 100-, 300-, 500-, 700-, and 900-ms SOAs. Although the Task 2 Difficulty ⫻ SOA interaction (i.e., the slack effect) was not significantly modulated by age, it is important to ask whether there is a trend toward larger slack effects for older adults than for younger adults. To provide a stable measure of slack effects, we combined the two shortest SOAs (50 and 100 ms) and combined the two longest SOAs (700 and 900 ms). The percentage decrease in the Task 2 difficulty effect from the two longest SOAs to the two shortest SOAs was 22.46% for younger

Table 3 Proportion of Errors on Task 1 and Task 2 for Younger Adults and Older Adults in Experiment 2 as a Function of Stimulus Onset Asynchrony (SOA) and Task 2 Difficulty SOA Task and difficulty

50

100

300

500

700

900

Younger adults Task 1 Easy Difficult Task 2 Easy Difficult

.042 .045

.038 .039

.024 .025

.020 .022

.021 .024

.012 .020

.031 .118

.029 .111

.019 .107

.020 .084

.028 .107

.020 .084

Older adults Task 1 Easy Difficult Task 2 Easy Difficult Note.

.033 .028

.029 .027

.012 .017

.009 .014

.0141 .014

.012 .012

.011 .040

.011 .053

.013 .041

.014 .049

.013 .042

.019 .041

For the Task 1 data, the easy and difficult conditions refer to the difficulty of Task 2.

AGING AND WORD RECOGNITION

adults and 22.51% for older adults. Thus, the slack effect was nearly identical for both age groups. The PE2 data revealed a main effect of age group, F(1, 46) ⫽ 18.77, p ⬍ .0001, MSE ⫽ 0.0363, indicating that PE2 was .03 higher for younger adults (.06) than for older adults (.03). Age group interacted significantly with SOA, F(5, 230) ⫽ 2.81, p ⬍ .05, MSE ⫽ 0.0043. Separate data analyses for each group on PE2 showed no PRP effect for older adults, F(1, 23) ⬍ 1.0, but a significant PRP effect of .023 for younger adults, F(1, 23) ⫽ 2.73, p ⬍ .05, MSE ⫽ 0.0065. Age group also interacted significantly with Task 2 difficulty, F(1, 46) ⫽ 22.79, p ⬍ .0001, MSE ⫽ 0.0137; the effect of Task 2 difficulty on PE2 was higher for younger adults (.08) than for older adults (.03).

Discussion Unlike Experiment 1, both younger and older adults showed a roughly additive interaction between Task 2 box-width difficulty and SOA. The percentage decrease in the Task 2 difficulty effect from the two longest SOAs to the two shortest SOAs was about 22.5% for both younger and older adults. This finding is consistent with our hypothesis that age-related differences in parallel processing (like those observed in Experiment 1) are restricted to lexical processes (and perhaps other processes for which older adults have an experience advantage). For a converging measure of parallel processing, we calculated the general savings scores in the same manner as in Experiment 1. For this analysis we used data from the difficult condition only, just as we did in Experiment 1. The age difference in the general savings scores was not significant, F(1, 46) ⬍ 1.0; the general savings was 162 ms for younger adults and 199 ms for older adults. We also computed the proportional savings (general savings divided by the overall RT; Equation 2) to correct for the fact that older adults— by virtue of having longer overall RTs— had more opportunity to produce savings. Analyzed this way, the trend was actually for younger adults to show more savings (.13) than older adults (.11), although the difference did not reach significance, F(1, 46) ⬍ 1.0. In summary, the savings data provide converging evidence that older adults have no special advantage in parallel processing when Task 2 is nonlexical.

General Discussion In the present study, we have examined age-related differences in parallel processing using a PRP design. Both experiments revealed larger PRP effects for older adults than for younger adults, replicating earlier aging studies (e.g., Allen, Smith, Vires-Collins, & Sperry, 1998; Glass et al., 2000; Hartley, 2001; Hartley & Little, 1999). Although this finding might appear to suggest that dual-task ability declines with age, a closer analysis suggests otherwise. According to the central bottleneck model, older adults’ longer RT1 should produce larger PRP effects even if older adults have no dual-task processing deficits. Indeed, the observed increase in PRP effects for older adults relative to younger adults was no larger than expected on the basis of the increase in RT1. Thus, there is no evidence in the overall data from the present experiments for a dual-task processing deficit in older adults. Furthermore, a finer grained analysis of the data suggests that there were isolated processes (e.g., word recognition) for which older adults actually showed improved dual-task processing.

441

To determine whether people can recognize a Task 2 word while central attention is devoted to Task 1, in Experiment 1 we factorially manipulated word recognition difficulty (by using low- and high-frequency words) and SOA. For younger adults, these variables produced additive effects in both the AV condition, replicating McCann et al. (2000), and in the VV condition. On the basis of locus-of-slack logic and the central bottleneck model, these results indicate that word recognition did not occur in parallel with Task 1 central stages. For older adults, however, there was an underadditive interaction between word frequency and SOA in both the AV and the VV conditions. This result replicates Allen et al.’s (2002) aging study using the VV condition and extends their finding to the AV condition. These results indicate that older adults are able to carry out word recognition in parallel with Task 1 but that younger adults are not.

The Effect of Input Modality and Spatial Attention In the introduction, we discussed the hypothesis that parallel processing of word recognition with another task depends critically on the input modality of that task. When Task 1 is presented visually in the same general location as the Task 2 word/nonword (as in Allen et al., 2002), spatial attention to the Task 1 stimulus might be directed to the Task 2 word/nonword as well. When the Task 1 stimulus is auditory (as in McCann et al., 2000), however, participants might intentionally direct spatial attention away from the location of the word/nonword of Task 2 to minimize potential interference from Task 2 processing. Diversion of spatial attention is a critical issue, because previous studies have provided evidence that words cannot be recognized without spatial attention (see Lachter et al., 2004; McCann et al., 1992). Thus, failures to recognize words without central attention in the AV condition of McCann et al. (2000) could have been due to the lack of spatial attention rather than the lack of central attention. In the present study, we have found no evidence for this hypothesis. For younger adults, the effects of word frequency and SOA were additive in the VV condition, even though spatial attention was presumably directed to the Task 2 word/nonword. Furthermore, older adults showed an underadditive interaction in the AV condition, even though spatial attention could have been directed away from the Task 2 word/nonword. Although spatial attention might be required for word recognition, we conclude that it was not a critical factor in our study, presumably because the words were spatially attended in both the AV and the VV conditions.

The Effect of Cognitive Aging The other hypothesis we have considered in the present study is that the ability to perform word recognition without central attention improves with age. Allen et al. (2002) found that older adults produced a stronger underadditive interaction between word frequency and SOA than did younger adults. The results were not entirely conclusive, however, because the authors failed to find a difference between older and younger adults in general savings from parallel processing. Furthermore, Allen et al. examined only the VV condition. In the present study, we have replicated Allen et al.’s (2002) finding of a stronger underadditive interaction between the word frequency effect and SOA for older adults than for younger adults and have shown that this finding does not depend

442

LIEN ET AL.

on Task 1 input modality. These findings provide further evidence that older adults are able to carry out word recognition without central attention, whereas younger adults are not. To provide converging evidence for this conclusion, we examined general savings (a measure of the overall degree of parallel processing) in Experiment 1 for both younger and older adults (see Equation 1). This analysis confirmed that older adults showed larger savings (434 ms) than younger adults (193 ms). The proportional general savings (Equation 2) also showed a similar trend (.20 for older adults and .14 for younger adults). The reasons we found an age-related difference in savings, whereas Allen et al. (2002) did not, might be that (a) we successfully deterred response grouping, which allowed us to obtain a truer measure of the time needed to complete Task 1 and Task 2 at long SOAs; (b) we included a shorter SOA, providing more opportunity for parallel processing and more opportunity to observe savings; and (c) we focused on low-frequency words, which should benefit the most from parallelism. In summary, we have obtained converging evidence that although older adults exhibit a longer bottleneck delay than younger adults (i.e., a larger PRP effect), older adults are better able to use this time to perform Task 2 word recognition (producing larger proportional savings). This conclusion is surprising, because it is rare for cognitive processing to actually improve with age. Our claim that older adults are better able than younger adults to perform word recognition in parallel with Task 1 might appear to conflict with the finding that older adults produced larger PRP effects. However, we are not claiming that older adults necessarily have a greater ability, in general, to perform tasks in parallel or that they have eliminated the central bottleneck. The large overall PRP effects suggest that there was still a central bottleneck—Task 2 response selection still must wait for Task 1 response selection to finish. Because older adults select responses to Task 1 more slowly, they would be expected to produce longer bottleneck delays and, hence, larger PRP effects. The hypothesized ability of older adults (but not younger adults) to perform Task 2 word recognition in parallel with Task 1 should reduce these age-related differences in the PRP effect but not entirely eliminate them. This analysis can explain why, for low-frequency words in Experiment 1, the difference in PRP effects (130 ms) between older adults and younger adults was much smaller than one would have expected given the large difference in mean RT1 (348 ms). Consequently, the big picture of cognitive aging in visual word recognition appears to reflect a complex interplay of stage-specific advantages and disadvantages for older adults compared with younger adults.

Source of Age-Related Differences in Parallel Processing Experiment 1 provides evidence that (a) it is possible to carry out word recognition without central attention and (b) whether word recognition requires central attention depends on age, not on Task 1 input modality. One logical follow-up question is why older adults typically can carry out word recognition for Task 2 in parallel with Task 1 but younger adults typically cannot. In the following sections, we discuss three possible explanations for this finding. Blocking. To minimize interference from Task 2 processing into Task 1 processing, participants might attempt to prevent word recognition from occurring during cognitive slack (see McCann et al., 2000). If the blocking is successful, then the interaction be-

tween word frequency and SOA on Task 2 should be additive, as evident in the younger adults’ data. Older adults, however, might have difficulty imposing a complete block on word recognition during cognitive slack and thus might produce an underadditive interaction between word frequency and SOA. According to this view, the increased parallel processing for older adults in Experiment 1 was due not to a rare processing advantage but to yet another cognitive deficit (reducing blocking ability). This blocking hypothesis is consistent with the findings of previous aging studies that older adults are less able than younger adults to inhibit or suppress irrelevant stimuli (e.g., older adults show a larger Stroop effect than younger adults; Hasher & Zacks, 1988; Zacks, Hasher, & Li, 2000). In particular, several aging studies explicitly suggested that inhibitory mechanisms for older adults fail to limit the presence of irrelevant information in working memory during encoding and retrieval of target information (e.g., Spieler, Balota, & Faust, 1996). In our study, word recognition was relevant to Task 2 but was irrelevant to Task 1. Consequently, older adults might have had difficulty preventing Task 2 word recognition from occurring while Task 1 was being carried out. If so, then word recognition could occur in parallel with Task 1 for older adults, producing an underadditive interaction between word frequency and SOA. Although this blocking explanation might account for many of our findings, there are several lines of evidence against it. First, the theory that older adults have a general inhibitory deficit is not always supported (for a summary, see McDowd & Shaw, 2000, pp. 268 –272). Second, if older adults have blocking problems, in general, then they should show underadditive interactions between SOA and other manipulations of Task 2 difficulty as well, even when younger adults show additive interactions between the same factors. The present Experiment 2, however, shows that is not always the case. This experiment used a nonlexical Task 2 involving a box-width judgment. To manipulate categorization time, we varied the width of the boxes. Although both groups showed a modest amount of underadditivity, which suggests some parallel processing, the amount was roughly equal for younger and older adults (22.5%). Furthermore, younger and older adults showed similar amounts of general savings. These findings are also supported by Maquestiaux, Hartley, and Bertsch (2004), who found an additive interaction between Task 2 stimulus–response compatibility (compatible vs. incompatible mapping) and SOA for both younger and older adults. These authors also found a strong underadditive interaction between letter contrast and SOA for younger and older adults, which suggests that neither group blocked letter categorization. Although it is impossible to rule out differential amounts of blocking specifically for lexical processing, there is no evidence for a general inability of older adults to block out Task 2 processing. Task coordination strategies. Another possible explanation for the findings of Experiment 1 is that older adults adopted different task coordination strategies (Meyer & Kieras, 1997). Whereas younger adults might have chosen to strategically defer word recognition, older adults might have allowed it take place in parallel with Task 1 (deferring at some later stage, e.g., response selection). This account is similar in many respects to the blocking account and thus has some of the same deficiencies. In particular, if older adults had adopted an aggressive task coordination strategy in Experiment 1, one might have expected them to also adopt an aggressive strategy in Experiment 2. Furthermore, Glass et al.

AGING AND WORD RECOGNITION

(2000) studied age differences in dual-task processing and actually reached the opposite (and more plausible) conclusion—that older adults adopted more cautious task coordination strategies than younger adults. Reading ability. Perhaps the most parsimonious interpretation of the present results is that visual word recognition and reading are skills that develop over time (i.e., as the cumulative experience with words increases). As word recognition skill reaches a certain point (obtainable by many older adults but not by many younger adults), individuals are capable of performing word recognition in parallel with other tasks. That is, older adults are able to begin lexical access for Task 2 even before central processing for Task 1 has been completed. For younger adults, conversely, lexical access still requires limited central resources, and thus Task 2 lexical access cannot take place before Task 1 central processing is complete. As a follow-up test of this hypothesis, we computed the correlation between each participant’s general savings score (a measure of parallel processing) and his or her WAIS–R Vocabulary score in Experiment 1. Consistent with the reading ability hypothesis, the correlation was strong for older adults (.516 for general savings and .534 for proportional savings; ps ⬍ .05); the best readers produced the largest savings. For younger adults, there was essentially no relation between savings and vocabulary scores (.005 for general savings and .014 for proportional savings; ps ⬎ .90). It may be that a great amount of cumulative experience with words is necessary before word recognition is truly automatic, even for good readers. On the basis of this reading ability hypothesis, one would also predict little correlation between general savings scores and WAIS–R scores in Experiment 2, in which a nonlexical Task 2 was used. As expected, the analyses revealed no significant correlations between general savings scores and WAIS–R scores for both older adults (.117 for general savings and .089 for proportional savings; ps ⬎ .60) and younger adults (⫺.228 for general savings and ⫺.303 for proportional savings; ps ⬎ .15). Although these correlational data from Experiments 1 and 2 are consistent with our reading ability hypothesis, it is premature to reach firm conclusions on the basis of these data alone. Further work is needed with larger samples of younger adults and a wider range of vocabulary scores.

Relations to Stroop Studies Our claim that lexical access requires central attention for younger adults appears to conflict with the conclusions from Stroop studies (see MacLeod, 1991, for a review). In the typical single-task version of the Stroop paradigm, younger adults cannot avoid reading the irrelevant word while they are naming the color in which the word is printed. Thus, contrary to our findings, Stroop data suggest that younger adults can recognize words while attention is devoted to another central operation. There are two ways to reconcile these seemingly discrepant findings. First, note that the studies of the Stroop effect typically use three or four relatively high-frequency color words (e.g., red, green, blue, and yellow) that are repeated many times during the experiment. It is therefore possible that lexical access can take place without central attention (even for younger adults) for words that have been repeatedly accessed in the recent past (and thus are highly activated) but not for less activated words in the person’s

443

lexicon. That is, we are suggesting that word recognition may occur without central attention in the Stroop task (in which each word is repeated many times), but not in lexical decision tasks such as ours (in which each word was presented only once). Second, when central attention is devoted to the color-naming task in the Stroop paradigm, it might accidentally slip to the irrelevant word name. That is, it might be difficult for central attention to name colors without also naming color words, because these operations are so closely related (see Lien, Ruthruff, Hsieh, & Yu, in press). When the central operations are more distinct, as in the present study (shape or tone classification for Task 1 vs. lexical decision for Task 2), central attention may be devoted exclusively to one operation or the other (e.g., to Task 1 but not Task 2). Further work is needed to test these two possible reconciliations.

Conclusion The present study examined under what conditions, if any, word recognition can proceed without central attention. We found evidence that, regardless of Task 1 input modality, word recognition on Task 2 can proceed while central attention is devoted to Task 1 for older adults but not for younger adults. One line of evidence was based on locus-of-slack logic: Older adults showed a strong underadditive interaction between word frequency and SOA, whereas younger adults did not. The second line of evidence was based on the fact that older adults showed larger general savings, a measure of the overall degree of parallel processing. These findings challenge the idea that older adults slow down because they suffer from a general resource decrement. Experiment 2, in which the lexical Task 2 was replaced by a nonlexical one (a box-width judgment), shows no evidence for improved dual-task abilities in older adults. Superior dual-task processing for older adults, therefore, appears to be restricted to processes for which older adults have much greater cumulative experience, such as lexical processes.

References Allen, P. A., Lien, M.-C., Murphy, M. D., Sanders, R. E., Judge, K. S., & McCann, S. R. (2002). Age differences in overlapping-task performance: Evidence for efficient parallel processing in older adults. Psychology and Aging, 17, 505–519. Allen, P. A., Smith, A. F., Vires-Collins, H., & Sperry, S. (1998). The psychological refractory period: Evidence for age differences in attentional time-sharing. Psychology and Aging, 13, 218 –229. Allen, P. A., Wallace, B., & Weber, T. A. (1995). Influence of case type, word frequency, and exposure duration on visual word recognition. Journal of Experimental Psychology: Human Perception and Performance, 21, 914 –934. Borger, R. (1963). The refractory period and serial choice-reactions. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 15(A), 1–12. Brown, T. L., Roos-Gilbert, L., & Carr, T. H. (1995). Automaticity and word perception: Evidence from Stroop and Stroop dilution effects. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 1395–1411. Cerella, J. (1990). Aging and information-processing rates. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (3rd ed., pp. 201–221). New York: Academic Press. Craik, F. I. M., & Salthouse, T. A. (Eds.). (2000). The handbook of aging and cognition (2nd ed.). Hillsdale, NJ: Erlbaum.

444

LIEN ET AL.

Glass, J. M., Schumacher, E. H., Lauber, E. J., Zurbriggen, E. L., Gmeindl, L., Kieras, D. E., & Meyer, D. E. (2000). Aging and the psychological refractory period: Task-coordination strategies in young and old adults. Psychology and Aging, 15, 571–595. Hartley, A. A. (2001). Age differences in dual-task interference are localized to response-generation processes. Psychology and Aging, 16, 47– 54. Hartley, A. A., & Little, D. M. (1999). Age-related differences and similarities in dual-task interference. Journal of Experimental Psychology: General, 128, 416 – 449. Hasher, L., & Zacks, R. T. (1979). Automatic and effortful processes in memory. Journal of Experimental Psychology: General, 108, 356 –388. Hasher, L., & Zacks, R. T. (1988). Working memory, comprehension, and aging: A review and a new view. In G. Bower (Ed.), The psychology of learning and motivation (Vol. 22, pp. 193–225). New York: Academic Press. Hazeltine, E., Ruthruff, E., & Remington, R. W. (2006). The role of input and output modality pairings in dual-task performance: Evidence for content-dependent central interference. Cognitive Psychology, 52, 291–345. Hazeltine, E., Teague, D., & Ivry, R. B. (2002). Simultaneous dual-task performance reveals parallel response selection. Journal of Experimental Psychology: Human Perception and Performance, 28, 527–545. Johnston, J. C., & McCann, R. S. (2006). On the locus of dual-task interference: Is there a bottleneck at the stimulus classification stage? Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 59(A), 694 –719. Johnston, J. C., McCann, R. S., & Remington, R. W. (1995). Chronometric evidence for two types of attention. Psychological Science, 6, 365–369. Kucˇera, H., & Francis, W. N. (1967). Computational analysis of presentday American English. Providence, RI: Brown University Press. Lachter, J., Forster, K. I., & Ruthruff, E. (2004). Forty years after Broadbent: Still no identification without attention. Psychological Review, 111, 880 –913. Lachter, J., Ruthruff, E., Lien, M.-C., & McCann, R. S. (2006). Is attention needed for object identification? Evidence from the Stroop paradigm. Manuscript submitted for publication. Lien, M.-C., McCann, R. E., Ruthruff, E., & Proctor, R. W. (2005). Dual-task performance with ideomotor-compatible tasks: Is the central processing bottleneck intact, bypassed, or shifted in locus? Journal of Experimental Psychology: Human Perception and Performance, 31, 122–144. Lien, M.-C., & Proctor, R. W. (2002). Stimulus-response compatibility and psychological refractory period effects: Implications for response selection. Psychonomic Bulletin & Review, 9, 212–238. Lien, M.-C., Proctor, R. W., & Allen, P. A. (2002). Ideomotor compatibility in the psychological refractory period effect: Twenty-nine years of oversimplification. Journal of Experimental Psychology: Human Perception and Performance, 28, 396 – 409. Lien, M.-C., Ruthruff, E., Hsieh, S.-L., & Yu, Y.-T. (in press). Parallel central processing between tasks: Evidence from lateralized readiness potentials. Psychonomic Bulletin & Review. Logan, G. D., & Schulkind, M. D. (2000). Parallel memory retrieval in dual-task situations: I. Semantic memory. Journal of Experimental Psychology: Human Perception and Performance, 26, 1260 –1280. MacLeod, C. M. (1991). Half a century of research on the Stroop effect: An integrative review. Psychological Bulletin, 109, 163–203. Maquestiaux, F., Hartley, A., & Bertsch, J. (2004). Can practice overcome age-related differences in the psychological refractory period effect? Psychology and Aging, 19, 649 – 667.

McCann, R. S., Folk, C. L., & Johnston, J. C. (1992). The role of spatial attention in visual word processing. Journal of Experimental Psychology: Human Perception and Performance, 18, 1015–1029. McCann, R. S., & Johnston, J. C. (1992). Locus of the single-channel bottleneck in dual-task interference. Journal of Experimental Psychology: Human Perception and Performance, 18, 471– 484. McCann, R. S., Remington, R. W., & Van Selst, M. (2000). A dual-task investigation of automaticity in visual word processing. Journal of Experimental Psychology: Human Perception and Performance, 26, 1352–1370. McDowd, J. M., & Shaw. R. J. (2000). Attention and aging: A functional perspective. In F. I. M. Craik & T. A. Salthouse (Eds.), The handbook of aging and cognition (2nd ed., pp. 221–292). Hillsdale, NJ: Erlbaum. Meyer, D. E., & Kieras, D. E. (1997). A computational theory of executive cognitive processes and multiple-task performance: Part 2. Accounts of psychological refractory-period phenomena. Psychological Review, 104, 749 –791. Pashler, H. (1984). Processing stages in overlapping tasks. Journal of Experimental Psychology: Human Perception and Performance, 10, 358 –377. Pashler, H. (1994). Dual-task interference in simple tasks: Data and theory. Psychological Bulletin, 116, 220 –244. Pashler, H., & Johnston, J. C. (1989). Chronometric evidence for central postponement in temporally overlapping tasks. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 41(A), 19 – 45. Ruthruff, E., Hazeltine, E., & Remington, R. W. (in press). What causes residual dual-task cost after practice? Psychological Research. Ruthruff, E., Pashler, H., & Klaassen, A. (2001). Processing bottlenecks in dual-task performance: Structural limitation or voluntary postponement? Psychonomic Bulletin & Review, 8, 73– 80. Ruthruff, E., Van Selst, M., Johnston, J. C., & Remington, R. W. (2006). How does practice reduce dual-task interference: Integration, automatization, or simply stage-shortening? Psychological Research, 70, 125– 142. Salthouse, T. A. (1996). The processing-speed theory of adult age differences in cognition. Psychological Review, 103, 403– 428. Schneider, W., Eschman, A., & Zuccolotto, A. (2002). E-Prime reference guide. Pittsburgh, PA: Psychology Software Tools. Schumacher, E. H., Seymour, T. L., Glass, J. M., Fencsik, D., Lauber, E., Kieras, D. E., & Meyer, D. E. (2001). Virtually perfect time sharing in dual-task performance: Uncorking the central attentional bottleneck. Psychological Science, 12, 101–108. Schweickert, R. (1978). A critical path generalization of the additive factors method: Analysis of a Stroop task. Journal of Mathematical Psychology, 18, 105–139. Spieler, D. H., & Balota, D. A., & Faust, M. E. (1996). Stroop performance in healthy younger and older adults and in individuals with dementia of the Alzheimer’s type. Journal of Experimental Psychology: Human Perception and Performance, 22, 461– 479. Telford, C. W. (1931). Refractory phase of voluntary and associative response. Journal of Experimental Psychology, 14, 1–35. Wechsler, D. (1981). Wechsler Adult Intelligence Scale—Revised. New York: Psychological Corporation. Welford, A. T. (1952). The “psychological refractory period” and the timing of high speed performance. British Journal of Psychology, 43, 2–19. Zacks, R. T., Hasher, L., & Li, K. Z. H. (2000). Human memory. In F. I. M. Craik & T. A. Salthouse (Eds.), The handbook of aging and cognition (2nd ed., pp. 293–358). Hillsdale, NJ: Erlbaum.

AGING AND WORD RECOGNITION

445

Appendix A Word/Nonword List 1 Four letters Words

No. occurrences

Five letters Nonwords

Words

No. occurrences

Six letters Nonwords

Words

No. occurrences

Nonwords

thase never houne foung woter thonk naght goint luter poirt groip givun whote lorge leest paweq thung lighe bigar sunse whule timas humen hends todey tuken deeth wirds fielt shilh miney whise majoy hearg stody knawn

before should during school number course always though public system better called toward social rather second things looked become within church seemed family turned matter having really period cannot behind making became street result reason change

1,016 888 585 492 472 465 458 442 438 416 414 401 386 380 377 373 368 367 361 359 348 332 331 320 308 279 275 265 258 258 255 246 244 244 241 240

befare stould doring schoon namber coulse alweys thoigh pablic systam botter callen towerd sociar rathen secand thengs looken bacome wathin charch seeted fomily turnel motter hoving rearly perios cannol behind meking becime streel resilt realon chenge

bried splin serit pourd gried scole beam pount baint rigil draft troce chast panil demay newer lebby lavel twest chass runch

reveal reform shorts dining colony combat cellar mutual lonely phases barrel select motive nerves merger jungle vacuum filling submit denial basket

30 30 29 28 28 27 26 26 25 24 24 23 22 22 21 20 20 19 18 18 17

remeal reford sharts bining calony combal cellan mutuad monely phasen burrel selece rotive derves ferger jengle wacuum liling submid denian lasket

High frequency must know last year come take went part left less find next give days room face case need four best mind done kind help name past feet body half word tell held sure keep free miss

1,013 683 676 660 630 611 507 500 490 438 399 394 391 384 383 371 362 360 359 351 325 320 313 311 294 291 283 276 275 274 268 264 264 264 260 258

mest knaw lasc yeal coms teke wunt pirt laft liss fint naxt guve deys rool foce cuse neel foun bost mond dobe kond halp nime pust feen bidy hald werd telt helt sare keed fren misk

those never house found water think night going later point group given white large least power thing light began sense whole times human hands today taken death words field shall money whose major heard study known

850 698 591 536 442 433 411 399 397 395 390 377 365 361 343 342 333 333 312 311 309 300 299 289 284 281 277 274 274 267 265 252 247 247 246 245 Low frequency

cuts meal barn sand cure chin gear wars pack jump bend lock tire herd beam trim slim span bell fame fuel

30 30 29 28 28 27 26 26 25 24 24 23 22 22 21 20 20 19 18 18 17

culs meaf barl sant bure chib geaf warb kack jume benk loce tirt hird leam trid slib spen beld fime muel

cried split merit pound dried scope beard mount faint rigid draft trace chart panic delay newer lobby label twist chase bunch

30 30 29 28 28 27 26 26 25 24 24 23 22 22 21 20 20 19 18 18 17

(Appendixes continue)

LIEN ET AL.

446

Appendix A (continued) Four letters Words

No. occurrences

Five letters Nonwords

Words

No. occurrences

Six letters Nonwords

Words

No. occurrences

Nonwords

refuse locate client valued superb gossip nuclei ballot cavity herald retain cement reject mister boiled

16 16 15 14 14 13 13 12 12 11 11 11 10 10 10

refust locade plient vanued superd gossil nucler hallot lavity hurald retair lement roject misten goiled

Low frequency (continued) damp kick rope shoe weep lime grin cane boil mill tide dash rust knit bolt Note.

16 16 15 14 14 13 13 12 12 11 11 11 10 10 10

domp kict rone shog welf lims gran cune doil cill mide dast hust knat solt

liver lined sweep gloom stare tower media bloom dimly haven wired blond theft lunar bacon

16 16 15 14 14 13 13 12 12 11 11 11 10 10 10

civer linet sween gloop stabe towen medil bloot gimly haren wiren flond thaft lunor sacon

No. occurrences refers to the occurrences in the Kucˇera and Francis (1967) frequency norms.

Appendix B Word/Nonword List 2 Four letters Words

No. occurrences

Five letters Nonwords

Words

No. occurrences

Six letters Nonwords

Words

closk voick frong cleab musit parth clasm norto leavn soung blace valum linev peack womet thirn basiv spaco moven wrotd browl costh hourg hearf staga milew makex trieo showe rivey doinp terme rangx brinf flooq finaz

future wanted center common policy figure strong modern living longer nature months needed ground values father spirit return recent beyong forces person taking report coming single simply trying higher walked passed county growth market police likely

No. occurrences

Nonwords

High frequency west turn cost seem wife girl town rate plan hard play near road gone book call fire kept view dark late hope live data read lost move hold sort rest care fine wall cent talk hall

235 233 229 229 228 220 212 209 205 202 200 198 197 195 193 188 187 186 186 185 179 178 177 173 173 171 171 169 164 163 162 161 160 158 154 152

wesp turb cosk seef wifs girb towb ratz plab harb plar neac roaj gonk bool calt firg kepd vieb darp latk hoce livn datu reaa losi movs holp sorn resb carv finu wals cenw tald halm

close voice front clear music party class north leave sound black value lines peace women third basis space moved wrote brown costs hours heart stage miles makes tried shown river doing terms range bring floor final

236 226 221 219 216 216 207 206 205 204 203 200 198 198 192 190 184 184 181 181 176 176 175 174 174 173 172 170 166 165 163 163 160 158 158 156

227 226 224 223 222 209 202 198 194 193 191 189 187 186 186 183 182 180 179 175 175 175 175 174 174 172 170 163 161 159 157 155 155 155 155 151

futurn wantel centen commor polict figurk strond moderl livink longel naturm monthy needeh grounp valuez fathei spiric returk recenk beyont forceq persoy takint reporr comina markej simplo tryinr highew walkex passel countj singlg growti polics likelz

AGING AND WORD RECOGNITION

447

Appendix B (continued) Four letters Words

No. occurrences

Five letters Nonwords

Words

No. occurrences

Six letters Nonwords

Words

No. occurrences

Nonwords

fance strig filse laigh nowly fiben storg trean deval brove swint nober deult fauld glort cheel slode crawn brict squar grast spran selar cubec prise grive steeg glabe rivan loisy glize joice moind creet foral fally

secure virtue guilty paused profit gentle repeat paying critic sitter wholly genius thrust wealth linear gather pursue timber buried ranged hunger closet finest picnic custom breeze sprang derive saloon deduct tender census shaken detect button poison

30 30 29 28 28 27 26 26 25 24 24 23 22 22 21 20 20 19 18 18 17 16 16 15 14 14 13 13 12 12 11 11 11 10 10 10

securt virtun gailty pausal profin gantle repead poying critil sitten whelly ganius thrusp weilth lineam gathed pursur tember baried renged honger closel finost pacnic castom breate sprant derave salood dedict tenden censul shiken ditect buttan poisor

Low frequency lane flux stem dawn fled dull card fist harm bore rank plug palm soap neat cape nest fury colt pole seal grab toll drag bark hawk cone pint tank deaf cult plea rake fake vent slug Note.

30 30 29 28 28 27 26 26 25 24 24 23 22 22 21 20 20 19 18 18 17 16 16 15 14 14 13 13 12 12 11 11 11 10 10 10

lene flix stam dawl fleg dall cerd fost horm bort renk plut pelm soal nead cipe nast furt calt polt sead grap tolt drad birk hawe cene pind tant deam culp prea roke fike vont slun

fence strip false laugh newly fiber storm treat devil brave swing nobel dealt fault glory cheek slide crown brick squad grasp spray solar cubic prose grove steep globe rival lousy glaze juice mound creep forum folly

30 30 29 28 28 27 26 26 25 24 24 23 22 22 21 20 20 19 18 18 17 16 16 15 14 14 13 13 12 12 11 11 11 10 10 10

No. occurrences refers to the occurrences in the Kucˇera and Francis (1967) frequency norms.

Appendix C Generalized Slowing and General Savings General savings ⫽ 共RT1 Longest SOA ⫹ RT 2Longest SOA兲 ⫺ 共SOAShortest ⫹ RT 2Shortest SOA兲. According to the central bottleneck model, Equation 1 can be rewritten as follows: General savings ⫽ 关共1A ⫹ 1B ⫹ 1C兲 ⫹ 共2A ⫹ 2B ⫹ 2C兲兴 ⫺ 关共SOAShortest兲 ⫹ 共1A ⫹ 1B ⫺ SOAShortest ⫹ 2B ⫹ 2C兲兴. In this equation, 1A, 1B, and 1C correspond to the prebottleneck (e.g., perceptual), bottleneck (e.g., response selection), and postbottleneck (e.g.,

response initiation) stages of Task 1, respectively, and 2A, 2B, and 2C correspond to the same processes for Task 2. This equation simplifies to General savings ⫽ 1C ⫹ 2A. Given a generalized slowing of all stages for older adults (e.g., Stages A, B, and C) by a multiplicative factor k, the general savings for older adults will be k times larger than the general savings for younger adults.

Received February 17, 2004 Revision received April 26, 2006 Accepted May 15, 2006 䡲