eif; 01jun99 - Semantic Scholar

Aging, Neuropsychology, and Cognition 1999, Vol. 6, No. 2, pp. 117-140

1382-5585/99/0602-117$15.00 © Swets & Zeitlinger

Aging and the Development of Automaticity in Feature Search* Benedict T. Anandam and Charles T. Scialfa University of Calgary

ABSTRACT A training experiment was conducted to investigate age differences in the learning that occurs when observers search for an orientation-defined target among a homogeneous set of distractors (i.e., feature search). Eighteen young (M age = 26 years) and seventeen old (M age = 62 years) participants completed seven practice sessions (3,024 trials), followed by a single session of full reversal. Training involved consistent mapping, varied mapping, or nonsearch, in which a precue predicted the target location with 100% validity. Younger and older observers demonstrated equivalent learning rates and equivalent disruption following reversal in all training conditions. Results are interpreted within models of visual attention and search.

Even though we may wish it to be so, practice seldom makes perfect. Practice does, however, allow us to improve our performance level; this is true in our youth as well as in later life. One area in which practice-based changes in performance have been studied extensively is that of visual search, the processes by which we localize, detect, and identify salient objects, often in visually cluttered environments. In the present study, we compared older and younger observers engaged in a feature search task, wherein the search target was defined as the object having a unique orientation in a field of otherwise homogeneous objects. The purpose of the study was to determine if the two age groups developed automatic responses to the target. Practice-induced change in the attentional demands of visual search have been reported for conjunctions of contrast polarity and shape (Theeuwes & Kooi, 1994), brightness and orien*

tation (Nakayama & Mackeben, 1989), and some combinations of stereopsis, orientation, vernier offset, and lateral separation (Steinman, 1987). Friedman-Hill and Wolfe (1995) found that display size effects were halved in as few as 1,000 conjunction search trials. As well, Nakayama and Mackeben (1989, Experiment 1), Scialfa and Joffe (1998, Experiment 3), and Scialfa, Jenkins, Hamaluk, and Skaloud (1999) reported that practice produces reductions in both the reaction times (RTs) and eye movements required for conjunction search. While Kramer, Martin-Emerson, Larish, and Anderson (1996, Experiment 2) did not find a significant reduction in display size effects for form by movement conjunction search, their observers’ search slopes were relatively shallow in the first session of practice, leaving little room for further reduction. In contrast to these findings for conjunction search, practice at easy feature

This research was carried out in partial fulfillment of the MSc requirements of Ben Anandam and was supported by a grant from the Natural Science and Engineering Research Council of Canada. The authors would like to thank Sharmini, Nisha, and Trisha Anandam for their loving support, along with Ken Joffe and David Thomas for their advice and comments during the development of the research. Thanks also to Frank Schieber, who provided the code for the acuity charts. Address correspondence to: Chip Scialfa, Department of Psychology, University of Calgary, Calgary, AB T2N 1N4, CANADA. E-mail: [email protected]. Accepted for publication: June 2, 1999.

118

B. ANANDAM AND C. SCIALFA

search tasks can produce general improvements in response latency (Ahassir & Hochstein, 1997; Kramer et al., 1996), but does not result in a reduction in the generally small display size effects obtained. It is not known if practice results in a diminished display size effect in more difficult feature search. Although some visual search studies have found that older adults benefit from practice as much as their younger counterparts (Kramer et al., 1996; Madden & Nebes, 1980; Salthouse & Somberg, 1982), work by Rogers, Fisk and their colleagues (Fisk, McGee, & Giambra, 1988; Fisk & Rogers, 1991; Fisk, Rogers & Giambra, 1990; Rogers, 1992; Rogers & Fisk, 1991; Rogers, Fisk, & Hertzog, 1994) has resulted in different conclusions. This latter set of investigations has relied strongly on the semantic category search task, and has generally supported the view that older adults have difficulty selectively allocating attention to target items. However, the substantial cognitive differences between semantic category search and visual search prompt the question of whether older adults can develop automatic responses to targets in a relatively difficult feature search task. Therefore, in the present study, younger and older observers were asked to search for an orientation-defined target embedded in varying numbers of homogeneous distractors. They practiced the search task under either consistent mapping (CM), varied mapping (VM), or nonsearch (NS) conditions. The questions addressed in the research were as follows. Does CM practice result in a reduction of the display size effect and if so, are these practice effects related to age? Do older adults show less disruption at reversal following CM training, as would be anticipated under the priority learning deficit hypothesis (cf. Rogers, 1992)? Are there performance improvements in VM training that serve to reduce the display size effect and if so, what are the implications of these changes for models of visual attention? Finally, are there age differences in a nonsearch detection task where priority learning is thought to be of minimal importance to practice-based improvement in performance?

Across various experimental tasks, substantial performance improvements have been observed in older adults as a result of practice. For example, Salthouse and Somberg (1982) examined young and old adults who underwent more than 50 hours of practice in signal detection, memory search, and visual search tasks. Older adults showed significant improvements in all three tasks, which is not surprising, and in both signal detection and memory search they made an even greater absolute improvement than did young people. Notwithstanding, significant age differences in absolute performance levels remained. Madden and Nebes (1980) had young and old participants undergo extended practice (2,592 trials) in a hybrid memory/visual search task, wherein both memory and display set size were greater than one. At the end of practice, both age groups continued to be affected by increases in set size, and the older group was affected to a significantly greater extent. There was a significant decline in the search set slopes for both age groups, but there was no age by practice interaction, suggesting that young and old adults benefited equally from practice. Similar to the Salthouse and Somberg (1982) investigation, however, age differences in the speed of search remained. It is not only in relatively simple perceptual tasks that this pattern of age constancy in performance improvement is seen. Charness and Campbell (1988) taught young, middle-aged, and old adults an algorithm to mentally square two-digit numbers. The expected age differences in mental calculation speed were observed among the three age groups over all the five sessions of practice. However, there was comparable improvement in RT among all groups, signifying equivalent practice benefits across the age groups. Recently, Scialfa et al. (1999) reported that older and younger observers showed equivalent development of automaticity in two orientation by contrast polarity conjunction search experiments. The two groups learned at the same rates and their eye movements showed feature-based selection to equal degrees. Disruption at transfer

AUTOMATICITY AND FEATURE SEARCH

was measured to index learning, which may show age deficits even when performance improvement rate is unrelated to aging. The amount of disruption seen at reversal was the same for older and younger observers. Thus, age constancy in the attentional mechanisms underlying practice effects has been seen in visual search as well as other cognitive and perceptual tasks. On the other hand, research generally involving semantic category search has produced very different results. Age differences in search skill acquisition have been examined extensively by Fisk and his associates (Fisk et al., 1988; Fisk et al., 1990). Fisk et al. (1988, Experiment 1) studied the effects of CM and VM semantic category search training in groups of young, middle-aged, and old adults. All participants were assigned three categories as the CM set and six categories as the VM set. A trial consisted of the person first studying a memory set which could have 1, 2, or 3 categories. The probe display was then presented. It consisted of two exemplars, each from a different semantic category. The task was to make a keyboard response as to whether an exemplar belonging to the memory set was displayed. There were 8,400 trials with CM and VM training manipulated between blocks of trials. As expected, CM performance was much faster than VM performance across all age groups. For the young people, performance in the CM task resulted in display size slopes that were not significantly different than zero. This was not the case for the old group. Moreover, the percentage improvement in their CM versus VM performance was lower than in the young group. Fisk et al. (1988) concluded that although old adults benefited from both practice and CM search, automatic detection of the stimuli did not occur for them as it did for young people. Fisk et al. (1988) also carried out a second experiment using letters and digits as the categories. This experiment provided observers with an easier task and thus tried to facilitate the development of automatic performance, especially in the old adults. The results were similar to those obtained in the first experiment in that old adults were still unable to reach automatic performance.

119

Fisk et al. (1990) extended the above findings to the situation where the response requirements were minimized. This was done by having two response conditions, a yes-only condition where the observer responds only if a target probe was displayed, and a no-only condition where a response would be made only if no target items were displayed. As found in the earlier study (Fisk et al., 1988) older people, unlike the young, were not able to achieve automatic performance. Fisk et al. (1988, 1990) have used strength theory (Schneider, 1985; Schneider & Detweiler, 1987) to account for the above-mentioned age differences in search performance. Strength theory asserts that the efficiency of visual search depends on two learning mechanisms. The first is associative learning, which is dependent on the consistency of pairing of responses and stimuli. The second is priority learning, which refers to the relative ‘‘strength’’ or priority of the targets and distractors in attracting attention. In CM visual search conditions, for example, a target is always attended or responded to each time it appears. This results in increased importance of the target, giving it a ‘‘high-priority tag’’ (Fisk, Lee, & Rogers, 1991, p. 276) or high strength. On the other hand, distractors in the CM condition become less important and have a reduced strength over practice, because increasingly their presence signals to the observer that these stimuli should be disregarded in searching for the target (Schneider, 1985). Priority learning then is the process of increasing the strength of targets and decreasing the strength of distractors over practice (Schneider & Fisk, 1984; Shiffrin & Dumais, 1981). Priority learning only operates in CM search conditions (Rogers & Fisk, 1991; Schneider, 1985; Schneider & Detweiler, 1987; Shiffrin & Czerwinski, 1988). In VM conditions, targets and distractors remain at intermediate strength because their strength levels increase on some trials and decrease on others (Rogers & Fisk, 1991). There have been several efforts to determine whether the locus of the age deficit in visual search was associative or priority learning. Rogers and Fisk (1991) compared old and young adults’ performance in semantic-category

120


search. There were three conditions: a CM condition where both associative and priority learning were maximized, a VM condition where associative and priority learning (if any) were minimized, and an attenuated priority (AP) condition where associative learning (categorization) could occur but priority learning was minimized. The major findings were that in the initial phase, the performance of the young in the AP condition was similar to that of the old in the CM condition. Also, the performance of the old in both the AP and CM conditions was worse than that of the young. More importantly, age differences in the CM condition were greater than in the AP condition. These findings, therefore, suggest an age deficit in priority learning. These results were corroborated in the transfer phase where the young exhibited more disruption than the old in the CM condition. In a second and related study, Fisk and Rogers (1991, Experiment 1) investigated practice effects on age-related visual and memory search deficits. Associative learning, as opposed to priority learning, was expected to predominate in the memory search task because performance improvements would depend on the ability to unitize the memory set into a single response category (Fisk & Rogers, 1991). Moreover, since there was only one display item, there seemed to be no way in which the attention-attraction strength of the target could change. In the visual search task, however, priority learning would dominate because performance improvements depended primarily on the ability to increase the differentiation of the target-distractor strength (Fisk & Rogers, 1991; Shiffrin & Czerwinski, 1988). Fisk and Rogers (1991) used both CM letter and CM semantic category search in memory, visual, and hybrid memory/visual search (where both memory and display set size are greater than one). The type of search (memory, visual, or hybrid) was a withinsubjects manipulation. They found that the set size manipulation in memory search did not differentially affect the young and old age groups, implying the lack of an age deficit in associative learning. In the visual search condition, however, the old were more affected by increasing display size. In the hybrid memory/visual search

task, compared to the younger people, an increase in the visual display size affected older observers more than an increase in memory set size. These results also suggested to the authors that the locus of the age deficit was in priority learning, which dominates visual search. Before accepting the priority learning deficit hypothesis, it is necessary to examine more closely the methods employed by Fisk and Rogers (Fisk & Rogers, 1991; Rogers & Fisk, 1991). Clearly, there are large differences between semantic category search and visual search. In the prototypical visual search study, the number of targets (i.e., the memory set) is kept at one and the number of items in the display is varied across a range from one to many. It is assumed that in many search tasks, observers form a mental representation of the target and engage in a combination of serial and parallel operations to compare items in the display to the target’s representation. When multiple targets are used, they are few in number and easily remembered. In contrast, in the semantic category search task used commonly to assess age differences in automaticity, the targets are numerous exemplars of several superordinate categories such as animals, furniture, appliances, and the like. Although exemplars are often high associates of the category name, observers come to the experiment with an implicit memory set that is often quite large and must be edited by the experimental experience before an episodic memory set can replace the larger set from which it originated. Thus, the memory component of semantic category search is much larger than in visual search tasks. As a result, age differences in semantic category search cannot be interpreted unambiguously as evidence for a priority learning deficit. The major research objective of the present study was to examine age-related differences in priority learning in a visual search task. In addition to the CM search condition, the study included two other task conditions: VM search and a NS condition where the position of the target was known in advance. There were three display sizes (2, 4, and 8) in each of the task conditions. In CM search, the priority learning deficit hypothesis (cf. Rogers, 1992) suggested


that while younger adults would achieve automatic levels of performance, older adults would not. Furthermore, disruption following reversal was expected to be of greater magnitude in the young people. In VM search, where neither priority learning (Schneider, 1985; Shiffrin & Czerwinski, 1988) nor associative learning could occur, young and old adults were expected to have similar patterns of performance (Fisk & Rogers, 1991; Fisk et al., 1990; Rogers & Fisk, 1991). Inclusion of NS necessitates some elaboration. In this condition, priority learning was not expected to occur (Logan, 1992) because attention was always directed to the target by the cue. There was, therefore, no competition from other stimuli for attention. Consequently, only associative learning between the target and distractor features and their appropriate responses was expected to operate. Even at relatively low levels of practice no age differences have been observed to occur in NS conditions (Farkas & Hoyer, 1980; Plude & Hoyer, 1986; Wright & Elias, 1979). Hence the NS functions relating RT to display size for both young and older adults ought to remain parallel over practice. METHOD Participants Initially 18 younger and 18 older adults were recruited for the experiment. There were equal numbers of males and females in both age groups. However, the data from one older female in the NS condition were considered to be deviant, and so were excluded from the analyses that follow (the rationale is detailed below). The findings in this paper therefore represent the data collected from 18 young (M = 26.06 years, range = 11 years) and 17 older (M = 61.88 years, range = 13 years) adults. The younger adults were mostly students at the University of Calgary, whereas the independently living older adults were from the University and the community at large. All participants were compensated monetarily for their participation at the rate of $5.00 per session with a cash bonus of $15.00 for completion of all sessions. After initially obtaining the participant’s informed consent, a questionnaire was administered to assess health status, especially that pertaining to visual health. This self-report indicated that all participants were

121

in fair to excellent health and free from any visual pathology. The self-report also screened for any drug use that could interfere with the person’s performance on the search task. Spherical and cylindrical correction were provided for the 45.7 cm test distance. Not surprisingly, young adults were found to have better visual acuity (0.77°) than the older adults (0.92°), F(l, 33) = 5.46, p = .026. Still, all observers had acuities of 20/20 or better. Intraocular pressure was also measured for all participants. It was found to be 14.93 mm Hg for the young and 15.15 mm Hg for the old, a difference which was not statistically significant, F(1, 33) = 0.04, p = .836. Participants also underwent a test of static visual perimetry out to 8.76°, which was a little larger than the maximum eccentricity of objects in the search displays. The young adults were found to be significantly more sensitive than older adults, F(l, 33) = 52.37, p < .001. Younger adults had a significantly higher level of formal education (M = 16.83 years) than the old (M = 13.82 years), F(l, 33) = 11.12, p = .002. However, the young (M = 53.06) and older (M = 55.65) observers did not obtain discernibly different raw scores on the Wechsler Adult Intelligence Scale (WAIS) – Vocabulary subtest, F(l, 33) = 1.60, p = .215. Stimuli and Apparati Stimulus items for the experiment were similar to those used by Enns (1989) and consisted of black line drawings of the uppercase Y figure. One object type was oriented at 315° and the other object type was oriented at 135°. At the viewing distance both these stimuli subtended a visual angle of 2.51°. In all conditions, stimuli had a minimum center-tocenter separation of 3.26° with a minimum clearance of 0.75°. The maximum eccentricity of the outer edge of the display elements was 8.69 °. An example of a typical target present display is shown in Figure 1. Display size was fixed at 2, 4, or 8 stimulus items in all search conditions. Objects were presented within a display of 24 different cells formed from a matrix of 6 columns by 4 rows. For display size 2 (DS2) the matrix was a subarray of 3 columns by 2 rows, for display size 4 (DS4) the matrix was a subarray of 4 columns by 3 rows, and for display size 8 (DS8) the matrix was 6 columns by 4 rows (essentially over the entire display). On each trial one of these subarrays was presented randomly, placed at different locations within the display. The placement of target and distractor stimuli within these subarrays was randomly varied over trials. With such an arrangement, the den-

122

Fig. 1.


Example of a target present trial with display size of 8 elements.

sity of stimulus presentation was kept constant. Stimuli were presented randomly subject to the above density requirement, and the restriction that target eccentricity be equal on average over all the three display sizes and search conditions. At the beginning of each trial a dot (0.63°) was presented in the center of the display to serve as both a warning and fixation stimulus for the trial. In the NS condition a smaller dot of 0.26° was subsequently presented to indicate target location. This smaller dot was displayed adjacent to the center of the target location. Participants received feedback for response accuracy at the end of each trial by a plus (+) sign for correct responses, and a minus (–) sign for incorrect responses. Luminance of all stimulus elements was 1.98 cd/m2 while the luminance of the display screen was 84.63 cd/m2. This provided a Michelson contrast of .95.

Stimulus presentation and data collection were under the control of a Macintosh SE computer using the VScope program (Rensink & Ochs, 1992). Optical correction was provided by a trial lens set (R.H. Burton), and the Canon R-1 Autorefractor was also used to measure refractive error when it was necessary. Acuity was measured using PostScript Landolt Cs with eight targets for each level of minimum angle of resolution (MAR) which ranged from .38 to 2.5° in approximately .05 log MAR steps. The measurement of intraocular pressure was carried out with the Reichert NCT II Noncontact Tonometer. Static visual perimetry was assessed with the Friedmann Visual Field Analyser. Luminance measures of the display elements and computer screen were taken using the Minolta LS110 photometer. Viewing distance was maintained with a height-adjustable chin rest.


Procedure Participants were randomly assigned in equal numbers to either CM search, VM search, or NS conditions. Trials in the CM and VM search conditions began with a 950 ms fixation dot. The target and distractor stimuli were then displayed immediately after the offset of the fixation dot. The target was present on only one half of all trials. This display remained on until the observer made a response or 5 s had elapsed. For one half of the observers, the 315° object was the target and 135° objects were the distractors. For the remaining observers, the target-distractor assignment was reversed. Participants responded by pressing the c key when they thought the target was present and the m key when absent. This assignment of keys to responses was counterbalanced across participants. Placards placed on the response-appropriate side of the monitor continuously informed observers of the target and distractor objects at that point in the study. Following a response, participants were presented with accuracy feedback. A new trial began after an interval of 2 s. Trials in the NS condition were exactly the same as in the CM and VM search conditions except that the target location was cued. This was accomplished through the presentation of a 100% reliable dot cue, which was displayed for 200 ms after the offset of the fixation stimulus. This 200 ms presentation of the cue was based on the findings of Madden (1985), who showed that older adults were able to utilize a cue of such a duration to allocate attention to a particular location in a display. There were 36 trials in a block, and the three display sizes were randomly varied with equal numbers within each block of trials. A total of 12 such blocks (432 trials) comprised a session. Participants were provided with accuracy feedback at the end of each block. Overall response accuracy collapsed across the entire 12 blocks was provided at the end of the session, which took about 40 min to complete. All participants completed a total of eight sessions. The first seven sessions comprised the training phase. In this phase participants in the CM search and NS conditions were presented with one object as the target, and the other object as the distractor. The VM search condition in this phase was the same as the CM search condition except that the target and distractor were reversed over consecutive blocks of trials. On completion of the training phase all participants carried out the same final transfer session consisting of CM search that for the CM group involved a full reversal of target and distractors. For all sessions participants were instructed to respond as quickly as possible while

123

keeping their error rate to 5% or less. There were up to two participants tested at a time in two adjacent and similar rooms. Generally, participants in both age groups took about two days between each session, the young with an average intersession separation of 2.08 days and the old with an average of 2.18 days. There was no significant difference between young and old adults in the time interval between sessions, F(l, 33) = 0.17, p = .682.

RESULTS Reaction times were excluded if they were more than two standard deviations from the person’s mean for that day and condition. The analyses of RT data include only those trials on which a correct response was given. Initial analysis revealed that many of the data points for one older female participant were outliers. Moreover, this person’s scores for all 6 conditions of Session 8 were found to be outliers. Therefore, as noted above, these data were excluded from all further analyses. The alpha level for all reported results was set at .05 and Bonferroni adjustments were made to maintain the family-wise Type I error rate at this level. The Geisser-Greenhouse adjustment was applied to the within-participants results of all univariate mixed-model ANOVAs to compensate for violations in sphericity (Maxwell & Delaney, 1990). It should be noted that the degrees of freedom of the Geisser-Greenhouse adjustment are fractional. To provide clarity and to assist the reader in the interpretation of the results, the unprotected degrees of freedom are reported, but with the p values associated with the Geisser-Greenhouse adjustment. Errors Error data are presented in Figure 2. A decline in error rates over the seven training sessions is clearly seen, with the older adults exhibiting the greater decline. For both age groups, the error rate across the seven training sessions was less for target absent trials than for target present trials. The above observations were supported by an Age (2) x Condition (3) x Presence (2) x Display Size (3) x Sessions (7) ANOVA. Main effects of

124

Fig. 2.


Error rates as a function of display size (DS). Left column, consistent mapping (CM); center column, varied mapping (VM); right column, nonsearch (NS). The upper two rows of figures depict young (Y) adult data and the lower two rows of figures depict old (O) adult data. TP = target present; TA = target absent.


Presence, F(l, 29) = 26.82, p < .001, Sessions, F(6, 174) = 14.09, p < .001, and an Age x Sessions interaction, F(6, 174) = 4.83, p = .002, were all significant. Additionally, the Display Size main effect, F(2, 58) = 16.51, p < .001, and Age x Display Size interaction, F(2, 58) = 11.55, p < .001, were significant. The significant Display Size, and Age x Display Size effects indicate that larger display sizes resulted in higher error rates, and that the older adults were affected to a greater degree. However, the interpretation of these findings is qualified by a fiveway interaction amongst all the factors in the analysis, F(24, 348) = 2.07, p = .014. To explore this five-way interaction, an Age (2) x Presence (2) x Display Size (3) x Sessions (7) ANOVA was conducted within each training condition. In the CM condition, there was a significant main effect of Sessions, F(6, 60) = 5.01, p = .006, and a marginally significant Age x Sessions interaction, F(6, 60) = 3.77, p = .020, with older adults showing greater improvement with practice. The main effects of Age, F(l, 10) = 1.79, p = .211, Display Size, F(2, 20) = 5.84, p = .033, and the Age x Display Size interaction, F(2, 20) = 2.13, p =.173, failed to reach significance. The VM error data showed that participants committed fewer errors with practice, F(6, 60) = 7.88, p = .003. The main effect of Presence was also significant, F(l, 10) = 16.80, p = .002, which indicated that participants were making more errors on target present trials. There was no age difference in the Presence effect, F(1, 10) = 7.56, p =.020. Error rate increased with larger display sizes, F(2, 20) = 17.55, p < .001, and the older adults were affected more than the young, Age x Display Size, F(2, 20) = 11.46, p = .003. In NS, the only significant effect was that of Presence, F(l, 9) = 14.51, p = .004, suggesting that participants were making more errors on target present trials. The only cause for concern is the significant Age x Display Size interaction in VM. However, given the generally small percentage of errors committed by both age groups (5% or less), as well as the presence of larger display size effects in RT for the old, there appears to be no speed-accuracy trade-off.

125

The error data for Session 8 (the transfer session) were analyzed by an Age (2) x Condition (3) x Presence (2) x Display Size (3) ANOVA. No age effects were found to be significant, which indicated that the number and pattern of errors were similar for both age groups. Reaction Times The mean RTs for correct CM trials as a function of sessions and display size are shown in the top row of Figure 3. Corresponding data for VM and NS conditions are shown in the central and bottom rows of that same figure. Training Sessions Not surprisingly, in all conditions younger participants were faster overall, and performance improved across training sessions for both age groups. Moreover, responses in both age groups were faster for target present trials. In both the CM and VM conditions large display size effects were obtained, with participants taking longer to respond when more distractors were present. The display size effects in both CM and VM were more pronounced on target absent trials and in the earlier sessions. Additionally, in these same conditions, the older adults appeared to be more affected by larger display sizes than did the young. In contrast, there were no such display size effects in the NS condition. Importantly, even with the differential effects of display size based on age in CM and VM, there does not appear to be any age difference in the rate of improvement with training within any of the three search conditions. A summary of the change in mean RTs from Session 1 to Session 7 is presented in Table 1. In the CM and VM conditions, larger display sizes correspond to larger relative improvements for both age groups. However, this relationship was not seen in the NS condition. The above observations were essentially supported when the RT data for the training sessions were analyzed employing an Age (2) x Condition (3) x Presence (2) x Display Size (3) x Sessions (7) ANOVA. All main effects were significant: Age, F(l, 29) = 20.83, p < .001, Condition, F(2, 29) = 25.33, p < .001, Presence, F(l, 29) = 48.92, p < .001, Display Size, F(2, 58)

126

Fig. 3.


Reaction time (RT) as a function of display size (DS) and age. Left column, target present data; right column, target absent data; top row, consistent mapping (CM); center row, varied mapping (VM); bottom row, nonsearch (NS). The ‘‘left column’’ of figures depict target present data and the ‘‘right column’’ of figures depict target absent data. Y = young; O = old.

127


Table 1. Mean Reaction Time (RT) Change between Session 1 and Session 7. Session 1

Session 7

RT change

% Change

559.92 628.84 719.68 610.08 602.83 618.63 704.05 749.55 906.81 711.93 762.41 801.02 406.05 414.07 415.08 383.41 376.90 379.55

187.43 277.44 512.05 46.07 145.77 265.64 178.34 313.95 534.02 102.80 193.63 239.40 228.14 208.67 231.44 177.79 185.54 200.70

25.1 30.6 41.6 7.0 19.5 30.0 20.2 29.5 37.1 12.6 20.3 23.0 36.0 33.5 35.8 31.7 33.0 34.6

639.81 747.72 1022.56 614.75 712.07 848.43 766.17 1018.59 1482.27 732.11 926.92 1084.95 602.99 641.09 626.90 552.14 552.24 561.59

229.40 377.29 595.54 154.51 252.79 401.12 271.58 317.86 360.16 200.71 273.27 395.37 171.80 194.97 223.12 136.91 166.02 160.22

26.4 33.5 36.8 20.1 26.2 32.1 26.2 23.8 19.5 21.5 22.8 26.7 22.2 23.3 26.2 19.9 23.1 22.2

Young adults CM

TA TP

VM

TA TP

NS

TA TP

DS2 DS4 DS8 DS2 DS4 DS8 DS2 DS4 DS8 DS2 DS4 DS8 DS2 DS4 DS8 DS2 DS4 DS8

747.35 906.28 1231.73 656.15 748.60 884.27 882.39 1063.54 1440.83 814.73 956.04 1040.42 634.19 622.74 646.52 561.20 562.44 580.25 Old adults

CM

TA TP

VM

TA TP

NS

TA TP


869.21 1125.01 1618.10 769.26 964.86 1249.55 1037.75 1336.45 1842.43 932.82 1200.19 1480.32 774.79 836.06 850.02 689.05 718.26 721.81

Note. RT change is calculated as: Session 1 RT – Session 7 RT. CM = consistent mapping; VM = varied mapping; NS = nonsearch; DS = display size; TA = target absent; TP = target present.

= 203.57, p < .001, and Sessions, F(6, 174) = 72.75, p < .001. In addition, significant interactions for Condition x Display Size, F(4, 58) = 47.08, p < .001, Condition x Display Size x Presence, F(4, 58) = 14.34, p < .001, and Age x Condition x Display Size, F(4, 58) = 9.70, p < .001, were found. There were no significant Age

x Sessions, F(6, 174) = 2.17, p = .119, Age x Condition, F(2, 29) = 25.33, p = .456, or Age x Condition x Sessions, F(12, 174) = .97, p = .433, interactions. However, there was a significant five-way interaction among all the factors analyzed, F(24, 348) = 2.93, p = .004.

128


A follow-up of the five-way interaction was carried out within each search condition via an Age (2) x Presence (2) x Display Size (3) x Sessions (7) ANOVA. In the CM condition the results suggested that participants responded faster on target present trials, F(l, 10) = 10.35, p = .009, and that RT improved with practice for both age groups, F(6, 60) = 24.54, p < .001. However, practice was found to benefit performance in both age groups similarly, F(6, 60) = 1.72, p = .213. Additionally, although both age groups were affected by larger display sizes, F(2, 20) = 62.15, p < .001, the effect was more pronounced for older adults, F(2, 20) = 9.66, p = .011. In the VM condition, young adults were found to be faster than older adults, F(l, 10) = 10.49, p = .009, and both groups improved over practice, F(6, 60) = 33.05, p < .001. Participants were also found to respond faster on target present trials, F(l, 10) = 30.66, p < .001. However, as in the case of CM, there was no Age x Sessions interaction, F(6, 60) = 1.16, p < .001, suggesting that with practice both age groups reduced their RTs at similar rates. Participants were also slowed when larger display sizes were presented, F(2, 20) = 199.29, p < .001, but older adults were found to be affected to a greater extent, F(2, 20) = 45.72, p < .001, especially on target absent trials, F(2, 20) = 15.40, p < .001. In the NS condition, young participants responded faster than older adults, F(l, 9) = 9.65, p = .013, and RTs for all participants were generally shorter on target present trials, F(l, 9) = 40.83, p < .001. Additionally, both age groups improved their performance with practice, F(6, 54) = 18.56, p < .001, and at approximately the same rates, as the Age x Sessions effect was found to be nonsignificant, F(6, 54) = 1.18, p = .330. The only other significant effect was an Age x Presence interaction, F(l, 9) = 9.11, p = .015, indicating that, compared to young adults, older adults were slowed on target absent trials. Importantly, in contrast to the CM and VM conditions, there were no significant Display Size, F(2, 18) = 1.53, p = .248, or Age x Display Size, F(2, 18) = 1.10, p = .324, effects, that is, performance in neither age group was affected by increasing the number of distractors in the display.

Last Training Session The data in Figure 3 and Table 1 provide helpful information about the performance of young and old adults at the completion of the training phase. Overall, the older adults were slower than young adults in all conditions, although they approached the RT of the younger group in the smallest display size in CM. The responses of both age groups were generally faster on target present versus target absent trials. It also appears that although there is a negative effect of increasing display sizes upon performance, this effect differs between the young and old, and among the three training conditions. In CM, older adults continued to be more affected by larger display sizes compared to the young. In VM, both young and old adults were slowed by larger display sizes for both target absent and target present trials, with older adults being more affected by larger display sizes on target absent trials. The high level of performance of young adults on target present trials was unexpected and should be noted. In particular, young adults scanned display items at rates approaching the level one would expect in CM performance. For instance, there was a mean RT difference of only 89 ms between responses in the smallest display size (2) and largest (8), corresponding to a search rate of about 15 ms/item. In the NS condition, there was little effect of display size upon RT in either target absent or target present trials for either age group. All these observations were essentially supported by the following analyses. An Age (2) x Condition (3) x Presence (2) x Display Size (3) ANOVA was conducted. The main effects of Age, F(l, 29) = 18.28, p < .001, Condition, F(2, 29) = 26.45, p < .001, Presence, F(l, 29) = 29.72, p < .001, and Display Size, F(2, 58) = 115.30, p < .001, were all significant. There were a number of significant interactions including the Age x Display Size, F(2, 58) = 37.65, p < .001, Age x Condition x Display Size, F(4, 58) = 10.11, p < .001, and Age x Presence x Display Size, F(2, 58) = 4.17, p = .030, effects. However, these interactions were influenced by the significant four-way interaction between all the factors analyzed, F(4, 58) = 3.99, p = .012.


Follow-up analysis entailed an Age (2) x Presence (2) x Display Size (3) ANOVA within each condition. The analyses revealed a significant Age x Display Size interaction in the CM condition, F(2, 20) = 13.61, p = .003, indicating that older adults continued to be affected by larger display sizes. In VM, older adults’ performance was found to be more affected than that of the young by larger display sizes, F(2, 20) = 26.14, p < .001, and especially on target absent trials, F(2, 20) = 8.96, p = .003. In the NS condition, although there was a significant main effect of Age, F(l, 9) = 20.46, p < .001, implying that the young adults were responding faster, no interactions involving Age were found. Transfer Session Table 2 presents the mean RT data for Session 7 and Session 8 over each of the three display sizes. There is also a computation of the magnitude of disruption based upon the following formula: RT Session 8 – RT Session 7 RT Session 7 The calculation of disruption scores in this manner results in an index of the relative changes occurring in Session 8. Consequently, the differences in RT among participants is equated relative to a baseline, RT for Session 7 (Rogers, 1992). These disruption scores plotted by display size are presented in Figure 4. Referring to this figure, and Table 2, the following observations can be made. Disruption in performance occurred for both age groups in both the CM and NS conditions, with the magnitude of disruption increasing with display size. On the other hand, performance improved at transfer in the VM condition. For both age groups there was also greater disruption in NS than in CM. Furthermore, the young adults in NS showed a larger disruption effect than old adults for DS2 and DS4. The magnitude of age differences in disruption is, however, unclear for DS8. The above observations were confirmed by an analysis of the disruption data. Initially, an Age

129

(2) x Condition (3) x Presence (2) x Display Size (3) ANOVA was conducted. This omnibus analysis revealed that the magnitude of disruption differed among training conditions, F(2, 29) = 49.75, p < .001, and increased with display size, F(2, 58) = 125.49, p < .001. Importantly, the results showed that overall, the magnitude of disruption in both age groups was similar: Age, F(l, 29) = 1.75, p = .196. Moreover, the Age x Condition interaction, F(2, 58) = 2.96, p = .068, was also nonsignificant, which indicated that no age differences in disruption were present in any of the training conditions. The greater-than-expected disruption in the NS condition relative to CM prompted an examination of the pattern of disruption data between the CM and NS conditions. This was conducted with an Age (2) x Condition (2) x Presence (2) x Display Size (3) ANOVA. Compared to CM greater disruption was found in NS, F(l, 19) = 23.12, p < .001. Moreover, the magnitude of disruption with larger display sizes was greater in NS than in CM, F(2, 38) = 27.48, p < .001. There were no significant differences in disruption between the age groups, F(l, 19) = 2.20, p = .152, and this result was consistent in both conditions, F(l, 19) = 2.81, p = .110. Moreover, although participants had more difficulty with the larger display sizes, F(2, 38) = 121.77, p < .001, both age groups were affected equally, F(2, 38) = .20, p = .694. Additionally, the Age x Condition x Display Size interaction, F(2, 38) = .39, p = .570, was found to be nonsignificant, suggesting that although both age groups experienced greater disruption with larger display sizes, there was equivalent disruption in both age groups. The omnibus analysis does not indicate that any age differences exist in the patterns of disruption within any of the three training conditions. The results of the analysis that included the disruption data from only the CM and NS conditions confirmed the results of the omnibus analysis in that no significant differences involving the age factor were revealed. The results also suggest that the magnitude of disruption in NS was significantly greater than that in the CM condition.

130


Table 2. Disruption Scores. Session 7

Session 8

RT change

Disruption

702.25 817.60 1114.06 735.82 838.36 999.85 642.77 695.08 801.12 608.49 673.42 725.30 748.76 925.87 1278.74 725.17 831.01 1018.98

142.33 188.76 394.38 125.74 235.53 381.22 –61.28 –54.47 –105.69 –103.44 –88.99 –75.72 342.71 511.80 863.66 341.76 454.11 639.43

0.25 0.30 0.55 0.21 0.39 0.62 –0.09 –0.07 –0.12 –0.15 –0.12 –0.09 0.84 1.24 2.08 0.89 1.20 1.68

754.51 1018.47 1654.23 760.54 1027.57 1367.22 733.53 957.30 1422.71 697.50 850.83 1051.46 849.70 1075.76 1640.42 771.61 929.50 1221.18

114.70 270.75 631.67 145.79 315.50 518.79 –32.64 –61.29 –59.56 –34.61 –76.09 –33.49 239.71 434.67 1013.52 219.47 377.26 659.59

0.18 0.36 0.62 0.24 0.44 0.61 –0.04 –0.06 –0.04 –0.05 –0.08 –0.03 0.40 0.68 1.62 0.40 0.68 1.17

Young adults CM

TA TP

VM

TA TP

NS

TA TP


559.92 628.84 719.68 610.08 602.83 618.63 704.05 749.55 906.81 711.93 762.41 801.02 406.05 414.07 415.08 383.41 376.90 379.55 Old adults

CM

TA TP

VM

TA TP

NS

TA TP


639.81 747.72 1022.56 614.75 712.07 848.43 766.17 1018.59 1482.27 732.11 926.92 1084.95 602.99 641.09 626.90 552.14 552.24 561.59

Note. Reaction time (RT) change is calculated as: Session 8 RT – Session 7 RT. Disruption: (Session 8 RT – Session 7 RT) / Session 7 RT. CM = consistent mapping; VM = varied mapping; NS = nonsearch; DS = display size; TA = target absent; TP = target present.

Useful Field of View Support for the finding of no age deficit in priority learning has thus far been provided by the disruption data. However, it is also important to consider the display size effect for the CM condition in Session 7. Because priority learning is associated with the development of automatic

performance, it can also be indexed by the display size effect. That is, automatic performance will have been attained, and priority learning substantially developed, if display size effects are found to be nonsignificant (Schneider, 1985; Schneider & Shiffrin, 1977).


Fig. 4.

131

Disruption as a function of display size and age. Top row, consistent mapping (CM); center row, varied mapping (VM); bottom row, nonsearch (NS). Y = young; O = old; TA = target absent; TP = target present.

132


By Session 7 only the young adults had achieved automatic performance on target present trials in the CM condition. Older adults continued to be adversely affected by larger display sizes, even on target present trials, suggesting an age deficit in priority learning not revealed in the disruption data. This was investigated further within the context of the useful field of view (UFOV) hypothesis (Ball, Roenker, & Bruni, 1990). The UFOV is ‘‘...the total visual field area in which useful information can be acquired without eye and head movements ...’’ (Ball, Beard, Roenker, Miller, & Griggs, 1988, p. 2210). The size of the UFOV is not constant but fluctuates depending upon task demands. Especially relevant in the present context is that the UFOV has been found to become constricted in visual search (Ball et al., 1988; Scialfa, Kline, & Lyman, 1987). Importantly, older adults have a restricted UFOV compared to the young (Ball et al., 1988; Scialfa et al., 1987). Moreover, the UFOV has been shown to increase with practice in visual search, with both young and old adults benefiting equivalently (Ball et al., 1988). These findings suggest that perhaps the older adults in the current experiment may have achieved automatic levels of performance in CM, but within a restricted portion of the display. To investigate this possibility in the CM condition, only the four central cells around fixation were considered. The eccentricity of the midpoint of stimulus items appearing in these four cells was 1.88°. Only the RTs corresponding to targets in one of these four cells were extracted for analyses on target present trials. Target absent trials extracted for the analyses had exactly the same distribution of stimulus elements over the display as in the selected target present trials. In this case, however, the target was replaced with a distractor. It should be pointed out at the outset that the inclusion of target absent trials revealed no interesting effects in this analysis. Training Sessions The RT data in the CM condition from the four central cells were subjected to an Age (2) x Presence (2) x Display Size (3) x Sessions (7)

ANOVA. As expected, the main effects of Presence, F(l, 10) = 24.60, p < .001, Display Size, F(2, 20) = 50.88, p < .001, and Sessions, F(6, 60) = 18.04, p < .001, were found to be significant. The only effect involving Age was the Age x Display Size interaction, F(2, 20) = 8.99, p = .011. The results, therefore, suggest that even when considering only the four central cells, the older adults in CM continue to be more adversely affected than the young with increasing display sizes over the training sessions. Last Training Session The RT data in CM from Session 7 were then analyzed within an Age (2) x Presence (2) x Display Size (3) ANOVA. The main effects of Presence, F(l, 10) = 18.61, p = .002, and Display Size, F(2, 20) = 30.18, p < .001, were significant. In addition, older adults were found to be more affected by display set size, F(2, 20) = 6.27, p = .014. Given the significant Age x Display size effect, follow-up analyses for Session 7 RTs were conducted. These were essentially one-way ANOVAs involving the Display Size factor. For the young adults, there was a significant Display Size effect on target absent trials, F(2, 10) = 10.47, p = .012, but not for target present trials, F(2, 10) = 0.52, p = .548, suggesting that automatic performance was achieved on target present (Mslope = –3.3 ms) but not target absent trials (Mslope = 28.5 ms). Older adults had a similar pattern of results. There was a significant Display Size effect for target absent trials, F(2, 10) = 28.75, p = .002, Mslope = 63.5 ms, but not for target present trials, F(2, 10) = 3.20, p = .096, Mslope = 6.0 ms. Thus, the results of the analyses indicate that the older adults in CM could identify the target automatically when it was present, as measured by the lack of a display size effect, by Session 7. These results are reflected in Figure 5, which shows the relation between RT and display size for target present trials in Session 7 for the CM condition. In particular, the RT changes for both young and old adults over display size for the peripheral 20 cells in the display and when only the four central cells are extracted, are plotted in the figure.


Fig. 5.

133

Reaction time (RT) as a function of display size, age, and eccentricity of the target. Y = young; O = old.

To investigate the effects of eccentricity upon the target present RT data in CM, further analysis utilizing an Age (2) x Eccentricity (2) x Display Size (3) ANOVA was conducted. Eccentricity in this analysis reflected the data from the four central cells (central) and those from the remaining 20 more peripheral cells. The analysis showed a significant interaction among all the three factors analyzed, F(2, 20) = 21.36, p < .001. The significant three-way interaction prompted follow-up tests within each age group. For the young adults there were no significant effects of Eccentricity, F(l, 5) = 5.57, p = .065, Display Size, F(2, 10) = 0.18, p = .714, or Eccentricity x Display Size, F(2, 10) = 1.96, p = .207. These results suggest that the performance of the young was not differentially affected by the target eccentricity. Older adults, on the other hand, were significantly slower in the periphery, F(l, 5) = 34.43, p = .002. Moreover, not only were they adversely affected by larger display set sizes, F(2, 10) = 27.64, p < .001, but this effect was dependent upon target eccentricity, as

evidenced by an Eccentricity x Display Size interaction, F(2, 10) = 39.96, p = .001. Transfer Session The next issue that had to be addressed was whether there was an age deficit in priority learning in the CM condition when only the four central cells were examined. To this end, the disruption data for the four central cells were calculated. These data are presented in Table 3 for both young and old adults. As Table 3 shows, the magnitude of disruption among the older adults is marginally larger than that for young adults, suggesting no age deficit in priority learning. An Age (2) x Presence (2) x Display Size (3) ANOVA was carried out on the disruption data. The only significant main effect was for Display Size, F(2, 20) = 32.49, p < .001, suggesting that both age groups experienced greater disruption with larger display sizes. There was also a marginally nonsignificant Age x Display Size interaction, F(2, 10) = 3.56, p = .066. However, not much weight can

134


Table 3. Disruption Scores for Central Cells. Session 7

Session 8

RT change

Disruption

716.59 757.92 1083.51 677.10 786.57 863.41

158.99 163.04 354.84 74.28 206.51 280.81

0.29 0.27 0.49 0.12 0.36 0.48

752.70 958.61 1614.41 759.08 928.87 1192.91

124.48 228.14 604.97 156.47 279.03 554.02

0.20 0.31 0.60 0.26 0.43 0.87

Young adults CM

TA TP

DS2 DS4 DS8 DS2 DS4 DS8

557.60 594.88 728.67 602.82 580.06 582.60 Old adults

CM

TA TP

DS2 DS4 DS8 DS2 DS4 DS8

628.22 730.47 1009.44 602.61 649.84 638.89

Note. Reaction Time (RT) change is calculated as: Session 8 RT – Session 7 RT. Disruption: (Session 8 RT – Session 7 RT) / Session 7 RT. CM = consistent mapping; DS = display size; TA = target absent; TP = target present.

be attached to this Age x Display because, as Table 3 indicates, any ences would favor the older adults! previous analyses, no age deficit learning was observed.

Size effect age differThus, as in in priority

DISCUSSION The most striking result of the present study is that there is little evidence of an age-related deficiency in learning a feature search task. The perceptual learning that occurred, especially in the CM condition, took the form of improvements over practice in the ability to attend to certain stimulus features in the display. There was, in fact, the development of an automatic response to target features. Disruption measures taken at transfer suggest that older and younger observers had automatized the allocation of attention to target features. The implication of these results in CM search is that the magnitude of priority learning that developed over the training sessions was similar for both age groups. The results from the NS condition point to the conclusion that older adults can match the level of associative learning found in younger adults.

The Consistent Mapping Search Condition If there was an age deficit in priority learning, then an age by display size interaction would be observed by the end of training in the CM condition. As well, younger adults would show more disruption at reversal. Although the transfer results in CM offer no support for the second implication, an assessment of the first requires more thought. The young adults on target present trials had by Session 7 achieved automatic performance in the CM search task. That is, they exhibited no display size effect. Older adults, however, continued to be affected by larger display sizes when making responses on these trials. To further investigate this finding, the possibility that older adults had a significantly smaller UFOV was entertained. A smaller UFOV in older adults would place them at a disadvantage relative to young adults when targets appeared at more peripheral locations in the display. Consequently, only the CM RT data for the four central cells were extracted for analysis. The analysis showed that, akin to the younger adults, older adults on target present trials showed no display size effect. Thus older adults were exhibiting automatic search perfor-


mance, but only within the central area of the display. This result, together with the eccentricity by display size effect in Session 7, demonstrates that the age by display size effect obtained earlier, when the entire display was included in the analysis, was due to the occasions when the target appeared in the more peripheral locations. The analysis of the disruption data associated with the four central cells affirmed the earlier finding that there was no age deficit in priority learning. Taken together, the results in the CM condition are inconsistent with the assertion of Fisk and Rogers (1991), Rogers (1992), and Rogers and Fisk (1991), of an age deficit in priority learning in visual search. The present study has also demonstrated that older adults can achieve automatic performance in a visual search task. The possible reasons for across-study differences in results are manifold but perhaps the most important concerns the definition of visual search. We asked observers to complete a visual search task that relied heavily on the ability to attend selectively to a single object feature (i.e., orientation). This is quite different than the semantic category search task typically used by Rogers and her colleagues, which relies to a much greater degree on unitization of a regularly changing memory set. The Varied Mapping Search Condition The results in VM for both age groups support the hypothesis that there would be no age differences in performance during training and little or no disruption at transfer. Training results showed that performance was similar in both age groups, despite the fact that the young adults were less affected by display size than expected. Moreover, little or no disruption at transfer was observed in either age group. Performance during training in VM was as hypothesized: slower RTs than in CM for all three display sizes, significant display size and age by display size effects throughout training, comparable rates of RT improvement for young and old, and lack of disruption at transfer. The training results on target present trials of young adults demand additional comment. Despite a statistically significant display size ef-

135

fect, the estimated time taken to search one item (15 ms) was unusually fast. One explanation for the faster search rates by young adults involves the development of a fundamentally different approach to the task. Czerwinski, Lightfoot, and Shiffrin (1992) proposed that over practice, people in VM may shift their search strategy from one based upon detection of a stimulus feature (such as orientation) to a higher one of featural overlap based upon stimulus homogeneity (Duncan & Humphreys, 1989; Fisher & Tanner, 1992). In the present case, distractor stimuli presented on a trial were all identical so that people could have grouped the distractors according to a common feature, such as orientation. Such shifts in strategies will of course increase search efficiency, because fewer comparisons will then need to be made. Interviews conducted at the conclusion of the present study are consistent with this view. The majority of young observers in the VM condition replied that by Session 7 they would respond on ‘‘seeing something different’’ in the display. The older people in the same condition reported otherwise; they would only respond upon ‘‘seeing the target’’. The older adults were presumably searching for a stimulus feature in the display that would suggest a target. Young adults, on the other hand, virtually discarded this method of search and instead capitalized on featural similarities and differences in the display. An important ramification of young adults carrying out such a search strategy in VM is that the stimulus-to-response consistency was modified (Duncan, 1986). The VM search task was plainly defined experimentally in terms of the level of the stimulus in that there was frequent switching between the two stimulus items as targets and distractors. By restructuring the task demands so as to emphasize the detection of local differences, young people would essentially have introduced a measure of CM, but at a more general level than the local stimulus level manipulated by the experimenter. Transfer could also have been facilitated for the young adults if they adopted the above rule-level strategy (Kramer, Strayer, & Buckley, 1991) during training. This is because they would have carried over the search strategy learned during training to the

136


transfer task. However, this suggestion must be tempered by the fact that no age effects were observed at transfer. The Nonsearch Condition Errors at NS transfer were greater than in the VM condition (see Figure 2) and RT disruption in the NS condition was significantly greater than in CM search. This was an unexpected outcome in light of the fact that only the associative learning mechanism was thought to be operating in the NS condition (Logan, 1992). As such, one would expect errors at transfer to be no greater than in VM search, which also involves associative learning, and less than in CM search because in the latter, both associative and priority learning were assumed to occur. A possible explanation for these results relies on the concept of interference, in particular contextual interference (Battig, 1979). A prevalent view about interference, as the name itself suggests, is that it applies to situations where learning is hindered in some way. Thus proactive interference is said to occur when previous learning hinders new learning. In contrast, retroactive interference is assumed to occur if new learning adversely affects previous learning. A major effect of these two forms of interference is that of negative transfer (Battig, 1972). An important point is that both these forms of interference utilize an interpolated task as the source of interference. Contextual interference, on the other hand, can occur in the absence of an interpolated task. It includes all sources of interference both extrinsic and intrinsic to the conduct of the principle task. These include the sequence of stimulus presentation, varying task demands, and practice schedules. Contextual interference, unlike the above two forms, is beneficial to skill learning (Magill & Hall, 1990) and effectively induces positive transfer (Battig, 1972). This is because, although the greater processing demanded by contextual interference slows performance during practice, this greater processing concurrently introduces more elaborate and distinctive processing of the task (Battig, 1979). Such processing would be expected to facilitate learning (Craik & Lockhart, 1972; Jacoby & Craik, 1979) com-

pared to learning under conditions of little or no contextual interference. This additional learning will be manifested as larger positive transfer effects. In the case of visual search, task demands (and contextual interference) can be manipulated by varying target-distractor similarity (Duncan & Humphreys, 1989), or simply by the inclusion of different display set sizes. In the present study, contextual interference was varied by the inclusion of different display set sizes in CM search. In contrast, in the NS condition, because the precue localizes the stimulus perfectly, observers effectively are presented with a display size of one. People in NS would face an easier and less variable task than those in the CM condition, leading to significantly less contextual interference during practice. Thus despite the poorer performance during practice in CM compared to NS, the learning which took place in the CM condition would, in fact, have been greater. This additional learning by CM observers would have been reflected in their significantly faster performance at transfer. Another possibility for the greater-than-expected disruption in the NS compared to the CM condition could have been that the task demands were changed at transfer, especially in NS. Not only were the target and distractor stimuli switched at transfer, but participants also had to carry out a scan of the display for the target, something that they appear not to have done over the practice sessions. Observers in the CM condition, on the other hand, continued to carry out a search at transfer despite the fact that the target and distractor stimuli were also switched. Consequently, performance would have been faster in the CM than the NS condition. Finally, our expectations regarding the NS condition were predicated on the assumption (Logan, 1992) that priority learning would be minimal in this condition because distractors do not compete with targets for attention. Perhaps it is the case that attention-attraction strength is modulated even in the absence of such featurebased competition. This might occur because the features of the target demand attention to the extent that they are associated with a particular motor response (Eriksen & Schultz, 1979).


Theoretical Perspectives The pattern of results obtained in this study supports a strength-based theory of visual search (Schneider, 1985). The fact that there was a significant amount of disruption at transfer for both age groups in the CM condition but none in VM search is a case in point. The disruption at transfer in CM occurred because on completion of training (Session 7) the accumulated attentionattraction strength of the target stimulus was very high relative to that of the distractors. Performance improvements occurred because of this increasing differentiation in target-distractor strength over training (Shiffrin & Czerwinski, 1988; Shiffrin & Dumais, 1981). Upon transfer, therefore, the new distractors (formerly targets) would now draw attention away from the new target because of their relatively greater attention-attraction strength. In the VM condition, however, the frequent switching between target and distractor stimuli results in these display elements maintaining an equivalent strength (Shiffrin & Dumais, 1981). Consequently, VM training produced little disruption at transfer. As mentioned above, disruption in NS could be explained under the assumption that even while distractors did not compete with the target for attentional resources, processing of target features that occurred prior to response selection was sufficient to induce a differential in attention-attraction strength. That is, even though the cue obviated the need for search in the NS condition, observers still needed to determine whether the target was present or absent in order to make a correct response. This decision required that attention be allocated selectively to stimulus features, with a consequent disruption at reversal. While strength theory was central to the design of this experiment, other theoretical perspectives can be extended to account for these results. In feature integration theory (FIT), when extended through the group scanning hypothesis (Treisman & Gormican, 1988), the ease with which feature search is conducted depends on the relative differences in activation distributions corresponding to target present and target absent trials. When the activation distributions are widely separated, displays can be searched

137

in groups of items. When activation distributions exhibit greater overlap, displays must be searched in smaller groups. With only limited exposure to a difficult feature search task, the distributions will have considerable overlap, effectively reducing signal strength and group size. Additional trials will serve to separate the distributions, allowing for larger groups to be searched in parallel. Thus, FIT can account for the changes in performance with CM training. The NS data do not pose a problem. Even if displays are searched in small groups in difficult feature search, the precue guarantees that selective attention is allocated to the target group on every target present trial and insulates observers from the display size effect. Explaining the VM results would require the invocation of a flexible inhibitory mechanism (Treisman & Sato, 1990) whereby ‘‘spatial selection can be achieved by changing the relative activation produced in the master map by one or the other of the distractor feature maps’’ (p. 461). Presumably this inhibitory plasticity would come about through experience with the task. FIT could also explain the disruption produced at reversal because any experienced-based activation or inhibition that allows for efficient selection will necessarily produce costs when applied inappropriately. According to the guided search model (Wolfe, 1994), bottom-up activation is unlikely to change with practice, but top-down activation might well be modulated by experience, leading to a general reduction in RTs and the display size effect in CM and VM performance. Topdown factors would also explain the NS data, because the serial component of search would not be required when a precue allows attention to be allocated to the target with perfect validity. The guided search model could also explain the disruption occurring at reversal because topdown activation that guided the serial process in CM search would, on transfer, guide observers to the wrong objects. Disruption in the NS condition might be explained by asserting that the need for an overt response compelled observers to match the object at the cued location to an internal representation of the target and this requirement induced a change in top-down activation that interfered with performance at reversal.

138


CONCLUSIONS Why have we found so little evidence of age deficits in CM feature search when Fisk, Rogers, and their colleagues have consistently found age deficits in semantic category search? First, consider differences in the stimuli used. In our study, targets and distractors were defined in such a way that feature-based selection could reduce the serial search component and the attentional involvement on which such a component relies. It may be that age differences in both initial and asymptotic performance are minimal under less attentionally demanding visual search conditions. Consistent with this view, Humphrey and Kramer (1997) found that age differences in conjunction search slopes vary considerably, and in easy conjunction search both age groups performed in a range generally considered to be indicative of parallel search. Rogers and her colleagues often use the semantic category search task in which observers must indicate which of several visually presented words is an exemplar of a target category (e.g., furniture). In such a task, it is unlikely that simple features such as orientation and contrast polarity can be used to select items, and even if it was possible the relatively small display sizes (typically 3 or less) and the relatively large memory component would make it difficult to generalize to other visual search tasks or to explain with contemporary models of visual attention and search (e.g., FIT, the guided search model). Second, consider the procedural differences across these studies. In Fisk and Rogers’ studies observers are often completing several cognitively heterogeneous search tasks in interlaced fashion. Under these circumstances, it may be difficult for older observers to rapidly adopt and consistently apply the most effective search algorithms. In contrast, we segregated the various search tasks so that cognitive procedures could be deployed continuously, and allowed older observers to avoid cross-task interference (Bailey & Lauber, 1998; Kray & Lindenberger, 1998) to which they may be more susceptible. Both stimulus- and procedure-based accounts require further study. If across-study differences

are due to variations in stimuli, then it becomes necessary to determine the stimulus features that are age-sensitive and those that are not, as exemplified in Humphrey and Kramer’s (1997) study of age differences in conjunction search. If procedural differences are the cause of these inconsistencies, then gerontologists will likely want to learn their hows and whys as well. Each path has its basic and applied bifurcation and, given the importance of learning to people of all ages, each will engage the interests of aging researchers for some time to come.

REFERENCES Ahassir, M., & Hochstein, S. (1997). Task difficulty and the specificity of perceptual learning. Nature, 387, 401-406. Bailey, A.M., & Lauber, E.J. (1998, April). Learning to task switch and aging. Paper presented at the Cognitive Aging Conference, Atlanta, GA. Ball, K., Beard, B., Roenker, D., Miller, R., & Griggs, D. (1988). Age and visual search: Expanding the useful field of view. Journal of the Optical Society of America, 5, 2210-2219. Ball, K., Roenker, D., & Bruni, J. (1990). Developmental changes in attention and visual search throughout adulthood. In J. Enns (Ed.), The development of attention: Research and theory (pp. 489508). New York: Elsevier Science. Battig, W.F. (1972). Intratask interference as a source of facilitation in transfer and retention. In R.F. Thompson & J.F. Voss (Eds.), Topics in learning and performance (pp. 131-159). New York: Academic. Battig, W.F. (1979). The flexibility of human memory. In L.S. Cermak & F.I.M. Craik (Eds.), Levels of processing and human memory (pp. 23-44). Hillsdale, NJ: Erlbaum. Charness, N., & Campbell, J.I.D. (1988). Acquiring skill at mental calculation in adulthood: A task decomposition. Journal of Experimental Psychology: General, 117, 115-129. Craik, F.I.M., & Lockhart, R.S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11, 671-684. Czerwinski, M., Lightfoot, N., & Shiffrin, R.M. (1992). Automatization and training in visual search. American Journal of Psychology, 105, 271316. Duncan, J. (1986). Consistent and varied training in the theory of automatic and controlled information processing. Cognition, 23, 279-284.


Duncan, J., & Humphreys, G.W. (1989). Visual search and stimulus similarity. Psychological Review, 96, 433-458. Enns, J.T. (1989). Three-dimensional features that pop out in visual search. In D. Brogan (Ed.), Visual search (pp. 37-45). London: Taylor & Francis. Eriksen, C.W., & Schultz, D.W. (1979). Information processing in visual search: A continuous flow conceptualization and experimental results. Perception & Psychophysics, 25, 249-263. Farkas, M.S., & Hoyer, W.J. (1980). Processing consequences of perceptual grouping in selective attention. Journal of Gerontology, 35, 207-216. Fisher, D., & Tanner, N. (1992). Optimal symbol selection: A semiautomatic procedure. Human Factors, 34, 19-95. Fisk, A.D., Lee, M.D., & Rogers, W.A. (1991). Recombination of automatic processing components: The effects of transfer, reversal, and conflict situations. Human Factors, 33, 267-280. Fisk, A.D., McGee, N.D., & Giambra, L.M. (1988). The influence of age on consistent and varied semantic-category search performance. Psychology and Aging, 3, 323-333. Fisk, A.D., & Rogers, W.A. (1991). Toward an understanding of age-related memory and visual search effects. Journal of Experimental Psychology: General, 120, 131-149. Fisk, A.D., Rogers, W.A., & Giambra, L.M. (1990). Consistent and varied memory/visual search: Is there an interaction between age and response-set effects? Journal of Gerontology, 45, 81-87. Friedman-Hill, S., & Wolfe, J.M. (1995). Second-order parallel processing: Visual search for the odd item in a subset. Journal of Experimental Psychology: Human Perception and Performance, 21, 531551. Humphrey, D.G., & Kramer, A.F. (1997). Age differences in visual search for feature, conjunction and triple-conjunction targets. Psychology and Aging, 12, 704-717. Jacoby, L.L., & Craik, F.I.M. (1979). Effects of elaboration of processing at encoding and retrieval: Trace distinctiveness and recovery of initial context. In L.S. Cermak & F.I.M. Craik (Eds.), Levels of processing in human memory (pp. 1-21). Hillsdale, NJ: Erlbaum. Kramer, A.F., Martin-Emerson, R., Larish, J.F., & Anderson, G.J. (1996). Aging and filtering by movement in visual search. Journal of Gerontology: Psychological Sciences, 51B, P201-P216. Kramer, A.F., Strayer, D.L., & Buckley, J. (1991). Task versus component consistency in the development of automatic processing: A psychophysiological assessment. Psychophysiology, 28, 425-437.

139

Kray, J., & Lindenberger, U. (1998, April). Age differences in internal cognitive control: The role of coordination and reconfiguration during task switching. Paper presented at the Cognitive Aging Conference, Atlanta, GA. Logan, G.D. (1992). Attention and preattention in theories of automaticity. American Journal of Psychology, 105, 317-339. Madden , D.J. (1985). Adult age differences in memory-driven selective attention. Developmental Psychology, 4, 655-665. Madden, D.J., & Nebes, R.D. (1980). Aging and the development of automaticity in visual search. Developmental Psychology, 16, 377-384. Magill, R.A., & Hall, K.G. (1990). A review of the contextual interference effect in motor skill acquisition. Human Movement Science, 9, 241-289. Maxwell, S.E., & Delaney, H.D. (1990). Designing experiments and analyzing data. Belmont, CA: Wadsworth Publishing Company. Nakayama, K., & Mackeben, M. (1989). Sustained and transient components of focal visual attention. Vision Research, 29, 1625-1641. Plude, D.J., & Hoyer W.J. (1986). Age and the selectivity of visual information processing. Psychology and Aging, 1, 4-10. Rensink, R., & Ochs, E. (1992) Vscope [Computer Program]. Vancouver, BC, Canada: University of British Columbia VSearch Labs. Rogers, W.A. (1992). Age differences in visual search: Target and distractor learning. Psychology and Aging, 7, 526-535. Rogers, W.A., & Fisk, A.D. (1991). Are age differences in consistent-mapping visual search due to feature learning or attention training? Psychology and Aging, 6, 542-550. Rogers, W.A., Fisk, A.D., & Hertzog, C. (1994). Do ability-performance relationships differentiate age and practice effects in visual search? Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 710-738. Salthouse, T.A., & Somberg B.L. (1982). Skilled performance: Effects of adult age and experience on elementary processes. Journal of Experimental Psychology: General, 111, 176-207. Schneider, W. (1985). Toward a model of attention and the development of automatic processing. In M.I. Posner & O.S. Martin (Eds.), Attention and performance XI (pp. 475-492). Hillsdale, NJ: Erlbaum. Schneider, W., & Detweiler, M. (1987). A connectionist/control architecture of working memory. In G.H. Bower (Ed.), The psychology of learning and motivation (Vol. 21, pp. 53-118). New York: Academic Press.

140


Schneider, W., & Fisk, A.D. (1984). Automatic category search and its transfer. Journal of Experimental Psychology: Learning, Memory, and Cognition, 10, 1-15. Schneider, W., & Shiffrin, R.M. (1977). Controlled and automatic human information processing 1: Detection, search, and attention. Psychological Review, 84, 1-66. Scialfa, C., Jenkins, L., Hamaluk, E., & Skaloud, P. (1999). Age differences in the development of automaticity in conjunction search. Manuscript submitted for publication. Scialfa, C.T., & Joffe, K.M. (1998). Response times and eye movements in feature and conjunction search as a function of target eccentricity. Perception and Psychophysics, 60, 1067-1082. Scialfa, C.T., Kline, D.W., & Lyman, B.J. (1987). Age differences in target identification as a function of location and noise level: Examination of the useful field of view. Psychology and Acting, 2, 1419. Shiffrin, R.M., & Czerwinski, M.P. (1988). A model of automatic attention attraction when mapping is partially consistent. Journal of Experimental Psy-

chology: Learning, Memory, and Cognition, 14, 562-569. Shiffrin, R.M., & Dumais, S.T. (1981). The development of automatism. In J. R. Anderson (Ed.), Cognitive skills and their acquisition (pp. 111-140). Hillsdale, NJ: Erlbaum. Steinman, S.B. (1987). Serial and parallel search in pattern vision. Perception, 16, 389-398. Theeuwes, J. & Kooi, F.L. (1994). Parallel search for a conjunction of contrast polarity and shape. Vision Research, 34, 3013-3016. Treisman, A., & Gormican, S. (1988). Feature analysis in early vision: Evidence from search asymmetries. Psychological Review, 95, 15-48. Treisman, A., & Sato, S. (1990). Conjunction search revisited. Journal of Experimental Psychology: Human Perception and Performance, 16, 459-478. Wolfe, J.M. (1994). Guided Search 2.0. A revised model of visual search. Psychonomic Bulletin & Review, 1, 202-238. Wright, L.L., & Elias, J.W. (1979). Age differences in the effects of perceptual noise. Journal of Gerontology, 34, 704-708.

eif; 01jun99 - Semantic Scholar

eif; 01jun99 - Semantic Scholar

Suggest Documents

fm1; 01jun99 - The Netlib

Co-expression of human eIF-4G and poliovirus ... - Semantic Scholar

eIF-2a - Europe PMC

Silencing of the Host Factor eIF(iso)4E Gene ... - Semantic Scholar

Eukaryotic Initiation Factors (eIF) 2a and 4E

EIF VC Survey 2018 - European Investment Fund

EIF Brochure - European Commission - Europa EU [PDF]

Evolutionary Relationship Between Translation Initiation Factor eIF ...

EIF Magazine - Ever Increasing Faith Ministries

EIF Brochure - European Commission - Europa EU [PDF]

MiCS-5521 CO/VOC Sensor - EIF Wiki

Erythroid Expression of the Heme-Regulated eIF-2a Kinase

The GCD10 subunit of yeast eIF-3 binds the ... - CiteSeerX

Call for Proposals under the IMOS (EIF) Five Year ...

The European venture capital landscape: an EIF perspective

The European venture capital landscape: an EIF perspective

Toward New Sources of Competitiveness in Bangladesh - EIF Logo

subunit of the rat protein synthesis initiation factor eIF-2B

The 3rd European Interdisciplinary Forum 2015 (EIF 2015):

Plant eIF4B, eIF4F, and eIF(iso) - Plant Physiology

The eukaryotic initiation factor (eIF) - Proceedings of the National ...

regulated eIF-2a Kinase in Gel-filtered Reticulocyte Lysate

RNA-Dependent eIF-2 Kinase DAI - Molecular and Cellular Biology

Phosphorylated eIF-2 on Translation Initiation - Molecular and Cellular ...