Visual search and memory search – p. 1 Preprint from the paper that appeared in Aging, Neuropsychology, and Cognition This article may not exactly replicate the final version published in the journal. It is not the copy of record. © Swets & Zeitlinger
VISUAL SEARCH AND MEMORY SEARCH
Age differences in efficiency and effectiveness of encoding for visual search and memory search: A time-accuracy study
Paul Verhaeghen Syracuse University
Published in Aging, Neuropsychology, and Cognition
Visual search and memory search – p. 2
Abstract In a time-accuracy study, encoding for visual search and memory search was compared in a sample of 23 younger and 26 older adults. Older adults were found to be slower than younger adults in two respects (viz., processing was delayed, and speed-of-processing was lower). There was no reliable age difference in asymptotic performance when analyzed at the level of proportion correct. Age differences in speed-of-processing were larger in visual search than in memory search. The results run counter capacity- or resource accounts of cognitive aging, but may fit the framework that predicts larger age differences in visuo-spatial than lexical tasks (Myerson & Hale, 1993)
Visual search and memory search – p. 3
Age differences in efficiency and effectiveness of encoding for visual search and memory search: A time-accuracy study One of the more fascinating regularities in the field of cognitive aging is the orderly relation between latencies of older adults and those of younger adults, known since the 1980s (Cerella, 1990; Cerella, Poon, & Williams, 1980; Myerson, Hale, Wagstaff, Poon, & Smith, 1990). This regularity can be captured quite well by a single highly linear function relating latencies of older adults to corresponding latencies of younger adults. This result has led to the formulation of a ‘general slowing hypothesis’, that is, the hypothesis that a single ratio of processing rates of older over younger adults can describe slowing in central cognitive processes (e.g., Cerella, 1990, 1994). One challenge for current theory and research in the field is to find exceptions to this pattern. Recently, distinct age-related slowing functions have indeed emerged for different types or ‘domains’ of tasks (Hale & Myerson, 1996; Kliegl, Mayr, & Krampe, 1994; Mayr, Kliegl, & Krampe, 1996; Myerson & Hale, 1993; Sliwinski & Hall, 1998). Such functions have been termed ‘dissociations’ (Perfect & Maylor, 2000). One of the more puzzling dissociations, discovered in a meta-analysis by Sliwinski and Hall (1998) is the dissociation between visual search and memory search. Sliwinski and Hall analyzed the data at the level of full task latencies; the result has since been confirmed at the level of processing rate (i.e., time needed to scan a single item) by Myerson, Adams, Hale, and Jenkins (2000). Slowing ratios are noticeably smaller in memory search than in visual search. This is a puzzling finding, because the bulk of cognitive aging theory would predict a dissociation in the opposite direction. In memory search, participants are presented with a series of stimuli (typically digits or letters), which is removed after a brief viewing period. A target is then presented and participants have to decide whether this target was present in the original search set or not. In visual search, the order of affairs is reversed: here the participant is first presented with a target, and then has to decide whether this target is present or absent in a subsequent search array that remains in view throughout the search process. Memory search is typically considered a fixed- or limited-capacity task (Gegenfurtner & Sperling, 1993; McLean, Palmer, & Loftus, 1997; Palmer, 1990; Shibuya & Bundesen, 1988; Sperling, 1960), that is, stimuli are processed (at least to some degree) interdependently, and a fixed or limited amount of attentional capacity is distributed among them. Visual search (like many other forms of perceptual processing), however, is typically considered to be less dependent on such limitations (Bennett & Jaye, 1995; Duncan, 1980; McLean, 1999; McLean, Palmer, & Loftus, 1997; Palmer 1994; Shaw, 1984; Shiffrin & Gardner, 1972). One illustration of this difference can be found in the slopes relating reaction time to set size, indicating how fast each item in the search set is processed: In visual search, the slope averages about 15 ms-peritem for target-present trials (for a meta-analysis, see Wolfe, 1998); in memory search, slope is close to 40 ms-per-item (e.g., Sternberg, 1966). If we assume that attentional capacity is restricted in older adults compared to capacity in younger adults (see McDowd & Shaw, 2000, for a review), then the result that adult age differences are smaller in memory search than visual search runs against the grain of aging theory, where it is typically claimed that more capacity-limited tasks yield larger deficits than tasks that are less capacity-limited (e.g., Hasher & Zacks, 1988). Moreover, many theorists (e.g., Hasher & Zacks, 1988; Kliegl, Mayr, & Krampe, 1994; Moscovitch & Winocur, 1995; Salthouse, & Babcock, 1991) have emphasized that aging is associated with specific deficits in tasks involving an active working memory component. Thus, one might predict a larger age deficit for memory search tasks (which require a matching process within
Visual search and memory search – p. 4
the short-term buffer) than for visual search tasks (which do not seem to require much manipulation in working memory). It would be useful to have a direct experimental comparison of the two tasks. Recently, McLean (1999; see also McLean, Palmer, & Loftus, 1997) introduced a paradigm that allows for this direct comparison. In this paradigm, accuracy is the dependent measure and the search array is presented only briefly. By varying the time between the onset of the search array and that of a mask, the functional relation between presentation time and accuracy can be mapped out (see Bergen & Julesz, 1983; Eriksen & Spencer, 1969; Sagi & Julesz, 1985; Shiffrin & Gardner, 1972; for applications in the field of visual search, and see Harpur, Scialfa, & Thomas, 1995; Madden & Allen, 1991; and Scialfa & Harpur, 1994; for applications in the field of adult age differences in visual search). McLean et al. exploited the surface similarities of memory search and visual search and presented their participants with identical search arrays, simply showing the target either before the onset of the search array (visual search) of after its offset (memory search), and then derived complete time-accuracy functions for each task. Note that this paradigm allows for the investigation of two aspects of search unavailable from traditional reaction time experiments. First, the focus is on how fast and how accurately information is gathered from the display, that is, the time-accuracy function captures efficiency and effectiveness of encoding of the exact same stimulus display for respectively visual search and memory search. For visual search, this presumably includes the scanning and matching process. For memory search, all scanning and matching happens after this encoding stage, because the target is only presented after the offset of the search display. Thus, in memory search, the time-accuracy function measures how fast and accurate the participants encode the set of stimuli into working memory. In a series of five experiments with samples of younger adults using this paradigm, McLean (1997) has concluded that encoding for memory search clearly conforms to a limited-capacity model; the capacity limits for encoding for visual search (even if conjunction search is necessary) are clearly smaller than those for memory search, and close (but not equal) to performance expected if no capacity limitations exist. Thus, from a capacity-restriction view on cognitive aging, one should expect smaller age differences for visual search than for memory search in this version of the task as well. Second, time-accuracy functions allow for the disentangling of effectiveness and efficiency of processing (e.g., McClelland, 1979; Verhaeghen, 2000). Effectiveness of processing is indicated by the level of accuracy that can ultimately be reached, that is, when processing time is unlimited; efficiency of processing is indicated by how fast this level can be reached. One can thus examine whether age differences are present not only in efficiency of encoding, as would be expected from a processing speed model of aging (e.g., Salthouse, 1996), but in effectiveness as well (and this may or may not follow from slowing; Salthouse, 1996). In this paper I present results from a study mapping out time-accuracy functions for encoding for visual search and memory search. Letters were used, because these stimuli are highly overlearned in both young and older adults, and lend themselves well to both visual and memory search. Note that using letters implies that the visual search task will not be automatic (Schneider & Shiffrin, 1977), nor will it rely solely on bottom-up processing or pop-out effects (Treisman & Gormican, 1989; Wolfe, 1994). The key questions of the present study are: (a) can the meta-analytically demonstrated larger age difference in visual search than in memory search be confirmed in a paradigm that offers a direct comparison between the two?; and (b) is the locus of the possible age effects situated in efficiency or effectiveness of encoding, or both? Method
Visual search and memory search – p. 5
Participants Participants were 23 younger adults (22 females), all undergraduate students at Syracuse University, who received course credit for participating; and 26 older adults (13 females) recruited through newspaper advertisements and presentations at community centers, who received $10 for participating. The younger adults were on average 18.8 years old (SD = 0.9), and had completed 11.7 years of education (SD = 4.6); the older adults were on average 71.5 years old (SD = 5.0), and had completed 15.4 years of education (SD = 2.6). The age difference favoring the old in education was significant (t(47) = 4.00), as was the age difference favoring the old in the WAIS Vocabulary test (t(47) = 2.75; the younger adults scored 50.3, SD = 5.9; the older adults scored 54.7, SD = 5.3. These participants were selected from a larger pool of 32 younger and 35 older adults. As explained below, we fitted time-accuracy functions to each individual’s data. A participant was removed from the original pool if she proved an outlier on one or more of the six parameters estimated. This procedure was used to allow for ANOVA analyses on individually estimated parameters; the number of discarded data is quite substantial, but not unusual for this type of research (see, e.g., Verhaeghen, Vandenbroucke, & Dierckx, 1998; Verhaeghen et al., 2000). Comparison of the presentation time by accuracy plots of the mean data for the original and selected sample did not reveal any obvious differences in the pattern of results between the two groups. Task and procedure The tasks were a consistently mapped visual search task and a consistently mapped memory search task. The search array consisted of four letters, drawn randomly (without replacement) from a pool of 20 letters (all consonants minus the letter ‘J’); the target was randomly drawn from the set ‘d’, ‘f’, ‘h’, ‘k’, and ‘x’. On half of the trials, the target was present; on the other half, it was absent. Targets were always presented at the center of the screen, in lowercase; the midpoint of the letters of the search array, presented in uppercase, were arranged at the corners of a 40 by 40 mm square at the center of the screen, with a fixation cross at the center of the square. A sans-serif font was used, the uppercase letters in the search display were 31 mm high and 24 mm wide, the lowercase letter used as the target measured 10 by 6 mm. Lowercase and uppercase letters were used to avoid simple perceptual matching effects. The masking display consisted of ‘@’ symbols, in the same location, size, and font as the letters they were masking. Because participants were allowed to chose a comfortable viewing distance from the screen, the visual angle subtended by the stimuli varied for each participant. For each of the tasks, 200 trials (20 for each presentation time of the search array) were presented. Presentation times used were 30, 40, 60, 80, 100, 150, 200, 300, 500, and 1000 ms. Presentation times were randomly mixed. All randomization was done once, and this random sequence was used for all participants and for both tasks. In other words, the identity and order of presentation of the search array and targets, and their presentation rates were held constant across tasks and individuals. The only difference between the two tasks types, then, was that in memory search the target was presented after the search array , whereas in visual search the target was presented before the search array; for both tasks the stimuli (i.e., targets and distractors), their order of presentation and the corresponding presentation times were completely identical. A target present/absent response format was used. The participant indicated her response by typing the appropriate key on the computer keyboard (the ‘z’ key, masked with red tape, for ‘absent’; the ‘x’ key, masked with green tape, for ‘present’). Participants were instructed to be as fast and accurate as possible. Auditory feedback (a low tone) was given whenever the participant made a mistake. Participants were allowed to choose a comfortable viewing distance from the monitor. Half of the participants received the memory task first; the other half started with the visual search task. Prior to each of the tasks, participants
Visual search and memory search – p. 6
received 5 practice trials, with decreasing presentation times (3000, 1500, 1250, 1000, and 500 ms). The order of events within a trial was as follows. For visual search, a fixation cross appeared on the screen for 500 ms, followed by the target (1000 ms), which was masked for 250 ms, then a 250 ms blank screen, followed by the presentation of the search array (variable presentation time), which was terminated by a 250 ms mask, after which a question mark appeared on the screen and the participant could make her response. For memory search, a fixation cross appeared on the screen for 500 ms, followed by the presentation of the search array (variable presentation time), which was terminated by a 250 ms mask, followed by a 250 ms blank screen, after which the target and a question mark appeared on the screen and the participant could make her response. (The presence of a mask, a delay, and the additional masking effect of question mark or target and question mark should preclude iconic read-out of stimuli at the time of responding.) The start of each trial was self-initiated by the participant by a key press. All stimuli were presented on a 16 inch computer monitor, in white on black. Timing of the display was controlled by a Turbo Pascal timer routine with millisecond accuracy (Brysbaert, Bovens, d’Ydewalle, & Van Calster, 1989); onset times of the stimuli were synchronized with the start of the refresh cycle of the screen. Alpha level for all statistical tests was set at the .05 level. Results Time-accuracy functions for proportion correct Figure 1 (left-hand panel) presents proportion correct as a function of search array presentation time, separated by task and age group, along with the psychometric functions obtained from averaging the parameters of the individual time-accuracy functions. Proportion correct was rescaled to correct for chance (given the yes/no format, the a priori chance level was 1/2) before plotting and curve fitting, so that the chance level in the Figure is set to be equal to 0 and the maximum level at 1. (This is a linear transformation, and hence this transformation has no impact on the subsequent model fitting and statistical tests.) The function used for fitting was the logistic curve, as usual in this type of perceptual task (e.g., Blaser, Sperling, & Lu, 1999; McLean et al., 1997). The equation used for fitting the logistic function (taken from Glantz & Slinker, 1990) was P = c / {1 + exp [-4 b (PT - a)] / c}. (1) The function is defined in the measured range, that is, between 30 and 1000 ms. This equation describes P, proportion correct, as a function of PT, presentation time (for model estimation, I scaled presentation time in seconds). The parameters estimated by the model are a, b, and c. The logistic function is an S-shaped curve. Parameter a is a location parameter, that is, it indicates the point in presentation time where the curve reaches 50% of the asymptotic level, or, in other words, the mid-point of performance. All other parameters remaining equal, it can be taken as an indicator of differences in time needed to start up the encoding processes. Parameter b is the dynamic parameter, indicating how fast performance is rising, once processing has started. It is the derivative of the function at the point in presentation time were 50% of the asymptotic level is reached, or in other words, the tangent slope at the point where the positive acceleration in the S-shaped curve stops and the negative acceleration begins. It can be thought of as a slope parameter, with higher values indicating a steeper slope. Parameter c is the asymptote of the function, that is, the proportion correct that will be reached if an infinite amount of time were available. The logistic function fitted the data quite well (as indicated by R2 values): average R2 for the total sample was 0.74 (SD = 0.10). The logistic function fitted the data better and with less interindividual variability than another frequently used function, the negative exponential (e.g., Verhaeghen et al., 1997, 1998; McClelland, 1979), where average R2 for the total
Visual search and memory search – p. 7
sample was 0.64 (SD = 0.18), t(48) = 3.55. There was no reliable difference in fit of the logistic curve for younger versus older adults: average R2 = 0.74 (SD = 0.12) and 0.73 (SD = 0.09), respectively, t(47) = 0.19. It should be noted that the curves represented in Figure 1 are obtained by averaging parameters from the curves obtained from each individual, and not by fitting the averaged data points for each presentation time (represented by the symbols in the figure). Given that the equation used for fitting the individual data is non-linear, averaging the parameter values to obtain the average curve does not produce the same result as fitting the equation to the average data (Estes, 1956), and hence the average observed data do not necessarily hug the average curve closely, even if the fit of individuals to the equation is quite good. A repeated measures ANOVAs was conducted for each of the parameters with task type (visual search vs. memory search) as the within-subject and age category (younger vs. older) as the between-subject variable. The main effect of type of task was reliable for all parameters (for a, F(1, 47) = 42.90; for b, F(1, 47) = 34.46; for c, F(1, 47) = 53.34), indicating that encoding for memory search was delayed, less fast, and ultimately less accurate than encoding for visual search. The main effect of age was significant for a, F(1, 47) = 15.04, and for b, F(1, 47) = 7.65, but not for c, F(1, 47) = 0.09. This indicates that encoding of older adults starts after a longer delay, and develops more slowly than that of younger adults, but that the level of asymptotic performance of older adults is not reliably smaller than that of younger adults. The age by task interaction was significant for the b parameter only, F(1, 47) = 7.93; for a, F(1, 47) = 3.60, for c, F(1, 47) = 1.20. A follow-up independent sample t test shows that the age effect on b is significant for visual search, t(47) = 2.86, but not for memory search, t(47) = 0.44. It can be argued, however, (e.g., Verhaeghen, 2000; Verhaeghen et al., 1997) that interactions in dynamic parameters (i.e., a and b) should be tested after the parameters have been subjected to logarithmic transformation, in order to take the multiplicative effects of general slowing into account. This transformation did not change the pattern of results: there was a significant age by task type interaction for log(b), F(1, 47) = 5.90, but not for log(a), F(1, 47) = 0.01. Therefore, a dissociation between performance of younger and older adults can be reliably demonstrated in the slope parameter only: The age difference in speed of encoding for visual search is larger than the age difference in speed of encoding for memory search. The right-hand panel of Figure 1 shows a plot of the same data as iso-temporal accuracy state traces (Verhaeghen, 2000), that is, performance in visual search as a function of corresponding performance in memory search for each of the 10 presentation times within each age group. Prior to plotting, a logit (a.k.a. log-odds) transformation was applied to the data (logit = ln[P/(1-P)]; with P = proportion correct); this transformation was applied to linearize the data. The logit transformation represents the most common and general link function for modeling proportions in linear models (McCullagh & Nelder, 1989). The graph illustrates the dissociation found in the slope parameter: both lines diverge from the diagonal, showing that visual search is an easier task than memory search. The deviation is steeper in younger adults, confirming the ANOVA finding that the task difference is larger in this age group than in older adults. Time-accuracy data for proportion correct as a function of stimulus position Figure 2 reports the time-accuracy data with proportion correct as the dependent variable, for each of the two tasks for each of the two age groups, separated by the position the target occupies in the display. Note that in these plots, proportion correct is hit rate, corrected for chance, that is, as in Figure 1, the data were rescaled so that zero indicates chance level. The plots suggest that encoding for memory search is a serial process, proceeding in a fixed order. The qualitative pattern of the data is very similar in younger and older adults.
Visual search and memory search – p. 8
Overall, the target is recognized earlier when it is presented in the upper left quadrant of the display; next, encoding proceeds to the element in the upper right quadrant; then to the element in the lower right quadrant; and lastly to the element in the lower left quadrant. In other words, participants are encoding the display in a clockwise fashion, starting at the upper left. This leads to real delay effects in the not-yet processed items, that is, performance stays close to chance level, or – in the case of older adults – even below chance level. The picture is less clear for visual search. There it seems that, for both younger and older adults, processing initially occurs in a parallel fashion, and only at around 100 ms for the young, and 175 ms for the old, a clear separation for the last two positions occurs. Interestingly, in older adults, it seems as if for the longest two presentation times, performance is the inverse from what could be expected from the order of processing in memory search, that is, performance is lowest for the item in the upper left quadrant, and best for the item in the lower left quadrant. This might imply that earlier answers get lost as processing progresses, or that attention lingers insufficiently at the locations visited earliest. Thus, with respect to asymptotic performance as a function of quadrant, visual search in older adults seems to be qualitatively different from both visual search performance in the young and their own memory search performance. As a simple formal test for order effects, I examined aggregated performance for the five shortest presentation times (i.e., 100 ms, or shorter). The results are depicted in Figure 3. This aggregate should give us some idea of which items are processed more fully at short presentation times, but it remains a necessarily crude index. A repeated-measures ANOVA was conducted, with quadrant and task type as within-subject factors, and age group as the between-subject factor. All main effects proved significant, indicating that older adults performed at a lower overall level, F(1, 47) = 37.80, MSE = 0.09; that visual search yielded higher performance than memory search, F(1, 47) = 18.79, MSE = 0.04; and that performance was not identical for all quadrants, F(3, 141) = 30.04, MSE = 0.03. The age by condition interaction proved significant, F(1, 47) = 6.72, MSE = 0.04, indicating that age differences are larger in visual search than in memory search. When quadrants were ordered clockwise from the upper left, the linear, F(1, 47) = 49.28, MSE = 0.05, but not the quadratic quadrant trend, F(1, 47) = 0.04, MSE = 0.05, was significant, suggesting that this clockwise order captures the overall order of processing quite well. The quadrant main effect was qualified by a reliable condition by quadrant interaction, F(3, 141) = 9.92, MSE = 0.02. As can be seen in Figure 3, the clockwise order emerges clearly in memory search, but is less clear in visual search. In fact, within visual search, the cubic trend was significant, F(1, 47) = 23.78, MSE = 0.01, due to the bump for the lower right-hand quadrant. Importantly, however, in the multivariate analysis there was no significant age by quadrant interaction, F(3, 141) = 0.87, MSE = 0.03, nor was the age by condition by quadrant interaction significant, F(3, 141) = 1.62, MSE = 0.02. Even when examined within the visual search task alone, the age by quadrant interaction failed to reach significance, F(1, 47) = 1.96, MSE = 0.04. This result indicates that processing order did not vary with age. Summarized, the results from visual inspection of the quadrant data in Figure 2 and those of the more formal test using short presentation times converge on three conclusions. First, encoding for visual search and memory search proceeds in a particular serial order. The analyses differ, however, with regard to the suggested order of processing: visual inspection suggests a consistent order from lower right to lower left; the formal test suggests that this is the order for memory search, but that the order for visual search is less clear. A second point of agreement is that older adults process the displays more slowly. A third uncontroversial conclusion, important for the aims of the present study, is that there is no interaction involving age and order of processing, indicating that younger and older adults process the elements of the displays in the same serial order.
Visual search and memory search – p. 9
Discussion This research was guided by two questions. The first question was whether the results from a meta-analysis, demonstrating that larger age differences are present in visual search than in memory search, could be replicated in a paradigm which offered a direct comparison between encoding processes for the two tasks. The answer is a simple yes, as explained below: Older adults are slower than younger adults in encoding information for visual search; they are as fast as younger adults in encoding information for memory search. The second question concerned the locus of age differences in visual search and memory search. I examined time-accuracy functions for both tasks in a group of younger and older adults, which allows for the examination of age differences in dynamic parameters of the curve (i.e., efficiency of processing), and in asymptotic performance (i.e., the ultimate effectiveness of encoding). Three main results concerning age differences in memory search and visual search emerge from this study. First, an age effect was found on the location parameter in both tasks. This implies that encoding of older adults for memory search and visual search is delayed when compared with encoding in younger adults. For visual search, the delay effect (obtained by subtracting the location parameters of the young from those of the old) equals about 50 ms; for memory search, it equals about 100 ms; this differential effect of task on age differences, however, did not reach significance. The reason for this age difference remains open. It may, for instance, reflect age differences in early stages of processing, such as visual perception (see Schieber & Baldwin, 1996, for an overview), or speed of iconic read-out (Coyne, Burger, Berry, & Botwinick, 1987; Gilmore, Allan, & Royer, 1986). The reader may note that these early processes appear to be important for good performance in the task: The curves do not rise from the origin, but performance is already elevated quite high above chance at a presentation time of 30 ms, presumably because of very fast transfer of part of the stimulus display to short-term memory. Consistent with this interpretation, the quadrant data reported in Figure 2 indeed suggest that participants perform well on the first element of the display (i.e., the upper left hand quadrant), even at very fast presentation times. Second, there was a main effect of age on the slope parameter. This implies that, once processing has started, older adults proceed more slowly in extracting information from the display. This result was qualified by a significant interaction. In fact, the effect was only significant for visual search, where the slowing factor (obtained by the young/old ratio of the slopes) is close to 2. In memory search, the slowing factor equaled 1.1, which is not significantly different from 1 (as can be derived from the results of the t test). These slowing ratios are very close to the young/old ratios obtained in the Sliwinski and Hall (1998) metaanalysis: 1.2 for memory search (not significantly different from 1), and 2.0 for visual search. This slope difference, then, may well be the locus of the age dissociation found in reaction time studies (Sliwinski & Hall, 1998): Relative to younger adults, older adults encode information more slowly in visual search, but they are equally fast in encoding information for memory search. Third, no age difference was found for asymptotic performance, that is, ultimately, young and older adults reach an identical level of performance in both tasks. When examined at the level of discriminability, it seems that visual search might yield an age difference in asymptotic performance, which would then be a further source of the previously reported dissociation, but because these data were derived at the group level, this assertion remains untestable. The prudent conclusion is that age effects in visual search and memory search are situated in the dynamics, or efficiency, of encoding, and not in final effectiveness. This conclusion, of course, only holds within the limits of the present task. It is feasible, for instance, that displays with a larger number of elements might lead to age differences in ultimate accuracy. The reader may note that the combination of an age difference in slope
Visual search and memory search – p. 10
with the absence of an age difference in asymptote is in accordance with findings that age differences grow smaller with longer presentation times in visual search (Harpur et al., 1995; Madden and Allen, 1991). It should be stressed that, given the bulk of cognitive aging theory, the dissociation in encoding speed is a surprising result. First, it is an example of an anti-complexity effect. It has often been asserted (e.g., Cerella et al., 1980) that more complex tasks (operationalized as task difficulty) yield larger age differences. Encoding for visual search, in both young and older adults, is an easier task than encoding for memory search. The processes relevant for the task start earlier (as demonstrated by a task type main effect in the location parameter), they proceed faster (as demonstrated by a task type main effect in the slope parameter), and they lead to higher asymptotic performance (as demonstrated by a task type main effect in the asymptote parameter). Thus, in the present case, it is clearly the less difficult task that yields the larger age difference. Second, McLean et al. (1997) have demonstrated (in samples of younger adults) that under the particular paradigm used in this study, memory search is clearly a capacity-limited task, whereas visual search places smaller demands on attentional capacity. Cognitive aging theory (e.g., Hasher & Zacks, 1988; Salthouse, 1991) typically states that age differences are larger in tasks that are capacity-driven than in tasks that are not. Consequently, the present result is in need of explanation. Essentially two types of explanations for age-related differences can be advanced: local (tied to task characteristics) or global (referring to similarities in age differences across broader task domains). The fact that our slowing factors for encoding closely resemble those found in a meta-analysis suggests that the dissociation found in the present study is not tied to the particulars of stimuli or procedure, but may be tied to broader characteristics of visual versus memory search. One very general and consistent age-related dissociation that may be relevant is the dissociation between lexical versus non-lexical (i.e., visuo-spatial) processing (Hale & Myerson, 1996; Lima, Hale, & Myerson, 1991; Myerson & Hale, 1993). Slowing factors are close to 1.5 for lexical tasks, and 2.0 for non-lexical tasks (Lima et al., 1991), but these slowing factors include the effects of sensorimotor slowing, which do not figure in the current slope measures. For the results of the present study to fit into this framework, we need to assume that visual search is in essence a visuo-spatial task, and memory search is a lexicalverbal task (a leap of faith also taken by Myerson, Adams, Hale, & Jenkins, in press). Note, however, that the lexical/non-lexical dissociation is merely an empirical generalization and that there is currently no convincing explanation of the underlying mechanism or mechanisms that drive this dissociation (see, e.g., Jenkins et al., 2000). Therefore, this explanation remains necessarily speculative at this point. If the memory search/visual search dissociation is indeed an instance of the lexical/non-lexical dissociation, then this suggests that the latter dissociation is driven mainly by the type of processing involved, and not by the type of stimuli used (see also Verhaeghen et al., 2001). One conclusion from this study is then that memory search performance in young and older adults is very similar. The only difference between the two age groups is the location parameter, implying that older adults merely need more time before processing starts. Moreover, a closer look at the data, separated by the position occupied by the elements in the display suggests that both groups encode the display in a serial, clockwise fashion, starting from the upper left. In both age groups, the delay effects between items are very similar. One possible age difference is that older adults seem to have more trouble in identifying items that are not situated at the current focus of attention, as exemplified by performance at or below chance for items not yet visited. For visual search, the order of processing seems less clear, maybe suggesting an initial parallel phase, which is then, after about 100 ms for the young and 175 ms for the old, supplanted by a serial mode (see also Wolfe’s Guided Search model, 1994). This result is
Visual search and memory search – p. 11
clearer in younger than in older adults; for older adults, the curves are much more entangled at short durations. At the longer durations, for older adults, but not for younger adults, an order effect emerges in level of accuracy. It is, however, the reverse order as given by the order of processing in the memory search task: anti-clockwise, starting from the lower left position. One possible explanation is that older adults, like younger adults, do try to process the display in a clock-wise fashion, but that performance suffers from the effects of item interference, item replacement, item decay, or combinations thereof, maybe because attention span is smaller in older adults (e.g., Ball, Owsley, Sloane, Roenker, & Bruni, 1993; Scialfa, Kline, & Lyman, 1987), or because of age-related difficulties with executive control over processing (e.g., Kliegl, Mayr, & Krampe, 1994). The attention span explanation seems unlikely, given the small size of the search set and the absence of age differences in the memory search condition, but it is possible that a different type of attentional mechanism operates during visual search than memory search. It must be noted that such explanations remain at the local level, that is, it puts the locus of the effect in task differences. Given that this dissociation seems to be operating within the wider context of a general verbal processing versus visuo-spatial processing dissociation, such local explanations may fall crucially short in a wider theoretical context. In sum, this study replicates the meta-analytic findings of smaller age differences in memory search than in visual search in a paradigm that offers a direct comparison between the two tasks, using the exact same search displays for both tasks. These differences are already present at the stage of encoding of the search array and appear to be due to speed of deployment of the encoding process itself.
Visual search and memory search – p. 12
References Ball, K. K., Beard, B. L., Roenker, D. L., Miller, R. L., & Griggs, D. S. (1988). Age and visual search: Expanding the useful field of view. Journal of the Optical Society of America A, 5, 2210-2219. Bennett, P.J. & Jaye, P.D. (1995) Letter localization, not discrimination, is constrained by attention. Canadian Journal of Experimental Psychology, 49, 460-503. Bergen, J.R. and Julesz, B. (1983). Parallel versus serial processing in rapid pattern discrimination. Nature, 303, 696-698. Blaser, E., Sperling, G., and Lu, Z-L. (1999). Measuring the Amplification of Attention. Proceedings of the National Academy of Sciences, USA, 96, 11681-11686. Brysbaert, M., Bovens, N., D'Ydewalle, G., Van Calster, J. (1989). Turbo Pascal timing routines for the IBM microcomputer family. Behavior Research Methods, Instruments, & Computers, 21, 73-83. Cerella, J. (1990). Aging and information processing rate. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (3rd. ed., pp. 201-221). San Diego, CA: Academic Press. Cerella, J. (1994). Generalized slowing in Brinley plots. Journals of Gerontology: Psychological Sciences, P49, 65-71. Cerella, J., & Poon, L. W., & Williams, D. M. (1980). Age and the complexity hypothesis. In L. W. Poon (Ed.), Aging in the 1980's (pp. 332-340). Washington: American Psychological Association. Coyne, A. C., Burger, M. C., Berry, J. M., & Botwinick, J. (1987). Adult age, information processing, and partial report performance. Journal of Genetic Psychology, 148, 219-224. Duncan, J. (1980). The locus of interference in the perception of simultaneous stimuli. Psychological Review, 87, 272-300. Eriksen, C. W., & Spencer, T. (1969). Rate of information processing in visual perception: Some results and methodological considerations. Journal of Experimental Psychology Monograph, 79(2), 1-16. Estes, W. K. (1956). The problem of inference from curves based on group data. Psychological Bulletin, 53, 134-140. Gegenfurtner, K. & Sperling, G. (1993). Information transfer in iconic memory experiments. Journal of Experimental Psychology: Human Perception and Performance, 19, 845-866. Gilmore, G. C., Allan, T. M., & Royer, F. L. (1986). Iconic memory and aging. Journal of Gerontology, 40, 183-190. Glantz, S. A., & Slinker. B. K. (1990). Primer of applied regression and analysis of variance. New York: McGraw-Hill. Green, D. M. and Swets, J. A. (1966). Signal detection theory and psychophysics. New York: Wiley, Hale, S., & Myerson, J. (1996). Experimental evidence for differential slowing in the lexical and nonlexical domains. Aging, Neuropsychology, and Cognition, 3, 154-165. Harpur, L. L., Scialfa, C. T., & Thomas, D. M. (1995). Age differences in feature search as a function of exposure duration. Experimental Aging Research, 21, 1-15. Hasher, L., & Zacks, R. T. (1988). Working memory, comprehension, and aging: A review and a new view. In G. H. Bower (Ed.), The psychology of learning and motivation (Vol. 22, pp. 193-225). San Diego, CA: Academic Press. Jenkins, L., Myerson, J., Joerding, J. A., & Hale, S. (2000). Converging evidence that visuospatial cognition is more age-sensitive than verbal cognition. Psychology and Aging, 15, 157-175.
Visual search and memory search – p. 13
Kliegl, R., Mayr, U., & Krampe, R. T. (1994). Time-accuracy functions for determining process and person differences: An application to cognitive aging. Cognitive Psychology, 26, 134-164. Lima, S. D, Hale, S., Myerson, J. (1991). How general is general slowing? Evidence from the lexical domain. Psychology and Aging, 6, 416-425. Madden, D. J., & Allen, P. A. (1991). Adult age differences in the rate of information extraction during visual search. Journals of Gerontology: Psychological Sciences, 46, 124126. Mayr, U., Kliegl, R., & Krampe, R. (1996). Sequential and coordinative processing dynamics in figural transformation across the life span. Cognition, 59, 61-90. McClelland, J. L. (1979). On the time relations of mental processes: An examination of systems of processes in cascade. Psychological Review, 86, 287-330. 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). Mahwah, N. J.: Erlbaum. McCullagh, P., & Nelder, J. A.. (1989). Generalized linear models (2nd ed.). New York: Chapman & Hall/CRC. McLean, J. E. (1999). Processing capacity of visual perception and memory encoding. Unpublished Doctoral Dissertation, University of Washington, Seattle. McLean, J. E., Palmer, J. & Loftus, G. R. (1997, November). Comparison of set-size effects in visual search and memory. Paper presented at the annual Meetings of the Psychonomic Society, Philadelphia, PA. Moscovitch, M., & Winocur, G. (1995). Frontal lobes, memory, and aging. In J. Grafman, K. J. Holyoak, & F. Boller (Eds.), Structure and functions of the human prefrontal cortex (pp. 119-150). New York: New York Academy of Sciences. Myerson, J., Adams, D. R., Hale, S., & Jenkins, L. (in press). Analyses of differences in processing speed: Brinley plots, Q-Q plots, and other conspiracies. Psychonomic Bulletin and Review. Myerson, J., & Hale, S. (1993). General slowing and age invariance in cognitive processing: The other side of the coin. In J. Cerella, J. Rybash, W. Hoyer, & M. L. Commons (Eds.), Adult information processing: Limits on loss (pp. 115-141). San Diego, CA: Academic Press. Myerson, J., Hale, S., Wagstaff, D., Poon, L. W., & Smith, G. A. (1990). The information-loss model: A mathematical theory of age-related cognitive slowing. Psychological Review, 97, 475-487. Palmer, J. (1990). Attentional limits on the perception and memory of visual information. Journal of Experimental Psychology: Human Perception and Performance, 16, 332-350. Palmer, J. (1994). Set-size effects in visual search: The effect of attention is independent of the stimulus for simple tasks. Vision Research, 34, 1703-1721. Perfect, T. J., & Maylor, E. A. (2000). Rejecting the dull hypothesis: The relation between method and theory in cognitive aging research. In T. J. Perfect & E. A. Maylor (Eds.), Models of cognitive aging (pp. 1-18). Oxford, UK: Oxford University Press. Sagi, D., & Julesz, B. (1985). Fast noninertial shifts of attention. Spatial Vision, 1, 141-149. Salthouse, T. A. (1991). Theoretical perspectives on cognitive aging. Hillsdale, NJ: Erlbaum. Salthouse, T. A. (1996). The processing-speed theory of adult age differences in cognition. Psychological Review, 103, 403-428.
Visual search and memory search – p. 14
Salthouse, T. A., & Babcock, R. L. (1991). Decomposing adult age differences in working memory. Developmental Psychology, 27, 763-776. Schieber, F., & Baldwin, C. L. (1996). Vision, audition, and aging research. In F. Blanchard-Fields & T. M. Hess (Eds.), Perspectives on cognitive change in adulthood and aging (pp. 122-162). New York: McGraw-Hill. Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human information processing: I. Detection, search, and attention. Psychological Review, 84, 1-66. Scialfa, C. T., & Harpur, L. L. (1994). Effects of similarity and duration on age differences Scialfa, C. T., Kline, D. W., & Lyman, B. J. (1987). Age differences in target identification as a function of retinal location and noise level: An examination of the useful field of view. Psychology and Aging, 2, 14-19. Shaw, M. L. (1984). Division of attention among spatial locations: A fundamental difference between detection of letters and detection of luminance increments. In H. Bouma & D. G. Bouwhuis (Eds.), Attention and performance X (pp. 109—121). Hillsdale, NJ: Erlbaum. Shibuya, H., & Bundesen, C. (1988). Visual selection from multielement displays: Measuring and modeling effects of exposure duration. Journal of Experimental Psychology: Human Perception and Performance, 14, 591-600. Shiffrin, R. M., & Gardner, G. T. (1972). Visual processing capacity and attentional control. Journal of Experimental Psychology, 93, 72-82. Sliwinski, M., & Hall, C. B. (1998). Constraints on general slowing: A meta-analysis using hierarchical linear models with random coefficients. Psychology and Aging, 13, 164175. Sperling, G. (1960). The information available in brief visual presentations. Psychological Monographs, 74, 1-29. Sternberg, S. (1966). High-speed scanning in human memory. Science, 153, 652-654. Treisman, A. M., & Gelade, G. (1980). A feature integration theory of attention. Cognitive Psychology, 12, 97-136. Verhaeghen, P. (2000). The parallels in beauty’s brow: Time-accuracy functions and their implications for cognitive aging theories. In T. J. Perfect and E. A. Maylor (Eds.), Models of cognitive aging (pp. 50-86). Oxford: Oxford University Press. Verhaeghen, P., Kliegl, R., & Mayr, U. (1997). Sequential and coordinative complexity in time-accuracy function for mental arithmetic. Psychology and Aging, 12, 555564. Verhaeghen, P., Palfai, T., Cerella, J., Buchler, N., Johnson, M. P., D'Eredita, M., Green, D. R., Hoyer, W. J., & Makekau, M. (2000). Age-related dissociations in timeaccuracy functions for recognition memory: Utilizing semantic support versus building new representations. Aging, Neuropsychology, and Cognition, 7, 260-272. Verhaeghen, P., Vandenbroucke, A., & Dierckx, V. (1998). Growing slower and less accurate: The effects of age on time-accuracy functions for recall from episodic memory. Experimental Aging Research, 24, 3-19. Wolfe, J. M. (1994). Guided Search 2.0: A revised model of visual search. Psychonomic Bulletin and Review, 1, 202-238. Wolfe, J. M. (1998). What can 1 million trials tell us about visual search? Psychological Science, 9, 33-39.
Visual search and memory search – p. 15
Author Notes Paul Verhaeghen, Department of Psychology and Center for Health and Behavior, Syracuse University. This research was supported in part by a grant from the National Institute on Aging (AG-16201). I would like to thank Geoffrey Loftus, Jennifer McLean and John Palmer for making the manuscript of the 1997 McLean et al. presentation at the Meeting of the Psychonomic Society and Jennifer McLean’s dissertation available. Silvie C. Semenec and David W. Steitz collected the data. I acknowledge helpful discussions with John Cerella, William J. Hoyer, Martin Sliwinski, and Tibor Palfai. I thank one reviewer for suggesting a working alternative to my original formal analysis of the quadrant effect. Correspondence concerning this article should be addressed to Paul Verhaeghen, Department of Psychology, 430 Huntington Hall, Syracuse University, Syracuse, New York 13244-2340. Electronic mail may be sent to
[email protected].
Visual search and memory search – p. 16
Table 1. Parameters of the best-fitting logistic functions relating proportion correct to presentation time (see Figure 1), along with standard deviations.
Parameter Location (a) Younger Older Slope (b) Younger Older Asymptote (c) Younger Older
Visual search Mean
SD
Memory search Mean
SD
0.028 0.071
0.034 0.050
0.101 0.203
0.114 0.110
7.61 3.82
5.96 3.02
1.89 1.76
1.20 0.84
.94 .90
.06 .09
.78 .76
.15 .12
Visual search and memory search – p. 17
Figure Captions Figure 1. Left-hand panel: Performance (proportion correct) as a function of presentation time of the search array, separated by age group and task, along with the best fitting average logistic functions, derived from individual fits. Right-hand panel: state-trace plot relating performance (logit of proportion correct) in visual search (VS) to performance in memory search (MS) for each of the two age groups (unfilled circles: younger adults; filled circles: older adults). Error bars represent 1 standard error. Figure 2. Performance (proportion correct) as a function of presentation time, separated by position of the target in the display, for each of the two age groups and the two tasks. Figure 3. Performance (proportion correct) as a function of target position in the display, for each of the two age groups and the two tasks, for short presentation times (smaller or equal to 100 ms) (V= visual search; M = memory search; UL = upper left; UR = upper right; LL = lower left; LR = lower right). Error bars represent 1 standard error.
Visual search and memory search – p. 18
1.0
4.0
0.9
3.5 3.0
0.7 0.6 0.5 0.4
Younger, VS Older, VS Younger, MS Older, MS
0.3 0.2 0.1
Logit (VS)
Percent correct
0.8
2.5
Younger
2.0
Older
1.5 1.0 0.5
0.0
0.0 0.0
0.1
0.2
0.3
0.4
0.5
0.6
Presentation time
0.7
0.8
0.9
1.0
0.0
0.5
1.0
1.5
2.0
2.5
Logit (MS)
3.0
3.5
4.0
Visual search and memory search – p. 19
Proportion correct
Visual Search, Younger Adults
Memory Search, Younger Adults
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0.0 0.0
0.2
0.4
0.6
0.8
1.0
0.0 0.0
Presentation Time
Proportion correct
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.4
-0.2 -0.4 0.0
0.2
0.4
0.6
0.8
Presentation Time
0.6
0.8
1.0
Memory Search, Older Adults
0.2
Upper Left Upper Right Lower Left Lower Right
0.0
0.4
Presentation Time
Visual Search, Older Adults
0.2
0.2
0.0 -0.2
1.0
-0.4 0.0
0.2
0.4
0.6
0.8
Presentation Time
1.0
Percent correct at presentation times up to 100 ms
Visual search and memory search – p. 20
1.0
Younger adults Older adults
0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0
MUL
MUR
MLR
MLL
VUL
VUR
VLR
VLL