Global Study For Intro 1

Experimental Aging Research, 26: 285 314, 2000 Copyright 2000 Taylor & Francis 0361-073X/00 $12.00 .
00
The Fate of Global Precedence With Age

Raymond Bruyer and Jean-Christophe Scailquin Cognitive Neuropsychology Unit (NECO), Louvain University (UCL), Louvain-la-Neuve, Belgium
The literature is virtually devoid of studies examining the e ect of aging on the ``global precedence e ect (Navon , D. [1977 ]. Cognitive Psychology, 9, 353 383 ). In this paradigm , global letters formed with local letters are shown, and the subject has to recognize either the local or the global letters. The relation between the global and the local letters is either congruent , neutral , or con icting. Five experiments are reported , with ve sets of 16 young and ve sets of 16 elderly, healthy , adult subjects . The global precedence e ect was observed in both age groups in the basic perceptual experiment as designed by Navon (Experiment 1). In addition , young and elderly subjects were able to process separa tely the global and the local shapes when attention was not directed towards a speci c level (Experiment 1a). However, subpopulations emerged, especially in the elderly, depending on their ability to ``resist to the interference of the global shape upon the processing of the local form. This could support the few indirect published data suggesting that global precedence tends to diminish or to disappear with age . In experi ments 2 and 2a, subjects were retained only if they did process the local level better than at random . Global precedence was con rmed in both young and elderly subjects . Again, elderly subjects manifested an increased sensitivity to interference . Moreover , a general e ect of age remained, even in the control (neutral ) conditions , suggesting difficulties of encoding in elderly. Therefore , in Experiment 3, encoding conditions of both age groups were equated by increasing exposure duration of the
Received 25 March 1999 ;accepted 15 October 1999. These experiments were made possible by the grants 92/97-158 and 98/03-215 (`Àctions de Recherche Concertee) from the Government of the French-Speaking Community of Belgium. Thanks are due to R. Perrera, I. Clercx, C. Mantovan, and C. Crombe, who collected part of the data. Address correspondence to R. Bruyer, UCL-NECO, Place du Cardinal Mercier 10, B-1348 Louvain-la-Neuve, Belgium. E-mail: bruyer @neco.ulc.ac.be
285
286
R. Bruyer and J.-C. Scailquin
material for elderly. Subjects of Experiment 3 were also submitted to a Stroop test, to verify whether the increased sensitivity to interference in the elderly was speci c to the ``Navon task or the expression of a general e ect of aging on inhibition mechanisms. The global precedence phenomenon was observed in young and elderly subjects , with no sign of an e ect of age under these conditions ; in elderly, furthermore , the interference e ect did not appear to result from a general aspeci c de cit of inhibition mechanisms. It thus appears that the global precedence phenomenon resists well the e ects of aging , but that subpopulations of elderly subjects should be considered in future studies .
The phenomenon of global precedence in the visual perception of complex stimuli has been hypothesized since the beginning of the 20th century, and has been rmly demonstrated by Navon (1977). Surprisingly enough, in spite of the considerable amount of experiments which have since replicated the phenomenon under several methodological variations, very little is known about the e ect of normal aging on global prevalence. The purpose of the present series of experiments was to ll this gap in the literature.
GLOBAL PRECEDENCE Studies in experimental psychology have shown, for a long time, that the visual perception of parts or local elements of a whole is in uenced by the perception of the whole, which indicates that the whole is ``more than the sum of the parts and that the part could even be the perceptual by-product of the whole. This kind of context e ect was clearly established by the Gestaltists between the two World Wars, as well as more recently (e.g., Palmer, 1980 ; Palmer & Bucher, 1981). Some current wellknown e ects can also be conceived within this framework: the word superiority e ect (Estes, 1975 ;Reicher, 1969), the object superiority e ect (Weisstein & Harris, 1974), and the face superiority e ect (Davido & Donnelly, 1990 ; Thompson, 1980 ; Valentine & Bruce, 1985 ; Young, Hellawell, & Hay, 1987). However, in these studies, the global form (the ``whole) di ers structurally from the local component elements (the ``parts) ;for example, small triangles (part) forming a line (whole), letters composing a word, an eye within a face, or tub in a bathroom. Therefore, the so-called ``superiority, ``primacy, ``priority, ``prevalence, or ``precedence of the whole over the parts could result from structural (complexity) and/or familiarity di erences, facilitating the discrimination
Age and Global Precedence
287
or identi cation of the whole over the discrimination or identi cation of the parts. In a seminal publication, Navon (1977) solved this uncertainty by designing particular ``compound (or hierarchical) stimuli where similar shapes (except for their size, of course) formed both the local and the global levels. Thus, Navon (1977) displayed brie y upper-case printed letters formed with smaller upper-case letters. For a given large letter, all the small component letters were identical, and they were either identical or not to the large one. A directed-attention paradigm was used. In one condition, the subjects (young adults) had to identify the large letter ; in the other condition, the small component letters. The global identi cation (large letters) was faster than the local one (small letters) and was una ected by the con icting condition (i.e., when the small component letters di ered from the large), whereas the identi cation of the small letters slowed down when they di ered from the large. This pattern of results was considered to be a demonstration of the ``global precedence phenomenon, that is to say, the perception of a complex pattern would proceed from global to local analysis. This e ect could result from an early advantage of the processing of low resolution (low spatial frequencies) information (Hughes, Nozawa, & Kitterle, 1996). This original study gave rise to many publications, including replications, generalization to other kinds of stimuli (nonverbal material, digits), populations (infants, children, brain-damaged subjects), and tasks (lateral visual hemi eld presentation, copy drawing, memory recognition, and recall), and triggered debates about the conditions (number, size, sparsity, and ``goodness of the stimuli) in which the e ect is and is not observed. The major part of this literature has been reviewed by Kimchi (1992) and will not be summarized here, except for the two speci c elds that follow. NEUROPSYCHOLOGICAL DISSOCIATIONS Neuropsychological investigations using the procedure designed by Navon (1977) not only support the global prevalence phenomenon but, in addition, suggest that di erent mechanisms underlie the processing of the global level and the processing of the local level. First, divided visual eld behavioral studies in normal subjects seem to indicate that local parts of a compound stimulus are better processed by the left (right visual hemi eld) than by the right cerebral hemisphere, whereas the global form is better processed by the right hemisphere (e.g., Hubner, 1998 ; Martin, 1979 ; Sergent, 1982). These designs and results have been used to assess hypotheses about schizophrenia (Ferman, Primeau, Delis, & Jampala, 1999 ; Granholm, Perry, Filoteo, & Bra , 1999). However, the picture is not completely clear for normal subjects. For instance, Boles and Karner (1996) and Van Kleeck (1989) point to
288
the discrepancy of results between (and within) studies of hemi eld presentation in normal subjects and clinical studies in unilaterally braininjured subjects (see below). Moreover, investigations of normal subjects by means of contemporary brain imagery techniques begin to contribute to the understanding of mechanisms underlying global and local processing, respectively. Thus, Han, Fan, Chen, & Zhuo (1997) recorded the event-related brain potentials during this kind of task and supported the view that the global precedence phenomenon takes place at early perceptual processing stages (see also Ridderinkhof & Van der Molen, 1995). Let us also note the studies by Fink and coworkers (1996, 1997a, 1997b, 1999) assessing, in normal subjects, the di erential hemispheric contributions to local (left) and global (right) processing by means of the positron emission tomography procedure (see also Martinez, Moses, Frank, Buxton, Wong, & Stiles, 1997, for a study using the functional magnetic resonance procedure). We nally note that this global precedence could not be a universal property, as Fagot and Deruelle (1997) brought to light the global advantage, with interference of the global over the local level, in humans but a local advantage and no interference in baboons, by using the same procedure for both species. Second, the di erential e ects of some di use or subcortical brain pathologies on local and global processing of compound stimuli have been investigated in group studies as well as in single-case reports. Filoteo et al. (1992) submitted normals and Alzheimers disease patients to an experiment involving digits as stimuli under two attentional conditions: directed (to process either the global or the local level ; the procedure used by Navon, 1997) and divided attention tasks (to process both levels at each trial). Patients were slower than normals in both tasks, but particularly in the divided attention task, and this was mainly due to their inability to disengage and shift their attention (Posner, Walker, Friedrich & Rafal, 1984) from one level to the other. A similar pattern, but with qualitative di erences, was found in parkinsonian patients by Filoteo et al. (1994). Filoteo et al. (1995) compared cortical (Alzheimer patients) and subcortical dementias (huntingtonian and parkinsonian patients) and found an impaired performance in the three samples (as compared to control subjects) that was due to di erent patterns: the shifting of attention was normal in huntingtonian patients, whereas parkinsonian patients were mostly impaired in maintaining visual attention, and Alzheimer patients were mainly impaired in the disengagement of attention. Finally, Delis et al. (1992) suggested two distinct subtypes, ``high verbal and ``high spatial, of Alzheimers disease patients on the basis of independent tests and it appeared that ``high verbal patients were particularly impaired in processing the local elements whereas ``high spatial ones were particularly impaired in processing the global shapes.
289
With regard to single-case studies of Alzheimer patients, one should note a study by Coslett, Stark, Rajaram, and Sa ran (1995) that showed that two subjects processed faster the local rather than the global level (a pattern that documented the hypothesized restriction of their spotlight of visual attention). According to Coslett et al. (1995), the patient with semantic dementia described by Breedin, Sa ran, and Coslett (1994) was also faster in processing the local rather than the global level. Let us note, however, that these subjects had been enrolled in a distributed attention condition, that the exposure duration was unlimited (up to the response), and that the target was verbally de ned at each trial. Third, experimental studies of groups of patients with a unilateral focal lesion deserve mention. In general, a defect in processing local parts of hierarchical stimuli is observed in groups of left-injured patients, whereas a defect in processing global elements is evidenced when rightdamaged subjects are considered (Delis, Robertson, & Efron, 1986 ; Lamb, Robertson, & Knight, 1989, 1990 ; Robertson, Lamb, & Knight, 1988). Fourth and nally, single-case studies of patients with a unilateral focal lesion must be mentioned. Polster and Rapcsak (1994) described two subjects su ering of a left hemisphere lesion and evidencing a precedence of the local over the global level in processing compound stimuli. More importantly, two independent studies revealed a double dissociation between the two levels. On the one hand, Marshall and Halligan (1995) described a patient with (left) hemineglect who processed normally global forms but neglected local forms in the left side of pictures. On the other hand, Doricchi and Incoccia (1998) described another patient with (left) hemineglect who neglected global forms of the left side and evidenced a general local precedence (on both sides of the material). In conclusion, the contribution of neuropsychological studies is interesting and twofold. First, neuropsychology con rms that one level the global one receives priority over the other the local one and that this priority can be disrupted or reversed by brain damage. Second and more importantly, both levels can be completely dissociated. Indeed, each level can be disrupted by pathology while the other is spared, and, moreover, each level seems to depend on distinct cognitive mechanisms and/or cerebral structures. The study of normal aging could, in turn, contribute to document such a dissociation. EFFECTS OF AGING Given these lessons o ered by neuropsychological studies, a developmental approach could indeed be fruitful, by showing the evolution of the relationship between the global and local processes over the lifespan. Such a dynamic view will, hopefully, detect subtle mechanisms underlying these processes and a possible di erential fate of global and local
290
TABLE 1 Control Groups of Some Neuropsychological Studies
Studies Polster & Rapczak, 1994 Filoteo et al., 1995 Lamb et al., 1989 Lamb et al., 1990 Coslett et al., 1995 Polster & Rapczak, 1994 Filoteo et al., 1994 Filoteo et al., 1995 Delis et al., 1992 Polster & Rapczak, 1994 Filoteo et al., 1992
Sample (n) 5 15 10 12 12 5 14 15 13 5 15
Mean age College students 49.3 50 55 59.4 `Ìn their 60s 65.8 67.1 69.2 `Ìn their 70s 72
Advantage Global No Global Local Local No No No Local Local No
Precedence Global ? Global ? ? No ? ? ? Local No
Note : Papers are ranked according to the age of the samples.
processing with normal aging. In other words, age can be another procedure that can document dissociations between the studied processes. Now, rather strangely, very few developmental studies employed compound stimuli with local and global shapes equal in complexity and familiarity. Therefore, it is worth noting that the global precedence phenomenon, as evidenced by means of hierarchical stimuli, is robust but, up to now, limited to the population of young adult subjects. As a matter of fact, to our knowledge, the sole study examining the e ects of age on the processing of hierarchic stimuli was carried out by Akshoomo , Delis, and Haist (1993), who reported a detrimental age e ect on the processing of local shapes. However, (a) this study used pairs of stimuli ; (b) stimuli were formed with letters, nameable shapes and non-nameable shapes ; and (c) the cognitive performance was approached by means of recall and recognition tasks. Fortunately, other but indirect pieces of information can be found by examining the performance of control groups in some of the neuropsychological studies mentioned above. Table 1 summarizes the ndings. Obviously, the global advantage and the global precedence usually observed in young adults vanish when older samples are considered. The study of Polster and Rapcsak (1994) is illustrative, even if each sample was formed with very few subjects: there appeared to be a global precedence in young subjects, no precedence in young-old subjects, and a local precedence in old subjects. The present study was planned to clarify this picture.
291
THE PRESENT STUDY The purpose of the present study was to explore the fate of the global precedence with age, by using a procedure very similar to that designed by Navon (1977 ; Experiment 3) and by comparing the performance of young to that of elderly adult subjects. Given the lack of published studies assessing the e ect of normal aging on this kind of task, a pragmatic goal of Experiment 1 was to check whether the ``global precedence e ect would still be observed in elderly subjects. It is well known that aging e ects are slight for many early, low-level perceptual processes (see Fozard, 1990, for a review), but the current literature shows that late, high-level visual processes are generally spared (see Bruyer, 1994, for a review). Therefore, elderly subjects of the present study were selected on the basis of a normal or corrected visual acuity, a short medical questionnaire (Christensen, Moye, Armson, & Kern, 1992 ; translated into French), and normal performance in a verbal evaluation of the intellectual level (the Mill-Hill vocabulary test, multiple-choice form ; French translation: Deltour, 1993). According to indirect observations summarized in Table 1, it was expected that global precedence aging will decrease to disappear with aging, or that a local precedence will appear. EXPERIMENT 1 Methods Stimuli and Experimental Design The stimuli (upper-case printed letters, font Geneva, size 10 for local letters, plain) were drawn to be as similar as possible to those of Navon (1977). The material was displayed in black on a white background, on the screen of a microcomputer (Macintosh) whose luminosity and contrast were maximal. The size of the global letters was 25 3 25 mm, which corresponds to an angular size of 3.11 3 3.11 degree at the viewing distance of 50 cm. Global letters were composed of local letters. The shape of local letters was identical to that of the corresponding global letters and their size was 1/8 of that of the global letters. All local letters forming a given global letter were identical. The eight stimuli were: a global H composed of local Hs, Ss, or Os (in short: HH, HS, HO) ;a global S also made up of local Hs, Ss, or Os (SH, SS, SO), and a global O composed of local Hs or Ss (OH, OS). `Ò was used in the neutral conditions, and the task was to decide whether the target letter ``H or ``S was displayed. In the global task, the subject had to make a decision about the global shape. In the local task , the decision concerned the local shapes. Each
292
task included three conditions: the consistent condition involved stimuli in which letter identity was identical at both levels (HH and SS), the con ictual condition involved stimuli in which the identity of the global letter con icted with that of the local letters (HS and SH), and the neutral condition involved the letter `Ò at the to-be-ignored level (i.e., OH and OS in the local task and HO and SO in the global task). Each task comprised 144 trials, randomly displayed. There were 48 trials for each condition (24 occurrences of each of the two available stimuli ; e.g., 24 HH 24 SS 5 48 consistent trials). The stimulus appeared randomly in one of the four quadrants around the xation point (each localization was equally used) and was centered 2.6 cm (3.24 degree) away from the xation point (a small cross 4 3 4 mm). Each stimulus was followed by a mask, a matrix of 32 3 32 dots, displayed in the same quadrant. The 144 trials were grouped in four blocks of 36 trials. In each task, the analysis took into consideration only the last three blocks (108 trials, i.e., 36 per kind of trial) ; the rst block served as a practice run and was not analyzed. Procedure Each trial consisted of the following sequence of events : warning signal for 1 s, display of the xation cross for 500 ms, display of the stimulus during 50 ms, immediate display of the mask until the response was given. The intertrial interval was 3 s. The subject produced the response by depressing one of two buttons on the keyboard (the two central keys above the space bar) by using the left or right index nger. One button was labelled ``H and the other ``S. Subjects Thirty-two subjects with normal or corrected visual acuity were recruited. The sample of young subjects comprised 16 participants (students recruited for credit courses), 8 females and 8 males. Their mean age was 19.25 years (range: 18 21). Their mean score in the Mill-Hill vocabulary test was 24.19 out of 34 (range : 12 30). The sample of elderly subjects was formed with 16 participants, 8 females and 8 males. Their mean age was 68.69 years (range: 65 74). Their mean score in the MillHill vocabulary test was 31.12 out of 34 (range : 28 34), better than that of young subjects (t (30) 5 5.99, p , .0001). These elderly subjects were retired but still active in several sociocultural associations and were members of the Third Age University. Each subject lled in a questionnaire concerning her/his medical history (Christensen et al., 1992 ; translated into French), and no subject failed any of the criteria of this medical screening. Within each age group, four males and four females were submitted to the global task before the local task and the inverse sequence was used
293
for the remaining four females and four males. Within these subsamples of four subjects, two subjects responded ``H with the left key and ``S with the right key, and the inverse applied to the remaining two subjects. For a given subject, the same assignment of keys applied to both tasks. Results Table 2 shows the mean number of errors (out of 36: mean of the last three blocks) and the mean latency of correct responses for each sample, condition, and task. Accuracy The number of errors was analyzed by means of an Age 3 Task 3 Condition analysis of variance (ANOVA) with two within-subject and one between-subject factors. All e ects proved to be signi cant. Thus, signi cant main e ects of age (F (1, 30) 5 9.16, p , .005), of task (F (1, 30) 5 32.47, p , .0001), and of condition (F (2, 60) 5 48.00, p , .0001) emerged. These e ects were quali ed by signi cant Task 3 Age (F (1, 30) 5 7.52, p , .01), Condition 3 Age (F (2, 60) 5 17.68, p , .0001), and Task 3 Condition interactions (F (2, 60) 5 34.32, p , .0001). Finally, the Age 3 Task 3 Condition interaction was also signi cant (F (2, 60) 5 14.10, p , .0001). In order to explore these e ects, a Task3 Condition ANOVA was computed on each age group separately. For young subjects, a signi cant main e ect of condition emerged (F (2, 30) 5 6.76, p , .004 ;more errors appeared in the con ictual than in the remaining two conditions, but the latter two conditions did not di er from each other). The condition main e ect interacted with task, as indicated by the signi cant Task 3 Condition interaction (F (2, 30) 5 6.22,
TABLE 2 Results of Experiment 1: Mean Number of Errors (Out of 36) and
Mean Latency of the Correct Responses (ms) as a Function of the Age Group, the Task, and the Condition
Task Global Condition Young Errors (/36) Latency (ms) Elderly Errors (/36) Latency (ms) Consistent 0.67 578 0.56 599 Neutral 0.44 582 0.88 559 Con ictual 1.00 608 1.56 593 Consistent 1.63 679 2.06 834 Local Neutral 2.94 693 5.75 919 Con ictual 5.31 831 17.31 986
294
p,
.005). The post hoc analysis of this interaction (p # .05) showed that the condition e ect was due only to the local task where the number of errors was greater in the con ictual condition than in the remaining two (which did not di er from each other) ; in addition, there was an advantage of the global over the local task in the con ictual and neutral conditions. For elderly subjects, the ANOVA disclosed a signi cant task main e ect (F (1, 15) 5 49.41, p , .0001) and a signi cant condition main e ect (F (2, 30) 5 42.69, p , .0001), favoring the consistent over the neutral and the con ictual conditions (which did not di er from each other). However, both main e ects were quali ed by the signi cant Task3 Condition interaction (F (2, 30) 5 28.43, p , .0001). The post hoc analysis of this interaction showed that the task e ect was due to the neutral and the con ictual conditions (no task e ect for the consistent condition), and that the condition e ect was due to the local task only, where advantages of the consistent over the neutral, and of the neutral over the con ictual condition, appeared. Finally, given the signi cant triple order interaction observed in the main analysis, age e ects were studied by means of separate t tests computed for each condition of each task. Only one of the six tests proved to be signi cant. Namely, in the con ictual condition of the local task, the number of errors was signi cantly greater in elderly than in young subjects (t (30) 5 4.13, p , .0003). Latency of Correct Responses The Age 3 Task 3 Condition ANOVA evidenced signi cant main e ects of task (F (1, 30) 5 18.62, p , .0002) and of condition (F (2, 60) 5 9.67, p , .0002). However, both main e ects were quali ed by the signi cant Task 3 Condition interaction (F (2, 60) 5 8.33, p , .0006). No other e ect was signi cant. In particular, age had no e ect on correct latency, either as a main e ect or as an interaction. The post hoc analysis of the interaction showed no signi cant e ect of the condition in the global task. In the local task, the con ictual condition di ered signi cantly from the central (p , .0001) and the consistent conditions (p , .0001) ; in addition, the neutral condition di ered from the consistent one in the local task (p , .046). The task e ect was signi cant for each condition (all p values , .0001). Discussion The ``global precedence phenomenon (Navon, 1977) was observed in young and elderly subjects for both dependent variables. In addition, the sole e ect of age concerned the con ictual condition of the local task on accuracy. Finally, a facilitation e ect of the consistent condition
295
appeared in the local task, and accuracy in elderly and on correct latency in both samples. Thus, it appeared that aging has little e ect on the global precedence e ect. However, the picture is a bit more complicated. Indeed, accuracy of elderly in the con ictual condition of the local task was very close to the random level (48.1% errors), which precludes the interpretation of correct latencies. As a matter of fact, this was due to a bimodal distribution of the sample. Indeed, by taking 45 and 63 errors out of 108, i.e., 15 and 21 errors (these values included) out of 36 per block, as limits of the random performance (as attested by chi-square tests) in the con ictual condition of the local task, it appeared that two elderly subjects fell within the random range (mean 5 15) ;eight other subjects produced less than 15 errors in this condition (mean 5 10.3), whereas the remaining six produced more than 21 errors (mean 5 27.3). In addition, both subgroups produced few errors in the remaining two conditions ( rst subgroup: 4.25 for the neutral and 3.5 for the consistent condition ; second subgroup: 6.4 for the neutral 0.83 for the consistent condition). It thus appeared that eight subjects clearly showed an interference e ect but did not base their response on the global letter (which is further supported by their performance in the neutral condition), and that six other subjects revealed an obvious interference e ect and did, in fact, base their response on the global letter (which was not possible in the neutral condition). Let us note that the two subgroups did not di er in age and in their Mill-Hill score, as con rmed by Student t tests. When young subjects were considered, we detected one subject performing at random (18 errors) and another subject performing worse than randomly in the con ictual condition of the local task (28 errors). It thus seems that subpopulations underly the sample of elderly subjects with respect to this kind of task. At the individual level, the random performance in the con ictual condition of the local task was due to only 2 subjects, whereas the remaining 14 formed two distinct subsamples. First, like young subjects, the elderly were obviously sensitive to the interference from the global shape during the processing of the local one. Second, this interference arose from the identity of the global letter, not from the presence of a global shape only, as seen by the performance of the elderly in the neutral condition. Third, it seems, however, that there were several kinds of sensitivity to interference underlying the responses of the elderly. Indeed, 50% of the elderly performed like young subjects and displayed the usual pattern of interference, that is to say, a slowing down of correct responses and a ``tolerable (i.e., not random) number of errors. Another two subjects were totally unable to discriminate the two target letters in the local task, provided the global shape was a possible target too (indeed, their performance was not random in the neutral condition) ;in addition, they performed adequately in the consistent condition. Finally, about a third of the elderly subjects were so sensitive to
296
interference that they responded to the global letter, even in the local task ; consequently, a good performance appeared in the consistent condition and a biased performance in the con ictual condition. And yet, they were able to process the local shapes, even embedded in a global one, provided this global form was not a candidate letter (i.e., in the neutral condition). These subsamples of elderly did not di er from each other in age and in Mill-Hill score but, even if we have no other explanation to o er for this heterogeneity, it is in good agreement with the often reported increase in between-subject variability with age (see Rabbitt, 1993, for a review) as well as the trend toward a reduction of the global precedence phenomenon with age (Table 1). However, caution is required because psychometric properties of the distribution are poorly known in this study. Experiment 2 was planned to clarify the matter by enrolling subjects performing better than at random in the con ictual condition of the local task. Moreover, before a clear statement about precedence e ect is possible, it is important to determine whether subjects are able to process both the global and local levels separately, i.e., that information is discriminable at the two levels when attention is not directed towards a particular level. For instance, the absence of interference of the local level in the processing of the global level (con ictual condition), as well as the absence of facilitation in the consistent condition (as compared with the neutral condition) of the global task could simply indicate that when attention is directed to the global level, the local elements are not detected. Experiment 4 of the seminal paper of Navon (1977) had also been designed to control this point. In short, one has to be certain that the prevalence of the global over the local level is a true perceptual phenomenon, and not related to decisional processes. Experiment 1a was designed to test such a possibility and, hopefully, support and validate the pattern of precedence observed in Experiment 1. The stimuli of Experiment 1 were again used, in a same/di erent discrimination task with two stimuli shown simultaneously at each trial, in two quadrants of the display. The two stimuli were identical in half the trials, di erent at the global level in 25% of the trials, and di erent at the local level in the remaining 25%. We acknowledge that the contribution of Experiment 1a should be considered with caution because, for obvious practical reasons, it was submitted to di erent subjects than those of Experiment 1. Experiment 1a The material was exactly the same as that of Experiment 1. Each subject was submitted to 240 trials randomly distributed in a single task. In each trial, two stimuli were simultaneously displayed, in two randomly selected (and equally used) quadrants and followed by a pair of masks.
297
A speeded same/di erent decision was required, by using the same keys as in Experiment 1. In 120 trials, the two stimuli were identical (15 times each pair of identical shapes). In 60 trials, the two stimuli di ered only at the local level (pairs HH/HO, HH/HS, HO/HS, SH/SO, SH/SS, and SO/SS, 10 times each). In 60 trials, the two stimuli di ered only at the global level (pairs HH/OH, HH/SH, OH/SH, HS/OS, HS/SS, and OS/SS, 10 times each). These 240 trials resulted from ve continuous repetitions of a block of 48 trials. The analysis will only consider the last four blocks (192 trials, i.e., 96 same, 48 di erent at the global level, and 48 di erent at the local level). The rst block served as a practice run. Each trial comprised the same events as in Experiment 1, with two simultaneous stimuli instead of one, a double mask, and a same/di erent kind of response. One button was labelled ``same and the other ``di erent. Instructions did not direct attention to a particular level. Thirty-two new subjects, selected in the same manner as those in Experiment 1, were enrolled. The mean age of the young subjects was 19.4 years (range: 18 24). Their mean score in the Mill-Hill vocabulary test was 26.44 out of 34 (range: 20 32). The mean age of elderly subjects was 68.9 years (range: 65 76). Their mean score in the Mill-Hill vocabulary test was 31.06 out of 34 (range: 27 33), again better than that of young subjects (t (30) 5 5.13, p , .0001). Within each age group, four males and four females responded ``same with the left key and ``di erent with the right key, and the inverse assignment applied to the remaining subjects. Given the purpose of this experiment, analyses were limited to di erent trials only. The proportion of errors was examined by an Age 3 Level (global vs. local) ANOVA. A signi cant main e ect of age (F (1, 30) 5 8.06, p , .008) was observed that favored young over elderly subjects (19.1 vs. 31.3%), as well as a signi cant e ect of the level (F (1, 30) 5 91.01, p , .0001), indicating that di erences were better detected at the global than at the local level (3.4 vs. 46.9%). The interaction was not statistically contributive. For global di erences, the number of errors differed from the random level in young (t (15) 5 106.79, p , .0001) as well as in elderly subjects (t (15) 5 32.75, p , .0001). For local di erences, young subjects performed better than at random (t (15) 5 2.16, p , .048), but elderly subjects did not (t (15) 5 1.15, NS). Obviously, the task was difficult, especially in the locally di erent condition, and even for young subjects. Actually, when thresholds of 18 and 30 errors out of 48 were de ned as the limits of the random area (these values included, as attested by chi-square tests), only 9 out of the 16 young subjects performed better than at random (mean 5 9.11 errors) ;5 subjects performed randomly (mean 5 24 errors), and another 2 performed worse than the random level (mean 5 40 errors). Note that these three subsamples did not di er in age and in their score in the Mill-Hill test, as attested by nonparametric Kruskal-Wallis tests. The pattern was
298
still more impressive for elderly subjects. As a matter of fact, their quasirandom performance in the ``locally di erent condition resulted from a subsample of subjects. Indeed, four subjects failed in the random area (mean number 5 20.5) and nine subjects performed ``worse than the random level (mean 5 36.3), so that only three subjects produced fewer errors than predicted by random decisions (mean 5 10). These three subsamples did not di er in age and in their Mill-Hill score, as attested by Kruskal-Wallis tests. For correct latencies, the Age 3 Level ANOVA revealed only a signi cant main e ect of the level (F (1, 30) 5 62.14, p , .0001), favoring the global over the local level (747 vs. 1114 ms). Clearly, the global precedence e ect observed in the young subjects of Experiment 1 cannot be explained by an inability to process the local level when the attention is not directed towards a particular level. However, we have to note that only 9 young subjects out of 16 were able to detect local di erences better than at random, and that di erent subjects were enrolled in Experiments 1 and 1a. The picture is a bit more complex for elderly subjects. Indeed, only 3 elderly subjects out of 16 were able to detect local di erences better than at random. Experiment 2a was planned to enroll subjects able to process the locally di erent pairs better than at random. Conclusion As a group, young adult subjects performed very much like those of Navon (1977), evidencing the global prevalence phenomenon in Experiment 1, i.e., there was an advantage of the global over the local task, and of the neutral over the con ictual condition in the local task only. In addition, this phenomenon was not accounted for by a decision bias, as young subjects were able to process both the global and the local level when attention was not directed toward a particular level (Experiment 1a). However, it remains that di erent subjects were enrolled for the two experiments, and that a within-subject design should clarify this point. For elderly subjects, the pattern was more complex, as 50% of them performed better than at random in the con ictual condition of the local task (Experiment 1) and three subjects only in the locally di erent condition of the matching task (Experiment 1a). One lesson of Experiments 1 and 1a is that subpopulations in the elderly have to be considered, even though they do not di er from each other in age or in verbal pro ciency, when high-level perceptual processes are investigated. This point will be considered in Experiments 2 and 2a, where we attempted to constitute samples of 16 young and 16 elderly subjects performing better than at random in the local condition. It is worth noting that this procedure raises the important issue of generalizability of the ndings with respect to understanding the aging
299
process. However, this issue should not be overestimated. Indeed, after all, this kind of procedure is systematically used for sake of methodology. For instance, in the literature, groups of elderly subjects are generally selected as being equivalent to the control group of young adults with respect to, say, sociocultural level, crystallized intelligence, or visual efficiency. In the local task (Experiment 1) or condition (Experiment 1a), accuracy was lower in elderly than in young subjects, so that a di erence in the level of difficulty for processing the signal could have introduced biases in the study of latency. This point will be considered in Experiment 3, where accuracy of the elderly was improved by allowing them to process the signal more efficiently (by manipulating the stimulus duration and the stimulus-mask interval). It was useful to carry out Experiments 1 and 1a, because the e ect of aging on the global prevalence phenomenon, as evidenced by the Navon paradigm, has never been reported in the literature. EXPERIMENT 2 Methods Experiment 2 was identical to Experiment 1, but we enrolled subjects in such a way that 16 young and 16 elderly subjects were able to perform the con ictual condition of the local task better than at random. As a matter of fact, the data of the 14 young and 8 elderly ``normal subjects of Experiment 1 were again used, and we attempted to nd 2 additional young ``normal subjects and 8 additional elderly ``normal subjects and submitted them to the corresponding hand and button conditions of the discarded subjects of Experiment 1. To do this, we were obliged to enrol only 2 young subjects, but 13 elderly subjects. The nal set of young subjects was 19.44 years old (range : 18 23 ; mean Mill-Hill score 5 24.12 ; range : 12 30) and the nal set of elderly subjects was 69.37 years old (range: 6.5 78 ; mean Mill-Hill score 5 31.12 ; range : 28 33). The elderly performed signi cantly better than the young subjects in the Mill-Hill test (t (30) 5 6.125, p , .0001). Results Table 3 shows the mean number of errors (out of 36: means of the last three blocks) and the mean latency of correct responses for each sample, condition, and task. Accuracy The Age 3 Task 3 Condition ANOVA evidenced signi cant main e ects of task (F (1, 30) 5 33.39, p , .003), condition (F (2, 60) 5 34.82,
300
TABLE 3 Results of Experiment 2: Mean Number of Errors (Out of 36) and
Mean Latency of the Correct Responses (ms), as a Function of the Age Group, the Task, and the Condition
p,
.0001), and age (F (1, 30) 5 10.92, p , .003), as well as signi cant Task 3 Condition (F (2, 60) 5 27.79, p , .0001), Age 3 Task (F (1, 30) 5 17.29, p , .0002), and Age 3 Condition interactions (F (2, 60) 5 6.93, p , .002). In addition, the Age 3 Task 3 Condition interaction was signi cant (F (2, 60) 5 11.87, p , .0001). Post hoc comparisons showed a signi cant advantage of young over elderly subjects in each condition of the local task (consistent: t (30) 5 2.19, p , .04 ; neutral: t (30) 5 2.52, p , .02 ; con ictual : t(30) 5 4.75, p , .0001), with no e ect of age in the three conditions of the global task. The triple interaction was decomposed into two Task 3 Condition ANOVAs separately computed on each age group. For young subjects, there was a trend towards an e ect of the task (F (1, 15) 5 3.46, p , .083). The main e ect of condition was signi cant (F (2, 30) 5 5.14, p , .015 ; more errors in the con ictual than in the remaining two conditions, not di ering from each other), and the Task 3 Condition interaction just failed to reach the critical threshold (F (2, 30) 5 3.14, p , .058). The post hoc analysis of this interaction (p # .05) showed that the number of errors was signi cantly higher in the con ictual condition of the local task than in the remaining ve Task 3 Condition combinations (not di ering from each other). So, the task e ect was only due to the con ictual condition, and the condition e ect appeared only in the local task. For elderly, there was a signi cant e ect of the task (F (1, 15) 5 30.46, p , .0001). The main e ect of condition was signi cant (F (2, 30) 5 37.99, p , .0001 ; more errors occurred in the con ictual than in the remaining two conditions, which did not di er from each other), as was the Task 3 Condition interaction (F (2, 30) 5 26.05, p , .0001). The post hoc analysis of this interaction (p # .05) showed that the number of errors was signi cantly higher in the con ictual condition of the local task than
301
in the remaining ve Task 3 Condition combinations, and higher in the neutral condition of the local task than in the neutral and consistent conditions of the global task. So, the task e ect was due to the con ictual and neutral conditions, and the condition e ect appeared in the local task only. To explore more speci cally a possible e ect of age on the global precedence, a last post hoc ANOVA was computed on the neutral and con ictual conditions of the local task, with age and condition as factors. Signi cant main e ects of age (F (1, 30) 5 20.43, p , .0001) and condition (F (1, 30) 5 44.65, p , .0001) were observed, as well as a signi cant Age 3 Condition interaction (F (1, 30) 5 13.00, p , .0015), showing that the interference was higher in elderly than in young subjects (additional 6.06 vs. 1.81 errors). Correct Latency The Age 3 Task 3 Condition ANOVA evidenced main e ects of task (F (1, 30) 5 10.90, p , .0025) and condition (F (2, 60) 5 7.94, p , .001), as well as a signi cant Task 3 Condition interaction (F (2, 60) 5 8.13, p , .0009). However, no trace of e ect of age was detected, either as a main e ect or in interactions. The Task 3 Condition signi cant interaction was analyzed further. It appeared a signi cant e ect of the task for each condition (all p values , .0001), no e ect of the condition in the global task, and a signi cant di erence between the con ictual and both other conditions (not di ering from each other) in the local task (both p values , .0001). The post hoc Age 3 Condition ANOVA on the neutral and con icting conditions of the local task veri ed the signi cant main condition e ect (F (1, 30) 5 9.07, p , .006), but failed to disclose a signi cant main age e ect and a signi cant Age 3 Condition interaction (the cost induced by the con ictual condition was 131 ms in young subjects and 84 ms in elderly subjects, NS). Discussion Young subjects displayed the classic ``global precedence phenomenon, a pattern observed with both dependent variables. The elderly, also, displayed this pattern in both measures. At rst glance, when ``good performers are selected, no sign of an age e ect was detectable in the results. However, we noted a detrimental e ect of aging in the local task, in particular an increased sensitivity to interference (con ictual condition), and this e ect was limited to accuracy. So, expectations were partially supported. However, because the e ect of aging appeared in the three conditions of the local task (the increased interference emerged
302
only from a very speci c post hoc ANOVA), it could be that age e ect is the result of a general increased difficulty of the local task. Experiment 3 was designed to solve this uncertainty. However, before starting Experiment 3, we wanted to clarify the picture observed in Experiment 1a, by enrolling young and elderly subjects able to process the locally di erent pairs better than at random, in a task where attention is not directed to a particular level. Experiment 2a Experiment 2a was similar to Experiment 1a, but we enrolled subjects in such a way that 16 young and 16 elderly subjects were able to process the locally di erent pairs better than at random. In addition, we recall that Experiment 1a was very difficult, because only three elderly and nine young subjects did reach this criterion of accuracy. This resulted from the perceptual conditions: subjects had to process two simultaneous stimuli at the two levels (global and local) under the same (and short: 50 ms) exposure duration as that of the single stimulus of Experiment 1 of which only one level had to be processed. In addition, preliminary attempts to select ``good performers failed (none of the ve young and ve elderly subjects enrolled were able to satisfy the criterion). Therefore, a methodological adaptation was provided, and submitted to new young and elderly subjects. Thus, Experiment 2a was a close adaptation of Experiment 1a, with 16 young and 16 elderly ``normal subjects in the locally di erent condition. To do this, and in spite of the methodological modi cation, we enrolled 19 young and 33 elderly subjects. The nal set of 16 young subjects was 19.9 years old (mean Mill-Hill score 5 26.69) and the nal set of 16 elderly subjects was 70.6 years old (mean Mill-Hill score 5 29.90 ; t(30) 5 2.87, p , .008). The methodological adaptation was as follows. The two stimuli of each trial were displayed sequentially instead of simultaneously, with the rst stimulus shown for 1 s and immediately followed by the second stimulus for 50 ms. Thus, the decision had to be taken on a single stimulus seen in the same conditions as in Experiments 1 and 2. Response latency was measured from the onset of the second stimulus. Once again, given the purpose of this experiment, analyses were limited to di erent trials only. The proportion of errors was examined by an Age 3 Level ANOVA. The main age e ect was signi cant (F (1, 30) 5 19.53, p , .0001) as well as the main level e ect (F (1, 30) 5 20.27, p , .0001), but both e ects were quali ed by the signi cant interaction (F (1, 30) 5 5.00, p , .03) which was analyzed in more depth. The e ect of the level was signi cant in each age group (young: 1.8 vs. 7.2% errors ; F (1, 15) 5 10.09, p , .007 ; elderly: 5.3 vs. 21.2%: F(1, 15) 5 13.01, p , .003). Moreover, there appeared a signi cant advantage of young over elderly subjects for
303
locally di erent trials (F (1, 30) 5 19.96, p , .0001), but no e ect of age for globally di erent trials (F (1, 30) 5 1.36). For correct latencies , the Age 3 Level ANOVA revealed a signi cant main age e ect (F (1, 30) 5 5.54, p , .026) as well as a signi cant main e ect of the level (F (1, 30) 5 15.28 ; p , .0005). In addition, both e ects were quali ed by the signi cant interaction (F (1, 30) 5 6.45, p , .02). There appeared an advantage of young over elderly subjects for locally di erent trials (1014 vs. 1262 ms ; F (1, 30) 5 9.98, p , .004) but not for globally di erent trials (978 vs. 1089 ms ; F (1, 30) 5 1.79). For young subjects, there was no signi cant e ect of the level (F (1, 15) 5 2.35) ; for elderly subjects, correct responses were faster for globally di erent than for locally di erent trials (F (1, 15) 5 12.98, p , .003). So, in Experiment 2a, the accuracy and speed of correct responses were better in young than in elderly subjects. Moreover, this was mainly observed when stimuli di ered at the local level. However, the main lesson of this experiment is that elderly (and, of course, young) subjects were able to process the local level even when their attention was distributed over both the global and the local levels. Conclusion When young and elderly subjects were carefully selected on the basis of performing better than at random, we con rmed, in both age groups, the presence of the classical e ects rst reported by Navon (1977). These include a global precedence phenomenon (Experiment 2), which could not be attributed to a failure to process the local level when attention is shared on both levels (Experiment 2a). We also con rmed the wellknown e ect of aging on cognitive performance, especially when speed of response is considered, i.e., aging slows down cognitive operations. It also appeared that elderly people seemed to be exaggeratedly sensitive to interference. Indeed, in the local task, the con icting condition was particularly difficult for them and there were signs of facilitation in the consistent condition. In addition, the elderly also di ered from young subjects in the neutral conditions, which might suggest some difficulties of encoding with aging resulting, perhaps, from the (short) exposure duration. In order to explore in more depth this hypothesis, Experiment 3 was therefore planned. First, subjects were submitted to an independent interference test (the Stroop task, see below) to examine whether the increased sensitivity to interference in old age was speci c to the ``Navon task or the manifestation of a more general phenomenon. Second, exposure duration of the stimuli was longer in elderly than in young subjects in an attempt to suppress age e ects in the neutral condition in such a way that encoding problems could be ruled out in explaining age e ects on the global precedence phenomenon.
304 EXPERIMENT 3
As in Experiment 2, the subjects of Experiment 3 were selected only if they were able to perform the con ictual condition of the local task better than at random, namely, by producing less than 15 errors out of 36 trials. Experiment 3 was identical to Experiment 2, with two modi cations. The rst modi cation concerned the selection of elderly subjects, as they were retained only when their accuracy in the neutral condition of the local task (used as baseline) did not di er statistically from that of young subjects. The criterion was the mean 2.5 standard deviations of the accuracy of young subjects, i.e., as shown below, no more than 3 errors out of 36. Indeed, we wanted to compare young to elderly subjects after having equated the performance of both samples for this baseline condition, i.e., by neutralizing possible age e ects in the encoding process (because age can a ect basic sensory visual processes ; see Fozard, 1990, for a review). This was made by displaying the stimuli for 200 ms in the elderly (against 50 ms in the young), and by providing a blank interval of 50 ms between the o set of the stimulus and the onset of the mask in the elderly (against 0 ms in the young). The second modi cation was the addition of an independent measure. Indeed, the elderly subjects of Experiments 1 and 2 seemed to have difficulties in inhibiting the `àutomatic processing of the global shape when they had to process the local shape. Now, an aspeci c de cit of inhibition is sometimes suggested as the basic mechanism of cognitive aging (e.g., Hasher, Stoltzufs, Zacks, & Rypma, 1991). To check whether or not this phenomenon was speci c to the ``Navon task, a Stroop test was added as a measure of inhibition. In the Stroop test (1935), the task requires subjects to name the color in which a word is written and to ignore the name (in fact, a color name). In such tasks, it is observed that reading is unavoidable ; in particular, reading times are slower when the word names a di erent color than the target, than when the word names a noncolor or the same color as the target. This ``Stroop e ect has been abundantly studied and used in the literature (see MacLeod, 1991 for review). Method We replicated Experiment 2, with (a) 16 young and 16 elderly new subjects, selected as above, whose performance was ``normal in the con ictual condition of the local task ; and (b) where the performance of elderly subjects did not di er from that of young subjects in the neutral condition of the local task (no more than the mean 2.5 standard deviation errors, out of 36, of young subjects). To do this, we enrolled 22 young and 19 elderly subjects. The nal set of 16 young subjects was 19.7
305
years old (mean Mill-Hill score 5 24.44), and the nal set of 16 elderly subjects was 72.6 years old (mean Mill-Hill score 5 31.37 ; t (15) 5 5.81, p , .0001). Six of the young subjects were replaced due to technical bugs (problems with data saving, ts of coughing, etc.) but all the remaining 16 subjects performed better than at random (less than 15 errors out of 36) in the con ictual condition of the local task. As will be seen below, their mean performance in the neutral condition of the local task was 1.25 errors out of 36, SD 5 1.065, so that the criterion to retain the data of elderly was 3 errors or less in that condition. Three of the elderly subjects were discarded due to misunderstanding instructions (one male subject), too many errors in the neutral condition of the local task (one male subject), or too many errors in the con ictual condition of the local task (one female subject). Subjects were also submitted to the Stroop test. In this test, the participant had to name as fast as possible the color of the stimuli. Each trial consisted of four rows of six stimuli each, and the dependent variable was the mean time to name all colors of one trial (i.e., of 24 stimuli). In the control condition, stimuli were strings of ``X (three trials). In the interference condition, stimuli were color names not congruent with the color of the word (three trials). In each condition, the recorded measure was the mean time to name the color of the 24 stimuli. In each subsample of both age groups, half the subjects were submitted to the Mill-Hill test before the experiment and to the Stroop test after the experiment, and the reverse applied to the remaining half. Results Table 4 shows the mean number of errors (out of 36: means of the last three blocks) and the mean latency of correct responses for each sample, condition, and task. Accuracy The Age 3 Task 3 Condition ANOVA computed on the number of errors revealed signi cant main e ects of the task (F (1, 30) 5 21.00, p , .0001) and of the condition (F (2, 60) 5 25.75, p , .0001), as well as a signi cant Task 3 Condition interaction (F (2, 60) 5 11.76, p , .0001). Post hoc analysis of this interaction showed a signi cant e ect of the task in the neutral (p , .04) and con ictual conditions (p , .0001), but not in the consistent condition. In addition, there was a signi cant e ect of the condition in the local task, with signi cant di erences between the con ictual and the neutral condition (p , .0001) and between the con ictual and the consistent condition (p , .0001), but not between the neutral and the consistent condition. However, not the slightest sign of an age e ect emerged, either as a main e ect or in interactions (all F values , 1). Moreover, the post hoc
306
TABLE 4 Results of Experiment 3: Mean Number of Errors (out of 36) and
Mean Latency of the Correct Responses (ms) as a Function of the Age Group, the Task, and the Condition
ANOVA computed on the neutral and con ictual conditions of the local task, with age and condition as factors, showed a signi cant main e ect of condition (F (1, 30) 5 19.76, p , .0001). However, age was not signi cant (F , 1) nor was the Age 3 Condition interaction (F , 1). By way of control, separate Task 3 Condition ANOVAs were computed for each age group, and the rst analysis was largely con rmed, with slight quali cations. Thus, for young subjects, there were signi cant e ects of the task (F (1, 15) 5 17.31, p , .0008), of the condition (F (2, 30) 5 13.06, p , .0001), and of the interaction (F (2, 30) 5 6.28, p , .006). The e ect of the task was signi cant for the con ictual condition only (p , .0001) ;there was no e ect of the condition in the global task. In the local task, there was a signi cant e ect of the condition, with signi cant di erences between the con ictual and the neutral condition (p , .0001) and between the con ictual and the consistent condition (p , .0001), but not between the neutral and the consistent condition. For elderly, there were signi cant e ects of the task (F (1, 15) 5 7.38, p , .016), of the condition (F (2, 30) 5 12.79, p , .0001), and of the interaction (F (2, 30) 5 5.50, p , .01). Post hoc comparisons showed (a) no e ect of the condition in the global task ; (b) in the local task, a higher number of errors in the con ictual condition than in the remaining two conditions (p , .0001) not di ering from each other ; and (c) an e ect of the task in the con ictual condition only (p , .0001). Latency of Correct Responses The Age 3 Task 3 Condition ANOVA evidenced main e ects of task (F (1, 30) 5 16.62, p , .0003) and condition (F (2, 60) 5 23.14, p , .0001), as well as a quasi-signi cant Task 3 Condition interaction (F (2, 60) 5 3.07, p , .054). There was a signi cant e ect of task in each condi-
307
tion (consistent: p , .0001 ; neutral: p , .0003 ;con ictual : p , .0001). In the global task, the con ictual condition di ered signi cantly from the consistent (p , .002) and the neutral conditions (p , .02), not di ering from each other. In the local task, the con ictual condition di ered signi cantly from the consistent (p , .0001) and the neutral conditions (p , .0001), not di ering from each other. However, no trace of an age e ect was detected, either as a main e ect (F , 1) or as an interaction (Age 3 Task: F (1, 30) 5 1.96, NS ; Age 3 Condition: F , 1 ; Age 3 Task 3 Condition: F , 1). The post hoc Age 3 Condition ANOVA on the neutral and con icting conditions of the local task con rmed the signi cant main condition e ect (F (1, 30) 5 28.81, p , .0001), but failed to show a signi cant main age e ect (F , 1) and a signi cant Age 3 Condition interaction (F , 1). By way of control, separate Task 3 Condition ANOVAs were computed for each age group, and the rst analysis was largely con rmed, but with slight quali cations. Thus, for young subjects, there were signi cant e ects of the task (F (1, 15) 5 12.83, p , .003) and of the condition (F (2, 30) 5 13.21, p , .0001). The interaction was not signi cant (F , 1). However, as in Experiment 2, the cost induced by the con ictual condition showed that the interference of the global over the local level was greater than the interference of the local over the global level (96 vs. 53 ms). In addition, the pattern of errors convinced us that the global precedence phenomenon was at work. For elderly, a trend towards a main task e ect (F (1, 15) 5 4.31, p , .056) appeared, as well as a signi cant main e ect of the condition (F (2, 30) 5 10.08, p , .0004). The Task 3 Condition interaction was not signi cant (F (2, 30) 5 2.40). However, the cost induced by the con ictual condition showed that the interference of the global over the local level was greater than the interference of the local over the global level (103 vs. 31 ms). In addition, the pattern of errors convinced us that the global precedence phenomenon was at work. Stroop Interference Stroop interference was observed in young subjects, with mean color naming times of 15.33 s in the control condition and 20.46 s in the interference condition (t (15) 5 9.29, p , .0001). No signi cant correlation emerged between the Stroop interference ([interference control]/ [interference control]) and the interference of the global level over the processing of the local level (r 5 .212) or the interference of the local level over the processing of the global level (r 5 .286) when correct latencies were considered. For errors, the correlation between the Stroop interference and the interference of the local level over the processing of the global level was marginally signi cant (r 5 .465, df 5 14, p , .05), but the correlation between the Stroop interference and the interference of the global level over the processing of the local level was not (r 5 .035).
308
Stroop interference was also observed in elderly, with mean color naming times of 17.96 s in the control condition and 26.79 s in the interference condition (t(15) 5 12.09, p , .0001). No signi cant correlation emerged between the Stroop interference ([interference control]/ [interference control]) and the interference of the global level over the processing of the local level (r 5 .211) when correct latencies were considered ;a signi cant, but slight, correlation appeared between the Stroop interference and the interference of the local level over the processing of the global level (r 5 .463, df 5 14, p , .05). For errors, the correlation between the Stroop interference and the interference of the local level over the processing of the global level was not signi cant (r 5 .251), and the same applied to the correlation between the Stroop interference and the interference of the global level over the processing of the local level (r 5 .048). The index of the Stroop interference was signi cantly higher in elderly (19.97%) than in young subjects (13.34%: t(30) 5 3.004, p , .0055). Accordingly, the Age 3 Condition ANOVA computed on the scores in both conditions of the Stroop test evidenced a signi cant interaction (F (1, 30) 5 16.39, p , .0003) : the e ect of conditions has already been reported in both age groups, and the e ect of age was marginal in the control condition (t(30) 5 1.95, p , .06), but robustly signi cant in the interference condition (t (30) 5 3.55, p , .0015). Obviously, the main e ects of age (F (1, 30) 5 8.72, p , .006) and of condition (F (1, 30) 5 232.51, p , .0001) were statistically signi cant. Discussion In Experiment 3, perceptual encoding of the material was made as easy for the elderly as for the young subjects, by increasing exposure duration and de ning a criterion of accuracy in the neutral condition of the local task. Under these conditions, the results were clear. First, for young subjects, the ``global precedence phenomenon was veri ed, no speed-accuracy trade-o was at work, and the precedence e ect was unrelated to the interference e ect as measured by means of the Stroop test. In addition, the (unexpected) interference of the local level on the processing of the global level when correct latencies were considered was explainable in part by a general sensitivity to interference, as testi ed by the correlation with the Stroop interference. Second, the global precedence phenomenon was observed in the elderly and no speed-accuracy trade-o appeared. Third, an important Stroop e ect was observed in the elderly, but this index of interference was unrelated to the global precedence e ect. Fourth, all signs of an e ect of age as noted in the main experiment disappeared, even if the Stroop e ect was greater in elderly than in young subjects.
309
Therefore, it appears likely that the e ects of age on the global precedence observed in Experiments 1a and 2a were entirely due to difficulties of encoding the material, which probably results from slight sensory decline due to ageing (Fozard, 1990) ;in particular, it is worth noting that the increased sensitivity to interference noted in Experiments 1 and 2 was no longer observed when encoding was made easier (Experiment 3). Moreover, elderly subjects displayed an increased sensitivity to interference in the Stroop task, but their sensitivity to the global shape in the local task of the experiment was not the result of a general and nonspeci c de cit of inhibition (Hasher et al., 1991), as attested to by the correlational analyses. In addition, as shown by a recent meta-analysis (Verhaegen & De Meersman, 1998), the e ect of age on the Stroop e ect could be an artifact resulting from the general slowing due to age ; this encourages the design of additional experiments, even if this point is marginal with respect to the purpose the present study. GENERAL DISCUSSION Since the seminal publication of Navon (1977), the so called ``global precedence e ect in the visual processing of compound stimuli under directed attention (i.e., global vs. local task) is characterized by: (a) an interference of the global level over the processing of the local level (con ictual condition as compared to neutral condition), together with, (b) no reciprocal and symmetric interference of the local level over the processing of the global level. Two additional, not mandatory features can be: (a) a processing advantage of the global over the local task, and (b) a facilitation e ect in the consistent condition of the local task (as compared to the neutral condition). This global precedence e ect is generally observed when analyzing the speed of correct responses, and it is robust, even if it can be modulated by many variables (see Kimchi, 1992, for a review). The present study was designed to ll a gap in the current literature, namely, an investigation of the fate of the global precedence e ect with aging. A series of three main experiments and two control experiments is reported, in which healthy elderly subjects were compared to young adults. Two dependent variables were studied, accuracy and correct latencies. First, we summarize the main ndings. Experiment 1 was a close replication of the original experiment (Navon, 1977). In young subjects, the expected e ects were exactly replicated for both accuracy and latencies. In the elderly, the e ect was observed in accuracy only and took the form of an `èxacerbation, as testi ed by the emergence of a facilitation e ect in the consistent condition of the local task, and by the observation that about one third of the subjects, in the con ictual condition, actually processed the global shape in the local task. This supports the trend
310
observed from indirect data (see Table 1), that suggests global precedence is reduced with age, or is even replaced by a local advantage phenomenon. Moreover, there was an advantage of young over ``normal elderly subjects only in the con ictual condition, and for accuracy only. Experiment 1a was a control discrimination task designed to check whether subjects were able to process the global and local levels together when attention was not directed towards a speci ed level. Young subjects were clearly able to perform such an operation. The picture was a bit more complex in the elderly, as about one half of them displayed difficulties in processing the local level (whereas the remaining half did not di er from young subjects). These difficulties resulted mainly from a high rate of false alarms, a pattern often associated with aging (e.g., Bartlett & Fulton, 1991 ;Bartlett, Strater, & Fulton, 1991 ;Fulton & Bartlett, 1991). Given this state of a airs, Experiments 2 and 2a were planned. They were designed exactly as Experiments 1 and 1a, but subjects were selected only if they performed better than at random in the processing of the local level (con ictual condition, in Experiment 2). In agreement with what was observed in Experiments 1 and 1a, no young subject, but about 50% of elderly, were eliminated from Experiment 2 ; for Experiment 2a, a small number of young subjects (n 5 3), but a substantial number of elderly subjects (n 5 17), were eliminated, whereas the experiment had been made slightly easier than Experiment 1a. The young and elderly remaining subjects of Experiment 2 displayed the global precedence phenomenon and elderly subjects evidenced an increased sensitivity to interference of the global level over the processing of the local level. The young and elderly remaining subjects of Experiment 2a showed their ability to process both the global and the local levels when attention was not directed toward a prede ned level. Second, we consider the results of young subjects . Young adults were able to process each level separately, as shown by Experiments 1a and 2a. They also displayed the ``global precedence e ect in both accuracy and latencies in the basic Experiments 1, 2, and 3. In this study, young samples were only used as control groups for the assessment of the e ects of age. It thus appears that our young subjects performed as those enrolled by Navon (1997) and many other researchers (see the review by Kimchi, 1992). Third and nally, we focus on the results of the elderly while considering the pattern observed in young subjects. At rst glance, the ``normal subsets of elderly enrolled in Experiments 1 and 2 were able to process both levels of stimuli and displayed the global prevalence phenomenon, together with an increased sensitivity to interference in the con ictual condition, and a facilitation in the consistent condition. However, Experiment 3 revealed that these e ects of age were artifacts resulting from encoding difficulties (probably due to slight sensory disturbances ; Fozard, 1990). Thus, apparently, the ``global precedence e ect per se is
311
not sensitive to age in normal adult subjects. Moreover, the sensitivity of the elderly to interference is not due to a general and nonspeci c e ect of age on inhibition mechanisms (Hasher et al., 1991). A major lesson of this series of experiments is that there are subgroups among the elderly. Indeed, from one third to one half of elderly subjects displayed an anomalous pattern of performance by performing at random or worse than at random in the con ictual condition of the local task (Experiments 1 and 2) or in the detection of local di erences (Experiments 1a and 2c). This performance could not be accounted for by age, visual acuity, or verbal intelligence. These outcomes cannot be explained within the framework of the present research and constitutes a challenge for future studies. Indeed, the next step of this research program should be to try to understand why some older subjects can do the task efficiently, whereas other cannot. For instance, it would be useful to check experimentally whether this phenomenon is speci c to the present experimental design, or if it is the manifestation of a more general increase of sensitivity to interference in some subsamples of elderly people. Indeed, it has been shown that aging could reduce the resistance to interference in very di erent kinds of experiments, such as the Stroop e ect (e.g., Cohn, Dustman, & Bradford, 1984), selective attention (e.g., McDowd & Birren, 1990) inhibition processes (e.g., Hasher & Zacks, 1988), or proactive interference in long-term memory tasks (e.g., Van der Linden & Bruyer, 1991). Experiment 3 began to clarify this point but, obviously, elderly subjects should be submitted to other experimental designs as a matter of control, in order to approach the nature of possible subpopulations of elderly people. Whatever the case may be, one has to take the possibility of several subsamples of elderly subjects into account when this population is enrolled for experimental cognitive tasks. REFERENCES
Akshoomo , N. A., Delis, D. C., & Haist, F. (1993). Age-related changes in memory for visual hierarchical stimuli. Developmental Neuropsychology , 9, 259 269. Bartlett, J. C., & Fulton, A. (1991). Familiarity and face recognition: The factor of age. Memory and Cognition , 19 , 229 238. Bartlett, J. C., Strater, L., & Fulton, A. (1991). False recency and false fame of faces in old age. Memory and Cognition , 19, 177 188. Boles, D. B., & Karner, T. A. (1996). Hemispheric di erences in global versus local processing : Still unclear. Brain and Cognition , 30, 232 243. Breedin, S. D., Sa ran, E. M., & Coslett, H. B. (1994). Reversal of concreteness e ect in a patient with semantic dementia. Cognitive Neuropsychology , 11, 617 660. Bruyer, R. (1994). E ets du vieillissement sur la perception visuelle des objets et des visages (E ects of aging on visual perception of objects and faces). In M. Van der Linden & M. Hupet (Eds.), Le vieillissement cognitif (pp. 233 270). Paris : Presses Universitaires de France.
312
Christensen, K. J., Moye, J., Armson, R. R., & Kern, T. M. (1992). Health screening and random recruitment for cognitive aging research. Psychology and Aging , 7, 204 208. Cohn, N. B., Dustman, R. E., & Bradford, D. C. (1984). Age-related decrements in Stroop color test performance. Journal of Clinical Psychology , 40, 1244 1250. Coslett, H. B., Stark, M., Rajaram, S., & Sa ran, E. M. (1995). Narrowing the spotlight : A visual attentional disorder in presumed Alzheimers disease. Neurocase, 1, 305 308. Davido , J., & Donnelly, N. (1990). Object superiority: A comparison of complete and part probes. Acta Psychologica , 73, 225 243. Delis, D. C., Massman, P. J., Butters, N., Salmon, D. P. Shear, P. K., Demadura, T., & Filotoe, J. V. (1992). Spatial cognition in Alzheimers Disease: Subtypes of global-local impairment. Journal of Clinical and Experimental Neuropsychology , 14, 463 477. Delis, D. C., Robertson, L. C., & Efron, R. (1986). Hemispheric specialization of memory for visual hierarchical stimuli. Neuropsychologia , 24, 205 214. Deltour, J. J. (1993). Echelle de vocabulaire de Mill-Hill de J. C. Raven : Adaptation francaise et normes compare es du Mill-Hill et du Standard Progressive Matrices de Raven (P M38 ). Braine-le-Chateau, Belgium: Application des Techniques Modernes. Doricchi, F., & Incoccia, C. (1998). Seeing only the right half of the forest but cutting down all the trees ? Nature , 394 , 75 78. Estes, W. K. (1975). Memory, perception, and decision in letter identi cation. In R. L. Solso (Ed.), Information processing and cognition : the Loyola symposium (pp. 3 30). Hillsdale, N.J.: Lawrence Erlbaum. Fagot, J., & Deruelle, C. (1997). Processing of global and local visual information and hemispheric specialization in humans (Homo sapiens ) and baboons (Papio papio ). Journal of Experimental Psychology : Human Perception and Performance , 23, 429 442. Ferman, T. J., Primeau, M., Delis, D., & Jampala, C. V. (1999). Global-local processing in schizophrenia: hemispheric asymmetry and symptom-speci c interference. Journal of the International Neuropsychological Society, 5, 442 451. Filoteo, J. V., Delis, D. C. Demadura, T. L., Salmon, D. P., Roman, M. J., & Shults, C. W. (1994). Abnormally rapid disengagement of covert attention to global and local stimulus levels may underlie the visuoperceptual impairment in Parkinsons patients. Neuropsychology , 8, 210 217. Filoteo, J. V., Delis, D. C., Massman, P. J., Demadura, T., Butters, N., & Salmon, D. P. (1992). Directed and divided attention in Alzheimers Disease : Impairment in shifting of attention to global and local stimuli. Journal of Clinical and Experimental Neuropsychology , 14 , 871 883. Filoteo, J. V., Delis, D. C., Roman, M. J., Demadura, T., Ford, E., Butters, N., Salmon, D. P., Paulsen, J., Shults, C. W., Swenson, M., & Swerdlow, N. (1995). Visual attention and perception in patients with Huntingtons disease: comparisons with other subcortical and cortical dementias. Journal of Clinical and Experimental Neuropsychology , 17, 654 667. Fink, G. R., Halligan, P. W., Marshall, J. C., Frith, C. D., Frackowiak, R. S. J., & Dolan, R. J. (1996). Where in the brain does visual attention select the forest and the trees? Nature , 382 , 626 628. Fink, G. R., Halligan, P. W., Marshall, J. C., Frith, C. D., Frackowiak, R. S. J., & Dolan, R. J. (1997a). Neural mechanisms involved in the processing of global and local aspects of hierarchically organized visual stimuli. Brain , 120 , 1779 1791. Fink, G. R., Marshall, J. C., Halligan, P. W. & Dolan, R. J. (1999). Hemispheric asymmetries in global/local processing are modulated by perceptual salience. Neuropsychologia, 37 , 31 40.
313
Fink, G. R., Marshall, J. C., Halligan, P. W., Frith, C. D., Frackowiak, R. S. J., & Dolan, R. J. (1997b). Hemispheric specialization for global and local processing: The e ect of stimulus category. Proceedings of the Royal Society of London, 264B , 487 494. Fozard, J. L. (1990). Vision and hearing in aging. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (3rd ed.) (pp. 150 170). New York : Academic Press. Fulton, A., & Bartlett, J. C. (1991). Young and old faces in young and old heads : The factor of age in face recognition. Psychology and Aging , 6, 623 630. Granholm, E., Perry, W., Filoteo, J. V., & Bra , D. (1999). Hemispheric and attentional contributions to perceptual organization de cits on the global-local task in schizophrenia. Neuropsychology , 13, 271 281. Han, S., Fan, S., Chen, L., & Zhuo, Y. (1997). On the di erent processing of wholes and parts : A psychophysiological analysis. Journal of Cognitive Neuroscience, 9, 687 698. Hasher, L., Stoltzfus, E. R., Zacks, R. T., & Rypma, B. (1991). Age and inhibition. Journal of Experimental Psychology : Learning, Memory , and Cognition , 17, 163 169. 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). New York: Academic Press. Hubner, R. (1998). Hemispheric di erences in global/local processing revealed by same di erent judgements. Visual Cognition , 5, 457 478. Hughes, H. C., Nozawa, G., & Kitterle, F. (1996). Global precedence, spatial frequency channels, and the statistics of natural images. Journal of Cognitive Neuroscience, 8, 197 230. Kimchi, R. (1992). Primacy of wholistic processing and global/local paradigm: A critical review. Psychological Bulletin, 112, 24 38. Lamb, M. R., Robertson, L. C., & Knight, R. T. (1989). E ects of right and left temporal parietal lesions on the processing of global and local patterns in a selective attention task. Neuropsychologia , 27, 471 483. Lamb, M. R., Robertson, L. C., & Knight, R. T. (1990). Component mechanisms underlying the processing of hierarchically organized patterns : Inferences from patients with unilateral cortical lesions. Journal of Experimental Psychology : Learning, Memory and Cognition , 16, 471 483. MacLeod, C. M. (1991). Half a century of research on the Stroop e ect: An integrative review. Psychological Bulletin, 109, 163 203. Marshall, J. C., & Halligan, P. W. (1995). Seeing the forest but only half the trees? Nature , 373 , 521 523. Martin, M. (1979). Hemispheric specialization for local and global processing. Neuropsychologia , 17, 33 40. Martinez, A., Moses, P., Frank, L., Buxton, R., Wong, E., & Stiles, J. (1997). Hemispheric asymmetries in global and local processing: evidence from fMRI. NeuroReport, 8, 1685 1689. McDowd, J. M., & Birren, J. E. (1990). Aging and attentional processes. In J. E. Birren & K. W. Schaie (Eds), Handbook of the psychology of aging (3rd ed.) (pp. 222 233). San Diego: Academic Press. Navon, D. (1977). Forest before trees : The precedence of global features in visual perception. Cognitive Psychology , 9, 353 383. Palmer, S. E. (1980). What makes triangles point : Local and global e ects in con gurations of ambiguous triangles. Cognitive Psychology , 12, 285 305. Palmer, S. E., & Bucher, N. M. (1981). Con gural e ects in perceived pointing of ambiguous triangles. Journal of Experimental Psychology : Human Perception and Performance,
314
R. Bruyer and J .-C. Scailquin
7, 88114. Polster, M. R., & Rapcsak, S. Z. (1994). Hierarchical stimuli and hemispheric specialization : Two case studies. Cortex, 30, 487497. Posner, M. I., Walker, J. A., Friedrich, F. J., & Rafal, R. D. (1984). Eects of parietal injury on covert orienting of attention. Journal of Neuroscience, 4, 18631874. Rabbitt, P. (1993). Does it all go together when it goes? The nineteenth Bartlett memorial lecture. Quarterly Journal of Experimental Psychology, 46A, 385434. Reicher, G. M. (1969). Perceptual recognition as a function of the meaningfulness of the material. Journal of Experimental Psychology, 81, 275280. Ridderinkhof, K. R., & Van der Molen, M. W. (1995). When global information and local information collide: A brain potential analysis of the locus of interference eects. Biological Psychology, 41, 2953. Robertson, L. C., Lamb, M. R., & Knight, R. T. (1988). Eects of lesions of temporalparietal junction on perceptual and attentional processing in humans. Journal of Neuroscience, 8, 37573769. Sergent, J. (1982). The cerebral balance of power : confrontation or cooperation? Journal of Experimental Psychology : Human Perception and Performance, 8, 253 272. Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18, 643662. Thompson, P. (1980). Margaret Thatcher: A new illusion. Perception, 9, 483484. Valentine, T., & Bruce, V. (1985). Whats up ? The Margaret Thatcher illusion revisited. Perception, 14, 515 516. Van der Linden, M., & Bruyer, R. (1991). Interference proactive chez des sujets ages vivant en institution ou a domicile. Archives de Psychologie, 59, 313 320. ` Van Kleeck, M. H. (1989). Hemispheric dierences in global versus local processing of hierarchical visual stimuli by normal subjects: New data and a meta-analysis of previous studies. Neuropsychologia, 27, 11651178. Verhaegen, P., & De Meersman, L. (1998). Aging and the Stroop eect: A meta-analysis. Psychology and Aging, 13, 120 126. Weisstein, N, & Harris, C. S. (1974). Visual detection of line segments: An object superiority eect. Science, 186, 752755. Young, A. W., Hellawell, D., & Hay, D. C. (1987). Congural information in face perception. Perception, 16, 747759.

Global Study For Intro 1

Uploaded by

Copyright:

Available Formats

Global Study For Intro 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Global Study For Intro 1

Uploaded by

Copyright:

Available Formats

Experimental Aging Research, 26: 285 314, 2000 Copyright 2000 Taylor & Francis 0361-073X/00 $12.00 .

The Fate of Global Precedence With Age

R. Bruyer and J.-C. Scailquin

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

TABLE 1 Control Groups of Some Neuropsychological Studies

Advantage Global No Global Local Local No No No Local Local No

Precedence Global ? Global ? ? No ? ? ? Local No

Note : Papers are ranked according to the age of the samples.

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

TABLE 3 Results of Experiment 2: Mean Number of Errors (Out of 36) and

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

TABLE 4 Results of Experiment 3: Mean Number of Errors (out of 36) and

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

Age and Global Precedence

R. Bruyer and J.-C. Scailquin

Age and Global Precedence

R. Bruyer and J .-C. Scailquin

You might also like