Advances in Cognitive Psychology Attentional

Visual selective attention and visual working memory (WM) share the same capacity-limited resources. We investigated whether and how participants can cope with a task in which these 2 mechanisms interfere. The task required participants to scan an array of 9 objects in order to select the target locations and to encode the items presented at these locations into WM (1 to 5 shapes). Determination of the target locations required either few attentional resources (“popout condition”) or an attention-demanding serial search (“non pop-out condition”). Participants were able to achieve high memory performance in all stimulation conditions but, in the non popout conditions, this came at the cost of additional processing time. Both empirical evidence and subjective reports suggest that participants invested the additional time in memorizing the locations of all target objects prior to the encoding of their shapes into WM. Thus, they seemed to be unable to interleave the steps of search with those of encoding. We propose that the memory for target locations substitutes for perceptual pop-out and thus may be the key component that allows for flexible coping with the common processing limitations of visual WM and attention. The findings have implications for understanding how we cope with real-life situations in which the demands on visual attention and WM occur simultaneously.

If visual attention and visual WM share common resources and, thus, interfere when engaged simultaneously, the question is how these limitations can be over- come. An answer to this question should have relevance for many real-life situations. For example, while looking at a map and following the route between two locations, one might have to memorize the visual information needed to reach the destination, while at the same time using attention to search and navigate through the map.
Thus, the aim of the present study was to investigate the strategies that allow participants to deal with such concurrent demands on visual selective attention and encoding into visual WM.
Participants performed a task that combined the classical features of visual search experiments, which have been widely used in the study of selective attention (Treisman & Gelade, 1980;Wolfe 1998a), with those of visual WM studies (e.g., Oh & Kim, 2004;Olsson & Poom, 2005;Wheeler & Treisman, 2002). In each trial, participants were presented with an array of nine objects and had to memorize only some of them (targets), while the others could be ignored (distractors). Determination of the target locations was based on an L-shaped item located in the center of the object, but only the outer shape of the object and its orientation had to be remembered (see Figure 1). Thus, the present procedure allowed us to manipulate independently the demands on encoding into visual WM and the demands on attention for visual search of target locations. Attentional demand was manipulated by implementing two stimulation conditions in which the L-shaped items had either unique features (i.e., color) and were highly discriminable from the distractors (resulting in perceptual "pop-out" [PO]) or shared the features with the distractors and were thus difficult to discriminate ("non pop-out" [NPO]) (Duncan & Humphreys, 1989;Treisman & Gormican, 1988;Wolfe, 1998b). Only in the latter case did we expect that the determination of the target locations would require the attention-demanding serial search, which is commonly indicated by a linear increase in search times as a function of the number of distractor items in the array (Treisman & Gelade, 1980;Treisman & Sato, 1990). To manipulate the load of WM encoding, the number of target items was varied in each array, which ranged from one to five.
In the classical visual search paradigm, the display remains visible until the participant responds: Response accuracy is usually high. Therefore, response time (RT) is the most important measure in this paradigm as it indicates the amount of time required to determine the presence or absence of a target that is presented among distractors (Duncan & Humphreys, 1989;Treismann & Gelade, 1980;Treisman & Sato, 1990;Wolfe, 1998aWolfe, , 1998b. This set-up was highly instrumental in the development of one of the most successful theories in psychology: the feature binding (feature integration) theory (Treisman, 1998;Treisman & Gelade, 1980;Treisman & Sato, 1990). In this paper the same concepts have been used to study the processes underlying the encoding of information into visual WM. Thus, the most important dependent variable was the presentation time of the stimulus array that participants needed to achieve good WM performance, and which they self-paced by a key press. We investigated how this time changed as a function of memory load and of attentional demand.
A similar dependent variable has been used in a recent study that investigated the role of visual WM for the formation of visual long-term memory (LTM, Nikolić & Singer, 2007). These authors first estimated the WM capacity for the locations of the target stimuli that either did or did not pop-out from the distractors, and then requested the participants to memorize accurately a number of target locations that grossly exceeded the capacity of WM. The participants self-paced the memorization process and the obtained encoding times were measured reliably (r > .90) and increased linearly as a function of target set size. Importantly, the changes in the slopes of these linear functions could be predicted accurately from the changes in the estimated WM capacities for the same stimuli. The authors concluded that the capacity of WM determined the speed with which visual LTM was created. This provided the missing evidence that visual WM played a pivotal role in the storage of information in visual LTM. Nikolić and Singer reported that the self-paced measure of the encoding times was reliable given that an immediate performance feedback was supplied at each trial, which, in turn, enabled the participants to learn quickly, on a trial-and-error basis, the minimum amount of effort (time) that was needed to achieve the required level of performance (95% correct in that study). In contrast, if such feedback was not provided, participants tended to shorten the encoding time and hence, trade the accuracy for speed. An imhttp://www.ac-psych.org portant advantage of using the presentation time as a dependent variable in the present study was, similarly to the analyses conducted in the previous studies (Nikolić & Singer, 2007;Treisman & Gelade, 1980), that we could describe and analyze the data quantitatively by simple mathematical functions based on linear fits of differing intercepts and slopes. Nikolić and Singer's study (2007) investigated the WM capacity for the locations of the target stimuli only, thus without any additional contents presented on the display. In that study, WM could be loaded with very short stimulus presentations of about 1 s. In the present study we investigated the WM for relatively complex objects that were presented at the target locations. Thus, participants needed not only to select the target locations but also to extract and memorize the various shapes that were presented at these locations. This required a much longer presentation time than 1 s, as the information could not be loaded "directly" but successful encoding required the participants to engage into a more elaborated processing. The main goal of the present study was to investigate the nature of these processing steps, and to this end, two types of strategy were considered.
In a "search-and-encode strategy" participants encoded each shape as soon as they selected a relevant location, thus interleaving the search process with the WM encoding. In this case, presentation time should be simply divided between the two task components, and the presentation time that participants need in the non pop-out condition should be the sum of the presentation time in the pop-out condition and the time needed to select the relevant locations in the non pop-out condition. Thus, as empirical support for the search-andencode strategy, we looked for evidence that the times for encoding and determination of target locations are additive.
The other considered strategy was postulated to involve two separate steps of encoding ("two-step encoding strategy"). In the first step participants selected and memorized only the locations of all target items and only then encoded the associated shapes at a later step.
The additional process of memorizing the target locations requires additional processing time. For that case, a super-additive combination of the times for encoding and determination of target locations in the non pop-out condition was predicted. The time needed to memorize the locations was directly measured and whether this time corresponded to the additional time required to encode the target shapes in the non pop-out condition was investigated.
Importantly, the two-step encoding strategy but not the search-and-encode strategy implies interference be-tween WM encoding and attention. A search-and-encode strategy should be possible if the two components need to be executed sequentially but do not interfere with each other, that is the search for a new target does not erase the contents stored previously in WM. As the existing evidence suggests that this is not the case (Awh et al., 1998;Barrouillet et al., 2007;Jolicoeur & Dell'Acqua, 1998Oh & Kim, 2004;Smyth & Scholey, 1994;Woodman & Luck, 2004), the two-step encoding strategy was considered as a possible tactic for overcoming this interference. Therefore, if empirical evidence favors one of the two strategies, the result also provides indirect information on whether, in this task, visual WM encoding and attention interfere.

Synopsis of experiments
We conducted five experiments in which the study phase always consisted of identical stimuli, the tasks differing only in the instructions and in the test displays.
Participants were debriefed at the end of each experiment and were asked about their subjective experience and strategies. In the main experiment (Experiment 1), participants encoded complex target shapes into WM, while determining their locations in a low or high attention-demanding visual search task (i.e., presence or lack of perceptual pop-out). WM performance was comparable across search conditions. Presentation time increased with increased WM load and, most importantly, with the lack of pop-out. Further experiments (Experiments 2 to 5) investigated the reason for the increase in the presentation time by contrasting the two, above described, strategies. Experiment 2 and 3 tested the hypotheses of additivity versus super-additivity of the times needed to encode and determine the target locations. In Experiment 2, the time needed for simple visual search was measured. These times could not explain the increased presentation time produced by the lack of pop-out in Experiment 1. Therefore, Experiment 3 tested whether the slower processing in the non pop-out condition in Experiment 1 could be explained by repeated searches, owing to a putative lack of memory for visited target locations (Irwin, 1992;Peterson, Kramer, Wang, Irwin, & McCarley, 2001) and the need to search the entire array. The need to search repeatedly was reduced by informing the participants at each trial about the upcoming number of targets. The time saved by this manipulation again could not explain the costs on presentation time produced by the lack of pop-out in Experiment 1. Therefore, the results from Experiments 1 to 3 indicated http://www.ac-psych.org Jutta S. Mayer, Robert A. Bittner, David E. J. Linden, and Danko Nikolić consistently super-additivity of the times for encoding and determination of the target locations, favoring the two-step encoding strategy.
In the remaining two experiments (Experiments 4 and 5) the two-step strategy was tested further.
The times were measured that participants needed to memorize the locations of the target items only and whether these times could explain quantitatively the difference between the pop-out and non pop-out conditions in Experiments 1 and 3 was investigated. Indeed, in Experiments 4 and 5, the times needed to memorize the target locations accounted well for the presentation time offsets between pop-out and non pop-out conditions in Experiments 1 and 3, respectively. These results again favored the two-step strategy.

EXPERIMENT 1
Experiment 1 was used to investigate whether and how participants can encode complex objects into WM, while engaging selective attention for a visual search task.
Participants memorized the shapes of only those objects whose center items matched the target items, and were instructed to ignore all the other objects. Determination  The display in the study phase consisted of nine different grey geometric shapes (each spanning approximately 1.1° × 1.1° of visual angle), arranged in a 3 × 3 matrix, and presented in the center of the screen and on a black background. The shapes were selected at random without replacement from a set of 12 shapes and each was oriented randomly in one of the four possible directions, so that in total it was necessary to discriminate between 48 different objects. In the center of each shape a small L-shaped item (0.3° × 0.3°) was placed. The Ls appeared in one of four different orientations (0, 90, 180, or 270°, clockwise) and were either blue or red in color (see Figure 1). Participants needed to memorize only the shapes associated with an L-oriented 90° (target items). The shapes associated with Ls of other orientations could be ignored (distractor items).

Participants, apparatus, and stimuli
The number of target items within each display varied randomly between one and five. In the pop-out condition target Ls always appeared in blue and distractors in red. Distractor Ls were always oriented at 270°. In the non pop-out condition each target and distractor was assigned randomly to either the color blue or red.
In this condition, the distractor items could be any of the remaining three orientations (0, 180, and 270°). In the test phase participants were presented with a single shape in the center of the screen and without the center item. The luminance of the shapes, the blue, and the red center items was 12.3, 6.01, and 9.87 cd/m², respectively. The background luminance was 0.01 cd/m².
During the delay period a white central fixation cross was presented on a blank screen (0.2° × 0.2°, 60.06 cd/m²).

Design and procedure
A 2 × 5 within-subjects factorial design was used, with two levels of attentional demand for determination of the target locations (pop-out and non pop-out) and five levels of WM load, determined by the number of targets (one to five targets). Each of the 10 experimental conditions was presented equally often (12 trials per condition). Pop-out (PO) and non pop-out (NPO) conditions were presented in separate blocks of 10 trials, with six blocks for each condition. This amounted to a total of 120 experimental trials per participant. The trials were fully randomized within blocks and pseudo-randomized across blocks and across participants. Before starting a new block, participants were always given instructions about the targets they needed to search for. At the beginning of the experiment participants performed two practice blocks of 10 trials, one for each of the two athttp://www.ac-psych.org tentional conditions. Each trial began with the presentation of the nineitem array, which remained visible until the participant pressed the response key. Participants had to determine the target locations and to memorize the shapes associated with the targets. The time they needed to achieve high memory performance, indicated by a key-press, was used as a dependent variable (presentation time).
Participants were also instructed to place emphasis on accuracy over speed in order to ensure that response accuracy was high and comparable across different attentional-demand conditions. After the display disappeared participants fixated a cross during a delay period of 8 s, which was followed by the presentation of a single test shape. Participants were then required to indicate whether the test shape matched in form and orientation one of the target shapes presented previously by pressing the "Y" or "N" key for match and non-match, respectively. Half of the trials were matches. In 50% of the non-matches the probe stimuli differed with respect to the shape, and in the other 50% with respect to the orientation. The non-matches probe stimuli were selected from the set of all possible shapes that were not used as a target in a given trial. After each response feedback was given ("wrong", "correct", or "no response"), which was followed by an inter-trial interval of 3 s. Analyses of presentation time included only correct trials (see Figure   1 for an illustration of the sequence of events at each trial). The experimental procedure lasted approximately 60 min for each participant. After the experiment, participants were asked within a semi-structured interview freely to recall the strategies they used to accomplish the task. They were asked the following questions: 1. What strategies did you use for searching the targets in the PO and NPO conditions?

2.
What strategies did you use for encoding the objects in the PO and NPO conditions?
3. What strategies did you use for memorizing the objects in the PO and NPO conditions during the delay period? Figure 1.

Accuracy at test
Overall, response accuracy for the WM task was high (on average 85% correct) and decreased with the number of shapes that needed to be encoded -from 93% correct, with WM load 1 to 75% correct with WM load 5 in the pop-out condition, and from 93% correct with WM load 1 to 78% correct with WM load 5 in the non pop-out condition (see Figure  According to Luck and Vogel (1997), the load-dependent decrease in accuracy is likely to reflect the limited ability of maintaining information in visual WM rather than the limitations of the encoding process. Thus, this drop in performance should not have affected the processes of encoding information into WM, which was the main focus of our analyses.

Presentation time
Participants were slower without than with perceptual pop-out and the presentation time increased with the number of targets that needed to be encoded (see  Importantly, the interaction between attentional demand and the number of targets was not significant, F(4, 140) = 1.2, p = .32, indicating that the slopes relating the average presentation time to the number of targets were practically identical in the two attentionaldemand conditions. The offset between the two slopes, that is the difference between non pop-out and pop-out conditions, ranged between 4008 and 4853 ms with an average of 4490 ms (see Table 1). Thus, the manipulation of attentional demand added considerable processing time but this time was constant across the number of targets. This result indicates that the manipulation of attentional mechanisms produced an effect on presentation time that was independent of the effect produced by the manipulation of WM load. Therefore, the results from Experiment 1 suggest that participants can achieve high memory performance despite the lack of pop-out but that this comes at the price of longer presentation time.

Reported encoding strategies
The majority of participants (32 of 36) reported that in the non pop-out condition they needed to use a two-step encoding strategy: In the first step they selected and memorized the locations of all the target items, encoding the associated shapes only in the second step.
Three participants reported using a search-and-encode strategy in the non pop-out condition, encoding each target shape immediately after selecting a target item and making only one sweep through the array. One participant did not report any specific strategy.
We found no significant differences in response ac- Five of the participants had also taken part in Experiment 1.
Participants were required to count the target items in the same stimulus array as used in Experiment 1. After completing the count, participants indicated the search time by pressing the "return" key on the computer keyboard. After this key-press a question mark appeared in the center of the screen that prompted the participants to enter the number of the counted targets. Participants were instructed to place the emphasis on accuracy over speed during the counting process and were informed that the time needed to enter the counted number of targets was irrelevant. After each response, the question mark disappeared and feedback ("wrong", "correct", or "no response") was provided and followed by an intertrial interval of 3 s. Only correct trials were included in the analyses of counting times. The experimental proce dure lasted approximately 30 min for each participant.
We used a 2 × 5 within-subjects factorial design with two levels of attentional demand for determination of the targets (pop-out and non pop-out) and five different counts (one to five targets).

Accuracy at test
Overall, response accuracy was high (on average  Table 1).
The  (Duncan & Humphreys, 1989;Treisman & Gormican, 1988;Wolfe, 1998b), it is consistent with reports that search time increases with the complexity of the items (Alvarez & Cavanagh, 2004). The slower search speed in our task than in standard visual search tasks cannot be simply explained by the need to select and count multiple targets because such tasks do not produce similar increases in response times (Horowitz & Wolfe, 2001). It can also be excluded that the prolonged search time was a result of the instruction to emphasize accuracy because, in one control experiment (not reported here), we instructed 10 participants to count the target items as quickly as possible and obtained only slightly faster search times (280 ms to scan each of the nine locations). Another reason why visual search was so slow in the present experiment might be that attention tends to be locked onto perceptual objects. When attention is voluntarily placed upon one feature of an object it automatically spreads to other features of the same object (Duncan, 1984;Scholl, 2001;Vecera & Farah, 1994

Reported search strategies
In the non pop-out condition all participants reported scanning the array serially, mostly from the upper left corner towards the lower right, and making one single sweep through the array. In the pop-out condition participants reported detecting the target items at a glance.

EXPERIMENT 3
The aim of this experiment was to assess whether repeated searches could explain the difference between the presentation time of the non pop-out and pop-out conditions of Experiment 1. Several studies have demonstrated that the temporary storage of previously searched target locations decays over time (Irwin, 1992;Phillips, 1974) and that participants sometimes need to repeat the search at target locations that they have already visited previously (Peterson et al., 2001).  A second question addressed by Experiment 3 concerned the role of verbal coding. The phonological store is highly efficient for serial recall, thus participants tend to recode visually presented items into a verbal code (Baddeley, 2000). Indeed, in Experiment 1, the majority of participants reported creating their own verbal labels for the complex shapes. As the aim of the present study was to investigate visual attention and WM, it was necessary to assess the role of verbal encoding during the encoding of the shapes into WM. To this end, an articulatory suppression task was implemented that is known to reduce, albeit not completely eliminate, subvocal rehearsal and the phonological encoding of visually presented material (e.g., Baddeley, 2000Baddeley, , 2003Besner, Davies, & Daniels, 1981;Murray, 1968

Participants, apparatus, stimuli, procedure, and design
Sixteen students and employees of the University of Frankfurt/M. (7 males, 9 females) participated in this experiment. The mean age of the participants was 24.6 years (range 18-44). Six participants also took part in Experiment 2, only one of them took part in Experiment 1.
The stimuli, procedure, and design were the same as in Experiment 1, apart from the following two differences. First, at the beginning of each trial a digit was presented at the center of the screen, for 2 s. This digit indicated the number of target items that would be presented in the upcoming stimulus array. Second, the articulatory suppression task required participants to repeat aloud a syllable la throughout the duration of the trial.

Accuracy at test
A repeated measures ANOVA revealed a significant main effect of number of targets, F(4, 60) = 13.4, p < .001, η² = .47, but no effect of attentional demand, F(1, 15) = 2.8, p = .12. The interaction between the two factors also reached significance, F(4, 60) = 3.3, p < .05, η² = .18, but the averages did not show any consistent relationships between the variables (see Figure 4, upper panel) and explained only 18.1% of the variance in the dependent factor. Therefore, this interaction was not used for further interpretation of the results.
These results are highly consistent with those ob-

Presentation time
Similarly  Table 1). Thus, as predicted, the presentation time One possibility, as suggested by the finding that WM and attention interfere (Awh et al., 1998;Barrouillet et al., 2007;Jolicoeur & Dell'Acqua, 1998Oh & Kim, 2004;Smyth & Scholey, 1994;Woodman & Luck, 2004), as well as by the subjective reports of our participants, is that they invested the additional time in the process of memorizing all target locations prior to encoding their shapes. This two-step strategy was investigated more directly in Experiments 4 and 5.

Reported encoding strategies
All 16 participants reported using the same two-step strategy as described by the majority of participants in Experiment 1.

EXPERIMENT 4
In Experiments 4 and 5 we explicitly tested the strategy that was reported by the participants during the debrief-

Participants, apparatus, stimuli, procedure, and design
Sixteen students and employees of the University of Eight participants took part in Experiment 1 and 2 of them also took part in Experiment 2.
The stimuli, procedure, and design were the same as those in Experiment 1, apart from the following two differences. Participants were instructed to determine and memorize the locations of the target items only http://www.ac-psych.org and thus to ignore the shapes of the associated objects.
In order to probe WM for target locations, the original stimulus array was presented at the test phase without the center items and with one of the shapes missing.
Participants needed to indicate whether the location of the missing shape matched one of the target locations.
After each response feedback was given (see Figure 1).

Accuracy at test
Overall, response accuracy was again high (on average 93% correct). A repeated measures ANOVA revealed only a significant main effect of number of targets, F(4, 60) = 5.8, p < .01, η² = .27. Neither attentional demand nor the interaction between the two factors was significant, F(1, 15) = 0.01, p = .96, and F(4, 60) = 0.7, p = .57, respectively. Thus, similar to Experiment 1, response accuracy decreased with the number of targets whose locations needed to be encoded and, again, did not differ between pop-out and non pop-out conditions (see Figure 6, upper panel

Reported encoding strategies
The majority of participants (15 of 16) reported integrating the target locations into one or two perceptual representations that could be described either as a spatial template, a shape composed of the individual locations, or as a chunk. One participant reported encoding discrete locations, one after another, without a particular perceptual organization.

EXPERIMENT 5
When informed about the upcoming number of targets in Experiment 3, participants also reported using http://www.ac-psych.org Jutta S. Mayer, Robert A. Bittner, David E. J. Linden, and

Participants, apparatus, stimuli, procedure, and design
Ten students and employees of the University of The stimuli, procedure, and design were the same as in Experiment 4, apart from the following two differences. First, the procedure from Experiment 3 was used to inform participants about the number of upcoming targets at the beginning of each trial. Second, the articulatory suppression task was implemented. Most importantly, the offsets in Experiment 5 did not differ significantly from those obtained in Experiment 3 (range 2145-3937 ms), F(1, 24) = 0.8, p = .37 (see Table 1). Next, we investigated the degree to which

Reported encoding strategies
The majority of participants (9 of 10) reported using the same chunking strategy as described by the majority of the participants in Experiment 4.

GENERAL DISCUSSION
The present study investigated whether and how partici- subjective reports about the strategy that they used to achieve the objectives of the task.
It might be argued that other processes than those related to the memorizing of target locations contributed to the additional time cost in the non pop-out condition.
WM suffers from a time-related decay as soon as attention is switched away and captured by concurrent activities (Barrouillet et al., 2007). Thus, the additional time cost in the non pop-out condition might also be related to an increased need to interleave the atten- reports, seemed to be most consistent with the two-step strategy that involved memorizing the locations of all the targets before memorizing the associated shapes.
Why would participants need to memorize target locations? One possible reason is that this is how they cope with the interference between WM and attention that would otherwise take place. Interference between selective attention and the storage of information in spatial WM has been well documented and interpreted in terms of common cognitive resources shared by these processes (Awh et al., 1998;Mayer et al., 2007;Oh & Kim, 2004;Smyth & Scholey, 1994;Woodman & Luck, 2004). The present findings suggest that interference between selective attention and WM encoding may not be restricted to the spatial domain, unlike the findings for WM maintenance (Oh & Kim, 2004;Woodman, Vogel, & Luck, 2001). Instead, it seems likely that in the non popout condition of the present experiment, interference occurred between the attentional resources needed for determination of the target locations (Treisman, 1998;Treisman & Gormican, 1988) and the WM resources needed to encode targets' shapes.
What is the common mechanism required by the visual search and the encoding of object information into WM? Selective attention appears to be that mechanism.
Representations of spatial locations are maintained in WM by keeping the spotlight of attention on these locations (Awh & Jonides, 2001;Awh et al., 1998 (Treisman & Gelade, 1980) and the storage of such information in WM requires capacity-limited attentional mechanisms as well (Wheeler & Treisman, 2002).
The implication of the present study is that the memory for locations may provide a coping mechanism for interference between search and memory. In the pop-out condition the unique elementary features attract the spotlight of attention by automatic bottom-up mechanisms (Treisman & Gelade, 1980). Along similar lines, the locations in the non pop-out condition, once memorized, might guide the attentional spotlight in an automatic-like fashion. Consistent with this notion, it has been proposed that in order to search for multiple targets efficiently, participants use spatial WM to keep track of identified targets (Horowitz & Wolfe, 2001).
It is possible that this storage of target locations was based on visual LTM because LTM is, in general, a tool for coping with capacity limitations. LTM is used during the chunking processes in WM (short-term memory) tasks (Chase & Simon, 1973;Cowan, 2001;Gobet et al., 2001;Miller, 1956) and is responsible for the development of skills and expertise in general (Chase & Ericsson, 1981;Hasher & Zacks, 1979;Shiffrin & Schneider, 1977). The main advantage of maintaining information in LTM, as opposed to WM, is that such storage does not seem to rely on capacity-limited resources (Ericsson & Kintsch, 1995;Phillips & Cristie, 1977). It has recently been shown that in a task similar to the present one, participants can readily store target locations into LTM when they need to memorize a number of locations that greatly exceeds the capacity of visual WM for such locations (Nikolić & Singer, 2007).
Real-life situations in which interference between WM and attention occurs may require similar coping mechanisms. One example of a cluttered visual scene, in which not only serial search but also other forms of spatial processing are needed, is map reading (e.g., Garden, Cornoldi, & Logie, 2002;Thorndike & Hayes-Roth, 1982). To find a suitable route, first the key locations (e.g., the origin and destination) need to be identi-fied, and only can then the rest of the route be explored.
If the route is non-trivial (multiple locations in-between and turns are involved), there might be at first interference between the memory for the examined part of the route and the search for the rest of the route. However, over time, as the route is being studied, knowledge will be acquired (including information about the sequence of landmarks along the route or about metric distances and angles that are integrated into a configural cognitive map), and the access to the route should gradually become easier. Similar processes should apply to other activities that involve visual WM and attention such as navigating through complex technical drawings or within one's environment (Foo, Warren, Duchon, & Tarr, 2005;Garden et al., 2002;van Asselen, Fritschy, & Postma, 2006). In general, memory for locations might be the very mechanism that allows us to extract and encode relevant information from complex visual scenes when obvious cues that automatically draw attention are not available.