Navigating web-based environments: Differentiating internal spatial representations from external spatial displays

Contemporary Educational Psychology 32 (2007) 551–568 www.elsevier.com/locate/cedpsych

Navigating web-based environments: Differentiating internal spatial representations from external spatial displays Neil H. Schwartz a,*, Michael P. Verdi b, Terra D. Morris c, Tiffany R. Lee d, Nikki K. Larson a a

Department of Psychology, California State University Chico, Chico, CA 95929-0234, USA b California State University, San Bernadino c University of Colorado d University of Washington Available online 27 June 2007

Abstract Fifty-five undergraduate students read pages on a website presenting text about familiar and unfamiliar geographic locations in the United States. Learners navigated the site by having available or unavailable navigational buttons showing the cardinal compass directions between the map locations in the presence or absence of a cartographic map enabling them to click specifically on the map locations. Learners navigated the site with the goal of learning as much information as they could about the geographic locations described on each page. Results revealed that learners remembered significantly more page content when the geographic map was present and the geography of the area was unfamiliar. However, when the geographic area was familiar, learners remembered more page content when the navigational map was absent regardless of whether the directional links to each page were present or not. The findings: (1) support the contention that maps are used differently as graphic displays for navigating a website than for comprehending associated text, and (2) reveal that learners develop a cognitive model of text and graphics when the content familiar is high, but simply use an image as a mnemonic when content familiarity is low. Published by Elsevier Inc. Keywords: Web navigation; Graphics; Comprehension; Maps; Text

*

Corresponding author. Fax: +1 530 898 4968. E-mail address: [email protected] (N.H. Schwartz).

0361-476X/$ - see front matter Published by Elsevier Inc. doi:10.1016/j.cedpsych.2007.03.004

552

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

1. Introduction Much of the research on learning in web-based environments is concerned with the ability of learners to develop a cognitive representation of the navigational structure of a website. By cognitive representation we mean the internal model a learner constructs from the semantic, spatial, and imaginal elements of the web-based environment. Navigational structure refers to the layout of the site, defined by the articulation of webpages via hyperlinks and nodes. Navigational structure is important to consider because learners spend much of their cognitive effort orienting to the structure of the site at the expense of the elaborative processes necessary for deriving deep levels of comprehension of website information (Eveland & Dunwoody, 2000). McDonald and Stevenson (1998), for example, showed that the learning outcomes of learners navigating a web environment are decreased when the navigational structure of the environment is unavailable. Even forcing learners to think about the structure of the environment does not necessarily lead to better comprehension (Nilsson & Mayer, 2002). The problem has led many instructional designers to design navigational maps—spatially based graphic displays intended to assist learners in deriving a website’s navigational structure and finding the locations of pertinent pages within it (cf. Brusilovsky & Rizzo, 2002). The maps are intended to decrease cognitive load (cf. Paas, Renkl, & Sweller, 2003). This investigation was designed to examine how learners cognitively process the graphics designed to help them navigate these sites. Theoretically, we were interested in how the graphics are cognitively represented, and how these representations are used during learning—whether they are encoded as a unitary image learners scan in working memory to navigate from one website location to the next; or whether they are integrated into a more complex representation where website semantic and location information must be integrated with the same information already represented in memory. We make the assumption that the representation contains: (a) previous knowledge about the spatial relations between locations, (b) semantic elements of knowledge previously associated with those locations, and (c) multiple idiosyncratic images integrated with the two. We also assume that the representation is fundamentally different for learners who have prior knowledge of the elements of information associated with the knowledge domain. Thus, the purpose of this investigation was to determine whether learners navigating a website use an internal cognitive representation constructed from the integration of text content and the spatial layout of the site, or an externally derived image from a map of the site, to encode the text-based information contained on site pages. In short, it is unclear how learners process spatially based graphics when the graphics are used for purposes of website navigation. It is equally unclear whether the graphic displays are processed in the same way as graphic displays accompanying text when the graphics are used exclusively for purposes of text comprehension. Two lines of research can be used to illuminate the role of graphics and text when graphics are used for purposes of navigation. One line addresses the role of graphics in the comprehension of text. The other is concerned exclusively with the navigational function of the graphics. While both lines of research have examined graphics and text within hypermedia environments, each has ignored the role of graphics when the graphics serve both navigation and text-comprehension functions. Thus, in the present investigation, we used geographic information in both webpage text and a site map graphic to tease apart the contribution of the graphic in text comprehension and web navigation. Our choice of

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

553

geography was based on the assumption that geographic information provides the ability to test the difference between a graphic functioning as: (a) a mnemonic (e.g., an image) for remembering semantic elements of text, and (b) an integral part of a cognitive model derived from the association of information in both referents. 1.1. Text–graphic comprehension Schnotz, Bannert, and Seufert (2002) have proposed a model for learning in an Internet environment that describes the functions of graphic displays with text. Their model states that there are two branches of representation generated when a learner encounters texts and graphics on the web. The first, a descriptive branch consists of the external text, the internal mental representation, and the propositional representations created by the learner from the text. The second, a depictive branch, contains the external graphic display, the visual perception of the display and the mental model of the depicted subject matter. The end result of a learner’s interaction with a map and text in an online environment is a mental representation that has aspects of both the display and the text material presented. Other researchers have investigated the way graphic displays function to help learners remember text. Kulhavy and his colleagues, in their Conjoint Retention Model (cf. Kulhavy & Stock, 1996; Verdi & Kulhavy, 2002), used Paivio (1986) dual coding theory to explain that maps, as graphic adjuncts to text, are processed in a visual store, while text is processed in the verbal store. The referential links between the two allow the propositions of text held in the verbal store to be placed at specific locations on the map in the visual store. Once placed there, the map image is used by learners to ‘‘unpack’’ the verbal propositions of text during the retrieval process. Thus, Kulhavy contended that the map serves as an organizational framework with which to encode, store and retrieve text elements in memory. That is, map–text elements are combined into one informational ‘‘chunk’’ and sent to long-term memory as a unitary representation (Kulhavy, Lee, & Caterino, 1985). While the research of Kulhavy and his colleagues was restricted exclusively to paper and pencil geographic maps, Verdi, Crooks, and White (2003) showed that the functions of these referents were maintained when they were presented in an Internet environment. That is, learners are able to use digital maps in the same way learners use paper and pencil maps. Mayer (2003), on the other hand, proposed a cognitive theory of multimedia learning suggesting that graphics and text are combined into a cognitive representation. The representation is the result of active knowledge construction by the learner in which information is selected, organized, and integrated from the pieces of visual and verbal information contained in the graphics and text. Mayer (2003) contended that the constructive process occurs within two separate processing systems—a visual system for processing visual knowledge and a verbal system for processing verbal knowledge. The end result is an integrated representation comprised from a one-to-one mapping between the elements, actions, and relations in the visual and verbal representations of working memory, each of which have been borne from the graphic and text, respectively. This representation also pulls from prior knowledge in an iterative fashion that guides the selection, organization and integration process. Thus, while Mayer (2003) cognitive theory is similar to Kulhavy and Stock (1996) conjoint retention model, in that they are both firmly planted in Paivio’s theory of dual coding, Mayer (2003) contends that the cognitive process of combining the two referents is generative—an active merging process between both referents and prior

554

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

knowledge that goes beyond a rather static linkage between the two. Kulhavy and Stock (1996) posit a more mnemonic, less constructive, process. It is important to note that Mayer (2003) theory accounts well for text–graphic comprehension among learners who have well-developed spatial skills, in addition to low and high prior knowledge of the material the referents present. That is, Mayer and his colleagues (e.g., Mayer, Steinhoff, Bower, & Mars, 1995) found that when learners are high in prior knowledge their meaningful learning, for example as demonstrated on tests of problem solving transfer (Mayer & Gallini, 1990), is similar whether graphics are present or not. Alternatively, when learners’ prior knowledge is low, learning performance is significantly better when illustrations are presented with text. Mayer et al. (1995) explained these findings by suggesting that students who possess high levels of prior knowledge create their own mental images when processing propositions from text and use those images to build connections between verbal and visual representations. In contrast, students who lack prior knowledge independently create useful mental images solely from the verbal materials. Kulhavy and his colleague’s research have restrictions as well. Their investigations produced findings applicable only to unfamiliar maps. Thus, the conjoint retention model (CRM) was applied only to geographic spatial displays for which learners had no prior knowledge. Other researchers, Robinson and his colleagues for example (cf. Robinson, Katayama, & Fan, 1996) found that the CRM explained other graphic adjuncts as well. However, these graphics were principally charts, diagrams and tables—graphics for which interpretation was difficult, or impossible, without the text. Griffin and Robinson (2005) also recently added that it is the mimetic nature (e.g., symbolic similarity) of the locations on an unfamiliar map, and not the integrated status of the locations by virtue of their placement within a map, that is responsible for better retention. The point is that, in all research examining the CRM, the graphics contained no features capable of summoning learners’ prior knowledge of the target content domain. In short, the CRM appears to apply only to graphics that convey little or no inherent information for which learners are already familiar. We believe the model proposed by Schnotz et al. (2002) is much more robust in accounting for graphics of all types, even when they are familiar and subject-matter prior knowledge is either high or low. According to the model of Schnotz et al., learners construct a sort of hybrid mental model derived from a dynamic interaction of inspection and evaluation between multiple levels of graphic–text representation—at the surface and propositional levels of the text, and the visual perception levels of the graphic. The inspection and evaluation processes, when driven by prior knowledge, are top down. That is, when domain knowledge already exists in the learner, the learner’s existing mental model is activated and serves to evaluate, through a top-down analogical mapping approach, similarities between the model and the text and graphic. When prior knowledge is low, the learner uses the surface, propositional and visual perception levels of the text and graphic, respectively, to construct the mental model via the same mapping approach, but this time, bottom up. In short, when prior knowledge is high, evaluation is made using the image of the graphic; when low, evaluation follows a process of construction derived from it (Schnotz, 1993, 2002). Still however, both models—Schnotz et al. (2002) and Kulhavy and Stock (1996)—do not address the role of graphics in deriving comprehension of accompanying text when the graphics are used for purposes of web-based navigation. The role of structure afforded by

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

555

a graphic—structure which allows for text-based propositions to be combined for comprehension (cf. Schwartz, 1997)—may be particularly different than the structure a map provides when used for navigational purposes on the web. Mayer and his colleagues, in work investigating the navigational role of graphics, provide evidence that this is true. 1.2. Text–graphic navigation According to Nilsson and Mayer (2002), when learners are afforded graphic information about the navigational structure of a website, the learners perform more poorly on content comprehension relative to learners who receive a graphic void of the site’s spatial structure. The finding was revealed in two experiments in which undergraduates learned more webpage content as cognitive load (induced by navigational graphics) was reduced. That is, when navigating a 150-page hierarchically arranged site, learners were provided with either a clickable or a non-clickable map in conjunction with an explicit or implicit organizer as a navigation aid. When the learners had benefit of the better aid, their navigation improved but their learning did not. Instead, learners’ comprehension suffered due to the cognitive load induced by the necessity to negotiate information of the site layout with information on the webpages. In short, the learners who received the graphic intended to facilitate navigation, performed better at navigating the site, but performed more poorly on their comprehension of site content. This suggests that the use of graphics for navigational purposes comes at a cost. Learners navigate the site more efficiently, but fail to acquire the information contained there. Schwartz and his colleagues (Schwartz, Andersen, Hong, Howard, & McGee, 2004; Scott & Schwartz, 2007) obtained data which shed light on Nilsson and Mayer (2002) finding. Scott and Schwartz revealed that learners must summon higher order cognitive skills in order to make use of a navigational map for dual purposes—that is, for the purpose of navigation, and the purpose of making meaning of hypermedia text. When learners process both the semantic and the spatial features of a navigational map, the text materials presented on the web are significantly better understood, but only when learners have the metacognitive skills necessary for evaluating and monitoring navigational map utility (Scott & Schwartz, 2007). Schwartz, Verdi, Larson, and Rittschof (2004) found that children searching a science website about ecology using a navigational graphic learned more about the content contained there, but again, only when they had better developed metacognitive skills. In short, learners engage executive processes of working memory to negotiate the functional utility of a graphic when learning on the web. 1.3. The present investigation In the present investigation, we were concerned with whether learners use a geographic map during web-based learning to derive a navigational structure of a site when they are familiar with the geographic area in which corresponding text material is based. That is, a geographic map (e.g., a black-and-white two-dimensional cartographic map of the United States) may serve as an organizational framework for cognitively representing text as Mayer (2003), Kulhavy and Stock (1996), Schnotz et al. (2002) suggest, but it may be unnecessary when learners are already familiar with the geographic area the map represents. Likewise, navigating a website via directional buttons (e.g., buttons corresponding to the cardinal directions between locations—north, northeast, east, southeast, south,

556

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

southwest, etc.) based on spatial knowledge of the geographic area would likely interfere with learners’ image of the domain—if learners actually rely on the image for understanding textual elements. Thus, we varied the familiarity of locations on a navigable cartographic map of the United States, asking learners to learn as much information from the site as possible. The information on each page comprised unfamiliar facts about the geographic locations. We also varied the presence of the graphic, as well as the presence of the geographically based directional navigation buttons. 1.4. Theoretical predictions We predicted that if, as Schnotz et al. (2002) suggests, learners build an internal mental model, having aspects of both the graphic display and the text material, then learners should learn more from the text corresponding to locations on a graphic when the locations are familiar. Alternatively, both Schnotz et al. (2002) and Kulhavy and Stock (1996) suggest that graphics serve as organizational frameworks with which to encode, store, and retrieve text elements in memory. If this is true, then learners should learn more information from text when the text is accompanied by a graphic. Likewise, there should be no difference in text learning when geographic locations are either familiar or unfamiliar. Mayer et al. (1995), on the other hand, suggested that learners high in prior knowledge show meaningful learning whether graphics are present or not. If Mayer et al. are correct, then learning should be best when text information is associated with familiar geographic locations regardless of the presence or absence of the graphic; if low prior knowledge renders a graphic higher in utility for learning, then learning should be best when a graphic is present than when it is not, but only for unfamiliar geographic locations. Mayer (2003) explained that students who possess high levels of prior knowledge create their own mental images when processing propositions from text and use those images to build connections between verbal and visual representations. Again, if Mayer is correct, the presence of a graphic should interfere with learning from text, but only when learners use their own image of familiar locations to move from one familiar location to the next. In contrast, if as Mayer suggests, students who lack prior knowledge independently create useful mental images solely from verbal materials, then learning should be expected to be best when learners derive images from locations that are unfamiliar, regardless of the presence of the graphic. This prediction is not inconsistent with Schnotz et al. (2002). However, Kulhavy and Stock (1996) would add that images derived from unfamiliar locations should yield higher learning outcomes only when the images can be conjointly retained with a graphic that is present. Finally, Nilsson and Mayer (2002) found a trade off between navigation success and comprehension of site content when the navigational graphic provided more scaffolding for webpage locations and directions between pages. We contend that this should only be observed if cognitive competition occurs—competition between a learners’ pre-existing internal cognitive representation of the spatial relations among locations within the entire site, and the image of the site provided by the site graphic. We would expect the competition to occur only if learners have available to them, directional buttons to navigate their internal map, at the same time they have available an external graphic showing the same geographic area.

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

557

2. Method 2.1. Design Three factors, navigation by direction, navigational map, and geographic familiarity were combined to yield eight experimental cells. The resulting design was a 2 navigation by direction (present vs. absent) · 2 navigational map (present vs. absent) · 2 geographic familiarity (familiar vs. unfamiliar) fixed ANOVA, with geographic familiarity manipulated as a within-subjects variable. 2.2. Participants One hundred and ten undergraduate psychology students volunteered for participation from a midsize university in the western US, and were randomly assigned to one of the four between-subjects conditions as they entered the test site. The participants’ primary language was English, and they had no apparent sensory or motor challenges that would preclude their full participation in the investigation. Two thirds of the sample in each between-subject group was female, with the mean age for the combined groups equal to 20.67 years. 2.3. Materials Two websites were used in this experiment. The first, the experimental site, prompted its visitors to navigate and study the webpage contents trying to remember as much as they could. The website was composed of a map of the United States along with 24 pages consisting of textual material about locations in the US. The webpage material consisted of interesting facts associated with the US cities—half of the cities in the Western and half in the Eastern portion of the US. Each page was composed of an average of two idea units. Western locations were designated as familiar since participants were sampled from the west, while locations in the east were regarded as unfamiliar. However, in order to ensure that Western locations were familiar and Eastern locations were unfamiliar, three estimates of location familiarity were assessed and tested. Specifically, students were asked to place as many of the 24 locations as they knew on a blank US map containing only its outline. The protocols were measured for: (1) the total number of cities correctly recalled; (2) the total number of cities correctly located; and (3) an accuracy index— the proportion of the number of cities correctly located to the number correctly recalled on the map. Across all measures, learners revealed significantly higher knowledge for Western over Eastern cities. The data are also corroborated in an investigation by Schwartz et al. (2004), in which a sample of western learners, identical to the ones here, were compared to a group of matched learners sampled from Georgia. In the Schwartz et al. (2004) study, the western learners showed the same pattern of results obtained here, and the eastern learners showed the pattern in reverse. The navigational map was an achromatic cartographic relief map of the United States, 1008 · 1008 pixels, showing the contiguous 48 states. (See Fig. 1) When learners navigated the website via the map, city locations were hyperlinked to their corresponding text page. The text page consisted only of the city name and the city facts. (See Fig. 2) When navigation was made by direction without the map, navigation buttons were located on the

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

Fig. 1. Web-page with navigational map.

558

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

559

Fig. 2. Web-page with navigational map link.

webpage in the position of their cardinal point on a compass (e.g., north, northeast, east, southeast, south, southwest, west, and northwest) (See Fig. 3). The button, when pushed, moved the learner to the corresponding text page of the city geographically closest in direction to the city from which the learner made the selection. The second site contained the Configurational Knowledge Questionnaire (CKQ). The CKQ is a measure of students’ ability to estimate distance and direction between city pairs, in addition to their reaction time in making those estimates. Distance estimates are made in two ways—by metric and non-metric judgments. Metric estimates require learners to estimate, in miles, the distance between city pairs ‘‘as the crow flies’’; non-metric distance estimates are made by moving a line on a bar to convey the relative distance between city pairs given a visual key. The key depicts two intersecting lines designating the relative north–south and east–west distance of the US from border to border and coast to coast, respectively. (See Fig. 4). Directional estimates require learners to place a line from the center of a circle through the circumference in order to designate the direction from one city to another. (See Fig. 5). The CKQ estimates were assessed in order to ensure that

Fig. 3. Web-page with navigation by direction links.

560

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

North-South X East-West Dimensions of the United States

Lansing, MI

Little Rock AR

Example of the cognitive requirement necessary to make a nonmetric distance judgment between Lansing, MI and Little Rock, AR. Fig. 4. Example of a non-metric distance estimate on the CKQ.

Little Rock, AR Fig. 5. Example of a direction estimate on the CKQ.

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

561

learners did not inconsistently vary across experimental treatment conditions in their spatial knowledge of the US. 2.4. Procedure Participants were randomly assigned to each experimental condition and completed the study tasks in groups of approximately 20. The experimenter instructed students to turn their computer monitors on and read the introduction to the website as the experimenter read the same text aloud. Participants submitted their informed consent electronically. The following page on the site prompted participants to select the condition indicated on the card they had chosen. Each person was asked to read the condition instructions to him or herself and to look up when finished. The instructions explained that participants would have 25 min to explore the website and try to remember as much as they could. Participants then pressed the ‘‘begin’’ icon at the bottom of their screens. When the time elapsed, participants were told to turn off their computer monitors, and complete a 1-min interpolated math task. Following the task, they had 20 min to recall on paper everything they remembered from the website. Participants were then instructed to return all testing materials to manila folders and given a 10-min break. Following a 10-min break, participants engaged in the tasks of the second website. The site was designed to measure configurational knowledge of the United States as measured by participants’ distance and direction estimates between city pairs. Participants read the instructions on the computer monitor as the experimenter read them aloud. Participants were encouraged to work at their own pace and when finished were debriefed, issued extra credit slips, and excused. 3. Results The number of idea units recalled were scored from the free recall protocols and entered into an ANOVA of the basic experimental design. An idea unit was defined as a single unit of meaning present in a sentence, clause, or phrase, and was scored correct if it’s presence in the recall protocol matched its corresponding unit of meaning in the text. In order to ensure the reliability of the idea units, two raters each independently scored 20 percent of the total protocols, resulting in an inter-rater reliability of 100%. All statistical tests were evaluated for significance at an alpha level below .05. As a first step in the analyses, it was essential to ensure that variation in webpage recall between the experimental conditions was not due to differences in spatial knowledge of the US. Thus, the three CKQ estimates (metric and non-metric distance estimates, and estimates of direction) were entered into a 2 navigation by direction (present vs. absent) · 2 navigational map (present vs. absent) · 2 geographic familiarity (familiar vs. unfamiliar) MANOVA. The analysis failed to show significant multivariate and univariate variation for all main effects and the interaction. In addition, all three estimates from the CKQ showed no relationship with the number of idea units learners recalled from the website. Thus, the findings ruled out the potential of effects being due to the co-variation of learners’ familiarity with the spatial domain used in this investigation. Next, we expected significantly more idea units to be remembered from the text when they were associated with graphic locations that were familiar than unfamiliar. The results supported this expectation. Significantly more idea units were recalled when they were

562

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

associated with cities in the Western US (M = 20.31; SD = 8.50) than when they were associated with cities from the East (M = 13.58; SD = 5.61), [F(1, 51) = 65.02, p = .000, g2 = .56]. In the CKQ analysis, learners showed no differences in their configurational knowledge of the US. However, learners found it easier to remember information about cities more familiar to them. Thus, learners seem to make use of the geographic knowledge they already possess in order to learn information related to the graphic of a familiar domain. Even though the US was familiar to all learners, information relative to cities farther away was more difficult for them to process. The size of this effect was large, yielding a Cohen’s d statistic = .94 and revealing that less than 48% of the recall distribution for Western cities overlapped with the distribution of recall for cities from the East. Next, we expected learners to use the geographic map for remembering information about geographic locations unfamiliar to them, but not for geographic locations that were familiar. The geographic familiarity · navigational map interaction supported this expectation, F(1, 51) = 4.40, p = .04, g2 = .08. When learners navigated the website using the navigational map, the map interfered with the number of idea units recalled from familiar cities (M = 19.48; SD = 8.17) compared to the condition when the map was absent (M = 21.11; SD = 8.88)—although the size of the effect was small (Cohen’s d = .19, effect percentile = 57). The opposite was true for learner’s memory of idea units associated with geographic locations farther away—those that were unfamiliar. When the map was present, the effect was comparatively stronger (Cohen’s d = .49, effect percentile = 68), with learners remembering more idea units (M = 14.52; SD = 5.20), than when it was unavailable (M = 12.68; SD = 5.93).

Fig. 6. Mean number of idea units correctly recalled for familiar cities as a function of navigational map and navigation by directions.

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

563

The finding suggests that learners may rely on an internal cognitive representation of the graphic when navigating a website (if the cities are familiar). When the cities are unfamiliar, the presence of an external spatial display is a beneficial adjunct for learning to occur. The differentiation is particularly salient when the factors of navigation by direction and navigational map are considered for familiar and unfamiliar cities separately. That is, when the cities were familiar, learners’ performance was roughly equivalent when the directional links were absent and the navigational map was available or not (direction absent/map absent M = 20.50, SD = 8.20; direction absent/map present M = 21.79, SD = 8.78), in addition to the condition where direction was present without the map (M = 21.71; SD = 9.79). However, when the buttons allowing navigation by direction were presented in conjunction with the navigational map, comprehension of the text was substantially impaired. (M = 17.00; SD = 6.98). The effect size for all three comparisons yielded a Cohen’s d statistic equal to .46, .60, and .55, respectively. Thus, the distribution of the group’s performance on text comprehension, when navigating by the buttons in the presence of the map, was exceeded in performance by approximately 33–38 percent of the distributions of the other three groups—a moderate size effect. (The means are plotted in Fig. 6). On the other hand, the importance of the navigational map was particularly influential for textual comprehension when the cities were unfamiliar. That is, learners navigating neither by direction nor map (M = 11.71; SD = 6.02) were only slightly lower in their performance to learners receiving both (M = 12.46; SD = 5.64). In addition, the presence of navigational directions alone did not substantially improve learners performance

Fig. 7. Mean number of idea units correctly recalled for unfamiliar cities as a function of navigational map and navigation by directions.

564

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

(M = 13.64; SD = 5.89). However, when learners received the navigation map only, their comprehension of textual elements was significantly improved (M = 16.43; SD = 4.07). The effect size for the comparison of learners in the latter group with the other three yielded a Cohen’s d statistic equal to .92, .81, and .55, respectively. Thus, the distribution of the group’s performance on text comprehension, when navigating only with the map, exceeded the distributions of performance of the other three groups by values ranging between approximately 35–52 percent. (See Fig. 7 for a graphic display of the means). Finally, it is important to note that the findings above could have been influenced by the number of pages learners visited, if the learners were unable to view all geographic location pages in the website. The finding also could have been influenced by time on page, if the learners spent more time on some pages over others. In order to control for the potential of these influences, the number of pages visited and the time spent on those pages were captured by the server, time-stamped, and evaluated. Examination of both variables revealed that all learners in all conditions visited all pages comprising the site, and spent an equivalent amount of time on each of those pages. 4. Discussion We sought to determine whether learners navigating a website use a cognitive model integrating textual elements with the spatial configuration of the site, or an externally derived image of the site to encode the text-based information contained there. Results of the investigation suggest that learners use both, depending upon their familiarity with the features on the graphic. We suggested that navigational maps may serve as an organizational framework for cognitively representing text, but the map may be unnecessary when learners are already familiar with the spatial domain the navigational map represents. We were guided in our prediction by Mayer et al. (1995) who suggested that students possessing high levels of prior knowledge create their own mental images when processing propositions from text—images from which connections are built between both verbal and visual representations. The data we obtained for text comprehension of familiar and unfamiliar cities, under conditions without navigational buttons or a map, lend support for Mayer et al.’s position. That is, when learners did not have the benefit of the map, they remembered significantly more information for familiar cities over unfamiliar ones. However, if learners derived their own image of the geographic domain, there should have been an interaction between navigation by direction, navigational map and geographic familiarity. Specifically, learners navigating by buttons should have comprehended significantly more idea units derived from familiar cities in the absence of the map—at least if the graphic– text connections explained by Mayer (2003, 1997) apply to navigation. The finding failed to emerge. Thus, graphics used for purposes of navigation appear to function quite differently then graphics used as adjuncts to text. In the present investigation, it is important to note that familiarity of cities was quite different from familiarity of the content in text. In all experimental conditions, text content was unfamiliar to learners. Mayer and his colleagues (e.g., Mayer & Gallini, 1990; Mayer et al., 1995) found that learning performance was significantly better when graphics were presented with text when learners’ prior knowledge was low, but undifferentiated by the presence of graphics when prior knowledge was high. The data obtained in our investigation support the former, but do not address the latter.

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

565

Nilsson and Mayer (2002) suggest that the use of graphics for navigational purposes comes at a cost. When learners are afforded graphic information about the navigational structure of a site, they fail to show performance increases in content comprehension relative to learners who receive a graphic void of the site’s spatial structure. That is, learners receiving a graphic intended to facilitate navigation perform better at navigating the site, but perform more poorly on their comprehension of site content. The data from the present investigation corroborate this finding, but add that navigational maps may interfere with learners’ representation of the navigational graphic, because learners may use their pre-existing cognitive representation of it, as well. All learners in the present investigation were familiar with the graphic of the United States since the US map is well known among American students. However, the cities on the graphic were presumably better represented within the familiar portion of the map, than the portion that was unfamiliar. Thus, we suggest that learners acting on the familiar cities of the map showed decrements in their performance because the graphic competed for working memory space with the representation they constructed from prior knowledge. Thus, graphics used as adjuncts to text seem to be functionally and cognitively different than graphics used for navigation. The two representations do not appear to be the same since they appear to compete for working memory space in both the familiar and the unfamiliar learning conditions. Indeed, learners were compromised in their comprehension of text elements when the elements were associated with familiar cities, but enhanced when the text was associated with cities that were unfamiliar. The data are consistent with the findings reported by Scott and Schwartz (2007). The comprehension of their learners was reduced but only when the learners attempted to use the navigational map for both text comprehension and navigation purposes. One of the questions we considered was how the graphic is represented in working memory when it is used for purposes of navigation in contrast to the comprehension of text. When considering graphic–text comprehension alone, we suggested that the model proposed by Schnotz et al. (2002) would be the most robust because it accounts for graphics of all types, even when the graphics are familiar and subject-matter prior knowledge is either high or low. Alternatively, we suggested that Kulhavy and Stock (1996) model was restricted because it addresses only graphics that are unfamiliar to learners. Kulhavy and Stock (1996) contended that there is a mapping of text propositions onto the structure of the graphic, and that it is the graphic’s image that guides comprehension and recall—at least when prior knowledge of the graphic and text are low. In Schnotz et al.’s model, when prior knowledge is high, both text and graphics are evaluated using the image of the graphic, but when low, evaluation follows a process of construction derived from the graphic. Thus, Schnotz et al. (2002) model is consistent with the model of Kulhavy and Stock (1996), but only under conditions when prior knowledge is low. Our results support both the models of Schnotz and Kulhavy but only under this condition. Learners remembered significantly more text elements in the presence of the map when cities were unfamiliar. However, Schnotz’ et al.’s model was not supported when learner’s prior knowledge of graphic locations was high. That is, learners failed to differ in their comprehension of text associated with familiar cities whether they received a map or not. While the graphic may function as an image, it appears to do so only when the familiarity of graphic elements is low.

566

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

When considering graphics used as navigational maps, the text–graphic model of Schnotz may provide a heuristic. Schnotz et al. (2002) contend that when domain knowledge already exists in a learner, the learner’s existing mental model is activated and serves to evaluate, through a top-down analogical mapping approach, similarities between the model and the text and graphic. It is highly likely that this occurred when learners were processing the familiar portion of the navigational map. Learners navigating the website by buttons without the benefit of a navigational map were significantly better in their comprehension of text when the text elements were associated with familiar over unfamiliar cities. Alternatively, navigation by buttons in the presence of the map yielded low text comprehension for text elements associated with both familiar and unfamiliar cities. Thus, if as Schnotz et al. (2002) suggests, learners access their own internal mental model of the map, the evaluation process may be responsible for the interference between the image they bring to the task and the image provided by the navigational map. 4.1. Implications for education Organized spatial displays, such as geographic maps, and texts that are related to them, are integral parts of most of the curricular materials used in schools today. Maps are routinely used to teach learners about historical events and trends, geographic characteristics of regions, and differences among people and places around the world. Maps are found in textbooks, on classroom walls, in television news, as well as content pages on the web. In short, geographic maps are the materials of choice under most instructional conditions where learners are required to learn about history, geography, and social studies in school. Indeed, the importance of technology and the need for its incorporation with the materials found in classrooms is clearly apparent. Today’s student is often asked to search the Internet while doing research for a term paper, when doing a web quest, or when creating an online portfolio to present their work. As more of these types of assignments and projects are being assigned to students, the more teachers, administrators and designers of curricular material must be aware how students are using them to acquire information when used alone or in conjunction with more traditional materials. The study presented here provides an early test of how web-based and traditional geographic materials operate together to improve learning for the students who use them. McCurley (2001) contended that geography plays a vital role in linking webpages on the Internet, pointing out that information retrieval based on geographic criteria is a very common task. For example, geographically linked webpages are widely used for ‘‘students who are studying another part of the world, intelligence analysts preparing a report on a given area, business planners investigating an area for expansion, or government planners who wish to focus on environmental or economic impact in a specific region’’ (p. 221). Larson (1996) also argued that geographic navigation structures (e.g., cartographic maps) provide a very sensible and effective way to link library resource material worldwide. The research presented here reveals that learners can and do use structures such as these; however, the efficacy of these structures is mediated by the knowledge learners bring with them to the task—knowledge related either to the information they seek from the domain, or knowledge they possess of the spatial organization of the information from that domain. This knowledge may be situated either in the content of the pages from which the site is built, or in the graphic display that is used to navigate the site. When in the pages, prior knowledge of the domain helps; when in the graphic, however, the

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

567

knowledge learners bring with them leads to a fundamentally more complicated condition of learning—at least, according to our findings, when the subject domain is geographic. The deployment of personal geographic knowledge of a relevant geographic domain competes for working memory space with the information afforded by the geographic map used for navigation. When geographic knowledge is low, interference borne from working memory competition is low; when it is high, interference is high. Thus, the situation in which learning is compromised, is the situation in which competition occurs in working memory for space from a learner’s pre-existing representation (e.g., domain knowledge of geography) and the information organized geographically throughout the site. This finding has direct implications for teaching. As classrooms become more and more diverse, the knowledge students bring to their learning tasks will be as varied as the students themselves. Thus, the nature of that knowledge will impact quite differently the learning students can accomplish when they navigate through a site. Therefore, teachers must obtain some measure of the type of knowledge their students possess, before the students attempt to learn from the web environments the teachers build or assign for this purpose. For example, students, whose background is the same as a foreign country or culture for which web-based study is assigned, will likely be disadvantaged when the website is organized by a map of the familiar geographic area. Thus, our data suggest that the function of maps appear to be quite different when they are used as adjuncts to text vs. website navigation. That is, learners appear to process adjunct maps to text quite differently than they do when using maps as guides to navigate between pages on the web. When maps are combined for both purposes (as they may be in teaching social studies on the web, for example), the findings revealed here suggest that instructional designers should be especially mindful of these differences—particularly in light of the geographic mental representations learners bring with them to the learning task. References Brusilovsky, P., & Rizzo, R. (2002). Using maps and landmarks for navigation between closed and open corpus hyperspace in web-based education. The New Review of Hypermedia and Multimedia, 9, 59–82. Eveland, W., & Dunwoody, S. (2000). Examining information processing on the WWW using think aloud protocols. Media Psychology, 2(3), 219–244. Griffin, M. M., & Robinson, D. H. (2005). Does spatial or visual information in maps facilitate text recall? Reconsidering the conjoint retention hypothesis. Educational Technology Research and Development, 53(1), 23–36. Kulhavy, R. W., Lee, J. B., & Caterino, L. C. (1985). Conjoint retention of maps and related discourse. Contemporary Educational Psychology, 10, 28–37. Kulhavy, R. W., & Stock, W. A. (1996). How cognitive maps are learned and remembered. Annals of the Association of American Geographers, 86(1), 123–145. Larson, R. R. (1996). Geographic Information Retrieval and Spatial Browsing. In L. Smith & M. Gluck (Eds.), University of Illinois (pp. 81–124). NewYork: GIS and Libraries: Patrons, Maps and Spatial Information. Mayer, R. E. (1997). Multimedia learning: Are we asking the right questions? Educational Psychologist, 32(1), 1–19. Mayer, R. E. (2003). The promise of multimedia learning: Using the same instructional design methods across different media. Learning and Instruction, 13, 125–139. Mayer, R. E., & Gallini, J. (1990). When is an illustration worth ten thousand words? Journal of Educational Psychology, 82, 715–726. Mayer, R. E., Steinhoff, K., Bower, G., & Mars, R. (1995). A generative theory of textbook design: Using annotated illustrations to foster meaningful learning of science text. Educational Technology Research and Development, 43, 31–43.

568

N.H. Schwartz et al. / Contemporary Educational Psychology 32 (2007) 551–568

McCurley, K. S. (2001). Geospatial Mapping and Navigation of the web. In Proceedings of the 10th International World Wide Web Conference, Hong Kong (pp. 221–229). McDonald, S., & Stevenson, R. J. (1998). Navigation in hyperspace: An evaluation of the effects of navigational tools and subject matter expertise on browsing and information retrieval in hypertext. Interacting with Computers, 10, 129–142. Nilsson, R. M., & Mayer, R. E. (2002). The effects of graphics organizers giving cues to the structure of a hypertext document on users’ navigation strategies and performance. International Journal of HumanComputer Studies, 57, 1–26. Paas, F., Renkl, A., & Sweller, J. (2003). Cognitive load theory and instructional design: Recent developments. Educational Psychologist, 38(1), 1–4. Paivio, A. (1986). Mental representations: A dual coding approach. New York: Oxford University Press. Robinson, D. H., Katayama, A. D., & Fan, A. (1996). Evidence for conjoint retention of information encoded from spatial adjunct displays. Contemporary Educational Psychology, 21(3), 221–239. Schnotz, W. (1993). Some remarks on the commentary on the relation of dual coding and mental models in graphics comprehension. Learning and Instruction, 3, 247–249. Schnotz, W. (2002). Towards an integrated view of learning from text and visual displays. Educational Psychology Review, 14(1), 101–120. Schnotz, W., Bannert, M., & Seufert, T. (2002). Toward an integrative view of text and picture comprehension: Visualization effects on the construction of mental models. In J. Otero & J. A. Leon (Eds.), Psychology of science text comprehension (pp. 385–416). Mahwah, NJ: Lawrence Erlbaum Associates. Schwartz, N. H. (1997). Human information processing of maps: A report to the geographic community. In R. G. Boehm & J. F. Petersen (Eds.), First assessment: Research in geographic education. San Marcos, TX: Gilbert M. Grosvenor Center for Geographic Education Press. Schwartz, N . H., Andersen, C., Hong, N., Howard, B., & McGee, S. (2004). The influence of metacognitive skills on learner’s memory of information in a hypermedia environment. Journal of Educational Computing Research, 31(1), 77–93. Schwartz, N. H., Verdi, M., Larson, N., Rittschof, K. (2004). The role of configurational knowledge in learning from maps and text. In Paper presented at the annual meeting of the American Educational Research Association, San Diego, CA. Scott, B. M., & Schwartz, N. H. (2007). Navigational spatial displays The role of metacognition as cognitive load. Learning and Instruction, 16, 89–105. Verdi, M. P., Crooks, S. M., & White, D. A. (2003). Learning effects of print and digital geographic maps. Journal of Research on Technology in Education, 35(2), 290–302. Verdi, M. P., & Kulhavy, R. W. (2002). Learning with maps and texts: An overview. Educational Psychology Review, 14(1), 27–47.