Examining Adolescents’ Strategic Processing During Online Reading With a Question-Generating Task

Previous Research

An increasing, albeit still small, number of literacy studies have examined in detail the nature of online reading strategies. A few studies have explored reading strategy type, analyzing readers’ verbal reports on their thinking and behaviors during online reading. For example, Coiro and Dobler (2007) examined 11 sixth-grade skilled readers’ online reading strategies as they responded to brief-response questions on a science topic. The analysis of the readers’ think-aloud protocols described a cyclical process of prediction, evaluation, and monitoring at a metacognitive level. Coiro and Dobler claimed that online reading may involve additional complexities, compared with print-based reading, due to the need to make multilayered inferences about links and texts prior to reading them closely. The study contributed to a new literacies debate on similarities and differences between print reading and online reading. However, the study’s results must be carefully interpreted, for its task design prevented a comparison of online reading with an equivalent case of print reading. The study also is limited in that it lacks explanation of how the complexities of strategic processing are related to the learning that the readers achieved through online reading.

Later, Cho (2014) investigated seven competent high school readers performing a critical online reading task. In this study, readers chose among a list of controversial topics (e.g., alternative energy, eco-friendly industry, the death penalty) and used the Internet to seek informational sources that would be useful in generating an inquiry question that they believed to be worthy of framing a classroom debate. The analysis of the readers’ verbal protocols resulted in an account of core reading strategies, including information location, meaning making, source evaluation, and self-monitoring. Further, quantification of the coded protocols indicated the importance of monitoring and evaluation in online reading, as had been described by Coiro and Dobler (2007). The analysis also indicated that in an unconstrained search context, information-locating strategies were used more extensively than meaning-making strategies; whereas in reading a finite set of websites, the tendency was shifted toward processing text information more deeply rather than continuing to find additional sources. The results supported the premise that a reader’s strategic patterns are bound to, and shaped by, the textual space that the reader identifies and specifies online. However, it is unclear in the study whether these strategic processes were positively associated with the quality of the questions that the readers generated using what they learned from online reading.

More recently, DeSchryver (2015) studied online reading strategies with a special focus on information (and knowledge) synthesis. In this case study, eight proficient Web users from a pool of doctoral and law students used the Internet to learn about people’s personal decisions related to climate change issues, while verbalizing their thinking processes. In the qualitative analysis, several integrative strategies were found to be involved in online reading, including locating information, refocusing and reinforcing evolving understandings, and creating knowledge. A conclusion was that a process of generative synthesis, when flexibly used online, is central to creative knowledge construction because it allows readers to revisit, update, and reconstruct their knowledge as it evolves during reading. The results can be interpreted as indicating that online readers’ moment-by-moment actions (e.g., website visits) must be coordinated with effective location and integration of information in a goal-directed way. Yet, as in other studies (Cho, 2014; Coiro & Dobler, 2007), DeSchryver’s study focused heavily on identifying synthesis strategies, but not on the contribution of these strategies to the knowledge and understanding that the readers gained anew or constructed from online sources.

On the other hand, some research has investigated process–product relations in online reading by adding outcome measures to analyses of strategic processing inferred from readers’ verbal reports. A notable study was done by Anmarkrud, Bråten, and Strømsø (2014). These researchers observed college learners reading within a Google-like environment that connected a preselected set of conflicting sources about cell phone use and possible health risks. The readers’ verbal reports and written essays made it evident that intertextual linking, source evaluation, and monitoring were all important to the comprehension of multiple digital sources. That is, the active use of sourcing strategies, in conjunction with intertextual linking strategies, was positively correlated with a better understanding of topic-relevant arguments. Along the same lines, Goldman, Braasch, Wiley, Graesser, and Brodowinska’s (2012) study observed college readers learning from a finite set of digital sources with the goal of analyzing the factors that cause volcano eruptions. Goldman and colleagues described these learners’ central strategies, including navigation, intertext connection, monitoring, and information evaluation, which are similar to those identified by Anmarkrud et al. (2014) and Cho (2014). It is worth noting that their comparative analysis demonstrated that the readers’ online processing skills were positively correlated with their knowledge gains and conceptual understandings upon completion of reading. In particular, proficient source evaluation was key to productive learning: Online readers who learned better habitually attempted to question information qualities and opportunistically find evidence to support their textual judgments, whereas their less productive counterparts tended to accept or reject sources outright without critical judgment.

The studies by Anmarkrud et al. (2014) and Goldman et al. (2012) offered evidence that an effective use of core strategies, including searching, sourcing, sense making, and monitoring, may enhance content learning in an online setting. The studies also demonstrated notable differences between proficient and nonproficient online readers’ processing of multiple digital sources, noting that such differences play a crucial role in determining the quality of learning. Nonetheless, these studies are limited with respect to the ecological validity of the tasks. The researcher-developed Google-like interface constrained reading to a predetermined, finite set of sources; the absence of such a constraint might have afforded observations of more nuanced, complex, or extensive uses of online reading strategies. In addition, the previous studies in general tended to constrain the outcome to one that could be measured with multiple-choice items, rather than looking for generative outcomes as evidence of the evolving knowledge that readers construct while navigating open Internet environments, identifing topic-related problems, and determining a scope of texts to read. For those studies that involved an essay, this constructed outcome facilitated an opportunity for readers to reflect on and engage in their learning and demonstrate meaning made in response to a researcher-created question, which is a useful approach especially in a school context. However, this approach can result in a constrained response due to the question being given to the reader, rather than the reader seeking to explore and identify information to develop her or his own question. A more generative reading task, such as identifying new problems and creating compelling inquiry questions, in an open textual environment may elicit the use of sophisticated online reading strategies beyond merely gathering and processing information to answer predetermined questions.

It is, therefore, critical to fill the gap between what the previous research has demonstrated and aspects of online reading that demand attention if we are to better understand students’ online reading strategy use and their learning with digital sources. Building on the previous studies, the present study examines high school students’ online reading performance, refining our understanding of online reading strategies and redefining the outcome of such strategy uses. The study builds a hypothetical model that explains the interplay of multiple reading strategies in an open Internet search environment. It then observes the workings of the strategies toward higher order learning, using a generative reading task intended to encourage readers to identify useful sources, construct meaning from cross-text ideas, and use the meaning to generate a compelling question that has the potential to open up new opportunities to learn.

Conceptual Framework

Online reading abounds with moment-by-moment situations of determining a manageable boundary of reading within which multiple links, texts, and information are selected and understood (Madrid, Van Oostendorp, & Melguizo, 2009; Protopsalitis, 2008; Salmeron, Canas, Kintsch, & Fajardo, 2005; Salmeron & Garcia, 2011; Sullivan & Puntambekar, 2015; Zammit, 2011). Previous research has documented multiple types of reading strategies that prominently feature in complex online reading (e.g., Anmarkrud et al., 2014; Cho, 2014; Coiro & Dobler, 2007; Goldman et al., 2012). These strategies can be generally grouped into information location, meaning making, source evaluation, and self-monitoring (Afflerbach & Cho, 2009). The core strategies interact with one another when a reader identifies and comprehends digital sources, as described in the following subsections.

Question-Generating Task

As research suggests that question generation is an important tool for improving text comprehension (Palincsar & Brown, 1984; Pressley & Afflerbach, 1995), strategic online reading may also be fostered if students are allowed to explore, examine, and manage information spaces with their own evolving questions (Leu et al., 2011). Question generation may encourage readers to get involved in critical thinking, such as hypothesis building and analysis of diverse online sources. When the generated question is expected to be relevant to a problem or issue that allows argument and debate, online readers must be active not just in finding general information to complement their prior knowledge, but in distilling emerging themes that are important to their understanding from multiple sources. Question generation can also help readers be more metacognitive in learning from online sources. With question generation as the goal, online reading is more likely to turn into a process of formulating, defining, and refining a compelling question that engages readers in the continual monitoring of how they learn and in strategic processes for choosing and processing potentially useful sources for their learning (Cho, 2014). Further, as research supports, question generation may motivate readers to impose coherence on sources and make high-level inferences from text ideas (Beck, McKeown, Sandora, Kucan, & Worthy, 1996; Raphael & Pearson, 1985). Therefore, to generate and elaborate a question that is critical and compelling to examine a certain problem or issue, readers in an online setting must be able to effectively use reading strategies—information location, meaning making, source evaluation, and self-monitoring—as mechanisms for building a coherent understanding from across multiple sources of information.

Built upon the theories and research synthesized above, Figure 1 presents a model of strategic online reading that depicts both the direct and indirect (or mediated) effects of strategic processes on reading outcome situated within a question-generating task. Above all, the model describes how question generation (i.e., reading outcome) in online reading can be explained by the individual and/or joint functioning of the four strategic processes. It is worth noting that in this association, meaning making plays a mediating role in the direct and indirect effects of self-monitoring, information location, and source evaluation conducted for the purpose of question generation. Source evaluation also has a mediating role that connects self-monitoring with information location and with meaning making. Furthermore, the model describes how self-monitoring as readers’ executive control can explain the strategic use of meaning making, information location, and source evaluation processes.

OPEN IN VIEWER

Figure 1.

Model of strategic processes during online reading within a question-generating task.

The Present Study

The present study uses the model of strategic online reading illustrated in Figure 1 as the point of reference for an analysis of high school readers’ strategic engagement in online reading. These readers were asked to generate a compelling question on a controversial topic, based on the information and knowledge gained from the online sources they located, accessed, and comprehended. The readers’ self-generated questions are incorporated into the analysis of strategic processes informed by their concurrent verbal reports during reading. Two research questions drive the study: (a) How do high school readers use the core reading strategies (i.e., information location, meaning making, source evaluation, self-monitoring) considered in the model of strategic online reading? (b) How do their strategic processes contribute to the qualities of their generated questions as an important outcome of online reading? With these questions, this study explores an integrative understanding of how online reading strategies operate interactively and the different degree of contribution that each of the strategies makes toward the construction of meaning in a multisource text environment.

Methods

Data Sources and Analysis Techniques

Verbal reports synchronized with screen recordings

Participants’ reading strategy use was assessed by analyzing the recorded Camtasia videos, which included their verbal reports and computer screen activities. A scoring rubric to evaluate the participants’ online reading strategy use was developed for the analysis, built upon the compendium of online reading strategies identified in previous studies (Afflerbach & Cho, 2009; Cho, 2014). The rubric comprised the four core strategy types (i.e., informaton evaluation, meaning making, source evaluation, self-monitoring), with each decomposed into four actions in the following manner:

Information location includes (a) applying search terms, (b) conducting information searches, (c) selecting relevant links, and (d) rejecting irrelevant links.

Meaning making includes (a) understanding important information, (b) building intertextual relationships, (c) elaborating a metalevel understanding, and (d) conducting further reading.

Source evaluation includes (a) identifying the author, (b) determining relevant sources, (c) discerning reliable sources, and (d) assessing each source’s significance to reading.

Self-monitoring includes (a) managing information searches, (b) regulating link selections, (c) deciding to stop or go further, and (d) adjusting meaning-making processes.

Furthermore, the rubric included the item “overall effectiveness” for each strategy type, in order to help the raters maintain attention to the big picture of each participant’s online reading performance as well as the isolated processes.

Four raters formed four pair groups for scoring with each rater participating in two different pair groups. Each pair watched a set of 10 participants’ recorded videos and scored the qualities of their strategic performance, which reflected each of the four strategy types. On the scoring rubric, each action—as listed (a), (b), (c), and (d) for each of the above-described strategy types—was worth 2 points. One point was given when it was evident in the verbal reports that the reader was engaged in planning and executing a particular strategic action at any point during reading; 1 additional point was given when a frequent use of the strategic action was noticed. Overall effectiveness of strategy use was worth 2 points per strategy type (i.e., information location, meaning making, source evaluation, self-monitoring); these 2 points were given when the raters judged that the particular strategy type used did indeed contribute to the participant’s online reading performance. To facilitate this scoring process, the researchers constructed a guiding scheme that included representative verbal reports indicating particular reader actions along with interpretation of the underlying processes. Finally, the rubric-based analysis of participants’ videos yielded four manifest variable scores corresponding to the four strategy types (maximum = 10 points per variable).

Written questions and justifications

The participant-generated written questions with their justification statements were analyzed and scored on an analytical rubric (Figure 2). The rubric was organized with three criteria (relevance, validity, and significance) and four quality descriptions (complete: 3, adequate: 2, partial: 1, lacking: 0). The criterion of relevance addressed how closely a written response was connected to the topic and the Internet sources. Complete relevance of a written response was expected to contain various topic-related comments, to not wander off the point, and to include evidence or notes on the sources used for the question generation. The criterion of validity assessed logical soundness. Complete validity required a close argumentative relation between the critical question and the supporting details provided in the justification statement. Finally, the criterion of significance addressed how well a critical question would promote critical thinking, provoke different perspectives, and initiate extended discussion related to the topic. Complete significance was achieved when a written response pinpointed important aspects of the problem or uncovered hidden perspectives and assumptions held by the stakeholders, with specific details to support the points. The scores that each participant earned for the three criteria were aggregated into one composite score (maximum = 9 points) for the statistical analysis.

OPEN IN VIEWER

Figure 2.

Rubric-based scoring of a participant’s written question with justification

Interrater reliability

Consistency among the four raters was established in the analysis of the participants’ online reading performance. Following a general assessment model (Pellegrino, Chudowsky, & Glaser, 2001), the raters initially had a series of in-depth discussions of the constructs to be measured; methods for observing the constructs, that is, extratextual factors related to both the online reading task and the verbal reporting; and data interpretation processes, including the scoring rubrics. Then, for practice, the raters sampled one participant’s (P001) recorded video and written response. The raters watched the video clip together and discussed the content of the verbal reports and screen behaviors. They then coevaluated the underlying strategic processes using the strategy use scoring rubric. Right after the video evaluation, the raters shared their evaluations of the participant-generated question and rationale statement, according to the scoring rubric, and discussed internal features of the written response and any evidence that implied an active use of sources.

An interrater reliability test was then performed to determine consistency among raters. Each rater was assigned to evaluate two randomly chosen participants’ (P021 and P043) performances using the scoring rubrics. The interrater reliability for strategic performance was measured with the intraclass correlation coefficient (ICC; McGraw & Wong, 1996) to test the degree to which the raters evaluated participants’ use of multiple strategies consistently. The resulting coefficient was in the excellent range (r = .878), indicating that the selected participants’ strategic performances were rated similarly across the raters. Regarding the written responses, the ICC for the four raters was in the excellent range (r = .861), indicating that the four raters reached substantive agreement, and the sampled participants’ critical questioning was rated similarly across the raters. A follow-up discussion was conducted to address inconsistent scorings, as an effort to ensure that the raters shared a common ground for their judgments.

Results

Descriptive Statistics and Correlation Analysis

The descriptive statistics for all measured variables show general tendencies in online reading performance among the participants (N = 43). Table 1 presents the mean and standard deviation scores for participants’ topic-related prior knowledge, the four types of online reading strategies, and question generation as the reading outcome. As Table 1 shows, participants possessed a low level of prior knowledge about MTR, which was expected, due to the novelty of the task topic. The quality of the participants’ performances varied according to the type of strategic processing, although they generally scored in the middle to lower middle range on the measures of both the reading strategies and the reading outcome. Relatively lower performance on self-monitoring and source evaluation was observed, compared with relatively higher performance on information location. This variation in aspects of their online reading performance can be interpreted as indicating that the different strategic processes might have different degrees of complexity and make different cognitive demands.

OPEN IN VIEWER

Table 1

Summary Statistics for Measured Prior Knowledge, Online Reading Strategies, and Reading Outcome in This Study (N = 43)

Variable	Mean	Standard Deviation
Topic-related prior knowledge (max = 30)	9.49	2.81
Strategy 1: Information location (max = 10)	6.58	1.89
Strategy 2: Meaning making (max = 10)	5.51	1.83
Strategy 3: Source evaluation (max = 10)	4.02	2.06
Strategy 4: Self-monitoring (max = 10)	3.65	1.93
Outcome: Question generation (max = 9)	5.14	1.98

Substantive bivariate correlations were observed between each strategic process and reading outcome. As shown in Table 2, this tendency was maintained after controlling for a topic-related prior knowledge effect on online reading performance. Question generation as the outcome variable was statistically significantly correlated with the strategy variables of meaning making, source evaluation, and self-monitoring. For information location, its relatively moderate correlation with question generation may reflect to some degree a weaker relationship of information location to reading outcome compared to the other strategies.

OPEN IN VIEWER

Table 2

Bivariate Correlations Between Information Location, Meaning Making, Source Evaluation, Self-Monitoring, and Question Generation

Variable	1	2	3	4	5
1. Information location	—	.557***	.723***	.404**	.322*
2. Meaning making	.544***	—	.660***	.607***	.592***
3. Source evaluation	.728***	.652***	—	.542***	.485**
4. Self-monitoring	.403**	.606***	.541***	—	.530***
5. Question generation	.327*	.590***	.488**	.531**	—

Note. Correlation coefficients (Pearson) are presented below the diagonal, and partial correlations after controlling for topic-related prior knowledge are presented above the diagonal.

*

p < .05. **p < .01. ***p < .001.

Path Analysis

The compressed associations were further examined in a path analysis with the five manifest variables considered in the model: four strategies (i.e., information location, meaning making, source evaluaton, self-monitoring) and one outcome (i.e., question generation). The Shapiro-Wilk test results for the normality of each measured variable indicated evidence of deviations from multinormality. Accordingly, the path analysis employed maximum likelihood parameter estimates with standard errors and a chi-square test statistic, which are robust to nonnormality; it was performed using Mplus 7.3.

The path diagram, shown in Figure 3, depicts the structure of the strategic patterns in which the participating readers engaged during the task of online reading for question generation. There was no statistically significant difference between the theoretical and observed model of strategic online reading: χ² (1, N = 43) = .576, p = .448, comparative fit index (CFI) = 1.000, root mean square error of approximation (RMSEA) = .000 (90% confidence interval [CI]: .000, .365), square root mean of the residual (SRMR) = .013. Despite the limited interpretation of the model-fit indices with the nearly saturated model in this study, this result suggests that the tested path model may offer an adequate description of the participants’ online reading strategy use and how strategic reading explains the quality of their generated question as the reading outcome.

OPEN IN VIEWER

Figure 3.

Path analysis of the four strategies (information location, meaning making, text evaluation, and self-monitoring) and the outcome (critical questioning) in the online reading of the participating adolescent readers.

Specifically, multiple inferences can be made with regard to the strategic relations based upon the observed direct effects. First of all, there was a statistically significant prediction of source evaluation by self-monitoring (R² = .293, p = .010); that is, self-monitoring was a significant positive predictor of source evaluation (β = .541, z = 5.140, p < .001). For example, an excerpt from P002’s verbal report demonstrated how self-monitoring of his attention to multiple levels and dimensions of text contributed to the construction of criteria for effectively evaluating sources. When viewing a discussion of MTR on the Sierra Club website (www.sierraclub.org), he engaged with the authors and questioned authorial intent and bias:

However, they word it in a way that seems to generalize and when you do that or when I see that it’s almost … it’s not credible because sure you can say that … but there’s not the information for the viewer to judge whether it’s true or not … I think a site should provide the information and have the viewer decide. [25:20–25:47]

In this example, the reader discussed his evaluation criteria and, by applying them, questioned the trustworthiness of the information. The textual judgment he made was based on his attention to the claims and evidence presented on different pages across the website. This excerpt also reflects an awareness of his role in determining the credibility of sources and how he regulated his active involvement in interpreting information.

In addition, the path analysis indicates that information location was significantly predicted by self-monitoring and source evaluation (R² = .530, p < .001). More specifically, source evaluation was a significant positive predictor of information location (β = .721, z = 7.952, p < .001), whereas self-monitoring did not predict information location significantly (β = .013, z = .103, p = .918). This result suggests a combined effect of source evaluation and self-monitoring in the location of online information, indicating that, without readers’ evaluation of the source’s reliability and usefulness in a critical manner, their self-monitoring alone may not contribute much to the effective use of information-locating processes. For example, P003 monitored her comprehension and noted the importance of the readability of a document to locate additional texts that would advance her understanding toward her reading goal. After accessing a detailed report on the National Resources Defense Council (www.nrdc.org) website about methods of financing MTR, she reported:

I don’t know a lot about bank and finance stuff … it is a little foreign to me, reading about this … I’m not sure if it [the report] is going to be too helpful … I am going to go back because it was a little bit heady for me … being a person [who] doesn’t know a lot about finance … I am going back to a different website … maybe the wording is a little bit easier for me to understand. [11:04–11:40]

Arriving at this moment, the reader was attentive to the point that understanding economic issues and factors is critical to understanding MTR. Then, in the excerpt above, she noted that the document she was currently reading would not advance her meaning making regarding this particular facet of MTR because the page lacked readability for her due to her own prior knowledge and the current focus of her reading. That is, the reader was not just active in assessing what she knew and could do but also attentive to judging the source’s usefulness against the result of self-monitoring. Therefore, she was able to maintain the goal of locating texts that explained the economic issues and sought alternative sources of information.

Another inference from the path analysis is that there was a statistically significant prediction of meaning making by a joint functioning of self-monitoring, source evaluation, and information location (R² = .525, p < .001). A positive prediction was made by both self-monitoring (β = .356, z = 2.820, p = .003) and source evaluation (β = .357, z = 2.127, p = .025), but not by information location (β = .141, z = 1.267, p = .227). This process is shown in part in P013’s analysis of a document found on the Kentucky Coal Association website (www.kentuckycoal.org) as he reflected on his current progress toward his goal and on the knowledge he had gained, while evaluating the situational nature of determining the legitimacy of information. Thus, he became aware of a possible alternative information location action:

Maybe Kentucky has more room to say no to things and has more room to expect more out of their post mining operations than West Virginia … so they just let the companies come in and do what they want … and they can’t stop them … so these facts could be just true for Kentucky and not for West Virginia … I might go back and try and find differences between Kentucky and West Virginia. [50:30–51:20]

In the above excerpt, the strategic interplay of monitoring the information that he had already gained from previous searches and evaluating the scope of the Kentucky-oriented source led this reader to effectively monitor his current understanding of the information and identify a possible alternative to address a gap that he identified in his understanding. Becoming aware of the need for information on differences between Kentucky and West Virginia for his overall understanding of MTR practices, the reader launched a new search but encountered difficulty finding a suitable source as indicated in the following excerpt:

I’m still thinking that West Virginia and Kentucky have different … results of their mining … and that people in West Virginia see a lot more harm done than good than people in Kentucky do … but I doubt I’ll find anything good about it … that’s kind of too specific for Google I think … I’d have to be looking at a certain source … then find information on that … one source … probably. [51:32–52:31]

In this excerpt, the reader’s information searching was not effortful and strategic. He attributed the difficulty of finding a suitable source to a lack of functionality of the web-search engine rather than conducting alternative searches (e.g., follow-up examinations of a few entries or more elaborated search terms). This excerpt demonstrates that although the reader’s source evaluation and self-monitoring were performed effectively, his attempt to locate the kind of text he wanted without using a strategic approach did not lead to an opportunity for enhanced meaning making.

Finally, the path analysis suggests a statistically significant prediction of critical question generation by a joint functioning of self-monitoring, meaning making, and source evaluation (R² = .403, p < .001). The prediction of meaning making toward critical questioning was marginally significant (β = .364, z = 1.828, p = .069). However, within the model, self-monitoring (β = .247, z = 1.543, p = .132) and source evaluation (β = .117, z = 0.599, p = .543) did not directly contribute to the prediction of critical questioning. In order to illuminate how these strategies influenced the meaning made, as shown through the reader-generated question outcome, two participants’ processes are juxtaposed in Figure 4. The first participant, P004, demonstrated successful strategy use and a high-scoring question; while the second participant, P028, was less strategic and generated a less sophisticated question. Figure 4 presents examples of specific strategy use that contributed to a specific understanding, as reflected in their generated questions.

OPEN IN VIEWER

Figure 4.

Two participants’ verbal reports and written questions that represent interrelationships of strategy use and question generation in online reading.

Discussion

This study provides evidence to support the view that online reading is a complex process in which multiple types of intertexual, metacognitive, and critical strategies are intertwined. At the outset, the study’s results suggest that more successful questioning can be achieved when readers are able to effectively use reading strategies to make multiple meanings from a variety of forms of texts that they find online. In the path analysis, readers’ use of meaning-making strategies positively intervenes in the joint functioning of information location, source evaluation, and self-monitoring. Consequently, the results imply that by engaging in a more productive comprehension process, readers can put themselves in a better position to generate higher quality questions that capture relevant and significant aspects of the topic at hand. Question generation is a complex task that demands the active identification and comprehension of different perspectives with different claims and evidence represented in multiple texts. As theories of reading multiple texts have described (Perfetti, Rouet, & Britt, 1999; Rouet & Britt, 2011), the process of intertextual comprehension entails the critical and argumentative thinking that underlies the creation of a compelling question. The results of this study, which are consistent with the findings from previous studies, shed light on how strategic readers integrate multiple online sources and transform information into a coherent understanding.

This study’s findings add an insight to new literacies debates (Leu et al., 2004). More often than not, discourses of new literacies mislead laypeople to underestimate the foundational processes for making meanings from texts in online reading, which readers must learn, use, and practice to achieve their reading goals. This tendency might stem from too great attention to the novelty that the Internet presents to readers, with its proliferation of information stored nonlinearly, and the emphasis on information-seeking behaviors that results from the increased access to information. Yet online reading can be better understood from a situative perspective that sees readers’ strategic actions as embedded in this highly accessible information space, but without positing a binary distinction of old and new processes. Online enviroments afford a vast information space that is textually mediated, and readers must tackle the information represented, both explicitly and implicitly, in a variety of forms of web sources. Strategic online readers manage a deliberate and thoughtful process of making meaning, while grasping important ideas from individual texts, contextualizing the understandings from each text within a larger boundary of textual space, and identifying the linkages that reside in and across diverse texts (Anmarkrud et al., 2014; Goldman et al., 2012). Online reading cannot be successful, without readers’ efforts to identify, analyze, and synthesize information.

Furthermore, this study describes readers’ meaning making as a thoughtful process made possible by their evaluation of each source they found. The mediating role of source evaluation in meaning making was observed in the path analysis of the measured strategy variables. Source evaluation has been widely claimed to be important as a higher order process for improved performance in learning (Wiley et al., 2009; Wineburg, 1991). Empirical studies on online reading have bolstered this claim as well (Cho, 2014; Goldman et al., 2012), reporting that, when readers evaluate information thoroughly, it helps them eliminate semantic noise, reject unwanted information, and reduce the uncertainties intrinsic to online texts. Conversely, if readers lack cognitive investment or are incapable of such evaluative reading, especially with complex and challenging texts, then their comprehension may suffer. A text-screening process is a prerequisite for building a coherent understanding of the text information, because incoherent accessing and collecting of texts undermines the ground on which information comprehension must stand. As previous research suggested (Braasch, Rouet, Vibert, & Britt, 2012; McCrudden at al., 2016), the utility of readers’ meaning-making strategies can be ensured by their critical decisions about texts at multiple levels (e.g., word choices, information citation, logical development of ideas, ways of representation, modal cohesiveness, author expertise, intended audience, hidden purposes and motives) and in multiple dimensions (e.g., relevance, reliability, accuracy, up-to-dateness, plausibility, usefulness).

Notably, the study’s results suggest that readers’ informed decisions about sources, in conjuction with self-monitoring, are pivotal to information location. In this study, a significant correlation of information location and self-monitoring disappeared when source evaluation was added to the association, while the effect of self-monitoring on source evaluation remained consistent. This is a novel observation because few studies have attempted to delineate the relationships between monitoring, evaluation, and location in the course of online reading, although previous studies in general have indicated the important roles of both metacognitive and evaluative processes in improving readers’ performance of information searching and online navigation (Coiro & Dobler, 2007; Goldman et al., 2012). In other words, self-monitoring at a metacognitive level is important in locating information, as it helps readers maintain the focus of information seeking when they use search terms and select links, but the utility of self-monitoring alone is limited without readers’ attempt to complete a more critical selection of sources based upon their judgments of source reliability and usefulness. For example, strategic readers in online information searches do not just reflect on information needs and alternative information-seeking plans (e.g., what to find, how to access it), but also frequently make forward inferences or predictions prior to clicking on a certain link and proceeding toward an in-depth reading of the connected text. Hence, strategic readers are tentative before they select information and thoughtful in judging the quality of the sources, and this mindset facilitates evaluative reading that results in goal-directed text location and navigation.

However, in contrast to what the conceptual model represents, the data show that information location was not strongly associated with the meaning-making processes of the adolescent readers in the current study. A positive interpretation of this result is that the readers observed in the study were in general skilled in locating online texts, and so information location did not contribute much to differences in the effectiveness of using meaning-making strategies among the readers. The performance assessment using the participants’ verbal reports in the study indicated that these readers earned higher scores, with smaller variation among participants, in information location than in any other strategy. Their generally skillful performance of information location might be due to their being already equipped with information-seeking tactics; because of the spread and impact of the Internet, today’s adolescents are often characterized as particularly adept at using the Internet as a major resource for both formal and informal learning on a daily basis (S. Fox & Rainie, 2014; Lenhart, 2015; Purcell & Rainie, 2014; Wells & Lewis, 2006). As previous research has demonstrated, adolescent readers can manage a set of technical skills in information searching, such as using web-search engines, browsing websites, checking links, and deciding whether to click on a link or not, in a very quick cycle of behaviors (Cho, 2014; Coiro & Dobler, 2007; Leu et al., 2008).

On the other hand, it also seems reasonable to consider a critical interpretation of this study’s result in regard to information location: The participants’ behaviors might have been shallow if they failed to engage in a deep level of reasoning about online sources. Uncritical readers in this study seemed to find it acceptable to perform actions in a disjointed manner, while ignoring emerging needs to coordinate all these isolated actions for productive learning. This tendency might be explained by the notion of a “representational bottleneck” that can occur in learning. Learners’ cognition is for action, which is situated in real time, and therefore, when readers face a novel task under time pressure, the representational bottleneck comes into play and performance suffers (Kirsh & Maglio, 1994). In this study, the projected task demands and constraints, such as processing limitless information about an unknown topic within a limited time, might have forced these readers to choose between the two important tasks of online reading, that is, information location and meaning making. As a result, these readers might have spent more time and effort finding sources, thus failing to achieve a balance with the time and effort needed to comprehend the sources. Or, conversely, they might have decided to focus on a couple of sources that they believed would be sufficient to complete their question-generating task. This could have made them reluctant to perform additional searches of alternative sources. In either case, these readers’ self-monitoring appeared to function as an off-loading strategy to minimize the cognitive burden caused by the dual task. This observation implies that a need exists for instruction that helps students effectively judge when and where to allocate their cognition and attention between information location and information comprehension at the metacognitive level. With carefuly designed materials, settings, and teacher scaffolding, these metacognitive strategies can be experienced, learned, and practiced, which would enable students to use these strategies to circumvent the representational bottleneck that occurs in online reading.

Alternatively, the tenuous relationship between information location and meaning making might have been caused by insufficient prior knowledge that could be utilized for information searching. It is evident that prior knowledge is one of the strongest nontextual factors that affects reading comprehension (Anderson & Pearson, 1984; Kintsch, 1988). Recent studies suggest that prior knowledge influences the depth of information searching and online navigation (Amadieu, Tricott, & Mariné, 2010; Lawless, Schrader, & Mayall, 2007). That is, readers with more knowledge tend to engage more in self-regulatory processes so they can visit relevant pages in a consistent manner (Taub, Azevedo, Bouchet, & Khosravifar, 2014). In contrast, for many of the participants in this study, mountaintop mining was an unfamiliar topic; for this reason, they might have concentrated more on finding general information about the topic to build their basic knowledge. While successful readers may bring to their reading a broader understanding of controversial arguments in general (e.g., economic benefits vs. environmental damages), a shallow focus on information location can lead nonproficient readers to accept or discard certain links and texts based only on a topic match between their search terms and the materials (Hofer, 2004). This observation suggests that providing topic knowledge is useful to facilitate systematic thinking during information location for productive comprehension, as long as it leaves enough room to provide a challenge and provoke a curiosity that students will want to satisfy.

Last but not least, self-monitoring cannot be ignored in strategic online reading. Building a coherent understanding of text(s) relies on a reader’s judgment of the texts pertaining to both initial and emerging goals for reading and the affordances and constraints of the textual environment in which the reading takes place. That is, while a cognitive strategy is carried out, the knowledge and awareness that result from self-monitoring continue to affect processing. The observations in this study that have emerged from examining the interplay of reading strategies support the idea that successful online readers are metacognitively competent at monitoring, modifying, and spontaneously revising what they know and how they learn (Brand-Gruwel et al., 2005; Hofer, 2004). It is noteworthy that information location, meaning making, and source evaluation were affected by each other when included in the model, while self-monitoring allowed a reader to engage in a system of strategic thoughts and actions involved in online reading. These results expand the findings from the previous studies that have reported the role of monitoring in online reading (Coiro & Dobler, 2007; Stadtler & Bromme, 2007), providing an integrative model of online reading strategies that accounts for how different strategies complement one another and contribute to the generation of new knowledge and perspectives.

Limitations and Conclusion

Some limitations should be noted in interpreting the results. A small sample size is an issue in path analysis, which in general requires a larger sample size, as recommended for structural equation modeling techniques. This study’s small sample size thus may have affected both the statistical power and the explanatory power of the analysis, and the interpretation of the results. In addition, the study’s participants from AP classes may not represent the entire population of high school students, so the results cannot be generalized, especially for unmotivated, low-skilled readers. Another limitation is related to the think-aloud task, which might have impacted readers’ cognitive engagement in both positive and negative ways. It is a methodological question whether the effects of a think-aloud process can or should be parsed out of an analysis of targeted cognitive processes. The study also did not consider potential bias or epistemological stances regarding, for example, certain topics or types of sources, which the readers may have brought to the task. Therefore, the possibility of such reader bias intervening in their strategic processes should be noted in the interpretation of the results. Finally, individual differences are an important consideration in the interpretation of data like those in this study. For example, noncognitive factors might exist that affected the participants’ reading performance, given that reading strategies are actions of human cognition, which is constantly reshaped by the readers’ feelings, emotions, motivations, interest, tolerance for uncertainty, willingness to take risks, and so forth.

Despite these limitations, the study contributes to our evolving understanding of online reading processes and how reading strategies work toward critical performance. In addition, an integrative account of students’ acts of reading in an uncertain online environment, as attempted in this study, can provide some implications for classroom practice. The study’s findings suggest that productive learning occurs when a reader makes informed decisions about digital sources, sources that often are complex and not vetted for credibility and trustworthiness. Thus, students need instruction that brings authentic textual settings (e.g., negotiating more than one source that can be self-identified) to classroom reading tasks (e.g., selecting reliable information, evaluating source authoritativeness, investigating what other sources might be helpful) to expose students to complex situations of online reading. Supportive materials (e.g., self-checklists) and teacher scaffolding (e.g., questioning, feedback) must be offered in a timely manner with an appropriate level of support and challenge, because it is extremely hard for many developing readers to internalize these critical strategies. The results of this study indicate that a particular focus on source evaluation will be beneficial due to the influence that identifying reliable sources has on students’ processes to locate the information they seek. Indeed, it may be useful for teachers to facilitate students’ understanding of source evaluation prior to in-depth instruction on the other strategies. Teaching one strategy at a time can be helpful, but ample opportunities must be provided for students to use the strategies together, reflect on their thinking and bias, and self-test new approaches to online reading. Finally, critical to strategy learning is teacher modeling and classroom talks that evolve around what students can do, know, and accomplish, which can be framed around the interrelation of the four strategy types in order to better facilitate student self-assessment and teacher identification of student progress and learning. Informative and dialogic interaction among class members and demonstrations of ways of thinking can help students experience and develop evaluative mindsets, willingness to challenge texts, and positive attitudes toward exploring new sources, knowledge, and perspectives beyond the boundaries of a given text.