Individual Differences in the “Myside Bias” in Reasoning and Written Argumentation

Abstract

Three studies examined the “myside bias” in reasoning, evaluating written arguments, and writing argumentative essays. Previous research suggests that some people possess a fact-based argumentation schema and some people have a balanced argumentation schema. I developed reliable Likert scale instruments (1-7 rating) for these constructs and conducted an evaluation of instrument validity and reliability. A myside bias in argumentative essays was predicted by the fact-based and balanced argumentation schema instruments using these individual-difference measures. Strength of opinion predicted the myside bias in generating reasons but not in writing argumentative essays. There was a weak but significant correlation between the myside bias in generating reasons and writing argumentative essays. In evaluating written arguments, the fact-based schema instrument predicted agreement and quality ratings for claims supported by factual but not nonfactual reasons. Ratings of the quality of rebutted arguments were predicted by the measures of individual differences in argumentation schemata.

Keywords

cognitive bias argumentation schema generative cognition argument psychology of writing

Recently, there has been a good deal of interest in the relationship between reasoning and argumentation (Mercier & Sperber, 2011). One type of cognitive bias found in the context of reasoning and argumentation is the “myside bias” (Wolfe & Britt, 2008). The myside bias was first identified by David Perkins and has been studied sporadically for the past 25 years (Baron, 1991, 1995; Macpherson & Stanovich, 2007; Perkins, 1985, 1989; Perkins, Bushey, & Farady, 1986; Perkins, Farady, & Bushey, 1991; Toplak & Stanovich, 2003). In studies of informal reasoning, the myside bias has been defined as the tendency to ignore information on the side that one disagrees with—the other side of an issue—in favor of information that supports one’s position, “myside.” Empirically, the myside bias has generally been operationalized as the tendency to generate more reasons in favor of one’s position than the other side’s (Macpherson & Stanovich, 2007).

In the context of written arguments, the purpose of an argumentative essay is to advance a position. Thus, it would be quite reasonable for an argumentative essay to include more and better arguments on myside rather than the other side. However, it would be problematic for the author of an argumentative essay to ignore the other side completely. While the optimal mix of myside and other-side arguments is affected by a variety of factors—including the topic, intended audience, context, style, and objectives—a complete lack of other-side information is generally problematic and demonstrably weakens arguments (Wolfe, Britt, & Butler, 2009). Thus, excluding other-side arguments entirely is a reasonable approach to operationalizing the myside bias in written argumentation (Wolfe & Britt, 2008).

Decades of research on reasoning indicate that people reliably differ from one another in epistemological beliefs about knowledge and thinking (Charney, Newman, & Palmquist, 1995; Hofer & Pintrich, 1997; Kitchener & King, 1981). Perry (1970) argued for a progression of stages of epistemological sophistication; however, the stage approach has been criticized on a number of grounds, with more recent researchers emphasizing epistemological styles (Charney et al., 1995), reflective judgments (King & Kitchener, 2004), and belief systems (Schommer, 1990) rather than stages. Although much is known about epistemological development in children, adolescents, and adults (Baxter Magolda, 1992; King & Kitchener, 2004; Kuhn, Cheney, & Weinstock, 2000), the current work does not address development of the argumentation schema over time. I share the reasonable assumption that individual differences in the argument schema are learned in cultural context, mutable, and subject to change over time (Wolfe, Britt, & Butler, 2009). I use the term individual differences not to advance any claim that such differences are rigid, unalterable, or biologically determined but rather to suggest that they are reliable and stable in the absence of powerful learning experiences.

Researchers have found important individual differences in argumentation. Ricco (2007) studied individual differences in reasoning and argumentation and found that deductive reasoning ability, higher-order epistemic beliefs, and the ability to overcome belief biases were associated with reduced fallacies in informal reasoning. Epistemic understanding was also found to be predictive of reasoning in a study of jury decision making, leading Weinstock (2009) to conclude that, “understanding what to do with knowledge may be more critical in informal reasoning competence than having a larger knowledge base” (p. 431). In a similar vein, Weinstock, Neuman, and Tabak (2004) found that awareness of the norms of argumentation predicted the ability to identify fallacious informal arguments. Mason and Scirica (2006) also discovered that epistemological understanding was a significant predictor of the ability to generate arguments, counterarguments, and rebuttals. Stanovich and West (1998) found that thinking dispositions predict some differences in reasoning. They argue that some thinking styles could be conceptualized as “stored rules with propositional content (e.g., ‘think of alternative explanations,’ ‘think of a reason against your position’)” (p. 180) while others are less rulelike. Many of these distinctions can be conceived as differences in the argumentation schema.

The Argumentation Schema

In essence, an argument is a claim supported by one or more reasons (Reznitskaya, Kuo, Glina, & Anderson, 2009; Voss, 2005; Voss, Fincher-Kiefer, Wiley, & Silfes, 1993; Wolfe, Britt & Butler, 2009), or “data,” using Toulmin’s (1958) terminology. When a person composes or comprehends a written argument, he or she makes use of an argumentation schema (Wolfe, Britt, & Butler, 2009): a learned, culturally derived set of expectations and questions about argumentative texts. In reading, the argumentation schema is typically evoked by a provocative claim (Britt & Larson, 2003). In writing, the argument schema is generally evoked by demands of an assignment, expectations about the audience, and the goals of the author (Wolfe, Britt, & Butler, 2009).

According to the Wolfe, Britt, and Butler (2009) argumentation schema model, a claim is associated with three expectations, or “slots” in the schema: the theme, side, and claim predicate (Britt, Kurby, Dandotkar, & Wolfe, 2008). The theme is the topic of the argument; the side is represented as either pro or con; and the predicate is the specific position taken by the author. The schema activates knowledge, attitudes, and beliefs relevant to the theme. Although one may logically consider arguments from many perspectives, the schema-based tendency is to simplify arguments into pro and con sides (Wolfe, Britt, & Butler, 2009). The argument predicate is the specific position taken by the author. Britt et al. (2008) found that memory for the theme and side is quite good while memory for the specific predicate is significantly worse. The principal expectation generated by the argumentation schema is for a reason addressing the question “Why should I accept this claim?” Reasons can be statements of fact, other texts, appeals to beliefs, or other arguments. A central claim of this paper is that individuals differ with respect to key characteristics of the argumentation schema and that these differences account for some differences in the myside bias in argumentation.

Evidence for individual differences in argumentation schemata stem from Wolfe and Britt (2008), in which participants wrote essays in the lab under various experimental conditions. Many participants failed to include any other-side information in their essays. In that study, post hoc tests revealed two predictors of the myside bias. When asked “What makes a good argument?” some participants provided evidence of a “fact-based argumentation schema” that was significantly associated with the myside bias. Some participants demonstrated a preference for balanced arguments and were less likely to demonstrate the myside bias. Wolfe and Britt manipulated information search and side assigned, and, interestingly, side assigned was not predictive of the myside bias—participants exhibited the bias at the same rate whether they personally agreed or disagreed with the position.

The fact-based argumentation schema can be contrasted to an evidence-based schema. To illustrate, readers of Written Communication are keenly interested in data and the evidence used to support various assertions. We respect a case supported by good data. By way of contrast, the fact-based argumentation schema is an uncritical belief that facts alone make an argument good. Individuals with a fact-based argumentation schema demonstrate little or no understanding of the audience or appreciation of context, the role of counterarguments, or alternative explanations. For these people, argumentation is simply a matter of presenting facts with little regard for other aspects of argumentation. To illustrate, one participant in the Wolfe and Britt (2008) study said, “A good argument is one that has plenty facts to back it up. Anyone can be won over if there are enough facts.” A balanced argumentation schema suggests a preference for arguments that acknowledge more than one side. For example, when asked what makes for a good argument, one participant said, “I feel a good argument gives opinions from both sides of the argument then shows why their side is correct, or better.”

The purpose of the current work is to provide a more solid empirical basis for our understanding of individual differences in fact-based argumentation schema and balanced argumentation schema. I examined these constructs in three contexts: writing argumentative essays, generating reasons, and evaluating written arguments. This paper compares the myside bias in written essays to the myside bias on reason generation tasks. It also assesses whether the myside bias differs reliably among individuals, and it sheds light on which individual-difference measures best predict the myside bias.

In previous work, Wolfe and Britt (2008) found evidence for fact-based and balanced argumentation schema through post hoc tests using statements about what makes a good argument to predict the myside bias in written arguments. In subsequent work, they found evidence that tutorial interventions designed to ameliorate the effects of the fact-based argumentation schemata reduced the myside bias in written essays (Wolfe & Britt, 2008; Wolfe, Britt, & Butler, 2009). My first task in this research campaign was to develop simple and reliable individual-difference measures for fact-based argumentation schema and balanced argumentation schema.

Study 1: Individual-Difference Measures

The purpose of the first study was to develop reliable Likert scale items to assess the fact-based argumentation schema and balanced argumentation schema. Many of the items were created by converting quotes from participants in the Wolfe and Britt (2008) study into Likert items. Sixty-three undergraduate participants at Miami University received course credit for responding to a number of Likert scale items interspersed with filler items. Participants in all of these studies were enrolled in an introductory psychology course, thus representing a broad cross section of Miami undergraduates. Typically, introductory psychology students are primarily lower-division students. The best items were selected for each purpose using Cronbach’s alpha as the selection criteria. For the fact-based argumentation schema, I developed a 10-item Likert scale instrument (see Appendix A) where participants had to agree or disagree with statements using a 7-point scale. Here are two examples of items where strong agreement is taken as evidence of a fact-based argumentation schema: “A winning argument is a claim supported by facts” and “I think a strong argument is based on facts that cannot be refuted. If there is solid evidence that cannot be disputed due to its truthfulness you have a solid argument.”

For the balanced argumentation schema, I developed an instrument that had 5 items (see Appendix A), again using participant statements from the Wolfe and Britt (2008) study to create Likert items—for example, “A strong argument presents both sides of the issue. In doing so it should point out the flaws in the opposing side, while highlighting the positive aspects of the side being promoted.”

It is important to note that few items produced strong disagreement (i.e., used the full range of the scale). Generally, with this sample of respondents, the survey items assess agreement that typically spans a range from strong agreement and moderate agreement. For example, Item I3 in Appendix A, “A winning argument is a claim supported by facts,” had a mean of 5.68 (SD = 1.23). On the 7-point scale (7 = strongly agree), 17 participants endorsed 7; 22 participants endorsed 6; 17 participants endorsed 5; 4 people responded 4; and responses 3, 2, and 1 were produced by 1 participant each. It would be inappropriate to claim that participants scoring low on fact-based argumentation schema are “against facts.”

In the first study, Cronbach’s alpha was .82 for the fact-based schema items and .79 for the balanced schema items (see Table 1). Regarding the next two studies, where I applied these instruments, for fact-based schema items, alpha was .84 in Study 2 and .78 in Study 3. For the balanced schema items, alpha was .80 in Study 2 and .81 in Study 3. I take this as reasonably solid evidence for the reliability of these measurers. I also attempted to produce an instrument measuring preference for rhetoric, a construct not investigated in previous research. For example, one item read, “In order to develop a strong argument you need to know about your readers (the audience).” Unfortunately, as can be seen in Table 1, I was unable to achieve satisfactory reliability for this instrument.

Table 1.

Cronbach’s Alpha for the Individual-Difference Measures in the Three Studies

Study	n	Fact-Based Schema	Balanced Schema	Preference for Rhetoric
1	63	.82	.79	.71
2	85	.84	.80	.74
3	100	.78	.81	.70

Study 2: The Myside Bias in Written Essays and Reason Generation Tasks

The second study served two purposes. The first was to assess the relationship between the myside bias in writing essays and reason generation tasks found in the literature. My prediction was that there would be a significant correlation suggesting some corresponding cognitive processes underlying biased reasoning and argumentation. The second was to assess the extent to which the balanced schema and fact-based schema instruments predict task performance. The key prediction was that both instruments would predict the myside bias on the essays. I further predicted that the balanced argumentation schema would predict the myside bias on the generation tasks.

Method

Eighty-five undergraduate participants at Miami University performed in the lab for about 2 hours in partial fulfillment of course requirements. The study was conducted using a within-subjects design with each participant receiving all tasks. Participants wrote persuasive essays after reading a booklet of pro and con arguments. Then they completed two argument generation tasks (described below) from the literature to assess the myside bias in reasoning. Participants rated their agreement with each proposition. Next they received a Likert scale instrument with the 10 fact-based schema items and 5 balanced schema items embedded among filler items.

For the argumentative essay, participants were asked to write an essay of about 500 words on a computer about a proposed new math requirement, requiring all students to take four math courses. The essay assignment, taken from Wolfe and Britt (2008), was designed to allow both pro and con arguments while producing reliable opposition—the majority of participants disagreed with the proposal. The specific proposition and essay instructions read as follows:

Miami University should impose a two-year (4 semester) math requirement for all students that includes at least two semesters of calculus, and another year of mathematics training in linear algebra and number theory. Students who are unprepared for these courses must take any prerequisites (such as pre calculus) before, and in addition to, these four courses. All students must earn at least a C in each course or the course will have to be repeated.

Please write a brief argumentative essay of about 500 words (at least 3 paragraphs) about the proposal in the space below. Feel free to consult the booklet of pro and con arguments as you develop and write your essay. Be sure to save your work. Please signal the research assistant when you are done.

For the reason generation tasks, the first one was adapted from Macpherson and Stanovich (2007) and applied to Miami University, Oxford, Ohio. The proposal suggests that all students should pay the higher rate of out-of-state tuition. Because approximately one third of Miami students are from out of state, participants were asked at the end of the experiment whether they were in-state or out-of-state students. The specific materials are presented as follows:

The difference between the actual out-of-state cost of an undergraduate education at Miami University and what an in-state student pays is approximately $12,500 per year. The Ohio taxpayer pays this $12,500 difference. All Miami undergraduates should pay the actual out-of-state cost of their Miami education, even Ohio residents. Write down arguments about this issue. Try to write as much as you can and please feel free to take your time.

The second problem was also taken from Macpherson and Stanovich (2007) and proposes that people should be able to share music over the Internet without paying for it. That problem reads as follows:

Music file sharing over the Internet is a growing phenomenon. People should be able to share music over the Internet without paying for it. Write down arguments about this issue. Try to write as much as you can and please feel free to take your time.

For each of the generation tasks, we asked participants to rate their agreement on a 6-point scale. After they completed the Likert tasks, including the items presented in Appendix A, participants were thanked and debriefed.

Results

One participant generated bullet points rather than an essay, and this person’s data were excluded from the analysis. Table 2 provides descriptive statistics pertaining to 84 essays. They average 446 words, or about 2.5 typed, double-spaced pages. Two trained judges independently assessed the essays on a number of dimensions, and interrater reliability was consistently above .90. The essays had an average of 3.0 claims supported by reasons and 2.9 unsupported claims. Eighty-seven percent had a thesis statement, and 17% used quotes from the booklets to bolster their case. The essays had a mean Flesch-Kincaid grade level of 11. The key dimension of interest is whether or not the essays included other-side information—that is, the myside bias. Of the 84 essays, 31 exhibited the myside bias, and the remaining 53 presented at least some other-side information. Of these, 39 (74%, or 46% of all essays) rebutted the other-side claims. These data suggest that despite the fact that the essays were not graded and were written in a laboratory environment, they were appropriate for students enrolled in an introductory psychology course.

Table 2.

Characteristics of 84 Essays

Characteristic	M (SD)	n (%)
Words	446 (97)
Sentences	20 (5.6)
Supported claims	3.0 (1.1)
Unsupported claims	2.9 (1.9)
Has a thesis statement		73 (87)
Rebutted other-side arguments		39 (46)
Presented quotes from the booklets		14 (17)
Flesch reading ease	48.0 (12.0)
Flesch-Kincaid grade level	11 (1.3)
Myside bias (no other-side information)		31 (37)

Note: Interrater reliability = .90 or greater for all judgments.

On the generation tasks, reasons were categorized as myside if they supported the side that participants endorsed on the agreement scale. Responses were characterized as other side if they supported the opposite side. On the tuition problem, participants generated a mean of 0.82 other-side reasons, and on the music problem, a mean of 1.39 other-side reasons. For the music-sharing problem, mean agreement was 3.91 (SD = 1.35). In-state and out-of-state students differed in their agreement with the tuition problem. In-state students (n = 45) produced a mean of 1.8 (SD = 1.08) and out-of-state students (n = 40) produced a mean of 3.58 (SD = 1.60), F(1, 83) = 36.68, p < .0001. As is typically found in the literature, in aggregate participants exhibited the myside bias on both problems. Table 3 presents the mean number of myside reasons generated, the number of other-side reasons generated, and the mean difference for each problem.

Table 3.

Number of Reasons Generated by Task: M (SD)

Task	Myside	Other Side	Myside – Other Side
Tuition	3.61 (1.87)	0.82 (1.03)	2.79 (2.29)
Music	3.31 (1.89)	1.39 (1.32)	1.91 (2.68)

One of the first questions of interest was whether or not there is an association between the myside bias in the essays and the myside bias in the reason generation tasks. I assessed the number of other-side reasons produced in each setting. As can be seen in Table 4, there was a significant correlation of .25, suggesting that about 6% of the variance in the myside bias in the essays could be predicted by the generation tasks. The correlation between the number of other-side reasons produced on the two generation tasks was .51.

Table 4.

Correlations Among the Essays and Generation Tasks for the Myside Bias

Other-Side Reasons Produced	R	R ²	p
Essay, generation tasks	.25	.06	< .03
Tuition, music generation tasks	.51	.26	< .0001

A central hypothesis was that items assessing the fact-based argumentation schema and balanced argumentation schema would predict the myside bias in the essays. This is what I found. Despite the fact that the two measurers were positively correlated with each other, r = .62, as predicted, (a) the higher the score on the balanced schema instrument, the lower the myside bias and (b) the higher the score on the fact-based schema instrument, the greater the myside bias. The balanced schema score had a negative beta weight (β = –.263) and significantly predicted the myside bias in the essays, p < .04. The fact-based schema score had a positive beta weight (β = .278) and significantly predicted the myside bias in the essays, p < .035. Preference for rhetoric was not significant, t = 1.54, p = .13.

Regarding the generation tasks, the fact-based schema and preference-for-rhetoric items were not predictive. However, as hypothesized, the balanced schema score was predictive of the myside bias such that the greater the score on the balanced schema instrument, the lower the myside bias (β = –.207), p < .03. Strength of opinion was measured as scores closer to the endpoints of 1 or 6 compared with those closer to the middle of the scale. Strength of opinion also predicted the myside bias. The stronger the opinion, the greater the myside bias (β = 2.05), p < .001.

There was not a significant relationship between opinion strength and myside bias in the essays. Agreement was measured dichotomously because there was virtually no variance, and for agreement and side advocated, χ²(1) = 1.12, p = .29. This prompted me to reanalyze the essays from Wolfe and Britt (2008), where agreement was measured on a 6-point scale, for any relationship between agreement and side advocated. Mean agreement for essays exhibiting the myside bias was 5.182 (SD = 1.21), and for those not exhibiting the myside bias, it was 5.135 (SD = 0.93), F < 1. There was no hint that strength of opinion corresponded to the myside bias. This appears to be a difference between the reason generation task and the task of writing a persuasive essay.

Discussion

As predicted, there was a significant correlation between the myside bias in the essays and on the generation tasks. The correlation was moderate, perhaps because there was relatively little variance in the amount of other-side information in the essays. This suggests that it is reasonable to use the term myside bias in both contexts. However, the data also highlight an important difference between the writing essays and the reason generation task. Strength of opinion was a good predictor of the myside bias on the reason generation tasks but not for essays in this and previously published research.

As hypothesized, the balanced schema and fact-based schema instruments predicted the myside bias in the essays. In previous studies (Wolfe & Britt, 2008), post hoc tests using statements about what makes a good argument were seen as evidence for fact-based and balanced argumentation schema. These post hoc measures predicted the myside bias in written arguments (Wolfe & Britt, 2008). In other experiments (Wolfe & Britt, 2008; Wolfe, Britt, & Butler, 2009), tutorials designed to minimize the effects of the fact-based argumentation schema reduced the myside bias in written essays significantly. Together, these studies provide solid converging evidence for individual differences in argumentation schemata associated with the myside bias in argumentation.

Study 3: The Myside Bias in Reading Arguments and Generating Reasons

The purpose of the third study was to examine individual differences in the myside bias in evaluating written arguments and generating reasons and the consequences of the fact-based and balanced argumentation schemata. Participants read and rated arguments supported by factual or nonfactual reasons and those for which other-side information was presented and rebutted, as well as those without other-side arguments. This enables us to examine the consequences of the myside bias and to determine the consequences of different kinds of support. My prediction was that arguments supported by facts would be higher rated than the same claims supported by other reasons. I also predicted that arguments in which other-side information was presented and rebutted would be rated as higher in strength or quality and would produce greater agreement. In other words, I predicted that the myside bias would weaken arguments. I further predicted that people operating with a fact-based argumentation schema would rate arguments supported by facts higher than others and that those with a balanced argumentation schema would rate balanced arguments higher.

The second study used just two reason generation tasks, and one produced significant differences between in-state and out-of-state students. Thus, I sought to better understand the myside bias in reasoning using six reason generation tasks with the same hypotheses tested in Study 2. The use of six generation tasks gave me the opportunity to determine which of several approaches to measuring the myside bias was the most reliable and whether conceiving the myside bias as the absence of other-side reasons is empirically feasible. The results of this methodological analysis are presented as Appendix B.

Methods

The third study assessed the myside bias on reason generation tasks and ratings of brief written arguments. One hundred undergraduates at Miami University participated for course credit in a within-subjects design. Participants were given six reason generation tasks from the literature. They also had to rate a number of arguments for agreement and the strength or quality of the arguments. They received the same individual-difference measures as in the previous study, interspersed among other tasks. The reason generation tasks and their sources in the literature are presented in Table 5. These tasks present a variety of themes, making it reasonable to generalize from any robust findings.

Table 5.

Reason Generation Tasks Used in Study 3 by Source

Reason Generation Task	Source
1. Tuition	Macpherson and Stanovich (2007)
2. Music file sharing	Macpherson and Stanovich (2007)
3. Providing more money for public schools would improve teaching.	Perkins, Farady, and Bushey (1991)
4. We should all strive to become less reliant on technology.	New for this study
5. Religious people are generally more honest than nonreligious people.	Stanovich and West (2007)
6. Students should have a stronger voice in setting university policies.	Stanovich and West (2007)

The argument evaluation task involved reading brief arguments and rating agreement and strength or quality on a 7-point scale. There were four blocks of 11 items with the order counterbalanced and the blocks separated by other tasks and filler items. Participants received four versions of each argument in counterbalanced blocks in a within-subjects design. The four versions of the argument had a claim supported by a factual reason, the same claim supported by a nonfactual reason, then each of these with other-side information presented and rebutted (i.e., balanced arguments). The Agree Fact Index comprises ratings of agreement supported by facts minus the same claims supported by nonfacts. The Quality Fact Index comprises the ratings of quality of claim supported by facts minus the same claims supported by nonfactual support. The Rebut Agree Index comprises ratings of agreement of arguments with other-side arguments presented and rebutted minus the same claims supported without other-side information. The Rebut Quality Index comprises ratings of quality of arguments with other-side arguments presented and rebutted minus the same claims supported without other-side information.

A: Factual support; no other-side information—Car emissions should be regulated because the EPA reports that 95% of carbon monoxide pollution in cities comes from car emissions.

B: Nonfactual support, no other-side information—Car emissions should be regulated because we need to prevent further global warming.

C: Factual support, with other-side information—Car emissions should be regulated because the EPA reports that 95% of carbon monoxide pollution in cities comes from car emissions. The automobile industry argues that regulating emissions raises the cost of cars because it is expensive to install pollution control devices that reduce emissions only slightly. However, the benefits of even small reductions in emissions are multiplied by the millions of cars on the road.

D: Nonfactual support with other-side information—Car emissions should be regulated because we need to prevent further global warming. The automobile industry argues that regulating emissions raises the cost of cars because it is expensive to install pollution control devices that reduce emissions only slightly. However, the benefits of even small reductions in emissions are multiplied by the millions of cars on the road.

Results

On the reason generation tasks, strength of opinion—measured as the absolute difference between agreement on the 6-point scale and 3.5, the midpoint of the scale—was predictive of the myside bias on all six problems (see Table 6). For each task, the greater the opinion strength, the greater the myside bias. Table 6 also shows that the number of myside arguments was predictive of the myside bias on all six problems. The greater the number of myside reasons generated, the more likely that participants produced no other-side reasons.

Table 6.

Myside Bias by Problem Predicted by Opinion Strength and the Number of Myside Reasons Generated

	Opinion Strength		Number of Myside Reasons
Problem	F(1, 96)	p	F(1, 96)	p
1. Tuition	6.8	< .01	14.8	< .0002
2. File sharing	13.9	< .003	6.7	< .01
3. Public school	11.9	< .0008	7.8	< .006
4. Technology	8.7	< .004	10.2	< .002
5. Religion	4.4	< .04	16.7	< .0001
6. Student voice	4.8	< .03	19.1	< .0001

The argumentation schema measures did not predict performance on the reason generation task. On only the problem about music file sharing was the myside bias predicted by the balanced schema items, F(1, 98) = 7.84, p < .006. This effect was found only for the file-sharing problem; however, this does replicate the same effect for the same problem in Study 2.

Regarding the evaluation of brief written arguments, participants rated their agreement with claims supported by facts significantly higher than the same claims supported by nonfactual reasons. On a 7-point scale (1 = strongly agree), factual arguments received a mean of 3.31 (SD = 0.64) and nonfactual arguments, a mean of 3.49 (SD = 0.67), t(99) = 4.82, p < .0001. There was a comparable effect for ratings of the strength or quality of an argument, with factual arguments receiving a mean of 3.38 (SD = 0.63) and nonfactual arguments receiving a mean of 4.66 (SD = 0.85), t(99) = 17.05, p < .0001. There was no difference in the rating of agreement with balanced arguments or those without other-side information presented and rebutted: both had a mean of 3.40. However, arguments with other-side information presented and rebutted were rated as significantly higher in strength or quality, 3.69 (SD = 0.74), than the same arguments without other-side information, 4.39 (SD = 0.76), t(99) = 9.98, p < .0001. The myside bias measurably weakened the perceived quality of these 11 arguments.

I hypothesized that the fact-based schema items would be predictive of agreement and quality ratings for claims supported by factual reasons but not for claims supported by nonfactual reasons. These hypotheses were supported. The fact-based argumentation schema significantly predicted ratings of agreement and quality for factual arguments: for agreement with factual arguments, F(1, 98) = 4.97, p < .03, and for quality of factual arguments, F(1, 98) = 7.12, p < .009. The fact-based schema did not predict differences in ratings on nonfactual arguments. For arguments supported by factual reasons, there was a .25 correlation between agreement and fact-based schema score (R² = .05) and a .27 correlation between quality ratings and fact-based schema score (R² = .07). For the nonfactual arguments, these correlations were not significant. These findings suggest that the fact-based argumentation schema is involved in the evaluation of brief written arguments.

Discussion

There are several reasonable ways to assess the myside bias in generating reasons (see Appendix B for a detailed account). In studies such as this, with five or more reason generation tasks, perhaps the best approach is to count the number of problems for which a participant fails to come up with any other-side reasons (i.e., the approach taken here). This approach has adequate statistical properties and is conceptually better suited to be labeled a bias. For studies using fewer generation tasks, counting the number of other-side reasons or determingin the percentage of myside reasons from total reasons is a justifiable strategy.

Strength of opinion consistently predicted the myside bias in generating reasons. This appears to be a robust difference from the task of writing argumentative essays where strength of opinion is not associated with the myside bias. Generating more myside reasons was also associated with the myside bias in generating reasons—even as defined as the absence of any other-side reasons, as was the case here. It may be that people have a cognitive “stop rule” for what constitutes an acceptable number of total reasons. It may also be the case that cognitive effort increases substantially after two or three reasons have been generated.

Across the board, people rated arguments that presented and then rebutted other-side arguments as stronger than those without any other-side information. In this study, arguments with other-side information were also longer, so it is logically possible that the preference for arguments where other-side information is presented and rebutted is simply a preference for longer arguments. Yet people who scored high on the balanced argument schema rated balanced arguments higher in quality than did those scoring low. In a similar vein, participants rated arguments supported by facts higher in agreement and quality than those same claims supported by nonfactual reasons. Yet people who scored high on fact-based argumentation schema rated factual arguments higher in quality and agreement than did those scoring low. These results can be interpreted as validation of the fact-based and balanced argumentation schema constructs and the instruments we used to assess them. People within this cultural context share a general argumentation schema, and yet individuals differ reliably in the ways that those argumentation schemata treat balanced arguments and factual support.

General Discussion

The myside bias is pervasive in generating reasons and developing written arguments, yet people differ in demonstrating the myside bias in generating reasons, writing essays, and evaluating written arguments. There are reliable differences among individuals in their schemata for argumentation, and these differences have consequences for both reasoning and argumentation. The fact-based schema predicts ratings of factual but not nonfactual arguments. There is less consistent evidence that those high on balanced argumentation schema favor reading balanced arguments over those without other-side information. In writing essays, the myside bias is predicted by fact-based and balanced argumentation schema instruments. These effects reflect and amplify findings in earlier research (Wolfe & Britt, 2008; Wolfe, Britt, & Butler, 2009).

Wolfe (2011) found that argumentative writing assignments are pervasive across the entire university curriculum. In a study of writing assignments in academic contexts as diverse as business, engineering, fine arts, humanities, and the sciences, Wolfe found that about 59% of the writing assignments required the student to make a written argument. An important implication of this research is that even bright students may have very different beliefs about the role of facts and support in argumentative discourse in these diverse academic contexts. Faculty aspiring to “reach all students” may find it beneficial to more self-consciously focus on the way that facts are used to support claims in their disciplines. Many students are sensitive to the need to support their assertions but naïve about the author’s responsibility to contextualize supporting evidence in a framework that includes competing claims, counterarguments, and contradictory evidence. Although interventions to directly teach the skills of counterargument have met with some success (Wolfe, Britt, Petrovic, Albrecht, & Koop, 2009), the myside bias may be a good candidate for contextually tailored writing instruction across the curriculum.

There is a reliable but weak correlation between the myside bias in written essays and on reason generation tasks. On the reason generation tasks, the myside bias is consistently predicted by strength of opinion and the number of myside arguments generated. These findings are consistent with those of Baron (1991) and Stanovich and West (2007). There is some evidence that the balanced argumentation schema affects the generation of reasons for some issues (i.e., music file sharing); however, the myside bias operates differently in the context of argumentation than it does in the context of generating reasons. This result contradicts strong claims that humans reason for the purpose of argumentation rather than to solve problems and make decisions (Mercier & Sperber, 2011).

These studies were quite time-consuming, taking about 2 hours per participant by having each participant write essays, evaluate arguments, and generate reasons on several problems in the lab. Thus, it is practical to have brief yet reliable measurement instruments for future research on the myside bias in written argumentation. Both the fact-based argumentation schema and the balanced argumentation schema are constructs that rest on solid empirical footing, and both play an important role in the myside bias in reasoning and written argumentation. One area that has not been explored is the extent and ways to which differences among people in argumentation schemata change over time and across educational and life experiences. Perhaps the research tools developed in these studies may enrich developmental, educational, and longitudinal studies of reasoning and argumentation.

“Authentic” arguments are those written for the purpose of persuading an actual audience, rather than fulfilling an academic assignment. Wolfe and Britt (2008) report that “the vast majority of these authentic written arguments included other side information (in other words, very few exhibited the myside bias). Typically, the purpose of including other side information in the essays was rebuttal” (pp. 2-3). The myside bias is a problem not just for those who wish to create a more open-minded society where contrary opinions are freely explored. It is also a problem for those who wish to more effectively advance their own arguments.

Recently, the Republican Party of Texas (2012) platform has come into a good deal of criticism in its opposition to the “teaching of higher-order thinking skills” and “critical thinking skills” (e.g., Harvey, 2012) in favor of “knowledge-based education.” I have no evidence that the fact-based argumentation schema is associated with any ideology. Rather than opening yet another front in the so-called culture wars, I (perhaps idealistically) hope that this research can help bridge the gap between educational approaches that emphasize domain-specific knowledge and those that emphasize critical thinking. Well-educated students—and truly all citizens in a healthy democracy—need both a large knowledge base and an understanding of how to use that knowledge in making and evaluating arguments.

Footnotes

Appendix A: Likert Items for Fact-Based Argumentation Schema and Balanced Argumentation Schema

Code numbers are based on their position in the larger pool of items in Study 1. For each construct, items are presented in the order of most rare to most frequent.

Appendix B: Five Approaches to Assessing the Myside Bias

With each participant completing six reason generation tasks, one is able to use the same data to calculate the myside bias five different ways. One can then examine the correlations among measures and determine which approach is most reliable using Cronbach’s alpha and the Kuder-Richardson formula 20 (KR-20). The myside bias can be assessed in any of the following ways:

Generally speaking, the myside bias is the tendency to produce more arguments on “my side” of the argument than the other side. Approach A is to count the total number of other-side arguments produced on all six problems, with the greater number of other-side arguments indicating less of the myside bias. Approach B counts the number of instances on six problems when there are more myside than other-side arguments. Each instance is scored as a 1, and the greater the number (from 0 to 6), the greater the myside bias. Approach C is to subtract the number of other-side reasons from the number of myside reasons and sum across the six problems, with larger numbers measuring greater myside bias. Approach D is to calculate the percentage of total reasons on myside as the number of myside reasons over total reasons generated, with higher percentages corresponding to greater myside bias. Finally, Approach E is to consider the myside bias only as instances where participants failed to generate any other-side reasons. Each of the six problems without any other-side information is scored as 1, and the greater the number (from 0 to 6), the greater the myside bias. This approach mirrors the tact previously taken measuring the myside bias in essays (Wolfe & Britt, 2008).

Regarding the best way to measure the myside bias on reason generation tasks, Table B1 presents the correlations among measures, with R² below the diagonal. Not surprising, these correlations are high, since they are using the same raw data. Approach D had the highest average correlation, .87, while approaches B, A, and E were similar with average correlations of .85, .82, and .82, respectively. Only Approach C was appreciably lower, with a mean correlation of .70.

The reliability of each approach was measured using (a) Cronbach’s alpha for those measures with a continuous variable for each problem (A, C, D) and (b) KR-20 for those where the myside bias on each problem was assessed as a binary variable (B, E). In each case, alpha was higher when the tuition problem was excluded, presumably due to the differences between in-state and out-of-state student participants. As indicated in Table B2, Approach A was the most reliable, α = .64. Approaches C, E, and D had similar scores: α = .59, KR-20 = .55, and α = .52. Although these scores would be disappointing for a test or measurement instrument, they are adequate for experimental materials requiring participants to reason about a variety of issues. Only Approach B yielded an appreciably lower KR-20 score, .34.

Table B3 presents the mean myside bias measure for each of the five approaches using only the five most reliable problems (i.e., excluding the tuition problem because the exclusion raised Cronbach’s alpha) along with the standard deviation and mode. Several distributions are quite skewed because many participants on most problems produced no other-side reasons. For Approach A, the most common finding was no other-side reasons generated over the five problems, and the mean was 1.98. The distribution is sharply skewed. Participants had more myside than other-side reasons (Approach B), with a mean of 3.97 of five problems, and the mode was to have more on all five problems. A majority had more myside reasons on four of five problems. Subtracting other-side reasons from myside reasons (Approach C) produced a mean of 6.9 over five problems and a mode of 7. This distribution is roughly bell shaped. The mean of myside reasons divided by all reasons (Approach D) was 0.82, indicating that 82% of reasons generated were myside and 18% were other side. The most common score was 1.0, indicating that for these participants, all their reasons were myside. The majority of scores were 0.8 to 1.0. Finally, when the myside bias was defined as those problems for which an individual produced no other-side reasons (Approach E), the mean was 3.70, and the most common response was 4 on five problems. Thus, the mode was not at the ceiling, with about as many participants showing the myside bias on all five problems as having one or more other-side reasons on three or more problems.

Regarding the correlations and the reliability measures, Approaches A, E, and D are the most appropriate. Although any of these are reasonable, perhaps the best case can be made for Approach E, the myside bias defined as no other-side arguments. Although Approach A had a better Cronbach’s alpha, Approach E had a .90 correlation with both A and D, and Approach E did not exhibit as much of a ceiling effect. More important, defining the myside bias as the absence of any other-side reasons on a given problem is more truly a reasoning bias, whereas Approach A makes greater distinctions at the “unbiased” end of the continuum and Approach D “penalizes” individuals who generate larger numbers of myside reasons. The analyses reported for Study 3 use Approach E; however, comparable results were obtained using other measures.

Acknowledgements

I wish to thank Milena Petrovic, Amy Humphrey, Jon Turpin, and Emma Hogg for their tremendous assistance in collecting and coding the data.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by a grant from the U.S. Department of Education Institute of Education Sciences (R305H020039).

Bio

Christopher R. Wolfe is a professor of psychology and the director of the Brain and Cognitive Sciences Program in the Department of Psychology at Miami University, Oxford, Ohio. He has conducted psychological research on the cognitive processes underlying written argumentation. With M. Anne Britt, he developed a web-based tutor to improve argumentation skills. Other research interests include cognitive technologies, medical decision making, and judgment and decision making.

References

Baron

(1991). Beliefs about thinking. In Voss

J. F.

Perkins

D. N.

Segal

J. W.

(Eds.), Informal reasoning and education (pp. 169-186). Hillsdale, NJ:. Erlbaum.

Baron

(1995). Myside bias in thinking about abortion. Thinking and Reasoning, 1, 221-235.

Baxter Magolda

M. B.

(1992). Knowing and reasoning in college: Gender related patterns in students’ intellectual development. San Francisco, CA: Jossey-Bass.

Britt

M. A.

Kurby

C. A.

Dandotkar

Wolfe

C. R.

(2008). I agreed with what? Memory for simple argument claims. Discourse Processes, 45, 52-84.

Britt

M. A.

Larson

A. A.

(2003). Constructing representations of arguments. Journal of Memory and Language, 48, 794-810.

Charney

Newman

J. H.

Palmquist

(1995). I’m just no good at writing: Epistemological style and attitude toward writing. Written Communication, 12, 298-329.

Harvey

J. T.

(2012). The terrifying Texas GOP platform. Forbes. Retrieved from http://www.forbes.com/sites/johntharvey/2012/07/01/texas-gop-platform/

Hofer

B. K.

Pintrich

P. R.

(1997). The development of epistemological theories: Beliefs about knowledge and knowing and their relation to learning. Review of Educational Research, 67, 88-140.

King

P. M.

Kitchener

K. S.

(2004). Reflective judgment: Theory and research on the development of epistemic assumptions through adulthood. Educational Psychologist, 39, 5-18.

10.

Kitchener

K. S.

King

P. M.

(1981). Reflective judgment: Concepts of justification and their relationship to age and education. Journal of Applied Developmental Psychology, 2, 89-116.

11.

Kuhn

Cheney

Weinstock

(2000). The development of epistemological understanding. Cognitive Development, 15, 309-328.

12.

Macpherson

Stanovich

K. E.

(2007). Cognitive ability, thinking dispositions, and instructional set as predictors of critical thinking. Learning and Individual Differences, 17, 115-127.

13.

Mason

Scirica

(2006). Prediction of students’ argumentation skills about controversial topics by epistemological understanding. Learning and Instruction, 16, 492-509.

14.

Mercier

Sperber

(2011). Why do humans reason? Arguments for an argumentative theory. Behavioral and Brain Sciences, 34, 57-111.

15.

Perkins

D. N.

(1985). Postprimary education has little impact on informal reasoning. Journal of Educational Psychology, 77, 562-571.

16.

Perkins

D. N.

(1989). Reasoning as it is and could be: An empirical perspective. In Topping

D. M.

Crowell

D. C.

Kobayashi

V. N.

(Eds.), Thinking across cultures: The third international conference on thinking (pp. 175-194). Hillsdale, NJ: Erlbaum.

17.

Perkins

D. N.

Bushey

Farady

(1986). Learning to reason. Unpublished manuscript, Harvard Graduate School of Education, Cambridge, MA.

18.

Perkins

D. N.

Farady

Bushey

(1991). Everyday reasoning and the roots of intelligence. In Voss

Perkins

D. N.

Segal

(Eds.), Informal reasoning (pp. 83-105). Hillsdale, NJ: Erlbaum.

19.

Perry

W. G.

(1970). Forms of intellectual and ethical development during the college years: A scheme. New York: Holt, Rinehart & Winston.

20.

Republican Party of Texas. (2012). 2012 state Republican Party platform. Retrieved from http://www.texasgop.org/about-the-party

21.

Reznitskaya

Kuo

Glina

Anderson

R. C.

(2009). Measuring argumentative reasoning: What’s behind the numbers? Learning and Individual Differences, 19, 219-224.

22.

Ricco

R. B.

(2007). Individual differences in the analysis of informal reasoning fallacies. Contemporary Educational Psychology, 32, 459-484.

23.

Schommer

(1990). Effects of beliefs about the nature of knowledge on comprehension. Journal of Educational Psychology, 82, 498-504.

24.

Stanovich

K. E.

West

R. F.

(1998). Individual differences in rational thought. Journal of Experimental Psychology: General, 127, 161-188.

25.

Stanovich

K. E.

West

R. F.

(2007). Natural myside bias is independent of cognitive ability. Thinking and Reasoning, 13, 225-247.

26.

Toplak

M. E.

Stanovich

K. E.

(2003). Associations between myside bias on an informal reasoning task and amount of post-secondary education. Applied Cognitive Psychology, 17, 851-860.

27.

Toulmin

S. E.

(1958). The uses of argument. Cambridge, England: Cambridge University Press.

28.

Voss

J. F.

(2005). Toulmin’s model and the solving of ill-structured problems. Argumentation, 19, 321-329.

29.

Voss

J. F.

Fincher-Kiefer

Wiley

Silfes

L. N.

(1993). On the processing of arguments. Argumentation, 7, 165-181.

30.

Weinstock

(2009). Relative expertise in an everyday reasoning task: Epistemic understanding, problem representation, and reasoning competence. Learning and Individual Differences, 19, 423-434.

31.

Weinstock

Neuman

Tabak

(2004). Missing the point or missing the norms? Epistemological norms as predictors of students’ ability to identify fallacious arguments. Contemporary Educational Psychology, 29, 77-94.

32.

Wilson

(2005). Constructing measures: An item response modeling approach. Mahwah, NJ: Erlbaum.

33.

Wolfe

C. R.

(2011). Argumentation across the curriculum. Written Communication, 28, 193-219.

34.

Wolfe

C. R.

Britt

M. A.

(2003, September). The processing of warrants in written arguments. Paper presented at the 44th annual meeting of the Psychonomic Society, Vancouver, BC.

35.

Wolfe

C. R.

Britt

M. A.

(2008). Locus of the myside bias in written argumentation. Thinking & Reasoning, 14, 1-27.

36.

Wolfe

C. R.

Britt

M. A.

Butler

J. A.

(2009). Argumentation schema and the myside bias in written argumentation. Written Communication, 26, 183-209.

37.

Wolfe

C. R.

Britt

M. A.

Petrovic

Albrecht

Koop

(2009). The efficacy of a web-based counterargument tutor. Behavior Research Methods, 41, 691-698.