Dr. Jon Bruschke

Communication Studies

PO Box 97368

Baylor University

Waco, TX 76798

817-755-1111 ext 6923


























This paper was presented at the 1998 National Communication Association convention in New York, NY.

This research was supported in part by funds from the Baylor University Research Committee.


Work in the field of argument on appellate legal communication has generally not been predictive while work in other fields, notably political science, has been. The purpose of this article is to integrate argument variables into predictive models and compare the relative contributions of each set of variables. An analysis of judicial votes revealed that argument variables were able to predict votes with 63.22% accuracy (n=348), political variables were able to predict votes with 60.76% accuracy (n=414), and a model that combined both argument and political variables was able to predict votes with
















































The field of argument has struggled mightily over definitional and boundary-defining issues. It is certainly not peculiar in this regard; most of the sub-disciplines that locate themselves within the communication rubric have endured torturous debates over their terms, territories, and jurisdictions. The litany of issues argument scholars have engaged is vast and profound (Is argument something you have or something you do? Is argument in people or between people? Does it involve any disagreements or only claims with reasoning?) and as with most definitional disputes there does not appear to be an easy resolution in sight for now or the immediate (and perhaps extended) future. Nonetheless, a few bodies of research have emerged that do show some streams of commonality within the field. First, argument scholars have been concerned with the structure of arguments and have developed systems for mapping out the configurations and development of argumentative components. Stephen Toulmin’s (1958) classic enthymeme-like model is one such effort; more recent schemes that examine the structure of interpersonal arguments have been developed by van Emeren and Grootendorst (1990) and, in separate work, Seibold, Canary, Brossmann, and others (e.g., Brossmann & Canary, 1990; Canary, Canary, Brossmann, Brossmann, & Weger, 1995; Canary, Brossmann, Sillars, & LoVette, 1987; Canary, Weger, & Stafford, 1991). Second, researchers from the interpersonal genre have studied a range of behavior that surrounds disagreements (e.g., Benoit, 1992; Kline & Oseroff-Varnell, 1993; Near, 1994) in order to understand how arguments occur, the motivations of the arguers, or how arguments function within relationships. While the potential exists for these behaviors to be linked to functional and dysfunctional outcomes, at present much of the research is descriptive. Third, research drawn from a wide and wildly disparate critical tradition has offered a host of analyses (e.g., Birdsell, 1991; Hellweg, 1993; Parson, 1993) that are difficult to categorize in any meaningful way. Critics have focused on forums (e.g., Kane, 1995), individual speeches, types of speeches, types of strategies, and types of rhetorical tools and issued a variety of judgments about the effectiveness, morality, and style of public communications. No attempt is made here to comprehensively or conclusively define what the study of argument entails, but as the field has evolved it seems safe to say that it includes, at least, the study of an argument’s structure, content, style, logic, and effect.

Many of these definitional battles have involved questions of what should be the appropriate object of study for the field. Although there is some dispute over what should be included at the periphery of study, examinations of the law and legal practice have always been central to the study of argument. The strength of the link between argument and law is not difficult to understand; argument is the sine qua non of the law and if lawyers do nothing else, they argue. So centrally is law connected to argument that when Stephen Toulmin took on the entire study of logic he did so by maintaining that argument could best be viewed with a "jurisprudential analogy" and essentially argued for a complete re-conception of logic based on legal reasoning. Since Toulmin’s re-conception of argument, a large corpus of argument work has examined issues of the law. Areas of study have included presumption (Alexy 1978/1989; Frankel, 1992; Mann, 1977), the amount of development a legal argument has (Snedaker et al., 1987), the degree to which an argument corresponds to public opinion (Newell & Rieke, 1986), the amount and quality of evidence offered in support of a legal claim (Hiday, 1983; Sheppard & Rieke, 1983), the precedents cited for a legal argument (Benoit & France, 1979; Soeteman, 1987; Songer, 1992), the fact patterns of a case (Aubert, 1963; Kort, 1963, 1966; Nagel, 1983; Rabin, 1978), and other legal variables too numerous even to list.

Despite the impressive gains in the sophistication and insight of these analyses, the scholarship seems schizophrenic in one engaging regard: The research that examines appellate decisions is not predictive, and the research that is predictive does not examine appellate decisions. One vein of research has focused on particular and usually appellate cases and described them in rich detail; the focused nature of this work makes the conclusions difficult to generalize to subsequent cases. Rieke (1987) and Newell and Rieke’s (1986) work in sex discrimination and First Amendment cases respectively provides an excellent case in point. The authors argued that the results of later cases could be found in the dissents of earlier cases, and apparently believe this to be a fairly general statement about legal functioning. While these explanations are rich and describe the cases under study in intriguing ways, scholars searching for insight into other cases remain at a loss. Will all dissents ultimately transform themselves into majorities over time? If not, what sorts of dissents will eventually win the day and which will simply remain as historical footnotes? Presuming that the dissents can transform themselves into majorities, when will the switch occur? How does the political composition of the court influence the transformation of a minority to a majority? It is not difficult to see how these analyses make it difficult to apply the concepts the authors discuss to predict the outcome of any specific case.

A second vein of research has examined a wider array of cases but has tended to focus on a particular type of case or ignored appellate decisions. Hiday (1983), for example, found that expert testimony was a powerful predictor of the outcomes of competency hearings, but the area of law that she examined was so limited it is difficult to generalize the results to other areas. Of course, if one were to expand the scope of review beyond that published in the relatively narrow list of argument-specific journals and publishing outlets it would be easy to discover an overabundance of research on legal communication. The research would encompass studies on a series of argument-related concepts ranging from language style to evidence strength to the ideal sequencing and content of closing and opening arguments. What is conspicuously lacking, however, is attention to argument factors at the appellate rather than trial level. This is not mentioned to demean the trial-level work; indeed, the bulk of legal interaction occurs at the trial level. However, while trial-level law is more common appellate law is clearly more influential and it is, after all, appellate decisions that form the body of precedent for lower courts. A point worth remembering is that the highest courts in any given jurisdiction are appellate courts and usually exclusively appellate courts. The fact that predictive work in this area has been bypassed by argument scholars is an oversight that is glaring.

While argument scholars have neglected predictive analyses of appellate decisions, other fields have taken up the study with vigorous enthusiasm. In fact, since the early 1920s, scholars from political science and sociology (e.g., Adamany, 1969; Allen, 1992; Aubert, 1963; Barber, 1969; Barnum, 1985; Baum, 1977; Dorff & Brenner, 1992; Epstein, Walker, & Dixon, 1989; George & Epstein, 1992; Hall, 1985; Songer, 1992; Ulmer, 1970) and from within the legal field (e.g., Cardozo, 1964; Frank, 1964; Lasswell, 1964; Lawlor, 1964; Nagel, 1973; Schubert, 1977) have sought to explain how judges make decisions and have tried to base predictions on those understandings. The lead article of the Political Research Quarterly written by Hall and Brace (1995) is an interesting example that implies much about current non-argument approaches and their strengths and limitations. Integrating political, case fact, and institutional factors, the authors were able to produce predictions that were roughly 70% accurate in death penalty cases. Furthermore, the authors found that interactions between the factors were necessary to obtain the greatest predictive accuracy. For example, the political environment of a state interacted with the institutional factor of judicial elections such that highly partisan states with elected judges produced the most death sentences. This work is fascinating and innovative in its interactional approach. Still, the enthusiasm should be tempered by what remains to be learned. While 70% accuracy seems a more than creditable starting point, the mode was around 64% in favor of the death penalty, and thus the improved accuracy represented less than a 6-point jump. Most importantly, and as integrated as the model was, it did not incorporate any argument-based variables in the analysis. The omission is typical; despite all the attention the study of law has received surprisingly few studies have ever attempted to measure, categorize, or codify the arguments issued by the lawyers.

The present state is thus one where argument factors might have tremendous explanatory potential in so clearly an argument-laden enterprise as law but they have not been studied in a way that allows for rigorous predictions. Political factors have been examined with predictive approaches but have not incorporated measures of argument. The obvious connection is to incorporate argument variables into predictive models. A few embryonic studies have given reason to believe that this approach offers much promise. Bruschke (1996a) was able to produce models with roughly 65% accuracy by relying solely on measures of argument quality. Although that figure is 5 points lower than the accuracy level obtained by Hall and Brace, it occurs in relation to a mode of 50% and thus is a 15-point improvement over the mode, roughly three times larger than that of Brace and Hall’s integrated model. In addition, "strong" argument predictions were accurate over 80% of the time (Bruschke, 1996b). Furthermore, argument factors might be relatively more important than their political counterparts: When political and argument models produced strong predictions and predicted differing outcomes the argument prediction was correct over 90% of the time (Bruschke, 1996b). Encouraging though this research was, its measures could only be described as rudimentary at best. One measure, for example, was the simple presence or absence of enumeration in the legal brief: Were the points in the brief numbered? Another measure simply counted the total number of citations included in the bibliography. These gross measures of argument were clearly tapping some aspect of argument quality but at the same time were quite primitive and never really examined questions of style.

The purpose of the present paper will be to try to again compare argument and political factors using more refined measures of arguments. While the concept "argument quality" certainly incorporates much more than simply the style of the argument, argument style undoubtedly is a central feature in an argument’s caliber. No less an authority than Aristotle, when introducing the subject in his Rhetoric, commented that: "Our next subject will be the style of expression. For it is not enough to know what we ought to say; we must also say it as we ought…" (335/1984, p. 164). More contemporary scholars have noted a similar centrality of features of style to studies of argument: "Connections between argument and style can be found in places as banal as beginning argument textbooks… places as enduring as the writings of Kenneth Burke… tomes as canonical as the work of Perelman and Olbrechts-Tytecha.. and discussions as recent as [journal] articles… and conference proceedings" (Bruschke, 1995, p. 19). To reiterate, the purpose here is not to define argument (much less argument quality) but to examine one widely accepted component of argument quality, style, in more depth. The value of such work is twofold. First, there is value in studying argument style in its own right and approximating its relative contribution to legal outcomes. Second, studying one particular feature of argument quality may yield insights into the functioning of argument quality writ large.

It is hoped that this analysis will add to a growing body of literature that seeks to explain legal outcomes by incorporating a more comprehensive group of argument factors into the analysis. Simply put, legal outcomes are likely to be the product of both argument and political factors and a full understanding of how argument functions in the legal context will not be complete until much more work that incorporates both sets of explanations is completed.



Political factors were measured in two different ways. First, judicial "ideological values" scores presented by Segal, Epstein, Cameron, & Spaeth (1995) were utilized. The scores measure the degree to which judges are liberal or conservative based on newspaper reports prior to the judges’ appointments. Scores range between +1 and -1 with lower scores indicating conservatism. Segal et al. speak enthusiastically of their measure of judicial attitudes: "our results… point to their strong predictive power, at least for aggregated votes in civil liberties and economic cases" (p. 822). In addition, the presence of amicus curie ("friend of the court") briefs were included as measures of political support for a given position (see Epstein, 1991) and have been empirically linked to litigant success at the Supreme Court level (McGuire, 1995). Amicus briefs are filed by interested third parties to legal cases in support of one of the litigants. The Sierra Club, for example, will often file an amicus brief in an environmental case although it is not a party in the suit.

Conceptually, two different stylistic factors were measured: Brief length and linguistic complexity. Length was measured with simple counts of the number of words, characters, paragraphs, and sentences in the document. Writing complexity was measured with the number of sentences per paragraph, words per sentence, characters per word, and four readily available readability indexes including the Flesch Reading Ease score and the Flesch-Kincaid, Coleman-Liu, and Bormuth grade levels. An additional measure of style was the percentage of sentences that took on passive verb forms. All measures were performed with Microsoft Word word processing software.

Beyond the opening briefs a number of supplemental briefs may be filed with the court, for example, while the appellant may file a reply brief in response to the respondent’s opening brief it is not mandatory that they do so, and as a result some cases have reply briefs and some do not. Since the inclusion of any other briefs would have resulted in a non-standard number of briefs across cases, only the first brief of the appellant and the first brief of the respondent was analyzed.

The argument variables were each entered individually and as relative advantage scores; relative advantage scores were obtained by subtracting the liberal score from the conservative score. For example, the number of amicus briefs filed for the liberal litigant was subtracted from the number of amicus briefs filed for the conservative litigant to obtain the relative advantage score; each individual score (liberal and conservative) was also included in the final analysis.


Case Selection

Fifty Supreme Court cases were randomly selected from the 1988 and 1989 terms. One case was not transcribed into the computer correctly and four cases could not be processed by the word processor due to computer memory limitations. The cases for which grammar data was not available were retained for the analysis of political factors. Following the procedure of Brace and Hall (1989), the individual votes of the judges served as the unit of analysis. Because not all judges participated in all cases fewer than nine votes were cast in some cases; in sum, 414 votes were available for the analysis of political factors and 348 votes were available for the analysis of argument factors.

Dependent Variable

The dependent variable was whether the judges voted for the liberal or conservative litigant. Litigant liberalism was coded according to well-established coding rules described in detail elsewhere (e.g., Brace & Hall, 1995; Bruschke, 1996a). Essentially, liberal litigants are anti-status quo and tend to be underdogs while conservative litigants are pro-status quo, tend to be upperdogs, and typically have more resources at their disposal. For example, a vote in favor of a litigant asserting a civil liberty claim is considered a liberal vote while voting against the civil liberty claim is considered a conservative vote. Any sort of victory for the appellant was scored as an appellant win, even if the decision resulted in only a partial reversal or a remand to a lower court. Trained undergraduate coders identified the liberal litigant in each case and agreed in 43 of 49 cases (88%); the remaining 6 cases were coded by the primary researcher.

Results and Interpretations

The analysis proceeded in three parts. In each part, a discriminant analysis was performed with several independent variables and the single dependent variable of liberal victory. In each part the analysis accounted for the prior probability that liberals would win and each discriminant equation used a stepwise procedure with Rao’s V as the criteria. In the first part, political factors were considered alone. In the second part, argument variables were considered alone. In the final part, both sets of variables were considered together.

Part One: Political Variables

The four political variables (judge ideology, amicus briefs for the liberal litigant, amicus briefs for the conservative litigant, and the difference in the number of amicus briefs for the liberal and conservative defendants) were entered into a discriminant analysis. Liberal litigants won 56% of the time. The analysis produced a single discriminant function (X2 = 8.6, df = 1, p < .005), but the only variable included in the equation was the number of amicus briefs for the conservative litigant. Not surprisingly, a larger number of briefs supporting the conservative litigant was associated with conservative litigant victory. The equation was able to correctly classify 60.76% of the judges’ votes, and thus the equation was able to improve prediction beyond the mode by a slight margin of roughly 4 points.

Part Two: Argument Variables

The eleven measures of argument quality produced 33 variables for analysis; eleven scores for the liberal brief, eleven scores for the conservative brief, and eleven measures of the differences between the briefs for each variable. Liberal litigants won 57.2% of the time. The analysis produced a single significant discriminant function (X2 = 38.5, df = 7, p < .0001). Seven variables were included in the final equation. Table 1 reports the pooled within-groups correlations between discriminating variables and the canonical discriminant function. These correlations represent the bivariate correlation with the discriminant function and are preferred for interpretive purposes (Klecka, 1980); higher scores indicate stronger relationships. In the Part Two analysis positive numbers are associated with liberal victory and negative numbers are associated with conservative victory.

Two patterns are apparent in the data. First, long arguments fare poorly. For both liberal and conservative litigants, measures of brief length were negatively associated with victory. For liberals, the total number of sentences and words was strongly associated with defeat. For conservatives, the total number of words was more modestly associated with defeat. Second, more eloquent briefs appear to fare better. A liberal advantage score on the Bormuth and Flesch-Kincaid grade levels was associated with liberal victory, although much more strongly for the Bormuth measure than the Flesch-Kincaid measure. In addition, dense paragraphs (many sentences per paragraph) were strongly associated with liberal victory while long words (more characters per word) were associated with conservative victory.

The equation was able to predict judges’ decisions with 63.22% accuracy, representing a roughly 6 point improvement over the mode. The 63.22% accuracy figure represents roughly a 3 point improvement over the political model.

Part Three: Argument and Political Variables Combined

The political variables analyzed in Part One and the argument variables analyzed in Part Two were combined into a single analysis. Liberal litigants won 57.2% of the time. The equation produced a single significant discriminant function (X2 = 58.3, df = 9, p < .0001). A total of nine variables were included in the final equation; seven were argument variables and two were political variables. Table 2 reports the bivariate correlations between the variables and the discriminant function. In the Part Three analysis, positive scores favor conservative litigants and negative scores favor liberal litigants. The sole exception is the judge ideology scores; because the judge ideology measures include negative values support for conservative litigants is indicated by negative bivariate correlations. In all cases, high scores indicate strong relationships.

Three trends are apparent in the data. First, the political variables do reasonably well. Although there are only two political variables, they rank first and third in terms of predictive power. Once again, a large number of amicus briefs favors the conservative litigant. Unlike the Part One analysis, however, conservative judge orientation is also associated with conservative victory. Second, the negative effect of brief length was less evident. A relative advantage in the number of paragraphs favored liberal litigants and long paragraphs hurt conservatives. No other measure of brief length was retained in the final equation. Third, the positive effect of eloquence again emerged. Long words favored conservatives, relative advantages in Bormuth grade levels favored liberals, and high scores in the Flesch Reading Ease and Coleman-Liu grade level measures favored conservatives.

The equation was able to predict judges’ decisions with 66.67% accuracy, a roughly 10 point improvement over the mode and roughly a 3 point improvement over the argument equation produced in Part Two.



Substantive Findings

At least two clear patterns emerge from the data and a third trend, while less straightforward, is provocative. First, political support does seem to make a difference. In both analyses in which it was included, the number of amicus briefs filed on behalf of conservative litigants was strongly associated with conservative litigant victory. When included in the final analysis with all other variables the number of amicus briefs emerged as the strongest single predictor. Although the effect was limited to conservative litigants, it should be noted that at the univariate level (not reported herein) the number of amicus briefs filed on behalf of liberal litigants produced significance, and thus its failure to be included in the multivariate analysis may simply have occurred because liberal amicus support shared variance with the conservative measure. The best interpretation is that amicus brief support does influence the Supreme Court and the effect may be somewhat stronger for conservative litigants. In addition, the measure of judicial ideology emerged as a strong predictor in the Part Three analysis, suggesting that a judge’s political orientation does play a role in the judges’ votes. Its failure to produce significance in the Part One analysis is puzzling, but its performance in the Part Three analysis and its predictive utility in prior research suggests that judicial ideology does play in role in case outcomes.

Second, well written briefs do tend to fare better than less well written briefs. For both Part Two and Part Three, relative advantage scores for the Bormuth grade level was associated with litigant victory. In addition, relative advantage scores for the Flesch-Kincaid grade level measure emerged in the Part Two analysis and simple high scores on the Coleman-Liu grade level favored conservative litigants in Part Three. In both Part Two and Part Three long words favored conservative litigants and in both Part Two and Part Three long paragraphs favored liberal litigants and worked against conservative litigants. Simply writing at a higher grade level (and thus in more complex prose) may not unequivocally improve a litigants’ chances, however: At the Part Three analysis, the Flesch Reading Ease measure favored conservative litigants.

Taken together, the findings suggest that (a) all writers should strive to write at higher grade levels, (b) conservatives should use long words, (c) liberals should use long paragraphs while conservatives should use short ones, and (d) writing in an easy-to-understand way slightly favors conservative litigants. The veracity of these findings is difficult to assess, however. Why would the use of long words favor only conservative litigants? Does an extensive vocabulary really assist a conservative advocate and more so than it would assist a liberal? Why do lengthy paragraphs so clearly favor liberals? Perhaps, if paragraph length is an index of the depth of the argument, in-depth arguments favors liberals while less developed (but complexly expressed) presentations favor conservatives. This interpretation may be plausible if liberal arguments tend to probe beyond simple statements of the law and consider underlying purposes and more grand theories of justice than do conservative arguments, which might be based more firmly on clear cut legal rules. For example, if it is true that liberal arguments tend to favor activist court decisions while conservative arguments adhere more strictly to principles of stare decisis, it is reasonable to expect that liberal arguments would need to explore more fluid aspects of the issue more fully to win support. Conversely, conservative arguments might fare better to the extent that they avoid probing the purposes of laws and instead rigidly adhered to "black letter" legal interpretations. In this view, an argument is successful to the extent that the prose reflects the underlying ideology of the argument. When, however, conservatives "play the liberal game" and stray from straightforward legal interpretations they become less persuasive. Similarly, when liberals "play the conservative game" and present their argument in simple, traditional legal syllogistic form they become less successful.

While this explanation is at least coherent, it may be strained. Paragraph length may not indicate the depth of argument or the depth of argument may be unrelated to ideology; conservative arguments may be as complex and purpose-probing as liberal ones. For now, the safest interpretation seems to be that the legal system does favor briefs written at higher grade levels and that stylistic choices may indeed favor liberals and conservatives differently, but any interpretation of how and why those differences exist must at this point take on the posture of interesting and tentative possibilities.

Finally, if simply writing at more advanced grade levels improves the brief (in relation to its chance of winning), how do the contributions of reading ease measures fit into the picture? Is there a simple, linear relationship between the complexity of the prose or at some point does the clarity of the ideas get lost in dense writing? Once again, these questions may weaken the substantive findings. Although the clearest finding in relation to the argument variables is that writing at higher grade levels favors litigants, a useful caveat is that simple complexity at the expense of clarity may not necessarily improve the argument quality. As always, writing is a challenging endeavor and balancing eloquence with clarity is a constant labor.

A third and less clear trend is the possibility that long briefs actually hurt litigants’ chances of winning. When only the argument variables were considered in Part One, brief length as measured by the total number of words and total number of sentences was negatively related to a liberal litigants’ success. The same was true for conservative litigants although the effect was less powerful and only emerged for total word counts. However, in the Part Three analysis, when argument and political variables were combined, the negative effect of brief length was largely absent. In the Part Three analysis, paragraph construction and the total number of paragraphs appeared as more important contributors to litigant victory. Several explanations for these divergent findings are possible. Perhaps the initial effect of brief length was simply spurious. Perhaps the effect of brief length shares variance with political factors, an explanation that is difficult to conceive of theoretically but on the other hand best fits the classic interpretation of regression-type analyses. At any rate, at a purely practical level lawyers would be well advised not to be overly verbose in their legal arguments: While there is some evidence here that brief length can negatively influence the way a judge processes an argument there is no evidence that brief length improves an argument’s chances of winning.

Comparison to Prior Research

There several important similarities between this effort and past research and at least one interesting difference. Initially, the roughly 66% predictive accuracy is almost identical to that obtained by Bruschke (1996a) and Hall and Brace (1995). While it was hoped that more accepted measures of written argument quality might improve predictions over those obtained earlier, such an improvement did not occur. Perhaps the argument and political variables combine to improve predictive accuracy only 10 points and the remaining variance is explained by other variables, such as institutional structures or some interaction between political, argument, and institutional structure variables (see Hall & Brace, 1995). Alternatively, it may still be the case that the measures of argument quality are still relatively crude. Surely there are more subtleties to written arguments than their length, grade level, and passive sentence counts. That question, at present, remains an open one for future research to explore. Much work needs to be done to develop incisive and reliable coding schemes of argument quality.

In addition, the ideological orientation measure did seem to predict judicial performance, although the relative strength of judge ideology in relation to amicus brief support poses an interesting question. Prior literature has been unabashedly proud of the predictive power of judge ideology (Segal, Epstein, Cameron, & Spaeth, 1995) while the influence of amicus support has been more modest (McGuire, 1995). In the present analysis, however, the amicus support variables clearly out-performed the judge ideology measures. While the present findings are not conclusive enough to support the determination that amicus support is the more important factor, it is interesting to note that the present study is the first to compare the two variables directly. An examination of the relative importance of amicus support as opposed to judicial ideology poses an interesting direction for future research.

A final similarity between the present project and past research is that argument variables seem, in general, to out-perform political factors. Argument factors improved predictive accuracy roughly six points above the mode while political factors obtained a gain only two-thirds that large. And while political variables did perform very well in the Part Three analysis, the influence of argument variables may be understated as there were a much larger number of argument measures and a larger number of variables tends to depress predictive accuracy (McLaughlin, 1980). While the current data are more equivocal about the primacy of argument as compared to political variables, prior research (see Bruschke, 1996) more conclusively makes the case for the strength of argument variables. The current data do seem to suggest that the conclusions of earlier efforts may need to be tempered somewhat, but it is apparent that argument factors are at least as important, and potentially more significant, than political influences.

In addition to these similarities, there is one interesting difference between this research and past efforts. Unlike prior research, the findings here suggest an additive contribution of the argument and political variables. Bruschke (1996a) found that either political or argument models could produce predictions that were 65% accurate, but when the variables from the two models were combined into a single analysis the predictive accuracy did not improve. In the present analysis, clear (if modest) gains in predictive accuracy were obtained by combining political and argument measures. Since the predictive accuracy of the combined model is virtually identical to the gains over the mode made by each model separately, the present data show an additive effect that has not emerged in prior research. This interesting pattern and the disparate nature of its occurrence across data sets makes it a provocative area for future inquiry.


While the results do generally comport with other research, there are limitations to this analysis. Although the overall n was sufficient to conduct the statistical tests contained here, it was nonetheless relatively small and based on a fairly limited number of cases. Although the cases were selected randomly, it remains possible that different patterns would emerge had the data set been larger and it is certainly the case that the conclusions presented here could be taken with a much higher degree of confidence if more legal briefs had been examined. In addition, the fact that some cases exceeded the memory capacity of the computer is troubling. Although only four cases could not be processed, it may very well be that the cases that were excluded were substantively different from the cases that we retained, especially in relation to the measure of brief length.

One characteristic of the data that suggests that the cases selected were atypical was the 57% liberal win rate in front of what is widely regarded as a conservative court. Although it is possible that the Supreme Court under study is simply more liberal than is typically assumed, another potentiality is that either the cases that were selected were atypical or that the coding scheme identified liberal litigants who were, in fact, not true representatives of modern liberalism. The data speak to neither possibility, but a more profound question is whether the current coding schemes, employed here and elsewhere, truly capture the political orientations of the litigants. For example, current coding schemes typically assume that all cases have equal political charging and thus litigants can be coded dichotomously as either liberal or conservative. In practice, however, some cases may be deeply political and others much more mundane or procedural. Under current coding systems, for instance, the OJ Simpson trial would code Mr. Simpson as the liberal and assume that the case had the same political import as, say, a run of the mill armed robbery case that made it to the Supreme Court based on jurisdictional questions. Even more troubling is that current coding schemes, for the sake of reliability and simplicity, define liberalism and conservatism by the positioning of litigants in various case types: Divorce-seekers are considered liberals in divorce cases, for example, and defendants are considered liberals in criminal cases. It is easy to imagine, however, instances where these coding rules could distort what researchers consider liberalism. Imagine, for example, a divorce case where a woman divorces her husband because she discovers his medical practice offers abortions, a procedure that offends her religious beliefs. Alternatively, imagine a criminal case where a Republican governor mandates prayer in schools and is accused of violating civil rights laws. At any rate, these possibilities may account for the counter-intuitive liberal win rate and at a minimum serve as a limitation to this research and the volumes of other studies that incorporate similar coding schemes.

The centuries long search to explain judicial behavior will not end soon. Despite the volumes of literature and decades of social science attention our best measures remain crude, and a host of potentially powerful argument factors have not yet been entered into any careful analysis. But scientific progress is measured in small steps and while much remains to be done the present work is of cognizable value. While neither was measured comprehensively, this study was one of the first to examine argument and political factors in light of one another. In addition, the results comport with a growing body of literature that is demonstrating that argument factors play a crucial, if not central, role in determining legal outcomes. Finally, the results are sufficiently encouraging to conclude that more refinement of measures will produce even more impressive research payoffs in the future. The search to understand how various inputs prod and influence the judicial process (and one another) is far from complete, but at least the investigation is now appreciably underway.






Table 1

Pooled within-groups correlations between discriminating variables and canonical discriminant functions for argument variables

Variable Correlation

Number of sentences/Liberal brief -.37

Characters per word/Conservative brief -.29

Bormuth grade level/Difference .26

Total words/Liberal brief -.24

Sentences per paragraph/Liberal brief .20

Total words/Conservative brief .07

Flesch-Kincaid grade level/Difference .04







Table 2

Pooled within-groups correlations between discriminating variables and canonical discriminant functions for Part Three analysis

Variable Correlation

Number of Amicus Briefs/Conservative .38

Sentences per Paragraph/Conservative -.26

Judge Ideology -.25

Characters per word/Conservative .24

Number of paragraphs/Difference -.16

Bormuth grade level/Difference -.21

Sentences per paragraph/Difference -.16

Flesch Reading Ease/Conservative .07

Coleman-Liu Grade Level/Conservative .04


Adamany, D. W. (1969). The party variable in judges' voting: Conceptual notes and a case study. The American Political Science Review, 63, 55-73.

Alexy, R. (1989). A theory of legal argumentation: The theory of rational discourse as theory of legal justification (R. Adler & B. MacCormick Trans.). Oxford: Carlendon Press. (Original work published 1978).

Allen, D. W. (1992). State supreme court voting behavior: A comparative analysis. The Social Science Journal, 29, 31- 51.

Aristotle (1984). Rhetoric (W., Rhys Roberts, Trans.). New York, NY: The Modern Library. (Original work published approximately 835)

Aubert, V. (1963). Conscientious objectors before Norwegian military courts. In G. Schubert (Ed.), Judicial decision making (pp. 201-219). London: The Free Press of Glencoe.

Barber, K. L. (1969). Partisan values in the lower courts: Reapportionment in Ohio and Michigan. Case Western Reserve Law Review, 20, 401-421.

Barnum, D. G. (1985). The Supreme Court and public opinion: Judicial decision making in the post-New Deal period. Journal of Politics, 47, 652-666.

Baum, L. (1977). Judicial specialization, litigant influence, and substantive policy: The court of customs and patent appeals. Law and Society Review, 11, 823-850.

Benoit, P. J. (Ed.). (1992). Interpersonal argument [Special issue]. Argumentaiton and Advocacy: The Journal of the American Forensic Association, 29(2).

Benoit, W. (1989). Attorney argumentation and supreme court opinions. Argumentation and Advocacy, 26, 22-38.

Benoit, W. L., & France, J. S. (1979). Analogical reasoning in legal argumentation. In J. Rhodes & S. Newell (Eds.) Proceedings of the summer conference on argumentation (pp. 49-61). Annandale, VA: SCA.

Birdsell, D. S. (Ed.). (1991). Presidential debating beyond the debate analogy [Special issue]. Argumentation and Advocacy: The Journal of the American Forensic Association, 27(3).

Brace, P., & Hall, M. G. (1995). Studying courts comparatively: The view from the American states. Political Research Quarterly, 48, 5-29.

Brossmann, B. G., & Canary, D. J. (1990). An observational analysis of argument structures: The case of Nightline. Argumentation, 4, 199-212.

Bruschke, J. C. (1995). Deconstructive arguments in the legal sphere: An analysis of the Fischl/Massey debate about Critical Legal Studies. Argumentation and Advocacy: The Journal of the American Forensics Association, 32, 16-29.

Bruschke, J. C. (1996). Explaining judicial decision-making: A comparison of argument and political factors. Paper presented at the Speech Communication Association Convention, San Diego, CA.

Canary, D. J., Brossmann, J. E., Brossmann, B. G., & Weger, H. (1995). Toward a theory of minimally rational argument: Analyses of episode-specific effects of argument structures. Communication Monographs, 62, 183-212.

Canary, D. J., Brossmann, B. G., & Seibold, D. R. (1987). Argument structures in decision-making groups. Southern Speech Communication Journal, 53, 16-37.

Canary, D. J., Brossmann, B. G., Sillars, A. L., & LoVette, S. (1987). Married couples’ argument structures and sequences: A comparison of satisfied and dissatisfied dyads. In J. Wenzel (Ed.), Argument and social criticism: Proceedings of the fifth SCA/AFA conference on argumentation (pp. 475-484). Annandale, VA: Speech Communication Association.

Canary, D. J., Weger, H., & Stafford, L. (1991). Couples’ argument sequences and their associations with relational characteristics. Western Journal of Speech Communication, 55, 159-179.

Cardozo, B. N. (1964). The discovery of the subconscious judicial mind. In G. Schubert (Ed.), Judicial behavior (pp. 14-18). Chicago, IL: Rand McNally & Company.  (Original work published 1921)

Dorff, R. H., & Brenner, S. (1992). Conformity voting on the United States Supreme Court. The Journal of Politics54, 762-775.

Dunbar, N., & Cooper, M. (1981). A situational perspective for the study of legal argument: A case study of Brown v. Board of Education. In G. Ziegelmueller & J. Rhodes (Eds.) Dimensions of argument: Proceedings of the second summer conference on argument (pp. 213-241). Annandale, VA: SCA.

Epstein, L. (1991). Courts and Interest groups. In J. B. Gates & C. A. Johnson (Eds.) American courts: A critical assessment (pp. 335-371). Washington, DC: Congressional Quarterly Press.

Epstein, L., Walker, T. G., & Dixon, W. J. (1989). The Supreme Court and criminal justice disputes: A neo-institutional perspective. American Journal of Political Science, 33, 825-841.

Frank, J. (1964). Law and the modern mind. In G. Schubert (Ed.), Judicial behavior (pp. 19-27). Chicago, IL: Rand NcNally & Company. (Original work published in 1930).

Frankel, D. K. (1992). The Hawai’I Supreme Court: An overview. University of Hawai’I Law Review, 14, 5-16.

George, T. E., & Epstein, L. (1992). On the nature of Supreme Court decision making. American Political Science Review, 86, 323-337.

Hall, M. G. (1985, Summer). Docket control as an influence on judicial voting. The Justice System Journal, 10, 241-255.

Hall, M. G., & Brace, P. (1989). Order in the courts: A neo-institutional approach to judicial consensus. Western Political Quarterly, 42, 391-407.

Hellweg, S. A. (Ed.). (1993). Political campaign debates [Special issue]. Argumentation and Advocacy: The Journal of the American Forensic Association, 30(2).

Hiday, V. A. (1983). Judicial decisions in civil commitment: Facts, attitudes, and psyciatric recommendations. Law and Society Review, 17, 517-530.

Kane, T. (Ed.). (1995). Argumentation and the U.S. Senate [Special issue]. Argumentation and Advocacy: The Journal of the American Forensic Association, 32(2).

Kienpointer, M. (1987). Towards a typology of argumentative schemes. In F. H. V. Eemeren, R. Grootnedorst, J. A. Blair, & C. A. Willard (Eds.) Argumentation: Across the lines of discipline. Proceedings of the conference on argumentation 1986 (pp. 275-287). Providence, USA: Foris Publications.

Klecka, W. R. (1980). Discriminant analysis. Beverly Hills, CA: Sage.

Kline, S. L., & Oseroff-Varnell, D. (1993). The development of argument analysis skills in children. Argumentation and Advocacy: The Journal of the American Forensic Association, 30, 1-15.

Kort, F. (1963). Content analysis of judicial opinions and rules of law. In G. Schubert (Ed.), Judicial decision making (pp. 133-197). London: Free Press of Glencoe.

Kort, F. (1966). Quantitative analysis of fact-patterns in cases and their impact on judicial decisions. Harvard Law Review, 79, 1595-1603.

Lasswell, H. D. (1964). Power and personality. In G. Schubert (Ed.), Judicial behavior (pp. 28-39). Chicago, IL: Rand McNally & Company (Original work published in 1963).

Lawlor, R. C. (1964). Stare Decisis and electronic computers. In G. Schubert (Ed.), Judicial behavior (pp. 492-505). Chicago, IL: Rand McNally & Company. (Original work published 1963)

Mann, R. A. (1977). Get the guest (statute): An analysis of the judicial determinants of their constitutionality. Jurimetrics Journal, 18, 150-185.

McGuire, K. T. (1995). Repeat players in the Supreme Court: The role of experienced lawyers in litigation success. Journal of Politics, 57, 187-196.

McLaughlin, M. L. (1980). Discriminant analysis in communication research. In P. R. Monge & J. N. Cappella (Eds.) Multivariate techniques in human communication research (pp. 175-204). New York: Academic Press.

Nagel, I. H. (1983). The legal/extra-legal controversy: Judicial decisions in pretrial release. Law and Society Review, 17, 481-515.

Nagel, S. S. (1973). Comparing elected and appointed judicial systems (Sage Professional Paper in American Politics 04-001). Beverly Hills, CA: Sage.

Near, M. R. (1994). Argumentative flexibility as a factor influencing message response style to argumentative and aggressive arguers. Argumentation and Advocacy: The Journal of the American Forensic Association, 31, 17-33.

Newell, S. E., & Rieke, R. D. (1986). A practical reasoning approach to legal doctrine. Journal of the American Forensic Association, 22, 212-222.

Parson, D. W. (Ed.). (1993). Dramatism and argumentation [Special issue]. Argumentation and Advocacy: The Journal of the American Forensic Association, 29(4).

Parkinson, M. G., Geisler, D., & Pelias, M. H. (1983). The effects of verbal skills on trial success. Journal of the American Forensic Association, 20, 16-22.

Rabin, D. A. (1978). Gottlieb’s model of rule-guided reasoning: An analysis of Griswold v. Connecticut. Journal of the American Forensic Association, 15, 77-90.

Rieke, R. D. (1987). Evolution of judicial argument in free expression cases. In F. H. vanEmeren, R. Grootendorst, J. A. Blair, & C. A. Willard (Eds.) Argumentation: Across the lines of discipline. Proceedings of the conference on argumentation 1986 (pp. 365-371). Providence, USA: Foris Publications.

Rieke, R. D., & Sillars, M. O. (1993). Argumentation and critical decision making (3rd Ed.). New York, NY: HarperCollins.

Schubert, G. (1977). Political culture and judicial ideology: Some cross- and subcultural comparisons. Comparative Political Studies, 9, 363-408.

Sheppard, S. A., & Rieke, R. D. (1983). Categories of reasoning in legal argument. In D. Zarefsky, M. O. Sillars, & J. Rhodes (Eds.) Argument in transition: Proceedings of the third summer conference on argument (pp. 235-249). Annandale, VA: SCA.

Songer, D. R. (1992). Integrating alternative approaches to the study of judicial voting: Obscenity cases in the US court of appeals. American Journal of Political Science36, 963-982.

Snedaker, K. H., VanCott, Cornwall, & McCarthy (1987). The content and structure of appellate argument: Rhetorical analysis of brief writing strategies in the Sam Sheppard appeal. In J. W. Wenzel (Ed.), Argument and critical practices: Proceedings of the fifth SCA/AFA conference on argument (pp. 315-323). Annandale, VA: SCA.

Soeteman, A. (1987). Deduction in law. In F. H. vanEmeren, R. Grootendorst, J. A. Blair, & C. A. Willard (eds.), Argumentation: Analysis and practice. Proceedings of the conference on argumentation 1986 (pp. 102-118). Providence, USA: Foris Publications.

Toulmin, S. E., (1978). The uses of argument. Cambridge: Cambridge University Press.

VanEmeren, F. H., & Grootendorst, R., (1990). Analyzing argumentative discourse. In R. Trapp & J. Schuetz (Eds.), Perspectives on argumentation: Essays in honor of Wayne Brockriede (pp. 86-106). Prospect Heights, IL: Waveland Press, Inc.