Assessing the Effects of Personal Characteristics and Context on U.S. House Speakers’ Leadership Styles, 1789-2006

Research on congressional leadership has been dominated in recent decades by contextual interpretations that see leaders’ behavior as best explained by the environment in which they seek to exercise leadership—particularly, the preference homogeneity and size of their party caucus. The role of agency is thus discounted, and leaders’ personal characteristics and leadership styles are underplayed. Focusing specifically on the speakers of the U.S. House of Representatives from the first to the 110th Congress, we construct measures of each speaker’s commitment to comity and leadership assertiveness. We find the scores reliable and then test the extent to which a speaker’s style is the product of both political context and personal characteristics. Regression estimates on speakers’ personal assertiveness scores provide robust support for a context-plus-personal characteristics explanation, whereas estimates of their comity scores show that speakers’ personal backgrounds trump context.

leadership behavior, such as the homogeneity of preferences within party caucuses; however, systematic measurement of leadership style presents a much more challenging enterprise. Still, one need not accept defeat. "[L]eadership is not an idiosyncratic residual that defies systematic analysis" (Schickler, 2001, p. 15). How, then, can we measure leadership style?
To answer this latest question, we focus on speakers of the U.S. House of Representatives from Frederick Muhlenberg (Federalist-PA) to Dennis Hastert (R-IL). Building on previous work, we seek to address the measurement problem by generating reliable measures of successive speakers' leadership styles by focusing specifically on their respective commitments to chamber comity and their individual leadership assertiveness when performing leadership and presiding roles within the House. 2 Congressional scholars have focused on both aspects of leadership and have often connected one aspect with the other to suggest that the most efficaciousand possibly the most effective-style of House leadership is one that combines a high commitment to comity with assertiveness. 3 For our purposes, a given speaker's commitment to comity involves actions or non-actions that have the effect of promoting chamber civility whereas assertiveness measures the strength and intensity of a speaker's interventions to promote his or the majority party's policy or procedural preferences in relation to legislation or majority control of the chamber (Cooper & Brady, 1981, pp. 411-425;Follett, 1896;Schraufnagel, 2005;Uslaner, 1991Uslaner, , 1993.

Measuring House Speakers' Styles: Commitment to Comity and Assertiveness
Previous research has examined longitudinal differences in successive speaker's behavior but, typically, this work has used small samples over relatively short periods of time and arrayed cases along fairly basic typologies of leadership style, for example, "centralized/decentralized" leadership power, "assertive/cautious" leadership styles, and more or less active (Palazzolo, 1992;Patterson, 1963;Strahan, Gunning, & Vining, 2006;Swift, 1998;Truman, 1959). Other work has arrayed leaders according to ideological "extremism" and/or levels of partisanship (Clausen & Wilcox, 1987;Polser & Rhodes, 1997;Sinclair, 1983). Nuanced measurement of other aspects of speakers' behaviors, however, has received less attention.
Congressional scholars cannot replicate the laboratory conditions that social psychologists may create to monitor and record leadership behavior in a detailed manner. Although denied direct access to our principals and to laboratory experiments, we can emulate some of the methods of social psychologists, which have generated data on leaders' personal characteristics, "at a distance" (Feldman & Valenty, 2001;Post, 2003;Winter, 2003Winter, , 2005. These techniques include using content analysis of leaders' verbal comments or written texts, employing experts to rate leaders on personality scales, and identifying behavioral patterns from political biographies (Christie & Geis, 1970;Hermann, 1980;Post, 2003;Winter, 1992).
A priori, the Congressional Record and predecessor publications might offer an attractive database for content analysis. However, these sources do not allow rigorous analysis for a project spanning 200 or so years. Apart from the simple fact that the Record's predecessors do not provide verbatim records of debate that might, for example, allow for the analysis of word use over time, then as now the record of floor proceedings does not necessarily yield the kind of data we need, although they can be important sources for corroborating speakers' actions or non-actions referred to in secondary accounts and analyses. Indeed, although previous studies of the House speaker have certainly quoted incidents recorded on the House floor, for good reasons, none relies exclusively on this source. Similarly, any analysis of roll call votes on the customary votes of thanks to speakers will not yield the kind of hoped for hard data necessary for systematic analysis. The same is true of data on caucus votes, party nominations for speaker, votes in speakership elections, procedural rulings by the speaker, and appeals against the chair. 4 Notwithstanding the undoubtedly hard status of much of these data, doubts will always remain, moreover, as to whether they effectively capture speaker assertiveness and/or commitment to chamber comity.
Our alternative approach is to rely on careful analysis of the accounts of historians, journalists, and other observers of our 50 speakers. We justify this method on two grounds: First, quantitative data alone are likely insufficient to capture all that we hope to learn about a speaker's commitment to comity and their assertiveness in office. Indeed, rarely have others relied exclusively on quantitative data in either contemporary analyses of recent House speakers or historian's analyses of past occupants of the office (Peters, 1997;Peters & Rosenthal, 2008;Strahan, 2002Strahan, , 2007. Second, while recognizing that a number of historical analyses provide post hoc evaluations of speakers, sometimes attributing only negative characteristics to those subsequently seen as failures and positive features to successful speakers, such biases can be offset by using a wider range of sources for each speaker and exercising careful judgment in avoiding biased accounts. We initiate the measurement of successive speakers' commitment to chamber comity and their personal assertiveness while leading and chairing the chamber by trawling the many biographies and biographic profiles written by historians, journalists, and others who interacted with these leaders on a regular basis. 5 Ultimately, we compiled a single 340-page bibliographic file into which were entered relevant quotations and passages of text taken from the numerous sources, including history monographs, doctoral dissertations, biographies of individual speakers, proceedings of the House from the Congressional Record and its predecessor publications, along with articles in learned journals, newspapers, and magazines. Relevant passages were also entered from key historical analyses of House speakers (Alexander, 1916(Alexander, /1970Brown, 1922;Chiu, 1928Chiu, /1968Follett, 1896;Fuller, 1909;Kennon, 1985;Peters, 1997). Only quotations that focused on a speaker's relevant behavior while occupying the speaker's chair were included in the file. Not surprisingly, the volume and quality of material on individual speakers varied, but in no case were there less than 15 relevant quotations available for any one speaker. 6 Having compiled the biographical file, we subject the passages on each speaker to content analysis and code each leader according to his commitment to comity and leadership assertiveness. Informed by the extant congressional and social psychology literature, we disaggregated each speaker's commitment to comity into four elements: generosity versus meanness of spirit toward other House members; public courtesy versus discourtesy toward other members; agreeableness (including a commitment to reciprocity) versus disagreeableness; and fairness versus unfairness in presiding over the House and in dealing with other members (especially the minority party). 7 In contrast, we define and measure leadership assertiveness along a single scale from strong to weak.
Once the contents of the bibliographic file were established, the two lead authors independently read each passage in the file and coded each speakers' leadership behavior along the five different 7-point Likert-type scales (1-7); four reflecting the multidimensionality of comity and the fifth tapping assertiveness. Particular care was taken to evaluate a speakers' behavior from the chair rather than how such behavior was received on the chamber floor, which might or might not be the result of the chair's action. By this method, comity and assertiveness scores were created by each author for 50 speakers.
Necessarily, evaluating and coding congressional leaders over more than 200 years of history risk applying contemporary behavioral and moral judgments to individuals living in different eras. Thus, it might be argued that late 18thcentury scholars, for instance, might score our subjects differently. However, personality researchers and political psychologists have for a number of years used this general technique across different historical contexts and with acceptable results (McCrae, 1996;Rubenzer, Faschingbauer, & Ones, 2002). A more serious concern with this type of research, however, must be inter-coder reliability (Caruso, 2000). Researching well-known-and not so well-knowncongressional politicians introduces the risk of bias, reputational or political. Certain speakers, such as Joseph Cannon (R-IL) and Newt Gingrich (R-GA), have acquired controversial reputations, whereas others, such as Henry Clay (Whig-KY) and Robert Winthrop (Whig-MA), are widely viewed more benignly. 8 Table 1 shows bivariate correlations between the two authors' first-cut coding of speaker style on the five different scales. In essence, our first coding attempts produce reasonable levels of agreement, the lowest correlation being for the assertiveness scale, but even here, there was a statistically significant association. Reliability analysis (or principal components analysis) on the 10 items (five from each author) extracted two components: the first from the eight responses of the two authors on the first four questions measuring speaker's commitment to comity, and the second from the author's item-response on the fifth scale tapping assertiveness. The eight items loading on the first component explain 60.98% of the inter-item variance, with no single item loading less than .733. Reliability estimates for all 10 items, based on the first-cut coding, yield a Cronbach's alpha coefficient of .92.
Satisfied by the reasonable levels of correlation and the results of the factor and reliability analysis, the authors then engage in a collaborative effort to obtain the "best" score for each speaker on each of the five scales. This "second-cut" effort entailed lengthy discussions on all 500 values assigned (i.e., 2 × 5 scales × 50 speakers) with particular attention paid to discrepancies where scores placed individual speakers on different sides of the scale median. 9 In these extensive discussions, the authors resolved also to reduce the difference between their coding of each speaker on any scale to a maximum of two points. As a result of this collaborative effort, 31.8% of the 500 values were changed, but 63.7% of these changes were by a single digit. The second-cut coding produced bivariate correlations that exceeded r = .85 (p < .001) for each of the five scales. 10 Once the authors had completed their collaborative coding efforts, the average of their scores on each of the five scales was then deemed the "best" value for each speaker.
To verify further the reliability of the authors' best values, three graduate students were hired to read the same 340-page bibliographic file and code the speakers separately along the same five scales. The three students came from two different academic institutions separated by considerable geography. To ensure each student coder thought carefully about the spatial placement of individual speakers along the different scales, and to avoid response set bias, the directions of one of .84** Assertive-Unassertive .47** n = 50 **p < .01. (two-tailed).
the comity scales as well as the assertiveness scale were deliberately reversed. Table 2 shows that the correlations between each student's score and the authors' best score for each scale were statistically significant. Combined with the strong first-and second-cut coding correlations, these additional significant correlations provide us with considerable confidence in the reliability of the authors' scores, which we use in the subsequent statistical analyses.
The possible values for a speaker's commitment to comity range from a low of 4 to a high of 28, and for assertiveness from a low of 1 to a high of 7. With the comity scores, lower numbers originally indicated a greater commitment to comity. These values were subtracted from 28 so that higher numbers would equal more comity. A revised 0 to 24 scale is the result. Now, higher scores indicate, respectively, a stronger commitment to comity and high assertiveness; low scores, toleration or encouragement of uncivil behavior and passivity. Average scores were 13.93 for comity and 4.6 for assertiveness. Comity scores ranged from 2.0 for Gingrich to 21.5 for Nicholas Longworth (R-OH, 1925(R-OH, -1931. Assertiveness ranged from a low of 1.0 for William Pennington (R-NJ, 1860-1861) to highs of 7.0 for James G. Blaine (R-ME, 1869-1875) and Thomas B. Reed (R-ME, 1889-1891 and 1895-1899).
One might expect that an assertive speaker with a strong commitment to comity would be most effective in leading the House, or at minimum, be held in high regard by House colleagues. Conversely, one who was neither committed to comity nor assertive might be less effective and held in less regard. Qualitative evidence from numerous historical sources suggests that the well-regarded Henry Clay falls into the first category (Colton, 1846(Colton, /2004Follett, 1896;Remini, 1991;Schurz, 1893). However, Thomas Reed-not known for his strong commitment to comity and denied the usual courtesy of a unanimous vote of thanks, but highly assertive-has also been judged to be very effective (Brown, 1922;Fuller, 1909;Robinson, 1930;W. H. Smith, 1928). Comparing the two speakers, Alexander (1916Alexander ( /1970 observes, What Clay did as Speaker in the first half of the [nineteenth] century; Thomas B. Reed did it for the last half. He had the courage to suggest parliamentary remedies and the personal force to apply and establish them. Men [sic] recognized him as a pre-eminently able leader . . . and while he lacked the tact and perhaps the wisdom of Clay, evidenced by the disclosure of prejudices and provincial narrowness, he left the Chair a legacy of power . . . (p. 284) It is not clear, then, which combination of comity and assertiveness scores would be most and least efficacious. Further research will be necessary to provide answers, with interesting implications for policy outputs. Figure 1 arrays the 50 Speakers' scores in two-dimensional space divided into quadrants separated by the respective means of comity and assertiveness. A higher proportion of speakers display above-average rather than below-average commitments to comity and just more than a majority demonstrate above-average assertiveness. The most populated quadrants are the bottom right (high comity and low assertiveness) with 16 speakers, including Schuyler Colfax (R-IN, 1863-1869), Sam Rayburn (D-TX, 1940(D-TX, -1947(D-TX, , 1949(D-TX, -1953(D-TX, , 1955(D-TX, -1961, and Carl Albert (D-OK, 1971(D-OK, -1977. In the top left quadrant (high assertiveness and low comity), we also find 16 speakers (including Reed, Cannon, and Gingrich). Just 12 speakers (including Clay, Banks, and Longworth) display the supposed happy combination of high assertiveness with high comity scores, whereas the presumably undesirable bottom left quadrant (low comity, low assertiveness) is occupied by just six speakers, almost all of whom few will recognize.

Exploring Speakers' Comity and Assertiveness: Personal Background and Context
We now move to our hypothesized conceptualization of leadership style as conditional on the context plus the leader's personal characteristics by testing models that draw on legislative socialization and experiential learning theory (Kolb, 1983). A wealth of political socialization literature shows how individuals learn and internalize certain values and norms very early in their lives, whereas the congressional literature shows how most members are assimilated into congressional norms, customs, practices, and procedures, thereby contributing to institutional loyalty and patriotism and helping moderate aberrant behavior (Asher, 1973;Bell & Price, 1975;Bullock, 1976;Clapp, 1963;Davidson, 1990, p. 52;Loomis, 1988;Rosenthal, 1998;Sigel, 1965;Wahlke, Eulau, Buchanan, & Ferguson, 1962). A third body of literature shows how an individual's ability to step on to a political leadership trajectory is determined by the host society's social and political opportunity structure, the individual's social origins, and the self-selection tendencies of certain personality types (Prewitt, 1965;Prewitt & Eulau, 1971). To evaluate our context-plus-personal characteristics interpretation of congressional leadership, we use multiple regression analysis to estimate the extent to which individual speakers' differentiated commitments to House comity and personal assertiveness are related to their pre-legislative careers, prior legislative experience, personal characteristics, party affiliation, and geographical location, as well as to several contextual variables, notably the prevailing partisan context in the chamber during their tenure as speaker.

Dependent Variables
We report the results of four different models, two estimating a speaker's commitment to comity and two their personal assertiveness. In the first of each pair of equations, the speaker is the unit of analysis, with a sample size of 50, and in the second model, the unit of analysis is each Congress. By using two different configurations of our data, we can better manage possible biases arising from the different lengths of service of our speakers. 11 This strategy will also provide more robust tests of the theoretical underpinnings of our regression estimates. The speaker-based data set likely serves to accentuate the relevance of personal background characteristics whereas using Congresses as the unit of analysis will likely advantage the significance of contextual considerations.
In each of the models, our dependent variable is a proportional measure of either comity or assertiveness scores (i.e., the authors' "best" commitment to comity and assertiveness scores for each speaker were divided by the maximum value for each scale). Thus, Speaker Longworth received a proportional commitment to comity score of .875 (his raw score of 21.0 divided by the scale's maximum value of 24), whereas Speaker Gingrich received a score of .083 (2 divided by 24). 12 Given the proportional nature of the dependent variable, beta regression is used instead of ordinary least squares to estimate the models. When each Congress is the unit of analysis, the stationarity of the dependent variable is checked using the augmented Dicky-Fuller test. In both the commitment to comity and assertiveness models, the dependent variables are stationary. The residuals of these models were also checked using the Durbin-Watson test, and there is no significant auto-correlation found.
To contend with the different limitations of our data sets, within the confines of our research design, some compromise was necessary. To accommodate a speaker presiding in multiple Congresses within the speaker data set, we used the mean measure of partisan context across the entire period he occupied the speaker's chair. Using mean values is problematic, particularly for Muhlenberg, Clay, John Taylor (Democrat-Republican-NY), Reed, Sam Rayburn (D-TX), and Joseph Martin (R-MA), whose tenure as speaker was not continuous. We avoid this problem when we run the analysis with each Congress as the unit of analysis. The Congress data set allows the partisan context and certain personal characteristics, such as Relative Age and Years in Congress, to vary across a speaker's time in office. However, in the Congress as the unit of analysis data set, we confront a different problem. In eight different instances, more than one individual served as speaker in a single Congress. We handle this by creating an additional observation for those Congresses (one for each speaker who served), which has the effect of increasing the sample size to 117, although the actual number of Congresses used is 109. Davidson (1990) has noted that although members of the Congress are more alike than different, thereby posing analytical problems when trying to isolate the effects of different backgrounds on members' behavior, that difficulty "should not blind us to the subtle power of background characteristics" (p. 52). Studies have demonstrated how professional education and socialization affect how individuals comprehend the world around them (Larson, 1977). Specifically, Geison (1983) argues that professions and professionals assume a "professional ideology and rhetoric" that shapes the way individuals understand their world. Moore and Rosenblum (1970) observe that professional socialization involves discernible changes in the individual so that he or she becomes part of an exclusive professional group, whereas Shanfield and Benjamin (1985) find that "professional schools . . . exert intense control by purposely influencing beliefs, values and personality characteristics of students; and law schools appear to be the most invasive among all graduate education" (p. 65).

Speakers' Personal Characteristics
These contentions lead to our first hypothesis concerning background characteristics. The high incidence of lawyerlegislators in the Congress is well known and has led to the suggestion that their behavior differs from that of non-lawyer peers (Morgan, 1966;Schlesinger, 1957). "Men who have made a special study of the laws," observed de Tocqueville (1956) of early 19th-century American legislators, "derive from this occupation certain habits of order, a taste for formalities, and a kind of instinctive regard for the regular connection of ideas" (p. 123). More recently, Miller (1995) has suggested, "Because of their legal training and professional socialization, lawyer-politicians seem much more concerned about following proper procedures" (p. 174), and, in a subsequent study, that state senators who were lawyers were more inclined to view themselves as trustees of the public good with high respect for the law and legal processes (Miller, 1993(Miller, , 1995. Extrapolating, we hypothesize that speakers who are Lawyers-in the sense defined by Miller-that is, lawyers with legal training who have actually practiced law for 5 or more years-are more likely to have been instilled with social norms that minimize any proclivity to act on emotional urges in response to popular or partisan demands and are therefore more inclined to adhere to comity norms (Miller, 1995). We also expect these individuals to be more confident and by extension more assertive speakers than their non-lawyer colleagues. A number of studies suggest that lawyers tend to be achievement-oriented, more aggressive, and more competitive than other professionals and individuals in general, and even more so the more time they spend in court (Chusmir, 1984;Houston, Farese, & La Du, 1992). We code speakers with more than 5 years of experience practicing law are coded "1" and anticipate a positive association both with a commitment to comity and assertiveness.
Our second independent variable taps a speaker's prior experience in state legislative politics. We expect a speaker with greater state legislative experience to be more familiar with the principles of lawmaking and more accepting of the practice and decorum of congressional governance, including responsibility for enacting legislation through reciprocity arrangements, compromise, logrolling, coalition building, and so forth (Fenno, 1997;Mason, 1938). Research reported by Berkman (1993) shows that when state legislators enter the U.S. House with considerable legislative experience and prowess, they are better able to adapt quickly and well to chamber life, and are "better prepared in the personal aspects of legislative politics" (p. 77). One House member's comment quoted by Berkman goes directly to comity: "[former state legislators] learn that when you insult somebody you are going to regret it later on." Another observed, "People who have not served do not know how to compromise." We code former State Legislators with more than 2 years of experience "1," and expect positive relationships with our dependent variables. That is, we hold these speakers will have a stronger commitment to comity score and to be more comfortable in their U.S. House role and by extension more assertive. We opt for a plus 2-year cut point, on the assumption that an individual likely served more than a single term and was therefore a member of the state legislature long enough to imbibe at least some of the norms of the legislative process. 13 For similar reasons, we expect a House speaker's Years in Congress to boost both his comity and assertiveness scores. We reason that longer congressional experience induces greater tolerance and indulgence of colleagues, whereas inexperienced or recently elected speakers are more likely to adopt a pace-setting leadership style less regarding of comity. Hibbing (1999) observes how House members acquire greater civility and a greater willingness to compromise over time. Clapp (1963) writes, "If their views regarding their colleagues have changed since they entered the House, the change has been in the direction of greater appreciation of their abilities" (p. 18; see also Loomis 1984, pp. 195-196). By the same logic, we suggest more congressional experience will induce greater confidence in leading the chamber and higher assertiveness scores. Ergo, we anticipate that the earliest speakers-who by definition had little or no experience in the House and, therefore, would not have reached that threshold of parliamentary expertise and comfort that ceteris paribus would give them confidence to be more assertive in the chair-will be associated with lower assertiveness scores. Our measure of congressional experience is simply the number of years served in Congress prior to a speaker's election to the chair. 14 Again, we expect both our comity and assertiveness scores to be associated positively with years of congressional experience.
Third, age stratification theory holds individuals in different age groups tend to share levels of physical energy and mental capacity, ability, and motivations related to their age (Riley, Johnson, & Foner, 1972). This hypothesized effect is conceptually, if not empirically, distinct from the effects caused by longer prior socialization and learning within the chamber itself. Accordingly, we hypothesize that Relative Age will be associated with more comity but less assertiveness in the chair. As Rager (1998) notes, "Uncle Joe" Cannon was 67 when he became speaker, and almost 74 by the time of the 1910 Revolt, by which time he "began to realize that he was no longer 'good old Uncle Joe' to all his colleagues in the House . . . But he did not fully come to grips with the reason for it" (p. 73). Younger liberal Democrats reached similar conclusions about the elderly John McCormack (D-MA) in the late 1960s (see Bolling, 1974). 15 We use Relative Age in each of our models, which we calculated as a speaker's age in years (including fractions of years) divided by the mean age of all House members at the time they began their service as chair of the chamber-thus giving Cannon a relative age value of 1.36 and R. M. T. Hunter (Whig-VA), the youngest member to assume the speaker's chair at the age of 30 in the 26th Congress (1839-1841), a value of 0.71. 16 Again, we anticipate a positive association with comity and a negative association with assertiveness.
Our fifth and sixth personal characteristics variables tap speakers' political-cultural backgrounds associated with the different parties and geographic regions. Writing at the turn of the 20th century, James Bryce (1910) observed that among the Federalists, and their successors the Whigs, and the more recent Republicans, there has never been wanting a full faith in the power of freedom . . . Neither [the Republicans] nor any American party has ever professed itself the champion of authority and order . . . Nevertheless, it is rather towards . . . the Federalist-Whig-Republican Party than towards the Democrats that those who have valued the principle of authority have been generally drawn. (p. 18) Even more directly, Brown (1922) insists that the power of the speakership that led to the 1910 revolt against Speaker Cannon "had been developed almost entirely by the Whig and Republican parties" (pp. 169-170). Similar cultural comparisons between the contemporary parties have been drawn (Freeman, 1986). In very general terms, the Republican Party is noted for a more top-down, hierarchical power structure, is found to be more homogeneous on average, and to display less tolerance toward political differences and the give and take of legislative life. Democrats, in contrast, comprise a more heterogeneous coalition, develop more decentralized organizational structures, and are more likely to be accepting of the values associated with legislative life (Fiorina, 1994;Loomis, 1988). We code speakers representing the Federalist, Whig, and the modern/post-Civil War Republican Party as members of a Pro-Authority Party with a value of "1," and we hypothesize that in presiding over and leading the House, these individuals will demonstrate lower levels of comity and greater assertiveness. 17 Our sixth personal characteristic variable taps the unique regional political culture and style of the South. From one perspective, the South is seen as purveying an aristocratic political culture that represents the epitome of gentility, courtesy, kindliness, and ease, and was embodied in heroes such as George Washington, James Madison, and Thomas Jefferson. Many analysts, however, have rejected what W. J. Cash called this "gone-with-the-windery," instead emphasizing within the region's historical political culture the manifestations of popular religion, the ill-mannered nature and inhumanity of much of Southern politics, and the relationship between violence and honor in early Southern culture, where honor was inextricably connected with physical courage and tenacity, and often with violence (Cash, 1941, pp. 20-21). It was believed that a man who would not defend his liberty and honor had neither (Cash, 1941;Nisbet & Cohen, 1996;Wyatt-Brown, 1986). We entered Southerners as a dummy variable with a value of "1" for speakers who represented any of the 11 states of the former Confederacy. We hypothesize that these individuals will demonstrate, on average, lower commitments to comity and greater assertiveness.
Finally, we enter relative Ideological Extremism scores for each speaker in our models, as evinced in their roll call voting before assuming the speaker's chair. We hold although a speaker's commitment to comity (or civility) is conceptually and empirically distinct from his or her partisanship and/ or ideological extremism, it is of course possible that the latter might influence the former. To provide a robust test for the influence of a speaker's personal background experiences and individual characteristics, we control for his own contribution to partisan and ideological polarization. Because the speaker most often does not vote while in office, we follow King & Zeckhauser (2002) and measure ideological extremeness of a speaker by the proportion of the majority party that has a DW-NOMINATE score more moderate/less extreme than his DW-NOMINATE score in the Congress preceding his election to the chair. 18 We expect a speaker's ideological extremism to be inversely associated with his commitment to comity and positively associated with his assertiveness.

Contextual Partisan/Institutional Factors
The three remaining independent variables tap the partisan and ideological contexts in which successive speakers presided over and led the House. As we noted earlier, many scholars argue that the strength of leadership exercised by a speaker is primarily a function of the majority party's internal homogeneity and the degree of polarization between the majority and minority parties (Aldrich & Rohde, 2000;Binder, 1996;Binder, 1997;Cooper & Brady, 1981;Rohde, 1991;Rohde & Shepsle, 1995;but see Schickler, 2001). According to the conditional party government thesis, when majority party preference homogeneity is high and the two parties are ideologically distinct, the majority party will delegate and centralize greater power in its leader; where these conditions do not obtain, leaders will be less assertive. Similarly, scholars have demonstrated a close correspondence between strong speakers and high inter-party ideological polarization (S. S. Smith & Gamm, 2001).
In her efforts to explain the introduction of new rules and procedures to advantage the majority party-and, by definition, strengthen the majority party leadership's hand-Binder concurs with the conditional party government thesis but opts for a concept of majority party capacity, which is determined by the relative cohesiveness and size of the majority party. We use a similar logic and opt specifically for Schickler's (2001) Ideological Capacity concept as our first contextual consideration. The measure is intended to capture the changing partisan-ideological context in which the speaker presided over the House. Using this measure allows us to separate two distinct political components of House polarization, and thus the context in which successive speakers operated; viz, majority size and majority partisan cohesion (see also Dion, 1997;S. S. Smith & Gamm, 2001). For the Congress data set, we calculated Ideological Capacity as the absolute value of the difference between the first largest party's median DW-NOMINATE score and the floor median and the absolute value of the difference between the second largest party's median DW-NOMINATE score and the floor median. For the speaker data set, we were obliged to use the mean of all values from each Congress in which the speaker occupied the chair. For both data sets, we assume that the majority party held more sway in those Congresses for which "ideological capacity" was high; and we anticipate that a speaker will be more assertive and his commitment to comity will be weaker. 19 Because Schickler's measure of ideological capacity does not directly tap the relative size of the speaker's party, we also enter Majority Size into our equations, although entering this variable yields somewhat mixed theoretical expectations. Contemplating the results of the 1908 House elections and anticipating a victory in the presidential election, future President William Howard Taft wrote to President Theodore Roosevelt expressing a wish for a small Republican majority of "eight or ten votes," so as to limit Speaker Cannon's power (Gwinn, 1957, p. 155). Twenty-six years later, newly elected Speaker Joseph Byrns (D-TN) echoed Taft's concern following the huge Democratic victories in the 1934 House elections, apparently viewing a smaller majority as more amenable to party discipline, and majority leadership influence (Irish, 2001). Dion (1997) goes further than Byrns: smaller majority parties are not only more cohesive but also seek to ensure their rule by reducing the minority party's rights to obstruct (pp. 31-37). Corroborating this finding, Smith and Gamm (2001) cite the assertive Reed's aggressive actions in the 51st Congress (1889-1891) in which Republicans held only an eight-seat majority, to enact new rules that significantly strengthened the majority party's hand and reduced the rights of the minority to obstruct. Speaker Hastert's highly assertive "take no prisoners strategy" during the 107th House (2001House ( -2002, when Republicans had only an eight-seat majority, also supports this proposition (see Owens, 2004, p. 134). The corollary to this partiesat-near-parity-equals-assertive leadership/lower comity is the contrary proposition: that, given a close party split, a prudent speaker may well adopt a less assertive posture and, concomitantly, a greater commitment to comity, in the expectation that such a posture will yield better legislative outcomes for his party, less opposition from the minority, and greater legitimacy for the chair. Thus, Democrats' slim majorities, respectively, in the 72nd (1931( -1933( ) and 78th Congress (1943( -1945 might explain the cautious leadership of John Nance Garner (D-TX) and Rayburn (see Gould & Young, 1998).
In the absence of a strong expectation one way or the other, we nonetheless enter Majority Size as an important control variable and subject our interpretation of its statistical significance to a two-tailed test. Operationally, the variable is defined as the proportional size of the speaker's party in the House (speaker's party size/chamber size) when the speaker was first elected to the chair in the Speaker Model. This consideration is allowed to vary when each Congress is the unit of analysis. 20 Our third and final contextual variable focuses on the partisan configuration of the separated system during the tenure of each speaker. Split Party Government has been a common occurrence in American history, whether in its "pure" or "quasi" forms (see Binder, 1997, pp. 74-75). Clearly, changing patterns of inter-branch relations over time depend to some extent on the degree to which the parties are internally homogeneous, as well as on inter-personal differences between presidents and congressional leaders. 21 Still, partisan and institutional conflict between the different branches of government will be reinforced by the configuration of election outcomes, specifically, whether they produce unified or split party control of government. Partisan pressures-whether they are in support or in opposition to the president-ought to be especially strong in the majoritarian House. Congressional histories repeatedly demonstrate the pressures placed on House speakers to assert their party's interests and policies vis-à-vis the president (Alexander, 1916(Alexander, /1970Follett, 1896;Fuller, 1909).
We contend that when the speaker is from a different party than the president, we should expect institutional competition between the branches to heighten, and within this process, we should expect the speaker-as the leader of the opposition-to be more assertive from the chair. Thus, when Jacksonian Democrats won a majority in the 20th House (1829-1830) 2 years after the Whig candidate John Quincy Adams was chosen president by the House, the new Jacksonian Speaker Andrew Stevenson (VA) boldly appointed majorities of Adams' Administration opponents to all House committees and aggressively "used the power of his office to implement the Jacksonian party's program" (Kennon, 1985, p. 69). The biographer of the assertive Democratic Speaker Samuel Randall notes that he was "at his best" when partisan control was split between a Democratic House and a Republican president and Senate (House, 1935, p. 90). Other scholars have noted the aggressiveness of 19th-century speakers Stevenson, Carlisle, Reed, and Charles F. Crisp (D-GA) and more recent strong speakers, such as Jim Wright (D-TX) and Gingrich, under conditions of split party control (see Barry, 1989;Fuller, 1909;Kennon, 1985;Peters, 1997;Rohde, 1991;Sinclair, 1995). We assign a value of 1 to Split Party Government when a different party from the president's controls the House, and anticipate under these conditions, ceterus paribus, that House speakers will be more assertive and evince a weaker commitment to comity. Tables 3 and 4 provide the results of our analyses. 22 Focusing  on Table 3, which estimates commitment to comity scores, the regression results are similar in both the Speaker and Congress Models. In both model runs, speakers' personal characteristics shine through as the dominant influence on a commitment to comity. Indeed, these effects are so strong they eliminate those from the contextual variables, none of which is significant in either model at the 90% confidence level. Having prior experience as a lawyer and as a state legislator, being from the South, and being a member of a pro-authority party (Federalist, Whig, or Republican) are statistically linked to the variance in speakers' comity scores in both models, whereas the test of the speaker's ideological extremism vis-à-vis his party is signed as expected in both models. However, years in Congress and the relative age of the speaker returned mixed results. Based on this analysis, one must wholeheartedly accept the null hypothesis that these considerations are not associated with a commitment to comity.

Results
Turning to Table 4, the results support a context-plus-personal characteristics interpretation of speakers' personal assertiveness. That is, a speaker's degree of personal assertiveness while presiding and leading from the chair is best explained by a combination of contextual and personal characteristics. In particular, Schickler's ideological capacity measure exerts a strong influence in both assertiveness models. As we anticipated, speakers were more assertive when the ideological balance of power on the House floor shifted toward the median position of the majority party. Split party government is also marginally significant and positively associated with a speaker's personal assertiveness in the By Congress Model (p < .08).
However, as in Table 3, speakers' personal characteristics also have an effect on assertiveness while in the chair, although the combinations of these factors and the overall competence of the models vary some. As Table 4 shows, the test of lawyer and state legislator return somewhat mixed findings (both were easily significant in the comity models). Contrary to the expectations extrapolated from Houston et al. (1992), the signs for lawyer in both models are in the wrong direction-although at the same time somewhat consistent with Chusmir (1984). He contends that lawyers, with the exception of criminal lawyers and judges, are not overtly interested in power. Having been a practicing lawyer engenders a commitment to comity but is inversely related to a speaker being personally assertive from the chair. Having been a state legislator was positively related to a commitment to comity and in the "By Congress" analysis linked to assertiveness. Results for both the Speaker and Congress Models also confirm that those speakers who assumed the chair at an older age relative to the House-such as Pennington and McCormack-tended to be less personally assertive. Relative youth entails greater assertiveness, the classic example being Henry Clay, elected speaker on his first appearance in the House. Last, in both models, we find that being a member of a pro-authority party and length of service before ascending to the chair are positively associated with a speaker's personal assertiveness, albeit years in Congress is only significant at a lower confidence level in the model that uses each speaker as the unit of analysis. 23 Taken together, all four models perform reasonably well. Given the large time span covered by our data and the general purpose of this research, these results are encouraging. Two different data sets, one a cross-sectional analysis and the other a time-series analysis, both demonstrate the importance of personal characteristics in determining speakers' behavior in the chair. Even so, our purpose was to present an empirical case for a context-plus-personal characteristics model of speakership leadership style. Our regression results provide validation of this model for speakers' assertiveness, but not for the variance in these leaders' commitments to chamber comity. No context variable was statistically significant in either of the comity models.

Conclusion
The systematic longitudinal analysis of leadership style has been understudied, partly because of the difficulty in quantifying "style." We began this research with the propositions that both commitment to comity and assertiveness are meaningful components of a speaker's leadership style and that these elements of speaker's behavior are influenced by various personal characteristics reflecting personality, prior socialization, and experiential learning processes, as well as the specific institutional and partisan contexts in which they seek to exercise leadership. The scores obtained from our exhaustive coding process resonate well with received wisdom on the reputations of House speakers. The results from our regression models, which include both personal characteristics and contextual variables, also comport reasonably well with our interactive interpretation of speakers' leadership styles. Context is important to understanding speakership behavior, but so is a speaker's personal background, prior experience, and other individual characteristics; exclusively so, with regard to a speaker's commitment to chamber comity.
In consequence, we now have evidence that House speakers acquire their distinct leadership styles through a variety of iterative developmental processes beginning with prior professional socialization before entering politics and, then, subsequently strengthened as members of state legislatures. Once they become speakers, their personal assertivenessderived from their state legislative experience, relative age, party affiliation, and ideological proclivities-interacts with  and is constrained by the partisan/institutional context in which they preside and seek to provide leadership. 24 The underlying premise of this research has been that context alone is likely an insufficient explanation of leadership style, as it relates to either a commitment to comity or assertiveness. Both the systematic analysis presented here and numerous historical accounts and biographies of House speakers attest to the importance of personality both in the initial selection of a party nominee for the speaker's chair and in his or her performance as presiding officer and party leader. These findings not only have implications for contemporary debate on congressional comity and leader recruitment but also future research on the House. Conceivably, the comity and assertiveness scores developed by this research might be useful as independent variables within extant models of House behavior, including legislative productivity, House-Senate relations, and so forth, over time and in light of the exogenous behavior of the speaker as leader.

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(s) received no financial support for the research and/or authorship of this article.

Notes
1. Indeed, most contextual scholars do not absolutely exclude the influence of leaders' personal characteristics. So, although Cooper and Brady (1981) argue that "Skill cannot fully compensate for deficiencies in the quality or quantity of inducements," they also observe that in the context of decentralized committee government in the mid-decades of the 20th century, "the personal, political skills of the leadership [Rayburn], rather than its sources of institutional power, [was] the critical determinant of the fate of party programs" (pp. 420, 423). Rohde (1991) also acknowledges that leaders' different styles matter, but he insists that "they will remain important [only] as long as the underlying forces that created this partisan resurgence persist . . . '[t] he strategies and tactics of leaders are shaped in large measure by the preferences and attitudes of the members they represent'" (pp. 172, 192, 272). Using principalagent theory, Sinclair (1995) similarly explains Speaker Jim Wright's early leadership style in terms of responding to "[s] pecial political conditions" and "[t]he members expectations," although beyond the confines of her theory, she too concedes a minor role for stylistic differences between different leaders (p. 272). 2. In what amounts to a third approach to understanding House leadership, Green (2010) uses goal theory to posit a much wider range of speakers' goals, including helping his or her party, winning re-election, passage of desired legislation, supporting the president, and defending the House's prerogatives within a highly complex typology, and a complex model that seeks to explain leadership behavior. He argues that the goals value on this variable equal to .97, indicating that 97% of the Republican majority was more moderate/less extreme than Reed prior to his election as speaker. Comparatively, Speaker Gingrich's score is .58, indicating that 58% of House Republicans in the 103rd House were more moderate than he. A speaker's DW-NOMINATE score from the Congress preceding his holding the chair is used as the measure of his Ideological Extremism, but for Muhlenberg, we use his DW-NOMINATE score from the third Congress. For Clay and Pennington, who became speaker almost immediately on being elected to the House, we use, respectively, Clay's score from the 16th Congress and Pennington's score from the 36th Congress. DW-NOMINATE data were obtained from http://www.voteview.com/DWNL.htm. 19. It is possible for this measure to take on a negative value suggesting that the second largest party's median score is actually closer to the chamber median than the first largest party. 20. In certain Congresses (e.g., the third, fourth, 27th, 31st, and 34th), the speaker's party did not have a majority of House members when he assumed the chair. 21. Not all periods in U.S. history are characterized by partisan division-notably, between the 14th and 19th Congresses (1815-1825), from the collapse of the Federalist Party to the so-called "era of good feelings," and during the 32nd and 33rd Congresses (1851-1855) when, as a consequence of the devastation of the Whig Party over slavery, the United States effectively had one-party government. 22. In reporting results, we purposely forego interpretation of coefficients of the "one unit change in x is associated with a z increase in y" variety for the simple reason that our Comity and Assertiveness scales, although rigorously vetted, are not particularly amenable to a precise reading of coefficient values. Hence, we discuss our results in the more general terms of statistical significance. 23. We also ran, but have not reported here, a number of other models using other contextual variables, including measures of majority party preference homogeneity and majority party capacity that do not perform any better in the equations presented in Tables 3 and 4. Similarly, when we break the by-Congress analysis into two periods, pre-and post-Speaker Reed, the contextual variables are not improved in either the comity or assertiveness models. 24. For an earlier exploration of an experiential learning/cognitive framework approach to understanding speakers, see Swift (1998).