Institutional Design and the Predictability of Judicial Interruptions at Oral Argument

Abstract

Examining oral argument in the Australian High Court and comparing to the U.S. Supreme Court, this article shows that institutional design drives judicial interruptive behavior. Many of the same individual- and case-level factors predict oral argument behavior. Notably, despite orthodoxy of the High Court as “apolitical,” ideology strongly predicts interruptions, just as in the United States. Yet, important divergent institutional design features between the two apex courts translate into meaningful behavioral differences, with the greater power of the Chief Justice resulting in differences in interruptions. Finally, gender effects are lower and only identifiable with new methodological techniques we develop and apply.

Introduction

Increasing attention is being paid to interruption behavior, particularly social norms relating to gender differences: not only are women more likely to be interrupted, but this is true even in contexts where women possess significant power, such as in Congress (Kathlene 1994; Mendelberg et al. 2014) and boardrooms (Dhir 2015). In recent years, scholars have observed these same interruption behavior patterns at U.S. Supreme Court oral arguments. Female Justices are disproportionately interrupted compared to their male counterparts (Jacobi and Schweers 2017; Patton and Smith 2017; Feldman and Gill 2019). A Justice’s ideology and experience also predict interruptive behavior, and interruption rates increase when a Justice is less likely to agree with a colleague or attorney (Jacobi and Sag 2019; Patton and Smith 2020).

Psychologists and sociologists have long found interruptions to be indications of dominance behavior between individuals (Watts 1991; Zimmerman and West 1975). Interruptions, then, have significance beyond social dynamics: if interruptions are reflective – and potentially reinforcing – of power imbalances, then interruptions can limit the contributions of some oral argument participants and promote others. The U.S. Supreme Court has seemingly recognized this significance, changing its structure of oral argument starting in the 2023 October term in response to findings of disproportionate interruptions of women (Deese 2021). Interruptions at apex court oral argument are particularly important because the transparency of oral argument – especially compared to the rest of the judicial decision-making process – contributes to an apex court’s legitimacy, and gender and other disparities at oral argument may harm that legitimacy and moral authority.

Accordingly, increasing scholarly attention is being paid to interruptive behavior in political and judicial institutions as a means of revealing structures of social hierarchy within deliberation. The systematic speech patterns discovered between genders, for example, raise concerns that greater representation of women on the Court, among both Justices and advocates, will not prevent women’s voices from being drowned out (Dhir 2015; Kathlene 1994; Mendelberg et al. 2014).

Despite the importance of the study of interruptive behavior as a proxy for power in political institutions, few studies look outside the U.S. judicial context to examine the institutional design features that might facilitate or constrain oral argument (but see Bentsen et al. 2021; Krehbiel 2016). The paucity of studies of oral argument beyond the U.S. Supreme Court means we have a limited understanding of how institutional design contributes to judicial behavior at oral argument. This lacuna in the judicial behavior literature stands in contrast to the robust study of the role of institutional design on deliberative behavior in legislative politics (Bäck et al. 2014; Goet 2019; Proksch and Slapin 2012) and among scholars of deliberative democracy when designing mini publics (Fung 2003).

This article examines the role of institutional design in shaping judicial interruptive behavior in oral argument. We examine interruptions at oral argument in the Australian High Court, an apex court that exhibits significant similarities in key institutional features to the U.S. Supreme Court (Aroney and Kincaid 2017). Further, a growing body of empirical scholarship has demonstrated that High Court Justices, like their American counterparts, vote in line with ex ante ideological and partisan preferences (Robinson et al. 2022). This is despite the Australian Constitution not containing the entrenched individual rights that are so often at the center of divisive cases before the U.S. Supreme Court. Relevantly, one recent study showed that case outcomes in the Australian High Court are similarly predicable as the outcomes in the U.S. Supreme Court based on oral argument (Jacobi et al. 2022). There is good reason, then, to believe that interruptions at the High Court may also mirror interruptive behavior at the U.S. Supreme Court, at least in the context of ideological interruptive behavior.

Yet the institutional contours of the Australian High Court differ markedly in two significant respects. First, gender is a significant driver of partisan judicial selection (Gramlich 2018) and judicial decision-making (Boyd et al. 2010) across the U.S. federal judicial system. However, in Australia, there is no comparable gender divide in partisan judicial selection – both major parties have selected women at similar rates; or in ideology – female High Court Justices do not tend to be more or less liberal than male Justices at the time of appointment; or in decision-making – there is no significant gender difference in the propensity to make liberal decisions in the Australian High Court (Robinson et al. 2022). Second, in contrast to the U.S. Supreme Court, judicial seniority is exceptionally important in the Australian High Court. Unlike the U.S. Supreme Court, the Australian Chief Justice has unusually powerful levers of control over both case dockets and oral arguments, including choosing the size and personnel of the panel that hears the case (Mason 2007). As such, the High Court is more hierarchical, leading to an expectation of deference to the Chief Justice and potentially less disruptive behavior by the puisne (i.e., Associate) Justices. Alongside the institutional similarities, these institutional differences provide the foundation to examine institutional-level drivers of interruptive behavior.

We analyze all oral arguments in the Australian High Court between 1995 and 2020. We find that institutional design does have an impact on interruptive behavior at the High Court. Specifically, first, in terms of the effect of ideology, Australian interruptive behavior at oral argument looks remarkably like that observed at the U.S. Supreme Court. This is as we as predict, given that the High Court and Supreme Court are institutionally similar in the ideology of judicial appointments and consequent decision-making, but it is contrary to orthodoxy. Second, while seniority is an important predictor of interruptive behavior in the U.S. Supreme Court, again as we hypothesize, the amplified role of the Chief Justice in the Australian High Court leads to significant differences in interruptive behavior vis-à-vis the Chief Justice. Finally, as expected, the relationship between gender and interruptive behavior in the Australian context is present but meaningfully less than in the United States, consistent with the fact that the gender identity of Australian High Court Justices is theoretically less salient than in the U.S. context due to the institutional features of High Court jurisdiction and the related impact on gendered judicial appointments.

The third result requires further explication. Using standard measurement techniques, as predicted, we find no effect of gender on interruptive behavior in the Australian High Court. Although this result is consistent with our own previous findings (Jacobi et al. 2020), it stands in contrast to studies of interruptive behavior across a range of institutional contexts. To be sure that we are fully and fairly testing whether there is a gender element to interruption behavior, we develop a novel methodological approach to thoroughly investigate the predictors of interruptions. Most studies of interruptive behavior control for an individual’s respective volubility, given that a person can only be interrupted if they are speaking (Johnson et al. 2009; Jacobi and Rozema 2018). But this could permit interruption patterns to be masked by differences in speaking patterns between the genders. For example, a potential interrupter may behave differently depending not only on how many words another person has spoken at oral argument but also how long any individual speech event has been going on, systematically reflecting the varying impatience of interrupters at different stages of the argument. Following this theory, we build in a prediction of the likelihood of an interruption every time any person is speaking. Deploying this approach, we not only confirm the strong impact of the other Justice characteristics we found under the traditional approach of controlling for volubility, but we also find a small but statistically significant gender effect. This result is still meaningfully lower than findings from the United States, as we predicted, but when accounting for the differences between the way men and women speak, there is a small gender effect on interruptions at the Australian High Court. Thus, our third hypothesis is supported, but we find there is more nuance than we, or others before us, anticipated.

Our findings make three key contributions to the literature on comparative courts generally and interruptive behavior specifically. First, by examining oral argument in an apex court beyond the U.S. Supreme Court, our findings contribute to the growing comparative judicial behavior literature mapping the landscape of comparative judicial institutions. Second, we show that interruptions are predictable based on varying institutional design features. Where institutional design is similar, judicial interruptive behavior is similar; conversely, where institutional design diverges, so too does judicial interruptive behavior. Third, we make a methodological contribution by developing a novel methodological approach to account for volubility and its interaction with interruptive behavior, one that we argue is better suited to studying and identifying drivers of interruptions generally. Conceiving of interruptions as predictable probabilistic parts of speech has significance for studies of interruptions generally and U.S. Supreme Court oral argument specifically. Using the probabilistic method, we show that longer arguments are associated with more interruptions, despite less competition for airtime in Australian arguments. This directly brings into question the theory that interruptions in the U.S. Supreme Court are a product of time constraints and the harried atmosphere at U.S. oral argument. This is not purely a theoretical contribution. In the 2021 term, the U.S. Supreme Court changed its structure of oral argument to include a stage where each Justice is permitted to question the advocate without interruption. This innovation was developed specifically to reduce interruptions (Deese 2021). As such, our findings have significance for both the study of U.S. Supreme Court argument and reforms happening on the ground.

Oral argument and interruptions in judicial proceedings

Studies of oral argument are almost universally based on analysis of the U.S. Supreme Court. These analyses of drivers of judicial behavior at oral argument are shaped by, and feed into, a debate over the role of oral argument. The traditional, legalistic view sees oral argument as a forum for information-seeking, allowing Justices to directly seek answers from advocates to questions that will help the Justices decide the case and controversy before them (Sullivan and Canty 2015). Oral argument also enables advocates to supplement their written arguments, respond directly to Justices’ queries, and emphasize the most important aspects of their arguments (Johnson 2001). Scholars have unearthed questioning patterns of Justices supporting this judicial information-seeking theory (Benoit 1989; Johnson 2001). For example, McAtee and McGuire (2007) suggest that Justices ask more questions in less salient cases because less external information is readily available and so Justices ask more questions to gather the information necessary to decide how to vote.

The more ideological or strategic view sees oral argument serving as a forum for persuasion among the Justices, who use argument in part to lobby each other and begin coalition building (Black et al. 2013; Johnson 2001, 2004; Johnson et al. 2006). This view has received the concurrence of some Supreme Court Justices, with Justice Breyer commenting, “the Court is having a conversation with itself through the intermediary of the attorney” and Justice Scalia noting, “It isn’t just an interchange between counsel and each of the individual Justices; what is going on is also to some extent an exchange of information among the Justices themselves” (cited in Johnson 2004, p. 61).

Scholars have demonstrated that interruptions affect this information exchange between Justices and counsel and are important for understanding and predicting case outcomes. Jacobi and Rozema (2018) show that interruptions are associated with conflict, such that each interruption between a pair of Justices is associated with a significant decrease in the chances of those Justices voting together in the ultimate outcome of the case. Furthermore, Jacobi and Sag (2019) show that the Justices interrupt the advocates they ultimately vote against at significantly higher rates than they interrupt the advocates they ultimately vote for.

Additionally, scholars have shown that interruptions are driven by individual Justice characteristics. As discussed, female Justices have been shown to be interrupted disproportionately compared to their male counterparts, both by their colleagues and their technical inferiors, the advocates (Jacobi and Schweers 2017). Justices also interrupt female lawyers more extensively and more frequently than male advocates (Patton and Smith 2020). Scholars have also shown that judicial ideology shapes interruptions, with cross-ideological interruptions more common than within-ideological-coalition interruptions (Black et al. 2012; Johnson et al. 2009; Patton and Smith 2020), and systemic differences between ideological camps, with conservative Justices interrupting more and being interrupted less (Jacobi and Sag 2019). Similarly, experience has been shown to be significant, with more senior Justices interrupting more and being interrupted less, though with much lower substantive significance than ideology or gender (Jacobi and Schweers 2017).

However, given that these studies are based in one specific jurisdiction, the extent to which these findings are contingent upon factors in institutional design are unknown. There is a growing recognition among comparative judicial scholars of the critical role of institutional variation in judicial power and behavior (Alarie and Green 2017; Black et al. 2018; Robinson 2013). A small body of work has begun to examine the impact of institutional features on behavior at oral argument, including whether the physical shape of the Supreme Court’s bench reduced interruptions during oral argument (Black et al. 2018) and the impact of online oral argument on judicial behavior during the COVID-19 pandemic (Ringsmuth et al. 2023).

Institutional Design and the Impact on Behavioral Expectations at Oral Argument

In this section, we construct a comparative theory of judicial interruptions according to similarities and differences in institutional design between the Australian High Court and the U.S. Supreme Court. Drawing on the extensive U.S. literature on judicial behavior at oral argument generally, and interruptive behavior specifically, we consider three key institutional design features: first, attitudinal expectations and judicial interruptive behavior, accounting for the mechanism of judicial selection; second, gender-based expectations of judicial interruptive behavior; finally, the role of seniority – with reference to Justice seniority, advocate seniority, the role of the Chief Justice, and case-Justice legal specialization – and related expectations for judicial interruptive behavior.

Institutional design and attitudinal interruptive expectations at oral argument

The politicized nature of judicial appointments in both the United States and Australia leads to an expectation of ideological judicial behavior, including at oral argument. Both the U.S and Australian constitutions grant the executive significant power over the appointment process, providing for strongly politicized apex court compositions. While the U.S. selection mechanism – presidential nomination with advice and consent of the Senate – is frequently pilloried as exceptionally political, the Australian selection mechanism is, in fact, more prone to political capture, as the Australian executive is even less encumbered in choice of nominee, with no legislative confirmation process required (Robinson et al. 2022). As a result, in Australia, judicial selection is said to be the gift of the Attorney General and Prime Minister: “The federal cabinet has a largely unfettered discretion to appoint almost any lawyer they want to the High Court. Nearly every aspect, including timing, candidates, relevant criteria, and consultation process, is secret” (McIntyre 2020). This unencumbered judicial appointment process in Australia is coupled with political cleavages reflective of those in the United States: like the United States, the Australian system features a two-party majoritarian system, with one center-left party (the Australian Labor Party) and an increasingly extreme right-wing bloc (the Liberal-National Coalition) as the predominant party institutions. Further, similar ideological cleavages that divide the two major parties in the United States are applicable in the Australian context, including the representation of women (Sawer 2013) and minorities (Medeiros and Noël 2014); responses to climate change (Crowley 2017); economic management and redistribution (Cameron and McAllister 2022); and the prominence of religion in public life (Medeiros and Noël 2014).

These institutional design elements have led to theoretical arguments and empirical findings that have evinced the impact of ideological division on both oral argument and interruptive behavior in the United States (Jacobi and Sag 2019; Jacobi and Schweers 2017; Johnson 2004; Patton and Smith 2020). Jacobi and Schweers (2017) show not only that ideology was a significant predictor of interruptions but also that conservative Justices were significantly more disruptive than liberal Justices. Jacobi and Sag (2019) confirm that this effect persisted throughout the last six decades, during which conservative Justices dominated the Court’s personnel.

While institutional design suggests ideology should impact oral argument, Australian legal scholarship counters that Chief Justices may use their authority to set the tone of oral argument. Some Chief Justices have favored a strongly Socratic approach (for example, Barwick and Isaacs), whereas others (Dixon and Mason) discouraged cross-examination (Bennett 2007). However, recent research suggests that adversarial debate has been the more common approach since the 1990s; High Court Justices tend to cross-examine the advocates whom they will ultimately vote against (Jacobi et al. 2021), and ideology correlates strongly with judicial decision-making (Robinson et al. 2022) – all of which is the case in the U.S. Supreme Court (Jacobi and Sag 2019). Consequently, the similarity in institutional design and prior findings leads to an expectation that High Court interruptions will be similarly predictable based on ideology.

Ideology Hypothesis A: A Justice’s ideology will be predictive of interruptive behavior.

Given the consistent U.S. findings that conservatives interrupt more and are interrupted less, as discussed (Jacobi and Schweers, 2017; Jacobi and Sag, 2019), we also narrow our test for systematic differences between the two ideological groups to a conservative bias toward interruptive behavior. Jacobi and Sag interpreted the finding that conservative Justices were significantly more disruptive than liberal Justices as likely a product of conservative dominance on the Court for the six decades studied, just as left-wing dominance in speaking seems a reaction to lack of influence. But at the Australian High Court, there has not been such conservative dominance. As such, if the same imbalance occurs in the Australian High Court, not only would this confirm the similarities in ideological behavior between Australian and American jurists, but it would call into question whether the difference found at the U.S. Supreme Court results from the historic ideological tilt of the Court rather than behaviors associated with a particular ideological camps. Therefore, we test the assertion that conservative Justices will drive interruptive behavior in the High Court of Australia.

Ideology Hypothesis B: Conservative Justices will interrupt more and will be interrupted less.

We propose an additional test for conservative bias through the interaction of ideology with the partisan identity of the Chief Justice. As we describe below, High Court Chief Justices are not simply “umpires” in oral argument (Roberts 2005) but important agenda-setting players with powers to set the panels for case hearings. If the ideological dominance theory explains conservative bias, we would expect that a liberal chief Justice would make liberal Justices more likely to interrupt.

Ideology Hypothesis C: Liberal Justices are more likely to interrupt when there is a liberal-aligned Chief Justice.

Institutional design and gender interruptive expectations at oral argument

Although the similarities in the politicization of judicial selection lead to the expectation of ideological behavioral similarities in both Australia and the United States, institutional differences imply that the association between gender and interruptions may be less prominent in the Australian context. The politics of gender has been correlated and causally linked to decision outcomes in the United States, identified in cases where gender is legally salient – for example, in Title VII sex discrimination (Boyd et al. 2010). In the Australian context, in contrast, the question of gender-based rights falls outside the scope of constitutional judicial review. The Australian Constitution does not provide a bill of rights, nor does it guarantee equality before the law. Further, even though statutory claims relating to gender, racial, and disability discrimination, as well as immigration claims involving racial and religious persecution, are within the jurisdiction of the High Court, statutory limits on the ambit of the High Court’s decision-making means these cases are limited in number as well as broader impact.¹

Further, judicial appointments in Australia are not driven by gender politics in the same way as appointments to the U.S. Supreme Court. A long-term goal of U.S. political parties has been to select Supreme Court Justices on the basis of their attitude toward Roe v. Wade 410 U.S. 113 (1973) (Ziegler 2014). There has been no analogous movement to select Justices based on attitudes toward gender-issues in the High Court of Australia, as the High Court does not have the constitutional jurisdiction to review legislation with respect to gender equality. Instead, women’s representation on the Australian High Court is largely bipartisan; both parties have selected similar numbers of women – at the time of writing, the left-wing ALP has selected three women, whereas right-wing Coalition governments have selected four – and there exists no significant difference between men and women in judicial ideology or decision-making (Robinson et al. 2022).

Although studies of oral argument in the U.S. Supreme Court have found female Justices are interrupted more than male Justices (Jacobi and Schweers 2017; Lindom et al. 2017), the limited studies on gender and interruptions in the Australian High Court have mixed findings.² Loughland (2019) used snapshot data to suggest that the same gender patterns emerge in the Australian Court as the U.S. Supreme Court, with female Justices being interrupted at a higher rate than their male counterparts. However, examining all en banc merits hearings – that is, 7-Justice panel decisions, which amounts to approximately 25% of the High Court’s workload – between 1995 and 2019, and using the same methods applied to the U.S. Supreme Court, Jacobi et al. (2021) show that Loughland’s result was a product of the few unrepresentative terms studied and that over a twenty-six year period, there was no such gender effect in Australia, at least during merits proceedings where the Court sat en banc.

The lack of both a gendered partisan divide and important gender-specific cases means we expect significantly less effect of gender on interruptions at High Court oral arguments. But we acknowledge that gendered interruptive behavior does exist across political contexts (Miller and Sutherland 2023; Vallejo Vera and Gómez Vidal 2022) and that volubility among women is reduced in the Australian Parliament (Dowding et al. 2021). Further, a recent study has found that other gendered behaviors, including women speaking less than men, hold equally at Australian High Court oral argument as at U.S. Supreme Court oral argument (Jacobi et al. 2021). Consequently, although we expect we could find some evidence of gender as a driver of interruptive behavior at Australian High Court oral argument, we expect the effect will be significantly less pronounced than at the U.S. Supreme Court. To put our prior results to the most thorough test, we explore the following hypothesis.

Gender Hypothesis: Female Justices and advocates will be interrupted more and interrupt less than male Justices and advocates.

Institutional design seniority-based interruptive expectations at oral argument

Seniority on the bench – referring to years of post-appointment experience – has been consistently shown to be a predictor of judicial behavior at oral argument in the U.S. Supreme Court, including interruptive behavior (Houston et al. 2021; Jacobi and Schweers 2017). With Australian High Court Justices facing mandatory retirement at age 70, there is reason to expect that future discounting would be higher in Australia, or even hyperbolic, compared to the United States, where there is no forced retirement. As such, we expect seniority to be at least as influential on judicial behavior at the Australian High Court. And indeed, recent studies have shown that High Court Justices were more likely to vote according to their ideology as their terms progress (Robinson et al. 2022). More directly, in a study of speech patterns at High Court oral argument, Jacobi et al. (2021) found that senior Justices speak more often and for longer periods than junior Justices. Consequently, we expect relative seniority among Justices to drive interruptive behavior at oral argument in the Australian High Court, at least as much as in the U.S. Supreme Court, where the effect is statistically significant but substantively small. We also expect similar patterns of deference will apply to the interruptions of advocates in oral argument, where senior counsel are more likely to command the respect of the bench and be permitted to speak for longer.

Seniority Hypothesis: Senior Justices and advocates will be interrupted less and will interrupt more.

A separate but related aspect of seniority is the role of the Chief Justice at oral argument. Unlike the U.S. Supreme Court, the Australian Chief Justice has unusually powerful levers of control over both case dockets and oral arguments, including choosing the size and personnel of the panel that hears the case (Mason 2007). Further, the conduct of oral argument partly depends on the preferences of the Chief Justice, with Justices and Chief Justices describing large changes in the level of dynamism at oral argument varying by Chief Justice era (Bennett 2007; Mason 2007). Jacobi et al. (2021) find that the Chief Justice tends to dominate oral arguments more than the other Justices. As such, we expect particular deference to the Chief Justice at Australian oral argument with respect to interruptions.

Chief Justice Hypothesis: Chief Justices will interrupt more and be interrupted less.

Finally, we test situational seniority among Justices – that is, whether or not a particular case sits in a Justice’s field of legal expertise. As the High Court sits in panels, unlike in the U.S. Supreme Court, Justices are more likely to sit on, and be seen as authorities in, cases that match their legal background (Robinson et al. 2022). This aligns with findings from the U.K. Supreme Court that Justice specialization is the primary explanatory factor in determining which Justices sit on which cases (Hanretty 2020). In addition, studies have demonstrated that a Justice’s issue-area expertise is a determinant of voting behavior and amplifies an individual Justice’s ideological propensities (Miller and Curry 2009; Robinson et al. 2022). Although the effect of specialization on oral argument has not been tested in the United States or elsewhere, we posit that Australia is a plausible case to observe legal specialization as a driver of interruptions in oral argument.

Case-Justice Specialization hypothesis: If a case sits in a Justice’s area of legal expertise, they will interrupt more and be interrupted less.

Data and measures

To test our hypotheses, we analyze the transcripts of every High Court oral argument between 1995 and 2020, where the argument on the merits was heard by a bench of two or more Justices.³ Oral argument transcripts from 1995 onward are freely available online and are recorded and compiled by the High Court’s internal reporting service (Howard 2007). The transcripts consistently identify case name(s), case number(s), participating Justices, who is speaking, who is interrupting, and the timestamp for each speech episode.

Dependent variable: Our dependent variable is the number of interruptions, which we examine in five models: Justice interrupted, Justice interrupter, Justice interrupts counsel, Justice interrupts Justice, and advocate interrupted.⁴ Our independent variables comprise individual (Justice and advocate) characteristics. The Justice characteristics variables are drawn from the Australian High Court Justices Biographical Database, which captures detailed biographical information on all Justices sitting on the Court since 1995. The advocate variables are drawn from each oral argument transcript, which identifies each advocate’s gender and seniority.

Justice characteristics variables: Ideology is measured using the Robinson et al. (2022) ideology score, an ex ante measure of judicial ideology generated from the content of newspaper coverage of each Justice for the six-month period prior to swearing in. The score ranges from 0 (most conservative) to 1 (most liberal). Gender is a binary variable, where each Justice and advocate is coded as 1 if female. Seniority denotes the number of years a Justice has sat on the High Court prior to any given case. Advocate experience captures the seniority of the attorney and is measured in three categories: Senior Counsel (SC) or Queen’s Counsel (QC), a rank awarded to the most senior and successful advocates by the relevant state bar association; a barrister or solicitor (denoting a less experienced attorney); or a self-represented litigant. IdeologyxALPCJ interacts the ideology of each puisne Justice (Ideology) with a dichotomous variable indicating whether the Chief Justice was appointed by an ideologically liberal Prime Minister (ALPCJ – that is, a Chief Justice appointed by the liberal Australian Labor Party (ALP)). To measure case-Justice specialization, we use the Robinson et al. (2022) measure of pre-appointment specialization, a hybrid measure of whether a Justice’s legal practice area prior to judicial appointment was in the issue area, or, where a Justice sat on a lower court, more than 50% of the 30 most recent lower court cases seen by a Justice were a match for the issue area of the case in question.

Control variables: Given judicial accounts of meaningful variation in the different eras defined by each Chief Justice’s leadership, we may expect that the impact of each Chief Justice could vary. Accordingly, we use Chief Justice fixed effects. Case- and outcome-level variables are taken from the Australian High Court Database, which codes all panel decisions of the High Court from 1995 to 2020 across 320 variables. Time is a continuous variable denoting the number of hours for oral argument in each case. Panel size denotes the number of Justices sitting on a panel, with the maximum panel size being 7, which represents the full bench of the High Court. Issue area is measured by deploying the issue categorization constructed by Robinson et al. (2022), which categorizes cases into one of six policy areas: economic, public law, criminal law and procedure, traditional common law, civil rights, and procedure and ethics.

Model choice and accounting for volubility. We utilize two different modeling approaches to estimate interruptions at Australian oral argument. First, we use a negative binomial regression to estimate changes in the rate of interruption both of and by Justices at the case level. We use this because it is an appropriate treatment of count data that exhibits over-dispersion (Lindén and Mäntyniemi 2011). In this approach, we control for volubility by including a word count for the total volubility of a speaker in each case. Second, we use a logistic regression model to predict the likelihood of an interruption at the Justice-case-speech level. This method incorporates volubility into the dependent variable itself, as we model changes to the probability of an interruption per speech episode rather than changes to the count of interruptions per case. We include coefficients to capture the effect of the word count of the speaker in each speech episode, as well as the speaker’s cumulative word count, capturing the verbosity of the speaker over the course of the entire hearing up to the point of their current speech.

Results

Predicting interruptions controlling for volubility

In Table 1, we present results from negative binomial regressions modeling the count of interruptions per Justice per case, controlling for Justice volubility. Tests for overdispersion relative to a Poisson regression were positive in each model, confirming that negative binomial regression is appropriate for the present data. Negative binomial coefficients represent a change in a conditional probability and cannot be interpreted directly. However, the positive or negative signs and the statistical significance are both immediately apparent, and we provide translations of the proportional effects of our key variables on the rate of interruptions.

Table 1. Negative Binomial Regressions of Interruptions at Oral Argument

Note: Significance at *p<0.1; ^**p<0.05; ^***p<0.01. Exponentiating coefficients gives the unit change in the rate of interruption per case, holding other coefficients constant.

The results in Table 1 demonstrate that the same Justice characteristics that are highly predictive of U.S. Supreme Court oral argument judicial behavior are also predictive of interruptive behavior at High Court oral argument. Volubility is statistically significant (p<0.01), and the coefficient translates into a substantial effect: holding the other variables at their means, speaking an additional 1,000 words in a case increases the rate of interruption by approximately 40%. We return to the substantive significance of this finding shortly, but it is worth noting that in methodological terms, this result confirms the importance of accounting for volubility at oral argument.⁵

Ideology is statistically significant for each of our models (p<0.01), confirming our Ideology Hypothesis. More specifically, the results also support the Ideological Division Hypothesis, that liberal Justices are interrupted more and interrupt less. Translating the coefficient in terms of model prediction, holding all other variable at their means, shifting from the ideology of a highly conservative Justice to a highly liberal Justice results in a 30% change in the rate of interruptions of a Justice, while interrupting almost 26% less often, in a typical case.⁶ This is consistent with the findings of Jacobi and Sag (2019) in the U.S. context. Since the Australian High Court has not experienced the more than half a century of conservative dominance as has the United States, historical right-wing tilted control of the Court cannot explain the same pattern. The explanation that Jacobi and Sag (2019) resisted – that there is something culturally linking conservatism and interruptions – is harder to rebut given this international similarity arising despite the varied context of ideological control.

Chief Justice is also universally significant and positive. A Chief Justice faces 10% fewer interruptions, even while Chiefs interrupt around 10% more often per case. Together, these results confirm the Chief Justice Hypothesis, since greater interruptions are generally a sign of dominance (Goldberg 1990). But these initial results show little evidence that liberal Justices are less likely to be interrupted when an ALP Chief Justice reigns, due to ideological affinity and the power of the Chief Justice. However, we find somewhat more evidence in support of the Ideology-Alignment Hypothesis in the second approach below.⁷

Seniority is also strongly predictive of interruptive behavior and significant (p<0.01) for the relevant categories (Justice interrupts Justice and Justice interrupts advocate). Each additional year on the High Court is associated with a 2% increase in the rate of being interrupted and a 4% increase in the rate of interruptions made, supporting the Seniority Hypothesis. These results suggest that Justices become more confident with interrupting advocates over time. The fact that these Justices are also interrupted suggests that some of this change may result from interruptions causing interrupting or from exposure to interruptions from speaking more, as explored more in our second methodological approach.

Judicial specialization, contrary to expectations, does not confer any protection against being interrupted, and Justices with specialization in a given field tend to interrupt slightly less frequently than those who do not, contrary to the expectations of this hypothesis.

Gender in models 2 and 3 mimics the results found in the United States with respect to women as interrupters. Women interrupt both fellow Justices and advocates significantly less – around 10% less often per oral argument. Yet, contrary to findings in the United States, the coefficient on female Justices being interrupted – model 1 – is effectively zero, indicating that women are not interrupted more at Australian oral argument. This is consistent with our own prior findings (Jacobi et al. 2020); however, given that we also found that female Justices talk significantly less than male Justices at High Court oral argument, it is possible that controlling for speech may not fully capture differences in activity levels between Justices. To fully interrogate our own prior and current results, we also measure interruptions as a probabilistic effect in any given speech episode below and find different results, consistent with a gender association with interruptions and consistent with our hypothesis of a reduced level when compared with the United States.

We include two case characteristic control variables in Table 1: Hours of Argument and Panel Size. Both are significant in most models, except in model 4, predicting Justice-to-Justice interruptions, which occur far less frequently. Time is consistently positive in all four models, and Panel Size is consistently negative. Interruptions tend to increase with more hours of argument, at around 2% per hour, as there is more time to be interrupted. This sheds light on the impact of limited versus unlimited oral argument: it may not be the case that interruptions at U.S. Supreme Court oral argument are so common because of the strictly limited time, with the Justices fighting to get their points across. Rather, interruptions increase with opportunity – that is, with more speaking time. These results likely indicate that increased debate time reflects increased case complexity, as time for oral argument in each case is negotiated by both parties and the Court once special leave has been granted. All types of interruptions of individuals decrease significantly with increased panel size, a mechanical consequence of increasing the number of debate participants while holding constant how much time is available to speak. The probability of a Justice being interrupted and interrupting are negative, meaning that the larger the panel size, the fewer interruptions of a given Justice that occur: an increase in panel size of one Justice is associated with a decrease in interruptions of between 5 and 6% in models 1, 2, and 3.

Predicting interruptions as a part of speech

Overall, our first approach shows that key Justice characteristics of the type shown to be highly predictive of interruptions at U.S. Supreme Court oral argument are similarly powerful in predicting Australian High Court oral argument, especially where the institutional comparison shows resemblance between the two systems. In particular, not only are ideology and seniority highly predictive, but the direction of ideology – the difference between conservative and liberal Justices – mirrors the direction of the United States despite the historical differences in ideological control of each Court. Yet other effects, primarily the association between gender and interruptions, were not found to be present when using comparable methodology to that typically used to study the U.S. context. But there is reason to pursue this matter further. Although our theoretical expectations point to a low or reduced level of gendered interruptive behavior by virtue of Australia’s constitutional and partisan arrangements, given prior findings that women speak differently to men at High Court oral argument, we must consider that our initial null finding could be driven by those speech pattern differences.

We therefore utilize more precise measures of volubility at a higher level of granularity. Simply controlling for the number of words spoken over the entire oral argument, although standard in interruption models, may fail to uncover gender differences if there are systematic differences in the way that men and women speak at oral argument in the High Court of Australia. For example, there may be differences in words-per-speech between male and female Justices. We confirmed that there are indeed such differences – male Justices speak more often, and women make slightly longer speeches on average (p<0.01).

Accordingly, rather than controlling for how voluble each participant is at oral argument, we now analyze every time a participant speaks, to predict the likelihood of an interruption occurring. While case-level count models of interruptions are statistically appropriate, there are important speech-level characteristics that might be critical to predicting each interruption. We therefore produce a larger N analysis (over 378,184 Justice utterances and 349,376 advocate utterances) to assess how well our proposed theoretical model predicts each interruption. Table 2 presents the results.

Table 2. Speech-Level Logistic Regression of Interruptions at Oral Argument

Note: Significance at *p<0.1; ^**p<0.05; ^***p<0.01.

Volubility could be associated with interruptions in two different ways. First, an interruption could result because a person is giving a particularly long speech, and another grows impatient to have their say. To test this, we include a measure of word count of the current speech unit. Alternatively, or in addition, an interruption could result because a person has talked extensively throughout the given argument, leading another to grow impatient, thinking the first person has more than had their say. To test this, we include a measure of the cumulative word count (i.e., the total number of words spoken by a given person in the case thus far, including the current speech episode). In specifications capturing interrupting, rather than being interrupted, we use a lag of each type of word count to capture the response effect. Table 2 shows the results of this approach.

Under this approach, we ascertain what will increase or decrease interruption levels. Consequently, the baseline (in terms of interruptions) is important. From Table 2, we see that Justices are interrupted infrequently (model 1), approximately one in every twenty speech episodes. In contrast, the base rate for advocates being interrupted (model 5) is significantly higher, at roughly one in four speech episodes. Most interruptions that occur are interruptions of counsel, so the base rate for Justices interrupting anyone (model 2) and the base rate for Justices interrupting counsel (model 3) is like model 5, around one in four. In contrast, Justice-to-Justice interruptions (model 4), the rarest kind of interruptions, occur on average only roughly once in 200 speeches.

All the word count coefficients except one are statistically significant, suggesting that per-speech volubility and case volubility should both be accounted for when predicting interruptions. (All differences in predicted probabilities reported in this section are significant (p<0.05)). Although the effect of volubility has a positive relationship with the likelihood of interruption for the most part, we find in model 1 that if Justices are interrupted (almost always by counsel), they will be interrupted after a relatively short speech, and earlier on in oral argument. Otherwise, we find that the effect of volubility, as expected, is associated with an increase in the likelihood of interruption. Justices are five percentage points more likely to interrupt a speech that reaches 1,000 words than a speech of only a couple of sentences, whereas Justices are only more likely to be interrupted by another Justice than baseline risk after a relatively long speech of around 1,500 words, a very rare occurrence.

The same effect is true for advocates being interrupted (model 5). Short speeches of only a few words are interrupted at a rate of 23%, whereas speeches of 1,000 words are interrupted at a rate of 26%. In terms of cumulative words spoken, advocates early into their argument (around 100 words) are interrupted at a rate of 23%, whereas advocates at 10,000 words are interrupted slightly more often, at a rate of 24%.

Inspecting the results for individual characteristics under the interruptive probability approach, the results for ideology, seniority, and specialization are consistent with the results from Table 1, where we simply control for volubility. In all four models concerned with Justice interruptions (models 1–4), ideology is statistically significant (p<0.01) and substantively meaningful. When looking at the probability of a Justice being interrupted conditional on speaking (model 1), a Justice with a highly conservative ideology score has a likelihood of being interrupted at just over 4%. In contrast, a Justice with a highly liberal ideology score will be interrupted at a rate of almost 7%, a 70% increase in the number of interruptions for a liberal Justice over a conservative Justice. Although the base rate of interruptions in any given speech event is low, this difference translates to a substantively significant expectation of more interruptions of liberal Justices.

A similar effect is captured in model 2 (whether a Justice interrupts). Being a liberal Justice is associated with a significantly lower level of likelihood to interrupt than being a conservative Justice. The most liberal Justices are predicted to interrupt around 21% of the time, whereas the most conservative Justices are predicted to interrupt 24% of the time, mirroring the American results, despite the historical differences in Court dominance. Finally, in Justice-to-Justice interruptions, the most liberal Justices are interrupted by other Justices 0.60% of the time, whereas the most conservative Justices are interrupted 0.37% of the time. Again, the base rate in model 4 is low, but the effect is nonetheless meaningful, with liberals interrupted almost twice as much as conservatives. Thus, when we consider interruptions occurring regularly as part of speech, we find that, as in the United States, Australian conservative Justices interrupt more and are interrupted less, conditional on speaking, as predicted in our Ideological Division Hypothesis. Ultimately, judicial ideology is the most powerful individual characteristic overall in determining interruptions at the High Court.

Being Chief Justice inures a judge from being interrupted by 35%, with a 3.7% rate of interruptions compared to 5.7% for other Justices overall. Justice-to-Justice interruptions occur less often but follow the same trend, at 0.27% compared to 0.61%, less than 45% as often. These results reconfirm the higher status of the Chief Justice. In fact, being Chief Justice is the most powerful predictor in model 3 and 4 and the second most powerful predictor in model 1, behind only judicial ideology, supporting the Chief Justice Hypothesis.

We find mixed evidence of an interactive effect of ideology conditional on the appointing party of the Chief Justice. In models 1 and 4, there is no interaction effect, but in models 2 and 3, it is significant. When predicting interruptions by Justices, we find that there is no difference in the rate for conservative Justices, but very liberal Justices were around 3% less likely to interrupt a speaker when the Chief Justice was appointed by a liberal Prime Minister. We see a similar shift evident in model 3 (Justices interrupting counsel). Thus, the effect here is more distinct than found in Table 1 under our first approach, but the effect only arises for some types of interruptions.

The results for seniority confirm that more experienced Justices interrupt more often, and they are interrupted marginally less often than less experienced Justices, as expected. When it comes to predicting being interrupted, as in the United States, the effect is statistically significant but substantively small. Those just beginning their tenure have a likelihood of interruption at 5.4%, whereas those at the end (at the maximum 20-year mark) have a 4.8% likelihood of interruption. This is significant statistically, but a half a percent difference over a twenty-year span is not striking; this is a very similar result to that found in the United States (Jacobi and Schweers 2017). In contrast, when it comes to predicting Justices interrupting (model 2), seniority is again statistically significant and the magnitude is more substantial. At the beginning of their tenure, new Justices interrupt at a rate 19.7%, whereas after 20 years, Justices interrupt in 26% of their speeches.

In terms of advocate seniority, for advocates being interrupted (model 5), the effect of experience is statistically significant and substantively meaningful. Ordinary advocates (barristers or solicitors) are interrupted around 24.5% of the time, whereas the more experienced and higher status Senior Counsel and Queens Counsel are interrupted less often, at around 23.5% of the time. Self-represented parties are interrupted at a rate of 26%, but this finding is not significantly different from the other two groups, likely because self-representation at the High Court is rare. Altogether, the Advocate Seniority Hypothesis is supported but is arguably less important than other results.

Where we see the most difference between the results using this methodology and the previous approach is that gender is significant in predicting interruptions in all models. This is contrary to both Table 1 and Jacobi et al. (2021), the only other comprehensive analysis of interruptions at Australian High Court oral argument. That study examined only en banc panels; here, using our new methodology and looking at all cases, when a female Justice speaks, she is significantly more likely to be interrupted: from 5.1% for male Justices up to 6.1% for female Justices. But the effect for gender is reversed and yet still statistically significant when looking at whether a Justice interrupts (model 2). All else equal, female Justices interrupt more often than male Justices per speech made, up from 23% for men to 24% for women – a novel finding.

That the effect for gender for model 1 (Justice interrupted) under this approach is statistically significant, but not previously, is not entirely surprising. In this model, each speech is a unit in our analyses; therefore, the number of observations has greatly increased to over 377,000. As such, smaller differences are more likely to show as significant, which makes the question of whether the effect is substantively significant even more pressing. The gender coefficient in model 1 translates to women being just under 20% more likely to be interrupted when speaking. However, the base rate is low, and so the real-life effect is that there is one extra interruption per case per woman on the bench. With each Justice speaking approximately fifty times on average per case, a man could expect to be interrupted 2.55 times, whereas a woman could expect to be interrupted 3.05 times, which is arguably a meaningful difference. When it comes to Justice-to-Justice interruptions, it is a harder case to make that the results are substantively significant: on average, a male Justice will be interrupted by another Justice 0.5% of the time and a female Justice 0.6%, a 20% increase. Since the base rate is low, a female and male Justice pair would have to sit in over ten cases before the model would reliably predict just one extra interjection for the female Justice.

For advocates being interrupted, there is no gender effect, which is normatively a positive result for the Australian High Court. Overall, the results for gender are compelling in their novelty and worthy of note, given their parallel to other experiences of women more generally. However, there are two important caveats. First, some of the effects are small, but they are similar in size to effects reported by others, such as Miller and Sutherland (2023), who report that women are 10% more likely to be interrupted than men in Senate committees, based on similarly sized differences. Second, the effect for women being more likely to interrupt is the strongest result of our gender effects. Thus, it is much harder to argue that being female is a disadvantage in terms of interruptions at the High Court, given the proportionality between interrupting and being interrupted.

Interestingly, length of oral argument measured in hours is not significant in this model, except for the chances that a Justice will interrupt (model 2 and 3). In the other three models, longer oral arguments have no significant difference from shorter arguments. This suggests that although Justices may grow impatient during longer arguments and be more likely to interrupt, other parties do not, and the rate of interruption per speech episode stays unchanged. Methodologically, it also suggests that accounting for volubility via the dependent variable may be a better way to account for significant variation in length of oral arguments.

Conclusion

This article has shown that the Australian High Court Justices display remarkably similar behaviors to their allegedly more political brethren, the U.S. Supreme Court Justices. The institutional differences between the two nations’ apex courts do result in differences; most notably, the additional powers of the Australian Chief Justice appear to impact the behavior of the Chief, who interrupts more than other Justices, and the response of the other Justices, who interrupt the Chief less than they do other Justices. But this strategic response of these Justices is hardly evidence of a less ideological or strategic court – rather, the opposite. And importantly, the most direct measure of political behavior, the effect of ideology, is highly predictive of interruptive behavior at High Court oral argument. Furthermore, interruptions are affected by ideology in the same direction as in the United States: conservative Justices interrupt more and are interrupted less than liberal Justices. This is so despite the lack of conservative dominance in personnel on the High Court. Likewise, advocates were interrupted by the Justices based on similar predictors to the Justices – effects that are probably unsurprising, given the results for judicial behavior, but have never previously been confirmed.

The only reliable predictor of interruptions at the U.S. Supreme Court that applies less clearly to the Australian High Court is the role of gender. Applying the standard mode of analysis, controlling for volubility, female Justices appeared to be no more interrupted than their male counterparts. Once we account for divergences in patterns of speech between men and women, we can discern differences: although the absolute rate of interruptions per case is greater for men than women, when we probe interruption patterns in varied lengths of speeches, women are interrupted more – a small but statistically significant effect. The fact that the effect is both more hidden and a much smaller substantial size is a positive sign, normatively, in contrast to the U.S. apex court, where women are at times interrupted three times as much as men. And Australia is exceptional in one regard: at the High Court, women interrupt significantly more – a very unusual result.

Of importance, this article applies a novel methodological approach to analyzing interruptions based on a rethinking of how interruptions occur. Examining interruptions as a variously probable outcome contingent on speech-level activity exposed effects that were missed by the standard approach of simply controlling for extent of speech. Notably, this approach revealed that disproportionate interruptions of women at the Australian High Court do occur; they are simply less visible and masked by differences between the way men and women speak. But equally important is the fact that some results consistently and strongly arose: judicial ideology was unswervingly significant, was substantively important, and had multifaceted effects as predicted, and the structural power of the Chief Justice was highly influential. These three elements – the contingency of the gender effect, the consistency of the ideological effect, and the power of the Chief Justice – contribute to the predictability of interruptions at the High Court. But each also provides new insight on interruptions at apex courts more broadly, illustrating, respectively, that conceptualizing interruptions as part of speech, rather than speech being an element to control for, can be revealing; that the claim of Australian or American exceptionalism is aspirational, not descriptive; and that how we structure institutions does affect behavior, even though there are consistent trends across different institutions.