Beyond Tokenism: Gender Equality
in the Dutch House of
Representatives
Master Thesis Name: Donna Lippes Student Number: 10331530
Supervisor: mw. Dr. A.M.C. (Anne) Loeber Second Reader: Dhr. dr. G. (Gijs) Schumacher Date: 22-06-2018
Abstract
Gender equality still remains an important topic on the agenda in 2018. Reports that aim to indicate gender equality often do this by displaying the number of women in powerful positions. Literature shows that this is problematic (Kanter, 1977), because the approach fails
to notice qualitative differences in power relations between men and women that have been created throughout history. Therefore, this research goes beyond counting men and women as an indication of equality and instead seeks to analyse gender differences in performances that
relate to dominance and power. In order to increase the impact of the results, this research analyses gender equality in the political sphere, since this resonates with the rest of society
(Easton, 1953). The following research question was formulated: To what extent are there gender differences during political debates in the Dutch House of Representatives in terms of
dominance? Dominance is measured by floor claiming behaviour and language use. Footage of committee debates in the Dutch Parliament that took place between June 2017 and April
2018 was observed and statistically analysed. It appeared that men claimed the floor more often than women, but there were no differences in language use between men and women. It
is concluded that gender norms leave a mark on dominance relations between men and women. However, if women have the autonomy to do so, they seem to break through these
Index
1. Introduction ... 5
2. Theoretical Framework ... 7
2.1 Historical overview ... 7
2.2 Tokenism ... 7
2.3 Social relevance of measuring dominance and power ... 9
2.4 Language and Power ... 10
2.5 Social Relevance of gender equality in the political domain ... 10
2.6 Empirical Approach ... 12
3. Research Questions and Hypotheses ... 13
3.1 Main research question ... 13
3.2 General expectations ... 13
3.3 Sub-questions ... 14
3.3.1 Sub-question 1 – Claiming the floor ... 14
3.3.2 Sub-question 2 – Language Use ... 18
3.3.3 Sub-question 3 – Debate context ... 20
4. Design and Methodology... 23
4.1 Case selection ... 23
4.2 Descriptives of Sample ... 24
4.3 Independent Variables ... 25
4.4 Dependent Variables ... 25
4.4.1 Dependent variables sub-question 1 ... 25
4.4.2 Dependent variables sub-question 2 ... 28
4.4.3 Moderating variables sub-question 3 ... 30
5. Results ... 32 5.1 Data handling ... 32 5.2 Descriptive Statistics ... 32 5.3 Regressions... 33 5.4 Models ... 34 5.5 Results ... 35
5.5.1 Total speaking time ... 35
5.5.3 Average. time predetermined speaking turn ... 36
5.5.5 Conventional interruptions ... 36
5.5.5 Reactions ... 36
5.5.6 Exceeded speaking time ... 36
5.5.7 Received interruptions ... 37
5.5.7 Uncertainty (ad)verbs ... 37
5.6 Interaction Models... 37
5.6.1 Topic of the debate ... 37
5.6.2 Group composition ... 38
6. Conclusion and Discussion ... 39
6.1 Conclusion ... 39
6.6.2 Conclusion sub-question 2 ... 40
6.1.3 Conclusion sub-question 3 ... 41
6.2 Gender norms ... 41
6.3 Discussion ... 42
1. Introduction
Suze Groeneweg became the first woman to be a member of the Dutch Parliament in 1918. Now, hundred years later, it seems the perfect moment to take stock of gender equality in the Netherlands. If one wants to get an idea of the status of gender equality in the Netherlands, it could be best to look at the facts, right? So, for example, if one compares the number of females and the number of males of the first two candidates of each electoral roll list, it will appear that the distribution is nearly half and half (14 males, 12 females). This does not sound too bad. However, how valuable is this information given the fact that the most first
candidates on the electoral roll are male and most second places on the list are filled in by women? Taking this even one step further; how valuable is this information considering that all the party leaders that went to the coalition negotiations (all male) did not take with them the second candidate on the list (mostly female), but a different candidate (all male!) as their deputy (Du Pré, 2017). As a result, women were not able to influence decisions that were taken at the negotiation table. In other words, the meaning of ‘facts’ sometimes seems to change if they are put in a different perspective.
This example attempted to show that gender inequality is often expressed in statistical facts, lacking context. It is one of many examples, as newspaper headlines read: "The number of women in top positions increases too slowly" (de Volkskrant, 2018), "Women still form a minority group in Parliament: currently 51 out of 150 MP's are female" (MetroNieuws, 2017). Even highly rated reports on gender (in)equality often rely on the percentage of women present in particular fields or organizations (e.g. see WEF, 2017; European Commission, 2017; UNESCO, 2014). The objective of this thesis is to look beyond the number of women in powerful positions as an indication of gender equality, and instead seek to analyse the impact of these women. This is considered a better method to measure equality because it takes into account historical relations that assumedly influence current gender norms. In the next chapter, it will be explained how this works based on the idea of tokenism.
The research question will focus on gender equality in the political sphere. In the next chapter, the theoretical framework of this thesis will be set out starting with a brief historical overview of the developments regarding women in the Dutch Parliament. Thereafter, the theoretical foundation will be laid out for how numerically approaches fail at indicating gender equality. Subsequently, alternative approaches will be discussed and it will be
explained how gender equality in the political sphere resonates with the entire society. The third chapter will translate the discussed theory into three research questions that aim to analyse gender differences in terms of dominance in the Dutch House of Representatives (‘De Tweede Kamer’). The fourth chapter discusses how the research questions will be conducted in practice and the fifth chapter will set out the statistical analysis of this thesis. The sixth and last chapter will set out the findings of this thesis and explain how these findings fit into the literature.
2. Theoretical Framework
This chapter starts by providing a brief historical overview of women in the Dutch Parliament. Thereafter, the concept of tokenism will be discussed in order to
explain why it could be misleading to indicate gender inequality based on numbers of men and women present at a certain place. The next paragraph will
discuss alternative approaches to indicate gender equality. Lastly, it will be set out why it is important to measure gender inequality in the political domain.
2.1 Historical overview
As said earlier, Suze Groeneweg was the first elected politician in the Dutch House of
Representatives (DHoR) in 1918, she was a member of the Social Democratic Workers Party (SDAP) and was engaged with the topic of education. Four years later, in 1922, women gained the right to vote during elections (‘Parlement & Politiek’, n.d.). Marga§
±%_ Klompe was the first female minister, she became minister of social work in 1956. Genuine increases in the number of female politicians got off the ground slowly up until around 1980. During the last elections of 2017, there was only one party, the Party for the Animals, with more women than men on their electoral roll. There are currently 51 of the 150 female MPs of which 2 MPs from the GreenLeft party are elected through preference votes (NPO1, 2018).
A paragraph describing the historical context of gender in the DHoR is crucial. However, as mentioned in the introduction of this thesis, these kinds of facts about numbers of women could be misleading. This could be understood by the idea of tokenism and will be explained in the next paragraph. The historical overview was predominantly added to show that the DHoR has known a history where men have for a long time been the only actors. For the remainder, in this thesis, it will be questioned what this means for prevailing gender norms in the Parliament.
2.2 Tokenism
Tokenism helps in understanding that merely presenting the number of present people of in terms of men and women in order to indicate gender equality is a risky approach. Tokenism is
the practice of symbolically recruiting people from oppressed and/or minority groups in order to make a good or diverse impression. George Simmel’s (1950) insights lie at the root of the idea that quantitative denotations of gender equality are problematic. This is because numbers do not contain specific information about social behaviour. Ignoring this has led to false and inaccurate conclusions (Kanter, 1977, p. 965). Kanter expanded this idea by revealing how homogeneous and heterogeneous groups differ in interactions. The number of people of one category compared to the number of people in another category influences the behaviour of the people within these categories. The members of the majority are the ‘dominants’ the members of the minority are the ‘tokens’. The term tokenism assumes that a specific person was token because of certain characteristics (e.g. gender, age, ethnicity, or sexuality). This could be the case, for example, in order to meet certain quota criteria. However, Kanter explains that the term ‘token’ (compared to ‘lone’, ‘solitary’, or ‘lonely’) was chosen because it marks the multidimensionality of the variable. Token groups are groups that carry with them a set of ascribed characteristics. Historical relations of minority and majority groups give meaning to these characteristics, e.g., female party leaders have been rare in the political history of the Netherlands. Therefore, female leaders nowadays are still looked at differently compared to male leaders. A female leader has much chance to be evaluated in relation to her gender, whereas male politicians’ behaviour is more likely to be ascribed to personality. In this perspective, and as Kanter (p. 968) clearly clarifies:
People can thus be in the token position even if they have not been placed there deliberately for display by officials of an organization. It is sufficient to be in a place where others of that category are not usually found, to be the first of one's kind to enter a new group, or to represent a very different culture and set of interactional capacities to members of the numerically dominant category. The term "token" reflects one's status as a symbol of one's kind.
Throughout history, it appears that women are quite new in the Dutch political arena. To say that “it is sufficient to be in a place where others of that category are not usually found” and “to represent a very different culture and set of interactional capacities” is to say that women in the Dutch Parliament encounter norms that are created by men. However, these norms are accepted as the only and standard norms, since women were absent in political debates for a long time. This also causes that behaviour is evaluated differently for men and for women (Heilman, 2001).
could be understood based on the concept of tokenism, it is important to take a look at alternative approaches that go beyond numbers. This will be done in the next paragraph.
2.3 Social relevance of measuring dominance and power
Instead of looking at numbers in order to indicate equality, there is evidence for the
importance to look at actual performances of people during group interaction. By all means, equality is a matter of power, since equality involves a balance of power (Emerson, 1962). Dominance is an important element of power that could be measured in order to make inferences about power relations (Tannen, 1990). Dominance is embedded in performances since people could carry out power-related performances. Power is defined by Henley (1977) as being in control of resources. The definition of these resources may differ per situation. For example, in some situations money leads one to power, in other situations, this might be knowledge.
Sylvia Shaw (2000) has put these theoretical ideas of measuring gender equality in terms of power relations into practice. In her study, she looked at power differences between male and female MPs in the House of Commons. Controlling the floor was considered the power resource. In the House of Commons, rules are designed in a way that every MP gets the same chance to control the floor. However, in practice, rules are sometimes violated. In this
perspective, political debates become situations in which power relations and pre-existing inequalities (e.g. gender inequality) can be exposed (Edelsky & Adams, 1990, p. 171). The data Shaw used consists of videotapes of these political debates. She looked at the behaviour of MPs regarding taking and giving away speaking turns. For example, during a speech of an MP, other MPs can indicate that they want to interrupt. The current speaking MP can decide whether he or she wants to give away the speaking turn. Subsequently, if an MP gives away a speaking turn, this speaker can decide if he or she reacts on or ignores the interruption of the other MP. Shaw has mapped these speaking turn decisions. She found that male MPs claimed more speaking turns than female MPs and had more power in the political debates. By far, most of the illegal interruptions were conducted by male MPs. Shaw concludes, in line with Kanter’s ideas, “that masculine discourse styles are treated as the interactional norm in debates and that this relates to the fact that traditionally women have not been represented in this institution, and continue to be under-represented” (Shaw, 2000, p. 416).
2.4 Language and Power
As yet, we have learned that power in political debates is symbolized by control over the floor. This could be in terms of claiming speaking turns and thereby speaking time, but power and dominance could also be exercised in linguistic expressions. Rowbotham (1973, in Elshtain, 1982, p. 603) explains: “Language conveys a certain power. It is one of the instruments of domination”. Literature demonstrates that groups can use certain strategies relating to speech style in order to preserve a powerful position (Coates, 2015). In practice, these speech styles are for example related to the use of sarcasm (Ducharme, 1994), filler words (or vocal pauses, i.e. “uhm”) (Laserna, Seih, & Pennebaker, 2014) , uncertainty verbs (e.g. perhaps, seems), intensive adverbs (e.g. very, really) (Pennebaker, Mehl, &
Niederhoffer, 2003) and there are many more elements that can shape a speech style (see for examples Llisterri, 1992).
So far, it is discussed that in order to examine gender equality, it is useful to look beyond language and to observe variables that give a more accurate indication of equality, namely power, or dominance. It appeared that this could be obtained by controlling resources. Previous research on gender equality during political debates demonstrated that claiming the floor represented the power resource. Other research showed that power relations could also be embedded in speech styles. This thesis will combine both claiming the floor and speech styles. No research has yet taken into account both factors in order to make inferences about dominance/power.
Group interactions take place everywhere, the next paragraph will discuss why it is relevant to focus on group interactions within the political domain.
2.5 Social Relevance of gender equality in the political domain
Previous studies have addressed gender differences concerning dominance and power during group interactions (see for examples Hawkins & Power, 1999; Sidanius, Pratto, Bobo, 1994). A smaller part of these studies focusses on political debates in specific. Shaw’s analysis is one of these researches. In this paragraph, it will be explained why it is highly relevant, if looking at gender equality in society, to examine gender equality in the political domain.
The DHoR is the highest political platform in the Netherlands. Gender equality seems crucial since practices that take place in the Parliament are of great significance for the rest of
society. In order to understand this, one needs to understand that Easton (1953) describes politics as the authoritative allocation of values for a society as a whole. Easton sees politics not as a self-contained concept but continuously in relation with its surroundings; with society as a whole. In the model, he distinguishes input, decision-making, output, and feedback. These four constructs are related to each other as could be seen in Figure 1.
Input could be seen as ways in which the average citizen engages in political life. This does not imply that every citizen has to be actively involved in politics, it includes any occurrence or feeling of citizens in relation to politics. For example, distrust in politics. This is input is related to decision-making. Easton’s conception of decision-making is commonly known as ‘politics’, however, Easton sees politics as an encompassing system and decision-making is just a part of this system. Decision-making encloses the practices that take place in, for
example, town halls or Parliaments. In case of this thesis, the Dutch Parliament in The Hague. Output could be seen as the results of the decision-making process, this includes all the
consequences of the decision-making part. For example, tax reduction. Next off, the circle starts once again, because output goes into input through what Easton calls feedback.
Feedback is the way in which civil society reacts to outcomes. Based on these reactions new inputs to the decision-making process are created. This generates a balanced reciprocal feedback loop.
In short, developments within the decision-making domain influence society and
developments within societal domain influence decision-making processes. This means that gender equality within the Parliament can also induce gender equality in civil society and the other way around. In the Dutch case, one has the national government on the one side and the civil society on the other side of this relationship. This exemplifies a relation between a group of 150 people and a group of 17 million people. Taking this into consideration, the
importance of this relatively small group of 150 MPs appears, since the reciprocal relation is a balanced one. Therefore, examining gender equality in the DHoR is paramount.
Figure 1. A schematic depiction of politics as conceptualized by Easton (1953).
2.6 Empirical Approach
In this thesis, it will be questioned to what extent gender differences exist in terms of dominance during political debates in the DHoR. Understanding the concept of tokenism helped in understanding that measuring gender equality in amounts of men and women can give a distorted depiction. A certain method overlooks social processes that take place during group interaction and reveal unbalanced historical relations. Subsequently, it is discussed that measuring behaviour relating to power and dominance during group interactions is a more accurate indicator of gender equality. Shaw’s research formed an example of how this is put into practice. From Easton’s idea of politics, it appeared that it is useful to measure gender equality in the political domain because of its strong reciprocal relationship with society. Within the political domain, the floor acts as a power resource. Other research demonstrated that power could also be embedded in speech styles. The perspectives of the previous chapter come together in the research question of this thesis. It is a unique research question in the sense that it is the first time that claiming the floor and language use as manners of indicating power are combined in one study in order to measure gender equality during political debates. It is also the first time that suchlike research will be conducted in the political domain of the Netherlands.
3. Research Questions and Hypotheses
This chapter translates the theoretical discoveries as discussed in the previous chapter into the main research question and 3 sub-questions and corresponding hypotheses. This thesis will examine gender differences in terms of dominance inthe DHoR. The first sub-question concerns a similar question to that of Shaw (2000), but will address gender differences in claiming the debate floor in the Netherlands, whereas Shaw did the same for the British House of Commons.
Sub-question 2 will add the use of language as a behaviour in which dominance is embedded. The third sub-question will examine the role of the context of the debate since the theory of tokenism includes the idea that group composition can influence group interactions. The last question also covers the possible effects of
the topic of the debate.
3.1 Main research question
Based on the above theoretical findings, the following research question has been formulated: RQ: To what extent are there gender differences during political debates in the Dutch House
of Representatives in terms of dominance?
3.2 General expectations
In general, the hypothesis is guided by the expectation that men are more dominant during political debates. This is because, for a long time, politics have been entirely dominated by men. This has left its marks (Shaw, 2000; Kendall & Tannen, 1997). The ways of interaction within Parliament have established in the absence of women. Men have invented the
interactional norms that over many years have been embedded and institutionalized in the system. Research even points in the direction that men’s language use is “not only seen as the better way to talk, but as the only way” (Lakoff, 1990, p. 188). These masculine norms are related to enforcing authority and legitimacy of the particular group that acts according to these norms (Gal, 1991). So naturally, is assumed that it is easier for men than for women to find their way in the masculine norms that prevail in political debates and subsequently, that they are more dominant than women. This finding is the main guideline for the hypotheses.
3.3 Sub-questions
The main research question will be divided into three sub-questions. The first sub-question will take up the issue of claiming the floor. The second sub-question will take up the issue of speech styles regarding dominance differences. The last sub-question aims at giving more understanding of whether possible gender differences found in sub-questions 1 and 2 are strengthened by the context in which they take place.
In the method section, it will be explained that footage of political debates from the Dutch Parliament will be watched. A pilot was conducted in which some footage was observed in order to get an idea of kind of interactions that take place during debates. This helped in making decisions regarding practicability. There are numerous behaviours that enclose power and dominance, however, many do not apply to the interactions that take place in during political debates. During the pilot, all kind of behaviours were monitored and it was checked whether the behaviour fits into the literature on power and dominance.
3.3.1 Sub-question 1 – Claiming the floor
The first sub-question considers claiming the floor. There different manners for MPs to claim the floor and this is obviously a way to obtain more speaking time. Measuring speaking time is an important facet of measuring dominance during debates in general. Bales, Strodtbeck, Mills, and Roseborough (1951) found that within a group conversation, dominance
hierarchies are dependent on to which extents group members participate in the conversation. Literature shows that speaking time is related to communication apprehension. It
demonstrates that individuals that are susceptible to anxiety tend to speak less (Allen & Bouris, 1996; Patterson & Rits, 1997). Shyness is associated with communication
apprehension and is the opposite of dominance. Mullen, Salas, and Driskell (1989) found that speaking time is also related to emergent leadership and research demonstrates a relation between longer speaking times and higher-status actors (Smith-Lovin & Brody, 1989). Speaking time could be affected in different ways. The pilots indicated that the manner of turn-taking, ignoring cautions of the chairperson, attempting to influence the chairperson, and receiving interruptions by other MPs were practices that occur on regular basis in the Dutch Parliament. These practices are related to speaking time.
order to ensure speaking time. The most brutal way of ensuring a speaking turn is interrupting the interlocutor without permission of the chairperson. Research shows that interruptions are clear examples of making effort to control the conversation. It is found that if a speaker gains a speaking turn by interrupting another speaker, this indicates dominance, power, and a high-status in the debate (Kollock, 1985; Smith-Lovin & Brody, 1989). However, in political debates, interrupting does not necessarily mean that someone breaks rules since interruptions could be approved by the chairperson. Therefore, a distinction will be made between
conventional interruptions, approved by the chairperson, and unconventional interruptions, not approved by the chairperson. Unlike pre-determined speaking turns and reactions, interruptions are always initiated by the MPs themselves. Pre-determined speaking turns are speaking turns that are established beforehand and reactions are speaking turns offered to an MP because it includes, for example, a question. Clearly, pre-determined speaking turns and reactions are not correlated with dominance since these speaking turns are carried out deliberately to a smaller extent in order to increase speaking time.
MPs can claim the floor by using different forms of turn taking. However, there are more ways in which MPs can ascertain speaking time. Most debates have time limits for speaking time so that MPs could not talk endlessly. Chairpersons keep track on these time limits and often warn MPs that exceed it. MPs can choose to obey or ignore the warning. Exceeding time limits and ignoring warnings are important components of extending one’s speaking time.
In examining speaking time, it is also important to take into account the received interruption. This has a negative effect on the speaking time. In theory, it should not be possible that an MP in the Dutch Parliament is interrupted by another MP. In practice, this happens from time to time when rules are violated. Receiving interruptions signifies the opposite of dominance (Shaw, 2000). It is an indication that a speaker does not have the complete control over one’s speaking turn; research shows that it indicates subordination (Zimmerman & West, 1996). The chance to get interrupted increases according to the degree to which a speaker shows expressions of hesitation (Brend, 1975). Patterns of hesitations will also be measured, this will be discussed in sub-question 2.
Lastly, the pilots showed that MPs attempt to influence the chairperson in order to convince them to give more speaking time. To be specific, this aspect concerns statements of MPs that
are directed to the chairperson that have the objective to manipulate rules concerning the speaking time of MPs. Exceeding speaking time, unconventional interruptions, ignoring cautions of the chairperson, and attempts to influence the chairperson could all be classified under subversive behaviour, because they are all attempts to not comply with the rules. Not complying with the rules is related to dominant behaviour (Edelsky & Adams, 1990). In the methodology section, examples will be given. All the components that are related to claiming the speaking floor are illustrated in Table 1. Sub-question 1 reads as follows:
SQ. 1: Are there gender differences during political debates in the Dutch House of Representatives in terms of claiming the floor?
Table 1. Components and descriptions of the dependent variables of sub-question 1.
Component Description
Total speaking time The total time that an MP has claimed the floor during a debate, also the sum of pre-determined speaking time, interruptions, and reactions. Speaking Turns:
Pre-determined Speaking turns that are pre-set in advance of the debate, often part of a first round of speaking turns in which every MP gets the floor Conventional interruption Speaking turns requested by an MP and approved by the chairperson
interruption
Unconventional interruption Speaking turns in the form of an interruption that is not approved by the chairperson. This is against the debate rules
Reaction Speaking turns offered to an MP because a specific answer is needed from this MP
Exceeded speaking time Speaking time can be exceeded if the agreed time limit is surpassed Subversive behaviour Collective term for behaviour that is related to not complying with the
rules and consists of different aspects including exceeded speaking time and unconventional interruptions
Ignoring cautions. Concerns cautions of the chairperson about not complying to the rules
Influence chairperson. Attempts to convince the chairperson to give an MP more speaking by changing the debate rules
Received interruptions Received interruptions. Unconventional interruptions that make the current speaking MP lose his/her speaking turn
Hypotheses sub-question 1
It is expected that men will take up more speaking time than women during political debates in Parliament. This is because previous research already found that in non-political debates men tend to talk more than women (Thorne & Henley, 1975). Regarding behaviours related to claiming speaking time, previous research found that gender influences the number of
interruptions that are conducted (Woods, 1988). Based on this research it is expected that men conduct more conventional interruptions. Besides, the literature indicates that men tend to break rules more often than women (Edelsky, Adams, 1990). It is expected that men influence the chairperson more often, make more unconventional interruptions, and receive more
cautions from the chairperson since these behaviours are all related to rule breaking. From interviews with female MPs, it appeared that they considered unconventional interruptions as a male activity in which they did not see themselves engage with (Shaw, 2000, p. 416). This again reflects perceived differences in behavioural norms for men and women. Previous research on received interruptions shows that women are more likely to get interrupted than men (Zimmerman & West, 1996). Therefore, it is expected that male MPs receive fewer interruptions than female MPs. Literature also demonstrates that influencing others is often seen as an act of masculinity (Leung, Fuli, & Zhou, 2012). Accordingly, it is expected that men will try to influence the chairperson more often than women. Since pre-determined speaking turns and reactions are not correlated to dominance it is expected that there will be no gender differences between men and women in the use of these speaking turns.
H1: Men take up more speaking time than women
H2 Men and women conduct an equal amount of predetermined speaking turns H3: Men make more conventional interruptions than women
H5: Men and women make an equal amount of reactions H6: Men exceed speaking time more often than women H7: Women receive more interruptions than men
H8: Men express more subversive behaviour than women
3.3.2 Sub-question 2 – Language Use
The second sub-question concerns gender differences with regard to language use. It involves the extent to which speech styles include patterns of dominance. From the pilots, a couple of linguistic aspects appeared to occur frequently. These are linguistic hesitations and personally directed statements are examples.
Linguistic hesitations are divided into non-fluent speech and linguistic uncertainty
expressions. These aspects can make a speaking turn more or less powerful. Examples of non-fluent language use that cause speakers being unable to produce their sentences properly are false starts, stuttering, and vocal pauses (“uh”) (Griffiths, 1991). Linguistic uncertainty expressions consist of uncertainty (ad)verbs, such as “it seems”, “sort of”, “maybe”, or “perhaps”. Also, disclaimers (“I might be not an expert, but…” or “It might not be a good idea, however…”) (Carli, 1989) will be categorized as linguistic uncertainty expressions. As the second component of language use that could be used to exercise dominance, personally directed statements are included. This concerns statements made by an MP that directly address another MP. These statements are in general indicators of combativeness and dominance (Gidengil & Everitt, 1999, p. 54). Moreover, according to the rules within the DHoR, MPs are obliged to address their pleas towards the chairperson. This means that it is not only an indication of aggressiveness but also a form of breaking the rules in the form of not speak through the chairperson. The components of sub-question 2 are illustrated in Table 2. Sub-question 2, reads as follows:
SQ. 2: Are there gender differences during political debates in the Dutch House of Representatives in terms of dominant speech styles?
Table 2. Overview of the components of the dependent variables of sub-question 2.
Component Description Example
Linguistic hesitations:
Non-fluent speech False repeats, the speaker has to restart (a part of) his/her sentence because
words come out influently Stuttering
Vocal pauses
“What is wh.. Why is that the case?”
“I.. I.. I want to say that…” “Uh”/”Uhm” Linguistic uncertainty expressions Uncertainty (ad)verbs Disclaimers
Statements that are not directed to the chairperson directly towards an MP.
This is against the debate rules.
“maybe”
“I might be wrong, but… “ “Misses Becker, have you
thought about…” instead of “Has misses Becker
thought ab…” Personally directed statements
Hypotheses sub-question 2
In non-political discussions, it appears that women’s use of language is in general perceived as “weaker and less effective than the speech of men” and is, therefore, inferior to ‘men’s language’ (Kramer, 1974, p. 82). This is in line with the argument that women are less powerful partly because of the way they speak (Thorne & Henley, 1975).
In specific, linguistic hesitations are more likely to be used by less self-confident speakers (Lakoff, 1990) and leads to subordination and powerlessness (Henley, 1977). As pointed out earlier, speaking with hesitation increases the chance to receive interruptions (Brend, 1975). In this way, hesitation is indirectly related to subordination. Thorne and Henley (1977) found a direct association between feminine manners of speaking and hesitations. All in all, it is expected that women express more linguistic hesitations during their speaking turns than men. Regarding personally directed statements it is expected that men use them more than women.
This is because personally directed statements in political debates is a form of ignoring the rules and research shows that men tend to do this more often (Edelsky, Adams, 1990).
H9: Women express more non-fluent speech in their speech than men
H10: Women use more Linguistic uncertainty expressions in their speech than men H11: Men use more personally directed statements than women
3.3.3 Sub-question 3 – Debate context
The third sub-question addresses the circumstances that could influence possible outcomes found in the first two sub-questions. These circumstances involve the topic and the gender composition of the debate. This sub-question is related to tokenism. We learned how
tokenism discloses how historical relations influence group interactions of different categories of people. The theory assumes that not only being female but especially being a female in a minority group, could influence group interactional behaviour (Kanter, 1977). This means that the setting is also important to take into account. In order to find out whether the possible effects of gender on dominance are dependent on the setting of the debates, sub-question 3 aims to measure whether possible outcomes found in sub-questions 1 and 2 are related to the group composition and the topic of the debates.
Research suggests that the gender composition of groups can have influence on speech
patterns of the participants that join a discussion (Smith-Lovin & Brody, 1989). However, it is hard to define the nature of that relationship. There are studies that suggest that minority categories within groups become more passive in their contribution to the group discussion (South, Markham, Bonjean, & Corder, 1987; Wharton & Baron, 1987). In other words, this would mean that women or men would become more dominant in relation to an increase of the present members of their own gender. However, other studies indicate a relationship between composition of the group and dominance in the sense that women become more supportive and less dominant when there is an increase in other female attendants and that men will become more dominant if there are more male attendants (Carli, 1989; Hall, 1984; Piliavin & Martin, 1978; Aries, 1976).
Besides, this sub-question issues the topic of the debate. The literature demonstrates that certain topics are associated with femininity or masculinity (Signorella & Vegega, 1984).
Stereotypical feminine topics within Parliament are for example health care, education, environmental protection, and family (Kahn, 1994; Huddy & Terkildsen, 1993a; 1993b). These two contextual aspects are encapsulated in sub-question 3:
SQ. 3: Does the setting of a debate play a role in gender differences in terms of dominance during political debates in the Dutch House of Representatives?
Table 3. Overview of the components of claiming speaking time
Component Description
Topic Debates are divided into 'feminine debates' (e.g. childcare) and 'masculine debates' (e.g. war)
Gender composition Debates are divided into 'female dominated debates' (i.e. debates in which there are more women present and 'male dominated debates' (i.e. debates where there are more men
present)
Hypotheses sub-question 3
Since previous studies on dominance and gender composition point out different relations, it is hard to determine the direction of this relationship in this thesis. However, since this thesis builds on the idea of tokenism, it is expected that people act more dominant when their group composes a majority. It is hypothesized that gender norms that benefit men are strengthened in debates with a masculine topic or in debates in which men form the majority. Regarding the topic of the debate, it is expected that women are relatively more dominant in discussions on alleged feminine topics and that men are relatively less dominant in discussions on alleged feminine topics than in discussions on a less alleged feminine topic. Figure 2 displays a visual scheme of how the research questions of this thesis are build up.
H12: Women score higher on dominance in female dominated debates than women in male
dominated debates
H13: Men score higher on dominance in male dominated debates than men in female
dominated debates
H14: Women score higher on dominance in debates with a feminine topic than in debates
H15: Men score higher on dominance in debates with a masculine topic than in debates with a
feminine topic
4. Design and Methodology
In this chapter discusses decisions related to the data collection. The data will be collected from committee debates because they contain more interaction. Purposive sampling was used in order to ensure the data contained debates from
different topics and different gender compositions. Descriptives of the characteristics of the MPs will be set out (i.e. age, party). Thereafter, it will be explained per variable how it is measured and the unit in which it is measured. Concluding with a handy overview of all the variables, their descriptives, and how
it is measured.
4.1 Case selection
For the analysis, video recordings were used of debates that took place in the DHoR between June 2017 and April 2018. These video recordings are freely available online
(https://debatgemist.tweedekamer.nl/). Conducting pilots helped to refine the methodological design of this thesis. During the pilots, different materials were analysed, without focussing on specific behaviours. For example, plenary and committee debates were watched, and minutes of plenary debates were read. Based on these pilots, certain aspects were evaluated, such as availability of the materials, the reliability of observed behaviours, and feasibility of the research methods. Despite the fact that plenary debates are more well-known than committee debates, it is chosen to use committee debates for the data collection. This is because the pilots showed that committee debates were more dynamic than plenary debates. For example, the fixed scheme of speaking turns is different; in committee debates, speaking time is divided over more speakers and there is also more room for interruptions. Those interruptions are also easier to carry out since MPs do not have to walk to the interruption microphones, but they can make them from their desk. Given the fact that every time a new speaking turn starts means an extra observation, committee debates afford more data within a shorter period of time. Besides, in the plenary hall, it is mainly the party leaders that are given the floor to. In the Netherlands, all party leaders are male, except for one since January 2017 and two since December 2018. Although this is an interesting fact in itself, in this thesis, the
focus is on differences between men and women. Therefore, committee debates are chosen for the analysis. Figure 3 gives an impression of how a committee debate looks like in the Dutch Parliament.
The average debate lasts for 3 or 4 hours so timewise, it is not possible to include a large sample of debates. However, in order to ensure that the data included debates on different (feminine and masculine) topics and that they have either male and female dominated group compositions, debates were selected based on these characteristics. This means that purposive sampling was used in order to select the debates.
Figure 3. Committee debate in the DHoR.
4.2 Descriptives of Sample
In total, 11 committee debates were observed. These debates included 31 hours of footage. The average debate lasted for 2 hours and 48 minutes, varying from 1.34 to 4.15 hours. In total, 723 speaking turns were collected carried out by 60 of the 150 members of Parliament. The average age of the MPs was 44.4 year. In total, 10 MPs were a member of CDA, 10 were a member of the VVD, 8 from GL, 8 from D66, 6 from SP, 5 from PVV, 4 from PVDA, 2 from PVDD, 50+, SGP, DENK, CU, and no one from FvD. The absence of FvD in the
party do not go to debates (Hadders, 2017; RTLnieuws, 2018). turns of ministers were not included, because ministers have a considerably different role in debates in a way that they could distort the data. For example, ministers could have a significantly larger amount of speaking at the end of the debate, but this is possibly more related to their role than to dominant behaviour.
4.3 Independent Variables
Gender is the independent variable. The gender of the speakers is simply determined based on the way the speakers are introduced by the chairperson, i.e. “Mister…” or “Misses…”.
4.4 Dependent Variables
The eventual goal of this thesis is to measure dominance as a dependent variable. However, dominance serves here as a latent variable. This means that it is not directly observed but inferred from other, more objectively measurable, variables. In this paragraph will be explained how every single dependent variable will be measured. In the first sub-question, actual speaking time, the manner of turn-taking, exceeded speaking time, cautions by the chairperson, received interruptions, and influencing the chairperson are measured.
4.4.1 Dependent variables sub-question 1
Speaking time indicates the time in seconds that a speaker holds the floor. Holding the floor means giving vocal input to the conversation, regardless of how the floor is claimed (i.e. legal or illegal).
The manner of turn-taking indicates the way a speaker obtains his or her turn. Four different ways of turn-taking are distinguished: ‘pre-determined’, ‘conventional interruption’
‘unconventional interruption’, and ‘reaction’. In the Dutch Parliament, it is common practice that the chairperson assigns speaking turns, some are already planned before the debate starts, those are the predetermined speaking turns. Besides, it is possible to obtain a speaking turn by indicating that one wants to interrupt. If the chairperson accepts this interruption and gives the floor to this speaker, the turn is marked as a conventional interruption. Conventional
interruptions generally give forth matter to react on, therefore, it is often the case that the chairperson automatically gives the floor to the speaker towards whom an interruption was directed, subsequently, these speaking turns will be marked as reactions. Lastly, speaking
turns will be indicated as unconventional interruptions if the speaker gains the floor without the approval of the chairperson. This is against the rules of the Dutch Parliament. Reactions, conventional interruptions, and pre-determined speaking turns are distinguished because they are related to dominance in different ways. Equal treatment could cause distorted data. This is because pre-determined speaking turns often have a time limit, whereas reactions and
interruptions should preferably as short as possible. Long interruptions and reaction,
therefore, seem more correlated with dominance. Besides interruptions are always initiated by an MP, reactions are speaking turns that are offered by the chairperson. Therefore, the amount of interruptions carried out by an MP during a debate says more about dominance than the number of reactions carried out. For each of the manners of turn-taking, it will be counted how often an MP uses them per debate.
Exceeded speaking time. Sometimes speaking turns in committee debates are tied to time limits. This means that the chairperson and the MPs agree in advance of the debate about a specific time that speakers get for their speaking turn. This often concerns the first, pre-determined, turn of every speaker. Debates often start by the opening pleas of the MPs. Turn by turn, every MP will get the floor and explains his or her position towards the discussed issue. Time limits often concern a duration of 5 minutes. The chairperson can use a stopwatch to keep track of this. In some debates, there were time limits for concluding remarks. In this case, all the MPs will be offered the floor for approximately 1 or 2 minutes to share their opinion of the debate. Lastly, MPs can exceed speaking time during any speaking turn if he or she continues talking after the chairperson has asked to round off. The exceeded speaking time will either be measured in seconds and accumulated per MP per debate. Besides, it will be counted how often a particular MP exceeds time for at least 1 second, this will be
measured as a part of subversive behaviour.
Received interruptions involve interruptions that a speaker receives during his or her speaking turn. Both successful and unsuccessful unconventional interruptions are tally marked. A successful interruption automatically means the end of a speaking turn. On the contrary to the other dependent variables of sub-question 1, this variable has a negative relation to
dominance. The more interruptions a speaker receives, the less dominant that speaker is deemed. It will be counted how often interruptions are received per MP per debate.
Subversive behaviour. subversive behaviour consists of behaviour that aims at ignoring or violating the debate rules. This variable is an accumulation of received cautions,
unconventional interruptions, attempts to influencing the chairperson, and amount of times that the time is exceeded. Exceeded speaking time and unconventional interruptions are explained above, received cautions and influencing the chairperson will be discussed below: Receiving cautions. This component of subversive behaviour includes behaviour related to exceeding the speaking time or attempts to interrupt unconventional that the chairperson gives a caution about. If an MP ignores the caution, it could happen that an MP received 3 or more cautions in a single speaking run. Cautions could be as follows: “Misses Karabulut, you should really let misses Dik-Faber speak without interrupting her.”, “Mister Fritsma, could you please wrap up quickly?”, “Dear Sjoerdsma, you should direct your plea towards me, otherwise I will have to ask you to stop talking”. It will be tally marked how often speakers receive such cautions per debate.
Influencing the chairperson. Influencing the chairperson indicates statements that aim to obtain more speaking time by challenging the chairperson. Examples of such statements might be: “Chairperson, you need to give me a little more time”, or “My argument is very important therefore it is crucial that I finish it”. Such statements will be tally marked per MP per debate.
Table 4. Overview of the components of the dependent variables of sub-question 1.
Component Description Unit of Measurement
Total speaking time The total time that an MP has claimed the floor during a debate, also the sum of pre-determined
speaking time, interruptions, and reactions.
In seconds
Speaking Turns:
Pre-determined Speaking turns that are pre-set in advance of the debate, often part of a first round of speaking turns in which every MP gets the floor
In seconds + amount of turns
Conventional interruption Speaking turns requested by an MP and approved by the chairperson interruption
Unconventional interruption Speaking turns in the form of an interruption that is not approved by the chairperson. This is
against the debate rules
In seconds + amount of turns
Reaction Speaking turns offered to an MP because a specific answer is needed from this MP
In seconds + amount of turns
Exceeded speaking time Speaking time can be exceeded if the agreed time limit is surpassed
In seconds
Subversive behaviour Collective term for behaviour that is related to not complying to the rules and consists of different aspects including exceeded speaking
time and unconventional interruptions Ignoring cautions. Concerns cautions of the chairperson about not complying with the rules Influence chairperson. Attempts to convince the chairperson to give an MP more speaking by
changing the debate rules
Accumulation of the number that the behaviours occur per
MP per debate
Received interruptions Received interruptions. Unconventional interruptions that make the current speaking
MP lose his/her speaking turn
Amount per MP per debate
4.4.2 Dependent variables sub-question 2
The second sub-question measures linguistical hesitations and personally directed statements. As explained in the previous section, all the components of linguisitc hesitations are
negatively correlated with dominance and the use of personally directed statements are positively correlated with dominance. Linguistical hesitations is divided into non-fluent speech and linguistic uncertainty expressions. Non-fluent speech consists of false starts, stuttering and vocal pauses. Linguistic uncertainty expressions consist of uncertainty (ad)verbs and disclaimers. All the observed variables will be discussed below: Non-fleunt speech
False repeats, stuttering, and vocal pauses will be tally marked and accumulated in order to measure non-fluent speech and a score will be calculated for each MP of a debate indicating the average use of these aspects of non-fluencies per minute. So for example, ‘misses
Ploumen has a score of 12.4 on non-fluent speech’ means that in a particular debate, Ploumen uses 12.4 false repeats, stutterings, and vocal pauses per minute.
False repeats. False repeats are slips that make that a speaker has to repeat a part of what has already been said or attempted to being said, e.g. “The report demonstrates that… that… that we should…” or “Therefo… For that reason, I…”. It will be tally marked
Stuttering. Discontinuities in talking as a consequence of uncontrolled repetitions of sounds, often consonants.
Vocal pauses. Vocal pauses are pauses in a sentence or the beginning of a sentence that a speaker fills up with the words “uh” or “uhm”.
Linguistic uncertainty expressions
In the same way as non-fluent speech, a score on linguistic uncertainty expressions will be calculated for MP per debate, indicating the average amount of uncertainty (ad)verbs and disclaimers that an MP uses per minute in a particular debate.
Uncertainty (ad)verbs. This includes verbs and adverbs that express doubt and make statements sound less convincing. Examples of such adverbs are: ‘perhaps’, ‘possibly’, ‘seemingly’, ‘presumably’. Examples of uncertainty verbs are: ‘think’, ‘might’, ‘seem’. Disclaimers. Disclaimers are often at the beginning of a plea or sentence and include a prior announcement of excuses for what is going to be said. It is often used as a tool to limit one’s responsibility of the statement that will follow and restrict the rights that others have to condemn that statements. Examples from the data are: “It might not be the best solution, but…” or “I am not completely sure about it, but…”
Personally directed statements. Personally directed statements are statements directed that directly address another MP. This is against the debating rules since MPs are obliged to speak towards the chairperson. Examples of personally directed statements are: “Can you explain
why this happens all the time?” instead of “Can mister Voordewind explain why this happens all the time?”
Table 5. Overview of the components of the dependent variables of sub-question 2.
Component Description Unit of Measurement
Linguistic hesitations:
Non-fluent speech False repeats, the speaker has to restart (a part of) his/her sentence because words
come out influently Stuttering
Vocal pauses
Sum of amounts of false repeats, stutterings, and vocal pauses an MP uses
on average per minute
Linguistic uncertainty expressions
Uncertainty (ad)verbs Disclaimers
Statements that are not directed to the chairperson directly towards an MP. This is
against the debate rules.
Sum of amounts of false repeats, stutterings, and vocal pauses an MP uses
on average per minute Average amount per MP
per debate Personally directed
statements
4.4.3 Moderating variables sub-question 3
Gender ratio and the topic of the debate are the two moderating variables. Both will be determined at the beginning of every debate.
Gender composition. debates will be categorized into male or female dominated debates. Debates with 75% or more women will be categorized as female dominated debates. Debates with 75% or more men will be categorized as male dominated debates.
Topics. Debates will be divided into feminine and masculine debates depending on the topic they discuss. Debates in the date that are categorized as feminine are dementia, childcare,
environment, education, and emancipation. Debates in the date that are categorized as masculine are ISIS, big data, financial sector, and material defence.
Table 6. Overview of the components of claiming speaking time
Component Description
Topic Debates are divided into 'feminine debates' (e.g. childcare) and 'masculine debates' (e.g. war)
Gender composition Debates are divided into 'female dominated debates' (i.e. debates in which there are more women present and 'male dominated debates' (i.e. debates where there are more men
5. Results
This section includes the statistical results of the analysis. First, the number of observations, removed variables, and the way the data is structured will be discussed in the paragraph data
handling. Subsequently, the means of men and women will be given for every dependent variable in the descriptive statistics paragraph. Next, it will be explained that the data will be
analysed with multiple linear regression models. Four different models will be included. Lastly, it will be set out for each of the dependent variables whether significant results were
found. The variables ‘total speaking time’, ‘predetermined speaking turns’, ‘reactions, ‘exceeded time’, and ‘received interruptions’, showed significant effects of gender. In the chapter Conclusion and Discussion, the results will be linked to the sub-questions and the
hypotheses.
5.1 Data handling
The data were aggregated based on speaker per debate. This means that a single value in the data provides information about the behaviour of one speaker during one debate instead of information about one speaking turn of one speaker. This was done in order to make the data more workable for the statistical analysis. This means that the number of observations decreased from N =723 to N =86. As a consequence, the variables ‘Member of Parliament’ and ‘party’ are not included in the statistical analysis. This would have given distorted data since the N is too small after the aggregation.
5.2 Descriptive Statistics
Table 7 shows the estimated values for men and women for each of the dependent variables. So, for example, the estimated speaking time for women during a debate is 288.38 (SD = 25.72) seconds and this is 76.63 seconds shorter than the estimated speaking time of men during a debate. At first glance, this difference in speaking time seems to explain by looking at the amount and length of speaking turns in the form of reactions and the estimated
exceeded speaking times. It is estimated that men (M = 8.88, SD = 3.6) exceed the speaking time more often than women (M = 8.88, SD = 3.6). The regression models will demonstrate whether the effects of gender are significant on each dependent variable. The means of total speaking time, average time unconventional turn, average exceeded time, Subversive
behaviour, Received interruptions, Linguistic hesitations, Uncertainty (ad)verbs, and Personally directed statements differ more than 20% of each other. Linear regressions are used in order to analyze whether these differences are significant.
Table 7. The Estimated Values, Standard Deviations and differences per debate of the Dependent Variables
Dependent Variable Men Women Difference
Total speaking time 365.01 (39.81) 288.38 (25.72) 76.63
Avg. time predetermined turns 269.43 (19.99) 293.59 (41.23) -24.16
Conventional interruptions 2.87 (0.6) 2.95 (0.57) -0.08
Avg. time conventional interruption 36.8 (2.09) 36.09 (4.04) 0.71
Unconventional interruptions 0.28 (0.17) 0.23 (0.12) 0.05
Avg. time unconventional interruption 0.69 (0.29) 1.7 (0.8) -1.01
Reactions 4.28 (0.65) 3.63 (0.46) 0.65
Avg. time reactions 53.43 (6.7) 46.79 (3.82) 6.64
Avg. Exceeded time 25.72 (4.84) 8.88 (3.6) 16.84
Subversive behaviour 2.39 (0.44) 1.63 (0.44) 0.76
Received interruptions 0.7 (0.2) 0.48 (0.15) 0.22
Linguistic hesitations 7.98 (0.57) 9.99 (0.99) -2.01
Uncertainty (ad)verbs 0.92 (0.18) 1.16 (0.18) -0.24
Personally directed statements 1 (0.34) 0.55 (0.1) 0.45
Note. Time is in seconds.
Note. N (men) = 46, N(women) = 40
5.3 Regressions
The general idea of a regression model is that it explains the relationship between at least one independent variable and a dependent variable. The elementary regression formula reads as follows:
The value of yi is the dependent variable. The a in the functions stands for the incept value,
that is the expected value of yi when xi = 0. B, unstandardized beta, is the regression
coefficient and indicates the expected increment of yi when xi increases by 1. A dependent
error variable is pointed out by eI . The formula can get more complex when independent
variables or interacting functions are added.
Tables of the outcomes of all regression models are included in the Appendix. The models of the average time of conventional interruptions, unconventional interruptions, exceeded speaking turns, and reactions, did not show significant results. Also, thee models of amounts of unconventional interruptions, linguistic hesitations, and personally directed statements did not result in significant outcomes. In this thesis, results are evaluated significant starting from a p-value of <.1 or less. This means that one could say that there is a 90% chance that if the effects are significant that the effect is actually true. A p-value of .05 is more common. However, because this research is conducted within social sciences in which countless
variables could play a role, a p-value of <.1 is accepted. Similar to the p-value, there has been
chosen for an interpretation of the R-squared that fits into the reality of social science. The R2
represents the percentage of variation of the outcome variable that is explained by the
regression model. For example, an R2 of 0.12 means that a proportion of 12 per cent of the
variance of the dependent variable is explained by the predictor variables. In this thesis, an R2
of .19 is evaluated as a weak effect, an R2 of 0.33 is evaluated as moderate and an R2 of 0.67 is
evaluated as a strong effect. These values are recommended by Henseler and Chin (2010) based on endogenous latent variables. Although the variables included in this thesis are not endogenous latent variables, similar criteria are chosen because this research is still in its infancy and intents to pave the way for more extensive research on this topic. Regression models in which significant relations were obtained will be discussed below.
5.4 Models
Multiple linear regressions were conducted for each of the dependent variables. A multiple linear regression was used to determine whether the effect of gender is significant. All the regression models control for age. Model 1 only includes gender (independent variable), age (control variable) and the dependent variable. Model 2 includes all the variables of Model 1 plus Topic as an extra control variable, whereas Model 3 adds Group Composition as a control variable to Model 1. Model 4 includes all the variables mentioned above.
Figure 4. Overview of the regression models.
5.5 Results
5.5.1 Total speaking time
The regression models for all the dependent variables are presented in Appendix A. The most important outcomes of the regressions are discussed here. The regression models in which the total speaking time served as the dependent variable showed all showed significant F-values. This means that the four models could provide better predictions than merely taking the mean of the response (i.e. the intercept-only model). However, Model 4 appeared to be the best
model to predict total speaking time (F(4, 8) = 4.17, p < .05) with an R2 of 0.12. However, the
coefficients of the separate regressors do not appear to be significant. Therefore, if we exclude group composition in the regression and look at Model 2, we see that this model is less, but still significant (F(3, 8) = 3.13, p < .1) also with an R2 of 0.12. Focussing on the individual
regressors shows that the effect of gender is significant if besides age also the topic of the debate is included in the model. The model shows that the estimated speaking time during a debate increases with 101.97 (SE = 49.58) seconds when the MP is male compared to female MPs. The model also shows that speaking time decreases the older a speaker is, this includes an estimated decrease of 1.9 (SE = 1.51). Regarding the topic of the debate, the model indicates that MPs speak 42.73 (SE = 42.73) seconds more in debates with a feminine topic compared to debates with a masculine topic.
5.5.2 Predetermined speaking turns
The analysis shows that all the four models for the estimated amounts of predetermined speaking turns per debate are significant (see table I, Appendix A). Especially the models in which the topic of the debate is included, contain a significant predictive value. Model 2 is
significant (F(3, 10) = 7.82, p <.001) with an R2 of 0.12. The model indicates that women
Furthermore, the model shows that MPs use by estimation 0.02 (SE = .1) more predetermined speaking turns by each year that the age increases. Besides, in debates with a feminine topic, it is estimated that speakers use 0.03 (.1) more predetermined speaking turns than in debates with a masculine topic.
5.5.3 Average time pre-determined speaking turn
Regarding the variable average time pre-determined speaking turn time all four models show significance. Model 2 (F(3, 8) = 6.89, p <.05, R2 = 0.31) and 4 (F(4, 8) = 5.4, p <.05, R2 = .31) seem to have the most predictive power. The model estimates that women speak 50.2 (SE = 40.94) seconds longer than men in predetermined speaking turns during a debate and that the estimated speaking time increases by 3 (SE = .74). seconds per year. The model also showed that MPs speak 103 (SE = 47,65) seconds less in debates with a feminine topic. The
regressors of age and debate topic showed significant relations.
5.5.5 Conventional interruptions
Model 3 (F(3, 10) = 2.94, p <.1, R2 = .03) and 4 (F(4, 8) = 2.81, p <.1, R2 = .12) of the regression on conventional interruption showed significant results. Model 3 explains 3 per cent of the variance of the observed values, whereas Model 4 explains 12 per cent of the variance of the observed values. The individual regressors are not significant.
5.5.5 Reactions
Model 2 (F(3, 8) = 3.05, p <.1, R2 = .13) and Model 4 (F(4, 8) = 2.82, p <.1, R2 = .13). Model 2 also showed a significant effect for gender; it is estimated that men perform 1.42 (.72) more reactions per debate. The model also indicates that MPs do 1.15 (SE = .73) more reactions in debates with a feminine topic and that reactions decrease with 0.02 (SE = .02) when the age of an MP increases with one year. However, the regressors age and topic of the debate are not significant.
5.5.6 Exceeded speaking time
All the four models on exceeded speaking time are significant (See Appendix A, table IX). Especially model 1 and 2 are relevant for the analysis since Mosel 1 (F(2, 10) = 10.42, p <.01, R2 = .13) shows a significant value (p > .01) for the effect of gender, indicating that women exceed the speaking time shorter with an estimated 16.07 (SE = 4.72) seconds per debate.
Model 2 (F(3, 8) = 58.54, p <.01) explains the 22 per cent of the variance of the observed
values (R2 = .22) and shows a significant effect for topic of the debate in the sense that MPs
exceed the speaking time more often in masculine debates with an estimated 13.58 (SE = 6.73) seconds.
5.5.7 Received interruptions
The regression model on received interruptions shows that Model 2 (F(3, 8) = 5.38, p <.05, R2
= .08) and Model 4 (F(4, 8) = 8.87, p <.01, R2 = .1) are significant. In particular Model 2 is interesting because it shows a significant effect for gender, indicating that women receive 0.43 (SE = .2) less interruptions than men in a debate. Furthermore, this model shows that the amount of interruptions decreases with 0.01 (SE = .01) the older an MP is and that less interruptions ( B= .36, SE = .22) take place during debates with a masculine topic.
5.5.7 Uncertainty (ad)verbs
With regard to the regression model on Uncertainty (ad)verbs, Model 4 (F(4, 8) = 1.99, p >.1,
R2 = .04), although not significant, showed a significant effect of group composition. In the
model, it is estimated that MPs use 0.57 (SE = .23) uncertainty (ad)verbs less in debates in which there are more female MPs. The model also estimates that women use 0.5 (SE = .34) more uncertainty (ad)verbs than men. However, this effect was not found significant.
5.6 Interaction Models
Interaction models were included the regressions in order to analyse whether the results of previous dependent variables account for all situations. By way of explanation, if an
interaction can give information about whether the estimated value of total speaking time for men is always 102 seconds more than for women, or if this is dependent on the topic or group composition of the debate. For example, men speak for 51 seconds longer than women during topics with a feminine topic and 151 seconds longer during debates with a masculine topic.
5.6.1 Topic of the debate
Two interactions were conducted, one between topic and gender and one between group composition and gender. Appendix B shows graphs of the estimated values for the dependent variables for men and women during debates with a feminine topic and debates with a
