To what extent do Democrats and Republicans use
different connectives when arguing for or against a
topic in presidential candidate debates, and can
topic be used as a predictor for connective use?
Evgenia Tincheva
by
Bachelor’s Thesis
Supervision: Jet Hoek
International Business Communication
Communication and Information Studies
Faculty of Arts
1
To what extent do Democrats and Republicans use different connectives
when arguing for or against a topic in presidential candidate debates, and
can topic be used as a predictor for connective use?
Abstract
This research aims to add to the current fields of political debates and discourse connectives. The main objective is to find a difference in the use of connectives by Republicans and Democrats based on stance and/or to try to predict connective use by looking into the topics discussed. In order to achieve this goal, the paper uses a corpus analysis of 6 political debates from 12 different speakers. After various statistical tests, some important results were discovered. Firstly, negative argumentation was the most frequent type for speech acts. Nexts, Temporal, Contingency and Comparison classes of connectives were most used when taking a negative stance, while Expansion connectives were more frequent for positive argumentation. Moreover, a significant relation was found between the topic discussed and the connectives used. Economy, Taxes, Healthcare and Iraq war often used Contingency connectives, while topics like Military included Comparison connectives. Finally, a significant difference was found between the topics discussed by Republicans and Democrats, supposing that topic could indeed be a predictor of connective use.
Key words: connectives, political debates, for and against, political party
Introduction
Political communication is a crucial type of discourse. The sole purpose of building electoral campaigns around candidates is to get their message across to as big of an audience as possible. Hence, almost all political phenomena happen in the form of text or talk. Speeches, meetings, publications, all revolve around communication from an office seeker to his or her voters. These interactions significantly mold the social construction of the politician’s image and affect the results of the polls. And because the stakes are so high, politicians are known to use writers for their speeches, so that every idea is appropriately formulated and tailored to the specific audience. The utterances are carefully written beforehand, patiently rehearsed and confidently delivered to the public to ensure maximum persuasion and support.
There is, however, one specific genre of public appearances which is slightly less staged - political debates. Those are formal discussions which are held after the political parties have nominated their candidates and where office seekers share and defend their opinions on formal matters. Although participants could be informed about the general topics that they will be asked to discuss prior to the event, they could
2 never know what exactly their opponent will comment on. Therefore, they have to come up with convincing argumentation to defend or oppose a viewpoint on the spot without scripted notes or memorized answers. This allows for a more natural look into politicians’ speech, the connectives they use and their opinion forming techniques.
Debates offer possibilities for research into persuasiveness, image, even body language. However, should the topic be analyzed from a discursive dimension, it has the potential to answer much more relevant questions in today’s politic-run world like whether connective use can influence the persuasiveness of a discourse, or help understand if office runners use different expressions for topics they support versus oppose, and whether that would lead to them winning or losing the debate. Therefore, the question that this research will focus on, implies reflection on three key topics: political language, connectives and political debates.
Literature Review
Starting in the field of discourse markers, of which connectives are a part of, it is important to indicate what the terms discourse markers and connectives suggest. Even today, the status of discourse markers is a bit vague, as various labels and lexical expressions have been used over the years to research these discourse phenomena. In an attempt to clarify what a discourse marker is and how it is to be defined, Fraser (1999) proposed that discourse markers are conjunctions, adverbs and prepositional phrases that signal a relationship between the segment they introduce and the prior segment. Although the article follows to give numerous examples of explanations or words/expressions that fall under the category of discourse markers, the essence of the terminology overlaps with other definitions. For example, Michael Swan defines discourse markers as "words or expressions which either connect a sentence to what comes before or after, or indicate a speaker's attitude to what he is saying” (Swan, 2005). In the case of debates, political candidates take different stances (either defending or opposing a topic), and therefore, adopt different attitudes towards the subject of speech. If indeed discourse markers portray a speaker’s mental outlook on a problem, then one might suppose that their use of connectives would vary in category or frequency depending on their attitude.
A number of studies have looked into the different uses of specific discourse markers. For example, the study of Fox-Tree and Schrock (2002) examines the use of
you know and I mean in regulating the collaborative nature of spontaneous talk and
expressing speaker involvement. Meanwhile, Overstreet and Yule (2002) found that the marker and everything is often used to predict or influence the interpretation of a situation or of the attitude expressed by the speaker (Overstreet & Yule, 2002). Following this pattern, a number of other articles have singled out discourse markers and examined their specific use in interaction. Nevertheless, not too many have touched upon researching the use of a specific group of discourse markers, namely connectives, in interaction. A good explanation of discourse connectives is offered by
3 Louwerse and Mitchell (2003) who define connectives as cohesive devices that hint coherence relations and mark transition points within sentences, between sentences, or turns, at the local as well as global levels of discourse. Another definition can be found in the article by Sanders, Spooren and Noordman (1992), stating that connectives “are linguistic markers, expressing the underlying conceptual relations that are of a cognitive nature” (p.3). However, despite other studies’ attempts to categorize linguistic expressions under exhaustive groups of connectives or markers, there is no universally acknowledged definition of either term. In fact, many articles refer to them as synonyms and use the two labels interchangeably. Nevertheless, it is generally considered that connectives are a smaller, more precise subgroup of discourse markers, often organized in tables or lists. Using a complete and definite list of connectives to study communication not only increases the external validity of any research, but also ensures valid and concrete results. Hence, this article will focus on connectives only, and try to evaluate political communication on the basis of the different types of connectives used.
In a more recent study, Sanders and Spooren (2015) investigated whether the use of the Dutch causal connective pair “omdat” and “want” was affected by the speaker’s language proficiency, the production context or the medium used and concluded that there is in fact a clearly different pattern of use. Their findings show that there are three main predictors of the decision to use one or the other connective. The most important predictor is Type of Relation between content and speech act, followed by Propositional Attitude of the first segment and the final decision concerns the Medium used (Sanders & Spooren, 2015). If the attitude is not a judgement, then
omdat will be used, and if the medium is written text or spontaneous conversation,
then omdat is again more likely to be used (Sanders & Spooren, 2015). Based on these results it would be interesting to investigate if Topic can be used to predict connective use, since the predictor Propositional Attitude has a lot in common with the Topic of a speech act. If the same predictor, namely Propositional Attitude, is to be applied only to the medium of spoken discourse in the form of political debates, the results found by Sanders and Spooren (2015) could hold true and a difference in connective use can be found.
As to the reasons behind why connectives should be investigated in political discourse specifically, quite a few studies could be used as exemplars. To begin with, as Ismail (2012) points out “the doing of politics is constituted in language” (p.1261). Moreover, the main tool used to achieve political goals is language, of which discourse markers are a part of and show what the speaker’s attitude is to what they are saying. This can be confirmed by Aijmer (1996), according to whom discourse markers perform various functions, but the two main ones are discoursal and interpersonal function. The first one signals relations between present, prior and future discourse units and the latter helps speakers express their stance (Aijmer, 1996). Hence, discourse markers, of which connectives are a part of, help politicians in building coherent speech acts, as well as better their argumentation in favor or against a topic.
4 The topic of political speech and debates is one that has undergone a considerable amount of research already. From focusing on one specific politician’s speech to reviewing a wide spectrum of different speakers, studies have found some insights into topics like persuasiveness, deception and personal speaking style of office seekers.
To begin with, many office seekers are subjects to such investigations, especially during their political campaigns. Donald Trump’s hypercompetitive speaking style and his win-lose approach to negotiating (Kapoutsis & Volkema, 2019), Hillary Clinton’s female-marked political discourse (Mukhortov & Malyavina, 2019) and Barack Obama’smore elaborate, structured and formal discourse (Reyes, 2014) are just a few examples of such investigations. However, research has yet to investigate if there is a certain type of tendency occurring specifically in the use of connectives by not just one, but various political figures with divergent attitudes and perspectives. Through distancing from who the speaker in particular is, the focus shifts to the use of words making way for potential discoveries like whether stance can affect connective use.
In political debates, communication is predominantly persuasive. Therefore, a number of papers have discussed different ways in which speakers try to influence their audience. For example, Innocenti and Miller (2016) tried to explain how speakers design political humor to coordinate public action and provided a theory-based rationale for designs of political humor. They discussed how humor, whether it is intended to be taken seriously or simply as a joke, always aims to influence voters to scrutinize arguments, by imposing on their attitudes, beliefs, etc. (Innocenti & Miller, 2016). Another way to persuade listeners is through argumentation. In their study Goodwin and Innocenti (2019) argue that making reasons for something apparent is a task pragmatically necessary for any audience effect and the single primary function of argumentation is just that. Having in mind that argumentation is a vital part of any political debate, it seems appropriate to research a bit more into how political candidates build their arguments through the use of connectives.
Another interesting element of research would be to see whether office seekers would use different connectives in their argumentation depending on whether they argue for or against a certain topic. One difference worth studying would be whether Republicans and Democrats build their arguments using different connectives based on the topic they are talking about. A study by Thomson and Froese (2016) looked into the opposing moral worldviews held by the two American coalitions and found a connection between faith and policies. Namely that “Democrats who believe in a judgmental God tend to support more conservative policies” (Thomson & Froese, 2016, p. 839). Moreover, they mention that the Republican party emphasizes on criminal punishment and military might and the Democratic party focuses on social sources of crime, including poverty, unemployment, and educational inequality (Thomson & Froese, 2016). Since Democrats and Republicans are known to defend not necessarily opposing, but nevertheless, different topics, it would be useful to know if they build their arguments differently based on their stance. That is to say, since
5 politicians in general are proven to use connectives as a means to express their stance on topics (Ismail, 2012), and having in mind that the two parties would be highly likely to hold differing stances, it seems natural that their connective use would be different as well. From a discourse perspective, it would be interesting to see whether the different topics discussed by candidates can be used as a predictor of their connective use.
Other studies have examined the role of deception in political communication. Results from Braun, Van Swol and Vang’s (2015) research showed that lying candidates on average used more words and negation statements than candidates telling the truth. Hence, a good topic of further investigation could be whether political candidates arguing in favor of a topic would on average use more connectives than office seekers arguing against a said topic. Moreover, Braun et al. (2015) discovered a difference in the responses given by politicians in scripted and interactive settings. On average office runners produce shorter utterances with simpler words in the interactive setting and longer, more complex ones in the scripted events (Braun et al., 2015). This goes back to the beginning of this article where the decision for using debates was justified. Debates are synchronous settings where speaker’s statements are prompted by questioning, and therefore, there is no time for preparation in advance or calculated use of words (Toma & Hancock, 2012). Because of time pressure, candidates also may find it hard to maintain proper decorum to respond as strategically as possible (Van Swol, Braun, & Malhotra, 2012). Hence, political statements taken from debates will indeed closely resemble politicians’ organic, unrefined speech. Consequently, by analyzing debates, the results of this and any other study will be more generalizable to common political communication, than if it was to use inauguration speeches, for example.
A type of research that comments on argumentation and connective use in debates is not only relevant for and contributing to the field of discourse, but to political and social sciences as well. It can serve as the foundation for further investigation on the topics of connectives and political argumentation as well as the interaction between the two. Investigating the use of connectives by various speakers could yield better understanding and differentiation between the different categories of connectives. A more comprehensive study of specific word use by politicians would be useful for future office seekers to improve their communication skills, or help voters better judge the candidates’ agenda. Knowing if there is any difference in the speaking styles between Democrats and Republicans based on connectives, would benefit parties and voters alike. Parties could better regulate and differentiate their image from one another and voters could understand what each side stands behind. The proposed study could even provide insight for a relationship between the topic discussed and the connectives used in a speech act. For all of the abovementioned reasons, this research aims to answer to following question:
6
against a topic in presidential candidate debates, and can topic be used as a predictor for connective use?
7
Methodology
A corpus analysis is used to describe the use of connectives in political speech. The speech units that were examined are types of connectives. The two main reasons as to why connectives were chosen over discourse markers are because they are practical and relatively easier to identify than discourse markers, and because they are the pieces that actually create coherent texts and speeches.
Materials
The corpus is built from transcripts of presidential debates between the leading candidates of the Democratic and Republican parties respectively in the United States. Starting from the latest presidential run in 2016 and going back to the debates in 2004, one presidential and one vice-presidential debate are analyzed from each quadrennium. Only the first debates of either category per year are included in the corpus as to avoid candidates referencing their arguments from previous debates and because it is supposed that the main topics of agreement/disagreement between candidates will be discussed in each first debate. Since the aim of this research is to disassociate from personal speaking style and rather focus on the general connective use of all political candidates, no debates were used from the 2008 presidential run of McCain and Obama. This was done to ensure that each candidate’s discourse was only analyzed once and to avoid any skewed data due to analyzing one person’s speaking style more than the rest. Following the aforementioned criteria leads to a corpus consisting of a total of six political debates: the first presidential debates from 2016, 2012 and 2004 together with the vise presidential debates of the same years. Therefore, the total number of politicians whose speech was analyzed equals twelve (six from the Republican party and six from the Democratic party).
Procedure
All debates were gathered in the same excel file in chronological order, with each speech act starting in a new cell. A speech act is a contribution by a speaker to a communicative exchange and it can either be one sentence or a couple of sentences uttered right after each other (Ludwig, 2020). All debates used in this research were already divided into speech acts from the source, so nothing was changed. Thus, for each speech act the following information was added: number of debate, year of debate, speaker, his/her political party and number of words in the speech act. What had to be coded was the general topic of the speech act (e.g. healthcare, immigrants, etc.), the stance of the speaker (for, against or neutral), the connectives used and what type they were.
The corpus was coded by a team of five coders, which are all students of the bachelor program International Business Communication at Radboud University,
8 Nijmegen. The subject of discussion for each speech act had to be coded for the variable
Topic. Each researcher did this individually for one debate. All topics would later be
analyzed and the most frequently used ones would also be used for the further statistical tests. For the variable Stance (i.e. for, against or neutral), double coding was performed by a first and second annotator in order to ensure intercoder reliability. Thus, the researchers grouped into five pairs and each coded whether the candidate was being in favour of the topic, against the specific topic or neutral in opinion. The double coding was performed for the variable Stance in the first 70 speech acts of every debate. Cohen’s Kappa was used to assess the level of agreement between each pair. Following McHugh (2012), a value below .59 is regarded as weak, from .60 to .79 is considered as moderate, between .80 and .90 as strong and above .90 – almost perfect level of agreement. Fortunately, the Kappa scores for all five pairs were between .711 and .885, which means that there was moderate to strong intercoder reliability for the variable Stance.
Next, the connectives used by each office seeker per speech act had to be noted and their type to be coded. This time no double coding was performed. Instead, the researchers decided to use a list of English Connectives from Prasad et al. (2007), which were presented in a table of one hundred connectives together with their type/sense (pp. 65-70). In order to differentiate whether these words were used as filling words and collocations (e.g. back and forth) or as actual connectives (e.g. And the truth is…) in the debates, the Hierarchy of Sense Tags model was used, again developed by Prasad et al. (2007). What the model does, is it divides connectives based on their sense in three different levels. The first level sorts connectives based on their belonging to one of the four major semantic classes, namely Temporal, Contingency, Comparison and Expansion. The Temporal class is used for temporally related situations, the Contingency class is used when there is a causal relationship between two situations, the Comparison class shows a relationship highlighting the differences between two situations, and the Expansion class shows relations which move the story forward (Prasad et al., 2007). The second level classifies connectives based on their class type, for example, Contingency has the types Cause and Condition. The last subtype level indicates the semantic contribution (Prasad et al., 2007). In order to recognize the different types of connectives, but not go into too much detail as per a qualitative study, this research was only focused on levels 1 and 2 from the table. Some connectives are only analyzed on the first level due to abstract or small differences between the second level categories. The table with all connectives and types can be found in the appendix.
For the final dataset, all moderators’ speech was deleted altogether since this research never intended to analyze moderator speech and for the sake of easier interpretation of future results. Hence, the database includes 2,692 speech acts by the 12 politicians, which accumulate to 89,164 words.
9
Results
Descriptive statistics showed that the amount of spoken words per politician does not differ too much (M = 7430.33, SD = 712.05), with a lowest score of 6 331 words and a highest of 8 457 words. To determine whether there is a significant difference in the amount of words spoken by Republicans and Democrats respectively, an independent sample t-test was performed on Party and Number of words. However, no significant difference was found (t (10) = .769, p = .459). Hence, all speakers, regardless of the political party they associate with, contributed a relatively equal amount of words to the corpus and no single speaker’s or party’s speech will prevail over the rest.
Further statistics showed that the data set contained 4 193 connectives, or in other words, 4.7% of all words spoken by the politicians were connectives. It is important to note that the list of connectives that was used to identify every word in the debate contained exactly 100 connectives. From those, 28 were never used by any candidate and 34 had been used less than 5 times total in all debates. Hence, due to their rare use and lack of a more complete sample, all connectives which were used 5 times or less in all 6 debates were left out of the analysis. This leaves the final data set with 38 connectives to be used in the analysis. In order to identify if any political group used more connectives a t-test was performed for Party and Connective. The result was insignificant (t (10) = .288, p = .779), concluding that there is no significant difference between the number of connectives used by Democrats and Republicans. Therefore, politicians from both political parties used a relatively equal amount of connectives.
Descriptive statistics also showed that from the 2 692 speech acts, 770 argued in favor of a topic, which equals 28,6%. From the rest, 892 speech acts or 33,1% were neutral or unclear and 1030 speech acts or 38,3% argued against a topic. Next, it is important to investigate whether politicians use more words when arguing for or against a topic, since it is supposed that longer statements would provide more room for the use of connectives. The longer the speech act, the higher the chance for the politician to use a connective. In order to check that, a t-test was performed for the variables Number of words and Stance. The results from the test showed some significant differences (t (1798) = 3.744, p < .001). On average, politicians used more words per speech act when arguing for a topic (M = 39.90, SD = 24.30) than when arguing against a topic (M = 35.52, SD =24.71). Results can be seen in Table 1 below. Consequtive t-tests showed a significant difference between the number of words used to argue in favor of a topic or neutral argumentation (t (1660) = 13.694, p < .001), and neutral argumentation and arguing against a topic (t (1920) = 10.349, p < .001). The results can be view in Table 2 and Table 3 respectively. The average number of words used when arguing both for and against a topic was higher than for neutral argumentation (M = 24.55, SD = 21.60). From here we can conclude that positive argumentation used the highest number of words per speech act, negative argumentation was shorter and neutral argumentation used the shortest speech acts.
10 According to the results, politicians used the longest speech acts when they took a positive stance, even though positive argumentation was the least frequent for speech acts (28,6%).
Table 1. T-test for number of words in positive and negative argumentation
Table 2. T-test for number of words in positive and neutral argumentation
11 In order to investigate if there is a relationship between the number of connectives used by politicians and the type of argumentation in which they appeared, the number of connectives had to be standardized per a 100 words. Because negative speech acts are a lot more frequent than neutral or positive speech acts, calculating the raw frequencies of connectives per stance would lead to a skewed result. Hence, after standardizing the number of connectives for each type of argumentation the following results were found: positive argumentation used an average of 4.97 connectives per 100 words, while negative and neutral argumentation used less, with average results of 4.56 and 4.53 respectively. Therefore, significantly more connectives were used when politicians took a positive stance, as compared to both the negative and the neutral one. Meanwhile, there was no significant difference in the number of connectives used in negative and neutral argumentation.
As a way to see if politicians used different types of connectives for the different topics depending on their stance, the same Chi-square test was used. In order to read the table correctly it is important to note which types of connectives correspond to the four classes. Numbers 1 and 12 are identified as Temporal connectives; numbers 2, 21 and 23 are Contingency; numbers 3, 31 and 33 are Comparison; and numbers 4, 41, 42, 43, 44 and 45 are Expansion connectives. Based on the 4 main classes of connectives, Expansion connectives were used most often (15,1% when arguing for and 14,4% when arguing against). Contingency connectives were the second most used, with 7% used in positive argumentation and 7,5% in negative argumentation. Next, from the Temporal connectives 4,2% were used when arguing for and 6,1% when arguing against a topic. Finally, Comparison connectives were the least used with only 3,4% in positive argumentation and 4,6% in negative argumentation. From the four classes, the only type of connectives used more in positive argumentation than in negative such, was Expansion connectives. Results are presented in Table 4.
12 Next, descriptive statistics for the variable Topic were performed. Because the topics were not double coded, there were a lot of synonyms used for one term and a lot of different terms, discussed between 1 and 2 times in all debates. This resulted in a very extensive table, which was very hard to read. Hence, from the table, only the top ten categories were left for further analysis, which represented 56.2% of the topics in the table. These were namely: Opponent, Taxes, Healthcare, Iraq war, Other, Terrorism, Homeland security, Military, Economy and Overseas Conflict. The rest of the topics were discussed in less than 2% of all speech acts and were therefore left out of the analysis.
In order to answer part of the research question, which is if different connectives are used when discussing different topics, a Chi-square test for the variables Type of connective and Topic was conducted. The results showed a significant relationship between the two variables, χ2 (140) = 523.603, p < .001. For all topics, the most frequent type of connective used was conjunction (41), which was used in 36.1% of the topics. The second most frequent connectives are the types of synchronous (12), cause (21) and contrast (31) connectives, which were used in 12%, 14.5% and 10% of the topics respectively. These exact connectives represent one main class of connectives each. Conjunction represents Expansion connectives, synchronous represents Temporal connectives, cause represents Contingency connectives and contrast represents Comparison connectives. Hence, by looking at the table we can conclude that Expansion connectives are used for all kinds of topics, Comparison connectives are frequently used in Military topics, Contingency connectives are often used in topics like Economy, Healthcare, Iraq war, Taxes and Other, and Temporal connectives are frequently used in Homeland security, Opponent, Overseas conflict and Terrorism topics. Results can be viewed in Table 5 below. The rest of the numbers and the types of connectives they stand for can be found in Appendix B.
14 Finally, a Chi-square test was conducted to research the relationship between the party of the speaker and the topics discussed. The Chi-square test showed a significant relationship between the variables Party and Topic, χ2 (10) =31.116, p = .001. Republicants discussed significantly more often the topics of Economy, Military, Other, Overseas conflict, Taxes and Terrorism. Democrats were fonder of discussing Healthcare, Homeland security and Iraq war. Both parties discussed the topic of Opponent, otherwise said – made comments about one another, an equal number of times. The results are presented in Table 6.
15
Conclusion
From the descriptive statistics it can be concluded that the dataset contains more or less equally distributed communication units. All politicians, in spite of their political orientation, on average used about the same amount of words and connectives in their debates. However, t-tests showed that while most speech acts argued against a topic, they contained less words per speech act. The longest speech acts were actually arguing in favor of a topic. One way to interpret these results is that positive argumentation flows more easily in the mind of the speaker and therefore, he or she does not need several speech acts to express their ideas, but rather a single coherent one. Another explanation could be that when politicians argue against a topic, they interrupt each other a lot more often, which results in shorter negative speech acts, but a greater amount of them. Neutral argumentation used the shortest speech acts with the lowest number of connectives. This could infer that when politicians have a neutral opinion about a topic, they prefer to be short-spoken and not link their ideas together. After standardizing the number of connectives used per 100 words for the three types of stances, the statistics showed that there were significantly more connectives used in positive argumentation, as compared to negative and neutral, which had almost no difference in-between. Hence, positive argumentation, although less frequent overall, used longer speech acts with the highest number of connectives per 100 words. Meanwhile, negative argumentation was most frequent in the debates, but used less connective on average.
A consequent Chi-square test showed the distribution of the four main types of connectives based on stance. For the Temporal, Contingency and Comparison classes more connectives were used when arguing against a topic. On the contrary, for the Expansion class more connectives were used when arguing for a topic, than when arguing against. These results are in support of the previous findings in a number of ways. As already mentioned, the Expansion class “expands” the discourse and helps the story move forward, therefore, adding more words to the speech act. This explains why positive argumentation, which uses Expansion connectives the most, contains the longest speech acts. Consequently, because negative argumentation uses more the other three types of connectives, linking so many ideas to one another leads to shorter speech acts with less connectives in general. The most important result is that connectives are in fact used differently depending on the stance of the speaker. If he or she is supporting the topic, they are likely to explain their view using longer statements and connecting ideas a lot more often with words like “also”, “and”, “moreover”, “or”. If they were arguing against a topic, they would be using shorter speech acts, and occasionally including words like “on the contrary”, “if”, “then”, etc.
As far as the topics of discussion are concerned, descriptive statistics showed that the most frequently discussed ones are Opponent, Taxes, Healthcare, Iraq war, Other, Terrorism, Homeland security, Military, Economy and Overseas Conflict. Next, a Chi-square test displayed a significant relationship between the variables Type of
16 connective and Topic discussed. The most frequently used types of connectives were representative of the four major classes of connectives. Hence, the most used connectives in any topic are Expansion connectives. Contingency connectives are the second most used in topics of Economy, Healthcare, Iraq war, Taxes and Other. Temporal types of connectives are the second most used in the topics of Homeland security, Opponent, Overseas conflict and Terrorism. Finally, Comparison connectives are most used when discussing the Military.
Even though descriptive statistics showed no significant differences between the use of connectives by Republicans and Democrats, the following findings suggest otherwise. Since there is a significant difference in the frequency with which Democrats and Republicans discuss certain topics and there is also a significant relationship between topics discussed and connectives used, it can be implied that based on the topic discussed politicians will use some types of connectives more frequently than others. That said, it can be suggested that Democrats will be using more Temporal and Contingency connectives when discussing Healthcare, Homeland security or the war in Iraq. Similarly, the Republicans may use more Compering connectives when talking about Military. However, both parties are expected to use Expansion connectives quite frequently, despite of the topic of discussion.
Discussion
This research uncovered some interesting findings about how connectives, topics of discussion and stance can affect political speech. To begin with, the analysis of Stance showed important differences between speech act construction, when arguing for or against something. Speech acts with positive argumentation were less frequent, but proved to be higher in word count and consequently, higher in use of Expansion class connectives. Moreover, positive argumentation included the highest number of connectives overall. Meanwhile negative argumentation was more commonplace and used shorter speech acts with more frequent use of Temporal, Contingency and Comparing connectives. However, the number of connectives included was no different than the one for neutral argumentation. One way to explain these results could be through the Pollyanna principle, which states that people tend to remember positive things better and for longer, than negatives ones. Perhaps, politicians know that there is a positivity bias and, therefore, only actively try to persuade listeners when they are taking a positive stance. In any case, the results confirm that there is indeed a difference in the use of connectives depending on the stance of the speaker.
Another curious result was that there is a significant relationship between the topic of discussion itself and the types on connectives used. Each class of connectives
17 was used in distinct topics of discussion. This finding could be explained by the nature of the connective classes and what each topic infers. For example, when discussing military topics, Comparison connectives are frequently used. This can be explained by the fact it is common practice for politicians to compare the military might of their own country to that of their enemies or potential threats.
Finally, the analysis showed that there is a significant relationship between the political party of the speaker and the topics he/she discusses. In particular, the findings that Republicans are more likely to discuss Military, Overseas conflict and Terrorism are in line with the findings of Thomson and Froese (2016) that the republican party emphasizes on military might and criminal punishment. Furthermore, this paper discovered that Democrats are more likely to discuss Homeland security, which again is confirmed by Thomson and Froese (2016). Furthermore, having results confirm that there is a significant relationship between the topics discussed and the connectives used, suggests topic may potentially be used as a predictor of connective use.
Some limitations of the present research are the following. Firstly, there is a certain level of ambiguity in the coding method used in this study, which can be bettered by double coding the entire debates. Another factor that influences the results could be the male/female ratio of politicians in this research, as there was only 1 female candidate, as opposed to 11 males. Although the political field is predominantly represented by men, further research could try and investigate a sample with an equal number of members from both sexes. Finally, a lot of the connectives this study was set out to investigate, turned out to be rarely to never used in the database. Hence, less than half of the initial connectives were analyzed.
Further researches should further investigate whether topic can be a predictor of connective use in presidential debates, since connective use can become easily predictable, if a clear range of topics is defined for every connective class. One more suggestion is to try and gather an even bigger corpus to analyze, so that a wider variety of connectives will be used more often and the reliability of results will increase. As this would be a very demanding task, another suggestion is to pick just one connective representative of each class of connectives and look for more specific distinctions or similarities between the class categorizations.
18
References:
Aijmer, Karin (1996). Conversational Routines in English: Convention and Creativity. London: Longman
Fox-Tree, J.E. and Schrock, J.C. (2002) ‘Basic Meanings of You Know and I Mean’, Journal of Pragmatics 34: 727–47
Fraser, B. (1999). What are discourse markers? Journal of Pragmatics, 31(7), 931–952. https://doi-org.ru.idm.oclc.org/10.1016/S0378-2166(98)00101-5
Goodwin, J., & Innocenti, B. (2019). The pragmatic force of making an argument. Topoi
: An International Review of Philosophy, 38(4), 669-680.
doi:10.1007/s11245-019-09643-8
Innocenti, B., & Miller, E. (2016). The persuasive force of political humor. Journal of
Communication, 66(3), 366-385. doi:10.1111/jcom.12231
Ismail, A. I. H. M. (2012). Discourse markers in political speeches: Forms and functions. Journal of the College of Education for Women, 23(4)
Kapoutsis, I., & Volkema, R. (2019). Hard-core toughie: Donald trump’s negotiations for the united states presidency. Negotiation Journal, 35(1), 47-63. doi:10.1111/nejo.12263
Louwerse, Max M., Mitchell, Heather H., 2003. Toward a taxonomy of a set of discourse markers in dialogue: a theoretical and computational linguistic account. Discourse Processes 35 (3), 243–281
Ludwig, K. (2020). What are group speech acts? Language and Communication, 70, 46– 58. https://doi-org.ru.idm.oclc.org/10.1016/j.langcom.2019.04.004
McHugh, M. (2012). Interrater reliability: The kappa statistic. Biochemia Medica, 22(3), 276-282
Michael T. Braun, Lyn M. Van Swol & Lisa Vang (2015) His Lips Are Moving: Pinocchio Effect and Other Lexical Indicators of Political Deceptions, Discourse Processes, 52:1, 1-20, DOI: 10.1080/0163853X.2014.942833
Mukhortov, D.S., & Malyavina, Y.S. (2019). Cross-situational consistency of female politicians’ language use. RUDN Journal of Language Studies, Semiotics and Semantics, 10(4), 1021—1036. doi: 10.22363/2313-2299-2019-10-4-1021-1036 Overstreet, M. and Yule, G. (2002) ‘The Metapragmatics of and Everything’, Journal of
Pragmatics 34: 785–94
Prasad, R., Miltsakaki, E., Dinesh, N., Lee, A., Joshi, A., Robaldo, L., & Webber, B. L. (2007). The penn discourse treebank 2.0 annotation manual.
19 Reyes, A. (2014). Bush, obama: (in)formality as persuasion in political discourse. Journal of Language and Politics, 13(3), 538-562. doi:10.1075/jlp.13.3.08rey
Sanders, T.J.M. & Spooren, W.P.M.S. (2015). Causality and subjectivity in discourse: The meaning and use of causal connectives in spontaneous conversation, chat interactions and written text. Linguistics 53(1), 53-92
Sanders, T.J.M., Spooren, W.P.M.S., & Noordman, L.G.M. (1992). Toward a taxonomy of coherence relations, Discourse Processes15(1), 1-35
Swan, Michael (2005). Practical English Usage. Oxford: Oxford University Press. p. xviii. ISBN 0-19-442098-1
Thomson, R., & Froese, P. (2016). God versus party: Competing effects on attitudes concerning criminal punishment, national security, and military service. Journal
for the Scientific Study of Religion, 55(4), 839-858. doi:10.1111/jssr.12293
Toma, C., & Hancock, J. T. (2012). What lies beneath: The linguistic traces of deception in online dating profiles. Journal of Communication, 62, 78–97
Transitional Devices // Purdue Writing Lab. Retrieved March 13, 2020, from:
https://owl.purdue.edu/.../transitions_and_transitional_devices/transitional _devices.html
Van Swol, L. M., Braun, M. T., & Malhotra, D. (2012). Evidence for the Pinocchio effect: Linguistic differences between lies, deception by omission, and truth. Discourse Processes, 49, 79–106.
20
Appendix A. Codebook for specific connectives:
Connective Code as:
Accordingly cause = 21 Additionally Conjunction = 41 After Synchronous = 12 Afterward Synchronous = 12 Also Conjunction = 41 Alternatively Alternative = 44 Although COMPARISON = 3 And Conjunction = 41 As Synchronous = 12 As a result Cause = 21 As an alternative Alternative = 44 As if EXPANSION = 4 As long as - Condition = 23 - Synchronous = 12 As soon as Synchronous = 12 As though - Comparison = 3 - Restatement = 43 As well Conjunction = 41 Because cause = 21 Before Synchronous = 12
21 Besides Conjunction = 41 But Contrast = 31 By comparison Contrast = 31 By contrast Contrast = 31 By then Synchronous = 12 Consequently cause = 21 Conversely Contrast = 31 Earlier Synchronous = 12 Either..or Alternative = 44 Else Alternative = 44 Except Exception = 45 Finally - Conjunction = 41 - Synchronous = 12 For cause = 21
For example Instantiation = 42
For instance Instantiation = 42
Further Conjunction = 41
Furthermore Conjunction = 41
Hence cause = 21
However Contrast = 31
If Condition =23
22
If.. then Condition = 23
In addition Conjunction = 41
In contrast Contrast = 31
In fact - Conjunction = 41
- Restatement = 43
In other words Restatement = 43
In particular - Instantiation = 42
- Restatement = 43
In short Restatement = 43
In sum Restatement = 43
In the end EXPANSION = 4
In turn Synchronous = 12 Indeed - Conjunction = 41 - Restatement = 43 Insofar as cause = 21 Instead Alternative = 44 Later Synchronous = 12 Lest - Alternative = 44 - Condition = 23 Likewise Conjunction = 41 Meantime Synchronous = 12 Meanwhile - Conjunction = 41 - Synchronous = 12 Moreover Conjunction = 41 Much as COMPARISON = 3
23 Neither...nor EXPANSION = 4 Nevertheless COMPARISON = 3 Next Synchronous = 12 Nonetheless COMPARISON = 3 Nor Conjunction = 41
Now that cause = 21
On the contrary Contrast = 31
On the one hand...on the other hand Contrast = 31
On the other hand Contrast = 31
Once Synchronous = 12 Or Alternative = 44 Otherwise Alternative = 44 Overall Restatement = 43 Plus Conjunction = 41 Previously Synchronous = 12 Rather - EXPANSION = 4 - Contrast = 31 Regardless Concession = 33 Separately Conjunction = 41 Similarly Conjunction = 41 Simultaneous Synchronous = 12
24 Since - cause = 21 - Synchronous = 12 So cause = 21 So that cause = 21 Specifically Restatement = 43 Still COMPARISON = 3 TEMPORAL = 1 Then Synchronous = 12 Thereafter Synchronous = 12 Thereby cause = 21 Therefore cause = 21 Though COMPARISON = 3 Thus cause = 21 Till Synchronous = 12 Ultimately Synchronous = 12 Unless Alternative = 44 Until Synchronous = 12 When Synchronous = 12
When and if - Condition = 23
- Synchronous = 12 Whereas Contrast = 31 While - Contrast = 31 - Synchronous = 12 Yet COMPARISON = 3 TEMPORAL = 1
